heavy vehicle systems optimization · pdf filereport less dependence on ... a study of...

A Strong Energy Portfolio for a Strong AmericaEnergy efficiency and clean, renewable energy will mean a stronger economy, a cleaner environment,

and greater energy independence for America. Working with a wide array of state, community, industry, anduniversity partners, the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy

invests in a diverse portfolio of energy technologies.

For more information contact:EERE Information Center

1-877-EERE-INF (1-877-337-3463)www.eere.energy.gov

2 0 0 4A N N U A L

P R O G R E S S

R E P O R T

L e s s d e p e n d e n c e o n f o r e i g n o i l , a n d e v e n t u a l t r a n s i t i o n t o

a n e m i s s i o n s - f r e e , p e t r o l e u m - f r e e v e h i c l e

HEAVY VEHICLESYSTEMS

OPTIMIZATION

FreedomCAR and Vehicle

Technologies Program

U.S. Department of EnergyFreedomCAR and Vehicle Technologies Program1000 Independence Avenue, S.W.Washington, D.C. 20585-0121

FY 2004

Annual Progress Report for Heavy VehicleSystems Optimization

Energy Efficiency and Renewable EnergyFreedomCAR and Vehicle Technologies Program

Approved by Dr. Sidney Diamond Technology Area Development Specialist

February 2005

Heavy Vehicle Systems Optimization Program FY 2004 Annual Report

iii

Contents

Foreword by Dr. Sidney Diamond, FreedomCAR and Vehicle Technologies Program,Energy Efficiency and Renewable Energy, U.S. Department of Energy ................................. 1

I. Aerodynamic Drag Reduction...................................................................................................... 3

A. DOE Project on Heavy Vehicle Aerodynamic DragLawrence Livermore National Laboratory; R.C. McCallen, et al. ......................................... 3

B. A Study of Reynolds Number Effects and Drag-Reduction Concepts on aGeneric Tractor-TrailerNASA Ames Research Center; B.L. Storms, et al. ................................................................. 9

C. An Experimental Study of Aerodynamic Drag of Empty and Full Coal Cars.......................... 19

D. Experimental Measurement of the Flow-Field of Heavy TrucksUniversity of Southern California; F. Browand, et al. ............................................................ 21

E. Continued Development and Improvement of Pneumatic Heavy VehiclesGeorgia Tech Research Institute; R.J. Englar........................................................................... 27

F. Heavy Vehicle Aerodynamic Drag: Experiments, Computations, and DesignLawrence Livermore National Laboratory; K. Salari, et al. ..................................................... 35A. An Expermental Study of Drag Reduction Devices for a Trailer

Underbody and Base............................................................................................... 35B. Investigation of Predictive Capability of RANS to Model Bluff Body

Aerodynamics ......................................................................................................... 39C. Splash and Spray Suppression ................................................................................ 43D. Computational Investigation of Aerodynamics of Rail Coal Cars

and Drag-Reducing Add-On Devices ..................................................................... 45

G. Computational and Analytical Simulation of Simplified GTS Geometries/Bluff BodiesSandia National Laboratories; L.J. DeChant, et al. ................................................................ 48

H. Commercial CFD Code Benchmarking for External Aerodynamics Simulations ofRealistic Heavy-Vehicle ConfigurationsArgonne National Laboratory; W.D. Pointer ........................................................................... 55

I. Bluff Body Flow Simulation Using a Vortex Element MethodCalifornia Institute of Technology; A. Leonard, et al. ............................................................ 65

II. Thermal Management .................................................................................................................. 70

A. Cooling Fan and System Performance and Efficiency ImprovementsCaterpillar, Inc.; R.L. Dupree, et al. ....................................................................................... 70


iv

Contents (Cont.)

B. Efficient Cooling in Engines with Nucleate BoilingArgonne National Laboratory; J.R. Hull .................................................................................. 77

C. Evaporative CoolingArgonne National Laboratory; S.U.S. Choi.............................................................................. 83

D. Nanofluids for Thermal Management ApplicationsArgonne National Laboratory; S.U.S. Choi.............................................................................. 89

E. Erosion of Materials in NanofluidsArgonne National Laboratory; J.L. Routbort ........................................................................... 96

F. Analysis of Ventilation Airflow and Underhood Temperatures in theEngine Enclosure of an Off-Road MachineArgonne National Laboratory; T. Sofu, et al. ......................................................................... 101

III. Friction and Wear ......................................................................................................................... 109

A. Boundary Lubrication MechanismsArgonne National Laboratory; O.O. Ajayi, et al. ................................................................... 109

B. Parasitic Engine Loss ModelsArgonne National Laboratory; G. Fenske, et al. ..................................................................... 115

C. Superhard Nanocrystalline Coatings for Wear and Friction ReductionArgonne National Laboratory; A. Erdemir, et al. ................................................................... 121

IV. Fuel Reformer Systems ................................................................................................................. 127

A. Diesel Fuel Reformer TechnologyArgonne National Laboratory; M. Krumpelt. ........................................................................... 127

B. On-Board Plasmatron Hydrogen Production for Improved Vehicle EfficiencyPlasma Science and Fusion Center, D.R. Cohn, et al. ............................................................ 130

V. EM Regenerative Shocks .............................................................................................................. 138

EM Shock AbsorberArgonne National Laboratory; J.R. Hull .................................................................................. 138

VI. Joining Carbon Composites ......................................................................................................... 142

Innovative Structural and Joining Concepts for Lightweight Design ofHeavy Vehicle SystemsWest Virginia University; J. Prucz and S. Shoukry.................................................................. 142


v

Contents (Cont.)

VII. Analysis .......................................................................................................................................... 151

Systems Analysis for Heavy VehiclesArgonne National Laboratory; L.L. Gaines.............................................................................. 151

VIII. Off-Highway .................................................................................................................................. 156

A. 21st Century Locomotive TechnologyGE Global Research; L. Salasoo, et al. ................................................................................... 156

B. Advanced Hybrid Propulsion and Energy Management System for HighEfficiency, Off Highway, 320 Ton Class, Diesel Electric Haul TrucksGE Global Research; T. Richter ............................................................................................... 164

IX. Particulate Matter Characterization ........................................................................................... 169

Morphology, Chemistry, and Dynamics of Diesel ParticulatesArgonne National Laboratory; K.O. Lee .................................................................................. 169

X. Brake Systems................................................................................................................................ 176

Advanced Brake Systems for Heavy-Duty VehiclesPacific Northwest National Laboratory; G. Grant .................................................................... 176

XI. More Electric Truck ..................................................................................................................... 183

Parasitic Energy Loss Reduction and Enabling Technologies forClass 7/8 TrucksCaterpillar, Inc.; W. Lane ......................................................................................................... 183

XII. Ultralight Transit Bus System...................................................................................................... 193

Vehicle System Optimization of a Lightweight, Stainless-Steel BusAutokinetics, Inc.; J.B. Emmons .............................................................................................. 193


vi


1

Foreword

The U.S. Department of Energy (DOE) effort on Heavy Vehicle Systems Optimization addresses the veryimportant area of non-engine losses that account for a significant amount of fuel consumption. These lossesare usually referred to as Parasitic Energy Losses. For example, aerodynamic drag accounts for about 53% ofthe non-engine losses of a fully loaded class 8 tractor-trailer traveling at 65 mph on a level highway; rollingresistance, 32%; drivetrain, 6%; and auxiliary losses, 9%. As aerodynamic drag is reduced, brakes thatalready operate at capacity can be seriously compromised. Therefore, the program also includes a safety lineitem that concentrates on improving brake systems. Funds in FY 2004 were approximately $10.8 M. Projectsare selected on the basis of (1) proposals addressing open solicitations issued by the FreedomCAR andVehicle Technologies (FCVT) Program or (2) proposals from national laboratories. Contracts with ourindustrial partners are cost-shared at least 50%.

The primary goal of this technology development area is to develop cost-effective technologies that willvastly improve the fuel efficiency of heavy vehicles and will eventually devolve into all vehicles. Theultimate goals are consistent with those of the 21st Century Truck Partnership: a 10-mph, fully loaded class 8tractor/trailer traveling at highway speeds. Note that a 1% increase in fuel efficiency for long-haul trucks willresult in a yearly fuel savings of 100 million gallons of fuel.

Individual goals are to reduce the aerodynamic drag coefficient by 25%, resulting in approximately a 12.5%increase in fuel efficiency at 65 mph (and increasing at higher speeds); a 40% decrease in rolling resistance; a30% decrease in drivetrain losses; a 50% decrease in auxiliary loads; and an 8% decrease in the size of theradiator, despite the higher cooling demands of higher-power engines and exhaust gas recirculation.

Steady and significant progress on achieving these goals has been made in FY 2004:

• On-track tests of a tractor-trailer combination with flat boat tails have demonstrated a 4.2% increasein fuel efficiency at 60 mph. Furthermore, rounding the leading edge of the trailer results in furtherincreases in fuel efficiency. Computations indicate that rounded boat tails and side skirts will reducethe drag coefficient by 20%.

• A recently signed contract with the Truck Manufacturers Association with four OEMs assubcontractors will fleet-test some of the near-commercial devices that are available for reduction ofaerodynamic drag.

• Road tests of the “More Electric Truck” have demonstrated a 2% increase in fuel economy obtainedby electrification of several of the belt-driven components. Furthermore, the on-board auxiliarypower unit will increase fuel economy by another 6% by reduction of engine idling time. The idlereduction campaign of FCVT, which consisted of analysis and publicity, has captured the interest ofthe U.S. Environmental Protection Agency (EPA), U.S. Department of Defense (DOD),U.S. Department of Transportation (DOT), and several state energy agencies, all of which have nowtaken a lead role in implementation.

• Tests under realistic conditions on a laser-glazed rail have shown that a reduction in friction of up to40% can be achieved. A consist management program was demonstrated to save about 2.5% fuel in a100-car train running between Kansas City and Los Angeles. Further savings will be achieved whenthe locomotives are fully hybridized.


2

• Wind tunnel tests have shown that the drag coefficient of empty coal cars is between 35 and 42%higher than that of loaded coal cars. Computations indicate that simple devices can reduce the dragcoefficient in empty coal cars by 10%.

• Superhard coatings have achieved friction coefficients of less than 0.03 under boundary-layer-lubrication conditions with virtually no wear that could contribute significantly to the increasedenergy efficiency of reciprocating and rotating vehicular components.

• West Virginia University has devised and analyzed alternative structural arrangements for the floor ofa heavy van trailer, leading to the conclusion that significant weight reduction is possible.

• Argonne National Laboratory has developed a model of heat transfer in nanofluids and found that akey mechanism of temperature- and size-dependent thermal conductivity is the dynamic interactionsbetween nanoparticles and liquid molecules. This finding could lead to a substantial downsizing oftruck cooling systems and a reduction of the parasitic energy loss of the systems.

Finally, four new industrial contracts have been initiated: Truck Manufacturer Association — fleetdemonstration of near-commercial drag reduction technologies; Eaton Corp — reduction of friction and wearof gears; Caterpillar — demonstration of a 5–11% improvement in the fuel efficiency of medium-duty trucksthrough the use of electrically driven components and the use of intelligent engine idling control and energystorage technologies; and Caterpillar — demonstration of increased fuel efficiency through improvements incooling system performance, air system management, and advanced power management.

The accomplishments mentioned above reflect only some of the progress achieved in FY 2004 in the area ofparasitic energy loss. A review of all of the sections of the annual report indicates that the progress toward thestated goals of the HV Systems Optimization Technology Development Area is significant and that they canbe met by continuing the same vigorous R&D that currently is contained within the research portfolio.

Finally, it is my pleasure to thank the dedicated scientists and engineers who work with the FCVT Program.It is their expertise, drive, and initiative that have made great strides in their technologies. Their innovationswill continue to contribute to advances in reduction of parasitic energy losses for heavy vehicles and to thesignificant reduction of U.S. dependence on petroleum fuels.

Dr. Sidney Diamond, Technology Area Development SpecialistHeavy Vehicle Systems OptimizationFreedomCAR and Vehicle Technologies ProgramEnergy Efficiency and Renewable EnergyU.S. Department of Energy


3

I. Aerodynamic Drag Reduction

A. DOE Project on Heavy Vehicle Aerodynamic Drag

Project Principal Investigator: R.C. McCallenLawrence Livermore National LaboratoryP.O. Box 808, Livermore, CA 94551-0808(925) 423-0958, e-mail: [email protected]

Principal Investigator: K. SalariCo-Investigators: J. Ortega, P. Castellucci, C. Eastwood, K. WhittakerLawrence Livermore National LaboratoryP.O. Box 808, Livermore, CA 94551-0808(925) 424-4635, e-mail: [email protected]

Principal Investigators: L.J. DeChantCo-Investigators: C.J. Roy, J.L. Payne, B. HassanSandia National LaboratoriesP.O. Box 5800, MS 0825, Albuquerque, NM 87185-0825(505) 844-4250, e-mail: [email protected]

Principal Investigator: W.D. PointerArgonne National Laboratory9700 S. Cass Avenue, NE-208, Argonne, IL 60439(630) 252-1052, e-mail: [email protected]

Principal Investigator: F. BrowandCo-Investigators: M. Hammache, T.-Y. HsuAerospace & Mechanical Engineering, University of Southern CaliforniaRRB 203, Los Angeles CA 90089-1191(213) 740-5359, e-mail: [email protected]

Principal Investigator: J. RossCo-Investigators: D. Satran, J.T. Heineck, S. Walker, D. YasteNASA Ames Research CenterMS 260-1, Moffett Field, CA 94035(650) 604-6722, e-mail: [email protected]

Principal Investigator: R. EnglarGeorgia Tech Research InstituteATASL, CCRF, Atlanta, GA 30332-0844(770) 528-3222, e-mail: [email protected]

Principal Investigator: A. LeonardCo-Investigators: M. Rubel, P. ChatelainCalifornia Institute of Technology1200 East California Blvd. MC 301-46, Pasadena, CA 91125(626) 395-4465, e-mail: [email protected]


4

Technology Development Manager: Sid Diamond(202) 586-8032, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, e-mail: [email protected]

Contractors: Lawrence Livermore National Laboratory, Sandia National Laboratories, ArgonneNational Laboratory, NASA Ames Research Center, Georgia Tech Research Institute, University ofSouthern California, California Institute of TechnologyContract No.: W-7405-ENG-48, DE-AC04-94AL85000, W-31-109-ENG-38, DE-AI01-99EE50559, DE-AC03-02EE5069, DE-AC03-98EE50512, DE-AC03-98EE50506

Consortium members road-testing base flaps at Crows Landing, California.Charles Radovich, USC

Fred Browand, USC Scott Johnston, PATHMathieu Boivin, Norcan Aluminum

Objectives• Provide guidance to industry in the reduction of aerodynamic drag of heavy truck vehicles.

• Establish a database of experimental, computational, and conceptual design information and demonstratepotential of new drag-reduction devices.

Approach• Develop and demonstrate the ability to simulate and analyze aerodynamic flow around heavy truck vehicles by

using existing and advanced computational fluid dynamics (CFD) tools.

• Through an extensive experimental effort, generate an experimental database for code validation.

• By using experimental database, validate computations.

• Provide industry with design guidance and insight into flow phenomena from experiments and computations.


5

• Investigate aero devices (e.g., base flaps, tractor-trailer gap stabilizer, underbody skirts and wedges, blowingand acoustic devices), provide industry with conceptual designs of drag reducing devices, and demonstrate thefull-scale fuel economy potential of these devices.

Accomplishments• The Program has demonstrated several concepts and devices that meet the 25% drag reduction goal.

• Insight from experiments and experimental database has provided clear guidance to industry on reliable,predictable experimental techniques.

• Computational results provide clear guidance and caution warnings on the use of steady Reynolds-averagedNavier Stokes (RANS) models for CFD simulations.

• Investigated aerodynamics of filled and empty rail coal cars and designed devices predicted to produce dragreductions of 10% for empty cars.

Future Direction• Continue to develop and evaluate drag-reducing conceptual designs and encourage and work with industry to

road-test the most promising drag-reducing devices.

• Continue experimental data reduction and analysis for the generic conventional model (GCM).

• Continue computations of flow around GCM, compare to experimental data, perform analyses, and provideguidance to industry on use of unsteady RANS and hybrid RANS/Large-Eddy Simulation (LES) methods.

• Develop and use an apparatus for studying wheel and tire splash and spray, pursuing ways to minimize this roadsafety hazard.

• Investigate airflow around rotating tires for improved brake cooling, as well as drag reduction.

• Collaborate with DOE Industrial Consortium, which will be conducting fleet tests of advanced aerodynamicdrag reduction devices. Schedule industry site visits and meetings to share findings and encourage considerationof effective design concepts for road testing.

• Leverage Program work and seek funding from other agencies.

IntroductionA modern Class 8 tractor-trailer can weigh up to80,000 pounds and has a wind-averaged dragcoefficient around CD = 0.6. The drag coefficient isdefined as the drag/(dynamic pressure x projectedarea). The higher the speed the more energy isconsumed in overcoming aerodynamic drag. At70 miles per hour, a common highway speed today,overcoming aerodynamic drag represents about 65%of the total energy expenditure for a typical heavytruck vehicle. Reduced fuel consumption for heavyvehicles can be achieved by altering truck shapes todecrease the aerodynamic resistance (drag). It isconceivable that present-day truck drag coefficientsmight be reduced by as much as 50%. This reductionin drag would represent approximately a 25%reduction in fuel use at highway speeds. Anestimated total savings of $1.5 billion per year (pre-2004 fuel prices) can be recognized in the United

States alone for just a 6% reduction in fuel use. Thisreduction represents 1% of all fuel use in the UnitedStates.

The project goal is to develop and demonstrate theability to simulate and analyze aerodynamic flowaround heavy truck vehicles by using existing andadvanced computational fluid dynamics (CFD)tools. Activities also include an extensiveexperimental effort to generate data for codevalidation and a design effort for developing drag-reducing devices. The final products are specificdevice concepts that can significantly reduceaerodynamic drag (and thus improve fuel efficiency)in addition to an experimental database andvalidated CFD tools. The objective is to provideindustry with clear guidance on methods ofcomputational simulation and experimentalmodeling techniques that work for predicting the


6

flow phenomena around a heavy vehicle and add-ondrag-reducing devices. Development of effectivedrag-reducing devices is also a major goal.

The following reports on the findings andaccomplishments for fiscal year 2004 in the project’sthree focus areas:

• Drag-reduction devices,• Experimental testing, and• Computational modeling.

A summary is given in the introduction portion ofthis report, and detailed reports from eachparticipating organization are provided in theappendices. Included are experimental results andplans by NASA, USC, GTRI, and LLNL inAppendices A through D. The computational resultsfrom LLNL and SNL for the integrated tractor-trailer benchmark geometry called the GroundTransportation System (GTS) model and trailerwake flow investigations are in Appendices D and E,from ANL for the Generic Conventional Model(GCM) in Appendix F, by LLNL for the tractor-trailer gap flow investigations in Appendix D, andturbulence model development and benchmarksimulations being investigated by LLNL and Caltechin Appendices D and G. USC also provides field testresults for the base flap device (Appendix B), GTRIcontinues its investigation of a blowing device(Appendix C), and LLNL presents results for baseskirts and wedges (Appendix D).

Drag-Reduction DevicesThere are three areas identified for aero dragreduction, and several drag reduction devices havebeen investigated

• Tractor-Trailer GapStabilizing devices, cab extenders

• Wheels/UnderbodySkirts/lowboy trailer (∆CD ~ 0.05), splitter plate

• Trailer BaseBoattail plates (∆CD ~ 0.05), base flaps (∆CD ~0.08), rounded edges, and pneumatics

Overview of AccomplishmentsThe Program has demonstrated several concepts anddevices that meet the 25% drag reduction goal.Specific devices have addressed base, gap, and

underbody drag reduction. Use of a simple base flapat the trailing edge of the trailer, side extenders orsplitter plate at the tractor-trailer gap, and a skirt or asimple short underbody wedge should provide dragreduction exceeding 25%. At highway speeds, fuelsavings around 12% should be recognized for a 25%reduction in drag. This would represent a savings of$3 billion/year in the United States (pre-2004 fuelprices).

The highly successful testing program has provideddetailed data for computational validation, resultedin guidance on device concepts, and establishedwind-tunnel testing guidelines. The detailed dataexceed what is typically available for careful codevalidation in a relatively complex flow and is thus ofinterest to the general fluid dynamics/aerodynamicsresearch and development community. The state-of-the-art in Particle Image Velocimetry (PIV) wassignificantly advanced in the efforts at the NASAwind tunnels. With the Ground TransportationsSystem (GTS) model in the 7-ft × 10-ft wind tunnel,NASA succeeded in being one of the first to use athree-dimensional (3D) PIV system in a productionwind tunnel. To use PIV in the 12-ft pressure windtunnel with the Generic Conventional Model(GCM), a new and innovative approach thatprovided remote control of the PIV system wasdeveloped. With this remote system, the tunnel wasnot opened — a costly and time-consumingprocedure — to re-position cameras.

The computational flow modeling has providedguidance in model definition, mesh refinement, andchoice of turbulence model for heavy vehicles.Computations have been used for both theevaluation of flow physics and to guide theconceptual design of devices. For example, it wasdemonstrated computationally that a splitter platethat partially closes the tractor-trailer gap isadequate to maintain the desired reduced drag,maintain symmetric flow condition, and avoid gapblow through. Previous designs assumed that fullgap closure from the tractor to the trailer wasnecessary.

The Program has successfully established industrycontacts and collaborations and internationalrecognition in the academic community. The 1st

International Conference on the Aerodynamics ofHeavy Vehicles: Trucks, Busses and Trains, which


7

was led by the DOE Aero Team, attracted world-renowned researchers and developers fromacademia, as well as significant industry interest andparticipation. It should also be emphasized that bycombining the best of academia and government labcapabilities, technical developments have beenleveraged across programs within DOE Labs,NASA, and university programs and DOE’s HeavyVehicle Program milestones have been delivered.Examples are the progress in the state-of-the-art in3D PIV, advances in turbulence modeling with theuse of hybrid RANS and LES models for efficientand accurate flow modeling, and the use ofbroadcast fuel rates during real-time, full-scaletesting. The Program’s achievements also includethe Team’s many publications and record-of-inventions or patents that have resulted from theDOE Heavy Vehicle Program work.

Future PlansFuture new areas being investigated are wheel andwheel well aerodynamics related to brake coolingand tire splash and spray, as well as an entirely newrelated area of investigation involving the evaluationof coal car aerodynamics with the objective ofidentifying drag reduction devices for filled andempty cars.

The Team will continue with its computational effortwhile enhancing its full-scale testing effort incollaboration with fleet owners and manufacturers.Substantial efforts to establish contacts with thefleets are planned. For example, the Teamrepresentative, Jim Ross, was invited to participatein a panel at the TMA meeting in NashvilleTennessee this September. The focus topic of thepanel is the recognized increase in fuel use duringthe cold weather season. Jim used this opportunity toshare the Team’s findings with fleet owners andoperators and seek their feedback on ways to getaero device technology on the road.

To successfully get aerodynamic devices on theroad, full-scale testing in collaboration with fleetowners and operators is needed. Testing locationsthat have been thus far utilized by the Team are theTRC in Ohio and Crow’s Landing in California.Both controlled and long road tests are needed atspeeds at or exceeding 65 miles per hour.

Computations of rotating wheels and investigationsinto the influence of underbody flow are planned andare recognized areas of interest to industry. Thiseffort is in addition to moving forward on full-vehicle simulations with advanced models. UnsteadyRANS and hybrid LES/RANS modeling of the fullGCM vehicle with comparison and analysis of the12-ft NASA Pressure Wind Tunnel data is planned.It is important to determine if unsteady RANS andhybrid RANS/LES turbulence modeling can captureprimary flow features. Guidelines for steady RANShave been openly shared and are available toindustry. Unsteady RANS and hybrid models needto be assessed for grid sensitivity and boundaryconditions. This assessment will provide specificguidelines for computations with advanced modelsand will assist in the further conceptual design ofdrag-reduction devices and an integrated vehicle.

Tractor manufacturers are also interested incomputational results for specific commercial tools,but it should be noted that guidelines for use ofspecific turbulence modeling approaches is notdependent on choice of computational tool.Currently, the National Lab participants are utilizingcommercial and NASA codes, as well as their ownin-house tools. There are advantages to each. For theaddition of new models and for response to R&Dissues, especially on large parallel machines forinvestigating model performance, the NASA and in-house tools provide the quickest and most flexibleapproach. However, for geometry and meshgeneration, the commercial tools tend to providesome desirable options.

It is recognized that further fuel savings are possibleand vehicle safety can be enhanced by leveraging theaccomplishments of the DOE Aero Team toinvestigate an integrated heavy vehicle system. Theeffect of aerodynamics on brake cooling and enginecooling will be considered. Air control for improvedbraking and engine performance is currently a highpriority for industry. Also, the initiated efforts inwheel, tire, and vehicle splash and spray willcontinue. This splash and spray investigation willprovide an understanding of this multiphase flowphenomena, thus leading to conceptual designs formitigation of splash and spray for improved vehicleand highway safety. Published research anddevelopment in the open literature appears to lackinformation in this area of interest.


8

The Team is also planning to continue its pursuit toimprove aerodynamics and reduce fuel use areaswith flow regimes similar to those of heavy vehicles.This year’s experiments and computations of railwaycoal cars have demonstrated the substantial increasein drag from full to empty railcars. The aerodynamicdrag of an empty railcar in the wind tunnel is 32%and 42% at 0 and 10° yaw, respectively, comparedwith that for a full railcar. We plan to continueworking with our contact, Jim Hart, of JohnstownAmerica Corporation (Johnstown, Pennsylvania) forguidance in conceptual designs that are automaticand durable for the 20-plus-year life of a coal car.Experiments and computations have been used forthe smart design of drag mitigating devices. Theseconceptual designs will condition the flow so thatthe empty car will mimic the flow of a full car,providing substantial fuel savings.


9

B. A Study of Reynolds Number Effects and Drag-Reduction Conceptson a Generic Tractor-Trailer

Principal Investigator: Bruce L. StormsAerospaceComputing, Inc.M/S/260-1NASA Ames Research CenterMoffett Field, CA 94035(650) 604-1356, fax: (650) 604-4511, e-mail: [email protected]

Field Project Manager: James C. RossNASA Ames Research CenterM/S/260-1Moffett Field, CA 94035(650) 604-1356, fax: (650) 604-4511, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4289, e-mail: [email protected]

ParticipantsDale R. Satran, James T. Heineck, Stephen M. WalkerNASA Ames Research Center

Contractor: NASA Ames Research CenterContract No.: DE-AI01-99EE50559

NASA’s effort consists of two experimental focus areas:

• A Study of Reynolds Number Effects and Drag-Reduction Concepts on a Generic Tractor-Trailer Objective

• An Experimental Study of Aerodynamic Drag of Empty and Full Coal Cars

The following describes the objectives, approach, accomplishments, and future direction for each of these focus areas.

Objectives• To investigate Reynolds-number effects on the flow field and resulting aerodynamic forces generated by a 1:8-

scale model of a class-8 tractor-trailer configuration.

• To provide quality experimental data on a simplified tractor-trailer geometry for CFD validation.

Approach• Vary the total pressure of the wind tunnel, thereby varying the Reynolds number from 500,000 to full-scale

values over 6 million, on the basis of trailer width.

• Measure the forces and momentum surface, pressure distribution, and off-body flow. Measurements were madeat various yaw angles to study the influence of crosswind and to calculate wind-averaged drag coefficients.

• Several drag-reduction concepts were studied to document their potential benefit as well as their Reynolds-number sensitivity.


10

Accomplishments• CFD validation data are now available for use by interested industry and government researchers.

• Reynolds number effects were found to be relatively small above a value of ~1 million. Care should be taken ininterpreting smaller-scale data.

• The results of the study were presented at the 34th AIAA Fluid Dynamics Conference and Exhibit in Portland,Oregon, on July 1, 2004 (paper number AIAA-2004-2251).

Future Direction• Additional drag-reduction devices will be examined for under-body flow control/drag reduction.

• Results from this study will be made more widely available via CD/DVD distribution to interested parties.

IntroductionFor a typical heavy vehicle at a highway speed of70 mph, the energy required to overcomeaerodynamic drag is about 65% of the totalexpenditure (which also includes rolling friction,transmission losses, and accessories). By altering thevehicle shape, it has been estimated that moderntruck drag coefficients may be reduced by up to50%, resulting in an annual national fuel savings ofthree billion gallons. This large potential savings,coupled with increasing fuel costs, has spurredrenewed interest in heavy-vehicle aerodynamics.

Recently, a series of experimental andcomputational studies has been funded by theDepartment of Energy. With the goal of CFDvalidation, the experimental efforts have focused onsimplified geometries at 1:8-scale and below. Earlyexperiments focused on the simplified geometry ofthe Ground Transportation System (GTS) modelrepresentative of a class-8 tractor-trailer with a cab-over-engine design. A 1:8-scale GTS model with notractor-trailer gap and no wheels was first studiedwith the addition of several ogival boattails andslants to the base of the trailer. The largest overalldrag reduction of 10% was obtained by an 8-ft ogiveconfiguration (full scale). The addition of boattailplates to the same model resulted in a 19% dragreduction, and particle image velocimetry (PIV)measurements behind the trailer document asignificant reduction in the wake size due to the flowturning provided by the plates. Variation of thetractor-trailer gap on a 1:15-scale model at zero yawrevealed relatively constant drag on the tractor whilethe trailer drag increases by a factor of three as thegap increases from zero to 1.55*√A. Reynolds-

averaged Navier-Stokes computations of thisgeometry include both grid-size and turbulence-model studies.

The Generic Conventional Model of the currentstudy was previously investigated at a Reynoldsnumber of 1.1 million in the NASA-Army 7-ft ×10-ft wind tunnel. The results include forces andmoments, surface pressures, and 3-D particle-imagevelocimetry. Measurements of two tractor-trailergaps (40 and 80 inches full scale) indicatedsignificantly greater drag for the larger gap at lowyaw angles (between ±4°) and reduced drag athigher angles. Several drag-reduction concepts wereinvestigated, including tractor side extenders,boattail plates, and a trailer belly box. Comparisonsof PIV data were presented in the tractor-trailer gapwith and without side extenders and in the trailerwake with and without boattail plates.

Experimental SetupThe investigation was conducted in the 12-FootPressure Wind Tunnel located at the NASA AmesResearch Center. This facility can be pressurizedfrom 0.25 to 6 atmospheres at Mach numbers from0.1 to 0.5. The test section has a circular crosssection 12 ft in diameter with four 4-ft-wide flatsurfaces centered about the horizontal and verticalcenterlines. A ground plane was installed 21 in.above the tunnel floor, providing a flat surface 10 ftwide and 18 ft long. Pressure taps were located onboth the ground plane (2 rows of 64 taps) and thetest-section walls (8 rows of 30 taps). A fairing wasinstalled to isolate the model-support hardware fromthe air stream, and speed-correction probes wereused to correct the facility speed as a result of the


11

blockage of the ground plane and fairing. There wasalso an additional pitot-static probe installed on theupper-left ceiling to measure the free-streamconditions in the test section. All of the datapresented are referenced to the Mach number basedon a wall tap at location 6.17 ft forward of the centerof rotation at an azimuth of 60° from vertical (twoo’clock looking downstream).

A photograph of the GCM baseline configurationinstalled in the wind-tunnel test section is shown inFigure 1. This 1:8-scale model is representative of ageneric class-8 tractor-trailer with the engine in frontof the cab. Designed for CFD validation, the modelincludes a number of geometry simplifications inorder to facilitate grid generation and avoid theassociated flow complexities. In particular, no effortwas made to duplicate the complex geometry of theundercarriage of either the tractor or trailer (bothwere approximated by flat surfaces). Similarly, thewheel wells of the tractor were not modeled, andonly the portion of the wheels below the tractorlower surface was included. Also, the tractorgeometry (designed by the Calmar Research Corp.)is a streamlined shape representative of a moderntractor design while omitting most small-scalesurface details and flow-through components. Thetrailer measures 45 ft in length (full scale) with

Figure 1. The Generic Conventional Model installedin the 12-Ft Pressure Wind Tunnel.

rounded front vertical edges (8-in. full-scale radius).The tractor-trailer gap for this study was heldconstant at the full-scale equivalent (40 in.). TheGCM was attached to the model-support hardwareby using four vertical posts that were 1.75 in. indiameter. The four posts were non-metric, with0.030 in. of clearance as they passed through thetrailer floor. The model was mounted with its wheels0.15 in. above the ground plane and centeredlaterally in the tunnel. The center of rotation of themodel was located 54.36 in. aft of the tractor frontbumper. The model frontal area of 1.6623 ft2 gives asolid blockage of 1.5%.

The overall model loads were measured by using asix-component internal balance that was mountedinside the trailer. The manufacturer-specifiedaccuracy of the internal balance in the axial (drag)direction was ±1 lb, but repeat runs indicated theexperimental uncertainty to be on the order of±0.5 lb. The tractor was suspended from the trailerthrough a set of flexures and two load cells thatmeasure the drag and yawing moment of the tractoralone. The specified accuracy of the load cellswas ±0.45 lb. The model was instrumented with200 pressure taps on the tractor and 276 taps on thetrailer. There were also 12 unsteady pressuretransducers mounted on the tractor rear surface,trailer front surface, and the trailer rear surface. Athree-component PIV system was used to obtainhorizontal-plane velocity measurements in thetractor-trailer gap. Details of the PIV systeminstallation are presented in Ref. 12. Pressuremeasurements will be documented in a future report.

The model was yawed through a range of anglesbetween ±14°. Except where noted, all data wereacquired at a Mach number of 0.15, which allowedfor Reynolds number studies with no Mach-numbereffects. With the tunnel pressurized to sixatmospheres, the Reynolds number was over6 million, on the basis of the trailer width, which iscomparable to a full-scale truck driving at 75 mph.At one atmosphere, the Reynolds number was1 million, based on the trailer width, which iscomparable to a full-scale truck driving at 15 mph.

Various add-on drag-reduction devices were testedon the bases of both tractor and trailer, as well as onthe trailer under-carriage. In this report, results willbe presented for tractor side and roof extenders,


12

trailer boattail and base flaps, and trailer belly boxand skirts. Details of each device will accompanythe discussion of the associated results.

Results and DiscussionThe results presented below detail the body-axisaxial force (drag) coefficient for the tractor-trailercombination and its components. This dragcoefficient represents the force along the axis of thevehicle in the direction of travel. The internalbalance also provided lift, side force, and momentmeasurements, which will not be discussed. Withthe objective of CFD validation, no wall correctionswere applied to the data, and all coefficients werecalculated on the basis of the static pressure at aknown point in the test section (as detailed above).Without wall corrections, the computed dragcoefficients will differ from those of the equivalentmodel in free air. However, the measureddifferences between configurations should berepresentative of the effects of the associatedgeometric modifications.

The drag data presented herein were acquired forincreasing yaw angle, unless otherwise noted. Usingthe variation of drag with yaw angle, wind-averageddrag coefficients were computed by using the SAERecommended Practice of Ref. 13. This practiceassumes that the mean wind speed in the UnitedStates of 7 mph has an equal probability ofapproaching the vehicle from any direction. Thismean wind speed and the vehicle velocity were usedto calculate a weighted average of the dragcoefficient at various yaw angles. The wind-averaged drag coefficients reported in this paperwere computed for a highway speed of 55 mph.

A. Baseline ConfigurationThe baseline geometry for this study isrepresentative of a modern tractor design with thestandard aero package less the side or roofextenders. Yaw angle sweeps for increasing anddecreasing angle revealed significant hysteresis inthe resulting drag measurements (Figure 2). Sincethe drag curve of a simplified geometry with no gapis relatively continuous, the discontinuities at highyaw angles are likely due to changes of the flowstructures within the tractor-trailer gap. The effect ofReynolds number is most evident at the yaw angles

0.3

0.4

0.5

0.6

0.7

0.8

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Re = 6.2 millionincreasing yawdecreasing

Re = 1.1 millionincreasing yawdecreasing yaw

PIV data, -10 deg

PIV, -10.5 deg

PIV data, 0 deg

crosswind = - 7 mph 7 mph

-> Yaw range for wind-averaged drag <-

Figure 2. Baseline hysteresis of drag coefficient for twoReynolds numbers.

greater than 8° where the peak drag and hysteresisare notably different. At lower yaw angles, however,the differences are relatively small and the hysteresispaths are nearly duplicated. For a mean crosswind of7 mph (-7.2 < ψ < 7.2) as shown in Figure 2, thewind-averaged drag coefficients at Reynoldsnumbers of 1 and 6 million (0.582 and 0.578,respectively) differed by less than 1%.

A three-component PIV system was employed forvelocity measurements in the tractor-trailer gap. Thevelocity vector maps (Figures 3–5) are averages of100 discrete measurements acquired at 2 Hz inintermittent bursts (due to computer limitations)over a period of three minutes. Although data wereacquired at 1/4, 1/2, and 3/4 trailer heights, theresults presented in this report are limited to thetopmost location. In these figures, the direction andmagnitude of the vectors indicate the in-planevelocities, while the color map indicates the out-of-plane (vertical) velocity. The velocity vector mapsfor zero yaw (Figure 3) reveal two counter-rotatingrecirculation regions in the gap and suggest minimalsensitivity to Reynolds number. The yaw angle of–10° (Figure 4), however, is in the hysteresis regionof the drag curves where there are significantdifferences between two Reynolds numbers. Bothmeasurements were obtained for decreasing yawangle that correspond to the upper curves, asindicated in Figure 2. Similar to the zero-yaw case,the flowfield at Re = 5 million exhibits tworecirculation regions, but they are strongly


13

Figure 3. PIV measurements in the tractor-trailer gap atyaw = 0°. Top view of horizontal PIV data plane at0.75 h.

Figure 4. PIV measurements in the tractor-trailer gap atyaw = -10°.

Figure 5. PIV measurements in the tractor-trailer gap forhigh- and low-drag states near -10°. Top view ofhorizontal PIV data plane at 0.75 h.

asymmetric and with much higher velocities. AtRe = 1 million, only the tight recirculation on thewindward (left) side is present, and there aresignificantly greater crossflow and downwardvelocities compared to the higher Reynolds number.The corresponding drag coefficient for Re = 1million is 2% higher than that for Re = 5 million.The lower measurement locations revealed similarflow structures to the topmost height, but withreduced Reynolds-number sensitivity.

The PIV data also provide insight into the drag-curve hysteresis. A look at two neighboring yawangles near –10° reveal drastically differentflowfields at one Reynolds number (Figure 5). Asindicated on the drag curves (Figure 2), these twomeasurements correspond to yaw angles of –10°(decreasing) and –10.5° (increasing) at Re = 1million. For the yaw angle of –10.5°, the windwardrecirculation region is absent, and the velocities inthe gap are significantly lower than the –10° case. Itis hypothesized that lower gap velocities yieldhigher pressures on the back of the cab (less drag)and a smaller separation region on the leeward(right) side of the truck (not visible in PIV images).As a result, the drag of the –10.5° case is 28% lowerthan that of the –10° case.

B. Side and Roof ExtendersSimilar to the components of a modern tractor aeropackage, side and roof extenders were attached tothe rear of the tractor, as shown in Figure 6. Theextenders were 1/8-in. thick (model scale), with fourdifferent lengths ranging from 30% to 60% of thetractor-trailer gap. For a length of 60% gap (asshown), the total drag coefficient as a function ofyaw angle (Figure 7) illustrates the dramatic effectof the extenders. At yaw angles less than 2°, the dragreduction is minimal, while at higher angles, thereduction increases dramatically to 35% at 10°. At aReynolds number of 6 million, the wind-averageddrag coefficient of 0.422 for the 0.6 g extenders was27% lower than that of the baseline withoutextenders. The low Reynolds-number measurementswere marginally higher than those of the full-scaleReynolds number until about 10° yaw, at whichpoint the curves cross and the low-Re drag drops offat the highest yaw angle of 14°. PIV data for thismodel with extenders were acquired previously inthe NASA Ames 7- × 10-ft wind tunnel at a


14

a) Baseline configuration

b) Side and roof extenders (0.6 g)

Figure 6. Close-up of tractor-trailer gap with and withoutside and roof extenders.

0.3

0.4

0.5

0.6

0.7

0.8

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Baseline, Re = 6.2 million

Side and Roof ExtendersRe = 1.1 million

Baseline, 1.1 million

ExtendersRe = 6.4 million

Figure 7. Effect of tractor side and roof extenders (0.6 g)on total drag coefficient.

Reynolds number of 1 million. The results indicate asignificant reduction in the gap cross flow due to thepresence of the side and roof extenders.

The wind-averaged drag reduction provided by theextenders is plotted as a function of extender lengthin Figure 8. For the four extender lengths testedbetween 30% and 60% gap width, there is aconsistent trend of increasing drag reduction withincreasing extender length. More specifically, thedrag reduction increases from 25% to 27% for anincrease in extender length from 30% gap to 60%gap. Although greater extender lengths may beimpractical from an operational standpoint, thesedata suggest that additional drag reduction may beobtained by completely blocking the gap crossflow.As demonstrated in Ref. 2, a centerline gap seal canbe more effective than the side extenders of astandard tractor aero package.

To determine the effect of Reynolds number, thefacility total pressure was varied while the Machnumber was held constant. The change in wind-averaged drag coefficient with Reynolds number(Figure 9) reveals differing sensitivities for thebaseline and extender configurations. In this figure,the error bars on the baseline data points show themagnitude of experimental uncertainty due to bothmeasurement resolution and repeatability. Error barswere of the same magnitude for the extenderconfiguration, but they were omitted from the figurefor clarity. For the baseline, the drag coefficient isobserved to increase by an average of 0.006 (1%) forReynolds numbers less than 4 million. The extenderconfiguration, however, does not indicate asignificant increase until below 3 million, with a

-0.16

-0.155

-0.15

-0.145

-0.140.2 0.3 0.4 0.5 0.6 0.7

²CD r

elat

ive

to b

asel

ine,

win

d-av

erag

ed

Extender length / tractor-trailer gap

Figure 8. Wind-averaged drag reduction due to side androof extenders at Re = 6 million.


15

-0.005

0

0.005

0.01

0.015

0.02

0.025

0.03

0 1 2 3 4 5 6 7

CD

- C

D(R

e=6E

06),

win

d-av

erag

ed

Reynolds Number, million

Extenders

Baseline

Figure 9. Reynolds-number sensitivity of wind-averageddrag for baseline and side extenders.

dramatic increase, close to 0.03 (6.9%), atRe = 500,000. The drag curves for the extenderconfiguration at several Reynolds numbers(Figure 10) illustrate a significant increase in dragthat increases with crossflow at Re = 500,000. Thisincrease is likely due to flow variations in thevicinity of the trailer leading-edge curvature, whichis more sensitive to Reynolds number than to sharpcorners.

C. Aerodynamic Boattail PlatesSince the side and roof extenders are first-generationdrag-reduction devices common to most moderntractor aero packages, the effect of the boattail plates(and all subsequent devices) will be measuredrelative to the extender configuration. Aerodynamicboat-tailing devices have several different variations,but boattail plates typically refer to panels mountedperpendicular to the trailer base and inset from theedges of the trailer. In this case, the same boattailplates that were studied on the simplified GTSmodel5 were applied to the rear of the GCM trailer,as shown in Figure 11. The plates extended 3.75 in.from the end of the trailer and were inset 0.625 in.from the sides and top of the trailer. The bottomplate was mounted flush with the bottom of thetrailer.

Relative to the side and roof extenders, the boattailplates significantly reduced the drag by a relativelyconstant margin between ±10° (Figure 12). At Re =6 million, the wind-averaged drag coefficient

0.35

0.4

0.45

0.5

0.55

0.6

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Re = 1.1 million

Re = 3.2 millionRe = 6.4 million

Side and Roof Extenders

Re = 0.5 million

Figure 10. Reynolds-number sensitivity of side and roofextenders on drag curves.

Figure 11. Aerodynamic boattail plates installed on thebase of the trailer.

0.3

0.35

0.4

0.45

0.5

0.55

0.6

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Re = 6.3 millionRe = 1.1 millionExtenders + Boattail

Extenders,Re = 6.4 millionRe = 1.1 million

Figure 12. Effect of boattail plates on total dragcoefficient.


16

was 0.364, which is 13.7% less than the extenders-only configuration. Relative to the high-Reynolds-number case, the drag curve for the boattail plates atRe = 1.1 million exhibited the same roll off at highyaw angles as that of the side extenders alone. Theresults were mixed at lower angles, with the low-Recurve lower than the high-Re curve at negative yawangles and higher at some positive yaw angles.These mixed results are evident in the resultingwind-averaged drag coefficients, which differ byonly 0.6%.

D. Trailer Base FlapsAnother method of aerodynamic boat-tailing is whatwill be referred to as base flaps. In this embodiment,the panels are attached to the edges of the trailerbase and angled inward. In the current study,measurements were made for a base-flap length of3.125 in. (25 in. full scale) at angles ranging from 0to 28°. The installation photo (Figure 13) shows thebase flaps with a 20° deflection mounted on the rearof the trailer. Note that the linkages connecting theflaps to the base were designed for easy anglechange and are not representative of the full-scalehardware.

The effect of the base flaps on the total dragcoefficient is presented in Figure 14 for a flap angleof 16°. Relative to the side and roof extenders alone,the addition of the base flaps provides significantdrag reduction that marginally increases with yawangle. The Reynolds-number sensitivity of the baseflaps is minimal for most yaw angles between ±10°.As in the previous configurations, the drag for Re =1.1 million rolls off at the higher yaw angles, whilethe drag at full-scale Reynolds number continues toincrease.

The effect of base-flap angle on wind-averaged dragreduction is presented in Figure 15. Differentsymbols are used to indicate the data from twoseparate wind-tunnel entries (four months apart) ofthe same model in the NASA Ames 12-Ft PressureWind Tunnel. The lower Reynolds number on thesecond entry was due to limitations of the coincidentPIV measurements. Although there is anunexplained offset in the two curves (marginallylarger than the experimental uncertainty), the trendsindicate that the optimum base-flap angle is around

Figure 13. Base flaps installed on trailer base (flap angle= 20°).

0.3

0.35

0.4

0.45

0.5

0.55

0.6

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Re = 6.3 millionRe = 1.1 million

Extenders,Re = 6.4 millionRe = 1.1 million

Extenders + Base flaps

Figure 14. Effect of 16° base flaps of total dragcoefficient.

-0.1

-0.08

-0.06

-0.04

-0.02

0-5 0 5 10 15 20 25 30

²CD r

elat

ive

to e

xten

ders

, win

d-av

erag

ed

Base flap angle, deg

2nd entry, Re = 4.9 million

1st entry, Re = 6.3 million

Figure 15. Effect of base-flap angle on wind-averageddrag reduction.


17

20°. The wind-averaged drag coefficient with20° base flaps is 0.340, which is 19.4% less than theextender configuration and 6.6% less than theboattail-plate configuration. Previous small-scaleexperiments2 at a Reynolds number of 1.25 millionyielded an optimum base-flap configuration with anangle of 15° and a full-scale panel length of 20inches. The difference between these twoexperiments is likely due to the differences in thebase-flap lengths since the Reynolds-number effectswere observed to be minimal. However, the effect ofReynolds number might be more significant athigher flap angles where the flow reattachment onthe base flaps is more sensitive.

E. Trailer Belly Box and SkirtsSeveral drag-reduction concepts were studied, withthe goal of reducing the lateral flow under the trailer.As in previous studies, trailer side skirts consisted offlat panels extending downward from the sides ofthe trailer between the end of the tractor bed and thefront of the trailer wheels. The full-skirtconfiguration extended to the rear of the trailercovering the trailer wheels and included lateralpanels at the trailer base and behind the tractor bed.The trailer belly box (Figure 16), so named becauseof its resemblance to the design of moving trailers,was identical to the full-skirt configuration, with theaddition of a horizontal surface at the bottom of theskirt to form a box on the undercarriage of thetrailer. All of these configurations had full-scale(11.8 in.) ground clearance.

The varying effect of the trailer belly box and skirtson the overall drag is presented in Figure 17 for Re= 6.3 million. Relative to the extender-only case, thetrailer belly box was the most effective, with a wind-averaged drag reduction of 11.8%. The side skirtswere somewhat less effective, with a drag reductionof 6.2%, while the full-skirt configuration increasedthe wind-averaged drag by 3.8%. The increase indrag of the full-skirt configuration is likely due tothe cavity flow that forms inside the skirt. The wind-averaged drag coefficients for these three cases were0.372, 0.396, and 0.438, respectively. Because oftime limitations, a Reynolds-number sensitivitystudy was not performed for these configurations.

a) Baseline (no belly box)

b) Trailer with belly box

Figure 16. Rear view of trailer with and without bellybox.

0.3

0.35

0.4

0.45

0.5

0.55

0.6

-15 -10 -5 0 5 10 15

Dra

g co

effic

ient

Yaw angle, deg

Extenders +Trailer Side Skirts

Side and Roof Extenders

Extenders +Full Trailer Skirt

Extenders + Belly Box

Figure 17. Effect of trailer belly box and skirts on totaldrag coefficient (Re = 6.3 million).


18

ConclusionsFor all configurations, the effect of Reynoldsnumber was most evident at high yaw angles (y > 8 °and y < -8°) where there was a significant reductionin drag at lower Reynolds numbers. However, thisdifference did not significantly affect thecomputation of the wind-averaged drag coefficientsat 55 mph, which uses data at lower yaw angles. Asa result, the variation of Reynolds number for thebaseline revealed an increase in the wind-averageddrag on the order of 1% for Re ≤ 3 million. TheReynolds-number sensitivity for side and roofextenders was more dramatic with an increase inwind-averaged drag of over 7% for Re = 500,000.Limited data for the boat-tail devices suggest thatthe Reynolds-number sensitivity is similar to that ofthe baseline. No Reynolds-number study wasperformed for the undercarriage flow barriers.

The PIV measurements in the tractor-trailer gapdocument significant crossflow velocities andrecirculation regions at yaw angles near 10°.Baseline measurements with reduced drag exhibitedsignificantly reduced crossflow and less-coherentrecirculation regions. Since the tractor side and roofextenders function to reduce the gap crossflow andincrease the tractor back pressure, the addition of theextenders to the baseline reduced the measured dragat all yaw angles. This reduction was dramatic athigh yaw angles (35% at 10°) and minimal at lowangles (2% at 0°). Of the four extender lengthstested (0.3–0.6 gap), the longest extenders were

most effective yielding a 27% reduction in the wind-averaged drag coefficient. Since extenders are astandard component of a modern tractor aeropackage, the effectiveness of the remaining drag-reduction concepts were measured relative to thisextender configuration. Note that no wall correctionswere applied to the experimental measurements inorder to facilitate comparison with CFD simulations.However, the influence of the tunnel walls wasminimized by calculating the tunnel speed on thebasis of the static pressure at a location adjacent tothe model.

Of the boat-tailing concepts (boattail plates and baseflaps), the base flaps were found to be mosteffective. Base-flap angles from 0 to 28° werestudied, resulting in an optimum angle of 20°. Thisresult is higher than a previous low-Reynolds-number study and may suggest some Reynolds-number sensitivity of the base-flap angle. The trailerbelly box was the most effective trailer-undercarriage concept, with almost double the dragreduction of simple side skirts. The belly box andfull skirt configurations were identical except for thelower-surface enclosure of the belly box. The dragof the full-skirt configuration, however, wasincreased while that of the belly-box configurationwas significantly reduced. This result stresses theimportance of eliminating cavity flows to minimizedrag.

For more details and a list of references, please seethe associated technical paper (AIAA-2004-2251).


19

C. An Experimental Study of Aerodynamic Drag of Empty and Full Coal Cars

Objective• To investigate the additional aerodynamic drag experienced by empty coal cars relative to cars full of coal and

methods to reduce the empty-car drag.

Approach• Small-scale wind-tunnel tests of empty and full coal cars and associated drag-reduction devices. 1:87th scale

models were used — drag measured by using a miniature 2-lb linear load cell.

Accomplishments• The drag penalty of empty coal cars relative to full cars varies from 35% to 42%, depending on the direction of

the prevailing wind (0–10°).

• Initial tests of vertical plates in the empty car showed a ~50% reduction in the drag penalty for empty cars.

Future Direction• Additional drag-reduction devices will be examined in collaboration with a coal car manufacturer and Lawrence

Livermore National Laboratory researchers.

As part of a DOE-sponsored study, the aerodynamicdrag of 1/87th-scale coal cars was measured in awind tunnel. The goal of the experiment was tomeasure the difference in drag between full andempty cars to determine if drag mitigation efforts arewarranted.

The measurements were made in the NASA-Ames15-in. × 15-in. low-speed wind tunnel. This is anopen-circuit, suction-type tunnel with at square testsection measuring 60 in. in length. Five coal carswere mounted on a scale train track (Figure 1), withthe middle car connected to the upwind car by a 2-lbload cell (Figure 2) and disconnected from thedownwind car. All cars (except the middle car) wereglued to the track to prevent motion along the track.This configuration was tested at a freestreamvelocity of 65 m/s (145 mph) with and withoutsimulated coal. The tunnel speed was chosen tomaximize the measured drag and minimizemeasurement uncertainty. The five-car combinationwas tested at yaw angles of zero and 10° todetermine the effects of crosswind. The blockage ofthe coal cars at zero yaw was 0.9%.

The small- and full-scale Reynolds numbers of thecoal cars (assuming full-scale velocity of 60 mph)

Figure 1. Five coal cars mounted in wind-tunnel testsection. Pitot-static probe used for tunnel speedmeasurement is visible above first car near the top of thephoto.

are 0.16 and 5.8 million, respectively. Because ofthe nature of bluff-body flow fields, this differencein Reynolds number is not expected to significantlyaffect the experimental results. The drag coefficientfor each configuration (Table 1) was calculated onthe basis of the square root of the zero-yaw frontalarea. The uncertainty based on measurementresolution and repeatability was 1.2–2.5%,


20

Figure 2. Load cell connecting metric car (at left) toupwind car (at right). The load cell is the thin verticalstrip at the center of the photo. Instrumentation wiringenters upwind car in the white cable. Simulated coal isvisible in both cars.

Table 1. Drag coefficients for full and empty coal cars at0 and 10° yaw (empty car is reference at each yaw angle).

Yaw (°) Cd empty Cd full %diff0 0.315 0.216 31

10 0.519 0.300 42

depending on the magnitude of the dragmeasurement. As indicated below, the drag on thefull coal car was significantly less than that of theempty car (32%). This difference was even greater(42%) for the 10° yaw condition in which the dragwas significantly increased relative to the zero-yawcase. These results suggest that significant fuelsavings could be obtained by reducing the drag ofthe empty coal cars.

Vertical dividers were installed to manipulate theflow inside the cars, with the goal of making theexternal flow more like it is when the car is loadedwith coal (Figure 3). The dividers had a significanteffect on the drag. The single divider reduced theempty-car drag by 15% at zero yaw. The threedividers reduced the empty-car drag by 21%. Morework is needed to identify the overall best approachto drag reduction. This effort will be considerablyhelped by the concurrent CFD being performed atLawrence Livermore National Laboratory.

Figure 3. Vertical plates installed in model cars. Note single and three evenly spaced vertical dividers.


21

D. Experimental Measurement of the Flow-Field of Heavy Trucks

Principal Investigator: Fred BrowandAerospace & Mechanical Engineering, University of Southern CaliforniaRRB 203, Los Angeles CA 90089-1191(213) 740-5359, fax: (213) 740-7774, e-mail: [email protected]


Contractor: National Energy Technology Laboratory, Pittsburgh, PA 15236-0940Contract No.: DE-AC03-98EE50512

Objective• Improve the performance of heavy trucks by reducing aerodynamic drag and by increasing safety.

Approach• Identify areas where improved use of aerodynamic design could decrease truck drag and consequently improve

fuel economy. These include:

– Gap between tractor and trailer — We have identified the importance of cross-gap flow as a source of dragincrease.

– Trailer base — We have performed wind tunnel testing to evaluate the performance of various flappeddevices to close the trailer-base wake more efficiently.

AccomplishmentsI. Minimizing drag due to cross-gap flow• Wind tunnel flow field studies document the appearance of violent cross-gap flows under certain conditions.

• The cross-gap flow has been studied by means of a special analysis procedure entitled Proper OrthogonalDecomposition.

• Suggestions are made to minimize this unwanted cross-gap flow by means of a simple splitter plate attached tothe front of the trailer and partially covering the gap.

II. Field measurement of fuel savings using a trailer-base add-on• Field tests directly measure fuel consumption of a Class-8 tractor and trailer with and without flaps attached to

the sides and top of the trailer. The most promising flap geometry for field test was chosen on the basis of windtunnel tests.

• The field test documents fuel savings of about 4%; to our knowledge, these savings are the best that have beenachieved for a simple, passive, trailer-base add-on.

Future Direction• Initiate a program and a new testing apparatus to study wheel/tire splash and spray.


22

This report describes the progress we have made ontwo separate aerodynamic problems (I) describingthe sensitivity of the drag to the geometry of the gapbetween tractor and trailer and (II) providing fieldtest results of high quality to demonstrate the fuelsavings to be realized for flaps attached to the baseof a trailer.

I. Minimizing Drag due to Cross-Gap FlowCertain aspects of the tractor-trailer gap flow-fieldhave been described in a paper (“Flow Structure inthe Gap Between Two Bluff-Bodies,” D. Arcas,F. Browand, & M. Hammache, AIAA PaperNo. 2004-2250) prepared for the AIAA meeting inPortland June 28–July 1, 2004. The gap data arepresented in the mathematical format of a ProperOrthogonal Decomposition, which will facilitatecomparisons between the experimental results andthe ongoing numerical computations at LLNL.

This work and earlier studies describing the gapflow (“On the Aerodynamics of Tractor-Trailers,”M. Hammache & F. Browand, Proceedings of theUEF Conference on: The Aerodynamics of HeavyVehicles: Trucks, Buses and Trains, Lecture Notesin Applied and Computational Mechanics,

Springer-Verlag, Heidelberg, September 2004) haveprovided a clear and quantitatively correct picture ofthe flow field within the gap for a variety of flowconditions. An example of the gap flow for an angleof yaw of 6° is shown in Figure 1.

The velocity field within the gap is determined fromDigital Particle Image Velocimetry (DPIV)measurements. Accurate mean-flow streamlines aresuperimposed upon several field quantities plotted incolor in the upper left of Figure 1. On the upper rightof Figure 1 is the mean velocity magnitude; to thelower left is mean, or average, vorticity; and to thelower right is the root-mean-square amplitude of theunsteady fluctuation. The flow crossing the gap canbe seen to be quite large — especially in the vicinityof the front of the trailer. The flow picture suggeststhat a short fin, or plate, protruding from the front ofthe trailer might serve to substantially or entirelydisrupt the cross-gap flow.

Splitter-Plate Gap StabilizationA follow-on experiment, performed by Eric Liu atUSC, shows the effect upon trailer drag of a splitterplate of various lengths placed at the nose of thetrailer in the center-plane. The geometry is shown in

Figure 1. Wind tunnel model of simple tractor-trailer gap flow at a gap of ½ trailer width.The cab outline to the left and the trailer outline to the right are yawed 6° to simulated aside-wind.


23

Figure 2, and photographs of the model in the USCwind tunnel are shown in Figures 3 and 4. The gap,G, is normalized by the square root of the cross-sectional area, √A = 181 mm.

G

632.5

308

214.3

154

y x

z

Figure 2. Geometry of wind tunnel model (dimensions inmm).

Figure 3. Head-on view of model in wind tunnel.

Figure 4. Detail of splitter plate installation.

The results of the test are given in Figure 5, whichshows the drag coefficient for the trailer as afunction of yaw angle for various lengths of splitterplate. There is little change in drag at small yawangles, but as yaw angle increases, the splitter platbecomes more effective. All three splitter platelengths have about the same performance at this gaplength, but at larger gaps (not shown), the longerplates show less improvement and can actually leadto drag increases. For this reason, the shorter platewould seem to be the desirable choice. Figure 6gives drag for the 1/3 gap length splitter plate forthree separate gap lengths with yaw angle as aparameter. The results are quite consistent inproducing a drag saving. We suggest that suchsimple fin geometries be given wind tunnel testing atfull-scale or be subjected to over-the-road testingwhere operation can be evaluated under actual fieldconditions.

0.2

0.25

0.3

0.35

0.4

0.45

0.5

-12 -10 -8 -6 -4 -2 0 2 4

Yaw Angle (degrees)

CD

(dim

ensi

onle

ss)

full length2/3 length1/3 lengthno plate

uncertainty

Figure 5. CD for three splitter plate lengths versus yawangle at G = 0.5 (G = gap/√A).

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7

Gap Length (normalized)

CD

(dim

ensi

onle

ss)

-10 degrees

-8 degrees

-6 degrees

-5 degrees

-4 degrees

-3 degrees

-2 degrees

-1 degree

0 degrees

Figure 6. CD for 1/3 gap length splitter plate as a functionof normalized gap length.


24

II. Field Measurement of Fuel Savings Usinga Trailer-Base Add-On

A very successful field test was performed at theNASA Crows Landing Flight Facility on May 17–22to evaluate the anticipated fuel savings associatedwith base-flap add-ons. USC organized andparticipated in the tests. The base flaps and thetrailer were provided by Norcan AluminumIncorporated, and the tests were conducted withtrucks and personnel from California PATH at UCBerkeley. The data acquired were of high quality fora field test. Testing showed an optimum fuelconsumption saving of about 4% at an optimum flapangle of about 13°, although the entire flap-anglerange from 10° to 16° would provide nearly thesame savings. The modification to the trailer base issimple and relatively inexpensive to implement. Thetest results, and the unique experimental procedurefor the tests, are documented for presentation at theSAE World Congress, April 2005 (paper # 05-B83,“Fuel Savings by Means of Flaps Attached to theBase of a Trailer: Field Test Results,” F. Browand,C. Radovich, & M. Boivin).

The Site at Crows LandingThe present tests are performed at the NASA CrowsLanding Flight Facility at the northern end of theSan Joaquin valley. The main runway isapproximately 2400 m in length and is orientedroughly north-south, as shown in Figure 7.

Truck and TrailerA single Freightliner 2001 Century Class truck isused for the tests. The truck is powered by CumminsN14 Celect engine developing a maximum of350 HP. The truck has an automatic transmission,Allison HD 4060 (six forward gears), and a rear axleratio of 4.63. In operation, the truck executesmultiple runs up and down the runway. A runconsists of an acceleration phase to a predeterminedspeed, a uniform speed phase, and a decelerationphase. First, an achievable speed trajectory isestablished for the truck on the runway. This desiredspeed trajectory is then programmed into an on-board computer that controls the truck and ensuresthat all runs are executed in identical fashion.

Figure 7. Plan view of the NASA CrowsLanding Flight Facility. Red bar marksmeasurement interval on runway.

The Base FlapsBase flaps are attached to the sides and top of therear of the trailer. The flaps are constructed from afiberglass-epoxy-resin material and are one-quarterof the base width in length (about 61 cm, or 2 ft).Figure 8 presents several of the flaps used for thetest. The side-flaps swing on piano hinges that arebolted directly to the rear side-edges of the trailer toproduce a sealed joint. This installation detail isdictated by the need to quickly remove the flaps forthe “no-flaps” control runs.

The flap angle is adjusted by means of twoaluminum supports. Holes are pre-drilled to allowthe five flap angles to be set quickly. In commercialapplication, the flaps are attached along the rear doorhinge lines, so that no gap appears at the jointbetween the flap and the side of the trailer. Also incommercial application, the flaps are constrainedonly by a short length of cable attached to the reardoor. Higher pressures on the trailer base and on theinside of the flap, compared to the stream side of theflap, are sufficient to keep the flap extended athighway speeds.


25

Figure 8. Views of flap installation.

The flap at the top of the rear door is split so thedoors can be opened. The two flap-halves aremounted by means of hinges and are kept in placeby means of adjustable turnbuckles (Figure 8, upperright). The split in the top flap is sealed with ducttape. The “no-flaps” control is obtained bycompletely removing flaps from the sides and tapingthe top flaps against the rear door.

Test ProcedureA typical run sequence starts at a fixed point at oneend of the runway. Computer control is initiated,data acquisition begins, and the truck accelerates.When the programmed acceleration rampterminates, the truck continues along the runway atthe preset cruise speed of 26.8 m/s (60.0 mph).Distance along the track is determined by integrationof the forward speed. At a pre-determined distance,the braking sequence is initiated, and the truck slowsto a stop at the far end of the runway. Dataacquisition stops, and the run file is logged in thecomputer. The truck is turned and made ready forthe return from a second fixed point on the track.Typically, a run and the return run are accomplishedwithin about 6 min. A total of 16 runs, or 8 run-pairs, are accumulated for each flap angle setting(about 45 min of test time) and constitute one dataset. When a data set is completed, the truck isreturned to the garage area, and a new flap angle ispositioned. Setting a new flap angle usually takes20–30 min. Four flap angles — corresponding to

four data sets — are accumulated each day duringthe 6:00–10:30 AM period, when wind andtemperature are most favorable. The total databaseover five testing days, May 17–21, consists of 304runs (152 run-pairs) — or 19 data sets — at the flapangles 10°, 13°, 16°, 19°, and 22°, as well as the no-flaps condition that serves as the control.

A total of 10 variables are recorded for each data set.The variables include time, integrated distance,engine torque, engine rpm, vehicle speed, andinstantaneous fuel consumption (broadcast fuel rate).All of these signals are commonly available on theJ1939 bus.

ResultsPlotted as the solid symbols in Figure 9 are theaveraged results — excluding the four data setsshowing the least internal consistency (data set rms> 1.5%). In addition, the dashed bars give theestimated 99% confidence interval for each flapangle. The confidence interval is determined fromthe standard deviation estimate at each flap angle.A 99% confidence interval suggests that if the testsare repeated under the same conditions, the averagedvalues will lay within the bounds of the confidencebars 99% of the time. While our repeated samplingover the week-long period does not alter thevariability inherent in the data, repeated samplingdoes provide a much more accurate estimate of themean (averaged) values.


26

Figure 9. Fuel consumption savings versus flap angle,various data reduction strategies.

Also shown are the two other data reduction choices— triangles indicate the result when all the data areutilized, and squares indicate the result when onlyruns at low wind intensity are kept. These threeresults are not substantially different — they all liewithin the 99% confidence bounds. Expressed as afraction of the fuel consumption without flaps, theaveraged values are accurate to about ±0.6%.

Implications for Fleet OperationThere is minimum fuel consumption at a flap angleof about 13°, but the minimum is broad — the fuelconsumed at 10° and 16° is only marginally greater.It would be advantageous from an operationsstandpoint to have such a broad minimum.

The savings in fuel consumption arising from theuse of base flaps at a 13º flap angle is1.63 L/100 km, or in gallons and miles,0.693 gal/100 mi. A dollar value can be placed onthe accumulated savings by assuming a price forfuel. Figure 10 shows the results for a fuel price of$2.20/gal for distances of 50,000 and100,000 highway miles traveled.

Figure 10. Potential dollar-savings from use of base flapson a single trailer.

These savings represent the increment associatedonly with the change in drag due to the presence orabsence of flaps. The result will hold for any truckof similar size and shape and engine performanceregardless of the loading of the truck or the rollingresistance.

The horizontal axis in Figure 10 is highway speed.Although the tests presented here are performed at60 mph, the results are easily extrapolated to otherspeeds because the fuel needed to travel a givendistance is quadratic in speed. Dollar savings fromthe use of flaps is greater at higher speeds, becauseaerodynamic drag is a larger fraction of the totalresistance. However, total fuel consumption willincrease with increasing speed.


27

E. Continued Development and Improvement of Pneumatic Heavy Vehicles

Principal Investigator: Robert J. EnglarGeorgia Tech Research Institute (GTRI)Atlanta, GA 30332-0844(770) 528-3222, fax: (770) 528-7077,e-mail: [email protected]

Technology Development Manager: Sid Diamond(202) 586-8032, e-mail: [email protected]

Technical Program Manager: Jules Routbort(630) 252-5065, e-mail: [email protected]

Contractor: Dept. of Energy – National Energy Technology Laboratory, Morgantown, WVContract No.: DE-AC26-02EE50691

Objectives• Previous smaller-scale model wind-tunnel evaluations at GTRI had demonstrated up to 15% reduction in

aerodynamic drag coefficient due to blowing and 10–12% due to the device’s corner rounding for a combineddrag reduction of 25–27%, not accounting for the fuel use by the blower. However, these tunnel results hadtranslated into less-than-expected fuel economy increases during full-scale on-track testing. Reasons for thisdifference need to be determined and improved aerodynamic characteristics identified.

• Continue this pneumatic heavy vehicle (PHV) technology development by employing new wind-tunnel resultsto modify the PHV test trailer. Conduct additional fuel economy evaluations of the blown test vehicle toconfirm improved drag reduction for parasitic energy loss reduction, fuel economy improvement, reducedemissions, and increased safety of operations for Heavy Vehicles.

Approach• Conduct additional experimental evaluations of modified wind-tunnel models to enhance the pneumatic

aerodynamic capabilities of the PHV configurations.

• Use these results to modify the DOE full-scale pneumatic test vehicle.

• Conduct preliminary on-road testing followed by on-track SAE Type-II fuel economy tests of the PHV testvehicle to verify and improve the drag-reduction and fuel-economy-increasing properties.

• Identify pneumatic aerodynamic and geometry improvements to increase fuel economy by an additional factorof 2 to 3 over that exhibited during our earlier Phase I SAE Type-II fuel-economy tests.

Accomplishments• Wind-tunnel tests of our modified smaller-scale PHV model have shown improved drag reductions for the

more-realistic current configuration with real-world components, such as axles, springs, under-ride bar, andjack stands. These tests confirmed that this new blown PHV model could reduce drag coefficient by 31% belowthat of the baseline HV configuration, with major geometry modification and not accounting for fuel use forblowing. New tests also demonstrated that active blowing control can reduce side-wind effects, and that thedevice has the potential to assist in braking and safety of operation for PHVs.

• Improvements thus needed for the Phase II full-scale PHV track test were identified, and the new test trailerdesign and modifications were completed. Primary here were improved fairings leading into the blownsurfaces, improved blowing surfaces, and improved cab gap extenders.


28

• On-road and on-track SAE Type-II testing were completed. Results show increased improvement infuel economy due to both blowing and the physical geometry of the new PHV truck. Results also show FuelEconomy Increase (%FEI) of up to 4.5 % at 65 mph due to blowing only, not accounting for fuel used by theblower.

Future Direction• Conduct additional development and demonstrations of pneumatic aerodynamic drag-reduction, braking, fuel-

economy, and safety of operation techniques to provide a confirming database allowing application of thistechnology to operational PHVs. Interact with DOE and Truck Manufacturers Association’s test and evaluationof these devices on their operational rigs.

IntroductionSince aerodynamic drag is the major component ofHeavy Vehicle (HV) resistance at highway speedand thus strongly impacts related fuel economy,GTRI has been applying advanced aircraftaerodynamic technology whereby blowing is used toreduce that drag generated by these bluff-basedhigh-drag vehicles. Using the pneumaticaerodynamic technology known as CirculationControl [Ref. 1] and certain geometry changes, wehave been able to reduce the drag coefficient (CD) inHV models due to blowing by up to 15% and toreduce the CD due to the device’s corner rounding inHV models by 10–12% for a total reduction of25–27%, not accounting for fuel use by the blower(see Figure 1), during a 5-yr tunnel test program forDOE [2, 3, 4]. We could also increase drag asneeded for braking during downhill operationwithout any moving aerodynamic parts by blowingonly select trailing-edge surfaces on the trailer. We

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

CD

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20Momentum Coefficient, Cµ

Blown Heavy Vehicle Drag Modifications,Wheels on,q=11.86 psf, V=70 mph, ψ=0°, α=0°

CD=0.627

CD=0.824

Top & Bottom Slots Only=Drag Increase for Braking

All 4 Slots Blown=Drag Reduction

Faired UnblownBaseline,No Gap, Square LE+TE

Unblown Baseline,Unfaired, Full Gap

2 Side Slots Only=Anti-Yaw for Stability

Blown Truck,Low Cab,No Gap, Round LE,0.375"R, Plates

0.5 psig1.0 psig

∆CD due to Gap

∆CD due to RoundedLE & TE

Figure 1. Drag reduction or dragincrease demonstrated by earlierGTRI GTS generic model PHV,depending on blowing slot activated.

could potentially reduce the huge drag increase andloss of stability that occur when an HV experiencesside winds or gusts. This multi-function potential ofthe blown configurations is seen in the wind tunneldata of Figure 1. Possible compressed air sources arean HV tractor’s turbocharger or an auxiliary enginesimilar to a refrigeration unit.

Full-scale fuel economy tests were conducted [3, 4]during our previous DOE program. These SAEType-II official test-track results on a Pneumatic HV(PHV) configuration were somewhat different fromthe tunnel models, which showed a measured %FEIof only 4–5%, not accounting for energy use forblowing. Since that first SAE test, the currentprogram has thus concentrated on determining thedifference between wind-tunnel results and the less-than-expected full-scale performance, correcting theblown configuration problem areas, and preparingfor a second fuel-economy evaluation with theimproved PHV vehicle.

Experimental Details and Results forUpdated Wind-Tunnel ModelExperimental wind-tunnel developments of thistechnology conducted on a smaller-scale PHVmodel under previous DOE funding [2, 3, 4, 5] hadled to two full-scale on-road Tuning Tests plus anSAE Type-II Fuel Economy Test conducted at the7.5-mi test track at the Transportation ResearchCenter in Ohio, with the results reported above.However, the wind-tunnel model employed here (thevery generic Ground Transportation System, GTS,configuration modified with our blowing systems[2, 3]) was geometrically quite different from theactual on-road and on-track test PHV configuration.This simple model had generated drag reductions ofup to 84% relative to a stock trailer configuration


29

[5]. Since the fuel economy increases from thesedrag reductions were found to be less for the track-test PHV vehicle than those predicted by the tunneldata, we returned to the tunnel this past year todetermine the reasons and possible corrections on amodel modified to be very similar to the full-scaleblown test truck. These results were reported inreferences 6 and 7 and are summarized here todemonstrate the significance of certain aerodynamiccomponents and features.

The new PHV model fabricated and tested is shownin Figure 2, where many of the new components arenoted. At DOE’s request, we replaced the earliergeneric GTS tractor with the more current GenericConventional Model (GCM) tractor shown inFigure 2. Note that while this tractor model isrepresentative of current on-road vehicles, it is notspecific to any one brand. The new trailer hasblowing surface components similar to those of thepreviews trailer but covering less vertical length (thetrailer floor is raised to the conventional level, not“low-boy” height). The model has many newcomponents typical of the real test vehicle:

• Trailer suspension, springs, brakes, axles,support feet (jack stand), and I-beam floor rails;

• Tractor differentials and suspension;• Mirrors;• Cab gap extenders (full or 60% coverage);• Trailer rear under-ride bar and mud flaps; and• Stock wheels spaced 4 per axle, plus other wheel

options.

Figure 2. New GCM tractor model with full cabextender (CE3”), 4 stock wheels per axle, jack stands,and differentials.

Details of some 325 new wind-tunnel runsconducted over ranges of tunnel speed, blowing rate,yaw (side wind) angle, and model configurationvariations are presented in references 6 and 7, andthe most significant findings are presented below.

The importance of cab extenders in counteractingthe adverse effects of asymmetric vortex shedding inthe gap between tractor and trailer is shown inFigure 3 for the unblown tractor/trailerconfiguration. Clearly, a gap fairing is needed here(see “Full Open Gap” curves), but since 100% full-coverage gap ex-tenders (CE3”) are not practicalduring real vehicle turning, we decided on a 60%gap closure (CE1.5”), which produces nearly thesame aerodynamic drag results as the 100% fullclosure but is functionally feasible (it leaves a 16”gap on the real vehicle).

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

CD

-15 -10 -5 0 5 10 15

Yaw Angle, ψ, deg

Drag vs Yaw Angle, α=0°, Wheels On,Square LE & TE on Trailer, Unblown,Effect of Gap Extenders

Nose right

Run 468, GCM Cab & Trailer,Full Open Gap, Whls B

Run 411, Trailer Alone Sq. TopLE, Wheels B

Gap Vortex

Gap Vortex

Run 475, GCM Cab & TrailerCE3" Ext'r, NO Gap, Wheels B

Run 481, GCM Cab & Trailer, CE1.5" , Wheels B

Figure 3. Measured drag of the new trailer andthe tractor/trailer combo at yaw angle forvarious gaps.

Figure 4 shows the results of blowing and the effectsof various components on drag reduction ascompared to the conventional GCM model withsquare leading and trailing edges on the trailer and afull open gap. Run 171 is the best blownconfiguration of the GTS “unrealistic” genericmodel from the previous tests.


30

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

CD

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20Momentum Coefficient, Cµ

Blown Heavy Vehicle Drag vs Blowing due to Configurations Modifications;

h=0.01", 0.375"R Circular Arc 90/30°TE, q=11.86 psf, V=70 mph, ψ=0°, α=0°

DASHED=Blown, GTS, MTF053, Run171, Best= 90/30, 4 slots, 1/2 whls

CD=0.603,GCM,Sq. LE &TE, Full OpenGap,Run 467

SOLID= Blown, GCM,Round LE, 90°/30° TE,Wheels B, All 4 slots blown

R584, Remove Trailer Wheels & suspensionsfrom R585, No Fairing

R604, Add I-beam Fairings to R584

R605, Remove Tractor Aft Wheels fromR604

R585, GCM, 90/30°. Trlr Wheels B,AFT; NOWhl Frngs, UR Bar, Jack Stand, nor I-beamfairings

R601, Add Wheel Fairings to R585

Expected Full-ScaleCmu

Figure 4. Measured drag reduction due to blowingall 4 slots of various configurations.

The top curve (Run 585) of Figure 4 represents thecorresponding blown trailing edge geometry withthe representative 4-per-axle wheels and suspensioninstalled on the GCM PHV model. Initial dragreduction due to Cµ flattens out and then risesslightly as the wake from the wheels interferes withthe jet turning, much as it did on the full-scale PHVdescribed in reference 4. As we faired the wheels(Run 601), the conditions improved until the newblown truck was very close to the dashed targetcurve from the previous generic tests (Run 171).This new configuration without the under-floordisturbance yields a drag reduction of 24% belowthe unblown baseline (Run 467) at the expected full-scale blowing coefficient of Cµ=0.065. Note,however, that if the trailer-wheel wake effects wereeliminated entirely (Run 584), and if the floor I-beams were faired over (Run 604), considerablybetter CD reduction will occur. In the extreme, if theaft tractor wheel effects were eliminated as well(Run 605), a drag coefficient of CD=0.33 is possiblefor this PHV configuration — this is on the order ofcurrent sports car coupes. This result furtheremphasizes the strong influence of the vehicleundercarriage. These results need be considered byother researchers conducting drag reduction efforts

on current HVs: it is very important to account forthe underbody and wheels of the entire vehicle.

Fairings covering the trailer suspension, axles, andwheels were tested, thus eliminating many of theundercarriage problems from Figure 4. Figure 5shows the drag reductions of the faired configurationcompared to the “baseline” HV, which has a dragcoefficient CD=0.702. The blown results, which aredue to several variations in slot height, are seen here.While all configurations have the same slot heighton the top and sides, the bottom surface slot heightis varied here to determine any gains fromimproving the disturbed lower surface flow byadding more mass flow there. Indeed, it is seen thatincreasing the lower slot height does reduce CD atthe same Cµ. In the extreme, too large a bottom slotcan reverse this trend. For the best arrangement, CDis reduced by 31% at Cµ=0.04–0.05 relative to thestock baseline configuration (CD=0.702), with majorgeometry modification and no accounting for fueluse for blowing. Thus, this latest wind-tunnelevaluation has provided a “real-world” configurationthat should be capable of about a 15–16% fueleconomy increase at highway speeds. These tests

0.40

0.42

0.44

0.46

0.48

0.50

0.52

0.54

0.56

0.58

0.60

0.62

0.64

0.66

0.68

0.70

0.72

CD

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20Momentum Coefficient, C µ

MTF053, 069--Blown Heavy Vehicle Modifications, FINAL CONFIG, h=0.01", 0.375"R 90°/30° TE,

q=11.86 psf, V=70 mph, ψ=0°, α=0°

Faired Unblown GTS Baseline, No Gap, Square TE,C D=0.627

DASHED, Blown GTS, MTF053, Run 171,Best= 90/30, 4 slots Blown,h=0.01"

GCM&Trlr, Sq. LE&TE, OpenGap, Run 467+ JackStand+Diffs+ Mirrors+ UR Bar +Mud Flaps; C D=0.702

SOLID, Blown GCM,MTF069, Round LE,0.375"R, 90°/30° TE, Full Whls, setB, All 4 slots blown

Run718, GCM, 90/30°. Final,h=0.010", all 4 slots

Run720, GCM, 90/30°. Final,h=0.010", h=0.04" bottom



Figure 5. Drag reduction due to blowing and geometryon the final PHV configuration.


31

also confirmed the ability of blowing to provideyawing moment to counteract side winds and thusprovide directional stability to these large-sidedvehicles [7]. A valuable lesson learned from thesetests was the considerable interference (separatedand reversed flow) effects produced by all of themechanical components and the wheels on the trailerunderside.

One last item of interest from these tests is theadditional drag and incremental horsepower requiredto overcome the protrusions into the flow of thecomponents shown in Table 1.

Table 1. Drag increments and corresponding required HPdue to external components.

Component ∆CD ∆HP @ 70 mphRear View Mirrors +0.043 +10.51Under-ride bar +0.049 +11.97Mud flaps +0.005 +1.22Jack stands (feet) +0.002 +0.49Tractor Different’l +0.001 +0.24

Figure 5 thus represents the drag of the full-scalePHV test vehicle configuration, which eliminates themajor Table 1 component items (under-ride bar andflaps are now enclosed) but still is hampered by thepresence of the required rear-view mirrors.

Full-Scale Trailer ModificationsAs a result of the above series of tunnel evaluationsand developments, a final blown PHV configurationto undergo fuel economy testing was determined andincludes the following:

• Ninety-degree (vertical side corners) and30-degree (top & bottom) blowing surfaces withvariable slot heights;

• 60% cab extender, 16-in. gap exposed;• Wheel and axle fairings on trailer and no

exposed trailer mud flaps;• Forward trailer wheel location;• Aerodynamic under-ride bar (airfoil fairing);• Stock wheels, four per axle on trailer;• Stock differentials, axles, and springs on tractor;

and• Side mirrors on tractor, as required.

The trailer modification was completed by GTRIand our teammate prototype shop Novatek, Inc., inearly summer 2004. It is shown at the test track inFigure 6. Not shown are the internal blowersconnected by ducting to the trailing edge blowingsurfaces and the internal diesel drive motorpowering these blowers. Air was entrained into theseblowers through the NACA inlets on the trailersidewalls shown in Figure 6. Preliminary checkouttesting was conducted at GTRI to measure internaland jet pressures, temperatures, and flow rates;degree of trailing edge jet turning; and data systemsoperation. When all systems where confirmed, thePHV trailer was picked up by teammate VolvoTechnology of America (VTA) and transported toVolvo’s facility in North Carolina.

Preliminary Tuning Test 3 (TT3)After arrival at the Volvo facility, the test truck wasagain evaluated to assure blowing jet turning, whichproved quite satisfactory, especially on the90-degree vertical surfaces. Tuning Test 3 (TT3)was then conducted on a four-lane highway toconfirm that all blowing and data systems wereoperating successfully on-road and to generatepreliminary fuel consumption data before theupcoming SAE Type-II test. On-road flow fieldattachment due to jet turning was very similar to thatshown in Figure 6 with blowing ON. With blowing

Figure 6. Tuft flow visualization showing flow turningwith blowing ON.


32

OFF, the tufts pointed aft and fluttered. This photogives a graphic demonstration of the blowingeffectiveness in preventing aft-surface flowseparation on the trailer aft corners.

Although not considered as truly indicative of fueleconomy determination, these on-road tuning testswe conducted yielded significant and informativetrends. To eliminate any side wind or elevationeffects, they were run in both north/south directionson a 2.9-mi length of four-lane highway using an on-board digital fuel readout based on recorded pulsesof the Volvo diesel engine’s fuel-injection system.Speed was set and maintained by the Volvo cruisecontrol at 65 mph between preset road signpostsonce the vehicle had achieved test speed, so noaccelerations/ decelerations were included. On-boardlaptop computers recorded truck engine parametersand fuel consumption plus blowing parameters. Thedata were averaged over the N/S runs to yield eachtest point, and then each test condition was repeatedat least once for consistency (29 runs wereconducted in three days in August 2004).

Fuel economy was determined by using severalmethods, with results shown in Figure 7 as functionsof blowing momentum coefficient, Cµ. (Data areplotted as percent miles/gallon change from the mpgof the baseline stock trailer, which was also testedwith the same tractor.) The engine parameters wererecorded digitally and integrated to give time-averaged mpg. The data shown in Figure 7, labeled“full distance,” were integrated over the entire2.9-mi run, averaged in each direction. These time-averaged data (striped bars) are compared here withthe trends of the wind-tunnel data “MTF069” (thistunnel data has been converted to %mpg increase byassuming that %CD reduction is roughly twice the%fuel economy increase [2, 3, 4]). These “on-road”integrated data thus show the %Fuel EconomyIncrease to range from 4-5% ±1%.

This preliminary TT3 was thus completed, andalthough it was not considered an official fueleconomy test, it confirmed that the PHV blown testrig was yielding appropriate drag reductioncharacteristics due to blowing and was thus readyfor SAE Type-II fuel economy testing.

0

2

4

6

8

10

12

14

16

18

20

22

24

% Chng in MPGfrom Base- line HV, N&S avg

0 0.01 0.02 0.03 0.04 0.05Momentum Blowing Coefficent, C µ

TT3 Fuel Economy Results at VTA (August ,2004)

%MPG, from MTF069 Wind-tunnel data (%MPG= -1/2 %CD reduction)

Tuning Test 3, Full Distance,N&S Avg

(Bars are averaged over same- condition runs at each Cmu;North & South Average;Blower fuel not included)

Figure 7. Tuning Test 3 Fuel economy results and wind-tunnel comparisons.

Phase II SAE Type-II TestsThe pneumatic test truck and a stock reference(control) tractor/trailer were transported by Volvo tothe Transportation Research Center (TRC) 7.5-mitest track in East Liberty, OH, for our Phase II SAEType-II fuel economy evaluations. These tests wereconducted by TRC drivers and personnel in strictaccordance with SAE J1321 procedures [8]. Foreach valid test point of fuel consumed, these requirethat three successive runs of six laps each (45 miaround the TRC test oval), at a constant speed andconstant blowing parameters or test configuration,be made by the test (T) truck and by the control(C) truck at the same time within certain allowabledeparture times and displacement distances. Fueleconomy is measured by weighing removable fueltanks and then comparing the test truck’s fuelburned with the control truck’s. This eliminatesvariations in temperature, side winds, an otherparameters. When the test/control fuel-burned ratiois within a required consistency of each other forthree measured runs, that data point is consideredvalid. A view of the entire PHV test vehicle,including the added 60% cab extenders and wheelfairings, is seen in Figure 8. The control tractor-trailer was a second Volvo/Great Dane combo withstock geometry, also shown in Figure 8.


33

Figure 8. PHV test trailer (top) at TRC with rebuilt cabextenders and control trailer (bottom).

For these test runs, our target data points based onTT3 and Figure 7 were Cµ=0.0. 0.02 and 0.04 at65 mph and Cµ=0.0, 0.02, and 0.035 at 75 mph.Plus, for comparison as a baseline vehicle, the PHVtest truck had to be disassembled on-site andreturned to the baseline (stock) configuration andthen run at 65 and 75 mph. This test program wasconducted at TRC in September 2004.

The %Fuel Economy Increase ratios (%FEI, same as%MPG increase) come from comparing the T/Cfuel-burned ratios for each test condition to the T/Cof the stock baseline truck at the same speed [8].These TRC fuel economy increases are seen for thetwo test vehicle speeds in Figure 9 (“PHV Total,”top solid curves), and are compared with the GTRIwind-tunnel-based data from Figure 7. The TRCdata have a very similar trend to the tunnel data interms of increased %FEI with blowing Cµ. Thelower solid curves show %FEI due to blowing only,where up to 4.5%FEI is seen, not accounting for fueluse by compression. Higher blowing than 0.03seems to cause a slight drop in fuel economy, just asit did in the tunnel data at higher Cµ, probablybecause of the lack of flow disturbance underneaththe trailer and degradation of the effects of higherblowing (see Figure 5 as well). However, relative toour previous SAE test on the first generation of thisPHV test truck [3, 4, 5], these results are more than2.2 times those earlier %FEI results. To put thefindings in perspective, a 1% increase in FEI

0

2

4

6

8

10

12

14

%FEI=% Chng in MPGfrom Base-line HV

0 0.01 0.02 0.03 0.04 0.05 Momentum Coefficient Cµ

TRC On-Track (9/04) SAE Type -II Fuel Economy Test Results

(Blower fuel not included.except as noted)

PHV Total, V=65mph

PHV Total, V=75mph

PHV, Blowing Only,V=65mph

PHV, Blowing Only,V=75mph

T/C = 1.00 forBaseline ReferenceTruck

PHV Total FuelIncluded, PulsedBlowing, V=75mph

PHV Total, FuelIncluded, PB,V=65mph

%FEI fromWind-tunnelData

Figure 9. TRC SAE Type-II fuel economy results for thepneumatic system components.

represents approximately 220–240 million gal ofdiesel fuel saved by the U.S. heavy vehicle fleeteach year.

In the above sets of data, fuel used to power theblower engine has yet to be included. In the middle(dashed) curves of Figure 9, we have includedblower fuel burned, in which we have also added theuse of pulsed (cyclic) blowing to reduce the blowingmass flow required to achieve these drag reductions(see reference 9 for details of this technology, whichGTRI developed with NASA). Results including thisnot-yet-optimized system still show approximately8–9%FEI for these blown configurations, includingblower fuel. Note also in Figure 9 that the data at thehigher speed (75 mph) show greater improvementfrom blowing than at 65 mph since drag there is themore dominant term over rolling resistance. Asnoted, the raw TRC data have been equalized toensure that the T/C ratios for the baseline referenceconfigurations (from which the test vehicle fueleconomy increasaes were derived) were the same(1.0) at both speeds.


34

ConclusionsTo advance the state of development of pneumaticaerodynamics for heavy vehicle drag reduction, fueleconomy, braking, stability, and safety of operation,GTRI and its team members have continued in 2004our previous program for DOE EERE. We haveconducted new model-scale wind-tunnelinvestigations to identify and correct aerodynamicproblem areas from our first fuel-economy test. Wehave also completed new full-scale on-road and test-track fuel economy validations of these advancedcapabilities on a full-scale PHV. Results of thisrecent effort include:

• We have identified aero problem areas existingafter our first PHV road test and how to correctthese; current wind tunnel data indicate that dragreductions of up to 31% result from the new“real-world” PHV configuration, from whichfuel economy increases of about 15–16% due toblowing and associated geometry improvementsshould result at highway speeds. On the basis ofthis wind-tunnel data, a new blown test vehiclemodification was fabricated and assembled.

• SAE Type-II fuel economy runs of this newPHV vehicle on the TRC test track showed animprovement of up to 4.5% at 65 mph as a resultof blowing only, not accounting for fuel used bythe blower.

• The PHV concept has now been verified both bysmaller-scale wind tunnel evaluations and byfull-scale on-road and on-track SAE testing tobe a promising means to reduce drag andincrease fuel economy of HVs. We must stilladdress and resolve the problems caused byundercarriage and wheel componentdisturbances that seem to cause lesser blowingeffectiveness at higher blowing.

References1. Englar, R.J., “Circulation Control Pneumatic

Aerodynamics: Blown Force and MomentAugmentation and Modification; Past, Presentand Future,” AIAA Paper 2000-2541,June 2000.

2. Englar, R.J., “Advanced Aerodynamic Devicesto Improve the Performance, Economics,Handling and Safety of Heavy Vehicles,” SAEPaper 2001-01-2072, May 14–16, 2001.

3. Englar, R.J., “Pneumatic Heavy VehicleAerodynamic Drag Reduction, SafetyEnhancement, and Performance Improvement,”Proceedings of the UEF Conference “TheAerodynamics of Heavy Vehicles: Trucks,Buses, and Trains,” Dec. 2002.

4. Englar, R.J., “Drag Reduction, SafetyEnhancement and Performance Improvement forHeavy Vehicles and SUVs Using AdvancedPneumatic Aerodynamic Technology,” SAEPaper 2003-01-3378, Nov. 2003.

5. Diamond, S., “FY2003 Annual Progress Reportfor Heavy Vehicle Systems Optimization,FreedomCAR and Vehicle TechnologiesProgram,” Feb. 2004.

6. Englar, R.J., “Continued Development &Improvement of Pneumatic Heavy Vehicles,Phase VI, DOE Quarterly Report 7,” April 15,2004.

7. Englar, R.J., “Continued Development &Improvement of Pneumatic Heavy Vehicles,Phase VI, DOE Quarterly Report. 8,” July 15,2004.

8. “Joint TMC/SAE Fuel Consumption TestProcedure-Type-II,” SAE J1321, Oct. 1986.

9. Jones, G.S., and R.J. Englar, “Advances inPneumatic High-Lift Systems through PulsedBlowing,” AIAA Paper 2003-3411, June 2003.


35

F. Heavy Vehicle Aerodynamic Drag: Experiments, Computations, and Design

Kambiz Salari, Jason Ortega, Paul Castellucci, Craig Eastwood, Rose McCallen, Kipp WhittakerLawrence Livermore National Laboratory7000 East Ave, L-098, Livermore, CA 94551(925) 423-0958, fax: (925) 422-3389, e-mail: [email protected]


Contractor: Lawrence Livermore National LaboratoryContract No.: EEW0046

LLNL’s effort consists of four experimental, computational, and design focus areas:

• An Experimental Study of Drag Reduction Devices for a trailer Underbody and Base• Investigation of Predictive Capability of RANS to Model Bluff Body Aerodynamics• Splash and Spray Suppression• Computational Investigation of Aerodynamics of Rail Coal Cars and Drag Reducing Add-On Devices

The following describes the objectives, approach, accomplishments, and future directions for each of these focus areas.

A. An Experimental Study of Drag Reduction Devices for a Trailer Underbodyand Base

Objective• This wind tunnel study investigates optimization of trailer base flaps and alternative forms of trailer skirts in an

effort to reduce the aerodynamic drag of heavy vehicles, thereby improving their fuel efficiency.

Approach• Low-speed wind tunnel measurements are made on a 1/16th-scale generic tractor-trailer model at a width-based

Reynolds number of 325,000.

• The model is fixed to a turntable, allowing the yaw angle to be varied between ±14o in 2o increments.

• Various add-on drag reduction devices are mounted to the model underbody and base.

• The wind-averaged drag coefficient at 65 mph is computed for each configuration, allowing the effectiveness ofthe add-on devices to be assessed.

Accomplishments• The most effective add-on drag reduction device for the trailer underbody is a wedge-shaped skirt, which

reduces the wind-averaged drag coefficient by 2.0%.

• For the trailer base, the most effective add-on drag reduction device is a set of curved base flaps having a radiusof curvature of 0.91 times the trailer width. These curved base flaps reduce the wind-averaged drag coefficientby 18.8%, providing the greatest drag reduction of any of the devices tested.

• Maximum drag reduction for the angled base flaps occurs when the top and side flaps have deflection angles of11.1° and 10.1°, respectively.


36

• When the wedge-shaped skirt and curved base flaps are used in conjunction with one another, the wind-averaged drag coefficient is reduced by 20%.

Future Direction• CFD simulations will be performed on a heavy vehicle with a wedge-shaped skirt to gain additional insight into

the performance characteristics of the wedge-shaped skirts.

• When future track tests are performed by using the angled base flaps, a recommendation will be made to testunequal deflection angles for the side and top base flaps.

IntroductionIn an effort to improve the fuel efficiency of heavyvehicles, this wind tunnel study investigates theoptimization of trailer base flaps and alternativeforms of trailer underbody skirt designs. Previousresearch [1, 2] on angled base flaps hasdemonstrated that this concept is capable ofreducing the drag by as much as 10% on a heavyvehicle when the flaps are deployed at equal angulardeflections. However, it is quite possible that anoptimum configuration may be one in which the topand side base flaps have slightly different angulardeflections. In the subsequent sections, we explorethis possibility. Additionally, we investigate thedrag-reducing capability of curved base flaps, whichhave previously shown to perform about as well asstraight base flaps [2]. To circumvent theshortcomings of straight side skirts, which limitaccess to the trailer underside, we investigate threevariations of a wedge trailer skirt concept that mayprovide the drag-reduction benefits of straight sideskirts.

Experimental SetupThe effectiveness of trailer underbody and base dragreduction devices is assessed by making axial forcemeasurements in a wind tunnel on a 1/16th-scalegeneric tractor-trailer model (Figure 1), which is asimple representation of a near-future tractor/trailer.The wind tunnel measurements are made in theNASA Ames Fluid Mechanics Laboratory open-circuit wind tunnel, which has a contraction ratio of9:1, a test section size of 813 mm × 1219 mm, andfreestream turbulence level of 0.15%. The modelmeasures 162 mm × 225 mm × 1238 m, giving anominal width-based Reynolds number of

ν/Rew Vw= = 325,000, where w is the modelwidth = 162 mm, ν is the kinematic viscosity of air,

Figure 1. Model of the cab-over engine tractor-trailerused in the wind tunnel study.

and V is the freestream velocity. For each modelconfiguration, the axial force measurements aremade at yaw angles, ψ, ranging from ±14° in 2°increments.

Four skirt designs (Figure 2) are tested on the trailerunderside: a long wedge skirt (Figure 2a), which hasan apex angle of 10°; a short wedge skirt(Figure 2b), which has an apex angle of 22°; a shortwedge skirt with an upstream center skirt(Figure 2c); and two conventional straight side skirts(Figure 2d), which are used as a reference formaking performance comparisons with the otherthree skirt designs. A set of angled flaps is tested asa means of reducing the base drag of the model(Figure 3a). The four curved base flap devices areconstructed by using a rapid prototyping technique(selective laser sintering), which forms a single-pieced design. The curved base flaps extend adistance of 48 mm from the trailer base and haveradii of curvature, R, of 52, 79, 148, and 288 mm. Inunits of the trailer width, the radii of curvature are0.32w, 0.49w, 0.91w, and 1.78w.

ResultsThe wind-averaged drag coefficients [3], Cdwa, forthe four trailer skirts are listed in Table 1. It shouldbe noted that the trailer skirts Cooper [1] testedyielded reductions in the wind-averaged drag


37

Figure 2. Skirt designs used to reduce the trailerunderbody aerodynamic drag: (a) long wedge skirt, (b)short wedge skirt, (c) short wedge skirt with center skirt,(d) straight side skirts. (Note that the gap between thecenter skirt and short wedge skirt in [c] is to allowclearance for the sting mount.)

Figure 3. Base flap designs: (a) angled base flaps; (b)curved base flap.

coefficient that are much larger than those shownpresently. The reason for this difference may bebecause the models Cooper used had complete axlesand wheels on the trailer, which would likely

contribute a much greater portion to the overallvehicle drag than simple half-cylinder wheels.Hence, the installation of the skirts on the morerealistic models could result in a greater dragreduction. Both short wedge skirts providenegligible reduction of the wind-averaged dragcoefficient. On the other hand, it can be seen that thelong wedge skirt provides the greatest dragreduction of the four trailer skirt designs. Thissuggests that the long wedge skirt is a design thatcan improve upon that of the traditional straightside skirts by not only yielding a greater reductionin the wind-averaged drag coefficient, but also byallowing easier access to the trailer underside.However, additional testing of the long wedge skirtwith a more realistic trailer underbody and a movingground plane is needed before a definite conclusioncan be drawn.

The angled base flaps on the sides and top of thetrailer are tested at independent angular deflectionsof 5°, 10°, 15°, and 20°, while the bottom flap ismaintained at a deflection angle of 0°. It is seen inthe data that there is a combination of αside and αtopthat yields a minimum in the wind-averaged dragcoefficient in the vicinity of αside ≈ 10° and αtop

≈ 10°. The wind-averaged drag coefficients in therange of 5°≤ αside ≤ 15°, 5°≤ αtop ≤ 15° are fit with asecond-order polynomial surface to estimate theangular deflections that yield a minimum value ofthe wind-averaged drag coefficient. Doing so gives aminimum in the polynomial surface at αside = 10.1°and αtop = 11.1°, at which location the wind-averaged drag coefficient is 0.493±0.004. Table 1shows that this value is 16.4±0.8% less than that ofthe baseline configuration.

As a comparison to the base flaps made of straightplates, four curved base flap configurations aretested. The minimum measured wind-averaged dragcoefficient occurs for the curved base flaps that havea dimensionless radius of curvature of R/w = 0.91.As shown in Table 1, the wind-averaged dragcoefficient for this configuration is 18.8±0.8% lessthan that of the baseline case. This reduction in thedrag coefficient is the largest for any single add-ondevice tested in this study.

Having analyzed the performance of the trailer skirtsand base flaps on an individual basis, we now assess


38

the effectiveness of combinations of these devicesand determine the configuration that results in thegreatest drag reduction. Of all the devices tested, thecombination of the curved base flaps (R/w = 0.91)with the long wedge skirt gives the greatestreduction in the wind-averaged drag coefficient.This combination results in a 20.0±0.8% reductionin the wind-averaged drag coefficient, the majorityof which is due to the contribution of the curvedbase flaps. The wind-averaged drag coefficients forthis configuration and that of the combination of thelong wedge skirt and the straight base flaps at αside =αtop = 10° are shown in Table 1.

A comparison of the wind-averaged dragcoefficients of the model with trailer skirts and themodel with base flaps indicates that the base flapsprovide a substantially larger drag reduction than dothe skirts. Assuming that the trends in the wind-averaged drag coefficients are applicable to full-saleheavy vehicles, these results could have importantimplications for the current day trucking industry.

Table 1. Wind-averaged drag coefficient, Cdwa, and thepercent reduction in the wind-averaged drag coefficientrelative to the baseline case for various modelconfigurations.

Configuration Cdwa(±0.004)

% Reduction inCdwa (±0.8%)

Baseline 0.590 ⎯Baseline w/longwedge skirt

0.578 2.0

Baseline w/shortwedge skirt

0.587 0.1

Baseline w/shortwedge skirt andcenter skirt

0.587 0.1

Baseline w/straightside skirts

0.582 1.4

Baseline w/angledbase flaps (αtop =11.1o, αside = 10.1o)

0.493 16.4

Baseline w/curvedbase flaps (R/w =0.91)

0.479 18.8

Baseline w/angledbase flaps (αtop =10o, αside = 10o) andlong wedge skirt

0.484 18.0

Baseline w/curvedbase flaps (R/w =0.91) and longwedge skirt

0.472 20.0

Clearly, if a trucking fleet decided to purchase asingle add-on drag device to obtain the greatest dragreduction, the choice of curved base flaps would bethe best alternative. However, the angled base flapsmay be more attractive from an investment point ofview since their rather simple design would be lessto manufacture and, thus, require a smaller initialinvestment.

ConclusionsThrough this study, we have investigated severaladd-on drag-reduction devices. The wind-averageddrag coefficient is used as a means of comparing theperformance of trailer skirts and base flaps. Of thetrailer skirts that are tested, the long wedge skirtprovides the greatest drag reduction. The angled andcurved base flap devices yield reductions in thewind-averaged drag coefficient approximately eightto nine times greater than that of the long wedgeskirt. It should be noted that the results of this windtunnel study are valid for the low Reynolds numberregime that was tested. Future tests will beconducted on actual heavy vehicles to determine theinfluence of Reynolds number and the movingground plane beneath the vehicle.

What is needed to get these devices onto operatingheavy vehicles is a committed effort between thegovernment, the tractor/trailer manufacturers, andthe trucking fleets. The government has seen theopportunity to improve the fuel economy of heavyvehicles and has taken the initiative to supportresearch and development in heavy vehicleaerodynamics. The involvement of tractor/trailermanufacturers can provide expertise in both roadtesting and design issues of these devices. Thetrucking fleets can give practical insight into howthese devices impact the operational capability oftheir fleets and foresee any potential concerns. Onlyuntil this collaboration is established will the nationbe able to benefit from the significant cost savingsthat these second-generation drag reduction devicescan provide.

References1. Cooper, K.R., “Truck Aerodynamics Reborn —

Lessons from the Past,” SAE Paper No. 2003-01-3376, 2003.

2. Cooper, K.R., “The Effect of Front-EdgeRounding and Rear-Edge Shaping on the


39

Aerodynamic Drag of Bluff Vehicles in GroundProximity,” Paper No. 850288, SAEInternational Congress, Detroit, MI, February25–March 1, 1985.

3. Ingram, K.C., “The Wind-Averaged DragCoefficient Applied to Heavy Goods Vehicles,”Transport and Road Research LaboratorySupplementary Report 392, 1978.

B. Investigation of Predictive Capability of RANS to Model Bluff BodyAerodynamics

Objectives• Investigate the applicability of state-of-the-art computational modeling and simulations to predict the flow field

around bluff bodies as related to the DOE Heavy Vehicle aerodynamics project. Specifically, determine thepredictive capability of several commonly used, steady, Reynolds Averaged Navier-Stokes (RANS) turbulencemodels.

• Provide tractor and trailer manufacturers with a knowledge base that describes the advantages anddisadvantages of each turbulence model when selected for use within a commercial code.

Approach• Perform simulations of a simplified tractor/trailer geometry for validating the drag prediction capabilities of

selected RANS turbulence models.

• Compare the force coefficients, as well as the surface pressures and flow features, to experimental data [1] ofthe Ground Transportation System (GTS) at 0° and 10° yaw conducted in conjunction with NASA AmesResearch Center in their 7' 10' wind tunnel.

Accomplishments• Compared the experimentally obtained floor boundary layer profile at the wind tunnel test-section entrance with

those predicted by each of the turbulence models.

• Compared computed aerodynamic force coefficients for the selected turbulence models to the experimental dataat yaw angles of 0° and 10°. Menter’s two-equation BSL RANS model shows the closest agreement to theexperimental data, generally predicting drag within 5%. However, all of the models fail to capture the structureof the wake flow near the trailer base. Consequently, all of the selected models fail to reproduce theexperimental base pressure coefficients.

Future Direction• Investigate the applicability of unsteady RANS, large-eddy simulation (LES), and combinations of the two, for

the prediction of these massively separated base flows.

IntroductionThe objective is to investigate the applicability ofstate-of-the-art computational modeling andsimulations to predict the flow field around bluffbodies as related to the DOE Heavy Vehicleaerodynamics project. Specifically, the goal is todetermine the predictive capability of severalcommonly used, steady, Reynolds Averaged Navier-Stokes (RANS) turbulence models. This informationwill provide tractor and trailer manufacturers with a

knowledge base that describes the advantages anddisadvantages of each turbulence model choice.

Computation Approach and ResultsSimulations are performed on a 1/8th-scalesimplified tractor/trailer geometry for validating thedrag prediction capabilities of selected RANSturbulence models. These models include the one-equation Spalart-Allmaras (SA) [2] and the two-equation Wilcox (KW) [3] and Menter (BSL) [4] k-ε


40

models. The force coefficients, as well as the surfacepressures and flow features, are compared toexperimental data [1] of the Ground TransportationSystem (GTS) at 0° and 10° yaw. The data wereacquired from an experiment conducted inconjunction with NASA Ames Research Center inits 7-ft × 10-ft wind tunnel.

Simulations of the GTS are run by using NASA’sOVERFLOW code. OVERFLOW is a fullycompressible, 3-D, finite volume code employingoverset grids. In all simulations, the flow is assumedto be fully turbulent and no attempt is made tomodel transition. Additionally, no wall functions areused, as all turbulence equations are integrated to thewall.

As the GTS model is exposed to the floor boundarylayer of the 7-ft × 10-ft wind tunnel, the accuracy ofa validation simulation depends on how well theupstream boundary layer is represented. Thus, aportion of the wind tunnel geometry is modeled,where careful attention is paid to matching thesimulation boundary conditions to the actual windtunnel conditions. Figure 1 presents a comparisonbetween the experimentally obtained floor boundarylayer profile at the wind tunnel test-section entranceand those predicted by each of the turbulencemodels.

Corresponding to the NASA experiment1 run 7,points 9 and 5, the GTS baseline configuration at 0°and 10° yaw are selected for simulation. Using theSA, KW, and BSL models, 0° yaw simulations areconducted on a 14-million-element grid. In addition,the BSL model is chosen to ensure grid convergenceof the solution on a coarser 11-million-element grid.In a similar fashion, 10° yaw simulations are run on19 and 14 million element meshes.

Tables 1 and 2 show the computed aerodynamicforce coefficients for the selected turbulence modelsand the experimental data at yaw angles of 0° and10° respectively. Among these models, Menter’stwo-equation BSL shows the closest agreement tothe experimental data, generally predicting dragwithin 5%.

Figure 1. Floor boundary layer velocity profile.

Table 1. Aerodynamic force coefficients at 0° yaw.

CD CL CSSA, 14 million elements 0.3173 -0.1317 5.1E-06KW, 14 million elements 0.2377 -0.1070 -4.3E-04BSL, 14 million elements 0.2638 -0.1214 -1.4E-05BSL, 11 million elements 0.2684 -0.1220 -3.0E-06NASA experiment 0.263 ± 0.01 -0.152 ± 0.01 0.007 ± 0.01

Table 2. Aerodynamic force coefficients at 10° yaw.

CD CL CSSA, 19 million elements 0.6361 -0.1162 1.1451KW, 19 million elements 0.5708 -0.0236 1.1431BSL, 19 million elements 0.5626 -0.0320 1.1376BSL, 14 million elements 0.5701 -0.0252 1.1329NASA experiment 0.542 ± 0.01 0.026 ± 0.01 1.200 ± 0.01


41

Differences among the turbulence models in thecomputed force coefficients are found to beprimarily due to contribution of the predictedpressure field around the GTS model. The computedpressure coefficients for each model are in closeagreement with the experimental data, except for thearea near the base of the trailer. Figures 2 and 3illustrate the agreement between the predictedpressure coefficients and those obtained along thecenterline of the top surface of GTS at 0° and 10°yaw. It is clear, however, that none of the turbulencemodels capture either the pressure field magnitudeor trend at the base of the trailer. Figures 4 and 5display the discrepancy between the computedpressure coefficients along the base centerline andthe experimental data for each yaw angle.

Figure 2. GTS top surface centerline pressure coefficientsat 0° yaw.

Figure 3. GTS top surface centerline pressure coefficientsfor 10° yaw.

Figure 4. GTS base centerline pressure coefficients at 0°yaw.

Figure 5. GTS base centerline pressure coefficients at 10°yaw.

The accuracy of the base flow predicted by theselected turbulence models is further illustrated byexamining the time-averaged PIV data available inthe wake of the trailer. Although hidden by thesimilarity in force coefficients, the selectedturbulence models predict very different separatedwakes, particularly in the low-velocity area verynear the base of the trailer. As expected from thecomputed pressure coefficients, none of the modelsfully capture the separated flow structure indicatedby the time-averaged PIV data. For example, alongthe model centerline, Figure 6 shows the BSL wakeat 0° yaw. For comparison, particle traces of thetime-averaged PIV data are provided in Figure 7.


42

Figure 6. BSL model centerline wake flow particle traces.

Figure 7. Particle traces of time-averaged PIV data withinthe frame in Figure 6.

ConclusionsOverall, the Menter BSL RANS model simulation iscloser to the experiment than either the KW or SAsimulations; however, all of the models fail tocapture the structure of the wake flow near the base.Consequently, all of the selected models fail to

reproduce the experimental base pressurecoefficients. Although the drag force coefficientsappear to be somewhat insensitive to the details ofthe flow near the base, to assess the efficacy of drag-reducing devices in the wake, steady RANS modelsare likely to prove to be inadequate. To consistentlyachieve drag values to within the incremental valuesproduced by trailer base treatments, unsteadysimulations may be required to fully represent thewake flow. Future plans call for investigation of theapplicability of unsteady RANS, large-eddysimulation (LES), and combinations of the two, forthe prediction of these massively separated baseflows.

References1. Storms, B., et al., “An Experimental Study of

the Ground Transportation System (GTS) Modelin the NASA Ames 7- by 10-Ft Wind Tunnel,”NASA/TM-2001-209621, Feb. 2001.

2. Spalart, P.R., and Allmaras, S.R., A One-Equation Turbulence Model for AerodynamicFlows, AIAA Paper 92-439, Reno, NV, 1992.

3. Wilcox, D.C., Turbulence Modeling for CFD,2nd Edition, DCW Industries, 1998.

4. Menter, F.R., Two-Equation Eddy-ViscosityTurbulence Models for EngineeringApplications, AIAA J., 32 (8), 1994, 1598-1605.


43

C. Splash and Spray Suppression

Objectives• Analyze the formation and spread of spray clouds generated around and behind heavy vehicles by using both

laboratory experiments and validated computational models.

• Develop a collection of predictive computational tools available to tire and truck manufacturers that will allowthe effects of add-on splash and spray suppression devices on vehicle aerodynamics and potential synergiesbetween add-on aerodynamic devices and spray mitigation to be thoroughly explored.

Approach• The droplet size, space, and velocity distributions determined as a function of tire speed and tread pattern from

our experimental investigation will be used as an initial condition for our computational simulations.

• Commercially available computational fluid dynamics (CFD) codes will be used to predict the aerodynamicatomization of these droplets. User-defined atomization subroutines will be added to the commercial packagesas necessary.

• Detailed computational investigations of the aerodynamics of rotating tires and wheel-wells and their effect onsplash and spray will be performed.

• The effects of splash and spray suppression devices on brake cooling will be investigated, and possibleenhancements to brake cooling resulting from well-designed add-on aerodynamic devices will be explored.

• Validated state-of-the-art large-eddy simulation (LES) models capable of capturing the spatial evolution of finedroplets in a turbulent flow will be developed and used to simulate the dispersion of mist around and behindheavy trucks.

Accomplishments• Reviewed the literature to determine both the existence and outcome of prior efforts toward splash and spray

suppression and to delineate the pertinent physics governing the problem.

• Coordinated experimental research plan with team at USC.

• By working with USC, developed an experimental test rig.

• Began investigating STAR-CD’s ability to track free fluid surfaces via its implementation of the volume offluid (VOF) method and to predict ensuing droplet formation.

• Began studying STAR-CD’s ability to accurately simulate the flow around a rotating tire, including bothaerodynamic and free-surface phenomena.

• Began investigating STAR-CD’s ability to capture aerodynamic droplet breakup (secondary atomization) andturbulent transport of sub-Kolmogorov scale particles.

Future Direction• This project is in its very nascent stages. All aspects of the project discussed above are under continuing

development.

IntroductionThe spray clouds generated around and behindheavy trucks traveling on highways during adverseweather conditions lead to reduced visibility andincreased anxiety for motorists traveling nearby and

pose a significant public safety concern. Ourobjective is to analyze the formation and spread ofthese clouds by using both laboratory experimentsand validated computational models.


44

ExperimentsFred Browand and co-workers at the University ofSouthern California have developed a unique test rigto investigate water pick-up by tire treads and theprimary formation of droplets in the wake of arotating tire. The test rig consists of a tubularaluminum frame supporting two spinning wheelsmounted pendulum-style upon swinging rigid arms.Commercially available, thirteen-inch-diameterwheels are used. Michelin of North America isproviding custom tires with specified tread designs.The two wheels are pressed together with a lateralspring-damper to form a well-defined, accuratelycontrollable tire patch. The wheels are rolled incontact with one another by using a sprocket andchain drive. The plane of symmetry of theexperimental facility replaces the road surface.Water is introduced from a solenoid-actuatedinjector placed just ahead of the tire patch and isinjected at a velocity matching the peripheral speedof the tires, as would be the case for a rolling tire ona wet road.

The formation of water droplets just downstream ofthe contact patch is studied by using a digitalcamera, pulsed laser light, and numerical algorithmssimilar to those developed for digital particle imagevelocimetry (DPIV). High-speed (4,000–16,000 fps)digital images of ligament formation and dropletejection from the tire surface are also feasible. Thisfacility allows the investigation of various treadgeometries to understand their function in wateruptake and primary droplet formation in a muchmore controlled manner than has been possible.

ComputationsThe droplet size, space, and velocity distributionsdetermined as a function of tire speed and treadpattern from our experimental investigation will beused as an initial condition for our computationalsimulations. Commercially available computationalfluid dynamics (CFD) codes will be used to predictthe aerodynamic atomization of these droplets. User-defined atomization subroutines will be added to thecommercial packages as necessary. Theaerodynamics of rotating tires and wheel-wells areof paramount importance to the secondaryatomization process. Detailed computationalinvestigations of these effects will be performed inconjunction with the atomization study. This will

allow us to simultaneously investigate the effects ofsplash and spray suppression devices on brakecooling and to explore possible enhancements tobrake cooling resulting from well-designed add-onaerodynamic devices.

Validated state-of-the-art large-eddy simulation(LES) models capable of capturing the spatialevolution of fine droplets in a turbulent flow will bedeveloped and used to simulate the dispersion ofmist around and behind heavy trucks. Ultimately, wewish to develop a collection of predictivecomputational tools available to tire and truckmanufacturers. It is desired, therefore, to implementthe specialized models developed in this phase of thestudy as user subroutines in widely available CFDcodes. The result will be a package that allows theeffects of add-on splash-and-spray-suppressiondevices on vehicle aerodynamics and potentialsynergies between add-on aerodynamic devices andspray mitigation to be thoroughly explored.

Accomplishments and ConclusionsWe began this study with an extensive review of theliterature to determine both the existence andoutcome of prior efforts toward splash and spraysuppression and to delineate the pertinent physicsgoverning the problem. We helped to coordinate theexperimental research plan with the team at USCand develop the experimental test rig. We beganstudying the suitability of commercial CFD codes to(1) simulate both primary and secondary formationof droplets and (2) simulate water entrainment andejection by rotating tires. In particular, we havebegun investigating STAR-CD’s ability to track freefluid surfaces via its implementation of the volumeof fluid (VOF) method and to predict dropletformation. We have also begun studying STAR-CD’s ability to accurately simulate the flow around arotating tire, including both aerodynamic and free-surface phenomena. We have also beguninvestigating STAR-CD’s ability to captureaerodynamic droplet breakup (secondaryatomization) and turbulent transport of sub-Kolmogorov-scale particles.

Future DirectionThis project is in its very nascent stages. All aspectsof the project discussed above are under continuingdevelopment.


45

D. Computational Investigation of Aerodynamics of Rail Coal Cars and Drag-Reducing Add-On Devices

Objective• To reduce the aerodynamic drag of an empty rail coal car through use of drag-reducing add-on devices.

Approach• The turbulent time-averaged flow field around a generic coal car at realistic Reynolds number is simulated by

using a RANS computational approach.

• Design drag-reducing add-on devices and test by using the same computational modeling technique.

Accomplishments• The turbulent flow field around the coal car was computed by Star-CD using a standard high Reynolds number

k-ε turbulence model with a wall function. A large recirculation area within the interior of the car that extendsover almost the entire length of the car is largely responsible for producing the high aerodynamic drag for anempty coal car.

• Designed a new add-on device to manipulate the interior flow structure of the car. Computations indicate a 10%drag reduction over the base empty coal car with this new device.

Future Direction• Further investigate the potential of LLNL drag-reducing add-on device by optimizing the number, location, and

height of the plates.

• Collaborate with NASA Ames on its experimental study of coal car aerodynamics and drag reducing add-ondevices

IntroductionThe objective of this work is to reduce theaerodynamic drag of an empty rail coal car throughthe use of drag-reducing add-on devices. Constraintsare imposed on the design of these devices, such asno interference with loading and unloading, noalteration of car internal volume, and no orminimum maintenance.

The turbulent time-averaged flow field around ageneric coal car at realistic Reynolds number issimulated by using the RANS computationalapproach. The predicated flow field is carefullyinvestigated for possible flow structures that aresusceptible to manipulation by add-on devices.Then, drag-reducing add-on devices are designedand tested by using the same computationalmodeling technique.

Computations and ResultsTo perform modeling and simulation on an emptycoal car, a generic model was needed. A CADdefinition of such a model was provided by JimRoss at NASA Ames for this investigation. NASA iscurrently studying the aerodynamics of loaded andunloaded coal cars through a series of experiments.As the NASA experimental data become available,the computational results can be validated. Thegeometry of the generic coal car is shown inFigure 1; the car’s dimensions are 6.84 m ×1.45 m × 1.80 m.

The turbulent flow field around the coal car wascomputed by Star-CD using a standard highReynolds number k-ε turbulence model with a wallfunction. The flow is assumed to be incompressibleand fully turbulent. A realistic flow Reynolds


46

Figure 1. Generic coal car geometry.

number of 4 million, which was based on the carwidth and a flow velocity of 70 km/h (43.6 mi/h),was used for all computations. Figure 2 presents theflow field around the empty coal car using particletraces. The stagnation area in front and flow over thecar are clearly visible. Figure 3 shows a largerecirculation area within the interior of the car thatextends over almost the entire length of the car. Thisflow structure is largely responsible for producingthe high aerodynamic drag for an empty coal car.The computed drag coefficient for this case is 1.28,and the lift coefficient is -0.1 (downward force).

Figure 2. Particle traces showing the flow field aroundthe coal car.

Figure 3. Particle traces on a symmetry plane includingthe interior space.

To reduce the aerodynamic drag of this generic coalcar, the flow recirculation within the car needs to bemodified. Figure 4 presents a new LLNL add-ondevice designed to manipulate the interior flowstructure of the car. Presently, the device consists oftwo thin plates that split the car into three equal-length sections, and they extend half way up asshown in Figure 4. The number of plates could rangefrom 1 to 10, and the height of each plate could varyfrom a one-third up to the full height of the car. Thedevice is extremely simple and should not requireany maintenance. Because of utilization of thinplates, the coal car capacity and loading andunloading are not affected. This is our first attemptto evaluate the performance of an LLNL drag-reducing device.

Figures 5 and 6 present the computed flow fieldaround the coal car caused by the presence of theaerodynamic add-on devices. Figure 5 shows a flowpattern similar to that in Figure 2, except thatparticle traces show lower flow deflection on top ofthe car. The reason for this is shown in Figure 6where the flow recirculation is confined to a regionbetween the two thin plates. For this geometry, thedrag and lift coefficients are 1.16 and -0.15,respectively. This represents a 10% drag reduction

Figure 4. Generic coal car with the aerodynamic add-ondevice.

Figure 5. Particle trace showing the flow field about thecoal car with the add-on device.


47

Figure 6. Particle traces on a symmetry plane, includingthe interior space with the add-on device.

over the base empty coal car. Because of the limitedscope of this effort, no attempt has been made tooptimize the number, spacing, and the height of theplates for maximum possible drag reduction.

ConclusionsThe potential of an LLNL drag-reducing add-ondevice will be further investigated by optimizing thenumber, the location, and the height of the plates.Another objective is to collaborate with NASAAmes researchers on their experimental study ofcoal car aerodynamics and drag-reducing add-ondevices.


48

G. Computational and Analytical Simulation of Simplified GTSGeometries/Bluff Bodies

Principal Investigator: Lawrence J. DeChant (PI), Basil Hassan, Jeff Payne, Matt Barone,Chris Roy (Contractor: Auburn U.)Affiliation: Sandia National LaboratoriesAddress: P.O. Box 5800, Albuquerque, NM 87185-0825(505) 844-4250, fax: (505) 844-4523, e-mail: [email protected]


Contractor: Sandia National LaboratoriesContract No.: DE-AC04-94AL85000

Objectives• Examine the effectiveness of the Detached Eddy Simulation (DES) approach for predicting tractor/trailer

wakes.

• Explore potential for reduced order drag coefficient model based upon combined Green’s function and Gram-Charlier series method.

Approach• Sandia’s SACCARA code was used to conduct DES simulations of a truncated version of the Ground

Transportation System (GTS) without boattail plates.

• Preliminary computations were run on a coarse (4 million cell) mesh.

• The DES simulations were compared to steady-state Reynold’s-Averaged Navier-Stokes (RANS) computationsand experimental data.

• A reduced-order model was developed by using self-similar, far-field, turbulent wake concepts to estimate the2-d drag coefficient for a range of bluff body problems.

Accomplishments• Statistically converged unsteady DES solutions were obtained for the truncated GTS model on the coarse mesh

at a Reynolds number of 2 million.

• The DES predictions were compared to steady-state RANS simulations by using the Menter k-omega model.

• The DES predictions were compared to experimental data obtained in the NASA-Ames 7-ft × 10-ft windtunnel.

• Comparisons include both time-averaged surface pressure and time-averaged wake velocities.

• A reduced-order model was developed by using self-similar, far-field, turbulent wake concepts to estimate the2-d drag coefficient for a range of bluff body problems.

Future Direction• Initial estimates suggest that the Sandia funding levels will be significantly reduced from prior levels.

• Auburn University will examine the DES simulations of the full GTS geometry by using finer meshes.


49

A. DES Simulations

IntroductionPreliminary results are presented for Detached EddySimulations (DES) of a generic tractor/trailergeometry at a Reynolds number of 2 million, on thebasis of trailer width. The DES simulations arecompared to both experimental data and steady-stateReynolds-Averaged Navier-Stokes (RANS)computations by using the Menter k-w turbulencemodel. These comparisons include both time-averaged base pressures and wake velocities. TheDES results with the truncated geometry do notprovide improved agreement with the experimentaldata relative to the steady-state RANS results withthe full geometry. The lack of improved agreementmay be due to either insufficient mesh refinement orthe use of the truncated geometry. Further details ofthis work are presented in reference 1.

ApproachThe computational fluid dynamics code used hereinis SACCARA, the Sandia Advanced Code forCompressible Aerothermodynamics Research andAnalysis. The SACCARA code was developed froma parallel distributed memory version of the INCAcode, originally written by Amtec Engineering. TheSACCARA code is used to solve the Navier-Stokesequations for conservation of mass, momentum,energy, and turbulence transport in either 2D or 3Dform. Prior code verification studies withSACCARA include code-to-code comparisons withother Navier-Stokes codes and with the DirectSimulation Monte Carlo method. These studiesprovide some confidence that the code is free fromcoding errors affecting the discretization.

For the simulations results presented herein, theturbulence transport equations are integrated all theway to the vehicle walls, thus no wall functions areemployed. In all cases, the distance from the wall tothe first cell center off the wall is less than unity innormalized turbulence distance (i.e., y+ < 1). Thesteady-state RANS model examined is Menter’shybrid model, which switches from a k-epsilonformulation in the outer flow to a k-omegaformulation near solid walls. The hybrid RANS/LESmethod developed by Spalart and co-workers hasbeen developed the furthest and is called Detached

Eddy Simulation, or DES. The DES approach usesthe unsteady form of the Spalart-Allmaras one-equation eddy viscosity model to provide the eddyviscosity for use in the sub-grid scale stress model.The Spalart and Allmaras one-equation eddyviscosity model provides the usual RANS-basededdy viscosity in the boundary layer, but it must bemodified to the appropriate eddy viscosity for LESoutside of the boundary layer.

The flow over the Ground Transportation System(GTS) model has been investigated experimentallyat a Reynolds number ReW of 2 million by Storms etal. [2], where W refers to the width of the trailerbase (0.32385 m). The experimental data set isunique in that it presents both ensemble-averagedsurface pressure data, as well as multiple planes ofinstantaneous and ensemble-averaged velocity datain the wake for this high Reynolds number flow.

To perform the computationally intensive DESsimulations, the front of the GTS was truncatedalong with the wind tunnel wall at x/W = 2. Notethat the wind tunnel wall and the rear posts areincluded in the simulation. This mesh consists ofapproximately 4 million grid points and was domaindecomposed and run on 32 processors of a Linuxcluster.

AccomplishmentsPreliminary results were presented for DES ofgeneric tractor/trailer geometry at a Reynoldsnumber of 2 million based on the trailer width forthe truncated geometry. The DES simulations werecompared to both experimental data and to steady-state RANS computations by using the Menter k-omega turbulence model. These comparisonsincluded both time-averaged base pressures andwake velocities. Although we fully expected theDES simulations to improve the agreement with theexperimental data, the results were clearly not betterthan the steady-state RANS computations (seeFigure 1).

The fact that the DES simulations did not provideimproved agreement with the experimental data, ascompared to the RANS results, could be due to anumber of reasons. The residuals for each time stepdid not obtain the desired iterative convergence levelbecause of the extremely fine mesh spacing near the


50

a) experiment b) RANS c) DES

Figure 1. Vertical streamwise cuts of u-velocity and streamlines from (a) experiment, (b) RANS, and (c) DES.

walls; thus, some iterative errors may be pollutingthe solution. The effects of truncating the forwardpart of the GTS geometry and the wind tunnel werenot assessed. The flow over the forward portion ofthe GTS does generate some streamwise vorticity,which is not included in the current simulations.

The back pressure was adjusted and then run for0.1 s before statistics were collected. Examination ofthe reference pressure indicated that the pressuretransients had not yet died out by the time statisticswere collected. In addition, the time window forcollecting statistics should be increased to ensurethat the time-averaged results are statisticallyconverged. Finally, another possible reason for thelack of agreement between the DES simulations andthe experiment is insufficient mesh refinement.Comparison of the unsteady pressure signal from theexperiment with the same signal from the DES

simulations showed that the DES signal had lessstructure and a larger amplitude. This behavior islikely related to the excessive dissipative errorsassociated with insufficient mesh refinement.

References1. C.J. Roy, J.C. Brown, L.J. DeChant, and

M.A. Barone, “Unsteady Turbulent FlowSimulations of the Base of a GenericTractor/Trailer,” AIAA Paper 2004-2255,34th AIAA Fluid Dynamics Meeting, Portland,OR, June 2004.

2. B. Storms, J.C. Ross, J.T. Heineck,S.M. Walker, D.M. Driver, and G.G. Zilliac,“An Experimental Study of the GroundTransportation System (GTS) Model in theNASA Ames 7- by 10-ft Wind Tunnel,” NASATM-2001-209621, 2001.


51

B. Reduced-Order Model

IntroductionIn this study, we extend self-similar, far-field,turbulent wake concepts to estimate the 2-d dragcoefficient for a range of bluff body problems. Theself-similar wake velocity defect that is normallyindependent of the near field wake (and hence bodygeometry) is modified by using a combinedapproximate Green’s function/Gram-Charlier seriesapproach to retain the body geometry information.Formally, a near-field velocity defect profile iscreated by using small disturbance theory and theinviscid flow field associated with the body ofinterest. The defect solution is then used as an initialcondition in the approximate Green’s functionsolution. Finally, the Green’s function solution ismatched to the Gram-Charlier series, yieldingprofiles that are integrated to yield the net form dragon the bluff body. Preliminary results indicate thatdrag estimates computed by using this method arewithin approximately 15%, as compared withpublished values for flows with large separation.This methodology may be of use as a supplement toCFD and experimental solutions in reducing theheavy computational and experimental burden ofestimating drag coefficients for blunt body flows forpreliminary design-type studies.

Drag estimates for strongly separated flow overblunt bodies is an essential piece of information formany engineering systems. An application thatdemands our particular attention is aerodynamicdrag forces on large ground transportation vehicles(i.e., tractor/trailer trucks). As noted by Roy et al.(2003), in common tractor/trailer, energy losses dueto rolling resistance and accessories increase linearlywith vehicle speed, while energy losses due toaerodynamic drag increase with the cube of thespeed. At a typical highway speed of 70 mph,aerodynamic drag accounts for approximately 65%of the energy output of the engine (McCallen et al.1999). Because of the large number oftractor/trailers on U.S. highways, even modestreductions in aerodynamic drag can significantlyreduce domestic fuel consumption. Lower fuelconsumption will result in a reduction in pollutionemissions and, significantly, a reduced dependenceon foreign oil.

Although most modern computationally basedaerodynamic drag reduction studies have focused onComputational Fluid Dynamics (CFD) methodsutilizing evermore sophisticated (and concurrentlycomputationally expensive) methodologies, thereremains a valuable role for reduced-complexity,analytically based models. Here we describe a modelthat provides an alternative to computationallyexpensive models.

2-d Bluff Body DragThe basic methodology recalls that there must be aself-preservation solution (Tennekes and Lumley,1972) — for example, self-similar construction forthe velocity field is necessary:

⎟⎠⎞

⎜⎝⎛=

−lyf

UUU

s

0 (1)

where Us is the maximum cross-stream variation inU. Note that for wakes, Us will be U(y=0) where therelevant coordinate system and associateddefinitions are shown in Figure 1.

Figure 1. Schematic diagram of wake flow withcoordinate, velocity, and length scale definitions.

Unfortunately, the solution obtained by Tennekesand Lumley cannot be strictly valid in the near fieldsince the form of the similarity solution chosen byTennekes and Lumley (1972) requires that

2/12/1 BxlAxU s == − , and A and B areconstants to be determined. Obviously, the Ussolution is not (and cannot be) valid for x<<1. Thislimitation poses no problem in the far-field, ofcourse, and is acceptable in an intermediate overlapregion as well, but it cannot be applied in the wakenear-field.


52

The connection to the velocity defect velocity fieldand the drag coefficient is given by (2-d dragcoefficient):

∫∞

∞−

== ηθ fdd

CD 2 (2)

The 2-d momentum thickness θ is introducedthrough equation (2) as well.

AnalysisThe classic, self-similar far field wake solution isnot valid in the near field of the bluff body.However, by carefully examining the process usedto derive the similarity solution, it is possible toachieve other formulations that are valid in the nearfield.

Approximate Green’s Function SolutionAn alternative approach to the physically basedsimilarity arguments presented in the text involvesthe mathematical analysis of a generalized problem.Here, we discuss a range of fundamental solutions tothe heat equation (the canonical form of thelinearized wake relationship). If, however, we arewilling to use the functional form of equation (2),we can form a solution to the governing equationsthat is self-similar:

( )( )∫

∞

⎥⎥⎦

⎤

⎢⎢⎣

⎡

−−

++−=

02

2

)(exp)(exp

)(1)( ηηη

ηηη

πη dff nearGF (3)

where η0 denotes the dimensionless length scaleassociated with the bluff body. Note that equation(3) will approximately satisfy the full range ofconditions required for solution of the wakeproblem.

“Initial Condition Velocity Field”;Connection to the Inviscid Flow FieldTo utilize equation (2) to obtain the defect velocityfield, it is necessary to be able to compute a near-field velocity defect profile (i.e., fnear). This functionmust provide the sum total of the bluff bodiesgeometric information. In terms of a practical result,it also must be readily obtainable and unique.Perhaps the most obvious closure for the near-field

defect solution that satisfies these requirements is toutilize the local inviscid potential flow solution.

For sharp-edged bluff bodies, such as the squarecylinder shown in Figure 1, where the separationlocation is well established, the defect velocity fieldis readily estimated by the discontinuous stepfunction:

⎩⎨⎧ ≤≤

=otherwise

fnear 0101 η

; where η=2y/d.

Substitution of this near-field relationship into

equation yields: ( ) ( )[ ]

211 −−+

=ηη erferffGF .

Note that when integrated, the near-field andGreen’s function solution give:

CD= ∫∫∞

∞−

∞

∞−

= ηη fddfnear =2.

Although this estimate for 2-d drag is quite good,since CD,exper=2.1 (White 1986), we note that theGreen’s function does not modify the net initialdefect velocity — hence, in terms of the dragcoefficient, the Green’s function relationshipprovides no new information. Our reason to utilizethe Green’s function form will become apparentlater, but we already note that the Green’s functionrelationship will satisfy several essential properties,including:1. Continuous, differentiable flow field valid over

full domain (i.e., -∞→+∞);2. Satisfies symmetry and far-field boundary

conditions; and3. Approximately satisfies governing linearized,

momentum (diffusion) equation.

Gram-Charlier SeriesIn the previous section, it was clear that it is thelocal velocity defect solution that provides anestimate of drag since the Green’s functionrelationship, equation (2), preserves the area underthe local velocity defect approximation. However,the classical far-field wake profiles derived byTennekes and Lumley (1972) should be taken intoaccount in any formulation.

We can achieve a far-field wake influence by notingthat the far-field wake might represent a single firstterm in a Gram-Charlier series:


53

...21exp)1(

21exp

21exp)(

222

21

20

+⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−−−

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−−⎟

⎠⎞

⎜⎝⎛−=

ηη

ηξηξ

a

aaf(4)

Computing derivatives and then evaluating andcollecting terms, we write:

012

1

20

3)0('')0('

)0(

aaafaf

aaf

GF

GF

GF

−−=−=

−=(5)

Note that for symmetric problems (the possibility ofasymmetry is included in the complete series), we

can write [ ])0('')0(321

0 GFGF ffa −= .

ResultsPreliminary drag-coefficient results for a range

of 2-d bluff body shapes are presented in Table 1.Comparison is made with published values given inWhite (1986) and Hughes and Brighton (1991).

Table 1. Preliminary comparison between published 2-ddrag coefficient results and the theoretically based modeldeveloped here. Notice that the theoretical modelcompares adequately with published values for rapidlyseparated flows (sharp-edged and laminar) but performspoorly for smooth bodies with delayed separation.

ShapeTheor.

CD

Pub.CD Reynolds #

Rel.Error(%)

Square Cylinder (stepfunction)

2.1 2.1 Independent 0

2:1 RectangularCylinder(num. inviscid)

1.85 1.7 Independent(Re>103)

10

Equilateral triangle(apex facing flow)

1.35 1.6 Independent 15

Circular Cylinder(laminar B.L.)

1.3 1.2 103<Re<105 8

Circular Cylinder(transition B.L.)

0.63 0.6 Re≈5x105 5

Circular Cylinder(turbulent B.L.)

0.51 0.3 Re>106 70

Parabolic Cylinder,f=x(1-x)(Small disturbance)(Turbulent B.L.)

0.15 0.2 Re>>>1 25

Although our interest is primarily focused onderiving an estimate for the integrated value CD, wecan also utilize the preceding analysis to obtainestimates of the flow field behavior. FollowingTennekes and Lumley (1972), and utilizing theirvariables 2/12/1 BxlAxU s == − and given that Aand B are constants to be determined, we obtain anapproximation for the centerline velocity behavior asa function of x: 1

016.3 −= dxUU s . Of course, thisexpression is not valid for x<<1, but we expect thatthe functional form (i.e., power of x, -1) to becorrect. By way of comparison, we consider thecenterline velocity data for flow over a squarecylinder square cylinder given by Lyn et al. (1995)in Figure 2.

0.00 2.00 4.00 6.00 8.00X/D

0.00

0.20

0.40

0.60

0.80

Uce

nt/U

inf

Fit Results

Fit 1: Power, log(Y)=B*log(X)+AEquation:log(Y) = -0.914379 * log(X) + -0.150117Alternate equation:Number of data points used = 14Coef of determination, R-squared = 0.987161

Centerline Velocity DataSquare Cylinder Wake

LERP 1995

Regression

Figure 2. Centerline velocity data for flow over a squarecylinder square cylinder given by Lyn et al. (1995) withregression analysis. Note that the curve-fit expressiondecays to the –0.91 power, which is a value that compareswell with the theoretical value, -1.

ConclusionsIn this manuscript, we have modified classical self-similar, far-field, turbulent wake concepts toestimate the 2-d drag coefficient for a range of bluffbody problems. Preliminary results indicate thatdrag estimates computed by using this method arewithin approximately 10–20% of the publishedvalues for flows with large separation. The potentialvalue of this method is that it is a way to utilizepoorly resolved simulation results to provide aninexpensive estimate of body drag or that it


54

functions as the basis of a physically consistentcorrelation scheme. This methodology may be of useas a supplement to CFD and experimental solutionsin reducing the heavy computational andexperimental burden of estimating drag coefficientsfor blunt body flows for preliminary design-typestudies.

References1. Lyn, D.A.; Einav, S.; Rodi, W.; and Park, J.H.,

1995, A Laser-Doppler Velocimetry Study ofEnsemble Averaged Characteristics of theTurbulent Near Wake of a Cylinder, J. of FluidMechanics, 304, pp. 285–319.

2. Tennekes, H., and Lumley, J.L., 1972, A FirstCourse in Turbulence, MIT Press, Cambridge,MA.

3. Townsend, A.A., 1976, The Structure ofTurbulent Shear Flow, Cambridge U. Press,Cambridge, UK.

4. Van Dyke, M., 1975, Perturbation Methods inFluid Mechanics, Parabolic, Stanford, CA.

5. White, F.M., 1986, Fluid Mechanics, McGraw-Hill, NY.

6. Hughes, W.F., and Brighton, J.A.,1991, FluidDynamics, 2nd ed., McGraw-Hill, Inc., NY.

7. McCallen, R.; Couch, R.; Hsu, J.; Browand, F.;Hammache, M.; Leonard, A.; Brady, M.;Salari, K.; Rutledge, W.; Ross, J.; Storms, B.;Heineck, J.T.; Driver, D.; Bell, J.; andZilliac, G., 1999, “Progress in ReducingAerodynamic Drag for Higher Efficiency ofHeavy Duty Trucks (class 7-8),” SAE Paper1999-01-223.

8. Roy, C.; McWherter-Payne, M.; and Solari, K.,2002, RANS Simulations of a SimplifiedTractor/Trailer Geometry, UEF.

9. Schlichting, H., 1979, Boundary Layer Theory,McGraw-Hill, NY.

10. Haberman, R., 1983, Elementary AppliedPartial Differential Equations with FourierSeries and Boundary Value Problems, PrenticeHall, Englewood Cliffs, NJ.

11. Logan, J.D., 1978, Applied Mathematics, Wiley,NY.


55

H. Commercial CFD Code Benchmarking for External AerodynamicsSimulations of Realistic Heavy-Vehicle Configurations

Principal Investigator: W. David PointerArgonne National Laboratory9700 S Cass Avenue, NE-208, Argonne, IL 60439(630)252-1052, fax: (630) 252-4500, e-mail: [email protected]


Contractor: Argonne National LaboratoryContract No.: W-31-109-ENG-38

Objectives• Evaluate capabilities in standard commercial computational fluid dynamics (CFD) software for the prediction

of aerodynamic characteristics of a conventional U.S. Class 8 tractor-trailer truck.

• Develop “best practice” guidelines for the application of commercial CFD software in the design process ofClass 8 vehicles.

Approach• Develop computational models for the Generic Conventional Model (GCM) geometry.

• Benchmark the GCM simulations for computed aerodynamic drag force and pressure field distributions withNASA Ames 7-ft × 10-ft wind tunnel data.

Accomplishments• Experimental measurements and computational predictions of the vehicle drag coefficient agreed within

experimental uncertainty for the 0° yaw simulations. Experimental measurements and computationalpredictions of the pressure distribution along the surface of the vehicle agree well everywhere, except in thevicinity of the trailer base and underbody and the gap.

• Comparisons of computational predictions of the vehicle at a yaw angle of 10° using coarse mesh modelsindicate that the accuracy of the steady RANS simulation capability is reduced at high yaw angles.

Future Direction• Extend evaluation of capabilities for prediction of aerodynamic drag in cases in which there is a cross-wind

component (e.g., wind tunnel experiments where the vehicle is placed at some yaw angle) to fine meshsimulations and lower yaw angles.

• Consider alternative GCM configurations using various add-on devices to examine capabilities for theprediction of changes in drag coefficient.

• Suggest potential drag reduction design options on the basis of knowledge gained from computational effort.


56

IntroductionIn collaboration with the U.S. Department ofEnergy’s Heavy Vehicle Aerodynamic DragConsortium, Argonne National Laboratory isdeveloping guidelines for the near-term use ofexisting commercial Computational Fluid Dynamics(CFD) tools by the heavy vehicle manufacturingindustry. These guidelines are being developed onthe basis of (1) measured drag coefficients and(2) detailed surface pressure distributions from windtunnel experiments completed at NASA AmesLaboratory by using a generalized 1/8th-scaleconventional U.S. tractor-trailer geometry, theGeneric Conventional Model (GCM) [1]. Studiesconsider the effects of selection of global and near-surface mesh-size parameters and selection ofturbulence modeling strategies.

Selection of Commercial CFD SoftwareThe guidelines developed by this program areintended to be generic advice for the application of acommercial CFD software package for theprediction of heavy vehicle aerodynamic dragcoefficients. Since this market is currentlydominated by finite volume formulations, theguidelines will focus upon software using thismethodology. Preliminary guideline developmentwill be completed by using the commercial CFDcode Star-CD [2]. The Star-CD software wasselected for this purpose largely as a result of theflexibility in computational mesh development thecode offers, along with the ability to utilizepolyhedral “cut” cells and recognize both integraland arbitrary interfaces between regions of thecomputational domain. It is anticipated that theapplicability of the general guidelines to othercommercial CFD codes will be examined and thatthe extension of the guidelines to alternatecommercial CFD software methodologies, such asLattice-Boltzmann, will be pursued following theinitial development stage.

The Generic Conventional ModelThe GCM is a generalized representation of aconventional U.S. tractor-trailer truck developed byNASA Ames Research Center. The model is 1/8thscale, with approximate dimensions of 97 in. long by13 in. wide by 21 in. high. The model is mounted atthe center of the ground plane of a 10-ft-wide by

7-ft-high wind tunnel test section. Instrumentationincludes a force balance, 476 steady pressuretransducers, 14 dynamic pressure transducers, andthree-dimensional Particle Image Velocimetry(PIV). The nominal configuration (see Figure 1) is arepresentative model of a current-generation tractor-trailer truck.

Figure 1. Generic Conventional Model (GCM) standardconfiguration.

Computational ModelThe computational model employed in these studieswas developed by using the ES-Aero tool foraerodynamic drag simulation that is available as partof the Star-CD software package. The surface of thestandard configuration GCM is defined by usingapproximately 500,000 triangular surface elementsbased upon CAD data representations taken fromoptical scans of the actual model. The computationaldomain is developed on the basis of this surfacedefinition by using a semi-automated process thatbegins by creating a hexahedral mesh that issuccessively refined in smaller zones aroundvehicle, with 4 cell to 1 cell coupling employed atthe interfaces between zones. The dimensions ofhexahedral elements that make up the innermostzone are specified by the user as the near-vehiclecell size. The mesh elements near the vehicle surfaceare then further refined on the basis of local surfacefeatures identified by the user or selectedautomatically on the basis of curvature or gap width.The user specifies a minimum allowable cell sizethat limits the refinement of the mesh in this step.

By using this locally refined hexahedral mesh, theoriginal surface is “wrapped” by projecting thehexahedral mesh onto the original surface. In thismanner, the multiple components defining the GCMare merged into a single surface. The “wrapped”surface definition is then volumetrically expanded tocreate a subsurface, which is used to cut away theportions of the locally refined hexahedral mesh thatfall inside the vehicle. A brick-and-prism cell-extrusion layer is then created to fill the gap betweenthe subsurface and the “wrapped” surface. In this


57

way, the non-hexahedral cut cells are removed somedistance from the surface. A final step further refinesthe wake region and the underbody region in orderto better capture important flow features in thoseregions. An example of the mesh construction of thecomputational domain used in the GCM simulationsis shown in Figure 2.

Figure 2. Example of computational mesh structure usedin the simulation of the aerodynamic characteristics ofGeneric Conventional Model (GCM) configurations.

When using locally refined, partially unstructuredcomputational domains with substantial numbers ofnon-hexahedral cells, the standard practice ofevaluating grid convergence by uniformly refiningthe entire mesh in all directions becomes intractable.In the computational meshes used in these studies(which are based on the sizes of surface features onthe vehicle), two separate parameters determine thesize of the mesh. The near-vehicle cell sizedetermines the bulk flow resolution surrounding thevehicle, and the minimum cell size determines thelevel of resolution allowed as a result of feature-based refinement around significant features of thevehicle’s surface. Mesh-sensitivity analysesincluded in these studies examine the effects ofchanges in these parameters upon the prediction ofthe drag coefficient; however, this is not equivalentto the traditional grid convergence study for tworeasons. First, the grid is not uniformly refined in alldirections throughout the domain. Second, thevehicle surface definition cannot be exactlymaintained for all models since the final surfacedefinition depends on the local refinement of thecomputational mesh.

Bulk Flow ResolutionFive unique computational domains were generatedon the basis of the standard GCM configuration inorder to evaluate the effects of the near-vehicle cellsize parameter on the prediction of the dragcoefficient. Near-vehicle cell sizes of 16.0, 12.0,10.0, 8.0, and 6.0 mm were considered. In each case,

the minimum cell size resulting from local feature-based refinements is 12.5% of the near-vehicle cellsize, and an additional restriction is set so that aminimum of 16 elements are required for thedefinition of a circle. To ensure that the quality ofthe vehicle surface is maintained, the cell layerimmediately adjacent to the surface is refined to25% of the original size before trimming. Thecomputational domain characteristics are shown inTable 1.

Table 1. Summary of computational domaincharacteristics for evaluation of bulk cell size effects.

Near VehicleCell Size

(mm)

MinimumCell Size

(mm)

Total Numberof VolumeElements

Number ofVolume

Elements onSurface

16.0 2.0 1012338 7357412.0 1.5 1737085 126119

10.0 1.25 2345640 175105

8.0 1.0 3282426 266666

6.0 0.75 5695622 400382

Each computational domain was considered in aparametric study for evaluation of the sensitivity ofthe prediction of the vehicle drag coefficient tochanges in the bulk flow resolution. A uniformvelocity of 51.45 m/s, corresponding to a Reynoldsnumber of 1.1 × 106, was specified at the inlet, and azero gradient condition is specified at the outlet. Inthese simulations, the standard high Reynoldsnumber k-ε turbulence model and a logarithmic wallfunction are employed for prediction of turbulentkinetic energy and dissipation.

For each case, 3,000 iterations were calculated byusing Star-CD’s standard conjugate gradient solverand the PISO predictor-corrector algorithm.Convergence criteria were set to ensure that all caseswould reach 3,000 iterations before stopping. At the3,000th iteration, all residuals are less than 10-4. Inaddition to standard flow variable residualmonitoring for the mass, momentum, and energyequations, the drag coefficient of the vehicle ismonitored as the solution develops to ensure that thedrag coefficient reaches a converged value. Totalcomputational time and clock time when using16 processors for each simulation are shown inTable 2.


58

Table 2. Summary of computational cost for each caseconsidered in the evaluation of bulk cell size effects.

Near-Vehicle CellSize (mm) Total CPU Time (s) Total Clock Time (s)

16 206072 1645412 390113 2939210 417686 321828 610958 449676 2720956 188577

Predicted drag coefficients from each of the fivecases are compared with experimental data fromwind tunnel tests in Table 3. While there is a trendof improvement with reduction in near-vehicle cellsize, the effects that lead to non-linearity in the trendare not immediately clear. More detailedcomparisons of pressure distributions on the surfaceof the vehicle provide better insight into thesensitivity of the predictive capability to the bulkflow resolution. The pressure coefficient distributionon the surface of the vehicle from the case using themesh based upon an 8-mm near-vehicle cell size isshown in Figure 3 as an example of a typicalpredicted pressure distribution on the vehiclesurface. Pressure coefficient data were extractedalong the centerline of the vehicle for each case andcompared with experimental data, as shown inFigure 4. These comparisons show that thedifference in the predicted drag coefficient betweenmodels using different near-vehicle cell sizes is aresult of small differences in the pressuredistribution over the entire surface rather than largelocalized differences.

Table 3. Effects of near-vehicle cell size parameter onaccuracy of drag coefficient prediction.

Near-Vehicle CellSize (mm)

Predicted DragCoefficient

Error in DragCoefficient

experiment 0.39816 0.449 12.012 0.441 10.310 0.418 4.98 0.415 4.26 0.405 1.7

Near-Wall ResolutionFollowing the assessment of the effects of the near-vehicle cell size parameter on the accuracy of thedrag coefficient prediction, the effect of the near-wall cell size parameter was also considered. Thenear-vehicle cell size was set to 8 mm, and theminimum cell size for local refinement was reduced

(b)

(a)

(a & b) (c)(c)

Figure 3. Predicted pressure coefficient distribution onthe vehicle surface. Shown are (a) the side view of the fullvehicle, (b) the front of the tractor, and (c) the base of thetrailer.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5Pressure Coefficient, Cp

Nor

mal

ized

Hei

ght,

z/w

6 mm8 mm10 mm12 mm16 mmExperimental Data

Radiator

Windshield

Trailer Top

TrailerBase

Gap

Hood Radius

CabFairing Radius

Figure 4. Comparison of predicted pressure coefficientdistributions on the vehicle surface for various values ofthe near-vehicle cell size parameter with experimentaldata for the GCM geometry.

from 1 mm to 0.5 mm. The change in the near-wallresolution increases the number of computationalelements from 3,282,426 to 4,264,232. Whenselecting near-wall cell size limits, it is important toconsider the appropriate limits of the parameter y+,which describes the thickness of the region near thewall where the logarithmic law of the wall functionis applied. For the turbulence model and wallfunction employed in these studies, the value of y+should fall between 20 and 200. A near-wallcomputational cell that is too small will result in avalue of y+ that is too small, and a near-wallcomputational cell that is too large will result in a y+that is too large. As shown in Figure 5, the value ofy+ falls within the appropriate range for theturbulence model employed.


59

Figure 5. Values of the y+ parameter along the surface ofthe computational model of the GCM geometry when thecomputational mesh uses a near-vehicle cell size of 8 mmand a near-wall cell size limit of 1 mm.

A simulation of the flow of air over the vehicle wascompleted by using the refined computational meshfor comparison with the previous simulations. As inthe previous cases, a uniform inlet velocity conditionand a zero-gradient outlet condition were specified,and the standard high Reynolds number k-ε modelwas utilized. Convergence criteria were set so that3,000 iterations were completed, and all residualsfall below 10-4 by the 3,000th iteration. The changein the computational mesh resolution results in anincrease in the total CPU time from 610,958 s to703,027 s.

The change in the near-wall refinement parameterresults in a reduction in the error of the dragcoefficient prediction from 4.2% to 1.0%, which iswithin experimental uncertainty. The predictedsurface pressure distributions along the vehiclecenterline for both cases are shown in Figure 6,along with the experimentally measured pressuredistribution. As in the assessment of the effects of

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6


Nor

mal

ized

Hei

ght,

z/w

1 mm limit0.5 mm limitExperimental Data

Radiator

Windshield

Trailer Top

TrailerBase

Gap

Hood Radius

CabFairing Radius

Figure 6. Comparison of predicted pressure coefficientdistributions on the vehicle surface for various near wallcell size limits with experimental data for the GCMgeometry.

the near-vehicle cell size parameter, comparisons ofsurface pressure data indicate that the predicted dragcoefficient between models using different near-wallcell sizes is a result of small differences in thepressure distribution over the entire surface ratherthan large localized differences.

Turbulence Model SelectionIn all simulations completed as part of thecomputational mesh sensitivity studies, the highReynolds number k-ε turbulence model was used inconjunction with a standard logarithmic wallfunction for the prediction of turbulent kineticenergy and eddy diffusivity. While the highReynolds number k- ε turbulence model is a robustgeneral purpose turbulence model, the strongadverse pressure gradients and large flowrecirculation regions associated with the GCMgeometry may limit the applicability of steady stateReynolds Averaged Navier-Stokes (RANS)modeling strategies. Therefore, the sensitivity of thedrag coefficient prediction to the choice of twoequation steady RANS turbulence model was alsoassessed.

By using the computational mesh with a near-vehicle cell size of 8 mm and a near-wall cell sizelimit of 0.5 mm, simulations of the aerodynamiccharacteristics of the GCM model were repeated byusing five steady RANS turbulence models and theirassociated wall functions:

1. The standard high-Reynolds number k-ε modelwith logarithmic wall function,

2. The Menter k-ε SST model,3. The renormalization group (RNG) formulation

of the k-ε model,4. The Chen formulation of the k-ε model, and5. The quadratic formulation of the k-ε model.

The standard k-ε model and the k-ε SST model areidentical in the far field, but the k-ε SST modelincorporates additional detail in the near-wall regionand in separated flow regions. The RNG model issimilar to the standard k-ε model, but it includes anadditional term to account for the mean flowdistortion of the dissipation. Chen’s model is alsosimilar to the standard k-ε model, but it includes anadditional term to more effectively account for theeffects of changes in the mean strain rate on the


60

energy transfer mechanism of turbulence. Thequadratic model is a higher-order model of the k-εtype that includes non-linear terms to allow for theanisotropy of turbulence in some flow fields.

As in the previous cases, a uniform inlet velocitycondition and a zero gradient outlet condition werespecified, and the standard high-Reynolds-numberk-ε model was utilized. Convergence criteria wereset so that 3,000 iterations were completed, and allresiduals fall below 10-4 by the 3,000th iteration.The drag coefficients predicted by using each of theselected steady-RANS turbulence models are shownin Table 4. More detailed comparisons of thepredicted pressure coefficient distributions whenusing each of the selected turbulence models areshown in Figure 7. In contrast to previous studies,the differences in the predicted drag coefficient arelargely a result of localized discrepancies in thesurface pressure coefficient predictions in theregions of separated flow, as shown in Figure 8.

Table 4. Results of the evaluation of two-equationturbulence models for prediction of drag coefficients forthe GCM geometry.

Turbulence ModelPredicted Drag

CoefficientPercent Error in

PredictionExperiment 0.398 --

High-Reynolds-Numberk-epsilon Model 0.402 1.0

Menter k-ε SST model 0.401 0.8RNG model 0.389 2.3

Chen’s model 0.3919 1.61Quadratic model 0.3815 4.32

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6


Nor

mal

ized

Hei

ght,

z/w

Chen's Modelk-epsilon ModelMenter SST ModelQuadratic ModelRNG ModelExperimental Measurement

Radiator

Windshield

Trailer Top

TrailerBase

Gap

Hood Radius

CabFairing Radius

Figure 7. Comparison of predicted pressure coefficientdistributions on the vehicle surface with experimentaldata for selected turbulence models.

Figure 8. Standard deviation of the surface pressuredistribution predictions by using the selected turbulencemodels.

The differences in the predictions of the surfacepressure distribution are a direct result of differencesin the predicted flow fields. The predicted velocitymagnitude at the centerline is shown for eachselected turbulence model in Figure 9. The primarydifferences in velocity field prediction occur in therecirculation zone under the trailer and in theinteraction between the underbody flow and theseparated flow region at the trailer base. Thelocation of these differences corresponds to thelargest discrepancies between the surface pressuredistribution predictions.

Full Vehicle versus Half VehicleTo evaluate the effects of considering only half ofthe vehicle rather than the full vehicle, two modelswere created by using the full vehicle geometry.These models use the same mesh parameter settingsas the two coarsest models considered in the meshsensitivity study. The full vehicle models are basedupon near-vehicle cell sizes of 12 mm and 16 mm,with minimum near-wall cell sizes of 1.5 mm and2.0 mm, respectively. As in all previous studies,3,000 iterations were completed for each steady-state simulation, and the convergence of the dragcoefficient was monitored.

As shown in Table 5, drag coefficient predictionsshow a slight improvement in agreement withexperimental measurements when the full-vehiclemodel is used. The comparison of more detailedpressure coefficient distributions along the vehiclesurface, as shown in Figure 10, reveals that the mostsubstantial discrepancies between the full and halfvehicle model predictions occur along theunderbody and in the gap between the tractor andtrailer. The GCM geometry is, in reality, slightlyasymmetric, and the consideration of this geometricasymmetry is likely the primary difference in themodels that contributes to these discrepancies.


61

(a)

(b)

(c)

(d)

(e)

Figure 9. Comparison of predicted steady-state velocity fields for five selections of turbulence model: (a) the high-Reynolds number k-ε model with logarithmic wall function, (b) the Menter k-ε SST model, (c) the renormalizationgroup (RNG) formulation of the k-ε model, (d) the Chen formulation of the k-ε model, and (e) the quadraticformulation of the k-ε model.


62

Table 5. Comparison of drag coefficient predictions fromhalf-vehicle and full-vehicle models.

Half-VehicleNear-Vehicle

Cell Size (mm)Predicted Drag


Prediction16 0.449 12.012 0.441 10.3

Full-VehiclePredicted Drag


Prediction16 0.441 10.312 0.426 6.7

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6


Nor

mal

ized

Hei

ght,

z/w Half Vehicle - 12 mm

Half Vehicle - 16 mmFull Vehicle - 12 mmFull Vehicle - 16 mmExperimental Data

Figure 10. Comparison of predicted pressure coefficientdistributions on the vehicle surface when the full vehiclemodel is used with predicted pressure coefficientdistributions when the half vehicle model is used and withexperimental data.

Vehicles at YawAn assessment of the capabilities available incurrent-generation commercial CFD software for theprediction of aerodynamic drag characteristics ofheavy vehicles at yaw angles greater than zero isunder way. A preliminary study using coarse-meshsimulations of the GCM geometry at a yaw angle often degrees has been completed. The 10° yaw anglewas selected because as the yaw angle of the vehicleincreases through 10°, a change from a high dragstate to a low drag state occurs (see Figure 11). Theprediction of aerodynamic drag coefficients in thisregion is expected to be more difficult than at loweror higher yaw angles.

As with the no-yaw case, the study begins with aninitial assessment of the sensitivity of the dragcoefficient prediction to the base cell size in thenear-vehicle region. Since the full vehicle must be

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

-15 -10 -5 0 5 10 15Yaw Angle (degrees)

Dra

g C

oeffi

cien

t

Figure 11. Experimental measurements of dragcoefficients for the GCM vehicle geometry at yaw anglesranging from -14 to +14.

modeled when the vehicle is placed at a yaw angleother than zero, only coarse mesh models with near-vehicle cell size parameters of 12 mm or 16 mmhave been considered to date. As with the previoussimulations, 3,000 iterations were completed foreach simulation. Results of these simulations areshown (along with the zero yaw angle results) inFigure 12. In going from a no-yaw condition of zerodegrees to a yawed condition of 10°, the error in thedrag coefficient prediction increases from 10.3% to33.8% for the 16-mm case and from 6.7% to 23.5%in the 12-mm case.

By using the coarsest mesh, which is based upon a16-mm near-vehicle cell size, a preliminaryassessment of the sensitivity of the prediction ofdrag coefficient to the selection of turbulence modelhas been completed. As in the assessment of steadyRANS models for the no-yaw case, five two-equation type models were considered:

1. The high-Reynolds number k-ε model,2. The Menter k-ε SST model,3. The RNG formulation of the k-ε model,4. The Chen formulation of the k-ε model, and5. The quadratic formulation of the k-ε model.

All five models use appropriate logarithmic wallfunctions to resolve the boundary layer region.Again, 3,000 iterations were completed for eachsimulation. Results of these simulations are shownin Table 6. The Mentor k-ε SST model, which uses ak-ε model in separated flow regions and a k-ε modelelsewhere, shows the greatest improvement over the


63

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

-15 -10 -5 0 5 10 15Yaw Angle (degrees)

Dra

g C

oeffi

cien

t

measured16 mm12 mm

Figure 12. Comparison of computational predictions withexperimental measurements of drag coefficients formodels based upon near-vehicle cell sizes of 16 mm and12 mm.

Table 6. Results of the evaluation of two-equationturbulence models for prediction of drag coefficients forthe GCM geometry at a yaw angle of 10°.

Turbulence ModelPredicted Drag


PredictionExperiment 0.72955 --

High-Reynolds Numberk-ε Model 1.027 33.8

Menter k-ε SST model 0.844 14.5RNG model 1.014 32.6

Chen’s model 1.052 36.3Quadratic model 1.001 31.3

standard high-Reynolds number k-ε model. TheMentor k-ε SST model also converges much morequickly than the other models, reaching the samelevel of convergence in fewer than half the iterationsrequired when using other models. The velocity fieldfor this simulation is shown in Figure 13.

Fine Mesh Models for Full VehiclesThe number of cells used by the semi-automaticmeshing tool in the process of creating the finalcomputational mesh increases rapidly as the near-vehicle cell size is reduced. Consequently, a 64-bitcomputational platform with significant memoryavailable to the processors running the meshingtools is needed to construct the fine mesh models ofthe full truck geometry, both at no yaw and a yaw

angle of 10°. The Star-CD software has beeninstalled on a 64-bit Itanium 2 workstation with24 GB of RAM to allow the construction of morerefined computational models. A significant efforthas been dedicated to identifying and addressingissues associated with this new port of the software.It is anticipated that assessments using fine mesheswill be completed in early FY 2005.

Future WorkOngoing efforts will continue to focus on theassessment of the capabilities within currentgeneration software by using simple steady RANSmodeling strategies for the prediction of changes indrag with changes in geometry or flow conditions.These efforts will consider the standardconfiguration of the GCM geometry at additionalyaw angles, as well as the alternative configurationsof the GCM geometry shown in Figure 1. Uponcompletion of computational studies for eachconfiguration of the geometry, predictions of dragcoefficient and surface pressure distributions will becompared with experimentally measured values forthat configuration.

ConclusionsThese studies are the initial component of anassessment of the capabilities for the prediction ofheavy vehicle aerodynamic characteristics by usingcurrent generation commercial computational fluiddynamics software. On the basis of the outcomes ofthese studies, guidelines are being developed for theimmediate application of these current generationtools by the heavy vehicle manufacturingcommunity. Initial assessments have shown that fullbody drag coefficients can be predicted within lessthan 1% of the measured value, which is withinexperimental accuracy. The surface pressuredistributions can be predicted with reasonableaccuracy by using fine mesh models. The coarsestmesh model of the full vehicle provides a predictionof the drag coefficient within 11% of the measuredvalue. When the vehicle is yawed to 10°, thecoarsest model provides a prediction within 19% ofthe measured value if the Menter k-ε SST model isused.


64

References1. Satran, D., “An Experimental Study of the

Generic Conventional Model (GCM) in theNASA Ames 7-by-10-Foot Wind Tunnel,”United Engineering Foundation Conference onThe Aerodynamics of Heavy Vehicles: Trucks,Buses, and Trains, United EngineeringFoundation, New York, 2002.

2. Star-CD, version 3.150A, CD-Adapco Group,Melville, NY.

Figure 13. Predicted streamlines across the surface of the GCM geometry when the vehicle is placed at ayaw angle of 10°.


65

I. Bluff Body Flow Simulation Using a Vortex Element Method

Principal Investigator: Anthony LeonardCalifornia Institute of Technology1200 E. California Blvd.Pasadena, CA 91125(626) 395-4465, fax: (626) 577-9646, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, fax: (630)252-4289, e-mail: [email protected]

Participants:Michael Rubel, Caltech, [email protected] Chatelain, Caltech, [email protected]

Contractor: California Institute of TechnologyContract No.: DE-AC03-98EE50506

Objectives• Study application of vortex particle methods to complex truck geometries at high Reynolds numbers.

• Investigate Large Eddy Simulation (LES) in the context of such solutions.

Approach• Develop physically and mathematically correct treatments for the generation of vorticity at complex boundaries.

• Extend boundary treatment to cases of spinning bodies, such as tires.

• Reduce the computational work required to arrive at solutions by novel time-integration techniques.

• Reconcile LES theory with this framework.

Accomplishments• Implemented near-wall vorticity elements with wall stress evaluation and additional Biot-Savart term for

spinning objects.

• Performed preliminary simulations of spinning sphere flows at Re=300 for a dimensionless spin rate of 0.5 and2 spin angles.

• Implemented a multiscale time integration algorithm with clear order of accuracy and convergence properties.

• Developed a new ensemble-averaging theory for LES.

Future Direction• Compute flows around tumbling objects and spinning tires.

• Perform detailed performance measurements on the multiscale time integrator; extend to parallel computation.

• Test ensemble averaging theory on turbulent flows; ascertain whether the theory will, in fact, be able to predictLES parameters.


66

IntroductionHeavy ground vehicles, especially those involved inlong-haul freight transportation, consume asignificant part of our nation's energy supply. It istherefore of utmost importance to improve theirefficiency, both to reduce emissions and to decreasereliance on imported oil.

At highway speeds, more than half of the powerconsumed by a typical semi truck goes intoovercoming aerodynamic drag, a fraction thatincreases with speed and crosswind. Thanks to bettertools and increased awareness, recent years haveseen substantial aerodynamic improvements by thetruck industry, such as tractor/trailer heightmatching, radiator area reduction, and sweptfairings. However, there remains substantial roomfor improvement as understanding of turbulent fluiddynamics grows.

Our group’s research effort focuses on vortexparticle methods, a novel approach forcomputational fluid dynamics (CFD). Wherecommon CFD methods solve or model the Navier-Stokes equations on a grid that stretches from thetruck surface outward, vortex particle methods solvethe vorticity equation on a Lagrangian basis ofsmooth particles and do not require a grid.

We are working to advance the state of the art invortex particle methods by improving their ability tohandle the complicated, high-Reynolds-number flowaround heavy vehicles. Specific challenges that wehave addressed in the past year include findingstrategies to accurately capture vorticity generationand resultant forces at the truck wall, handle theaerodynamics of spinning bodies (such as tires),reduce computation time through improvedintegration methods, and conduct theoreticaltreatment work on large eddy simulation (LES)turbulence modeling.

Near-Wall VorticityWe developed a representation of near-wall vorticityby means of an attached regularized sheet. Thissheet has several roles. It interacts viscously with therest of the flow, receives contributions of elements

close to the wall during a redistribution, and helps incapturing the high-vorticity gradients near the wall.

This last role is critical if one needs to accuratelymeasure stresses at the wall. In addition, weintroduce a correction that takes into account thegradient of vorticity, which can be estimated fromthe solution of the panel solver.

Spinning BoundariesThe flow around spinning objects is of particularinterest because it is encountered around the wheelsof heavy vehicles and will interact with the rest ofthe flow. It is also interesting because of its impacton the problem of splash-and-spray. Because we usea vorticity-based formulation along with acomputation of velocities by Biot-Savart, we need toaccount for the vorticity inside any rotating object.This term, a volume integral, is not the best suitedfor our method, which uses a surface mesh torepresent boundaries. We thus switch to a surfaceintegral by application of Gauss’s theorem. In 2004,preliminary results were gathered for twoconfigurations involving a spinning sphere: therotation axis of the sphere was aligned with thestream or set perpendicular to it. Both cases werecomputed for Re=300 and a dimensionless spinvelocity WR/U∞ of 0.5, where W is the angularvelocity. The case of a perpendicular axis is ofparticular interest because of its resemblance to thegeometry of a spinning tire. The problem is notsymmetric and does not need to be perturbed totrigger shedding; Figure 1 shows vortex structures

identified by contours of )(21 uuTrQ ∇⋅∇−= .

Even though the problem is still marked by theinitial transient, the sphere sustains an importantamount of lift (Figure 4, Figure 3) and develops awake that comprises a system of counter-rotatingvortices.

The stream-wise rotation is also of interest becauseof the various wake structures that are observedacross the ranges of Reynolds numbers and spinrates. Our results show the transition from the axi-symmetric wake typical of a low Reynolds numberto an asymmetric periodic wake (Figure 2).


67

T = 6 T =7.125

T = 8.25 T = 9.375Figure 1. Spinning sphere at Re=300, span-wise rotation (axis of rotation is perpendicular to free stream direction):vorticity structures (q=0 surfaces).

TimesteppingBecause contemporary CFD is limited by the powerof available computers, it is of interest to reduce thework necessary to compute a given flow. One majorarea of inefficiency that remains largely untapped isthe time integration process.

In Figure 5, one sees a frequency distribution of thestrengths of vortex particles from one snapshot of avery low Reynolds number (1000) truck modelsimulation. By dimensional analysis, the localtimescale is inversely proportional to the local strainrate tensor norm, which, for the purpose of thisillustration, is taken to be particle strength (a choicethat is approximate in that it neglects the symmetricpart of the tensor). In a conventional timestepper,even an adaptive one, the CFL condition limitsintegration rate according to the strongest gradient inthe flow. However, even at this unrealistically smallReynolds number, the mean strength is hundreds orthousands of times smaller than that of the strongest

particles, so most of the flow is being over-resolvedby the same factor. Performing timesteps that areadaptive per-particle, rather than per-step, couldpotentially reduce the computational workload byorders of magnitude.

Some multiscale integration techniques areavailable, but they are not suitable for vortex-basedfluid flow problems, which operate over acontinuous range of scales and involve fairlycomplicated tree-based right-hand-side evaluation.The goal of this phase of research has beendevelopment of a new multiscale time-integrationscheme that is tailored to vortex particle methods.

Such a method has been developed and refined overthe course of several years, and it is now beginningto yield results. In Figure 6, one sees in the leftcolumn several snapshots of a simple vortex particleflow developing in two dimensions, withcorresponding particle-specific timesteps on the


68

T = 7 T = 9

T = 11 T = 13Figure 2. Spinning sphere at Re=300, stream-wise rotation (axis of rotation is aligned with free stream direction):vorticity structures (q=0 surfaces).

Figure 3. Spinning sphere at Re=300, span-wise rotation:drag coefficient, shear component (dotted), pressurecomponent (dashed), total (solid) versus time.

Figure 4. Spinning sphere at Re=300, span-wise rotation:lift coefficient, shear component (dotted), pressurecomponent (dashed), total (solid) versus time.


69

Figure 5. Vortex particle strength distribution aroundGTS truck model.

Figure 6. Stages in asynchronous time integration of twovortex patches: strengths (left) and timesteps (right).

right. The most significant challenge in developingthe method was achieving decent scaling for largenumbers of particles; the latest incarnation scaleslinearly with the total number of timesteps across allparticles, as required.

Rigorous order-of-accuracy estimates have beenderived (the method can be made accurate to anyorder) and a number of successful tests have beenperformed, although more will be required. A paperdetailing the method is in progress.

Theoretical LES WorkThere is ongoing debate on the relationship of LESand Reynolds-averaged Navier-Stokes (RANS)solutions to the filtered or time-averaged directnumerical simulations (DNS) they are designed tomodel. Because of the chaotic nature of turbulence,the modeled solution is not generally the same resultone would obtain by applying its simplifyingassumption to an exact solution. The problem is notmerely an academic one; understanding how amodel relates to the flow being modeled is essentialfor choosing parameters correctly, which, in turn, isessential for finding and interpreting computedturbulent flows in the context of heavy vehicleaerodynamics.

We are investigating the implications of a newtheoretical concept that treats LES as an explicitensemble-averaging procedure. This is still a newidea, but there is hope that it will be applicable tochoosing parameters for LES models. It has beentested to an extent on the Lorenz equations. The firstnontrivial test on 1-D Burgers’ equation is expectedto be complete soon; if the test works, the methodshould be straightforward to extend to 3-D Euler andNavier-Stokes equations.

ConclusionsVortex method development continues mostlyaccording to plan; developments in FY 2004 mean itshould now be possible to simulate complicatedflows around truck bodies, including those aroundrotating tires. Time integration techniques haveimproved, although these improvements are not yetbackported into the main code. Work began ondevelopment of Large Eddy Simulation ensembletheory, and preliminary tests to prove or disprove itsusefulness will be conducted in the near future.


70

II. Thermal Management

A. Cooling Fan and System Performance and Efficiency Improvements

Principal Investigator: R.L. DupreeCaterpillar, IncorporatedTechnical Center, Building EP.O. Box 1875, Peoria, Illinois 61654(309) 578-0145, fax: (309) 578-4277, e-mail: [email protected]

Investigator: D.C. MessmoreCaterpillar, IncorporatedTechnical Center, Building EP.O. Box 1875, Peoria, Illinois 61654(309) 578-8715, fax: (309) 578-4277, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Philip S. Sklad(865) 574-5069, fax: (865) 576-4963, e-mail: [email protected]

Contractor: Caterpillar, IncContract No.: DE-FC04-2002AL68081

Objective• Develop cooling system fans and fan systems that will allow off-highway machines to meet Tier 3 emissions

regulations and reduce spectator sound levels with improved fuel efficiency and within the functionalconstraints of machine size.

Approach• Task 1 is to develop a large, high-performance axial fan.

• Task 2 is to develop an active fan shroud for use with large axial fans.

• Task 3 is to develop fan modeling techniques to improve the ability to predict fan performance and componentairflows in cooling systems with side-by-side heat exchangers.

• Task 4 is to demonstrate the performance of a lab-developed swept blade, mixed flow fan in an off-roadmachine.

• Task 5 is to develop a high-efficiency continuously variable speed fan drive for use in high-horsepower fansystems.

• Task 6 is to develop cooling system air filtration concepts that can allow more dense (higher performance) heatexchangers to be used in off-road machines.

Accomplishments• Task 1 — Second generation fan has been built and tested from designs developed by using generic algorithms.

Performance of second-generation design still does not meet performance goals for flow and efficiency.

• Task 2 — Michigan State University has completed performance measurement of a radiused inlet/radiusedoutlet fan shroud, with and without secondary airflows. The addition of a secondary airflow to the shroud does


71

not provide any more performance in the axial flow region of the fan performance curve, but it doessignificantly change the stall point of the fan.

• Task 3 — Caterpillar has completed initial correlation with fan flow data measured at Michigan StateUniversity. Results are the basis of this report.

• Task 4 — Task was completed in 2003. Fan was not able to meet performance goals because of significantperformance loss when it was installed close to the front of the engine.

• Task 5 — Task was terminated in 2003 during a performance review. Dual ratio clutch is not able to meetefficiency goals, and alternative CVT concepts will not fit within the space allowed.

• Task 6 — Completing initial lab evaluation of the use of grounding, application of both ac and dc voltages,ionization of the incoming air, ultrasonics, and the use of mechanical vibrations to reduce dust accumulation onheat exchanger surfaces.

Future Direction• Task 1 — Genetic algorithms cannot provide an acceptable fan design. Will proceed forward with classical fan

design methods.

• Task 2 — Michigan State University is evaluating the performance of the active fan shroud concept with twovariations of conventional sheet metal fans to determine the sensitivity of shroud performance to fan design.

• Task 3 — Caterpillar is completing a design of experiments to provide final quantification on the sensitivity ofCFD performance predictions to various modeling parameters.

• Tasks 4 and 5 — Tasks completed. Final documentation due.

• Task 6 — Results of initial lab evaluations are being reviewed to select the technology most appropriate forfurther lab evaluation.

IntroductionAccurate CFD analysis of axial fan performance inmachine cooling systems is critical to the ability tomeet upcoming Tier 3 emissions regulations in off-highway machines. The analyst must be able toprovide accurate estimates for both overall fanairflow and individual component airflows insystems with side-by-side heat exchangers, or thefinished cooling system will not be able to providethe required inlet manifold temperatures needed forTier 3 emissions compliance. Many aspects of theCFD modeling process can affect the final solution,but to date there has not been a determination ofwhich aspects are most critical, nor has there been adefined set of modeling guidelines available to theanalyst. Toward this end, an extensive set ofmeasured data was obtained on an axial fan in an“installed condition” at Michigan State University in2003. The data were used as a baseline from which aset of CFD modeling guidelines could be developed.The purpose of this project is to determine thecritical modeling factors, and the degree to whichthey must be controlled, in order to obtain consistent

and accurate airflow predictions of axial fans andindividual heat exchanger airflows in off-highwaymachine cooling systems.

Project DeliverablesAt the end of this project, the modeling elementscritical to consistent, accurate CFD models of axialfans in installed conditions will have been developedand documented.

Baseline ModelCooling system CFD analysis typically involvessimulation of cooling system geometry that includesan axial fan within a shroud with various heatexchangers and other flow obstructions. Models arecreated in commercially available Fluent® CFDsoftware by using the multiple reference frame(MRF) formulation for steady-state analysis ofrotating machinery. There is no “standard” methodfor generating these CFD models, and analysts eachhave their own techniques that they use in anyparticular modeling situation. Overall system flowrate and fan performance degradation (i.e., what


72

percentage of the fan airflow is reduced when a fanis in an installed condition vs. when it is evaluated inan industry standard free inlet/free outlet [FIFO]performance test such, as AMCA® [Air Movementand Control Association] Standard 210) are thetypical areas of primary concern, with a lesseremphasis on prediction of airflow through individualheat exchangers mounted in parallel (side-by-side).The individual component airflow data are importantbecause air-to-air aftercooler (ATAAC) cores areoften much more restrictive to fan airflow thantypical radiator cores. The net result is that theanalyst could accurately predict the overall flow, butinaccurately predict the flows through the individualcores, resulting in an incorrect estimation of theperformance of both the radiator and the ATAAC.Unfortunately, test data are seldom available againstwhich to validate these CFD models. On theoccasion that such test data do become available, itis usually some months after the simulations havebeen completed, and the models are generally not re-exercised in order to improve on their predictivecapability.

To standardize the modeling process, an extensivetesting program was undertaken with Michigan StateUniversity (MSU). (The MSU test facility and testmethods are described in “PerformanceMeasurements and Detailed Flow FieldObservations for a Light Truck Cooling Fan” byJ.F. Foss [SAE paper 971794].) A 482.6-mm-diameter axial fan was inserted into a 6.35-mm-thickknife-edge shroud at 50% projection (50% of theprojected width was upstream of the upstream faceof the shroud), which was then placed in one face ofan air chamber. Downstream of the fan, in the airchamber, a helper fan regulated the amount ofdownstream restriction against which the test fanworked. Pressure rise (from atmospheric pressureupstream of the fan to static pressure in thedownstream air chamber) and total airflow throughthe system were measured under a variety ofconditions. Figure 1 shows the fan mounted withinthe test facility.

The testing would represent an AMCA® FIOF fanperformance test. The fan would then be modeled inFLUENT® as an MRF model. The second stepwould be to test (and model) a fan system consistingof heat exchangers mounted in front of the fan, todevelop an acceptable performance model. The

Figure 1. Diagram of fan in test facility.

third step would be to develop a design ofexperiments to determine the impact of the variousmodeling variables on the accuracy of the CFDpredictions.

Initially, the fan was modeled in the free inlet/freeoutlet condition (FIFO, no flowstream obstructionsother than the fan shroud). Figure 2 shows the fanfrom below, installed in the top of the exit chamberwith the attached driveshaft.

The first critical element to be evaluated was the sizeof the MRF volume. Initially, it was thought that asmaller MRF volume would produce acceptableresults, but subsequent simulations proved this not to

Air outlet

Drivesh

Air

Figure 2. Image of fan as installed in the top of exitchamber.


73

be the case. Three iterations were made to MRF-volume size: (1) both the axial and radial dimensionswere increased, (2) the axial dimension was furtherincreased while maintaining the new radialdimension, and (3) both the axial and radialdimensions were further increased. Figure 3 showsthe progression in MRF size.

Figure 3. Progression in MRF size.

Initial simulations were run at relatively lowpressure rise (204 Pa), resulting in primarily axialflow from the fan. The third MRF size brought theoverall flow rate to within about 5% of the measuredvalue. At this point, the model was run with 700 Papressure rise, which produces significant radial flowfrom the fan. Poor correlation led to the MRF’s finaldimensions. The effect on solution accuracy isshown in Figure 4a and b.

The significant error at high pressure rise (radialflow) is likely because the MRF was expanded tothe point at which it included non-axisymmetricstationary features. This introduces some falsevelocities resulting from the mathematicallyimposed fluid rotation, in addition to reflecting someflow back into the MRF, which will be discussedlater. The final MRF was used throughout theremaining modeling effort as the best compromisebetween MRF size and fan flow prediction accuracyover the full range of fan performance. Table 1summarizes the recommended MRF volume.

A FIFO fan performance curve showing MichiganState test data and CFD predictions is shown inFigure 5. AMCA®-recommended tolerance bandsfor flow measurement variability from facility to

Effect of MRF Diameter on Airflow

-70-60-50-40-30-20-10

0102030

0.5 1 1.5 2

MRF Dia/Fan Dia

Per

cen

t V

aria

tion

fro

m A

vg. M

eas.

204 Pa Press Rise 700 Pa Press Rise

a.

Effect of MRF Axial Length on Airflow

-70-60-50-40-30-20-10

0102030

0 0.02 0.04 0.06 0.08 0.1

Distance, Blade-to-MRF Face/Fan Dia

Per

cen

t V

aria

tion

fro

m A

vg. M

eas

204 Pa Press Rise 700 Pa Press Rise

b.

Figure 4. Effect of solution accuracy on fan parameters.

Table 1. Recommended MRF volume.

MRF Diameter as aPercentage of Fan

Diameter

Axial Length from Leading/TrailingEdge of Fan to MRF Face as a

Percentage of Fan Diameter150% 10%

facility from its Publication 211-94 are shown forreference (approximately 3–5% error band). Exceptright at the transition zone, CFD predictions with thebaseline model fell within the AMCA® -recommended tolerance.

Simulated InstallationThe CFD model was then modified for the installedconditions. Upstream of the fan (outside of the airchamber) was a mock-up simulating a machine


74

Figure 5. FIFO fan performance.

cooling system with side-by-side heat exchangercores. The mock-up was divided into two sectionsalong the axis of the fan vertical centerline. Themock-up, shown in Figure 6, contained a five-row,11-fin/in. heat exchanger, as well as slots for holdingperforated sheet metal screens. The use of varyingnumbers of screens allowed the core restriction to bechanged, but the radiator itself did not need to bechanged for the duration of the testing.

A sketch of the mock-up mounted above the exitplenum is shown in Figure 7. Inlet air comes fromthe atmosphere above the mock-up, is drawnthrough the fan, is then discharged into the exitplenum, and finally exits out through the flowmeasuring system under the floor.

“Installed conditions” consisted of the heatexchanger alone in the mock-up (low symmetricalrestriction), the heat exchanger plus two screens oneach side of the mock-up (high symmetricalrestriction), and the heat exchanger plus two screenson one side of the mock-up and the heat exchanger

RadiatorScreens

Figure 6. Mock-up of machine cooling system.

Figure 7. Illustration of mock-up mounted aboveexit plenum.

alone on the other side (asymmetricalrestriction).The symmetrical conditions simulatedfull-plane cores in front of the fan. Theasymmetrical condition simulated partial-planecores, resulting in uneven loading across thediameter of the fan. CFD predictions were againcompared to MSU test data. Correlations betweenpredicted and measured values for both low-symmetric and high-symmetric loading are shown inFigures 8 and 9. Notice also that restriction curvesfor specific diameter (Ds) values of 1.6 and 1.8 areincluded. (See the Appendix for a description ofspecific diameter.) It is particularly important thatthe model correlate well in this region.

The AMCA® -recommended tolerance bands areagain shown around each set of test data. In general,the CFD model under-predicted flow and pressure atany given point, with the prediction falling off moresharply at higher flows and lower pressure. Tocompensate for this, fan speed was increased by 5%above measured speed (a 5% “boost” to 2100 rpm).This resulted in a 5% increase in overall airflow,with a corresponding increase in pressure rise. It canbe seen that just a 5% boost results in considerableimprovement in the correlation between predictedand measured. The reason the adjustment isnecessary has to do with inaccuracies related to theMRF formulation for steady-state solution ofrotating flows.


75

Figure 8. Correlations between predicted and measuredvalues for low-symmetric loading.

Figure 9. Correlations between predicted and measuredvalues for high-symmetric loading.

In the present case, the shroud and the framing ofthe plenum all work to reflect air coming radially offthe fan back into the MRF volume. This situationwould likely exist in actual underhood installationsas well. Flow traversing the MRF boundaries in bothdirections is known to cause problems for the MRFformulation for rotating fluids. Fluent® recommendsusing sliding mesh methods under thesecircumstances. One operating condition, low-symmetrical loading with 190-Pa pressure rise, wasmodeled by using the sliding mesh method. Byusing MRF, the model predicted an overall flow ratethat was 13% below the measured flow rate. Withthe 5% fan boost, predicted flow rate was 5% flowrate below the measured flow rate. By using thesliding mesh, without fan boost, the predicted flow

rate was 0.4% higher than measured flow rate. Thiswas only one simulation point, but the result isencouraging. However, because of the transientnature of the solution, computing time was severaltimes longer than the steady-state solution withMRF, with the fan requiring two completerevolutions before converging to a steadystate flow.Future software and/or computer developments mayreduce the computing, but the present state of the artdoes not allow the cost and CPU time required touse sliding mesh models for cooling systemperformance analysis. An MRF model with a 5% fanboost provides a more economical solution.

MRF was also used to predict fan performance withdifferent fan projections. However, radial airflowfrom the fan reflecting off the shroud back into theMRF volume resulted in incorrect prediction of boththe performance values and trends. This is asituation where sliding mesh may be able to provideacceptable performance predictions, once solutiontime becomes more reasonable.

The CFD model was also exercised by usingasymmetric fan loading. The objective was not onlyto match the fan performance curve, but also tomatch measured flow through each side of themock-up. Figure 10 shows the fan performancecurve correlation.

Again, it can be seen that the prediction drops offquickly as flow rate increases and pressure risedecreases. However, using a 5% fan boost results inacceptable predictions at and near the Ds range of

AXIAL INSTALLED Asymmetric Restriction at 2000 RPM

0

100

200

300

400

500

600

700

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Volume Flow Rate (m^3/s)

Pres

sure

(Pas

cals

)

TSFL Asymmetric LTLUTLCFD PredictionCFDw/ 5% fan boostDs=1.8Ds=1.6

Figure 10. Fan performance curve correlation.


76

interest. Flows through the individual heatexchangers were also compared with measured data.Measured test data showed a fairly consistent split of58:42; that is, 58% of the flow went through thelow-restriction side (heat exchanger alone), and 42%of the flow went through the high-restriction side(heat exchanger plus screens; see Table 2). Althoughthe CFD predictions displayed a slight flow-rate/pressure rise effect, the error from the measuredvalues was less than 5%.

Table 2. Total flow through each side of mock-up.

Percent of Total Flow through Each Side of Mock-Up(at a given total flow rate)

m3/s Low-restriction side High-restriction side1.41 56.0 44.01.30 55.8 44.21.12 55.2 44.80.94 54.8 45.20.69 54.3 45.7

SummaryHighlights of the progress made to date include:

• Identified MRF volume size as an extremelyimportant parameter affecting fan performanceprediction accuracy for both FIFO and installedfan performance predictions.

• Successfully developed a model of an axial fanwith upstream flow restrictions installed in theMichigan State University test facility.

• Exercised the model for upstream symmetricallow- and high-restriction and upstreamasymmetrical restriction.

• By using a 5% fan boost, CFD modelpredictions fell within AMCA® -recommendedtolerance bands.

• The existing model provided a basis fordetermining what factors go into creating asuccessful model.

• A design of experiments was generated toevaluate factors that contribute to a successfulCFD model and to determine the most criticalmodeling factors.

ConclusionsA CFD model of an axial fan in an installedcondition has provided good correlation with anextensive set of detailed flow and pressuremeasurements taken under a variety of conditions.This model has led to the development of a design ofexperiments to evaluate the factors that contribute toa successful fan model. The design of experiments isbeing run, and results will be evaluated to determinethe modeling critical success factors and the extentto which they must be controlled.

AppendixSpecific diameter relates the relative fan size to thegiven restriction. This is evident by noting thedefinition of the system restriction coefficient (K)and its relationship to specific diameter:

K = Pt / Q2σ

Ds = k3DK0.25

Physical factors may be incorporated; thesegenerally follow from flow and pressurecoefficients:

Specific diameter = Pressure Coeff0.25/Flow Coeff0.5

Some sources use two other interpretations ofspecific diameter, which should be clarified to avoidconfusion.

Specific diameter may be used to indicate thediameter fan required for a unit pressure, unit flow,and unit density, assuming the fan operates atstandard density. This results in an additional factorof density to the one-fourth power.

Often specific diameter of a fan refers to singleoperating point in terms of specific diameter forwhich fan efficiency is maximum. A given fandesign is then given a representative “specificdiameter” number. This specific diameter number isrelated to the aerodynamic fan type.

Some sources also use specific radius as analternative to specific diameter.


77

B. Efficient Cooling in Engines with Nucleate Boiling

Principal Investigator: J.R. HullArgonne National Laboratory9700 South Cass Avenue, Building 335, Argonne, IL 60439(630) 252-8580, fax: (630) 252-5568, e-mail: [email protected]



Objectives• Investigate the potential of two-phase flow in engine cooling applications.

• Determine the limits on two-phase heat transfer (occurrence of critical heat flux or flow instability).

Approach• Experimentally determine heat transfer rates and critical heat fluxes in small channels with water and 50%

ethylene glycol in water mixture.

• Perform experiments over a concentration range of ethylene glycol in water.

• Perform experiments with alternative fluids.

Accomplishments• Completed experimental tests and data analysis on single- and two-phase heat transfer over a concentration

range of ethylene glycol in water mixtures (concentrations 40/60, 50/50, and 60/40).

• Developed a general correlation for heat transfer coefficients for the prediction of boiling heat transfer rates offlow boiling in small channels including refrigerants, water, and ethylene glycol/water mixtures.

• Modified Chisholm’s correlation with a concentration factor to better predict pressure drops for ethyleneglycol/water mixtures.

Future Direction• Build a vertical-flow test section to study the impact of vertical vs. horizontal flows on two-phase heat transfer.

• Perform two-phase heat transfer experiments involving ethylene glycol/water mixtures with vertical flows toprovide complete information for nucleate-boiling cooling system design.

• Perform systematic experiments with alternative fluids.


78

IntroductionAnalyses of trends in the motor vehicle developmentsector indicate that future engine cooling systemswill have to cope with greater heat loads because ofmore powerful engines, more air conditioning, morestringent emissions requirements, and additionalauxiliary equipment. Moreover, there is considerableinterest in reducing the size of cooling systems toobtain a better aerodynamic profile. To meet theseconditions, it is necessary to design cooling systemsthat occupy less space, are lightweight, have reducedfluid inventory, and exhibit improved performance.Among various new cooling systems proposed byresearchers, the nucleate-boiling cooling system hashigh potential to meet these challenges. Order-of-magnitude higher heat transfer rates can be achievedin nucleate-boiling cooling systems when comparedwith conventional, single-phase, forced-convectivecooling systems. However, successful design andapplication of nucleate-boiling cooling systems forengine applications requires that the critical heatflux and flow instability not be reached. Therefore, afundamental understanding of flow boilingmechanisms under engine application conditions isrequired to develop reliable and effective nucleate-boiling cooling systems.

Cooling engine areas such as the head region oftencontain small metal masses that lead to small coolantchannels. This geometry, in turn, leads to low massflow rates in order to minimize the pressure drop.Although significant research has been performed onboiling heat transfer and the critical heat fluxphenomenon, results on the conditions necessary forengine cooling systems are limited. The purpose ofthe present study is to experimentally investigate thecharacteristics of coolant boiling, critical heat flux,and flow instability under conditions of smallchannel and low mass flux.

The test apparatus used in this investigation wasdesigned and fabricated to study boiling heattransfer, critical heat flux, and flow instability offlowing water, ethylene glycol, and aqueousmixtures of ethylene glycol at high temperatures (upto 250oC) and low pressures (<345 kPa). Figure 1shows a schematic of the apparatus. It is a closed-loop system that includes two serially arrangedpumps with variable speed drives, a set offlowmeters, an accumulator, a preheater, a testsection, and a condenser. The flowmeter set,including various types and sizes, was chosen tocover a large range of flow rates and was calibratedwith the weighing-with-stop-watch technique. Theestimated uncertainty in the measurements of flow

ToutE

3" 3" 3" 3" 3"1" 1"6"6" 1"

I

6"

Ta Tb Tc Td Te Tf Tg Th TiTj

3'

∆p

Tin

440 micron filter

Pyrex window

Condenser Water out

Water in

Relief valveFill port

ISO SG

TP RP

RGA1 RGA2

Refractometer

Drain port

2" 2" 2"

T1 T2 T3 T4

ISO

Sorensen DCR 16-625T

EMHP 40-450-D-11111-0933

Vent port

Drain port

Expansion tank

Pressure gauge

Nitrogen/Air in

TFM

Pump 2Rotameter

Turbometer

T5

pout

Ta'

17"12.5"3"

Tb' Tc'

Tin '

Max

Pump 1

∆p

Figure 1. Schematic diagram of nucleate-boiling cooling test apparatus.


79

rates was ±3%. The bladder-type accumulator allowsfor stable control of system pressure. The preheaterprovides a means to set the inlet temperature of thetest section at various desired levels. The preheaterand the test section were resistance-heated withcontrollable DC power supplies. Provisions weremade to measure temperatures along the test sectionfor calculating heat transfer coefficients. Thepressures and temperatures at the inlet and outlet ofthe test section were also measured. Pressuretransducers and thermocouples were calibratedagainst known standards.

The estimated uncertainties in the measurements ofpressures and temperatures were ±3% and ±0.2°C,respectively. As a safety precaution, the preheaterand test section were provided with high-temperature limit interlocks, to prevent them fromoverheating. After leaving the test section, the two-phase flow was condensed into a single-phase flow,which returned to the pumps to close the system.

A data acquisition system, consisting of a computerand a Hewlett-Packard multiplexor, was assembledto record outputs from all sensors. A data acquisitionprogram was written to include all calibrationequations and conversions to desired engineeringunits. The data acquisition system provides not onlyan on-screen display of analog signals from allsensors and graphs of representative in-stream andwall-temperature measurements, but also a means ofrecording temperature measurements and pertinentinformation such as input power, mass flux, outletpressure, pressure drop across the test section, andoutlet quality, for further data reduction.

Results and DiscussionDuring FY 2004, both experiments and analysiswere completed on boiling heat transfer withethylene glycol/water mixtures over a concentrationrange (concentrations 40/60, 50/50, 60/40). Themain results are reported below.

Boiling CurveFigure 2 shows the heat flux as a function of wallsuperheat for boiling water and ethyleneglycol/water mixtures in small channels. As it can beseen from Figure 2, the saturation boiling in smallchannels can be divided into three regions, namely,the convection-dominant-boiling region, the

nucleation-dominant-boiling region, and thetransition-boiling region. Both convective heattransfer and boiling heat transfer exist in all threeregions, but their proportions are different in thesethree regions. In the convection-dominant-boilingregion, the wall superheat is low, usually less than afew degrees centigrade. Although there is boilingheat transfer, the dominant mechanism is convectiveheat transfer. As a result, the mass quality and heattransfer rate are quite low, in comparison to those inthe other two regions. In the nucleation-dominant-boiling region, the wall superheat is higher than thatin the convection-dominant-boiling region but lowerthan certain upper limits that depend on mass flux.Opposite to the convection-dominant-boiling, theboiling heat transfer in the nucleation-dominant-boiling region is so developed that it becomesdominant, and the heat transfer rate is much higherthan that in convection-dominant-boiling region. Ascan be seen from Figure 2, the heat flux in thisregion is independent of mass flux and can bepredicted with a power-law function of wallsuperheat. This characteristic was used in correlatingthe heat transfer data. In the transition-boilingregion, the wall superheat is relatively high. Theheat flux in this region is also high and close to thecritical heat flux. The boiling in this region isunstable, and a small change in the heat flux willresult in a big change in wall superheat. If the heatflux increases further, it is possible for the system toreach the critical point, producing an undesirablylarge jump in the wall superheat.

The above discussion shows that nucleation-dominant boiling is desired in engineeringapplications for both high heat transfer rate andstable flow boiling without reaching the criticalpoint.

Two-Phase Pressure DropThe concept of two-phase multipliers proposed byLockhart and Martinelli and the correlation of thosemultipliers by Chisholm were used to compare withthe present experimental data. As can be seen fromFigure 3, the experimental data are in reasonableagreement with the Chisholm predictions both invalues and trends, although the Chisholm correlationslightly overpredicts the experimental data.


80

1

10

100

1000

0.01 0.1 1 10 100

G = 50 kg/m2sG = 76 kg/m2sG = 103 kg/m2sG = 129 kg/m2sG = 151 kg/m2s

Hea

t flu

x (k

W/m

2 )

Wall superheat (oC)

convection dominant

nucleationdominant

3

transition boiling

p = 200 kPa, d = 2.98 mm

(a) water

1

10

100

1000

0.01 0.1 1 10 100

G = 38-57 kg/m2sG = 64-76 kg/m2sG = 86-105 kg/m2sG = 115-135 kg/m2sG = 140-175 kg/m2s

Hea

t flu

x (k

W/m

2 )

Wall superheat ( oC)

convection dominant

nucleationdominant

transition boiling

p = 145 kPa, d = 2.98 mm

(b) 40/60 ethylene glycol/water mixture

4.50

1

10

100

1000

0.01 0.1 1 10 100


Hea

t flu

x (k

W/m

2 )

Wall superheat (oC)

convection dominant

nucleationdominant

transition boiling

4.50

p = 160 kPa, d = 2.98 mm

(c) 50/50 ethylene glycol/water mixture

1

10

100

1000

0.01 0.1 1 10 100


Hea

t flu

x (k

W/m

2 )

Wall superheat ( oC)

convection dominant

nucleationdominant

transition boiling

p = 142 kPa, d = 2.98 mm

(d) 60/40 ethylene glycol/water mixture

4.50

Figure 2. Heat flux as a function of wall superheat.

10 0

10 1

10 2

10 3

10 0 10 1 10 2 10 3

Chi

shol

m p

redi

cted

(φ FL

)2

Experimental ( φFL

) 2

Figure 3. Two-phase frictional pressure gradient.

To better predict the experimental data and to takethe concentration factor into account, the constantparameter C =12 in Chisholm’s correlation wasmodified into a function of the volumeconcentration, v , of ethylene glycol/water mixtures.Thus modified, Chisholm’s correlation becomes:

h =1+12 1− 2.8v(1− v)[ ]

X+

1

X 2 .

This correlation reduces to Chisholm’s correlationfor both pure water ( v = 0) and pure ethylene glycol( v =1). In Figure 4, the experimental data arecompared with the predictions of the modifiedChisholm’s correlation. This modification improvesthe predictions both in values and trends.


81

10 0

10 1

10 2

10 3

10 0 10 1 10 2 10 3

Mod

ified

Chi

shol

m p

redi

cted

(φ FL

)2

Experimental ( φFL

) 2

Figure 4. Two-phase frictional pressure gradient.

Heat Transfer CoefficientIn the present study, the nucleation-dominant boilingdata have the following characteristics:

(a) Although both convective heat transfer andnucleate-boiling heat transfer exist, the dominantheat transfer mechanism is nucleate boiling. Sincenucleate-boiling heat transfer rate is much higherthan convective heat transfer, the latter can beneglected.

(b) As shown in Figure 2, the boiling heat transfer isdependent on heat flux but almost independent ofmass flux, which means that, for a certain fluid, theboiling heat transfer coefficient can be expressed asa function of heat flux.

(c) The heat transfer coefficients have differentdependences on heat flux for different fluids.Therefore, to get a general correlation for boilingheat transfer coefficients, it is necessary to includefluid properties in the correlation.

(d) Argonne researchers employed thedimensionless parameter combination form ofBoiling number, Weber number, and liquid-to-vapordensity ratio in developing different predictedcorrelations for boiling heat transfer coefficients todifferent fluids; and the predicted results are quitegood.

On the basis of the above facts, we extended theproperty term to include the liquid-to-vaporviscosity ratio and were able to correlate boiling heat

transfer data for water, 50/50 ethylene glycol/watermixture, refrigerant 12, and refrigerant 134a.

h =135000(BoWel0.5)0.5 (ρl ρv )−0.5(µl µv )0.7[ ]−1.5

In the above equation, the Boiling number, Bo, andthe Weber number, Wel , are defined respectively byBo = ′ ′ q (Gifg ) and )(2 σρ ll DGWe = , where ρis the density, µ is the viscosity, G is the mass flux,D is the diameter, and σ is the surface tension. Forthis equation to be used for the predictions ofexperimental data of ethylene glycol/water mixtureswith concentrations other than 50/50, we furthermodified the above correlation with a concentrationcorrection factor, which reduces to 1 forconcentrations v = 0 and v = 0.5 . The newcorrelation can be expressed as:

[ ] [ ] 5.17.05.05.05.0* )()()()5.0(61 −−−+= vlvllBoWevvhh µµρρ

where h* is a characteristic heat transfer coefficientof 135 kW/m2K above all of the data.

Figure 5 shows the experimental data and thepredicted values obtained with the correlation for allfluids. The predictions of the equation are in goodagreement with the experimental data, and most ofthe predictions are within ±30% of the experimentaldata. It should be noted that the comparisons areonly for the data within the nucleation-dominant-boiling region. The success of the correlation inpredicting the heat transfer coefficients of all fluidsboiling in small channels is directly related to thetrend, as presented in Figure 2, that the heat transferdata are dependent on heat flux but not mass flux.The fact that the equation is also heat flux but notmass flux dependent is in accordant with theexperimental data.

ConclusionsExcellent progress has been made on both theexperiments and analysis during FY 2004:

(a) Two-phase frictional pressure gradients ofethylene glycol/water mixtures follow similar trendsas those of water. The results are in reasonableagreement with the predictions of Chisholm’scorrelation. A modification has been made to


82

0

2000

4000

6000

8000

10000

12000

0 2000 4000 6000 8000 10000 12000

40/60 ethylene glycol/water mixture50/50 ethylene glycol/water mixture60/40 ethylene glycol/water mixtureExperimental +30%Experimental -30%

Pre

dict

e he

at tr

ansf

er c

oeffi

cien

t (W

/m2 K

)

Experimental heat transfer coefficient (W/m2K)

+30%

-30%

Figure 5. Heat transfer coefficient comparisons (nucleation-dominant-boiling region).

Chisholm’s correlation, which reduces toChisholm’s correlation for concentrations v = 0 andv =1. This modified Chisholm’s correlationimproves the predictions of the pressure drop datafor ethylene glycol/water mixtures.

(b) The experiments show a very high heat transferrate with ethylene glycol/water mixtures, which ispromising for engine cooling applications. A generalcorrelation has been developed based on data forwater, ethylene glycol/water mixtures(concentrations 40/60, 50/50, and 60/40), andrefrigerants. This correlation predicts theexperimental data quite well, and most of thepredicted values are within ±30% of theexperimental data.

(c) It was found that the boiling heat transfer ofethylene glycol/water mixtures is mainly limited byflow instability rather than critical heat fluxes, which

are usually the limits for water boiling heat transfer.Tests show that stable long-term two-phase boilingflow is possible for ethylene glycol/water mixturesas long as the mass quality is less than a certaincritical value (approximately <0.2). The heat transferrate at this mass quality is significantly higher thanthat of conventional, single-phase, forced-convectiveheat transfer.

(d) In the applications of engine cooling, bothhorizontal and vertical flows exist. Therefore, it isnecessary to investigate the impact of vertical flowsvs. horizontal flows on two-phase heat transfer.During FY 2005, the test facility will be modified torun vertical flows. Experiments are planned usingethylene glycol/water mixtures for boiling two-phase heat transfer of vertical flows. The tests willprovide essential information for the design ofnucleate-boiling cooling systems.


83

C. Evaporative Cooling

Principal Investigator: S.U.S. ChoiArgonne National LaboratoryArgonne, IL 60439(630) 252-6439, fax: (630) 252-5568, e-mail: [email protected]

Technology Development Manager: Sid Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4289, e-mail: [email protected]

Contractor: Argonne National LaboratoryContract No.: W-31-109-Eng-38

Objectives• Improve air-side heat transfer of the radiator.

• Reduce radiator size.

• Reduce radiator weight.

Approach• Conduct evaporative cooling tests on flat and cylindrical heaters.

• Characterize evaporative cooling with microporous surfaces.

• Develop improved techniques for making microporous and nanoporous coated surfaces.

• Conduct evaporative cooling tests on prototype radiators.

Accomplishments• Investigated the effects of coating particle sizes and coating thicknesses on evaporative cooling performance.

• Initiated the development of an improved coating technique called thermally conductive microporous coating(TCMC).

Future Direction• Develop improved techniques for making microporous and nanoporous coated surfaces, including the TCMC

technique.

• Conduct evaporative cooling tests on prototype radiators.

IntroductionEvaporative cooling is an effective and economicalmethod to cool heated objects requiringcomparatively high heat flux dissipation fornumerous industrial applications. Since evaporativecooling involves a liquid-vapor phase change, theincrease of heat transfer is significant compared to

single-phase forced convection of air. The heat ofvaporization augments heat transfer when the tinywater droplets encounter and evaporate from theheated surface. In previous research, we successfullydesigned, fabricated, and operated an experimentalfacility to study evaporative cooling with plain andcoated surfaces for flat and cylindrical heaters. Oneof our key findings is that the combination of


84

evaporative cooling and microporous coating on aflat heater enhances the heat transfer coefficient byup to 400% relative to the reference case (dry aircooling with uncoated, plain surface). Furthermore,the microporous coating extended the dryout heatflux significantly (~21 kW/m2) over the plainsurface (15 kW/m2). These findings clearly showthat the performance on the air side of the radiatorcould be improved if evaporative heat transfer couldbe used. Therefore, one of the general strategies forimproving the thermal management performance inheavy vehicles is to make better use of evaporativecooling.

In FY 2004, we investigated the effects of waterflow rates, coating particle sizes, and coatingthicknesses on evaporative cooling performance. Inaddition, we initiated the development of animproved coating technique called thermallyconductive microporous coating (TCMC).

Results and DiscussionEffect of Microporous Coating Particle SizesThree ranges of particle size, 8-12 µm, 30-60 µm,and 100-300 µm, were employed in formingmicroporous structures on the flat heater. Thecoating thicknesses were measured to be ~50 µm for8-12 µm particles, ~150 µm for 30-60 µm particles,and ~500 µm for 100-300 µm particles. The resultsshown in Figure 1 reveal that the tested particle sizesproduced similar heat transfer curves. The capillarypumping phenomenon should be stronger as thecoating particle size decreases. On the other hand,fluid flow resistance for the soaked water within themicroporous passageways increases as the particlesize decreases. Also, the wall temperature increasesdue to the additional thermal resistance ofconduction as the layer thickness increases for agiven heat flux. Therefore, the possible enhancementfrom less fluid flow resistance due to larger cavity

Figure 1. Effect of microporous coating particle sizes on heat transferperformance of flat heater (water flow rate: 1.75 mL/min).


85

size is believed to be counterbalanced bydegradation due to additional thermal resistancefrom conduction and less capillary pumping power.

Effect of Microporous Coating ThicknessesThe effect of microporous coating thickness wasinvestigated using the smallest particle size(8-12 µm). Figure 2 shows the heat transferperformance with coating thicknesses of 50, 100,150, and 200 µm on a flat heater during tests withspray cooling and a water flow rate of 1.75 mL/min.Microporous coatings with 150 and 200 µmthicknesses showed smaller enhancement in sprayheat transfer than coatings with 50 and 100 µmthicknesses. The 100 µm coating thickness attainedthe highest heat transfer coefficient among the testedthicknesses. This thickness produced up to ~50%increase in evaporative heat transfer coefficient overthe 200 µm thickness. The thicker microporouscoating retains more water droplets as the coatingthickness increases, due to the larger number ofmicroporous cavities. However, the evaporative heat

transfer coefficient degrades extensively due to theadditional thermal resistance to conduction as thethickness exceeds 100 µm.

ABM Coating TechniqueThe University of Texas at Arlington (UTA) hasbeen collaborating with Argonne in the conduct ofbasic research on the effects of microporouscoatings and other techniques on evaporativecooling. The ABM coating technique, named fromthe initial letters of the three components of themicroporous coating: aluminum/brushableceramic/methyl-ethyl-ketone, is one of themicroporous coating methods previously developedto make a porous structure on heat transfer surfaces.After the carrier (M.E.K.) evaporates, the resultingcoated layer consists of microporous structures withaluminum particles (8 to 12 µm) and a binder(Devcon® brushable ceramic) having a thickness of~50 and 100 µm, which was shown to be anoptimum thickness for FC-72 and water,respectively.

Figure 2. Effect of microporous coating thickness on heat transferperformance of flat heater (water flow rate: 1.75 mL/min).


86

The previous study on evaporative cooling wasachieved by the ABM coating technique. The ABMcoating enhances the heat transfer coefficient ofevaporative cooling mainly due to capillary pumpingaction within the microporous structures. However,the epoxy component of the microporous structurehas a poor thermal conductivity, and thus potentiallyincreases the conduction thermal resistance throughthe layer, especially when the coating becomesthick. Also, use of epoxy appears to prohibit theporosity variation within the microporous structureseven if the aluminum particle size was changed. Toovercome these inferior thermal characteristics inevaporative cooling, a new innovative conductivemicroporous coating method needs to be developed.

TCMC Coating TechniqueWe have therefore developed a thermally conductivemicroporous coating (TCMC) technique thatconsists of metal particles with various micron-levelsizes and thermally conducting binding materialsthat bond the metal particles to produce numerousmicroporous cavities on a target surface. The newTCMC technique is similar in concept to the ABMcoating (many small metal powders jointed by a thinbonding layer around the particles). However, athermally conductive layer will be used for TCMCtechnique, instead of the epoxy layer used for theABM technique to form the microporous coating.Different particle materials are being tested for thiscurrent technique, considering corrosioncharacteristics and compatibility with various fluids.The microporous coating methods, including theABM technique, are inexpensive and easy processescompared to sintering/machining methods.However, the coating methods are inferior tosintering/machining methods due to the low thermalconductivity of binders (epoxy). The mainadvantage of this improved coating technique is thecombination of an inexpensive and easy coatingprocess and the low thermal resistance of thermallyconductive binding materials. Another advantage isthat heat transfer is insensitive to coating thicknessdue to the high thermal conductivity of the bindingstructures. In addition, the coating technique isapplicable to various working liquids, highly wettingto poorly wetting, simply by changing the size ofmetal particles, since the surface tension of liquidsdetermines the sizes of porous cavities that optimizeheat transfer performance.

Figure 3 shows a SEM image of a TCMC-coatedsurface with aluminum particles (8 to 12 µm). Asshown in the figure, the coating produced numerousmicroporous cavities. Figure 4 shows the boilingperformance comparison between TCMC and ABMwith 30-50 µm particles. It is clearly shown thatTCMC generates higher boiling heat transfer thanABM. This boiling enhancement could be possiblyachieved due to the thermally conducting binderoption, which generates very low thermal resistanceat high heat flux compared to nonconductingbinders.

The TCMC technique using a thermally conductivebinder has significant advantages compared to theABM method:

• ~100 times higher thermal conductivity of thebonding layer than the epoxy binder.

• Thicker coating with higher porosity is possibleto strengthen capillary pumping through thecoating without degrading thermal performance.

Therefore, the TCMC technique is believed toperform better than the microporous plate inevaporative cooling heat transfer.

Future DirectionIn the future, we will develop an improved TCMCtechnique and conduct evaporative cooling tests onprototype radiators. The work will include thefollowing major tasks and subtasks:

1. Develop improved techniques for makingmicroporous and nanoporous coated surfaces,including the TCMC technique.1.1 Select a powder material and binder

combination.1.2 Coat the surface of one flat heater with

TCMC coating.1.3 Perform dry and evaporative cooling

experiments.1.4 Compare evaporative cooling data for the

plain, ABM-coated, and TCMC-coatedsurfaces.

1.5 Optimize evaporative cooling with TCMCsurfaces by investigating the effect of spraywater flow rate, particle size, and coatingthickness.


87

Figure 3. SEM image of thermally conductive microporous coatedsurface with aluminum particles 8–12 µm in diameter.

Figure 4. Comparison of heat transfer performance between ABMand TCMC with 30–50-µm aluminum particles.


88

1.6 Characterize the structure of the selectedTCMC surfaces using SEM.

2. Conduct evaporative cooling tests on prototyperadiators.2.1 Evaluate a number of methods to

incorporate evaporative cooling intoprototype radiators.

2.2 Conduct experimental study of evaporativecooling with a TCMC coating on aprototype radiator.

ConclusionsWe have characterized evaporative cooling withplain and microporous surfaces with a focus on theeffects of particle sizes and thicknesses of coatingson evaporative cooling. The size of coating particleshas a negligible effect on heat transfer performance.Therefore, a nanoporous surface is recommended forfurther study of the particle size effect. The 100-µmcoating thickness gave the maximum heat transferperformance.

To overcome the inferior thermal characteristics ofthe ABM coating in evaporative cooling, a newinnovative conductive microporous coating methodneeds to be developed. We have initiated the

development of a thermally conductive microporouscoating (TCMC) technique that consists of metalparticles and thermally conducting bindingmaterials. The microporous coating methods,including the ABM technique, are inexpensive andeasy processes compared to sintering/machiningmethods. However, the coating methods are inferiorto sintering/machining methods due to the lowthermal conductivity of binders (epoxy). The mainadvantage of this improved coating techniqueinvolves combining an inexpensive and easy coatingprocess with the low thermal resistance of thermallyconductive binding materials. Another advantage isthat heat transfer is insensitive to coating thickness,due to the high thermal conductivity of bindingstructures. Therefore, the TCMC technique isbelieved to perform better than the microporousplate in evaporative cooling heat transfer.

FY 2004 PublicationJ.H. Kim, S.M. You, and S.U.S. Choi, “EvaporativeSpray Cooling of Plain and Microporous CoatedSurfaces,” International Journal of Heat and MassTransfer, Vol. 47, pp. 3307–3315, 2004.


89

D. Nanofluids for Thermal Management Applications

Principal Investigator: S.U.S. ChoiArgonne National LaboratoryArgonne, IL 60439(630) 252-6439, fax: (630) 252-5568, e-mail: [email protected]

Technology Development Manager: Sid Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4289, e-mail: [email protected]


Objectives• Exploit the unique properties of nanoparticles to develop heat transfer fluids with ultra-high thermal

conductivity.

• Characterize the thermal properties and heat transfer performance of nanofluids.

• Develop nanofluid technology for vehicle thermal management.

Approach• Develop simplified models of a nanofluid.

• Measure the thermal conductivity of nanofluids.

• Measure the heat transfer coefficient of nanofluids.

Accomplishments• Developed a nanofluid model and found that the dynamic interactions between nanoparticles and liquid

molecules is a key mechanism of the temperature- and size-dependent thermal conductivity. Our paper waspublished May 24, 2004, as the cover story of the journal Applied Physics Letters.

• Predicted, by using our nanofluid model, that the thermal conductivity of nanofluids increases with decreasingparticle size.

• Published, in collaboration with researchers in several universities, the first and second review articles onnanofluids.

Future Direction• Produce and characterize nanofluids containing nanoparticles less than 10 nm.

• Conduct nanofluid flow and heat transfer experiments in laminar and turbulent flow.

• Modify the one-step nanofluid production system for a larger quantity of nanofluids.

• Characterize metallic and oxide nanofluids in pool boiling for two-phase engine cooling applications.

• Refine models of nanostructure-enhanced and nanoparticle-mobility-enhanced thermal conductivity ofnanofluids.


90

IntroductionThe heat rejection requirements are continuallyincreasing as a result of trends toward more poweroutput and stringent emission requirements forengines. Therefore, cooling is one of the toptechnical challenges facing the transportationindustry. The conventional method to increase heatrejection rates is to use extended surfaces (such asfins). However, current designs have alreadystretched this technology to its limits. Furthermore,conventional heat transfer fluids used in today’sthermal management of vehicles, such as lubricantsand engine coolants, are inherently poor heattransfer fluids. Therefore, a strong need exists fornew and innovative concepts to achieve ultra-high-performance cooling in thermal managementsystems for vehicles.

Modern fabrication technology provides greatopportunities for actively processing materials at thenanoscale. The thermal, mechanical, optical,magnetic, and electrical properties of nanophasematerials are superior to those of conventionalmaterials with coarse grain structures. Consequently,the research and development on nanophasematerials has drawn considerable attention frommaterial scientists and engineers alike. Taking adifferent tack from extended surface approaches andexploiting the unique properties of nanoparticles,Argonne National Laboratory (ANL) has pioneereda new kind of ultra-high-thermal-conductivity fluid,called nanofluids, by uniformly and stablysuspending a very small quantity (preferably <1%by volume) of nanoparticles in conventionalcoolants [1].

The potential benefits of using nanofluids include:

• Improve heat transfer of engine coolants andoils,

• Reduce heat exchanger size and weight,• Reduce heat transfer fluid inventory,• Reduce vehicle emissions,• Improve wear resistance,• Reduce friction coefficient, and• Reduce pumping energy in existing system.

The objectives of the project are to exploit theunique properties of nanoparticles to develop heattransfer fluids with high thermal conductivity, to

characterize the thermal conductivity and heattransfer behavior of nanofluids, and to developnanofluid technology for increasing the thermaltransport of engine coolants and lubricants.

ApproachOur approach to overcoming the poor heat transferrates of conventional heat transfer fluids is tosignificantly increase the thermal conductivity of theliquids. An attractive method in that regard is the useof a suspension of solid nanoparticles in a liquid.Solids with thermal conductivities orders ofmagnitudes higher than those of the liquids arechosen. For example, the thermal conductivity ofcopper at room temperature is about 3000 timesgreater than that of engine oil. In fact, numeroustheoretical and experimental studies of the effectivethermal conductivity of dispersions that containsolid particles have been conducted since Maxwell’stheoretical work was published about 100 years ago[2]. However, studies on thermal conductivity ofsuspensions have been confined to millimeter- ormicrometer-sized particles. The major problem withthe use of such large particles is the rapid settling ofthese particles in fluids. Another is the abrasivenature of the suspension. In contrast, use ofnanoparticles has the potential to produce theincreased thermal conductivity sought in heattransfer fluids without the problems of settling andabrasion. Maxwell’s concept of enhancing thethermal conductivity of fluids by dispersing solidparticles in them is old, but what is new andinnovative with the concept of nanofluids is the ideaof using the nanoparticles that have becomeavailable to us only recently.

The ANL nanofluid team has developed twotechniques to make nanofluids: the single-step directevaporation method, which simultaneously makesand disperses the nanoparticles directly into the basefluid, and the two-step method, which first makesnanoparticles and then disperses them into the basefluid. Although the two-step technique works wellfor oxide nanoparticles, it is not as effective formetal nanoparticles, such as copper. For nanofluidscontaining high-conductivity metals, the single-stepdirect evaporation technique is preferable to gas-condensation processing. ANL has already producedoxide and carbon nanotube nanofluids by the two-step technique and metal nanofluids by the single-


91

step technique. In particular, it was demonstratedthat stable suspensions can be achieved bymaintaining the particle size below a threshold level.

The thermal conductivity of nanofluids with lowparticle concentrations (1–5% by volume) wasstudied experimentally. The experimental resultsshow that these nanofluids have substantially higherthermal conductivities than the same liquids withoutnanoparticles. For example, a 20% improvement inthe thermal conductivity of ethylene glycol was seenwhen 4 vol.% copper oxide was dispersed in thisfluid. However, recent measurements show thatmetallic nanoparticles or carbon nanotubes (CNTs)can increase the thermal conductivity by asignificant amount over the oxide particles. Forexample, the dispersion of a tiny amount (<1 vol.%)of copper nanoparticles or carbon nanotubesdispersed in ethylene glycol or oil improves theirthermal conductivities by 40 and 150%, respectively[3–4].

Inspired by our discoveries, other groups across theworld have begun nanofluid research and providedfurther breakthroughs. Among them, Das et al.explored the temperature dependence of the thermalconductivity of nanofluids containing Al2O3 or CuOnanoparticles [5]. Their discovery of a two- to four-fold increase in thermal conductivity enhancementfor nanofluids over a small temperature range (21–51°C) implies that nanofluids could be smart fluids,sensing their thermal environment. This new featuremakes nanofluids even more attractive as coolantsfor devices with high energy density. You et al. [6]measured the critical heat flux (CHF) in a boilingpool of nanofluids containing an extremely smallamount of alumina nanoparticles. They discovered adramatic increase (up to about 200%) in CHF whenthe nanofluid is used as a cooling liquid instead ofpure water.

Since the pioneering work of Choi [1], the four keyfeatures of nanofluids have been found to be thermalconductivities far above those of traditionalsolid/liquid suspensions [3–4], a nonlinearrelationship between thermal conductivity andconcentration [4], strongly temperature-dependentthermal conductivity [5], and a significant increasein critical heat flux (CHF) [6]. These key featuresmake nanofluids strong candidates for the next

generation of coolants for improving the design andperformance of thermal management systems.

Results and DiscussionThe discoveries of ultra-high thermal conductivityand CHF clearly show the fundamental limits ofconventional models for solid-liquid suspensionsinvolving nanoparticles. At present, mechanisms forenhanced CHF are not understood. Severalmechanisms that could be responsible for enhancedthermal conductivity in nanofluids have beenproposed, including ballistic conduction innanoparticles, as well as the thermal conductivitybeing dependent upon the nanoparticle structure andmobility.

We have developed, for the first time, a simplifiednanofluid model and found that the dynamicinteractions between nanoparticles and liquidmolecules is a key mechanism of the temperature-and size-dependent thermal conductivity [7]. Ourpaper was published as the cover story in the May24, 2004, issue of the journal Applied PhysicsLetters (see Figure 1). Our nanofluid model predictsthat the thermal conductivity of nanofluids increaseswith decreasing particle size (see Figure 2). This is amajor milestone in the development of nanofluidtechnology. For this technology to become a reality,new techniques are needed for manufacturingnanoparticles smaller than 10 nm. Just several

Figure 1. Modes of energy transport in nanofluids. Thefirst mode is collision between base fluid molecules, thesecond mode is the thermal diffusion in nanoparticlessuspended in fluids, the third mode is collision betweennanoparticles (not shown), and the fourth mode is thermalinteractions of dynamic or dancing nanoparticles withbase fluid molecules.


92

Figure 2. Nanoparticle size-dependent thermalconductivity of nanofluids at a fixed concentration of3 vol.% and room temperature, normalized to the thermalconductivity of the base fluid. Although experimentaldata are very limited, model predictions (solid curve a)agree with experimental data (solid squares) for water-based nanofluids containing 13-nm Al2O3 nanoparticlesand 38.4-nm Al2O3 nanoparticles. The same is true ofmodel predictions (solid curve b) and the experimentaldata point (solid circle) for ethylene glycol-basednanofluids containing 24.4-nm CuO nanoparticles.Surprisingly, the conductivity of the base fluid increasesby more than a factor of two with <10-nm Cu particles inethylene glycol (curve c). In striking contrast, the insetshows that the Maxwell model, together with thenanoparticle thermal conductivity calculated from Chen’sequation, predicts decreasing conductivity of nanofluidswith decreasing particle size for water-based nanofluidscontaining Al2O3 nanoparticles.

months later another dynamic model of a nanofluidwas published in the journal Physical Review Letters[8] and confirmed the key role of Brownian motionof nanoparticles in nanofluids. Still, a morecomprehensive theory is needed to explain thebehavior of nanofluids.

The Nanofluid Heat Transfer Test Facility has beenconstructed to experimentally determine heattransfer rates to flowing nanofluids under a varietyof flow conditions and using a variety of nanofluids.A major constraint on the test facility design is theneed to keep the fluid inventory to a bare minimum(~240 mL). Consequently, we applied methodsdeveloped over many years in small-scale heat-transfer testing to the nanofluid situation. For over adecade, we have been involved with experimentalheat transfer to single-phase, boiling, and

condensing of one- and two-component fluids. Testchannels have been on the order of 3 mm insidediameter for the most part, with some channels assmall as 0.5 mm. To meet the low-fluid-inventorydesign constraint for the Nanofluid Heat TransferTest Facility, a test section was chosen with aninside diameter of 2 mm. All components werecarefully sized accordingly. Instrumentation anddata acquisition techniques developed for previoustests in this small channel size range were adapted tothe Nanofluid Heat Transfer Test Facility. Aschematic of the facility is shown in Figure 3. It is aclosed-loop system that includes a variable-speeddrive pump and meter, a preheater, test section, andcooler. The electrically heated test channel is astainless-steel tube with a length of 508 mm andinside diameter of 2 mm. Inlet pressure (P) anddifferential pressure across the channel (dP) aremeasured by means of a piezoresistive-typetransducer and strain-gauge-type transducer,respectively. Stream thermocouples measure thefluid temperature at the inlet and outlet of the testchannel, and the surface thermocouples placed onthe test channel wall measure wall temperature atfive locations along the channel length. A dataacquisition system (DAS) reads and records allsensor-output voltages.

Figure 3. Schematic diagram of nanofluid heat transfertest facility.

ANL has inaugurated a work-for-others project withNuvonyx. The objective is to develop several water-based nanofluids so that Nuvonyx can assess thefeasibility of applying nanofluid technology tomicrochannel cooling. Since nanofluids inmicrochannels are expected to dramatically enhancethe convection heat transfer and provide a basis for


93

state-of-the-art applications of nanofluid technologyto advanced vehicle thermal management systems,this project would directly benefit nanofluidtechnology development funded by DOE.

FY 2004 Publications1. Jang, S.P., and Choi, S.U.S., “Role of Brownian

Motion in the Enhanced Thermal Conductivityof Nanofluids,” Applied Physics Letters, Vol.84, No. 21, pp. 4316–4318 (May 2004). Also inVirtual Journal of Nanoscale Science andTechnology, May 24, 2004.

2. Choi, S.U.S.; Zhang, Z.G.; and Keblinski, P.,“Nanofluids,” Encyclopedia of Nanoscience andNanotechnology, Vol. 6, pp. 757–773,H.S. Nalwa, ed., American Scientific Publishers,Stevenson Ranch, CA, 2004.

3. Eastman, J.A.; Phillpot, S.R.; Choi, S.U.S.; andKeblinski, P., “Thermal Transport inNanofluids,” Annual Review of MaterialsResearch, Vol. 34, pp. 219–246, 2004.

4. Choi, S.U.S.; Yu, W.; Hull, J.R.; Zhang, Z.G.;and Lockwood, F.E., “Nanofluids for VehicleThermal Management,” SAE Transactions J.Passenger Cars: Mech. Systems, Vol. 111,No. 1021, pp. 38–43, 2003.

5. Xue, L.; Keblinski, P.; Phillpot, S.R.;Choi, S.U.S.; and Eastman, J.A., “Effect ofLiquid Layering at the Liquid-Solid Interface onThermal Transport,” submitted to Int. Journal ofHeat and Mass Transfer, 2004.

6. Yu, W., and Choi, S.U.S., “The Role ofInterfacial Layers in the Enhanced ThermalConductivity of Nanofluids: A RenovatedHamilton-Crosser Model,” submitted toJ. Nanoparticle Research.

Future DirectionFuture work will focus on the following five majortasks:

1. Produce and characterize nanofluidscontaining nanoparticles less than 10 nm.Characterization of the thermal conductivity ofnanofluids has led to the discovery that it increaseswith decreasing nanoparticle size. In contrast, solid-solid composites have the reverse size effect: thethermal conductivity of solid-solid nanocompositesdecreases with decreasing particle size. Since ourdynamic model shows that particle size is of primary

importance in the development of nanofluidtechnology, we will produce and characterizenanofluids containing nanoparticles less than 10 nm.

2. Conduct nanofluid flow and heat transferexperiments in laminar and turbulent flow.A major focus in nanofluids has been on theirthermal conductivity. Nanofluids will be producedand characterized to generate a database on theirthermal conductivity. However, a new focus will beon the flow and heat transfer of nanofluids. Webelieve that it is critical to characterize the flow andheat transfer behavior of nanofluids in tubes in orderto explain and exploit nanoparticle size-, shape-, andconcentration-dependent pressure drop and heattransfer coefficient of nanofluids in tubes. Forexample, theories of flow in tubes or porousmaterials have been developed at the macroscopiclevel. However, we believe that it is vital to connectthe structure and mobility of nanoparticles at thenano- or micro-scale to the properties of nanofluidsmeasured at the macroscopic level. Therefore, wewill explore the flow characteristic and measure theheat transfer coefficient of nanofluids in tubes. Wewill then relate enhanced thermal conductivity andflow characteristic to enhanced heat transfercoefficient of nanofluids. The new Nanofluid HeatTransfer Test Facility will be carefully scrutinizedfor instrument accuracy, data acquisition process,and test procedure. Control tests will be performedwith water as the test fluid. Subsequently, nanofluidflow and heat transfer experiments will beconducted. The flow and heat transfer data fornanofluids will be used to determine whether theflow and heat transfer correlations developed forhomogeneous fluids could be applied to nanoparticledispersions with aspect ratios different from one.

3. Modify the one-step nanofluid productionsystem for a larger quantity of nanofluids.Nanofluids hold significant potential forrevolutionizing industries that are dependent on theperformance of heat transfer fluids. However, thereis one big barrier to commercialization ofnanofluids: production scale-up methods to producenanofluids in volume and at low cost. Finding a wayto produce small (1–10 nm) nanoparticles cheaplyand disperse them without agglomeration is the keyhurdle to commercialization. As a first step towardproduction scale-up, we will modify ANL’s one-stepprocess to produce more nanofluids.


94

4. Characterize metallic and oxide nanofluids inpool boiling for two-phase engine coolingapplications.Most studies carried out to date are limited to thethermal characterization of nanofluids withoutevaporation or condensation. However,nanoparticles in nanofluids can play a vital role intwo-phase heat transfer systems. Therefore, there isa great need to characterize nanofluids in boiling andcondensation heat transfer. The high critical heatflux of nanofluids is important, particularly for two-phase engine cooling. The nanofluid tested by Youet al. [6] contains alumina (Al2O3) nano-particlesdispersed in distilled and deionized water. We willdesign and fabricate a pool boiling test facility andcharacterize metallic and oxide nanofluids in poolboiling.

5. Refine models of nanostructure-enhanced andnanoparticle-mobility-enhanced thermalconductivity of nanofluids.We have developed simple models of nanostructure-enhanced [9] and nanoparticle-mobility-enhancedthermal conductivity of nanofluids [7]. However,much is still unknown about the mechanisms of theanomalous thermal behavior of nanofluids.Therefore, we need to understand the fundamentalsof energy transport in nanofluids. As we discovernew energy transport mechanisms that are missingin existing theories, we will refine the simple modelsor develop a new model of energy transport innanofluids integrating new mechanisms. A betterunderstanding of the mechanisms behind thethermal-conductivity enhancement will likely lead torecommendations for nanofluid design andengineering for transportation applications.

ConclusionsNanofluids have four key features: (1) thermalconductivities far above those of traditionalsolid/liquid suspensions, (2) nonlinear relationshipbetween thermal conductivity and concentration,(3) strongly temperature-dependent thermalconductivity, and (4) significant increase in criticalheat flux. These key features make nanofluids strongcandidates for the next generation of coolants forimproving the design and performance of thermalmanagement systems.

To facilitate the development of nanofluidtechnology, we will focus in FY2005 on thermalconductivity, flow, and heat transfer experimentsusing present and developmental nanofluids. Theimproved thermal properties of nanofluids wouldallow the use of significantly smaller and lighterheat exchangers for vehicles and other performance-driven systems.

References1. Choi, U.S., “Enhancing Thermal Conductivity

of Fluids with Nanoparticles,” Developmentsand Applications of Non-Newtonian Flows,D.A. Siginer and H.P. Wang, eds., TheAmerican Society of Mechanical Engineers,New York, FED–Vol. 231/MD-Vol. 66, pp. 99–105 (Nov. 1995).

2. Maxwell, J.C., 1881, A Treatise on Electricityand Magnetism, 2nd ed., 1, 435, ClarendonPress.

3. Eastman, J.A.; Choi, S.U.S.; Li, S.; Yu, W.; andThompson, L.J., “Anomalously IncreasedEffective Thermal Conductivities of EthyleneGlycol-Based Nanofluids Containing CopperNanoparticles,” Appl. Phys. Lett. 78(6), pp. 718–720, 2001.

4. Choi, S.U.S.; Zhang, Z.G.; Yu, W.; Lockwood,F.E.; and Grulke, E.A., “Anomalous ThermalConductivity Enhancement in NanotubeSuspensions,” Appl. Phys. Lett. 79(14), pp.2252–2254, 2001.

5. Das, S.K.; Putra, N.; Thiesen, P.; andRoetzel, W., “Temperature Dependence ofThermal Conductivity Enhancement forNanofluids,” ASME J. Heat Trans. 125,pp. 567–574, 2003.

6. You, S.M.; Kim, J.H.; and Kim, K.M., “Effectof Nanoparticles on Critical Heat Flux of Waterin Pool Boiling of Heat Transfer,” Appl. Phys.Lett. 83, pp. 3374–3376, 2003.

7. Jang, S.P., and Choi, S.U.S., “Role of BrownianMotion in the Enhanced Thermal Conductivityof Nanofluids,” Appl. Phys. Lett. 84, pp. 4316–4318, 2004.

8. Kumar, D.H.; Patel, H.E.; Rajeev, K.V.R.;Sundararajan, T.; Pradeep, T.; and Das, S.K.,“Model for Heat Conduction in Nanofluids,”Phys. Rev. Lett. 93, pp. 144301-1-4, 2004.


95

9. Yu, W., and Choi, S.U.S., “The Role ofInterfacial Layers in the Enhanced ThermalConductivity of Nanofluids: A RenovatedHamilton-Crosser Model,” submitted toJ. Nanoparticle Research.


96

E. Erosion of Materials in Nanofluids

Principal Investigator: J.L. Routbort (co-worker: D. Singh)Argonne National Laboratory9700 S. Cass Avenue, Argonne, IL 60439-4838(630) 252-5065, e-mail: [email protected]

Technology Development Manager: Sid Diamond(202) 586-8032, e-mail: [email protected]


Objectives• Determine if the use of fluids containing a variety of nanoparticles and nanotubes result in erosive damage to

radiator materials.

• Develop models to predict the erosive damage.

Approach• Develop an experimental apparatus to measure erosive loss.

• Conduct experiments to study erosive damage of fluids on typical radiator materials.

• Develop methods to analyze the results.

Accomplishments• Built apparatus to study liquid erosion.

• Completed calibration of sample impact area as a function of fluid velocity.

• Gathered baseline data with ethylene glycol and water solutions, and this effort is ongoing.

• Conducted preliminary experiments with nanofluids.

Future Direction• After accumulation of a baseline data by using ethylene glycol and water, fluids containing a variety of

nanoparticles and nanotubes will be tested on typical radiator materials, varying both the angle and velocity ofimpact, volume percent of nanoparticles, and temperature.

IntroductionMany industrial technologies face the challenge ofthermal management. With ever-increasing thermalloads due to trends toward greater power output forengines and exhaust gas recirculation for dieselengines, cooling is a crucial issue in transportation.The conventional approach for increasing coolingrates is the use of extended surfaces (such as finsand microchannels). Reducing radiator size willreduce the frontal area and hence the aerodynamic

drag. However, current designs have alreadystretched this approach to its limits. Therefore, anurgent need exists for new and innovative conceptsto achieve ultra-high-performance cooling.Nanofluids seem to show enormous potential as acoolant for radiators. Choi et al. have shown thatfluids containing 1 vol.% Cu nanoparticles increasethermal conductivity by 40% [1], while 1 vol.%carbon nanotubes increase thermal conductivity by250% [2].


97

To utilize the enhanced thermal conductivity, it mustbe shown that liquid erosion of typical radiatormaterials will be tolerable. Hence, the FreedomCarand Vehicle Technologies Program has funded aprogram on liquid erosion of nanofluids.

Results and DiscussionAs reported earlier, an apparatus was designed andconstructed to allow the erosion rates of fluidscontaining various nanoparticles and nanotubes to bedetermined from ≈ 1 m/s, velocities that are used inheavy-vehicle radiators, to about 15 m/s. Angle ofimpacts can be varied from 0 to 90°C, andtemperatures up to ≈ 95°C can be obtained. Aphotograph of the apparatus is shown in Figure 1.Figure 2 is a view of the specimen chamber with thecover removed to show the specimen and the nozzle.The experimentally determined variables are(1) pump RPM, (2) temperature, (3) pressure, and(4) weight loss. The velocity is determined bymeasuring the amount of fluid for a given time at afixed RPM by using a given nozzle diameter. Such acalibration is shown in Figure 3.

During experimentation over the past year, it wasdetermined that velocity at a fixed RPM during thecourse of an experiment is dependent on severalfactors, such as wear of pump gears and clogging ofthe mesh filter placed in the fluid. To ensure thatvelocity-RPM behavior remains consistent,modifications were made to the erosion set-up.These modifications included additional pressuregages placed at the pump exit and before the nozzle

Figure 1. Photograph of the liquid-erosion apparatus withvarious components labeled.

Figure 2. Close-up view of experimental chamber withthe cover removed to show the specimen and the nozzle.

0

1

2

3

4

5

0 200 400 600 800 1000 1200 1400

Vel

oci

ty (

m/s

)

RPM

Nozzle Diameter - 2.5 mm

Figure 3. Plot of velocity, determined by measuring thevolume of fluid for a fixed time, as a function of pumpRPM.

(Figure 4). Stability of the pressure readings at theselocations ensured the constant fluid velocity at afixed pump RPM per the calibration profilesdeveloped.

Figure 5 shows the initial result of weight loss of3003 aluminum sample as a function of time atvarious velocities. These tests were conducted byusing equi-volume solution of ethylene glycol andwater. As expected, the weight loss increases withincreasing velocity. However, at 3 m/s, noappreciable weight loss was measured. Moreover, atextremely high velocities (> 11 m/s) there is a rapidinitial weight loss followed by a decreased rate ofweight loss, which indicates saturation in the erosionbehavior.


98

Figure 4. Newly installed pressure gages to ensureappropriate pressures and velocities in the erosion system.

0

1

2

3

4

5

6

7

8

0 100 200 300 400 500 600 700

Wei

gh

t L

oss

(m

g)

Time (h)

V = 3 m/s

V = 4.5 m/s

V = 6 m/s

V = 9 m/s

V = 11.5 m/s

Figure 5. Plot of measured weight loss of 3003 Al vs.time, obtained at ambient temperatures for normalincidence by using ethylene glycol/water solution.

Increasing fluid velocity increases area over whichfluid impacts the sample surface. Thus, to comparethe erosion behavior at various velocities, it isnecessary to normalize slope of the weight-loss timeplot with area of erosion at specific velocities. Thisnormalized value can be defined as erosion rate(ER). To identify the area of impact, an aluminumsample was spray painted and exposed to the fluid atthe specific velocity.

Figure 6 shows a typical damage pattern that isgenerated and used for determining the fluid impactarea at a fixed velocity. Similar patterns wereobtained at other velocities at which erosionbehavior has been studied.

Figure 6. Typical damage zone obtained on a spray-painted aluminum sample at a velocity of 10 m/s. Noticethe stagnation zone at the center.

Table 1 provides the erosion rates for Al 3003 underdifferent velocities (3–9 m/s). Also, expectedrecession rate (rate at which material recedes,ER/aluminum density) is listed.

Preliminary experiments using nanofluids containingcopper oxide nanoparticles have been initiated.Nanofluid containing 1 vol.% CuO in ethyleneglycol/water solution was provided by S. Choi/ANL.For the preliminary study, the fluid was furtherdiluted to 0.25 vol.% nano-particles by volume.Figure 7 shows the weight loss of Al 3003 withtime. These results suggest that the weight loss ratewith nanofluids is somewhat similar to that of plainethylene glycol/water solutions, as shown inFigure 5, at similar velocities. At longer times, theweight loss appears to be saturated (data forV=4 m/s). However, at 2 m/s, weight loss ismeasurable with nanofluids as opposed to nodiscernible loss measured with plain fluid atV=3 m/s. It needs to be pointed out that there isadhesion of agglomerated CuO nanoparticles on thesample surface (as shown in Figure 8) that couldhave affected the weight loss measurements. Forfuture experiments, nanofluids with appropriatedispersants will be used.

Table 1. Erosion and recession rates of Al 3003 as afunction of velocity obtained at ambient temperatures atnormal incidence.

V (m/s)Erosion Rate

(g/h-m2)Recession Rate

(µm/h)3 0 04.5 0.005 0.0026 0.031 0.0129 1 0.370


99

0

0.5

1

1.5

2

2.5

3

0 100 200 300 400 500 600 700

Wei

gh

t L

oss

(m

g)

Time (h)

V = 2 m/s

V = 4 m/s

Figure 7. Plot of measured weight loss of 3003 Al vs.time, obtained at ambient temperatures for normalincidence by using 0.25 vol.% CuO nano-fluid.

Figure 8. Al 3003 surface showing agglomerated CuOnanoparticles adhering to the surface.

It is well known that Al 3003 is susceptible tocorrosion in the presence of ethylene glycol/watersolutions [3]. In the presence of small amounts ofimpurities, Al metal can corrode in ethylene glycolsolutions by localized pitting. Evidence of extensivepitting corrosion (Figure 9), on the unimpactedsurface, was found in one of the samples thatshowed weight loss with time. It is believed thatthere is a galvanic couple formed between thesample holder (steel) and the sample. In the presenceof impurities in the fluid, test temperature corrosionof Al sample occurs. However, corrosion is expectedto be independent of the hydrodynamic conditions[3]. Thus, the weight loss observed as a function ofvelocity is probably from a combination of erosiveand corrosive actions.

Figure 9. Evidence of corrosion pit with corrosionproducts in Al 3003 eroded with ethylene-glycol/watersolution.

Issues and Future DirectionAt present, the cause of the observed sample weightloss is unclear — it may be due to erosion,corrosion, or a combination of erosion/corrosionmechanisms. Understanding the mechanism ofweight loss is vital to developing predictive models.It is believed that the difficulty associated withrepeatability of experiments is related to theoperating mechanism(s). Distinction betweenerosion and corrosion behaviors will be made byusing a corrosion-resistant material or sample holderand/or controlling the temperature during theexperiment. Furthermore, the role of aluminumoxide layer in erosion behavior will be investigated.This will require some modifications to theapparatus. Baseline erosion data will be obtained forvarious impact angles and velocities. Finally, afamily of erosion rates as a function of angle andvelocity will be measured and correlated withdispersed nanoparticle loadings. Computational fluiddynamics will be used to predict the damage zone asa function of velocity.

ConclusionsAn apparatus to measure the erosion of radiatormaterials using nanofluids has been constructed. Anethylene glycol/water solution has been used tocollect baseline data on normal incidence impact.Experimentation using nanofluids has been initiated.


100

References1. J.A. Eastman, S.U.S. Choi, S. Li, W. Yu, and

L.J. Thompson, “Anomalously IncreasedEffective Thermal Conductivities of EthyleneGlycol-Based Nano-Fluids Containing CopperNano-Particles,” Applied Physics Letters, 78,718–720 (2001).

2. S.U.S. Choi, Z.G. Zhang, W. Yu,F.E. Lockwood, and E.A. Grulke, “AnomalousThermal Conductivity Enhancement in Nano-tube Suspensions,” Applied Physics Letters, 79,2252–2254 (2001).

3. D. Wong, L. Swette, and F.H. Cocks,“Aluminum Corrosion in Uninhibited EthyleneGlycol-water Solutions,” J. Electrochem. Soc.,126, 11–15 (1979).


101

F. Analysis of Ventilation Airflow and Underhood Temperatures in the EngineEnclosure of an Off-Road Machine

Principal Investigators: Tanju Sofu, Jimmy Chang, John HullArgonne National Laboratory9700 S. Cass Avenue, Argonne, Illinois 60439(630) 252-9673, fax: (630) 252-4500, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Jules Routbort(630) 252 5065, fax: (630) 252 4289, e-mail: [email protected]


Objective• Assess a novel simulation technique based on combined use of 3-D CFD and 1-D network flow models for

predicting underhood airflow and temperatures in a separated and sealed engine compartment of an off-roadmachine.

Approach• A prototypical test-rig that includes an engine and other installation hardware was built by the engineers at

Caterpillar, Inc., and the experiments were conducted with controlled ventilation air inlet locations and flowrates.

• The analytical effort focused on combined use of a 1-D network flow model with four loops to simulate theengine and the cooling system, and a 3-D CFD model for the ventilation air circulation in the test rig.

• The goal has been to provide an integrated analysis capability that needs only the flow rates and inlettemperatures as boundary conditions to study the effects of ventilation on heat rejection and componenttemperatures.

Accomplishments• Studied five different inlet configurations were studied for various ventilation flow rates and calculated the

corresponding flow field and temperature distributions.

• Used the results to determine the pressure drop through the test-rig, the flow splits around the engine, the heat-transfer coefficients on the engine and other heat-generating component surfaces, convected heat with theventilation air, and the air and surface temperatures on various thermocouple locations.

Future Direction• By using the data from Caterpillar experiments, the calculated temperatures using the 1-D network flow and the

CFD models will be compared with the measured temperatures in the next phase of the project.


102

IntroductionMostly because of sound-reduction requirements,modern off-road machines are designed withseparate engine and cooling system compartments.Since high underhood temperatures can reducecomponent durability and life, the assessment ofcomponent temperatures in a relatively well sealedengine enclosure is important. The main thermalmanagement challenge for such a configuration is todetermine the ventilation airflow needs in the engineenclosure and its effect on the thermal balance. Toprovide insight, a prototypical test-rig that includesan engine and other installation hardware was builtby engineers at Caterpillar, Inc. The experimentswere conducted with controlled ventilation air inletlocations and flow rates to help understand theeffects of ventilation on heat rejection andcomponent temperatures.

In parallel, an assessment of various simulationmethods was initiated by analysts at ArgonneNational Laboratory (ANL) for predictingunderhood ventilation airflow and temperatures inthe test-rig. The analytical effort consisted ofcombined use of a 1-D network flow model tosimulate the engine and the cooling system and a3-D CFD model for the ventilation air circulation inthe test rig. The goal has been to provide anintegrated analysis capability that needs only theflow rates and inlet temperatures as boundaryconditions to assure adequate air cooling and to helpidentify potential hot-spots.

CFD Model DevelopmentThe test rig included an engine installed into amock-up representing a typical medium-size off-highway machine with a full engine enclosure.Small doors installed at five sides of the enclosure(front, back, intake side, exhaust side, and bottom)served as ventilation air inlet locations. An externalblower installed in the outlet pipe at the top wasused to control the ventilation airflow rate. Thiscombination of fixed outlet and varying inletlocations helped researchers understand the effect ofinlet location on critical component temperatures.The engine coolant temperature and compressed airtemperatures at the intake manifold were maintainedby laboratory heat exchangers and instrumented toaccount for the heat rejection closely. The

ventilation air and component surface temperatureswere measured at various locations.

A CAD model of the test rig was prepared andprovided by the Caterpillar team to help assist themodel development. This CAD model was used todevelop a 3-D CFD model of the test rig by usingthe commercial CFD code STAR-CD. To capturethe ventilation airflow distribution at the enclosureinlet accurately, a large inlet plenum was includedfor each inlet configuration to represent ambientconditions for pressure and temperature. The CFDmodel consisted of approximately 1.34 million fluidcells with a 3-mm-thick cell layer surrounding theengine and enclosure surfaces.

To allow the data to be scaled for different enginecompartment configurations, the ventilation airflowrate was normalized with respect to the enginecombustion airflow rate. This ratio of the ventilationairflow rate to the engine combustion airflow ratewas used as the basis for comparing the analyticalresults with experimental data. The airflow ratio wasvaried from 0.5 to 3.75 during the experiments. In atraditional underhood compartment with no servicewall to separate engine from cooling systems, theairflow ratio is believed to be approximately 10.

Each of the five configurations was studied for fivedifferent ventilation airflow rates corresponding toairflow ratios of 0.5, 1.0, 1.5, 2.0, and 2.5. Thedesired flow rate through the enclosure was assuredby imposing a proportional uniform flow field at theplenum inlet as the boundary condition. An optionwith two-pressure boundary conditions (pressure-driven flow between the inlet plenum and outletpipe) was also investigated to assess the sensitivityof the flow field to the boundary conditions. Theparametric studies did not indicate a strongdependence of thermal mixing inside the enclosureto the turbulence boundary conditions at the inlet. Ineach case, 10% turbulence intensity and half-window-width mixing length were specified at theplenum inlet. Engine and enclosure boundaries werespecified as smooth “non-slip” walls, but the effectof surface roughness on the pressure drop was alsoevaluated.

A parametric study for inclusion of the buoyancyforce in the thermal-fluid calculations revealed thatthe effect of density variation on the overall flow


103

and temperature fields is negligible. Therefore, theventilation airflow field was simulated as a steadyincompressible flow, decoupled from the energyequation by using constant thermo-physicalproperties. This allowed a two-step analysis inwhich first the flow field is established inside theenclosure and then the temperature distributionswere obtained for a given set of thermal boundaryconditions.

Most of the simulations were performed by using thehigh-Reynolds number k-epsilon turbulence modelin conjunction with logarithmic wall functions;however, a few other turbulence models (such as therenormalization group [RNG] formulation, Chen’svariant, and low-Re model), as well as the quadraticand cubic k-epsilon options, were also evaluated.A set of transient calculations was studied toinvestigate temperature fluctuations observed duringthe experiments and to assure that the calculatedflow field was steady with no small frequencyoscillations. The results indicated a negligibledifference between the transient and steady-statesolutions. The calculations were performed on aLinux cluster, mostly using 4–6 processors.

1-D Network Model DevelopmentAs part of the analytical efforts, a 1-D thermal-fluidnetwork model was also developed by using thecommercial software FLOWMASTER. In the 1-Dmodel, the air, coolant, and oil loops wererepresented as different thermal subsystemsinteracting with each other and the engine structurethrough discrete heat transfer paths. The 1-D modelserved as a tool to analyze the interactions of theengine with the air, coolant, and oil loops forpredicting the complete thermal system performanceand to account for overall energy balance.

Engine Metal Structure: In the network model ofengine metal structure, the engine metal structure iscooled by the coolant, oil, and air. FLOWMASTERmodels the combustion gas as a heat source with aknown heat rejection to the cylinder head, liner, andpiston. The exhaust is treated as gas flow throughthe muffler, so the muffler temperature is at thetemperature of the exhaust gas. Any componentslinked to the muffler participate in the modeling ofthe heat rejection from the exhaust.

Cooling Air Circuit: The entering air gains heat asit passes through individual surface points on theengine mentioned above. FLOWMASTERcalculates the amount of heat gain based on thesurface area in contact and the heat transfercoefficient. The heat gain by air circuit is calculatedthrough the temperature variation between thecomponent inlet and the component outlet.

Lubrication Oil Loop: In the oil loop of 1-Dnetwork model, the oil is supplied in three separatebranches: (1) to the turbo and back to the sump;(2) to the cylinder head where the oil gains heatfrom the valve cover, the cylinder head, and theupper engine block and then returns to the sump;and (3) to the engine block where it gains heat fromthe piston rings and the lower engine block and thenreturns to the sump. The oil circuit does not removeany heat from the system; instead, it distributes theheat in the system.

Coolant Water Loop: In the coolant water loop, thewater coolant, at a given flow rate, passes throughthe oil cooler to cool the lubrication oil in oil circuit,circulates inside the upper engine block, and thencirculates through the cylinder head. The radiator issimply modeled by a flow source with constant flowrate and by a pressure source with knowntemperature and pressure.

The 1-D computation has the air inlet at front,bottom, rear, intake, and exhaust side, separately,with assigned airflow rates at all branches. In eachcomputation, the various temperatures in the modelare calculated on the basis of the certain physicaldimensions and the heat transfer coefficients at theengine-air interface. Some physical data on theengine were supplied by Caterpillar and input in themodel. The airflow paths/flow rates and the heat-transfer coefficients were obtained from CFDcalculations to compute the system temperaturedistributions. Steady-state simulation at a givenengine rated condition with conduction, convection,and radiation models was carried out. The air entersat the enclosure inlet (front of the enclosure) withgiven temperature, pressure, and flow rate and thenexits at enclosure outlet (top of the enclosure) withtemperature, pressure, and flow rate. The heat loadswere calculated at the engine surfaces fortemperature distribution.


104

Interface between the 3-D CFD and 1-DNetwork Flow ModelsThe ventilation air entering the enclosure wasconsidered to split around the engine block atproportions specified by the CFD model and gainheat as it passes through individual surface points onthe engine.

The surface heat flux boundary conditions for theCFD model were deduced from the calculatedsurface and bulk ventilation air temperatures fromthe 1-D model by using

( )bs TThAq

−=

where q/A is the surface heat flux (the boundarycondition for the CFD model), h is the heat transfercoefficient (input to the 1-D network flow model),Ts is the surface temperature, and Tb is theventilation air temperature near a specific engine orcomponent surface. In return, the results of the 3-DCFD analysis were used to provide the 1-D modelwith improved ventilation airflow paths and theresulting heat-transfer coefficients between the heat-generating components and ventilation air. Since the3-D CFD model requires component heat fluxes(q/A) from the 1-D network flow model while the1-D model requires accurate estimates of heattransfer coefficient (h), an iterative procedure wasemployed between the two models to obtainconsistent solutions. The iterations between the 3-DCFD and 1-D network flow models continued untilthe difference in the estimated heat transfercoefficients between two subsequent steps werenegligible. For most of the underhood components,the convergence was established after a fewiterations between the 1-D and CFD solutions.

ResultsThe ventilation airflow paths around the engine andthe surface and ventilation air temperatures in thetest-rig are calculated for all five inlet configurationsand five different airflow ratios. In agreement withthe experimental observations, the CFD resultsindicate a well-mixed flow inside the enclosure, withno significant difference in component temperaturesfor different inlet locations. The flow field suggestssignificant turbulence inside the enclosure since theaverage viscosity is two orders of magnitude greaterthan the molecular viscosity of air.

To provide the 1-D thermal-fluid network modelwith an accurate measure of ventilation airflowpaths around the engine for different inletconfigurations, the computational domain isconsidered in six interconnected zones around theengine, and the flow splits are determined from theflow rates through the surfaces separating thesezones. When normalized with respect to the inletflow rate, the calculated splits for each inletconfiguration are found to be independent of theinlet flow rate. Since the CFD results do not indicatepreferred (unidirectional) flow paths around theengine, the calculated flow splits based on net fluxacross each side of the engine tend to underestimatethe near-surface flow rates. An assessment of theimportance of underestimating the flow rates nearthe component surfaces will be made aftercomparing the analytical predictions with theexperimental data.

To predict the temperature distributions in the testrig by using only the inlet boundary conditions, thethermal calculations are performed by using thesurface heat fluxes as the boundary conditions basedon the results of the 1-D simulations. Consistentwith the nodalization of the 1-D model, the constantheat flux values are specified for the entire surfaceof heat-generating components. The portions of theenclosure wall near the muffler are also consideredas heat transfer surfaces because of exposure toradiation from elevated muffler temperatures. Thefinal heat-transfer coefficients, as estimated by theCFD model and used in 1-D simulations, follow anexpected trend and increase almost proportionallywith increased airflow ratio within the studied range.At the airflow ratio of 0.5, the heat-transfercoefficient estimates vary between 2 and22 W/m2•°C, with an average value of 10 W/m2•°C.At the airflow ratio of 2.5, the heat-transfercoefficients vary between 10 and 62 W/m2•°C, withan average value of 21 W/m2•°C.

As examples of the results, the ventilation airtemperatures for the front inlet case and airflow ratioof 1.5 are shown on several vertical planes inFigure 1. For all inlet configurations, thetemperatures are significantly higher on the exhaustside of the engine, particularly near the most heatgenerating components, the muffler and turbocharger. At low-ventilation airflow rates, asubstantially large temperature difference (on the


105

order of 50°C) exists with little thermal mixing ofthe hot and cold jets near the entrance to the outletpipe. The difference is smaller but still noticeable(about 25°C) at the highest airflow rate studied.

The convected heat from the engine enclosure withthe ventilation air is calculated by evaluating thefollowing summation for all the cells at the outletpipe exit

( )∑ −ρ= inoutp TTvAcq

where q is the rejected heat via ventilation air, ρ isthe air density, v is cell velocity, A is cell surfacearea, cp is specific heat of air, Tout is celltemperature, and Tin is ventilation air inlettemperature. The results, summarized in Figure 2,show a similar trend for all inlet configurations andsuggest a proportional increase in the convected heatwith the increase in ventilation airflow rate. On theother hand, the area averaged outlet temperatures atthe outlet pipe exit seem to exhibit a nonlinearbehavior and saturate at an airflow ratio of 2.0, asshown in Figure 3. With various inlet airflow ratiosand different air inlet locations, the 1-D simulationresults of both surface- and air-temperatures at somecomponents are shown in Figure 4.

ConclusionsAn assessment of combined 1-D and 3-D simulationmethods was initiated for predicting ventilationairflow and underhood temperatures in an enclosedengine compartment of an off-road machine. A CFDmodel was built to determine the 3-D flow field andthe rate of heat transfer between engine andventilation air inside the enclosure. Consistent withthe experimental observations, the CFD resultsindicate a well-mixed flow inside the enclosure withno significant difference in component temperaturesfor different ventilation inlet locations.

The calculations performed to date indicate that theCFD model generally under-predicts the pressuredrop based on the current geometric definition of thetest rig. This is perhaps, in part, due to omitteddetails in the model (various hoses and instrumentwires). Although the agreement between thecalculations and experiments on system restriction isgood for front inlet configuration, it is possible thatthe CFD model over-predicts the pressure drop inthat case because of the ignored effect of high-rpmrotation of the crank-shaft pulley right in front ofinlet window.

A 1-D model was used mainly to determine thetemperature distributions in engine structurecomponents, air loop, oil loop, and even coolantloop. The predictions show the consistency intemperature distribution with variation of the airflowratio for all air inlet locations.

By using the data from Caterpillar experiments, thecalculated temperatures resulting from the1-D network flow and the CFD models will becompared with the measured temperatures in thenext phase of the project. On the basis of thepreliminary assessments (by comparing thecalculated results with the only known experimentaldata for the front-inlet case and airflow ratio of 1.5),the combined model also under-predicts theventilation air temperatures. After the detailedassessment of the results and comparisons withexperiments, the model discretization issues andother uncertainties will be addressed to improve theaccuracy of predictions.

Heavy Vehicle System

s Optim

ization ProgramFY 2004 Annual Report

106

Figure 1. Temperature contours in vertical planes with equal increments from front to back. The results for front inlet configuration with airflow ratio of 1.5.


107

6000

8500

11000

13500

16000

0.5 1 1.5 2 2.5

Airflow Ratio

Hea

t Rej

ecte

d (W

)

frontbackbottomintake sideexhaust side

Figure 2. Heat rejected via the ventilation air through the engine enclosure as a function of airflow ratio.

40

45

50

55

60

65

70

75

0.5 1 1.5 2 2.5

Airflow Ratio

Ave

rage

Out

let T

empe

ratu

re (°

C)

frontbackbottomintake sideexhaust side

Figure 3. Area averaged outlet temperatures at the outlet pipe exit as a function of airflow rate.


108

0

10

20

30

40

50

60

70

0.5 1 1.5 2 2.5

Surface Temperature at ECMFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

20

40

60

80

100

120

0.5 1 1.5 2 2.5

Surface Temperature at Cylinder HeadFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

10

20

30

40

50

60

70

0.5 1 1.5 2 2.5

Surface Temperature at ECMFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

20

40

60

80

100

120

0.5 1 1.5 2 2.5

Surface Temperature at Cylinder HeadFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

(a) (b)

0

10

20

30

40

50

60

0.5 1 1.5 2 2.5

Air Temperature at Cylinder HeadFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

5

10

15

20

25

30

35

40

0.5 1 1.5 2 2.5

Air Temperature at ECM Front InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

10

20

30

40

50

60

0.5 1 1.5 2 2.5

Air Temperature at Cylinder HeadFront InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

0

5

10

15

20

25

30

35

40

0.5 1 1.5 2 2.5

Air Temperature at ECM Front InletBottom InletRear InletIntake-Side InletExhaust-Side Inlet

Air Flow Ratio

(c) (d)Figure 4. 1-D network predictions: surface temperatures at (a) cylinder head and (b) ECM and airtemperatures at (c) cylinder head and (d) ECM for various airflow ratios under different air inlet locations.


109

III. Friction and Wear

A. Boundary Lubrication Mechanisms

Principal Investigators: O.O. Ajayi, J.G. Hershberger, and G.R. FenskeArgonne National Laboratory9700 South Cass Avenue, Argonne, IL 60439(630) 252-9021, fax: (630) 252-4798, e-mail: [email protected]


Contractor: Argonne National Laboratory, Argonne, IllinoisPrime Contract No.: W-31-109-Eng-38.

ObjectivesDevelop a better understanding of the mechanisms and reactions that occur on component surfaces under boundarylubrication regimes. Specific objectives are to:

• Determine the basic mechanisms of catastrophic failure in lubricated surfaces in terms of materials behavior.

• Determine the basic mechanisms of chemical boundary lubrication.

• Establish and validate performance and failure prediction methodologies for lubricated components.

• Integrate coating and lubrication technologies for maximum enhancement of lubricated-surface performance.

Approach• Characterize the dynamic changes in the near-surface material during scuffing. Formulate a material-behavior-

based scuffing mechanism.

• Determine the chemical kinetics of boundary film formation and loss rate by in situ X-ray characterization oftribological interfaces at the Advanced Photon Source (APS) at Argonne National Laboratory (ANL).

• Characterize the physical, mechanical, and tribological properties of tribo-chemical films, including the failuremechanisms.

• Integrate the performance and failure mechanisms of all the structural elements of a lubricated interface toformulate performance and/or failure prediction methodology; this will include incorporation of surfacecoatings.

Accomplishments• Conducted extensive characterization of microstructural changes during scuffing of 4340 steel, by using

scanning electron microscopy (SEM) and X-ray analysis.

• Developed a model of scuffing initiation on the basis of an adiabatic shear instability mechanism.

• Formulated a preliminary model of scuffing propagation on the basis of a balance between heat generation andheat dissipation rates.

• Characterized the mechanical properties and scuffing resistance of a graded nanocrystalline surface layerproduced by the scuffing process.


110

• Using X-ray fluorescence, reflectivity, and diffraction at the APS, demonstrated the ability to characterize tribo-chemical films generated from model oil additives.

Future Direction• Integrate scuffing initiation and propagation models for a comprehensive scuffing prediction method.

• Experimentally validate the comprehensive scuffing theory for various engineering materials, includingceramics.

• Design and construct a tribo-tester for in situ characterization of tribochemical surface film at the APS.

• By using various techniques, continue to characterize tribochemical films formed by model lubricant additives.

• Characterize the physical, mechanical, and failure mechanisms of tribochemical films with nano-contact probedevices.

• Evaluate the impact of various surface technologies, such as coating and laser texturing, on boundarylubrication mechanisms.

IntroductionMany critical components in diesel engines andtransportation vehicle subsystems are oil lubricated.Satisfactory performance of these components andsystems is achieved through a combination ofmaterials, surface finish, and lubricant oilformulation technologies. To that end, an Edisoniantrial-and-error approach is often followed. Indeed,experience is likely the sole basis for new designand solutions to failure problems in lubricatedcomponents. With more severe operating conditionsexpected for component surfaces in advancedengines and vehicles, the trial-and-error approach toeffective lubrication is inadequate and certainlyinefficient. Departure from this approach willrequire a better understanding of the fundamentalsof both boundary lubrication and surface failuremechanisms.

A major area of focus for the U.S. Department ofEnergy in the development of diesel enginetechnology is emission reduction. Some essential oillubricants and diesel-fuel additives, such as sulfur,phosphorus, and chlorine, are known to poison thecatalysts in emission-reducing after-treatmentdevices. Although reduction or elimination of theseadditives will make emission after-treatment devicesmore effective and durable, it will also render thesurfaces of many lubricated components vulnerableto catastrophic failure. Many of the effectivemethods to reduce diesel engine emissions may alsorender critical-component surfaces vulnerable to

catastrophic failure, thereby compromising theirreliability.

Increases in vehicle efficiency will require anincrease in power density, resulting in increasedseverity of contact between many components in theengine and powertrain systems. This, again, willcompromise the reliability of various criticalcomponents, unless they are effectively lubricated.The efficacy of oil additives in protectingcomponent surfaces depends on the nature andextent of the chemical interactions between thesurface and the oil additives.

In addition to reliability issues, the durability oflubricated components also depends on theeffectiveness of oil lubrication mechanisms.Components will eventually fail or wear out overtime. Failure mechanisms that limit the durability oflubricated surfaces are wear, in its various forms,and contact fatigue. Wear is the gradual removal ofmaterial from contacting surfaces, and it can occurby various mechanisms, such as abrasion, adhesion,and corrosion. Also, the repeated contact stresscycles to which component contact surfaces aresubjected can initiate and propagate fatigue cracksand, ultimately lead to the loss of a chunk ofmaterial from the surface. This damage mode isoften referred to as “pitting.” Wear and contactfatigue are both closely related to boundarylubrication mechanisms. Antiwear additives inlubricants are designed to form a wear-resistantprotective layer on the surface. The role of lubricantadditives on the contact fatigue failure mode is not


111

fully understood, although it is clear that thelubricant chemistry significantly affects contactfatigue. Again, lack of an adequate, comprehensiveunderstanding of basic mechanisms of boundarylubrication is a major obstacle to a reasonableprediction of component surface durability.

The desire to extend the drain interval for dieselengine oil, with an ultimate goal of a fill-for-lifesystem, is increasing. Successful implementation ofthe fill-for-life concept will require optimization ofsurface lubrication through the integration ofmaterials, lubricant, and, perhaps, coatingtechnologies. Such an effort will require an adequatefundamental understanding of surface materialbehavior, chemical interactions between the materialsurface and the lubricant, and the behavior ofmaterial and lubricant over time.

Some common threads run through all of thechallenges and problems in the area of surfacelubrication of engine components and systemsbriefly described above. The two key ones are(1) lack of adequate basic and quantitativeunderstanding of the failure mechanisms ofcomponent surfaces and (2) lack of understanding ofbasic mechanisms of boundary lubrication (i.e., howlubricant chemistry and additives interact withrubbing surfaces and how this affects performance).To progress beyond the empirical trial-and-errorapproach for predicting lubricated componentperformance, a better understanding is required ofthe basic mechanisms regarding the events thatoccur on lubricated surfaces. Consequently, theprimary objective of the present project is todetermine the fundamental mechanisms of boundarylubrication and failure processes of lubricatedsurfaces.

The technical approach taken in this study differsfrom the usual one of posttest characterization oflubricated surfaces, and, rather, will involvedeveloping and applying in situ characterizationtechniques for lubricated interfaces that will use theX-ray beam at the APS located at ANL. Using acombination of X-ray fluorescence, reflectivity, anddiffraction techniques, we will study, in real time,the interactions between oil lubricants and theiradditives and the surfaces they lubricate. Such studywill provide the basic mechanisms of boundarylubrication. In addition to surface chemical changes,

materials aspects of various tribological failuremechanisms will be studied.

Results and DiscussionEfforts during the past year (FY 04) were devoted tothe development of a scuffing propagation model,study of near-surface microstructural changes andtheir implication for enhancing scuffing resistance,and refinement of the X-ray characterization oftribochemical boundary films.

Scuffing PropagationScuffing, which is a sudden catastrophic failure ofsliding surfaces, is perhaps the least understood ofthe various failure mechanisms in lubricatedsurfaces. Numerous phenomenological observationsabout scuffing have been made over the years byvarious investigators. It is known that sometimesscuffing will initiate but not propagate. In somecases, scuffing occurs so fast that it is difficult toidentify distinctive stages in the process.Nonetheless, on the basis of various observations onscuffing, a useful approach to study the basicmechanism is to separate the process into theinitiation and propagation stages. This approach canadequately account for the many cases of initiationwithout propagation.

In the previous year, we developed a model forinitiation of scuffing based on adiabatic plastic shearinstability, which occurs when the rate of thermalsoftening exceeds the rate of work hardening. Thelarge amount of heat generated by the severe plasticdeformation of shear instability is the driving forcefor scuffing propagation. Thermal softening of thematerials in the vicinity of the initiation sites couldbe large enough to propagate the scuffing front. Likethe scuffing initiation process, which involves acompetition between work hardening and thermalsoftening, scuffing propagation involves acompetition between heat dissipation and heatgeneration. We propose that scuffing will propagatewhen the rate of dissipation of heat produced byadiabatic shear instability is less than the rate of heatgeneration by the same process. If the rate of heatdissipation exceeds the rate of heat generation, thescuffing process can be “quenched,” in which caseonly microscuffing events are observed. Scuffingpropagation can thus be controlled either by


112

increasing the rate of heat dissipation or by reducingthe rate of heat generation.

On the basis of balancing the heat dissipation andheat generation from shear instability, the velocity(V) of propagation of the scuffing front can bedescribed by the following simplified equation in1-D:

δ

δρλ

ργβτ

δ

XT

XT

XT

CCdtdV

∂∂

⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

−∂∂

−==

2

2

2

2&

(1)

where:β = fraction of plastic work converted to heat

(0.9),τ = shear stress,λ = thermal conductivity,γ = strain rate,ρ = density, andC = heat capacity.

In Equation 1, the first term in the numeratordescribes the rate of heat generation, while thesecond term describes the rate of heat conductionacross the boundary of the scuffing propagationfront. When the propagation front velocity V ispositive, scuffing propagates. When the velocity iszero or negative, scuffing does not propagate.

In conjunction with the initiation model, it should bepossible to predict the occurrence of scuffing failurefrom pertinent material properties and operatingconditions. Some of our future efforts will bedevoted to the integration of the initiation andpropagation models for scuffing.

Graded Nanophase Surface LayerMicrostructural characterization of the near-surfacelayer of a scuffed steel surface showed the formationof a 20-µm-thick graded nanophase layer. The grainsize in the top layer is about 20–30 nm. This layer isproduced by a severe plastic deformation processassociated with the scuffing failure. It is well knownthat one method of grain refinement to thenanometer size range in metallic material is bysevere plastic deformation.

The mechanical properties of the nanocrystallinesurface layer produced on a hardened 4340 steelsurface were characterized by the nano-indentationtechnique. As shown in Figure 1, the hardness of thetopmost layer with the finest grain size issubstantially higher than that of the original material(13 GPa compared to 3.5 GPa). Similarly, the elasticmodulus of the top layer is increased from theoriginal 235 GPa to 294 GPa. Indeed, the hardnessand elastic modulus of the nanocrystalline top layerproduced during scuffing are comparable to thevalues for structural ceramic materials. Thisobservation represents a new approach to surfaceengineering.

H E (Gpa Hv RWC) (GPa)

First Layer 13.1 1336 85? 294.7

Second layer 9.0 918 67 232.2

Third layer 3.4 347 35 216.3

Bulk 3.5 367 36 235.8

Figure 1. SEM micrograph of graded nanocrystallinesurface layer on a steel surface after scuffing. Mechanicalproperties (hardness and modulus) are shown in table nextto the micrograph.

In view of the changes to the microstructure andsurface mechanical properties resulting fromscuffing, we evaluated the effect of the new ormodified surface layer on subsequent scuffingfailure. Figure 2a shows the statistical distribution ofscuffing occurrence in polished, hardened 440Csteel, and Figure 2b shows the result for the samematerial that was previously scuffed. The modifiedlayer showed scuffing resistance equal to or betterthan that of the unmodified surface. Moresignificantly, the tendency toward premature failurewas significantly reduced in the modified layer. Thebetter predictability of scuffing in the modified layerwill certainly reduce the challenge of designingtribological sliding systems against prematurescuffing failure.

X-Ray Characterization of Tribo-ChemicalFilmsCharacterization of tribochemical films formed bylubricant additives with the various X-ray techniquesavailable at the APS is ongoing. Films formed by


113

0

2

4

6

8

10

12

14

<120 120 180 240 300 360 420 480

As Polished

Num

ber

of O

ccur

renc

es

RPM at Scuffing

a)

Figure 2a. Histogram of scuffing occurrence at variousspeeds for a polished steel surface.

0

2

4

6

8

10

<120 120 180 240 300 360 420 480

Tribosynthesized and Smoothed

Num

ber

of O

ccur

renc

es

RPM at Scuffing

Figure 2b. Histogram of scuffing occurrence at variousspeeds for a steel surface with graded nanocrystallinelayer from prior scuffing.

two methods with zinc diathiophosphate (ZDDP) oiladditive were analyzed with glancing-incident-angleX-ray fluorescence, diffraction, and reflectivity. Thegoal is to assess the ability of these techniques todetect small differences in tribochemical filmsformed by the same additives.

Figure 3 shows the variation of Zn content in filmsformed by either tribo contact (mechanical) orthermally for different concentrations of ZDDPadditive in polyalphaolefin (PAO) lubricant. Theresults clearly show that the compositions of themechanical films and the thermal films are different.This observation is very significant, as experiencehas shown that small differences between

0 100

2 1017

4 1017

6 1017

8 1017

1 1018

1.2 1018

0.0 0.1 0.2 0.3 0.4 0.5

ThermalMechanical

Zn

atom

s per

cm

2

ZDDP Concentration

Figure 3. Areal density of Zn atoms in tribochemicalfilms formed thermally and mechanically as a function ofZDDP additive concentration.

tribochemical-films composition and structure cantranslate into significant differences in performance.The ability to detect small differences in tribo-chemical films is critical to a better understanding oftheir performance and optimization.

ConclusionsDuring the past year, a preliminary model forscuffing propagation was developed. This is the nextstep in the separation of the scuffing process intoinitiation and propagation stages. While the scuffinginitiation is by adiabatic shear instability, scuffingpropagation is governed by the balance between heatdissipation and heat generation as a result of severeplastic deformation of shear instability. When therate of heat dissipation exceeds the rate of heatgeneration, scuffing does not propagate.

Mechanical properties of the graded nanocrystallinesurface layer from scuffing were characterized. Boththe hardness and the elastic modulus of the newsurface layer were significantly increased to valuescomparable to those of structural ceramics. This newlayer was also shown to reduce the tendency towardpremature scuffing failure. It was also demonstratedthat the suite of X-ray characterization techniquesavailable at APS can detect differences in thetribochemical surface films formed from the sameoil additive by two methods.


114

Publications1. J. Hershberger, O.O. Ajayi, and G.R. Fenske,

“Improvement in Scuffing Resistance due toTribosynthesis of a Modified Surface Layer,”Thin Solid Films (in press).

2. O.O. Ajayi, J. Hershberger, J. Zhang, H. Yoon,and G.R. Fenske, “Microstructural Evolutionduring Scuffing of 4340 Steel – Implication forScuffing Mechanism,” Tribology International(in press).

3. J. Hershberger, O.O. Ajayi, and G.R. Fenske,“Composition of ZDDP Films FormedThermally and Mechanically,” TribologyInternational (in press).


115

B. Parasitic Engine Loss Models

Principal Investigator: George Fenske, co-workers: Ali Erdemir, Layo Ajayi, andAndriy KovalchenkoArgonne National Laboratory, Energy Technology DivisionArgonne, IL 60439(630) 252-5190, fax: (630) 252-4798, e-mail: [email protected]

Bill Brogdon, Troy Torbeck, Isaac Fox, and Jim KezerleRicardo, Inc.7850 Grant StreetBurr Ridge, IL 60527-5852(630) 789-0003, fax: (630) 789-0127


Contractor: Argonne National Laboratory, Argonne, IllinoisContract No.: W-31-109-ENG-38

Objectives• Develop and integrate a set of software packages to predict the impact of advanced low-fiction coating

technologies on parasitic energy losses from diesel engine components.

• Develop a suite of computer codes to predict the wear and durability of engine components when exposed tolow-viscosity lubricants.

Approach• Predict fuel economy improvements over a wide range of oil viscosity.

• Predict change in wear loads due to reduced oil viscosity.

• Develop superhard and low-friction coatings that can reduce friction and wear in low-viscosity oils.

Accomplishments• Modeled the impact of low-friction coatings and low-viscosity lubricants on fuel savings (up to 4%), and

predicted the impact of low-viscosity lubricants on the wear/durability of critical engine components.

• Developed experimental protocols to evaluate the friction and wear performance of advanced engine materials,coatings, and surface treatments under prototypical piston/ring environments.

Future Direction• Apply superhard and low-friction coatings on actual engine components and demonstrate their usefulness in

low-viscosity oils.

• Optimize coating composition, surface finish, thickness, and adhesion to achieve maximum fuel savings.


116

IntroductionFriction, wear, and lubrication impact energyefficiency, durability, and environmentalsoundness of all kinds of transportation systems,including diesel engines. Total frictional losses ina typical diesel engine may alone account for morethan 10% of the total fuel energy (depending onthe engine size, driving condition, etc.). Theamount of emissions produced by these engines isrelated to the fuel economy of that engine. Ingeneral, the higher the fuel economy, the lower theemissions. Higher fuel economy and loweremission in future diesel engines may be achievedby the development and widespread use of novelmaterials, lubricants, and coatings. For example,with increased use of lower-viscosity oils (thatalso contain the lowest amounts of sulfur andphosphorus-bearing additives), the fuel economyand environmental soundness of future enginesystems can be dramatically improved.Furthermore, with the development and increaseduse of smart surface-engineering and coatingtechnologies, even higher fuel economy and betterenvironmental soundness will be feasible.

The main goal of this project is to develop a suiteof software packages that can predict the impact ofsmart surface engineering and coatingtechnologies (e.g., laser dimpling, near-frictionlesscarbon, and superhard nitride coatings) onparasitic energy losses from diesel enginecomponents. The project also aims to validate thepredictions by using experimental friction andwear studies at Argonne National Laboratory.Such information will help identify critical enginecomponents that can benefit the most from the useof novel surface technologies, especially whenlow-viscosity engine oils are used to maximize thefuel economy of these engines by reducingchurning and/or hydrodynamic losses. A longer-term objective is to develop a suite of computercodes capable of predicting the lifetime/durabilityof critical components exposed to low-viscositylubricants.

In FY 2004, Argonne and Ricardo, Inc., workedtogether to identify engine components that canbenefit from low-friction coatings and/or surfacetreatments. The specific components consideredincluded rings, piston skirt, piston pin bearings,

crankshaft main and connecting rod bearings, and cambearings. Using computer codes, Ricardo quantifiedthe impact of low-viscosity engine oils on fueleconomy. Ricardo also identified conditions that canresult in direct metal-to-metal contacts, which, in turn,can accelerate engine wear and asperity friction.Efforts were also initiated to identify approaches tovalidate the predictions under fired conditions.

Argonne also worked on the development and testingof a series of low-friction coatings under a wide rangeof sliding conditions by using low- and high-viscosityengine oils. These coatings (such as near-frictionlesscarbon), as well as laser-textured surfaces, weresubjected to extensive tests using bench-top frictiontest rigs. The test conditions (i.e., speeds, loads, andtemperatures) were selected to create conditions inwhich direct metal-to-metal contacts will prevail, aswell as situations in which mixed or hydrodynamicregimes will dominate. Using the ranges of frictionaldata generated by Argonne, Ricardo estimated theextent of potential energy savings in diesel enginesand identified those components that can benefit themost from such low-friction coatings and/or surfacetreatments. Argonne developed a test rig to simulateengine conditions for piston rings sliding againstcylinder liners — one of the major sources of parasiticenergy losses identified in Ricardo’s studies. Weardata generated by Argonne can be used to developmodels and computer codes that predict thelifetime/durability of diesel engine components.

ResultsIn a series of computer simulations, Ricardo analyzedthe effects of lubricant viscosity on engine wear loads,engine friction, and fuel consumption. With drasticreductions in oil viscosity, it was found that in additionto other engine components, crank-shaft bearingswould become vulnerable to high friction and wearlosses. The situation with crank shaft bearings wasanalyzed by using ORBIT software, which takes intoaccount non-circular bearing/journal geometry,distortions, cavitation, and boundary-lubricated slidingconditions, where asperity contact andelastohydrodynamic lubrication may dominate. Inaddition to the ORBIT software, PISDYN,RINGPACK, and VALDYN simulations were carriedout under both baseline and reduced viscosityconditions. The objective was to determine the impactof oil viscosity on overall engine friction and wear.


117

Engine Friction StudiesThe Ricardo suite of engine component simulationcodes was run at varying levels of lubricantviscosities (and four levels of boundary friction –baseline, and 30, 60, and 90% reductions) at eightload and speed conditions. Hydrodynamic andasperity friction predictions were tabulated foreach case. The predicted optimum lubricant gradefor the baseline piston was SAE 20. With a 30%reduction in asperity friction, the allowedreduction in lubricant viscosity did not result in alarge reduction in frictional mean effectivepressure (FMEP). A 60% reduction in asperityfriction allowed a larger reduction in lubricantviscosity. With a 90% reduction in asperityfriction, the predicted optimum viscosity waslower than SAE 5, and the reduction inhydrodynamic friction was significant. Thesefindings are summarized in Figure 1.

Change in Fuel Consumption versus Viscosity Grade and Asperity Friction

Reduction

-5%

-4%

-3%

-2%

-1%

0%

1%

2%

3%

0 5 10 15 20 25 30 35 40 45 50 55SAE Viscosity Grade

Pre

dict

ed C

hang

e in

Fue

l Con

sum

ptio

n

0% Reduction

30% Reduction

60% Reduction

90% Reduction

Figure 1. PISDYN predictions of FMEP vs. SAElubricant viscosity grades and asperity frictionreduction.

According to these predictions, for the piston skirtalone, with a 90% reduction in asperity friction,reducing lubricant viscosity reduced parasitic pistonlosses by nearly 50%.

Piston Wear StudiesAs lubricant viscosity was reduced, wear loads on thepiston skirt and cylinder liner were increased in bothmagnitude and extent, thus increasing the vulnerabilityof these components to rapid wear. As shown inFigure 2, the total average wear load per cycle isabove four times higher for SAE 5 than SAE 40 oil.These findings suggest that, to allow the use of SAE 5oil, a surface treatment is needed to provideapproximately four times the wear resistance of thebaseline engine material.

Coating StudiesAt Argonne, systematic studies are being performed todevelop low-friction and high-wear resistant coatingsfor use in critical engine parts and components. One ofthe coatings studied extensively was near-frictionlesscarbon (NFC), which can provide friction coefficientsof 0.001–0.01 under dry sliding conditions, especially

Normalized Total Wear Load on Piston and Liner

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

0 5 10 15 20 25 30 35 40 45 50 55

SAE Viscosity Grade

Rel

ativ

e W

ear

Load

Figure 2. The effect of SAE lubricant viscosity grade onrelative wear load index of piston and liner in an engine.


118

in inert gas environments. Another coating(superhard nanocomposite – SHNC) underdevelopment is extremely hard (67 GPa hardness)and has a surface chemistry that enhances itsperformance in the presence of commercial oiladditives. A third approach at Argonne (inconjunction with Technion University) involvesthe use of laser-textured surfaces to modify theStribeck response of the surface — essentiallylowering frictional losses under boundarylubrication regimes. Under boundary lubricationconditions, NFC and SHNC coatings have beenshown to reduce friction by up to 90%. Laser-surface textured-treated surfaces to reduce frictionby similar amounts.

Recent studies at Argonne have been initiated toinvestigate the impact of coatings on wear. Inthese studies, a laboratory test rig is used toimpose unidirectional or reciprocating motion ontest coupons to investigate the wear (and friction)properties. Figure 3 shows a schematic (andphotograph) of the high-frequency reciprocatingrig. In the tests presented below, the ball was52100 steel (either uncoated or uncoated), and theplate was hardened H-13 steel (uncoated orcoated). The 20-N load produced an initialHertzian contact pressure of 1.05 GPa. The oil wasan unformulated basestock poly-alpha olefin(PAO), a formulated PAO, or an unformulatedparaffin oil.

Tests were performed under a wide range ofconditions (speed, load, lubricant, and coating).For the NFC coatings, three grades of coatingswere used: NFC2, NFC6, and NFC7, which showdistinct differences in terms of friction, hardness,

Figure 3. Schematic and photograph of a reciprocatingtest rig used to study friction and wear.

and hence wear. The wear rates for uncoated 52100steel balls sliding against coated (and uncoated) H-13steel plates are shown in Figure 4. The wear of theNFC-coated plates during the 1-h test was below thelimit of detection (approximately 50 µm), except forthe NFC6 coating exposed to the basestock PAOlubricant. Nevertheless, the results demonstrate thatthe use of a low-friction, wear-resistant coating canreduce wear rates by several orders of magnitude —sufficient for the factor of 4 increase in wear loadseverity for the low-viscosity lubricant.

1.E-11

1.E-10

1.E-09

1.E-08

1.E-07

1.E-06

1.E-05

NFC-2 NFC-6 NFC-7 Uncoated

Coating TypeW

ear R

ate

(mm

3/N

-m)

Synthetic

Paraffin OilMobil 1

Lower Limit of Detection

Figure 4. Wear rate of NFC-coated (and uncoated) H-13steel plates reciprocating against 52100 steel balls.

Even larger improvements in wear resistance havebeen observed (up to five orders of magnitude). Theextent of improvement depends on the coating andtribological environment. Preliminary results on theSHNC coatings are very promising — the highhardness, coupled with the surface chemistry betweenthe SHNC and the oil additives, suggests an extremelyhigh level of wear resistance.

DiscussionPreliminary results from computer simulations showedthat hydrodynamic friction decreased with reductionsin lubricant viscosity, while wear loads and asperityfriction increased. This trend is mainly due to thereduction of churning losses associated with the shearrheology of high-viscosity oils. However, the declinein hydrodynamic friction is eventually offset by theincrease in asperity friction. For each level of asperityfriction reduction, there is a lubricant viscosity thatprovides the minimum overall fuel consumption(Figure 1). For SAE 5 grade oil, simulations predict anoverall fuel savings of greater than 4%, if a surfacetreatment is used to reduce boundary friction by 90%.Contact severity and wear loads are substantiallyincreased in such low-viscosity environments, and


119

they would need to be mitigated to avoid increasedwear by using superhard coatings.

Work at Argonne has resulted in the developmentof NFC films as well as SHNC coatings, both ofwhich provide not only very low friction but alsomuch needed wear resistance in low-viscosity oils.With further development and optimization, thesecoatings may, indeed, prevent wear during slidingunder conditions where low-viscosity oils areused. They can also provide low friction to helpfurther increase fuel efficiency. Laser dimpling ofselected engine components offers additionalbenefits in terms of reducing friction and thusincreasing the fuel economy of engines. Some ofthese surface treatments and/or coatings will soonbe applied on portions or segments of real pistonrings and cylinder liners to further substantiatetheir effectiveness under conditions that closelysimulate the actual engines. In FY 2004, work atArgonne has focused on developing areciprocating test rig that uses segments of heavy-duty piston rings and cylinder liners to measurefriction and wear.

ConclusionsComputer simulation studies to date haveconfirmed that using low-viscosity engine oils willincrease the fuel efficiency of future dieselengines. However, wear load and asperity frictionwill increase substantially. The use of a low-friction, superhard coating on critical enginecomponents should mitigate asperity friction andreduce wear, thereby enabling the use of low-viscosity oils to increase the fuel efficiency.

Efforts are in progress to experimentally validatethe computer simulations. These efforts include(1) lab-scale tests to provide experimental frictionand wear data on NFC, SHNC, and texturedsurfaces and (2) single-cylinder fired engine testsdesigned to directly measure friction forcesbetween the liner and piston skirt and rings.

The lab tests will use segments of productionrings, piston skirts, and liners, the surfaces ofwhich have been modified with NFC and SHNCcoatings and/or laser texturing. The testconfiguration will replicate the reciprocatingmotion of the rings and pistons. Information will

be recorded continuously on the friction coefficient,and the wear data will be determined from post-testcharacterization of the components. Figure 5 showsphotos of a ring and liner mounted in test fixtures.

The Argonne rig will provide continuous frictionmeasurements during the reciprocating tests. Post-testoptical and interferometry will provide wear data. Therig shown in Figure 5 will be used to provide wear andfriction data under boundary and mixed lubricationregimes for use by Ricardo in its simulation studies.Tests will be performed with uncoated, coated, andtextured surfaces, as well as with formulated andunformulated oils.

Ricardo examined a number of potential fired-enginetechniques to validate the simulation studies. Theseincluded (1) direct fuel consumption measurements inmass-production engines, (2) friction-forcemeasurements in an instrumented mass-productionengine, and (3) friction measurements in aninstrumented single-cylinder engine. The thirdapproach was selected, and efforts are in progress toinstall a fixed-sleeve adaptor on a Ricardo single-cylinder Hydra engine. Load cells mounted on thefixed sleeve will enable continuous, directmeasurement of the friction forces between thering/piston assembly and the cylinder liner duringfired operation. Tests will be performed under a rangeof operating conditions (load and speed) to comparethe piston assembly friction losses of modified pistonsand rings against a baseline assembly.

Figure 5. Photographs of ring and liner segments mountedin a high-frequency reciprocating rig.


120

Publications1. Fox, I., “Numerical Evaluation of the Potential

for Fuel Economy Improvement due toBoundary Friction Reduction within Heavy-Duty Diesel Engines,” ECI International Conf.on Boundary Layer Lubrication, CopperMountain, CO, Aug. 2003.

2. Fenske, G., Erdemir, A., and Ajayi, O., “LowFriction Coatings and Lubricant SYSTEMS,”presentation at the 95th Annual AOCS Meeting& Expo, Cincinnati, OH, May 2004.


121

C. Superhard Nanocrystalline Coatings for Wear and Friction Reduction

Principal Investigators: A. Erdemir, L. Ajayi, O. Eryilmaz, K. Kazmanli, and I. EtsionArgonne National Laboratory, Energy Technology DivisionArgonne, IL 60439(630) 252-6571, fax: (630) 252-4798, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Jules Routbort(630) 252 5065, fax: (630) 252 4289, e-mail: [email protected]


Objectives:• Design, develop, and optimize low-friction, super-hard, nano-composite coatings for use in diesel engines and

drivetrain components of heavy vehicles.

• Demonstrate scalability and large-scale production of such coatings by using a large physical vapor depositionsystem.

• Characterize structural, mechanical, and tribological properties of these nano-composite coatings.

• Explore integration of these coatings with laser-textured surfaces.

• Demonstrate application and superior performance of these coatings in diesel engine and drivetraincomponents.

Approach:• Prepare test samples for the deposition of superhard and low-friction coatings.

• Identify and optimize deposition parameters that are critically important for the synthesis of nano-compositeand super-hard coatings on these samples.

• Assess structural, mechanical, and tribological properties of coatings.

• Analyze test results and examine sliding surfaces.

• Integrate optimized superhard coating technology with laser surface texturing.

• Report results on friction and wear properties.

Accomplishments:• Successfully demonstrated large-scale production of superhard coatings in a commercial system.

• Optimized deposition conditions (i.e., bias voltage, substrate temperature, deposition pressure, etc.) anddeveloped a deposition protocol that can provide consistent reproducibility and reliability on coated samples.

• Achieved friction coefficients of less than 0.03 under boundary lubricated sliding conditions on highlyoptimized superhard coatings.

• Virtually eliminated wear of coated surfaces.

• Successfully produced superhard coatings on a series of laser-dimpled surfaces and achieved very uniformcoverage and strong bonding to steel substrate.


122

• Demonstrated superior wear performance for sliding surfaces that were laser textured and then superhardcoated.

• Produced superhard coatings on cut segments of rings and liners in order to assess their friction and wearbehavior under boundary lubricated sliding conditions and at elevated temperatures.

Future Direction:• Further optimize coating microstructure, chemistry, property, and performance.

• Control size of grains and chemistry of grain boundaries to achieve even better mechanical properties.

• Perform more detailed characterization studies by using both the surface and structure-sensitive techniques(i.e., HRTEM, AES, XPS, nano-mechanical characterization).

• Produce laser-textured dimples on superhard coated-steel surfaces and assess their tribological behavior underboundary lubricated sliding regimes.

• Perform long-duration friction and wear tests by using a reciprocating ring-on-liner test machine anddemonstrate usefulness of new coatings with and without laser texturing under conditions that are typical ofactual engine components (load, speed, temperature).

• Evaluate effect of viscosity on friction and wear performance of laser-textured plus superhard coated surfaces(several more samples were sent to Technion University in Israel for laser texturing).

• Elucidate lubrication mechanism by using surface sensitive techniques; determine the chemistry of boundaryfilms.

• Explore other promising coatings systems (other than the current one).

• Transfer this technology to diesel engine industry.

IntroductionIncreasing demands for higher fuel economy, loweremissions, and longer durability in diesel and otherengine systems are pushing current materials andlubricants to their limits. Reduction of oil viscosityas well as sulfur- and phosphorous-bearing additivesin future engine oils are making current enginematerials increasingly more vulnerable to severewear and high frictional losses. Therefore, it is veryimportant to develop new materials and/or coatingsthat can resist wear and, at the same time, providelow friction when used in future diesel engines anddrivetrain components. Accordingly, the major goalof this project was to design, develop, and optimizenovel superhard and low-friction coatings for use incritical diesel engine parts and components, inparticular piston rings and cylinder liners. Anotherimportant goal of this research was to integratesuperhard coatings with a laser surface texturingprocess in order to achieve much higherperformance and durability under boundarylubricated sliding conditions where direct metal-to-metal contacts occur.

Synthesis and Large-Scale Production ofSuperhard CoatingsOne of the major goals of this project was todemonstrate the production of superhard coatings onselected test samples and engine components byusing a commercially viable deposition system.With the procurement of a production-scale sputterion plating system (Model # CC800/9XL, seeFigure 1) from CemeCon-USA, Inc., we have madegreat strides in meeting this goal. Specifically, wehave successfully synthesized high-qualitysuperhard and low-friction coatings at very largescales. These coatings were jointly conceived byscientists at Argonne and Istanbul TechnicalUniversity and were produced originally by using anarc physical vapor deposition (arc-PVD) system. Inrecent months, we were able to produce similarquality films by using the commercial-scale sputter-ion plating system shown in Figure 1. This system isequipped with two sputtering targets and/or cathodesso that the ceramic-based films can be doped withvarying amounts of alloying elements and/orchemical compounds to result in a nanostructuredand nano-composite microstructure. These alloying


123

Figure 1. General view of CemeCon CC800/9XLCoating system.

elements are strategically selected to enhance thechemical reactivity or responsiveness of coatingsurface to additives in oils and thus produce low-friction, scuff-resistant boundary films duringlubricated rolling, sliding, or rotating contacts.Ceramic phases within the films are expected toincrease hardness and thus resist adhesive andabrasive wear.

Structural CharacterizationFigure 2 shows the cross-section SEMphotomicrographs of a nano-composite filmproduced on a H13 steel substrate. High-resolutionSEM studies of ion-etched film cross-sectionsrevealed clear evidence of a nano-compositemorphology. Combined X-ray diffraction and TEMstudies have further confirmed that the individualgrains within the films were around 50 nm and madeof the hard ceramic phase. The softer metallic phasewas found at the grain boundaries and wasextremely thin. To produce such a dense and nano-composite film morphology, we identified andeffectively controlled several important depositionparameters (such as bias on sputtering target andsubstrate holder, chamber pressure, depositiontemperature, sputtering rate, plasma gascomposition, and the rotational speed of sampleholder) that seemed to have a very strong influenceon both film morphology and chemistry.

Substrate

Nano-Grains

Cross-Section electronMicroscopy image ofNew, nano-structuredCoating

(a) (b)

Figure 2. (a) General and (b) high-resolution SEMphotomicrograph of cross-section of a superhard coating.

Mechanical and TribologicalCharacterizationInitial mechanical characterization of nanostructuredcoatings was carried out by using a Fischersope-H100 nanohardness test machine. Hardnessmeasurements have revealed that these coatingswere truly superhard, with typical hardness valuesranging from 40 to 60 GPa. We have also noticedthat there were certain regions or phases within thesurveyed coating surface that exhibited hardnessvalues as high as 70 GPa.

To assess film-to-substrate adhesion, we used aRockwell C indentation method. Specifically, adiamond indenter was pressed against the superhardcoated surface until a spherical indent was created.Mainly because of the very large elastic and plasticstrain in and around the indented region, if the filmadhesion was poor, then one would expect widespread film delamination or detachment from theseregions. However, as shown in Figure 3, there weresome cracks near the indented spot but no signs offilm detachment, thus verifying that the film-to-substrate adhesion was very strong.

Tribological testing of superhard coatings was carriedout by using pin-on-disk and reciprocating testmachines. These tests were performed under oil-lubricated sliding conditions at very low velocities andunder very heavy loads in order to create a boundary-


124

Figure 3. Optical photomicrograph of an indent verifyingthat there is no film detachment from substrate steel.

lubricated sliding regime in which direct metal-to-metal contact occurs. As a baseline, we have alsorun a few tests under dry sliding conditions. Thefriction coefficients of nano-composite films underdry sliding conditions were between 0.2 and 0.4against steel or ceramic counterfaces. However,when tested under boundary-lubricated slidingconditions, their friction coefficients weresignificantly lower (i.e., 0.03–0.07) compared tosteel/steel test pairs, which provided frictioncoefficients of 0.1–0.15 under boundary-lubricatedsliding conditions. Figure 4 shows the actualfrictional behavior of a nano-composite coatingunder the boundary-lubricated sliding regime. Notethat the steady-state friction coefficient of thiscoating is less than 0.03. It is obvious that thepresence of such a coating on one of the sliding

Figure 4. Typical friction behavior of superhard coatingduring sliding against a steel pin under boundarylubricated sliding condition.

surfaces can reduce friction by 75%. Such animprovement in lubricity of sliding surfaces canhave a broad positive impact on actual engine partsand components. Because of their excellentcompatibility with or favorable response tolubricated test environments, these coatings can beregarded as smart or highly adaptive tribologicalcoatings. These coatings may further be optimized toprovide low friction and wear, even in the nearabsence of sulfur and phosphorous-bearing additivesin future engine oils.

Combining Superhard Coatings with LaserTexturingDuring this work period, we have also explored thesynergistic effects of combined uses of superhardcoatings and laser-textured surfaces on friction andwear. Laser texturing produces shallow dimples onmetallic or ceramic surfaces and when used underlubricated sliding conditions, these dimples cansignificantly improve mixed and hydrodynamiclubrication behavior of such surfaces. Shallowdimples can also trap abrasive wear debris that arecreated and thus reduce wear damage on slidingcontact surfaces. Earlier work in our laboratory hasverified that dimpled surfaces work extremely wellunder mixed or hydrodynamic sliding regimes oflubricated contacts; however, under boundarylubricated sliding conditions, they provide marginalimprovements in friction but may cause increasedwear losses. We felt that by applying a superhardcoating over dimpled surfaces, we can also achievevery high wear resistance. Accordingly, during thiswork period, we have prepared a batch of laser-textured H-13 steel flats and later coated them with asuperhard coating. It is expected that when appliedon a textured surface, superhard coatings mayfurther improve friction and wear performance ofsuch surfaces, especially under high-load, low-speedsliding conditions that can lead to scuffing andhence major wear losses. Figure 5 shows a dimpledregion with superhard coating on the surface.

Some of the dimpled and superhard coated-steelsamples were subjected to friction and wear by usinga reciprocating test machine under lubricated slidingconditions. Figure 6 compares the frictionperformance of as-received base steel, laser-dimpled


125

Figure 5. Hard coating applied over a dimple andmagnified details of a segment showing coating and steelsubstrate.

Figure 6. Friction behavior of as-received, laser-dimpled,and laser dimpled plus superhard coated surfaces underboundary lubricated sliding conditions.

steel, and superhard coated plus laser-dimpled steelsurfaces under boundary lubricated slidingconditions.

On the basis of the results shown in Figure 6, it isclear that laser-dimpled and superhard coated-steelsurfaces give fairly low friction coefficients, whilebase steel and combined laser-dimpled plussuperhard coated surface provide same levels offriction. Surface studies after the tests indicated thatdespite their relatively higher friction coefficients,laser-dimpled plus superhard coated surfacessuffered the least amount of wear regardless of thecontact loads; thus, verifying that the combined usesof superhard coatings and laser-textured surfaces canalleviate the wear problems of these surfacesobserved during sliding under boundary lubricatedsliding conditions.

In addition to sliding friction and wear studiesmentioned above, we have explored the possibilityof applying superhard coatings on cut segments ofactual piston rings and cylinder liners for friction

and wear studies by using a dedicated ring-on-linertype test machine.

Figure 7 shows cut segments of a ring and a linerthat were coated with superhard coatings. This workis still in progress, and we expect to obtainpreliminary friction and wear data very soon.

Figure 7. General layout of cut segments of dimpledpiston ring and cylinder liner placed in its holder in areciprocating test machine.

ConclusionsWe have successfully demonstrated large-scaleproduction of superhard coatings in a commercial-size deposition system. We have also optimized thedeposition conditions (i.e., bias voltage, substratetemperature, deposition pressure, etc.) and henceachieved excellent reproducibility and reliability oncoated samples. We have fully characterized thesecoatings by using a variety of mechanical andtribological test machines. Highly optimizedsuperhard coatings were able to provide frictioncoefficients of less than 0.03 under boundarylubricated sliding conditions. The amount of wear ofsliding surfaces was difficult to measure. We havealso produced superhard coatings over laser-texturedsurfaces and demonstrated their superior wearproperties under boundary lubricated slidingconditions. Optimized superhard coatings were alsoproduced on cut segments of piston rings andcylinder liners, and they are currently being tested ina ring-liner type test machine.


126

Patents and PublicationsSo far, we have filed two invention disclosures onthe work that we have done under this project. Oneof them (titled: low-friction, long wear coatings forconveyor assemblies) deals with the application ofsuperhard coatings on chain links of automotiveassembly plants. The second disclosure (titled:method to produce modulated composite surfaces)deals with the application of low-friction, superhard,and soft coatings over laser-textured surfaces.

During this period, we published and/or presentedseveral papers on the subject. Below is a list of mostsignificant ones.

1. Erdemir, A., “Review of EngineeredTribological Interfaces for Improved BoundaryLubrication,” Tribology International, In Press,2004.

2. Erdemir, A., “Smart Surface Engineering forImproved Boundary Lubrication,” InvitedKeynote Paper, Published in the Proc. of the 14th

International Colloquium on Tribology andLubrication Engineering, Stuttgart, Germany,January 13–15, 2004, Pages 13–20.

3. Donnet, C., and Erdemir, A., “HistoricalDevelopments and New Trends in Tribologicaland Solid Lubricant Coatings,” Surface andCoatings Technology, 180–81:76–84, 2004.

4. Kazmanli, K.; Eryilmaz, O.L.; Ajayi, O.O.;Erdemir, A.; and Etsion, I., “Effects of Soft andHard Coatings on Tribological Behavior ofLaser Textured Surfaces,” paper submitted forpresentation at 2005 Annual Meeting of theSociety of Tribologists and LubricationEngineers, May 15–19, 2005, Las Vegas, NV.

5. Erdemir, A., “Engineered TribologicalInterfaces for Improved Friction, Wear andLubrication in Engines,” Invited PanelPresentation at the Engine and DrivetrainSession of the 59th Annual Meeting of theSociety of Tribologists and LubricationEngineers, Toronto, Canada, May 17–20, 2004.

6. Eryilmaz, O.L.; Urgen, M.; and Erdemir, A.,“Production and Characterization of Nano-composite Coatings Produced by Arc andMagnetron Sputtering Physical VaporDeposition Techniques,” Presented at theInternational Conference on MetallurgicalCoatings and Thin Films, San Diego, CA,April 19–23, 2004.


127

IV. Fuel Reformer Systems

A. Diesel Fuel Reformer Technology

Principal Investigator: M. KrumpeltArgonne National Laboratory9700 S Cass Ave., Argonne, IL 60439(630) 252-8520, fax: (630) 252-4176, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Jules Routbort(630) 252-5065, e-mail: [email protected]

ParticipantsHual-Te Chien, Argonne National LaboratoryShuh-Haw Sheen, Argonne National Laboratory

Contractor: Argonne National LaboratoryContract No.: 49099

Objectives• Develop an autothermal reforming process to convert diesel fuel into hydrogen-rich gas on-board a heavy-duty

vehicle in a small fuel processor.

• Develop a fuel-air mixing device to achieve perfect mixing and reduce coke formation.

• Determine operating parameters for establishing “cool flame” conditions and evaluate the effects on reformingprocess.

Approach• Evaluate engineering issues that need to be better understood with regard to diesel fuel reforming. The main

issues are how to avoid pre-ignition and coke formation and how to achieve catalyst stability.

• Construct a test facility with single-nozzle diesel fuel reforming.

• Develop a fuel/steam/air injection and mixing system that can achieve homogeneous mixing and stable coolflame conditions.

• Conduct diesel fuel-reforming tests to evaluate the catalytic autothermal reforming process.

• Develop a compact diesel-fuel reforming system that complies with industrial needs.

• Conduct field tests and document the results.

Accomplishments• Completed construction of the fuel/steam/air mixing test facility.

• Completed the Argonne safety review of the facility.

• Evaluated the fuel-injection control system provided by International Truck and Engine Corporation (ITEC).

• Designed and constructed a steam/air injection device.


128

Future Direction• Complete development of the fuel/steam/air mixing apparatus.

• Conduct fuel injection/mixing tests to determine the operating parameters for cool flame.

• Modify the facility for autothermal reforming tests.

• Conduct autothermal reforming tests.

IntroductionConverting diesel fuel into a hydrogen-rich gas haspotential applications in auxiliary power units(APUs) for heavy-duty vehicles and emissionscontrol of diesel engines. An APU of 3–10-kWelectric capacity can provide enough electric powerto operate the air conditioners and heaters for thecabin and the cargo space while the vehicle isparked. It is common practice in the truckingindustry today to keep the diesel engines running togenerate electricity while drivers rest or are held atcustomer sites. The efficiency of diesel engines isonly 9% in this idling mode. Significant fuel savingsand emission reductions would be achieved if kW-capacity APUs for trucks were available and theengines could be turned off. Argonne NationalLaboratory (ANL) has developed a catalyticautothermal reforming process that may be used as acompact and potentially cost-effective technologyfor diesel-fuel reforming. However, because dieselfuel is more difficult to process because of highsulfur content and large aromatic molecules, theautothermal reforming process may encounter pre-ignition and coke formation problems. Operating athigher temperatures mitigates these problems butleads to rapid catalyst deactivation. This programwill explore the feasibility of achieving the diesel-fuel reforming by optimizing fuel/exhaust-gasmixing dynamics. Several engineering issues, suchas how to avoid pre-ignition and coke formation andhow to maintain catalyst stability, will be examined.

Diesel Fuel Injection/Mixing FacilityIn FY 2004, we completed the construction andsafety review of the fuel-injection/mixing testfacility. Figure 1 shows the schematic diagram ofthe facility, which consists of a fuel-injectionassembly, simulated exhaust gas generating system,and fuel-exhaust-gas mixing apparatus. Thefacility’s injector system is a commercially availableG-2 fuel injector provided by the International

Engine and Truck Company (ITEC). The injector iscontrolled by the ITEC’s PCM (Powertrain ControlModule) simulator and can be operated under twoinjection modes: single shot and pulsed injection.The pulsing rate varies from 10 to 70 Hz, and theinjection duration is typically between 1 ms and20 ms, giving a maximum fuel injection volume ofabout 8 cm3/s.

ResultsThe injector-controlling software provided by ITECwas installed in a desktop PC and run on theLabview platform. Injection tests were conducted,and a CCD camera (Canon PowerShot A80) that hasa resolution of 72 pixels/in. when operated under thedynamic mode was used to capture the sprayappearance of diesel fuel. Figure 2 shows a pictureof the spray structure generated with 7-MPainjection pressure into ambient conditions. Theestimated fuel flow rate under the single-shot modeof injection is 30 mm3/s. The figure shows primarilya two-dimensional (2-D) image of the generalstructure of the dense spray region. The stable fueljet extends to about 200 times the nozzle diameter(~ 0.18 mm). On the basis of the reported data [1],this finding indicates that the liquid volume fractionat the boundary of the dense spray region is less than10%. Detailed characteristics of the diffuse phasecannot be resolved with the present CCD camera.

Future PlanIn FY 2005, we will focus on development of asteam/air injection device that can deliver controlledvolumes of steam and hot air into the mixingchamber. The state of fuel/steam/air mixing will bemonitored with a 2-D array of temperature sensors.Tests will be conducted at different air temperature,fuel injection rate, and steam concentration. Resultsfrom the tests will be used to determine the optimalconditions for producing a cool flame from thediesel fuel.


129

Reference1. G.A. Ruff and G.M. Faeth, “Nonintrusive

measurement of the structure of dense sprays,”Progress in Astronautics and Aeronautics, Vol.166, “Recent Advances in Spray Combustion:Spray Atomization and Drop BurningPhenomena, Vol. 1,” Chapter 11, 1996.

SapphireWindow

HPN2

Tank

PressureRegulatorPressure

Multiplier

FuelFilter

HumiditySensor

FuelInjector

Pre-heated

Air

CompressedAir

T

P T H

Fuel Heater

PC

P T H Data

ValveControl

Data

MixingChamber

HValve

InjectorControlModule

EngineOil

High PressureAccumulator

LPN2

Tank

CoolingAir

Oil Filter

DisposalTank

Blower

Exhaust GasOutlet

6”-4” ReducerReformerChamber

Exhaust Gasfrom Hood

Valve

WaterReservior

SolenoidValve

DieselFuel

MistTrap

Figure 1. Diesel fuel injection and mixing facility.

Figure 2. A typical spray structure of G-2 injector.


130

B. On-Board Plasmatron Hydrogen Production for Improved Vehicle Efficiency

Principal Investigators: Daniel R. Cohn, Leslie BrombergPlasma Science and Fusion Center, Massachusetts Institute of TechnologyMIT PSFC NW16-104, 77 Massachusetts Avenue, Cambridge MA 02139(617) 253-5524, fax: ( 617) 253-0700, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-803, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Philip S. Sklad(865) 574-5069, fax: (865) 576-4963, e-mail: [email protected]

Contractor: Massachusetts Institute of TechnologyContract No.: DE-AC03-99EE50565

Objectives• Develop attractive applications of plasmatron fuel reformer technology to internal combustion engine vehicles

using diesel, gasoline, and biofuels.

• Develop means of reducing NOx and particulate matter emissions from diesel engines by using plasmatronreformer generated hydrogen-rich gas.

• Investigate conversion of ethanol and bio-oils into hydrogen-rich gas.

• Develop concepts for the use of plasmatron fuel reformers for enablement of HCCI engines.

Approach• Optimization of plasmatron fuel reformer configurations, including gas and fuel management, plasma

geometry, and reactor chamber variations.

• Optimization of catalytic and other materials-enhanced plasmatron reforming.

• Determination of reforming characteristics of pure and hydrated ethanol, refined and unrefined bio-oils, andwith commercial-grade diesel fuel.

• Development of instrumentation for investigating fast transients when using the plasmatron in a pulsedconfiguration, as would be the case of NOx trap regeneration.

• Modeling of plasmatron fuel reformer operation by using computational studies of the chemistry of simplegaseous hydrocarbons (mainly methane) because of the complicated effects from liquid fuels and CFDcalculations of the flows upstream from the plasma discharge.

• Test of plasmatron fuel-reformer-enhanced NOx regeneration both in laboratory and on vehicles.

Accomplishments• Demonstrated efficient conversion of biofuels to hydrogen, CO, and light hydrocarbons by using non-catalytic

reforming, opening new potential applications of these fuels in the transportation sector.

• Demonstrated rapid response and high hydrogen yields for ethanol reforming.

• Demonstrated increased hydrogen production by using Ni catalysts and Rh catalysts on an alumina substrate.

• Demonstrated increased hydrogen production by using metallic plates in the reactor.

• Developed an improved method for the use of mass spectrometer for the quantitative measurement of flow ratesof gases of interest.


131

• Demonstrated hydrogen-rich gas as an effective reductant for converting NOx in NOx traps to N2, in vehiculartests carried out by our industrial partner, ArvinMeritor.

• Showed that hydrogen-rich gas significantly improves NOx trap catalyst operating temperature range:– Demonstrated decreased fuel penalty for diesel NOx emissions exhaust treatment and– Demonstrated advantages of hydrogen-rich gas from plasmatron fuel reformers for desulfation of NOx

traps.

• Developed a power supply and an electrical configuration for the elimination of electromagnetic interference(EMI), which had affected measurements during tests in laboratory.

• Continued collaboration with ArvinMeritor with the goal of commercialization of plasmatron fuel reformertechnology.

• Developed additional MIT intellectual property and filed patent applications; also, some previously filed patentapplications have been granted in 2004.

Future Directions• Demonstrate through experiments and modeling increased capability of plasmatron fuel reformers for higher

flow rates (4 g/s of fuel) and applications in on-line regeneration of NOx traps and lean-NOx catalysts.

• Investigate new approaches to increase hydrogen yield for enhanced reforming from metallic materials.

• Investigate new approaches of thermal control of catalyst in plasmatron fuel reformers.

• Investigate operation of plasmatron reformers over a range of oxygen-to-carbon ratios (O/C), from O/C ~1(stoichiometric partial oxidation) to full combustion to optimize regeneration of diesel particulate filters.

• Investigate use of pulsed operation of plasmatron reformer to obtain variable flow rate.

• Initial tests of plasmatron fuel converter regeneration of diesel particulate filter.

• Investigate concept of hydrogen regeneration of SCR catalyst.

IntroductionTo meet stringent U.S. emissions regulations thatwill be implemented in the 2007–2010 diesel vehiclemodel years, new after-treatment technology isbeing developed. Both particulate and NOxemissions after-treatment technology will benecessary, as it appears that engine in-cylindertechniques alone will be unable to meet theregulations.

MIT, in collaboration with ArvinMeritor, has beendeveloping a plasmatron fuel reformer for vehicularon-board applications, including engine fuelconditioning as well as diesel engine after-treatment.The MIT technology has been licensed byArvinMeritor. The DOE program is aimed atrealizing the broad potential of plasmatron fuelreformer technology for hydrogen-based dieselengine exhaust after-treatment and for expandingopportunities for use of renewable energy biomassderived fuels.

Plasmatron fuel reformer technology has thepotential to be an important enabling technology fora range of practical vehicular applications. It canprovide unique capability for reforming diesel andbio-oils into hydrogen-rich gas for vehicularapplications. Plasmatron hydrogen generationtechnology has the potential to play a major role inthe development of practical NOx and dieselparticulate after-treatment technology needed tomeet the 2010 EPA mandate for emissionsreductions. Recent concerns about the possiblecontribution of diesel soot to global warmingprovide additional impetus for development of thesetechnologies.

The potential value of the hydrogen-based vehicleenhancement has been validated by successfulvehicular tests of diesel exhaust after-treatment byusing plasmatron hydrogen-enhanced lean-NOx traptechnology in a bus and pickup truck byArvinMeritor, our industrial collaborator.ArvinMeritor, a $7-billion/yr U.S. automobile and


132

truck components manufacturer, has licensed theplasmatron fuel reformer technology from MIT.

Plasmatron Fuel ReformerPlasmatron fuel reformer technology consists in theuse of a special plasma discharge for the initiation ofreformation process. Proper placement of the plasmadischarge is used in a flowing air/fuel mixture that isvery rich. Proper preparation of the fuel and fuel/airmixture is required for optimal operation.

The plasmatron fuel reformer technology isattractive because of capability for fast response,elimination or relaxation of catalyst requirements,and fuel flexibility. Fast response is needed for on-board applications in SI engines, where on-demand,wide-dynamic-range hydrogen-rich gasrequirements are stringent, as well as inaftertreatment applications, where short bursts ofhydrogen-rich gas are needed. Moderate hydrogenyields (defined as the ratio between the hydrogengas in the reformate to the hydrogen in thehydrocarbon fuel) have been achieved without theuse of catalysts (about 40%), The yield increases to~70% when hydrogen in light hydrocarbons is alsoincluded. The use of a catalyst downstream from theplasma region can increase the hydrogen yield to~70%, as discussed below. Experiments have beenconducted with gasoline, diesel, ethanol and hard-to-reform bio-oils.

To increase hydrogen yield, a catalyst can be placeddownstream from the homogeneous reformingregion. The homogeneous stage of the plasmatronfuel reformer converts the liquid hydrocarbons intohydrogen, CO, and light hydrocarbons, plus a smallamount of carbon dioxide and water. The catalyst isuseful in converting light hydrocarbons intoadditional hydrogen.

Diesel ReformingIn 2004, the input fuel flow rate of a dieselplasmatron fuel reformer at MIT was operated withan increased diesel fuel flow rate of 2 g/s,corresponding to 80 kW of fuel reformate. Thisreformer produced about 1.5 L/s of hydrogenwithout the use of a reforming catalyst for additionalhydrogen generation. This type of higher-flow-rateoperation can facilitate quick regeneration of single-leg NOx trap systems.

The characteristics of the plasmatron diesel fuelconverter and the reformate it produces are shown inTable 1. The goal of reformation of diesel by theplasmatron fuel converter is the conversion of theheavy diesel compounds into hydrogen, carbonmonoxide, and light hydrocarbons for use in after-treatment applications. The diesel reformateproduced by the plasmatron fuel converter has littleresidual oxygen and little or no soot. The dieselplasmatron can be cycled on and off quickly(~3–5 s). For diesel after-treatment applications, thedynamic performance of the plasmatron fuelreformer is important because of the low duty cycleof operation. For NOx catalyst regenerationapplications, the plasmatron may be turned on forseveral seconds with a duty cycle of ~10%. Fordiesel particulate filter (DPF) applications, theplasmatron could well be operated for a few minutesevery few hours, resulting in much smaller dutycycles.

Hydrogen is a powerful reductant, and its use for theregeneration of NOx traps has been researched incatalyst development laboratories. The goal of theplasmatron reformer program is to determine theadvantages of hydrogen-assisted NOx trapregeneration, by using the plasmatron fuel reformeras the source of on-board hydrogen. Tests atArvinMeritor have demonstrated the advantages ofusing hydrogen-rich gas for NOx regeneration anddesulfation of NOx traps. Substantially higherregeneration of NOx traps with hydrogen can beobtained with hydrogen-rich gas. In contrast to theuse of diesel fuel for regeneration, adequateregeneration can be obtained at low exhausttemperatures (in the 150–200°C range), which cansignificantly increase the range of vehicle conditionsin which the technology can be used. In addition, alower fuel penalty is observed when comparinghydrogen-rich gas regeneration to diesel fuelregeneration. The temperatures required fordesulfation are also substantially decreased.


133

Table 1. Experimental results of homogeneous dieselreforming.

Parameter ValueElectric power (W) 200O/C 1.2Diesel flow rate (g/s) 0.8

Corresponding chemical power (kW) 36

Reformate composition (vol %)H2 7.6O2 1.3N2 64.0CH4 2.4CO 13.0CO2 4.4C2H4 2.2C2H2 0.0H2O 7.1

Energy efficiency to hydrogen, CO andlight HC (%)

65

H2 flow rate (L/min [STP]) 20Soot (%; opacity meter) 0.0

Enhanced Reforming with Metallic MaterialsUse of conventional catalysts can be hampered bylimits on reliability and lifetime. We have studiedthe use of metallic materials for enhanced reformingas an alternative to conventional catalysts.Reforming experiments of diesel and ethanol havebeen carried out by using a bed of metallic spheresof different materials and sizes in the reactor. Theseexperiments resulted in a hydrogen concentration ofup to 18% in the reformate gas (in contrast to 8%without the metallic material) and a high energyefficiency. However, none of the spheres used lastedvery long, as the material deteriorated and turnedinto a metallic dust. On the basis of these results,ceramic (alumina) bids have been tested to counterthe destruction effect of heat. The results yielded alow hydrogen concentration in the reformate,indicating that the reforming phenomenon onsurfaces could not be linked only to surface area andheat.

On the basis of the same idea of using non-conventional materials for homogeneous reforming,metallic plates of different forms, shapes, and

materials have been used. Experiments using steelplates yielded a hydrogen concentration of 18%, butthe plates deteriorated just as in the case of thespheres. Inconel plates have also been tested andwere also shown to deteriorate. Finally, nickel platesshowed the most promising results: the hydrogenconcentration obtained with these plates was 17%,and the material did deteriorate much less than othermaterials.

Instrumentation DevelopmentA new approach was developed that allowedquantitative fast measurements of flow rates. Sincethe flow rates of air are known (from the inputs),and air contains substantial amounts of Ar (about1%), the flow rate of Ar in the reformate is known.By taking ratios between the signals of the gases ofinterest and Ar, the flow rates of the gases of interestcan be determined. This avoids problems of absolutecalibration of the mass spectrometer, which issubject to variations and drifts due to temperatureeffects, partial blockages of the capillary, and otherenvironmental effects. The setup can be used fordetermining flow rates with appropriate choice of Arconcentration in the calibration mixture.Alternatively, the N+ ion can be used instead of Ar,but our preliminary results indicate that lower noiseis obtained with Ar as the benchmark. The devicewas optimized for making measurements of 10 ionsevery 200 ms. A technical paper is being preparedthat describes this approach.

Reformation of BiofuelsReformation experiments of renewable energybiomass derived fuels have been carried out. Ethanoland vegetable oils, (including soy and canola oils)have been efficiently reformed, with no sootproduction at input rates up to 1 g/s, correspondingto about 34 kW of heating power. Both non-catalyticand catalytic (with a catalyst downstream from ahomogeneous reforming zone) have been studied.On-board conversion of biofuels into hydrogen-richgas opens up a range of opportunities for reducingpetroleum consumption.

Bio-OilsDependence of the homogeneous reforming processon the ratio of oxygen to carbon (O/C ratio) in thereagents indicates almost constant yield over a broad


134

range of O/C ratios. As the O/C is increased, thetemperature of the reformate increases as a result ofincreased exothermicity of the reaction,compensating for the fact that a fraction of the fuelis converted into hydrogen.

The hydrogen yield for the reformation of corn andsoybean oils in the presence of a catalyst is higher,as is the conversion, than that for homogeneousreforming, with little difference between the fuels.In this case, there is a broad maximum at O/C ratiosclose to 1.5. The catalyst is useful in converting lighthydrocarbons into additional hydrogen. The cornand soybean catalytic experiments were done with avolume of 85 cm3 of catalyst, whereas only 25 cm3

of catalyst was used during the canola oilexperiments (to explore the effect of space velocity).A nickel catalyst was used.

Studies of effect of catalyst on the hydrogen yieldindicate that for homogeneous reformation ofsoybean oil, the hydrogen yields are about 30%,while the yield increases to about 70% in thepresence of a catalyst. The difference between theyield of 70% and the ideal yield of 100% is dueprimarily to the loss of hydrogen resulting from fulloxidation into water. For homogeneous reformation,the ratio of the heating value of the hydrogen, CO,and light hydrocarbons by-products to the heatingvalue of the fuel was typically between 60 and 70%.

The non-catalytic investigations indicate that theplasmatron converts the biofuels into hydrogen,carbon monoxide, and light hydrocarbons, withminimal soot production, and virtually eliminates allfree oxygen. The catalyst then takes the oxygen-freeplasmatron gas and roughly doubles the hydrogenyield, performing CO2 and possibly steam reforming(the CO2 concentration decreases downstream thecatalyst; water is not monitored).

It should be stressed that the performance of thesystem was not optimized, and higher hydrogenyields could be possible by converting thesubstantial amounts of C2s present in the gasdownstream from the catalyst.

For all of the experiments performed with veggieoils, the opacity was below the sensitivity limit ofthe instrumentation.

EthanolTwo sets of experiments have been conducted withethanol. The first set of homogeneous and catalyticexperiments used the same experimental protocol asdescribed for the bio-oils. The maximum hydrogenyield obtained for catalytic reforming of a fuel flowrate of 0.35 g/s was 75%, as shown in Figure 1.From this figure, it appears that the hydrogen yielddepends upon the O/C ratio and the maximum valueis obtained for an optimum O/C of 1.6.

0.00

0.20

0.40

0.60

0.80

1.00

1.00 1.20 1.40 1.60 1.80 2.00O/C ratio

H2

yiel

d

Figure 1. Hydrogen yield for the plasma-catalyticreformation of ethanol vs. O/C ratio.

Other experiments have also shown that thehydrogen yield depends upon the fuel flow ratethrough the reactor. Figure 2 gives the hydrogenyield for catalytic reforming of ethanol versus thefuel flow rate. It shows that the hydrogen yieldincreases with decreased flow rate. This indicatesclearly that higher hydrogen yields for the catalyticreforming of ethanol are possible for lower spacevelocities. The space velocity is defined as the ratioof the airflow rate at STP that passes through thecatalyst to the volume of this catalyst.

0.00

0.20

0.40

0.60

0.80

1.00

0.00 0.20 0.40 0.60 0.80 1.00Ethanol flow rate, g/s

H2

yiel

d

Plasma 120W

Plasma 270W

Figure 2. Hydrogen yield vs. ethanol flow rate forplasma-catalytic reforming.


135

The second set of catalytic experiments wasconducted by using rhodium catalyst deposited onalumina cylindrical substrate beads. The beads were2 mm long by 1 mm in diameter, and the rhodiumcontent was 0.5% by weight. The catalyticexperiments were performed by using 50 g ofcatalyst.

Plasmatron cold start, with the reactor initially atroom temperature, was investigated for ethanolreforming by using the mass spectrometer describedabove.

Figure 3 shows the concentration (vol. %) ofhydrogen, carbon monoxide, ethane, and methanefor the case of homogeneous reforming of ethanol.The hydrogen concentration is about 6% right atstart, and it increases slowly to about 12% (in 150 s).These results have been obtained by using adischarge power of 200 W with an ethanol flow rateof 0.9 g/s, corresponding to 26 kW of heating power.The O/C was ramped from an initial value of 2.3 atstart-up to 1.73 after 12 s of reforming. It should benoted that at O/C ~1.73, a good amount of ethanol iscombusted, and some of the hydrogen is turned intowater. The hydrogen yield could be increased andresponse time could be decreased by optimizing theO/C start-up ramp.

Combined homogeneous and catalytic reformingwas explored as means of increasing the hydrogenyield. The electrical power in the discharge is thesame as that for homogeneous reforming, as is theethanol flow rate. The ramp of the O/C ratio,however, is slightly different, as the final O/C ratio

Figure 3. Concentration of H2, CO, and C2H6 in thereformate as a function of time for homogeneousreforming, with variable O/C ratio during the first 12 s.

in the case of catalytic reforming is lower than thatfor homogeneous reforming.

In the case of plasma catalytic reforming, the O/Cratio is ramped from O/C ~ 2.3 at start-up to O/C~ 1.56 at 12 s. The hydrogen concentration andyields are higher than in the case of homogeneousreforming, even at start-up, with an initial hydrogenconcentration of 10%. In addition, the steady-statevalue of the hydrogen concentration and yields aresubstantially higher than in the case of homogeneousreforming. In addition to higher yields, the time toachieve high conversion is significantly shorter thanin the homogeneous case, reaching near steady stateat the end of the 12-s ramp-up. The energyefficiency (ratio of the heating value of thereformate to the heating value of the ethanol) isgreater than 70% at steady state.

DPF RegenerationOptions of using reformate from a plasmatron fuelreformer for the controlled regeneration of DPFs arebeing explored.

Hydrogen-rich gas could provide importantadvantages for controlled regeneration of DPF. Thereformate can provide clean-burning fuel to initiateand control the burnup of soot in a trap duringregeneration. Alternatively, the use of hydrogen-richgas may result in DPF regeneration at lower sootlevels. This regeneration scheme could allow morefrequent regenerations, the advantage of whichwould be a decreased chance of large uncontrolledregeneration thermal excursion that could negativelyimpact the performance of the trap, thus resulting inincreased reliability and longevity.

Active techniques for the regeneration of DPF use aburner to achieve the temperatures at the catalystrequired for filter regeneration. This technologyburns fuel to generate heat, carbon dioxide, andwater. Alternatively, plasmatron technology couldbe used not only as an igniter to combust part of thefuel using the oxygen present in the diesel exhaustand generate heat, but also as a means to producehydrogen-rich gas. The fuel combustion can takeplace even under conditions where the free-oxygen-to-fuel ratio is larger than unity (i.e., lean-burnconditions in the plasmatron).


136

The advantage of combusting the fuel in the exhaustis that it is possible to provide the heat forregeneration while minimizing the amount of extraair introduced into the system. The hydrogen-richgas with exhaust gas as oxidizer results in a netdecrease of the amount of free oxygen in theexhaust, decreasing the combustion rate of the sooton the DPF and decreasing the possibility ofuncontrolled regeneration. This technology is alsoattractive because the plasmatron hardware mayeventually already exist on-board the vehicle forregeneration of NOx absorber traps.

The plasmatron fuel reformer can thus be used as apowerful combustor, operating with exhaust as theoxidizer. There is thermal output from thecombustor, but no hydrogen or light hydrocarbons.Alternatively, the reformer operates underconditions close to partial oxidation (free-oxygen-to-carbon ratio of 1), producing hydrogen, CO, lighthydrocarbons, and limited thermal output (~ 1/5 ofthe thermal output is the fuel is combusted). Thesetwo are extremes, with the plasmatron operatinganywhere in between, generating both easilycombustible gases and thermal output.

Reformate requirements for achieving DPFregeneration have been calculated on the basis ofsimple models. It has been determined that the flowrates requirements for a 6-L turbocharged engine arewithin the range of today’s plasmatron capabilities.Larger engines can be regenerated by using(1) multiple plasmatrons or (2) thermalmanagement/non-uniform regeneration techniques.

It has been determined that the hydrogen-rich gaswill not spontaneously combust before reaching thesoot trap under normal operating conditions of theengine. Thus, slow, uniform regeneration of the DPFcould be achieved by using both the thermal effectfrom the reformate and the reducing capabilities andlocalized thermal effect of the hydrogen rich gas.

Industrial CollaborationMIT is collaborating with ArvinMeritor indeveloping the plasmatron fuel reformerregeneration of NOx absorber catalysts as a means ofcontrolling emissions from trucks, buses, and onvehicles. MIT has licensed plasmatron fuel reformertechnology to ArvinMeritor for their applications.

The intellectual property includes both thetechnology relevant for plasmatron fuel reformersoperating on diesel fuel, as well as system patentswhere plasma-based reformers are used in theregeneration process.

ArvinMeritor has studied the implications ofhydrogen-rich gas regeneration of NOx absorbercatalysts in the laboratory, by using hydrogen-richgas from bottled gas. These tests demonstrated theusefulness of using hydrogen-rich gas instead ofdiesel fuel for regenerating the catalysts. Theadvantages were both in decreased fuel penalty atmoderate temperatures and in enabling regenerationat lower temperatures (where diesel fuel isineffective). After these experiments in whichsynthetic reformate was used, the systems wereplaced on vehicles to investigate issues ofintegration and packaging. Successful tests havebeen performed on a bus. ArvinMeritor is alsodeveloping plasmatron fuel reformer technology forincreasing the efficiency of gasoline engines.

The substantial ArvinMeritor investment inplasmatron fuel reformer technology represents astrong leveraging of a much smaller DOEinvestment. The integrated DOE investment over thetime that is has supported the MIT plasmatronprogram has already led to an ArvinMeritorinvestment that is more than five times greater.Further strong interaction between MIT andArvinMeritor is anticipated.

SummaryHighlights during FY 2004 are:• On-board testing of plasma-fuel-reformer-based

hydrogen-rich gas regeneration of NOx traps hasbeen carried out in a bus at ArvinMeritor’s testtrack.

• Significant benefits (relative to diesel fuelregeneration) have been demonstrated, includingdecreased fuel penalty and lower regenerationtemperature.

• Desulfation of traps by using hydrogen-rich gasfrom a plasmatron diesel reformer, with lowertemperature regeneration, has beendemonstrated by ArvinMeritor.

• Fast turn-on reformation from diesel andethanol, with instantaneous (< 1 s) yield of 32%to hydrogen and light hydrocarbons and 20% to


137

hydrogen, has been demonstrated by usinghomogeneous reforming, facilitated by fuelvaporization, enlarged volume reactioninitiation, and enthalpy addition.

• High yields into hydrogen and lighthydrocarbons were obtained in homogeneousreforming of ethanol and bio-oils.

Conclusions• On-board generation of H2-rich gas can provide

important benefits for diesel engine NOx exhaustaftertreatment.

• Plasmatron fuel reformer is a promisingtechnology for practical on-board generation ofhydrogen-rich gas for other applications wherediesel and gasoline are used.

• Plasma fuel reformers can facilitate increaseduse of biofuels

Patents and PublicationsFour patents were issued since May 2003:

• US Patent 6793899: Plasmatron-CatalystSystem, L. Bromberg, D.R. Cohn,A. Rabinovich, and N. Alexeev, issuedSeptember 21, 2004.

• US Patent 6718753: L. Bromberg, D.R. Cohn,and A. Rabinovich, Emission Abatement SystemUtilizing Particulate Trap, issued April 13,2004.

• US Patent 6655324: High Compression Ratio,Hydrogen Enhanced Gasoline Engine System,D.R. Cohn, L. Bromberg, A. Rabinovich, andJ.B. Heywood, issued December 2, 2003.

• US patent 6560958 Emission Abatement System,L. Bromberg, D.R. Cohn, and A. Rabinovich,issued May 13, 2003.

Publications include:• L. Bromberg, D.R. Cohn, K. Hadidi,

J.B. Heywood, and A. Rabinovich, PlasmatronFuel Reformer Development and InternalCombustion Engine Vehicle Applications,presented at the Diesel Engine EmissionReduction (DEER) Workshop, 2004, CoronadoCA, August 29–September 2, 2004.

• K. Hadidi, L. Bromberg, D.R. Cohn,A. Rabinovich, Plasma Catalytic Reforming ofBiofuels, report PSFC-03-28;http://www.psfc.mit.edu/library/03ja/03JA028/03JA028_abs.html; presented at World EnergyConference, Denver, CO, August 2004.


138

V. EM Regenerative Shocks

EM Shock Absorber

Principal Investigator: John R. HullArgonne National LaboratoryArgonne, Illinois 60439(630) 252-8580, fax: (630) 252-5568, e-mail: [email protected]



Objectives• Develop model to predict performance of regenerative shock absorber.

• Conduct experiment on a small-scale vehicle to verify the energy recovery from road-induced vehicle motion.

• Compare the experimental results with theoretical prediction.

• Develop road data to use as input to the model.

Approach• Mount regenerative shock on small vehicle and accrue test data.

• Instrument truck and run it over local roads to obtain data.

Accomplishments• Tested experimental electromagnetic shocks on all-terrain vehicle and used the data from it to verify model of

regenerative shock performance.

• Instrumented truck and accumulated data on acceleration over local roads.

Future Direction• More road data need to be accumulated for a complete database.

• The road data need to be used in the regenerative shock model to predict expected performance.

• Explore potential of developing technologies for dynamic control of vehicle.

IntroductionThe up-and-down motion of a vehicle traversing aroad is unwanted from a passenger comfortperspective. In present vehicles, where the motion isdamped by shock absorbers, the energy provided by

the prime power plant is diverted from its intendedpurpose and gets dissipated in heat. Prevention ofthis motion in the first place could be accomplishedby construction and maintenance of smooth roads.However, in the absence of such a societal program,the energy diverted into the vertical motion can


139

potentially be recovered by a regenerative shockabsorber that converts the up and down motion intoelectrical energy. The goal of this project is toascertain whether this can be accomplished in a cost-effective manner.

To estimate the power generated by the EM shocks,experimental regenerative shocks were mounted inan ATV, as shown in Figure 1. This vehicle wastested on local roads and on a specially constructedobstacle course to determine the performance of theexperimental shock absorber.

Figure 1. ATV used for EM shock testing.

In a separate series of experiments, the performanceof experimental regenerative shocks has beenobtained by subjecting them to forced oscillations ofa shaker, as shown in Figure 2.

Figure 2. Shaker testing of Mark 1 EM shock.

ResultsBy using data from both experimental programs, amathematical model of regenerative shockperformance was established and verified. Sampledata used in the model are shown in Figure 3. In thisinstance, the ATV was driven over a 4-by-4 beam.

To use the model to predict actual performanceunder real conditions, it is necessary to have theexpected forcing function of a vehicle traveling on

Figure 3. Response of Mark 2 (rotary) shock absorber transversing 4-by-4 beam.


140

real roads. Examination of existing databasessuggested that the required data were not sufficient.Therefore, we instrumented a road vehicle to acquirethe data.

DiscussionAn LVDT was mounted in a dump truck betweenthe axle and the chassis, as shown in Figure 4. Thepurpose of this experiment was to record thesuspension (conventional) relative displacement toget an estimate of the available energy.

Figure 4. LVDT mounted in the trunk.

The voltage output from the LVDT was sent to atape deck on board the truck cabin (as shown inFigure 5) as the truck was driven around a roadinside Argonne National Laboratory.

Figure 5. Tape recorder inside the trunk.

The tape gain was set to 2 so that the actual voltageis twice the voltage shown in the graph (i.e., 2-Vpeak in the graph implies 4-V peak signal fromLVDT). A segment of the time domain signalrecorded is shown in Figure 6. The LVDT scalefactor was 3.3 V/in. and the frequency spectrum ofthe above signal is shown in Figure 7.

From Figure 7, a 0.374 mv at 4 Hz implies 0.226 inpeak displacement at 4 Hz, which is approximately5.6 in./s or 0.142 m/s. Since the actual value of thesuspension damping constant of the dump truck isunavailable, a suspension damping constant of 89kN•s/m, which may be typical of a single-axle flatbed truck, is assumed. This leads to power of 900 Wat 4 Hz. However, if we consider the time domainsignal of Figure 6 (which accounts for contributionfrom all frequencies), it seems maximum velocity(after considering a 2:1 factor of tape and an LVDTscale factor of 3.3 V/in.) is approximately 13.3 in./s(0.338 m/s). Again assuming the same dampingconstant, peak power is approximately 5 kW.However, it may be cautioned that this 5-kWcalculation is for peak power (when the truck hits apot hole) and not average power, and so designingan electromagnetic shock with such a dampingconstant may not be easy.


141

Figure 6. Time domain signal (with 2:1 reduction).

Figure 7. Frequency spectrum of the signal.


142

VI. Joining Carbon Composites

Innovative Structural and Joining Concepts for Lightweight Design of HeavyVehicle Systems

Principal Investigators: J. Prucz and S. ShoukryDepartment of Mechanical and Aerospace EngineeringWest Virginia UniversityMorgantown, WV 26506-6106(304) 293-3111 ext. 2314, fax: (304) 293-8823, e-mail: [email protected]

Investigators: G. William, T. Damiani, T. Evans, P. ShankaranarayanaWest Virginia UniversityMorgantown, WV 26506-6106(304) 293-3111, fax: (304) 293-6689


Contractor: West Virginia UniversityContract No.: EE50692

Objectives• Devise and evaluate lightweight structural configurations for heavy vehicles.

• Study the feasibility of using Metal Matrix Composites (MMC) for critical structural components and joints inheavy vehicles.

• Develop analysis tools, methods, and validated test data for comparative assessments of innovative design andjoining concepts.

• Develop analytical models and software for durability predictions of typical heavy vehicle components made ofparticulate MMC or fiber-reinforced composites.

Approach• Devise integrated design concepts for critical structural components of van trailers.

• Develop innovative, fastener-less joints for composite and dissimilar materials.

• Prototype and evaluate experimentally new composite design and joining concepts.

• Model damage effects at the mesoscale level to predict the durability of composite materials in heavy vehiclestructures.

Accomplishments• Alternative structural arrangements for the floor of a heavy van trailer have been devised and analyzed, leading

to the conclusion that sandwich panels allow minimum weight designs for a variety of core configurations.


143

• Alternative material selections have been considered for the structural floor of a heavy van trailer, leading to theconclusion that carbon-carbon composites enable the greatest weight savings, depending on the specific floordesign configuration.

• Alternative fastener-less joining methods between sandwich panels in various sections of a typical van trailerstructure have been devised and evaluated, both through theoretical modeling and actual prototyping.

• Double-lap-bolted joints made of different MMC materials have been evaluated both through standard ASTMtests and failure simulations based on finite-element analysis, leading to the conclusion that such joints arelikely to fail early in the bearing mode, because of the brittle behavior of MMC materials.

Future Directions• Develop and validate integrated, minimum-weight design concepts of full-scale structural assemblies for van

trailers, based on sandwich panels with optimized face sheet and core configurations.

• Build a full-scale baseline prototype of a lightweight van trailer for extensive instrumentation and testing inrealistic operation scenarios, as well as for showcasing the new concepts to producers and operators of heavyvehicles.

• Establish a knowledgebase for reliable durability predictions in terms of either fatigue life or damage toleranceof composite materials, for decision support in both the design and repair of structural components in heavyvehicle systems.

IntroductionThis report summarizes the second year of a projectinitiated at West Virginia University (WVU) toinvestigate practical ways of reducing the structuralweight and increasing the durability of heavyvehicles, through judicious and innovative use oflightweight composite materials. While the projectwas initially focused on specific components in atrailer structure and on a specific Aluminum/Silicone Carbide (Al/SiC) MMC, namely the“LANXIDE” material [1], the second-year effortwas expanded from the component to the systemlevel and from MMC to other composite materialsystems. Broadening the scope of this research iswarranted not only by the structural and economicaldeficiencies of the “LANXIDE” MMC material, asoutlined in the previous report [1], but also by thestrong coupling that exists between the material andthe geometric characteristics of the structure. Suchcoupling requires a truly integrated design approach,focused on the heaviest sections of a van trailer. It isobvious that the lightweight design methodsdeveloped through this work will not beimplemented by the commercial industry unless theweight savings are, indeed, impressive and areproven to be economically beneficial in the contextof Life Cycle Costs (LCC).

Integrated Design Concepts for CriticalStructural ComponentsThe chassis assembly contributes about 73% of theoverall weight of a typical van trailer (15,100 lb fora 48-ft-long trailer), of which 47% is contributed bythe oak floor panels and the cross beams that supportthe floor [1].

The baseline loading scenario assumed forintegrated structural design of the chassiscomponents consists of structural dead loads and thelive load applied by a moving loaded forklift [1].The corresponding baseline minimum weight designof a structural floor built on steel I-cross beams wasevaluated in terms of its factor of safety bycalculating the stress and deformation fieldsassociated with such loading.

Alternative design concepts for the structural floorof a van trailer were devised to reduce its weightbelow that of the current baseline configuration. Allof these lightweight designs rely on sandwich panelswith various material and geometric characteristicsof the core. The main design objective stated fortheir comparative evaluation is an optimal trade-offbetween the overall weight and stiffness of the floor.The alternative floor design configurations have tosatisfy the following design criteria:


144

• The factor of safety in flexure must not be lowerthan 2.0.

• The free-edge deflection of any cross beam mustnot exceed that calculated for a similar steelbeam currently used in the baseline floorconfiguration.

Alternative 1: The core of the floor consists in thiscase of I-cross beams made of various candidatematerials, spaced at a distance of 1 ft apart, andconnected by two fiberglass “bearing” bars“running” along the trailer through holes drilled inthe web sections of the cross-bars (Figure 1).

96 in.Plan View

Section Side View A-A

12 in. 12 in.

bf

d

tf

tw

Figure 1. I-cross beam flooring.

For each alternative material selection for the I-crossbeams, Table 1 specifies the minimum standarddimensions required for the I-beam cross-section inorder to meet the design criteria outlined above. Inaddition, Table 1 displays the factors of safetycorresponding to these dimensions, the edgedeflection, and the weight of the unit floor areacorresponding to every material option for theI-cross beams.

The results presented in Table 1 reveal that thecurrent weight of a baseline van floor can be reducedby as much as 69% or 66% when the steel I-crossbeams are replaced, through an integrated designapproach, by I cross-beams made of carbon-carboncomposite or magnesium alloy, respectively. Theseresults demonstrate the drastic reductions instructural weight that can be achieved throughrational applications of lightweight materials inheavy vehicles that integrate the layout andgeometric design with the material selection process.This conclusion is further supported by similarstudies on three other alternative design concepts forthe trailer floor, as follows:

• Alternative 2: Sandwich panel consisting of topand bottom fiberglass faceplates and a coreformed of transverse C-channel cross beams, asshown in Figure 2-a.

• Alternative 3: Sandwich panel built of ribbedfiberglass faceplates with a core consisting ofhollow cross tubes of either rectangular orcircular cross-section, as shown in Figure 2-b.

• Alternative 4: Floor constructed from sandwichpanel with a homogeneous, lightweight core.

Table 1. Alternative material solution for I-beam floor.

SIZEd bf tw tf

Material in. in. in. in.

UltimateStrength

(ksi)

Factorof

Safety

Max. EdgeDeflection

(in.)

UnitWeight(Lb/ft3)

Weight(lb/ft2)

STEEL 4 2.5 0.16 0.25 80 4.38 4 2.5 0.16Aluminum 8 2.25 0.13 0.19 35 3.38 8 2.25 0.13EXTREN 525 8 4 0.38 0.38 30 8.70 8 4 0.38Carbon-Carbon 6 3 0.25 0.25 155 17.78 6 3 0.25Nitronic 19D Stainless Steel 4 2.5 0.16 0.25 103.6 5.69 4 2.5 0.16Nitronic 60 Stainless Steel 4 2.5 0.16 0.25 160 8.84 4 2.5 0.16Nitronic 30 Stainless Steel 4 2.5 0.16 0.25 117 6.46 4 2.5 0.16Magnesium 6 2.5 0.16 0.25 26.8 3.0 6 2.5 0.16


145

12 “ 12” 12”

Face Plate

C-Channel @1 ft spacing

a. C-Channel Floor

Tube Sections @1 ft spacing

Fiberglass Face Plate

b. Tube Core FloorFigure 2. Alternative floor designs.

The results of minimum weight, integrated designstudies for all of the above four alternative sandwichpanel configurations of the trailer floor aresummarized in Table 2 for eight different materialselections for the core of the panel. Both themaximum deflection and the minimum weight perunit area shown in Table 2 for every design optionconsidered here meet the design criteria definedearlier in terms of the factors of safety and deflectionlimits.

The results displayed in Table 2 indicate that, forany core material selection, the best design

configuration for maximum weight savings is that ofsandwich panels with a light homogeneous core. Onthe other hand, the sandwich floor panel with a coreformed of cross C-channel beams may even increasethe required weight of the floor for certain materialchoices for the core C-channels. However, for mostof the material candidates listed in Table 2, thisstructural arrangement appears to provide higherstiffness than the other options compared here.Carbon-carbon composites allow the largest weightreductions and the minimum deflections for anydesign configuration. Obviously, the benefits ofusing carbon-carbon cores are strongly dependent onthe structural configuration of the floor. Emergingtechnologies for producing low-cost carbon-carboncomposites are expected to make them affordable forhigh-volume applications in heavy-vehicle structures[2–5].

(Note: Every structural arrangement evaluated abovecould be further optimized by altering, for example,the spacing between cross beams, the number oflongitudinal bars in Figure 1, or the characteristicsof the face sheets. However, the main objective ofthis study is to assess the predicted trade-offsbetween weight savings and stiffness for alternatecore material selections and not the optimization ofany one particular structural arrangement oranother.)

Table 2. Weight and deflection comparison of structural-material integrated design concepts.

I-Cross BeamsConnected by Bars

through WebCenters

(Alternative I)

FiberglassFaceplates

C-Channels Core(Alternative II)

Ribbed FiberglassFaceplates withCore of Hollow

Cross Tubes (III)

Sandwich Panelswith Light Core(Alternative IV)

MaterialDeflection

(in.)Weight(lb/ft2)

Deflection(in.)

Weight(lb/ft2)

Deflection(in.)

Weight(lb/ft2)

Deflection(in.)

Weight(lb/ft2)

STEEL 0.11 6.47 0.08 6.70 0.08 12.70 0.09 5.14Aluminum 0.09 6.50 0.08 4.03 0.08 5.83 0.13 1.77EXTREN 525 0.11 4.61 0.07 5.50 0.25 6.99 0.07 4.10Carbon-Carbon 0.05 1.98 0.06 2.21 0.06 6.00 0.09 1.04Nitronic 19D Stainless Steel 0.13 6.38 0.08 6.59 0.08 12.49 0.10 5.06Nitronic 60 Stainless Steel 0.12 6.50 0.09 6.72 0.09 12.72 0.11 5.15Nitronic 30 Stainless Steel 0.13 6.30 0.08 6.51 0.08 12.32 0.10 5.00Magnesium 0.17 2.18 0.13 2.59 0.13 3.76 0.10 1.28


146

Innovative Joining Concepts for CompositesThe development of advanced joining concepts forheavy trailer structures requires both theoreticalmodels and experimental results. This effort wasdirected mainly toward adhesive bonding ofsandwich panels and bolted joints of MMC platestrips. It is difficult to join sandwich panels since theload transfer has to be distributed over large areas inorder to prevent highly localized, concentratedstresses [7–9].

Various forms of joints between sandwich panelshave been devised and prototyped in the process ofconstructing a scaled, all-composite van-trailer, asdescribed in the next section of this report. All ofthese joints are free from mechanical fasteners andutilize different combinations of adhesive bondingand special connector parts to join (1) the side wallsof the trailer to its floor and roof, (2) segments of theside walls, or (3) segments of the roof to each other.An example of a three-dimensional finite-elementmodel (3D FEM) developed for theoretical analysisof a corner joint between sandwich panels(Alternative 4 in Table 2) is depicted in Figure 3.Since the panels connected through such a jointrepresent a corner section of a box trailer, their freeends in Figure 3 are constrained by symmetryboundary conditions. Both the vertical andhorizontal panels penetrate the tailored connectorpart to an insertion depth of “L,” along which aperfectly bonded interface is assumed to exist on allsides. Such a model provides an effective tool fortrade-off studies of various material and geometriccharacteristics of the joint before an optimalconfiguration is prototyped and tested. Typicalresults for theoretical analysis of such a joint areillustrated, at two different loading levels, inFigure 4 for sandwich panels consisting of 0.02-in.-thick aluminum faceplates and 0.75-in.-thick,continuous wood core. They indicate that thebending stress in the sandwich panel reaches itsmaximum value just outside the corner connector,but its magnitude can be slightly reduced byincreasing the “interlocking” length, “L.” However,this design parameter has no significant effect on themaximum deformations depicted in Figure 4 forcertain loading levels.

Figure 3. FE model of corner joint.

P= 100 lb

Fringe Level, ksi Fringe Level, ksi

P= 200 lb

-11.28

-7.536

-3.734

-0.052

3.630

7.432

11.17

-25.10

-16.31

-7.527

1.258

10.04

18.83

27.61

x

zy

Figure 4. Fringes of bending stress.

Three-dimensional finite-element modeling(3DFEM) was also used for failure simulation ofdouble-lap-bolted joints made of “Duralcan” and“LANXIDE“ Metal-Matrix Composite (MMC).Both materials consist of an aluminum matrixreinforced by silicon carbide particles in variousconcentrations. The 3DFE models were built tomatch bolted joint specimens of different geometricconfigurations that were tested to failure underunixial tensile loading, as illustrated in Figure 5.

Figure 5. Testing setup of bolted joint.


147

The material properties used in the 3DFE model forthe Duralcan and LANXIDE composites wereverified experimentally through tests conducted inaccordance with the ASTM standards E8M andD-70. The results obtained from these materialcharacterization tests are summarized in Table 3.

Table 3. LANXIDE vs. Duralcan properties.

Lanxide 30% Sic Duralcan 20 % SicMeasured Published Measured Published

Young’sModulus(Msi)

15.45 15.97 17.73 14.6

UltimateStrength(Ksi)

15.6 32.6 25.39 42.0

Table 4 displays a comparison between predictedand measured failure characteristics of Duralcan andLanxide double-lap-bolted joint specimens ofvarious “e/w” configurations (where “w” is thewidth of the face-plate, and “e” is the distance of thehole from the closest end edge). Excellent agreementcan be observed between the theoretical andexperimental results for Duralcan specimens.However, the test results for the LANXIDEspecimens deviate significantly from thecorresponding 3DFEM predictions, possibly becauseof inconsistent material properties and the existenceof manufacturing defects in these specimens. The3DFE model developed for design and failureanalysis of MMC bolted joints was validatedexperimentally not only by the correlation betweenthe calculated and measured failure loads, but alsoby comparing the predicted and the actual modes offailure, as illustrated in Figure 6.

Net Section Shear Pullout

Figure 6. Failure modes of MMC joints.

Prototyping of Scaled Trailer ModelA 1:4-scale model of a van trailer was designed andfabricated, as shown in Figure 7, to investigate newjoining concepts and assess their feasibility andbenefits as alternatives to conventional riveted,bolted or welded joints in heavy vehicle structures.The entire prototype was constructed manually fromcommercial, standard parts, such as glass fiber-reinforced polymer-matrix composite (PMC) panels,beams, rods, angles, and anodized aluminum “H,”“J,” “U,” and corner channels, as a preliminary test-bed for evaluating innovative design, assembly, andfabrication methods.

The fabrication of the scaled trailer model requiredjudicious combination of aluminum parts withtailored fiberglass composite panels, I-beams, rods,and angles to fit design specifications aimed atbuilding a lightweight, modular prototype, withstrong structural integrity. This was a step-by-stepprocess whereby the feasibility of new concepts wasexplored one-by-one, incrementally. The rear sectionof the model was completed in the first phase, thus

Table 4. Comparison of test and 3DFE model results for Duralcan and LANXIDE bolted joints.

Duralcan Failure Load (Lb) LANXIDE Failure Load (lb)Specimen

e(in.)

w(in.) Measured 3D FE Difference Measured 3D FE Difference Failure Mode

Specimen 1 2.5 2.0 3561 3776 6% 2277 1350 -41% Net SectionSpecimen 2 1.5 2.0 2516 3500 39% 1740 1350 -22% Net SectionSpecimen 3 0.6 2.0 1294 1348 4% 516 700 36% Shear pulloutSpecimen 4 2.5 1.5 2139 2360 10% 2121 1300 -39% Net SectionSpecimen 5 0.6 1.5 1353 1264 -7% 490 674 38% Shear pullout


148

b. Top edgejoint

a. Corner joint

c. Flat joint

Figure 7. Views of the van trailer model and joining details.

providing ideas and lessons that were subsequentlyapplied to the rest of the trailer.

The fiberglass panels are used for the roof and side-walls of the trailer and for floor covering. I-beamsare the main load-bearing elements of the floor.These cross-beams are reinforced by fiberglass“bearing bars,” which run perpendicular to andthrough the web of each I-beam. They also provideconnection points to the bogey, landing gear, andkingpin on the underside of the trailer. It is obviousthat the design configuration selected for the floor ofthe prototype trailer is similar to the designAlternative I (Figure 1) discussed in this report.Aluminum H-channels and edge corners were cutand manufactured to provide connections betweenadjacent side panels, as well as connections betweenthe side panels and the top panel. The H-channelsare anodized aluminum with 0.25-in. openings toaccept the thickness of the side panels. The H-channels had to be cut to the proper height andtrimmed to allow proper spacing with the corneredge trim. The U-channels and J-channels are usedas trim for the rear door of the trailer.

The side-walls of the trailer were segmented toprovide a reasonable level of flexibility forabsorbing dynamic forces. Bonding was utilized

extensively along the contact surfaces between thevarious aluminum channels and the fiberglass panelsin order to secure the joints and enhance the overallstability, or integrity, of the model. The bondingprocess requires thorough cleaning of the aluminumchannels and roughing the fiberglass panels withsandpaper. Araldite 2021 toughened methacrylateadhesive has been used to bond the fiberglass to themetal. Araldite 2021 provides a strong bond that fillsvoids between the mating parts and exhibits elasticbehavior in its cured state.

The prototype model was designed and constructedin a modular configuration that includes threeremovable or replaceable middle segments. Theoverall length of the model can therefore be changedin steps of one foot within the 9–12-ft rangecorresponding to the dimensions of the 1:4-scalemodel. In a real trailer, this feature would allow theadjustment of the trailer length to the volume ofcargo that has to be transported in a particular run.Furthermore, it allows modular repair orreplacement of damaged sections, without affectingthe entire trailer.

In conclusion, the prototype model provides ascaled-down physical representation of an actualtrailer that is used to assess the feasibility, benefits,


149

and deficiencies of alternative joining concepts,floor designs, side and top panel configurations, andbonding techniques. It has stimulated additionalcreative ideas for developing effective design,assembly, and construction methods for heavy vantrailers.

Durability Predictions of Particulate MMCMaterialsA new task has been initiated in the second year ofthis project to develop a predictive model for theeffects of intrinsic and/or externally induced multi-scale damage on the residual service life of fiber-and particulate-reinforced composite materials incritical components of heavy vehicle structures. Themodel relies on the periodic microstructure of suchmaterials and leverages earlier research performed atWest Virginia University in the area of ContinuumDamage Mechanics (CDM) of fiber-reinforcedPolymer Matrix Composites (PMC). The majorcontribution of those efforts so far is thedevelopment of a CDM model formulated in termsof parameters that can be measured easily fromstandard ASTM tests of single lamina couponspecimens and utilized for predicting the non-linearresponse of macro-scale laminates in the presence ofdamage [10, 11].

The main objective of the current task is to developan analytical model for analyzing the behavior ofdamaged particulate-reinforced metal matrixcomposites (PMMC) subjected to monotonicloading, by expanding and modifying the prior workon fiber-reinforced composites. This approachallows cost-effective quantification and analysis ofmaterial property loss associated with damageinitiation and accumulation in structural componentsmade of PMMC or other composite materials. Toachieve this goal, the model will be developed as auser-defined input to a commercial finite-elementsoftware package, such as ANSYS. A survey ofpertinent literature dealing with the onset oflocalized micro-scale crack initiation and growth inextruded PMMC materials [12, 13] reveals thatdamage initiates at the micro-scale from clusteringof the reinforcement particulate in certain regionsthroughout the PMMC material [14]. Consequently,this task is currently focused on the implementationof periodic microstructure techniques, along withCDM modeling, in order to describe the constitutive

behavior of PMMC materials subject to monotonicloading through the analysis of clustering regions.Periodic microstructure techniques are used todetermine the effective properties of the PMMCmaterial with homogenized clustering regions, asdepicted in Figure 8.

Homogenized particle clusters in a periodic distribution

Surrounding even particle distribution

0σ 0σ

0σ0σ

Homogenized particle clusters in a periodic distribution

Surrounding even particle distribution

0σ 0σ

0σ0σ

Figure 8. Homogenization of particleclusters in MMC model.

References1. J. Prucz and S. Shoukry (2003). “Structural

Characterization and Joining of MMCComponents for Heavy Vehicles.” 2003 AnnualProgress Report on Heavy Vehicle SystemOptimization, U.S. Department of Energy,Washington, D.C., pp. 115–120.

2. C.F. Leiteen, W.L. Griffith, and A.L. Compere(2002). “Low-Cost Carbon Fibers fromRenewable Resources.” 2002 Annual ProgressReport on Automotive LightweightingMaterials, U.S. Department of Energy,Washington, D.C., pp. 115–119.

3. H. Dasarathy, C.L. Leon, S. Smith, andB. Hansen (2002). “Low-Cost Carbon FiberDevelopment Program.” 2002 Annual ProgressReport on Automotive LightweightingMaterials, U.S. Department of Energy,Washington, D.C., pp. 121–132.

4. D.G. Baird, A.A. Ogale, J.E. McGrath,D.D. Edie (2002). “Low-Cost Carbon Fiber forAutomotive Composite Materials.” 2002 AnnualProgress Report on Automotive LightweightingMaterials, U.S. Department of Energy,Washington, D.C., pp. 133–136.


150

5. F.L. Paulauskas (2002). “Microwave AssistedManufacturing of Carbon Fibers.” 2002 AnnualProgress Report on Automotive LightweightingMaterials, U.S. Department of Energy,Washington, D.C., pp. 137–142.

6. J.R. Vinson (1999). “The Behavior of SandwichStructures of Isotropic and CompositeMaterials.” Technomic Publishing Company,Inc. Lancaster, Pennsylvania.

7. I. Kuch and F. Henning (2000). “ThermoplasticSandwich Structures with High Content ofRecycled Material — An Innovative One-StepTechnology.” 21st SAMPE Europe InternationalConference, Paris, France, April 18–20, 2000.

8. S.E. Mouring and R.P. Reichard (2001).“Fabrication of a Composite SuperstructureUsing a new Assembly Method.” Proceedings ofthe 11th International offshore and PolarEngineering Conference.

9. Hexel Composites. Adhesive Bonding ofHoneycomb Structures. http://www.heselcomposites.com

10. E.J. Barbero and L. DeVivo (2001). “AConstitutive Model for Elastic Damage in Fiber-Reinforced PMC Laminae.” J. of DamageMechanics, 10(1) 73–93.

11. E.J. Barbero and P. Lonetti (2001). “DamageModel for Composites Defined in Terms ofAvailable Data.” Mechanics Of CompositeMaterials And Structures, 8(4), 299–315.

12. D.L. Davidson (1991). “Fracture Characteristicsof Al-4% Mg Mechanically Alloyed with SiC.”Metal. Trans. A. 18A, pp. 2115–2138.

13. L. Mishnaevsky, M. Dong, S. Honle, andS. Schauder (1999). “ComputationalMesomechanics of Particle-ReinforcedComposites.” Computational Materials Science,v. 16, pp. 133–143.

14. J. Brockenbrough, W. Hunt, and O. Richmond(1992). “A Reinforced Material Model UsingActual Microstructure Geometry.” ScriptaMetal. Mater., 27, pp. 385–390.


151

VII. Analysis

Systems Analysis for Heavy Vehicles

Principal Investigator: Linda GainesArgonne National LaboratoryArgonne, IL 60439(630) 252-4919, fax: (630)252-3443, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected]


Objectives• Analyze transportation technologies and fuel-efficiency measures objectively.

• Provide critical and unbiased evaluation of transportation or materials-related projects.

• Serve as a key idling-reduction resource for DOE.

Approach• Analyze technical and economic characteristics of competing technologies.

• Represent DOE at idling-reduction meetings.

• Coordinate efforts by DOE and other agencies.

• Supply technical input to policy-makers.

• Maintain awareness with latest technical and political developments related to idling.

Accomplishments• Revised and expanded technology sections of idling-reduction report.

• Planned and acted as technical chair for National Idling Reduction Planning Conference.

• Provided tutorials on idling reduction at numerous Clean Cities and other meetings.

• Completed short study of dynamic brake energy recovery for locomotives and presented at RailroadEnvironmental Conference.

• Managed awarding of contract to Antares for analysis of wear due to idling.

Future Direction• Complete revised and expanded technical report.

• Lead Technology Working Group (with Lee Slezak).

• Complete and promulgate National Idling Reduction Plan (with other working group chairs).

• Evaluate technology options for domestic liquid fuel production.


152

Truck Idle-Reduction TechnologyComparisonBefore comparing energy savings or economics, it isuseful to clarify the basic differences among theavailable devices for reducing long-duration idling.They do not all supply the same services, and sotheir relative merits will depend on several factors,including services required and number of hours ofidling that would be avoided. Table 1 summarizesthe basic differences among the technologies andpoints out any particular advantages ordisadvantages of each.

Table 1. Benefits and drawbacks of truck idle-reductiontechnologies.

Technology Benefits DrawbacksAutomatic start/stop Intermittent services

anywhereNoise disrupts rest;uses main engine

Direct-fired heater Heat anywhere; smalland inexpensive

Cannot supply cooling;uses battery

Auxiliary power unit(APU)

HVAC and poweranywhere

High cost and weight

Truck stopelectrification (TSE)

All services; no localemissions

Only at fixed locations;limited potential

Integrated electricalaccessories with APU

HVAC and poweranywhere

In development (2005intro); only on newtrucks

Energy Use and EmissionsEnergy UseDiesel fuel saved by reducing idling is fairly simpleto estimate; it is simply the difference between thefuel that would have been used idling and thatconsumed by the alternative system. For theautomatic start/stop systems, the fuel saved is simplythat percent of idling fuel use that corresponds to theperiod when the engine is off. Emissions aresimilarly reduced. If the idle-reduction technologyburns diesel fuel, savings are reduced by thedevice’s own fuel consumption, typically 5–25% ofthe savings. Emissions may be reduced by more orless than the same percentage, depending on howclean the idle-reduction technology is. Such devicesall meet appropriate small engine standards, butthese are generally less stringent than those for truckengines. Emissions from several devices have beenmeasured by the EPA. For devices that plug in, all ofthe idling diesel fuel use and emissions are avoided,but the fuel use and emissions from electricitygeneration at the power plant must be accounted for.

Several earlier publications have neglected to dothis.

Finally, there is the more complicated case ofdevices that run off of power from batteries thathave been charged during the vehicle’s operation.No information was found about the additional fuelconsumed due to battery charging. Because systemsthat run off a battery do not burn fuel while they areproviding services, vendors generally do notrecognize and report fuel use for their systems. Weestimate it here. If the unit draws a maximum of8 Amps at 12 V, and we assume that it operates at75% capacity for 8 h, for an average power draw of72 W, it consumes 2074 KJ of energy directly. Fromthe Bosch Automotive Handbook, automotivealternators are approximately 50% efficient —meaning that 50% of the rotational energy from theaccessory-drive belt is converted into electricalpower. The charging efficiency of typicalautomotive-style charging systems is between50 and 65%, meaning that 50–65% of the electricalpower delivered to the battery actually goes intoincreasing its state of charge, and the rest is lost asheat. The faster the battery charges, the moreinefficient the process becomes. We assume that thebattery is charged relatively slowly while the truck isoperating, and charges efficiently (65%).

That means that the engine-work-to-battery-chargeefficiency is 32.5%. Therefore, the amount of energythe engine must provide to make up for what the airconditioner consumed is 6382 kJ. That amounts to221 W of engine power. This is the most optimisticcase, assuming a separate battery pack for thispurpose. If the starting battery is used, the minimumlevel of charge needs to be maintained to ensurestarting, so the charging rate must be higher, whichmeans lower efficiency.

The 6383 kJ of energy is required at the drive belt.This energy is supplied by the truck’s diesel engine,which is about 40% efficient in turning fuel energyinto shaft power. Therefore, it needed to burn15,958 kJ (15,196 Btu) of fuel, or about 0.112 gal offuel (0.014 gal/h). This is a best-case estimate;actual fuel use could easily be doubled, dependingon conditions and weather (battery-chargingefficiency is very low in cold weather). This is stillvery low energy use, because the 72 W powerdemand is less than that of most light bulbs. Even if


153

the unit drew 350 W, total fuel use would only be0.14 gal/h in cold weather.

EmissionsUntil recently, one had to rely on measurements byequipment manufacturers to estimate emissionreductions achieved by use of alternative devices.However, recent experiments funded by the U.S.EPA, New Jersey DOT, and DOE at the Army’sAberdeen Test Center carefully measured emissionsof CO2, CO, NOx, HC, particulates, and aldehydesfrom idling trucks, and from installed auxiliarypower units and heaters, under several sets ofconditions. Measurements were taken at severalidling speeds, in hot, cold, and mild atmosphericconditions, and with several different trucks. Theresults of this work clearly demonstrate theenvironmental benefits of reducing truck idling. Thedetailed results are presented in two papers [1,2]; wesummarize them and draw some conclusions here.The exact numbers vary with truck model; we choseto provide as an example the impact reductionsachieved with the newest truck tested (2001Freightliner), because of the rapid turnover of truckfleets. We determined that it was most meaningful tocompare impacts from a heater or APU at 0°F tothose from the truck idling at the speed at which itwould normally idle to supply heat (700–800 RPM).Therefore, we interpolated the detailed data providedand estimated truck heating emissions at 750 RPMand similarly estimated idling emissions duringcooling at 90°F at 900 RPM. Note that theconditions chosen are somewhat extreme; impacts atmore moderate temperatures are expected to belower. An accurate estimate of annual impacts andbenefits would require integration over typicalannual conditions and loads. It would also requiremeasurement of APU impacts at 65°F, when itwould be run for electrical loads only.

The emissions for the truck and alternatives at thesetwo extreme conditions are shown in Table 2. Asexpected, all of the impacts are very low from theheater, but it is only supplying heat and noelectricity.

All of the impacts from the APU are lower thanthose from idling, for both heating and cooling. It isinteresting to note that the emission reductions aregreater on a percentage basis (in most cases) thanthe fuel use reductions (as represented by the CO2emissions). This could be because idling represents anon-optimal engine condition, compared to that ofthe auxiliary device. The one glaring exception isthe PM emissions from the APU when supplying airconditioning. These are reduced by only 50%, eventhough fuel use is reduced 75%. Since only oneAPU was tested, this may or may not be typical.However, note that the NOx emissions have beenreduced disproportionately, and it is likely that theAPU’s engine could be adjusted to trade NOx forPM.

Impacts from those devices that run off the batterywould be equal to the marginal emissions fromincreasing fuel consumption on the road. Theimpacts from shore power are simply the emissionsfrom generation of the electricity that they draw.Although this actually varies by region with fuelmix, we will use the national average for technologycomparisons in this report.

EconomicsThe economic benefits of idling-reductiontechnologies depend on the costs avoided by notidling and on the costs incurred to purchase or leaseand use the idling-reduction technology. Benefits areexpected to increase as the number of idling hoursdisplaced increases. The largest avoided cost, that

Table 2. Emissions from idling and alternatives.

Device ConditionCO2

g/h (% red.)NOx

g/h (% red.)HC

g/h (% red.)PM

g/h (% red.)CO

g/h (% red.)Truck Heat 7500 158 26.2 3.43 115APU Heat 2146 (71) 8.7 (94) 7.8 (70) 0.478 (86) 25.0 (78)Heater Heat 445 (94) 0.2 (99+) 0.04 (99+) 0.055 (98) 0.1 (99+)Truck AC 9500 137 35.7 2.0 60APU AC 2351 (75) 11.4 (92) 4.2 (88) 0.995 (50) 10.8 (82)


154

for fuel, can be calculated simply for on-boardtechnologies that burn diesel fuel (e.g., APUs andheaters). For these, the fuel savings are simply thedifference in hourly fuel use (in gallons per hour)between idling and APU use, multiplied by thenumber of hours of idling displaced during theperiod in question, times the fuel cost per gallon. Fordevices that plug in, the avoided fuel cost is simplythe cost of the diesel that would have been burnedidling minus the cost of the electricity purchased.The situation is more complicated for systems thatrun off of batteries that are charged while the engineis running. In that case, it is appropriate to estimate,and bear the cost for, the extra fuel burned duringtruck operation in order to charge the batteries, aswas estimated above.

In addition to savings from reduced fuelconsumption, not idling a truck reduces maintenancecosts. Oil changes can be performed less frequentlyif the engine is operated for fewer hours; thus, thecost for oil changes can be reduced. The Technologyand Maintenance Council (formerly the TruckMaintenance Council) of the American TruckingAssociations has published two RecommendedPractice (RP) Bulletins to help truckers estimatemaintenance costs due to idling. The first, publishedin 1985, was superceded by another (RP1108) in1995 [3], in which the cost estimates weredrastically reduced due to the reduction in fuel sulfurand changes in idling practice. Care must be taken touse the updated version; several recent publicationshave cited the older one. The main difference is thenumber of “equivalent miles” represented by onehour of idling. The revised RP estimates this fromthe fuel consumption (i.e., each gallon of fuel usedidling is accounted for as if it were used at thetruck’s average miles per gallon). So one hour ofidling is typically equivalent to about 6 or 7 miles ofdriving. Then, if an oil change costs $150 and isdone every 35,000 miles, the cost per hour of idlingis about $0.03. This calculation should be done morecarefully, based on the actual fuel economy and oilchange cost and frequency for the truck or fleet to becosted; this can be done by using the worksheetincluded here as Figure 1.

A similar calculation can be done for the cost due tooverhaul. If a truck requires an overhaul after500,000 miles, and the overhaul costs $10,000, the

Figure 1. Idling cost worksheet.

cost per hour of idling is $0.14. Again, this is simplyan example. Some trucks just need partreplacements, costing about $2500, while othersmight need a full engine-out overhaul, costing asmuch as $20,000. There is also considerablevariation in mileage to overhaul; a typical mileageused to be 300,000, but now may be up to1,000,000 miles, and no overhaul may ever beperformed.1 The worksheet allows the user to selectappropriate entries for these parameters. It would bevery useful to obtain actual data on the effects ofidling on oil quality and engine wear.

Note that many truck owners, especially fleetowners, sell their trucks before it would be time foran overhaul, and therefore may not choose toinclude this component in their cost estimates.However, if the buyer can see the truck’s idlinghistory in its computer records, the purchase pricewill theoretically be reduced for a truck that hasidled excessively. However, mileage does not seemto be a key parameter cited in ads for used trucks.

Once the fuel and maintenance costs for idling havebeen estimated, the savings and payback times forthe various idle-reduction technologies can also beestimated. Figure 2 shows the payback time forseveral alternatives, as a function of idling hoursdisplaced. The times in the table are durations ofdevice operation; for example, if the device is aheater, those hours are during the winter only, so themonths until payback are also winter only, and itmay take two years to accumulate 12 months ofpayback time. Payback times for a truck with low


155

0

10

20

30

40

50

60

1000 3000 5000

Hours of idling displaced

Mon

ths

to p

ayba

ck

APUHVACStart/stopEngine/cab heaterAC

Figure 2. Reduction of payback time with increasedidling hours.

yearly idling time are very long. Thus, a firm thathas reduced its idling by changing the operatingpractices of its drivers might not find it economicalto install equipment to further reduce idling.

Truck stop electrification (TSE) costs are slightlycomplicated because some are borne by the truckowner and some by the owner of the electrifiedsystem. Both must see an appropriate payback if thesystems are to be installed and used. For standardTSE, the trucker sees the cost of the equipmentrequired to electrify the truck, and the marginal costof using an electrified space at a truck stop (the extracost over standard space rental). This is presumablygreater than the actual electricity price. The truckstop owner must pay (1) to install and maintain theelectric plug system and (2) for the actual utility bill.There is little incentive at this time to electrify truckstops; the few that have been demonstrated haverather low utilization because so few trucks are ableto use the spaces so far. 2

Advanced TSE is a win for the truck owner; theavoided idling costs are as above, and the cost to usethe system is simply the small initial cost of thewindow template (about $10) plus the hourly chargeof $1.50 ($1.25 for subscribers). The economics ofthe system itself are less clear. The system isinstalled by its manufacturer, with the truck stopowner receiving a portion of the proceeds. Theinstallation cost is as much as $17,000 per space andis generally shared by a government agency in orderto demonstrate the technology. Operating revenuesare the fees collected for the service, minus any cutto the truck stop owner, minus any maintenancecosts. These may not be small, given the reported

12–20 staff members required per 100 parkingspaces. Detailed costs have not been published, butit is likely that high utilization would be required tomake the project economical.

ConclusionsIn summary, although numerous technologies areavailable to reduce idling of heavy-duty trucks, thereis no one perfect device that will satisfy the needs ofevery trucker. The appropriate device to choosedepends on the region that is traveled (for instance, asouthern route might need just air conditioning), theschedule (shorter hauls might not justify APU costs),and the equipment available (highly traveled routeswill install electrification faster). Better operatingdata and continued technological development willallow truck owners to make purchase decisions withmore confidence in the future.

References1. Particulate Matter and Aldehyde Emissions from

Idling Heavy-Duty Trucks, J. Storey, et al.,SAE Paper 2003-01-0289 (2003).

2. Study of Exhaust Emissions from Idling Heavy-Duty Diesel Trucks and CommerciallyAvailable Idle-Reducing Devices, H. Lim,U.S. EPA Office of Transportation and AirQuality, EPA420-R-02-025 (October 2002).

3. TMC Recommended Practice 1108, Analysis ofCosts from Idling and Parasitic Devices forHeavy Duty Trucks, American TruckingAssociations, March 1995.

1 Personal communication, A. Laible, Cox Transfer,

January 2004.2 T. Perrot et al., Transportation Research Board,

January 2004.


156

VIII. Off-Highway

A. 21st Century Locomotive Technology

Principal Investigator: Lembit SalasooGeneral Electric Global Research1 Research Circle, Niskayuna, NY 12309(518) 387 5024, fax: (518) 387 6675, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Jules L. Routbort(630) 252-5065, fax: (630) 252-4289, e-mail: [email protected]

Participant: Paul K. Houpt, General Electric Global Research

Contractor: General Electric Global ResearchContract No.: DE-FC04-2002AL68284

Objectives• Develop and demonstrate locomotive system technologies targeting gross fuel efficiency gains.

• Demonstrate a modular energy storage system for hybrid locomotive system on a hybrid locomotive platform.

• Develop optimization software for minimizing fuel consumption in both hybrid and non-hybrid locomotivesand demonstrate the consist optimizer on the test track.

• Demonstrate the combination of a modular energy storage system and a fuel optimizer on a hybrid locomotivewith a full-scale energy storage system.

Approach• Specify and develop advanced modular energy storage units and lab test prototype modules.

• Design and bench test advanced hybrid locomotive energy management system (EMS) controls anddemonstrate controls on a hybrid locomotive on a test track.

• Develop suitable models for dynamic optimization of fuel use.

• Develop practical, robust fuel-use-optimization algorithms capable of locomotive implementation.

• Demonstrate fuel use optimization algorithms in off-line interactive simulation environment.

• Complete field demonstrations of selected fuel saving algorithms.

• Fabricate and bench test full-scale energy storage modules and associated controls.

• Integrate full-scale energy storage modules to hybrid locomotive.

• Design, fabricate, and bench test hybrid fuel optimizer controls.

• Integrate fuel optimizer controls on hybrid locomotive; demonstrate on test track.

• Measure and verify fuel savings benefit.


157

Accomplishments• Downselected advanced battery technology for hybrid locomotive and initiated development of advanced

modular energy storage units.

• Initiated advanced battery technology long-term life testing.

• Designed and bench tested advanced hybrid locomotive energy management system controls. Integration tohybrid locomotive under way.

• Conducted successful 3400-mi field test of Consist Manager system, demonstrating fuel savings of 1–3% (1.5%average).

• Designed and evaluated robust optimal and greatly simplified near-optimal approaches to compute fuel-savingdriving plans.

• Developed real-time simulation platform for interactive demonstration of Consist Manager and Trip Optimizer.

Future Direction• Fabricate and evaluate locomotive-worthy hybrid energy storage modules.

• Design the integration of a hybrid energy storage system to the hybrid locomotive platform.

• Fabricate full locomotive set of hybrid energy storage modules, integrate system to the hybrid locomotive, andtrack test.

• Develop “bullet-proof” and simple operator interface for fuel optimization.

• Conduct on-locomotive test of hybrid fuel optimization controls.

• Measure and verify fuel performance of combined system.

IntroductionA systems view of the locomotive and its consistreveals significant opportunities for double-digitreduction in locomotive fuel use through thefollowing:

• Add new means to store and recover brakingenergy that would otherwise be dissipated asheat and

• Exploit new ways to coordinate train driving andenergy storage through advanced controls,which leverages energy management throughoutthe trip.

When applied to the North American Class 1railroad fleet, advanced hybrid-electric propulsiontechnology for storing and recovering the brakingenergy would save 450 million gallons of diesel fuelannually, consist fuel optimizer techniques wouldsave 55 million gallons annually, and trip fueloptimization together with hybrid technology wouldsave an additional 230 million gallons annually.

Two major fuel optimization strategies are beingdeveloped: (1) Consist Manager — control logic thattakes the commanded power by the driver (throttlenotch) and distributes it among operatinglocomotives in the most fuel efficient manner and(2) Trip Optimizer — control logic that works incollaboration with the driver and dispatcher to planand drive a trip, taking into account the powerconsist, terrain, freight load, and operatingconstraints along the way. FY 2004 effortsculminated in the highly successful demonstration ofconsist optimizer fuel savings on a group of Class 1railroad revenue missions. As a direct result of thisDOE project, a commercial Consist Manager systemproduct has been launched. In addition, the bestbattery technology solution was downselected fromthe available advanced hybrid battery technologies.Development started of a locomotive-worthy systemfor freight locomotive demonstration in 2006.


158

Advanced Modular Hybrid Energy StorageDevelopmentTypical hybrid operation scenarios were determinedfrom existing locomotive field data in order todetermine the requirements of the hybrid energystorage system’s performance. Studies to addressbattery system technology developments that wouldmeet the hybrid locomotive requirements werecarried out by advanced battery vendors. Acomparative assessment of the results was carriedout, in terms of footprint (weight and volume)driven by life, power, and energy requirements;thermal and electrical management requirements;and readiness for locomotive deployment. It wasseen that the capabilities of the vendors and energystorage technologies were clearly differentiated. Onetechnology and one vendor were selected fordevelopment and delivery of enhanced-capabilitymodules. Agreement was secured with the advancedenergy storage vendor to work to meet the projectschedule and deliver prototype modular energystorage units for evaluation in the next project yearand to subsequently deliver the full set of advancedmodular energy storage units for on-locomotivetesting in FY 2006. Battery compositionmodifications were evaluated for customization tomeet locomotive requirements.

The life of hybrid locomotive battery systems iscritical to acceptance by the railroad industry.Battery cycling life depends on many operatingparameters, such as duty cycle, depth of discharge,and operating temperature, and there is a significantgap in published data applicable to candidateadvanced battery technology performance in hybridlocomotive application. A battery subassemblytesting program was set up to develop detailed datato extend understanding of the life in the locomotiveapplication. A systematic life-testing protocol wasdeveloped to establish the baseline cycling lifeentitlement of the advanced energy storagetechnology. It was expected to continue long-termtesting for at least 12 months.

Initial results of life-cycle testing indicated thatsome of the electrochemical cells under test haddrifted away from their operating point due toinaccurate measurement of charge. Figure 1 showsmodule voltage over selected charge-dischargecycles. It can be seen that the end-of-discharge

2021222324252627282930

0 500 1000 1500 2000

Cycle-1Cycle-100Cycle-200Cycle-300

Battery test module voltage vs time (sec)

2021222324252627282930

0 500 1000 1500 2000

Cycle-1Cycle-100Cycle-200Cycle-300

Battery test module voltage vs time (sec)

Figure 1. SOC drift of battery test module.

voltages at approximately 750 s and end-of-chargevoltages at approximately 1500 s are trending lowerfrom cycle 100 to cycle 300. This indicates that thecycle mean state of charge is trending down and isnot being maintained constant as is required by thelife-testing protocol. Assessment of the condition ofthe modules indicated that some had been exercisedbeyond manufacturer’s recommended limits, and sothey may have suffered premature aging and thusshould not be used further for life testing. The testprotocol was revised and a new set of batterymodules was obtained for restarting life tests withappropriate control of operating point.

Advanced Hybrid Locomotive EnergyManagement System ControlsDetailed analysis was carried out of hybridlocomotive field test data to confirm performancegaps in battery management. It was decided thatbattery management enhancements would addressthermal compensation and accurate internal statedetermination.

Extensive laboratory testing of battery modules ofthe types to be used in the upcoming advancedhybrid EMS locomotive track test was carried out todevelop a database for use in battery stateprediction. The performance of the battery statecomputation in a hybrid locomotive duty cycle wasassessed by exercising the battery modules with datacollected from the previous field test. The firstbattery state prediction algorithm was notsuccessful, and so a follow-up round of laboratorytesting was performed in more detail. A temperaturecompensation algorithm was successfullydeveloped, which will enable better utilization ofenergy storage battery capabilities.


159

The hybrid locomotive advanced energymanagement system track test will be carried out ona more advanced platform than presented in theproposal. In the April 2002 21st CenturyLocomotive proposal, the platform fordemonstrating an advanced energy managementsystem was shown to consist of a towed energystorage container mounted on a flatcar, electricallyconnected to a GE AC-4400 locomotive. This hybridtestbed configuration has been upgraded by GEthrough introduction of nickel-metal-hydridetechnology of the type developed by theU.S. Advanced Battery Consortium, and the hybridenergy storage system was mounted onto thelocomotive platform itself (Figure 2).

This “on-board hybrid” locomotive is suitable forfield testing in freight hauling service. Future testingunder task 3 of this program will have enhancedvalue, as it will be using more advanced batterytechnology and a realistic locomotive platform.GE Rail is completing required software andhardware modifications to the hybrid locomotive(at no cost to the program) to measure key hybridbattery parameters and interface these to theadvanced EMS controls.

The DOE team was able to inspect the GE Rail on-board hybrid locomotive and ride the locomotive onthe GE test track during the May 13, 2004, projectreview in Erie, PA.

A test plan for track testing of the hybrid locomotivewith advanced EMS control was prepared. Anextensive evaluation of the installed hybrid batterieswas performed. After consultation with the batterymanufacturers (including an on-site visit), thebatteries were cleaned, overhauled and tested to

EnergyStorageEnergyStorage

Figure 2. Tow-behind and on-boardhybrid locomotive configurations.

establish their existing capability. When the batteryparameter measurement modifications to the hybridlocomotive are completed, the track test of theadvanced EMS algorithms will be carried out.

Models for Fuel Use Dynamic OptimizationHybrid energy storage capability was added to ourconventional locomotive and train analysis tools.This enabled setting up the hybrid trip optimizationproblem, which can be formulated in several ways.In one approach, the energy storage control istreated as an additional optimization variable. Thisapproach will enable us to explore generalizedstrategies of hybrid energy management that may benon-intuitive. The integrated model allows study ofthe impact of energy management strategies on thetrip optimization problem.

Trip Optimizer AlgorithmComputing a driving strategy for a freight consist tominimize fuel subject to operating constraints isdifficult because of the many constraints andconflicting objectives. For example, over a 100-mijourney, speed limits may change 50 times and havethousands of decision variables (e.g., 50 m.p.h.average speed with 1-s control updates). To form anumerically reliable baseline in solving theseoptimal control problems, the goal is to developnumerically robust solution techniques,accomplished with proprietary software. In thisproblem formulation, the objective function caninclude fuel consumed, total emissions generated,target arrival time, rate of change of throttle position(e.g., as a penalty against induced slack-action), andother performance-related variables. Constraintsincorporated can include speed restrictions along theroute and power limits. Information about the trackgrade and curvature is part of locomotive traindynamic model. We have achieved major progressin development of reliable methods for computingoptimal driving plans in Trip Optimizer and havedeveloped a simplified way to re-optimize a partiallycompleted plan in an efficient way.

We have completed simulation studies of optimizedtrajectories on a variety of terrains, grades, andspeed restriction profiles. Our analysis shows thetypical trade-off of fuel savings versus journey timethat results. While the exact shape of the trade-offcurve varies with terrain, the load, and the specific


160

speed constraints, they all exhibit the large reductionin fuel consumption per marginal hour of additionaltravel time. This means even small increases injourney time taken when slack time is available orslow orders are ahead have large potential fuelsavings.

Sample results are shown in Figures 3 and 4 for a100-mi trip segment on BNSF’s southwest territory.This is for a 3500-ton consist pulled by 2 GE Dash 9locomotives operating on a relatively flat stretchbetween Kansas City, MO, and Wellington, KS. Asthe travel time objective varies, Figure 4 shows theresulting fuel consumption. While it is not a typicalgoal for a railroad to slow down, the curve doesshow the strong sensitivity of optimal fuel use tosmall changes in travel time, without hybrid energystorage, and is illustrative of potential for savings inthe real world where speed restrictions are commondue to traffic, work crews, or other problems.

Figure 3. Example optimal driving trajectory(1.9 h travel time).

Figure 4. Example of fuel use vs. travel time.

Simplified Trip Optimization AlgorithmsThis year, we have developed a new searchalgorithm that we think can radically reduce thecomplexity of applying Trip Optimizer to longertrips and, more importantly, to allow on-the-flyrecalculation of plans following an interrupt. Ourapproach partitions the journey segment and thenmatches up and links the subsegments back togetheron each journey segment to meet the overall timeobjectives with the least average fuel. Since no largeoptimizations are carried out, it can be done in realtime or nearly so. Moreover, if conditions change atsome point during a journey, the plan can be re-optimized over the remaining segments, taking intoaccount the changed conditions.

Figure 5 shows just one example of the effectivenessof this approximation method for a 200-mi trip on aportion of the BNSF Kansas City to LA route again.For this trip, our benchmark optimization softwarefound the optimal fuel use to be 6830 lb. Theapproximation technique was able to complete thetrip with 6880 lb, which is only 0.7% worse, but theproblem could be solved three orders of magnitudefaster. Note that all of the speed limits are preservedin the approximate solution.

Fuel Consumption

lbs x 103

Optimal = 6.83

Approx = 6.88

Speed Limit

Optimal

Approx

Fuel Consumption

lbs x 103

Optimal = 6.83

Approx = 6.88

Speed Limit

Optimal

Approx

Speed Limit

Optimal

Approx

Figure 5. Comparison of optimal solution withapproximate technique.

Real-Time Interactive Demonstration ofOptimizer AlgorithmsA real-time simulation system has been constructedto serve two purposes: (1) demonstrate ConsistManager in a simulation close to its commercialembodiment and (2) provide a platform for later


161

interactive simulation demonstration of TripOptimizer. Figure 6 shows a block diagram of thecomputing platform used.

With this system, we can use our engineeringsimulation to model drivetrains with and withoutoptimized solutions concurrently. In addition, wecan show a fully animated, real-time 3D viewoutside the locomotive to convey some realism toanyone operating the consist through a controlinterface panel. While this display is not essential tocalculating the quantitative performance in eitherConsist or Trip Optimizers, it will be an importantelement in the demonstration of the Trip Optimizerbefore implementation in a locomotive during thisproject. The demonstration system was firstprototyped for the 2004 RSI Rail Show in Chicago(Figure 7), which was the first public offering of theConsist Manager system developed in part underthis DOE contract. By allowing customers directreal-time access to the Consist Manager controller, itwas easy to show the simplicity and efficacy of thenew product.

Consist Manager Field DemonstrationThe goal was to build a prototype of the ConsistManager system and field test it on a Class 1railroad. BNSF railroad agreed to host GE on across-country trip in a revenue service trainoperating in high-priority service. The ConsistManager control strategy was implemented on arevenue service “Z-train” on BNSF’s southwest“Transcon” route (Figure 8). The 1–3% fuel savings

Figure 6. Simulation environment for Consist Managerand Trip Optimizer.

Figure 7. Consist Manager display setup for RSI RailShow (Chicago, September 2004).

Figure 8. BNSF route for Consist Managertest.

demonstrated, together with flawless operation ofthe prototype system, have generated strong interestby the participating railroad.

Four GE Dash 9 locomotives and a 3000-toncontainer and trailer on flatcar cargo comprised theoverall 4500-ft train consist. The consist was aregularly scheduled priority cargo train that runsfrom Kansas City to Los Angeles and back with ashort layover, a total distance of 3400 mi at anaverage speed of 41 mph. The Consist Manager wasoperational with no problems for the entire journeyin both directions.

By using algorithms prototyped and tested against adynamic train simulation, software was coded forreal-time execution by using an “xPC” targetplatform. “xPC” is a PC-class computer controllerinterfaced to the locomotive consist by usingcommercial off-the-shelf digital and analog I/O(Figure 9), which is physically located in the secondof the four locomotive units.

42” LCD Simulation display

Graphics Computer

Cube

Graphics Computer

Cube

LaptopComputerLaptop

Computer

20” LCD Control Display

LocoControl Lever

Panel

LocoControl Lever

Panel

USB Port

Speakers/Amplifer

Speakers/Amplifer

VGA OutVGA Out

AudioOut

AudioIn

Network Interface

KeyboardKeyboard

42” LCD Simulation display

Graphics Computer

Cube

Graphics Computer

Cube

LaptopComputerLaptop

Computer

20” LCD Control Display

LocoControl Lever

Panel

LocoControl Lever

Panel

USB Port

Speakers/Amplifer

Speakers/Amplifer

VGA OutVGA Out

AudioOut

AudioIn

Network Interface

KeyboardKeyboard


162

Figure 9. xPC control computer and I/O on locomotive.

While the ultimate implementation of ConsistManager is likely to be within the locomotivemicroprocessor control system software, this versionallowed rapid prototyping of the controller withoutthe inconvenience of having to modify locomotivewiring. The wiring changes consisted of interfaces tothe master control throttle in the lead unit and theMU-train line and CAB controllers in the both thefirst and second units. The trailing third and fourthunits required no changes. All changes were quicklyreversible. The control computer was monitoredfrom a computer host located in a test car attached tothe rear of the power consist.

In a conventional multi-unit locomotive consist, thethrottle commanded in the lead unit is relayed to allthe trailing units. With Consist Manager, thecommanded power to each locomotive is computedto deliver the total required power to within aspecified tolerance with the best overall fueleconomy: it exploits the fact that specific fuelconsumption varies with locomotive power output.Part of the control strategy is the dynamic logic thatmanages transitions from one optimized notch toanother for best performance. Thus, one goal of therevenue test was to evaluate our strategy formanaging the transient transitions.

The prototype Consist Manager system was operatedthroughout the run from Kansas City to Los Angelesand back. All hardware and software operatedreliably for close to 100% of the trip, allowingConsist Manager to be in the loop for more than3000 revenue service operation miles; more than1 GB of data on performance was collected.

Fuel savings were estimated to be between 1% and3%, with an average of 1.5%. These estimates wereobtained by using a model for the fuel use vs. notchfor a typical Dash 9 locomotive, multiplied by theduty cycle in each notch as observed on the trip withoptimizer enabled compared to what the fuel wouldhave been with all locomotives in the same notch.To avoid overstating the fuel savings, the estimateassuming un-optimized operation was normalized sothat both calculations of savings had the samehorsepower-hours:

where fu is fuel used in pounds and Eu the totalenergy (hp-h) for unoptimized operation; “o”subscripts denote the corresponding optimizedquantities.

Ten different crews and their managers, includingthe BNSF Vice President for Operations, had anopportunity to drive the train with the system active.They reported only minor issues and had no problemadjusting their driving style to the minor changes intransient response with Consist Manager active.Detailed questionnaires from all crew were collectedand are being analyzed. We observed qualitativelythat some crews had an easier time adapting to thechanges in power response with the ConsistManager enabled than others — in particular, thosewho anticipate power changes better could holdspeed and get better performance, suggesting thatwith experience, these fuel savings projections maybe surpassed.

BNSF provided a 24-hour window of draw-bar buffand draft forces, which showed that no unusualtransient slack-action forces were produced with theConsist Manager active.

ConclusionsA comprehensive locomotive hybrid technologydevelopment program is well under way. The uniqueenergy storage requirements for a hybrid locomotivehave been specified, and the appropriate batterychemistry solution has been selected. Acomprehensive hybrid energy storage systemtechnology development has been initiated with the

[ ] [ ]u

ou

EEE

/f/f/f

savingsu

ou −=∆


163

vendor, who will deliver locomotive-worthyevaluation systems in the upcoming year and a fullsystem for hybrid locomotive installation early inFY 2006. Advanced energy-management techniqueshave been developed this year and are beinginstalled in a hybrid locomotive.

We have met or exceeded all the goals set out for theproject fuel optimization task, particularly with therevenue service locomotive demonstration ofConsist Manager. This test has built credibility withour partner railroad BNSF. The 1–1.5% average fuelsavings offered by Consist Manager has provensufficiently attractive that it is on track forcommercial release as a product during the firstquarter of 2005.

It is a big step to go to the Trip Optimizer, requiringchanges among crew and dispatchers to achieve thefuel saving potential of 4% or more. Our confidencein robust and efficient ways to calculate optimaltrajectories has greatly increased, and we will beginapplication to models with hybrid storage next year.Building on the database developed in the January2004 Consist Manager test, we will continue toconduct case studies with varying terrains, consists,and loads representative of realistic Class 1 railroadfreight operations. Our latest results are showing thateven with the same travel time, it is possible toachieve significant (4–5%) fuel savings by carefully

managing power and braking. These will be used toform fuel use reference baselines, enabling us tocompare the range of benefits: from ConsistManager “assistance,” to optimal and sub-optimalTrip Optimizer solutions, to the same TripOptimized consists working with hybrid storage.

Our belief is that the viability of Trip Optimizerhinges on a bullet-proof and simple operatorinterface. To allow design and evaluation ofpotential operator interfaces for Trip Optimizer, wewill also be expanding the real-time interactive workstation described above. This real-time facility willuse our Simulink simulation models with actual cabhardware control inputs. Outputs from the simulatorwill be displayed in a 3D graphics display with an“out the window” world view. This will allowdesign and testing of trip planning scenarios with asimplified but realistic cab environment that a drivermight experience, to gain insight into the best way tointeractively carry out a plan and also handle thetrain safely.

Our major task for next year will be to captureenough of the core functionality in trip planning,with and without hybrid storage in the locomotive,and conduct an on-locomotive field test. We areexploring options with several Class 1 railroads toperform the field test again on a revenue-servicetrain.


164

B. Advanced Hybrid Propulsion and Energy Management System for HighEfficiency, Off Highway, 320 Ton Class, Diesel Electric Haul Trucks

Principal Investigator: Tim RichterGE Global Research, 1 Research Circle, EP110C, Niskayuna, NY 12309(518) 387-5670, fax: (518) 387-6675, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected]

Field Technical Manager: Jules Routburt(630) 252-5065, fax: (630) 252-4289, e-mail: [email protected]

Contractor: GE Global ResearchContract No.: DE-FC04-2002AL68080

Objective• Reduce the fuel consumption of off-highway vehicles, specifically large

tonnage mine haul trucks, as shown to the right. (A hybrid energy storageand management system will be added to a conventional diesel-electrictruck that will allow capture of braking energy normally dissipated in gridresistors as heat. The captured energy will be used during acceleration andmotoring reducing the diesel engine load, thus conserving fuel.)

Approach• The project will work towards an on-board demonstration of the hybrid system by first selecting an energy

storage subsystem and energy management subsystem.

• Laboratory testing at a subscale level will evaluate these selections and then a full-scale laboratory test will beperformed.

• After the subsystems have been proven at the full-scale lab, equipment will be mounted on a mine haul truckand integrated with the vehicle systems.

• The integrated hybrid components will be exercised to show functionality, capability, and fuel economyimpacts in a mine setting.

Accomplishments• Completed subscale battery testing. Testing confirmed operation of battery cells for mine haul applications and

provided validation for the vehicle battery model.

• Improved vehicle model by applying coastdown data collected from several trucks at Komatsu’s ProvingGround.

• Preliminary energy system enclosure and mounting designs for several truck models is completed.

• Started design of full-scale prototype and specified components. Several components have been received andassembly of prototype has commenced.

Future Direction• Following the on-board demonstration, work could continue investigating the commercialization paths and

next-generation energy storage systems.


165

• Beyond improving fuel economy, the use of hybrid technology for reducing emissions would also bedeveloped.

• By considering the energy storage system and related hybrid components during design, the next-generationhybrid mine haul truck will provide increased benefits in many areas.

IntroductionThe conventional mine haul truck, also referred to asan Off-Highway Vehicle (OHV), uses a dieselengine to turn an alternator that generates alternatingcurrent (AC) electricity. The electricity is rectified toDC voltage that is applied to the main electrical buscalled the DC Link. Power electronics convert theDC link voltage to variable frequency, three-phaseAC that drives the wheel motors during motoring.When braking is required, the wheel motors functionas generators and electrical power is directed to theBraking Grid Resistors, also known as the “GridBox,” to be dissipated as heat.

In addition to traction motors, the diesel engine mustpower auxiliary loads, such as radiator cooling fans,operator air conditioning, steering, hydraulics, andcontrol circuits. These auxiliary loads are poweredby mechanical means and with a smaller, secondalternator connected to the vehicle’s 24-V batterysystem.

Figure 1 shows a block diagram of the plannedhybrid vehicle system. Three primary componentswill be added to the conventional OHV: EnergyManagement System, Energy Storage ElectronicInterface, and Energy Storage System. Eachcomponent plays a specific role in the recovery ofbraking energy.

Figure 1. Hybrid electric system block diagram.

The Energy Management System (EMS) performsthe high-level supervisory functions of the hybridvehicle, primarily controlling the balance of enginepower, grid dissipation, and battery charging ordischarging. Performing these functions requiresintimate connections to all of the truck systems.

The Energy Storage Electronic Interface, alsoreferred to as the Hybrid Control Group, acts toconvert electrical power on the DC link into levelscompatible with the Energy Storage System. Thehybrid control group is essentially a high-powerDC-DC converter.

The Energy Storage System (ESS) consists of thebatteries and manual disconnects, fuses, and relatedsafety and protective components.

Subscale Battery TestingTo investigate the performance of the target batteryfor the demonstration, individual cells were tested ina subscale apparatus. Subscale testing allowsthorough instrumentation of battery cells and tightcontrol of battery current. Figure 2 shows thelaboratory setup for testing the “hardware-in-the-loop” experiments.

Power flowData & control

Engine+

Alternator+

Rectifieremulation

TractionDrive emulation

DC-DC Converter

Na-NiCl2 batterymodule

NiCd battery modules

Controller withembedded mission analysis. “Drives” the mission profile

Engine Power

Voltage, current and temperature

Battery Power

Typical battery power traces

Power flowData & control

Engine+

Alternator+

Rectifieremulation

TractionDrive emulation

DC-DC Converter

Na-NiCl2 batterymodule

NiCd battery modules

Controller withembedded mission analysis. “Drives” the mission profile

Engine Power

Voltage, current and temperature

Battery Power

Typical battery power traces

Figure 2. Subscale laboratory configuration.


166

One DC source emulates a diesel engine/alternator-rectifier that supplies electrical power to the DClink. Another bidirectional DC power sourceemulates the traction drive, with capability to act inmotoring (power sink) and retard (power source). Atwo-channel bidirectional DC-DC converterindependently controls power flow between thebattery modules and the DC link.

The controller commands the alternator and tractiondrive power flows as defined by the hybrid truckmission. The controller also monitors power flowwithin the system and implements the energymanagement algorithms. Data are acquired andstored for display on a graphical user interface andfurther analysis.

Figure 3 shows one instance of hybrid performancecompared to the baseline vehicle speed asdemonstrated by using the subscale test equipment.The hybrid system is switched on during the uphillportion of a mine haul while the truck is fullyloaded. The additional power supplied by the hybridsystem results in additional motoring poweravailable at the wheels and allows approximately3 MPH faster speed-on-grade.

180 200 220 240 260 280 300 320

6

8

10

12

14

16

Spee

d (m

ph)

Hybrid switched on at 253sec

Time (sec)

6

8101214

16

200 300

Hybrid

Baseline

180 200 220 240 260 280 300 320

6

8

10

12

14

16

Spee

d (m

ph)

Hybrid switched on at 253sec

Time (sec)

6

8101214

16

200 300

Hybrid

Baseline

Figure 3. Hybrid performance.

The higher speed-on-grade is important whenconsidering the impact of hybrid system weight onproductivity. Any hybrid system weight must besubtracted from the working payload of the truck tomaintain the same gross vehicle weight (GVW).Because the truck is moving downhill as fast aspossible, productivity, expressed in tons/hour, canonly be maintained by decreasing the uphill haultime. The hybrid system allows this while savingfuel.

The initial configuration was planned to utilize twobattery banks, each comprised of different

chemistries. This configuration was chosen toreduce risk in using an advanced battery.

Figure 4 shows the results of one subscale test oftwo battery types. One complication of this systemis in balancing the state-of-charge (SOC) for eachsystem. Charge balance must be enforced during theoverall haul cycle or the system will deplete orovercharge. The right-hand graph of Figure 4illustrates one system maintaining proper balancewhile the other (magenta) overcharges.

Battery Power

-30

-25

-20

-15

-10

-5

0

5

10

0 200 400 600 800 1000

Time (sec)

Pow

er (K

W)

P (Zebra)P (NiCd)

Battery Power

-30

-25

-20

-15

-10

-5

0

5

10

0 200 400 600 800 1000

Time (sec)

Pow

er (K

W)

P (Zebra)P (NiCd)

Terminal Energy

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0 200 400 600 800 1000

Time (sec)

Ener

gy (K

WH

)

Energy (Zebra)Energy (NiCd)

Terminal Energy

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0 200 400 600 800 1000

Time (sec)

Ener

gy (K

WH

)

Energy (Zebra)Energy (NiCd)

Figure 4. Two-battery ESS.

Controlling this SOC balance adds complexity to theEMS. One solution is to utilize one batterytechnology for the ESS. A single-battery system iscurrently being planned for the truck demonstration.

The battery application has an impact on life. Thesubscale tests have completed over 12,500 cyclesover 9 months, demonstrating excellentperformance. The life tests were cut short because ofa long-term drift of SOC that resulted inoverdischarge of the cells.

Full-Scale Battery TestingFollowing the subscale evaluation, full-scale testingwill be performed at GE Transportation in Erie, PA,where the testing facility has recently beenupgraded. Facilities are available to emulate allvehicle systems in a controlled environment,allowing instrumentation of the system for analysisand troubleshooting (Figure 5).

Components for the full-scale testing have beenspecified, and most have been ordered and received.Electrical schematics were generated to show theintegration of the hybrid components with theexisting vehicle design.

Batteries were received and tested for functionalityand performance before they were installed in thefull-scale setup. Part of the testing checked the


167

Figure 5. Control room and powerelectronics cabinet.

operation of the communication bus (CAN) used bythe batteries and supervisory control. Several issueswere discovered in integrating lab equipment, thetruck’s vehicle controller, and the EMS. The batteryvendor has been especially helpful in resolving thecommunication gaps. A bank of four batteries hasbeen connected successfully in parallel for testing.Hardware to allow the vehicle controller to interfaceto the EMS and ESS has been received and will betested shortly.

Full-scale evaluation work will continue intoFY 2005, leading up to the integration of the hybridcomponents on the mine haul truck (Figure 6).

Figure 6. Mine haul truck with hybrid equipment shownin purple.

Vehicle IntegrationIn preparation for the truck integration, layouts ofhybrid components on the truck have been created.Several configurations are being discussed, andKomatsu is analyzing structural requirementsneeded to support the additional hybrid weight.

Five general areas of the truck are available to addhybrid equipment: RF upper deck, front bumper,behind operator cap, and between the frame rails.The preferred configuration for mounting the hybridsubsystems is to locate one battery bank on the RFdeck. A second battery bank will be located on thefront bumper. The hybrid control group will sitbehind the operator cabinet, close to the maincontrol group.

The hybrid system will require at least one reactor tosmooth current from the hybrid control groupswitching electronics. The reactor will be located theinductors between the frame rails, beneath the dumpbody.

Another design uses the space between the framerails for a battery bank, but this moves the inductorsto the upper deck in front of the control cabinet. Thisconfiguration is not desirable as it clutters the deckand will make access to the main control cabinet andupper battery bank difficult.

ConclusionsThe work performed in FY 2004 has shown thefeasibility of the hybrid demonstration, and theexpected benefits are sufficient to continue worktoward the on-board demonstration.

The subscale laboratory testing has providedvaluable experience with advanced batterytechnology and has added confidence to the vehiclemodeling and performance estimates.

Initial designs of the vehicle integration show thereis sufficient space on the truck for mounting ahybrid energy system and controls. The designssuggest a clean installation with room for changes asthe design matures.

Full-scale laboratory testing is well under way, withprimary components purchased and received andschematics complete. Construction of test equipment


168

specific to the full-scale testing has begun andacceptance testing of the batteries is nearlycomplete.

The demonstration at Komatsu Proving Grounds,shown in Figure 7, will show operation of the hybridOHV and validate model estimates for energysavings.

Figure 7. Komatsu Proving Grounds.


169

IX. Particulate Matter Characterization

Morphology, Chemistry, and Dynamics of Diesel Particulates

Principal Investigator: Kyeong O. LeeArgonne National Laboratory9700 S. Cass Ave.Argonne, IL 60564(630) 252-9403, fax: (630) 252-3443, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-803, fax: (202) 586-1600, e-mail: [email protected] Program Manager: Jules Routbort(630) 252-5065, e-mail: [email protected]


Objectives• To investigate the evolution of diesel particulates along the exhaust pipe.

• To provide technical data for the development of advanced diesel aftertreatment systems.

Approach• Collect diesel particulates along the exhaust pipe from a light-duty diesel engine.

• Analyze morphological parameters along the exhaust pipe.

Accomplishments• Developed a thermophoretic soot sampling device that can collect soot particles at four different positions along

the exhaust pipe.

• Analyzed the morphology, sizes, fractal geometry, and chemistry of diesel particulates along the exhaust pipe.

• Identified the significant changes in particle morphology, sizes, and fractal geometry of diesel particulatescaused by the oxidation catalysts, as well as other exhaust components along the exhaust pipe.

• Found catalytically reacted particulates immediately after the oxidation catalyst, indicating the importance ofthis research for development of advanced diesel aftertreatment systems.

Future Direction• Analyze morphological parameters immediately before and after a diesel particulate filter installed on the

exhaust system.

• Analyze parameters as different properties of diesel fuel are used, e.g., different levels of sulfur, paraffins andaromatics concentrations, oxygen concentration, and Cetane numbers.

• Compare those parameters with those measured for gasoline engine particulates.


170

IntroductionParticulate matter (PM) is a chemically andphysically sophisticated substance that is formed viaa series of rapid thermal and chemical reactions as aby-product of diesel engine combustion.Measurement of the physical dimensions of dieselparticulates is an important issue, because theirnano-size characteristics are directly related tohuman health, particularly in the pulmonary system.The size measurements need to be carried out fortwo different types of particles: primary andaggregate particles. In combustion physics, thegrowth of individual primary particles is responsiblefor the total mass of particulate emissions, while theaggregate particles, formed through agglomerationof primary particles and smaller aggregates, areemitted to the atmosphere and may affect humanlife. Indeed, the particle growth throughagglomeration is irrespective of the change in totalparticulate emissions. Therefore, measurement ofboth sizes is necessary.

Many investigators reported the size measurementsof diesel particulates, typically in a form ofaerodynamic or mobility diameter. These sizemeasurements of diesel aggregate particles wereconducted mostly by using commercial instruments.In principle, these instruments measure theequivalent diameters (volume-based) of aggregateparticles; they assume that the particles arespherical. These measured particle sizes are, ingeneral, classified into two different size categories– nucleation and accumulation modes – at athreshold size of 50 nm, (i.e., the particles smallerthan 50 nm are considered to be in the nucleationmode and the ones larger than 50 nm are in theaccumulation mode). In our previous experimentsperformed for a light-duty diesel engine, the averagesizes of primary particles were appeared to be in arange of 20–30 nm in diameter at various engine-operating conditions, under which the primaryparticles should undergo the growth and oxidationprocesses during combustion. The nucleation ofincipient soot particles occurs at a very early stageof soot formation; the sizes of these particles aremuch smaller than the average sizes of the particlesthat are typically measured after the termination ofmajor combustion reactions. In combustionexperiments for a laminar diffusion flame, theincipient soot particles in nucleation, which are

usually detected in the low-temperature region,appear to be extremely small and are known to bechemically unstable. To reveal the detailedproperties of complex diesel particulates, therefore,accurate measurements of physical dimensions andmorphology of diesel particulates are necessary.

In addition to the measurement of particle sizes, theefficient control of diesel particulate emissions isanother important issue that has been discussedextensively. Many researchers have attempted toreduce diesel particulate emissions using varioustechnologies: the development of in-cylindercombustion control systems, improvement of enginehardware, and development of a new concept ofdiesel engine combustion. Recently, the use ofaftertreatment systems, such as particulate filters oroxidation catalysts, has been considered as a moresubstantial and practical technology to control dieselPM emissions. Detailed characterization of dieselparticulates along the exhaust system will benecessary to develop advanced aftertreatmentsystems.

In this investigation, we observed the morphology ofdiesel particulates on the PM samples collected atdifferent locations along the exhaust system byusing a high-resolution transmission electronmicroscope (TEM). The samples were collected byusing a thermophoretic soot sampling device firstdeveloped at Argonne. The sizes of both primaryand aggregate particles were measured along theexhaust pipe. The geometry of the particulates wasstandardized at each sampling location throughfractal analysis. Detailed investigation of particlesize and morphology along the exhaust system israre in the literature.

Experimental DescriptionsEngine Specifications and PM SamplingA 1.7-L light-duty diesel engine was used to samplePM at four different positions along the exhaustsystem. The specifications of the engine used arelisted below:

Displacement (4 cyl.): 1.7 LAspiration: Turbocharged and inter-

cooledCompression ratio: 19:1Rated power: 66 kW at 4,200 rpm


171

Rated torque: 180 Nm at 1,600–3,200 rpmMaximum speed: 4,800 rpmFuel injection system/ Common-rail, direct-injection pressure: injection/200–1,400 barTest fuel: California low-sulfur

commercial diesel fuel,HF-328

A thermophoretic sampling system was used tocollect particulates from the engine exhaust system.Detailed descriptions for the use of this systemappeared in our previous publications. The samplingtechnique offers several unique advantages, such asrapid sampling (as fast as 20 ms max.), minimalinterference of sampling events to the exhauststream, no need for air dilution, and no extratreatments for subsequent morphological andchemical analyses. The samples collected fromengines are directly inserted into the TEM toobserve the morphology of collected particles andobtain their images.

Diesel PM was collected from four differentpositions along the exhaust system: right after theexhaust valves (Port 1 or P1), between theturbocharger and first oxidation catalyst (Port 2 orP2), between two oxidation catalysts (Port 3 or P3),and further downstream, removed from the secondcatalyst by about 200 cm (Port 4 or P4), as shown inFigure 1. The engine operating condition forsampling was 2,500 rpm and 25% of the maximumtorque. To obtain consistent data, the engine waswarmed up to stabilize its operation at a mediumspeed/load condition for about 40 minutes prior toeach experiment. During each sampling experiment,other important parameters, such as the temperatureand pressure of exhaust gas, cooling water, and fuel,and equivalence ratios, were monitored to ensureconsistent sampling conditions. Particles werecollected at the four different positions at one time.Because the exhaust gas temperature at eachsampling position was different, the residence timeof the sampling probe in the exhaust stream had tobe optimized to enable us to collect a sufficientnumber of particles while avoiding possibleoverlapping or additional agglomeration betweenparticles.

Figure 1. Schematic of PM sampling positions along theexhaust system.

TEM Image Analysis and Energy DispersionSpectroscopyA Philips CM30 TEM was used to observeparticulate samples and photograph the images.Then the TEM micrographs were digitized with ahigh-resolution CCD camera and a dataacquisition/image processing system, and detailedmorphological properties of each particle — such asthe primary particle diameter, radius of gyration, andfractal dimension — were statistically measured toprovide average values. An energy dispersionspectrometer (EDS) was used to analyze chemicalelements dissolved in sample particles with adetection area as small as 95 nm in diameter.

Results and DiscussionMorphological ObservationNumerous soot particles were observed through aTEM and photographed to characterize the evolutionof particle morphology and sizes along the exhaustsystem. Figure 2 shows the TEM images of thediesel particulates collected at the four differentsampling positions depicted in Figure.1. All fourmicrographs have the same magnification (21,000×)to allow a direct comparison of particle propertiesamong the images. Each micrograph displayed wascarefully selected to exhibit a representative trend ofmorphological characteristics of the particulatescollected at a specific sampling position. As seen inthe figure, a common morphological characteristicwas found: small near-spherical primary particlesagglomerate to form a larger aggregate particle, andthose aggregate particles appear to be stretched,chain-like structures. An interesting observation isthat the particles collected at Port 1 (P1, on theexhaust manifold) appeared to be larger in overallaggregate size and higher in particle number densitythan the ones collected at other sampling ports. In


172

(a) (b)

(c) (d)

Figure 2. TEM images of diesel particulates collected atdifferent positions along the exhaust pipe, (a) Port 1,(b) Port 2, (c) Port 3, and (d) Port 4.

particular, the number density and sizes of particlescollected at Port 3 and 4 were quite different fromthe others. This observation indicated that theoxidation catalysts installed on this engine had asignificant influence on the size and concentration ofparticulate emissions, which will finally reduce thetailpipe PM emissions.

Particle SizesMeasurement of primary particle sizes is importantbecause soot yield is determined basically by thesizes and number of particles. Also, detailedmechanisms of the particle growth and oxidationthat occur during combustion processes can possiblybe explained with further analysis of themicrostructures and chemistry of the particles. Tobetter understand particle sizes, a schematic of anaggregate particle was developed (Figure 3).A chain-like aggregate particle consists of anagglomeration of tens to hundreds of near-sphericalprimary particles, where the size of a spherule isrepresented by a diameter of dp. The size of anaggregate particle is represented by the radius ofgyration, Rg, which is more reasonable to represent

Rg

dp

Figure 3. Schematic of an aggregate particleconsisting of near-spherical primary particles.

the size of complex-shaped diesel particulates thanother equivalent diameters.

The sizes of primary particles were measured fromTEM micrographs by using a digital imageprocessing system. From the soot images, more than200 primary particles were randomly selected fromdifferent aggregates to determine an averagediameter of primary particles at a specific samplingposition. In most cases, the size distribution ofprimary particles represented a standard Gaussiandistribution. Figure 4(a)–(d) shows the histograms ofprimary particle diameters measured at Ports 1through 4, respectively. As seen in the figures,primary particles in a size range of 6–42 nm wereselected to measure average diameters at fourdifferent sampling positions, which appeared to be28.6 nm, 26.8 nm, 21.8 nm, and 19.1 nm at Ports 1through 4, respectively. It is apparent in the figurethat the sizes of particles representing a majority ofcounts shifts toward smaller sizes as the samplingposition moves downstream along the exhauststream. The average diameter measured at Port 4,where sampled particulates should have experiencedoxidation after traveling through two oxidationcatalysts, was decreased by 33% compared to theone measured for engine-out particulates at Port 1. Itwill be interesting to compare the reduction ofprimary particle size with measurements of overallconcentrations of particulate emissions, in light ofassessing the contribution of particle size reductionthrough oxidation to total particulate emissions.

The spheroidal shape of primary particles wasverified by measuring the diameters of primaryparticles constituting an aggregate particle at twodifferent viewing angles, 0° and 50°. The


173

0

10

20

30

40

50

60

70

14 18 22 26 30 34 38 42 46

dp (nm)

Avg dp = 28.6 nm

0

10

20

30

40

50

60

70

14 18 22 26 30 34 38 42 46

dp (nm)

Avg dp = 26.8 nm

(a) (b)

0

10

20

30

40

50

60

70

80

4 12 20 28 36 44

dp (nm)

Avg dp = 21.8 nm

0

10

20

30

40

50

60

70

4 8 12 16 20 24 28 32 36

dp (nm)

Avg dp = 19.1 nm

(c) (d)

Figure 4. Size distributions of primary particles sampledat (a) Port 1 through (d) Port 4, respectively.

sample was rotated in the TEM to obtain sootimages at a 50o viewing angle. Figure 5 shows thesize distributions of primary particles at 0° and 50°viewing angles, respectively. The particulatesamples were collected at the exhaust manifold at2,500 rpm and 70% load. The measured averageparticle diameters of 24.6 nm and 24.4 nm differ byless than 1%, which implies that the measurement ofprimary particles is not sensitive to orientation andthat primary particles are fairly spherical in shape.

The sizes of aggregate particles were measured interms of radius of gyration that represents a physicaldimension. The radius of gyration (Rg) of a singleaggregate particle is defined by the followingequation:

∑=

=n

iig r

nR

1

21 (1)

where rI is the distance between the center of anindividual primary particle and the centroid of theassociated aggregate (see Figure 3), and n is thenumber of primary particles within a singleaggregate particle. For a specific engine condition,

0

20

40

60

80

100

0o angled

p,avg= 24.4 nm

50o angled

p,avg= 24.6 nm

1 2 3 4 5 6 7

dp (nm)

10-15 15-20 20-25 25-30 30-35 35-40

Figure 5. Size distributions of primary particles measuredat different viewing angles, 0° and 50° (2,500 rpm/70%load).

more than 200 aggregate particles were randomlychosen to determine the number average of Rgs.

Figure 6 shows the size distributions of aggregateparticles sampled from four different sampling ports.A bin of the particle size representing a majority ofparticles tends to shift to a smaller radius ofgyrations along the exhaust pipe; for example, from80 nm at Port 1 to 40 nm at Port 4. The largestparticle size measured at each port appeared to besmaller than Rg = 300 nm (a diameter of 600 nm),although the particles larger than this size were notselected for measurement because of their rarity.These size measurements revealed that the physicalsizes of a majority of diesel particulates, particularlynear the tail pipe (Port 4), are distributed in quite anarrow size range, close to the lower limit of theU.S. Environmental Protection Agency’s (EPA’s)size standard of PM1.0 and far from the upper limit of1.0 micron.

Indeed, the sizes of a majority of particles at thetailpipe are close to those of PM0.1 particles. On thebasis of these physical size data, none of theparticles in the size ranges of EPA standards PM2.5or PM10 were emitted from this diesel engine.

Fractal GeometryThe morphology of highly complex dieselparticulates can be standardized by evaluating fractaldimensions in fractal analysis. Fractal dimensionsdemonstrate the growth mechanism that dominatesthe formation of aggregate particles during a


174

0

50

100

150

200

0 40 80 120 160 200 240 280 320

P# 1P# 2P# 3P# 4

Rg (nm)

Avg. Rg (nm)

101.683.756.149.6

Figure 6. Size distributions of aggregate particles at fourdifferent positions along the exhaust system.

combustion process. The fractal dimension (Df) isdefined as

fD

p

gf d

Rkn ⎟

⎟⎠

⎞⎜⎜⎝

⎛= (2)

where n is the total number of primary particleswithin an aggregate particle, kf is a prefactor, Rg isthe radius of gyration of an aggregate, dp is theaverage primary particle diameter, and Df is themagnitude of fractal dimensions. By plotting ln(n)versus ln(Rg/dp), the fractal dimension can beobtained by evaluating the gradient of linearregression line.

Figure 7 shows the linear regression lines for theparticulate data collected at the four differentsampling positions. The hollow circles display thedata for particulates sampled at Port 1 (other datawere omitted). The gradients of the regression lineswere evaluated at 1.62, 1.52, 1.73, and 1.53 alongthe exhaust pipe, respectively (The regressioncoefficients for all the fitting lines are above 0.95.).The magnitudes of these fractal dimensions lie in thesame range as the ones previously measured forlight-duty diesel engine particulates at variousengine-operating conditions. This result supports theformer conclusion that the light-duty diesel engineproduces more chain-like stretched particles thandoes the heavy-duty diesel engine.

Figure 8 displays data for the average diameters ofprimary particles, the radius of gyrations ofaggregate particles, and the fractal dimensions at allfour sampling positions. As shown in the figure, the

0

1

2

3

4

5

6

-0.5 0 0.5 1 1.5 2 2.5 3

P#1P#2P#3P#4

ln(Rg/d

p)

Df

1.621.521.731.53

Figure 7. Evaluation of fractal dimensions of theparticulates collected at four different positions along theexhaust pipe.

0

20

40

60

80

100

0

1

2

3

4

Port 1 Port 2 Port 3 Port 4

Rg

dp

Df

Figure 8. Average sizes of primary and aggregateparticles and fractal dimensions at four different samplingpositions.

primary particle size (dp) gradually decreased alongthe exhaust pipe: the size was reduced by 33%, from28.6 nm at Port 1 to 19.1 nm at Port 4 (tailpipe). Theradius of gyrations of aggregate particles alsodecreased quite significantly as particles movedownstream along the exhaust pipe: about a 50%reduction along the entire exhaust system (101.6 nmto 49.6 nm). These reductions in particle sizesoccurred after particles traveled through oxidationcatalysts, where they should have undergonecatalytic oxidation. Detailed information about theoxidation catalysts is not available at present.Meanwhile, the magnitude of fractal dimensions (Df)slightly increased at Port 3, while the Df values atother ports are quite even. The larger Df indicates


175

that these particles are more spherical in shape thanthose sampled from other ports.

These analyses trigger an interesting discussionabout the possible mechanisms of particle evolutionin the exhaust system. The particles formed incombustion chambers were emitted to the exhaustsystem and underwent further thermal reactions andaerodynamic interactions with a fairly hightemperature of gaseous emissions and particulates inthe system. As a result, their sizes were significantlyreduced: the sizes of aggregate and primary particleswere reduced by 50% and 33%, respectively, alongthe exhaust pipe. Note that the reduction percentageof aggregate particle size is larger than that ofprimary particle size. The following hypothesis mayaccount for the particle evolution: both a partialbreakdown (or disintegration) of single primaryparticles from an aggregate and the reduction ofinterstitial distances between primary particles in theaggregate can reduce the aggregate particle size.These two physical processes are explained mainlyby aerodynamics in the exhaust stream and particleoxidation in the catalysts, respectively. Ourmicroscopic observations support this hypothesis:the particulates emitted immediately from thecatalysts appeared to be smaller in size and morespherical and compact in shape, particularly atPort 3, where the catalytic oxidation should haveoccurred.

The chemistry of particulates sampled at an idlingspeed was analyzed by using an EDS. As seen inFigure 9, the major portion of diesel PM consisted ofelemental carbons, while a few percent of sulfur wasalso detected from this particulate sample. Althoughquantitative measurement of these chemicalelements is still required for comparison, the sulfurcontent measured from these particulates was quitesignificant, considering the sulfur concentration of110 ppm in the fuel. An accurate calibration with astandard specimen of carbon and sulfur is requiredto measure the absolute amount of sulfur dissolvedin diesel PM. The copper and silicon were detectedfrom the soot sampling grid, not from particulates.

Figure 9. Spectrum of chemical elements in dieselparticulates measured by an EDS.

ConclusionsThis microscopic study, which basically used aseries of PM sampling and measurement systems,including a thermophoretic soot sampling system, aTEM, and an image processing system, wassuccessfully conducted to analyze the morphology,sizes, and fractal geometry of light-duty dieselparticulates at various positions along the exhaustsystem. The sizes of both primary and aggregateparticles were significantly reduced, mainly byoxidation catalysts. The degree of size reduction ofthe aggregate particles exceeded that of the primaryparticles through the entire exhaust system. It islikely to be attributed to particle breakdown byaerodynamic interactions in the exhaust stream andreduction of interstitial distances between primaryparticles caused by catalytic oxidation. The EDSanalysis revealed that a majority of the dieselparticulates consists of elemental carbons, alongwith a significant amount of sulfur.

AcknowledgmentsThe authors would like to acknowledge strongsupport from Dr. Sidney Diamond at theFreedomCAR and Vehicle Technologies Program ofthe U.S. Department of Energy. We also thank Dr.James Eberhardt at DOE for in-depth discussionsrelevant to the environmental safety and healthimpacts of diesel PM.

C

OCu

Si S

Element Atomic % C 96.8 O 0.7 S 0.1 Si 0.2 Cu 2.2


176

X. Brake Systems

Advanced Brake Systems for Heavy-Duty Vehicles

Project Manager: Glenn GrantPacific Northwest National LaboratoryP.O. Box 999, Richland, WA 99352(509) 375-6890, e-mail: [email protected]

Technology Development Manager: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected]

Field Technical Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4798, e-mail: [email protected]

Contractor: Pacific Northwest National LaboratoryContract No.: DE-AC06-76RL01830

Objective• Assess the potential of alternate disc brake designs and technology concepts, in conjunction with emerging

friction materials. The expected outcome of this project is identification of an appropriate disc brake systemdesign and friction pair sets, with improved thermal, wear, and durability properties that can be applied toheavy-duty vehicles for enhanced braking performance and improved safety. The emerging innovative conceptswill be examined in collaboration with brake suppliers and/or truck original equipment manufacturers (OEMs).

Approach• Disc brake design and thermal management, which will be achieved through a combination of theoretical

modeling and experimental investigation and validation.

• Enhanced performance and durability of friction pairs, which will be achieved through a combination oftheoretical modeling and experimental investigation and validation.

• Alternate or auxiliary retarding/stopping methods, which will be achieved through an evaluation of emergingbraking systems and a concept feasibility study of novel methods to scrub kinetic energy from a movingvehicle. Methods that show promise will be presented to brake suppliers and OEMs and investigated further asappropriate.

Accomplishments• Developed a disc-brake rotor model for thermal performance studies and evolution of rotor design concepts and

waste-heat recovery methods.

• Selected candidate friction materials.

• Initiated friction and wear evaluation of candidate materials at Tribo Materials and Rockwell Scientific.

• Demonstrated friction stir reaction processing (FSRP) as a manufacturing method for friction materials thathave a selectively graded, wear-resistant structure.


177

Future Direction• Investigate multi-disc brake rotor arrangements and other new design concepts and determine benefits and

sensitivity of each to braking performance.

• Identify alternate energy absorption/conversion and heat management/rejection methods for potential systemconcept redesign.

• Evaluate of system response to noise-vibration-harshness (NVH) for friction pair combinations.

• Further down-select of friction materials based on wear rate, thermal stability, friction characteristics, NVH,and cost.

• Coordinate with OEMs and create a project steering committee.

IntroductionReducing parasitic losses (primarily drivetrainfriction and rolling resistance) and aerodynamicdrag improves the fuel efficiency of trucks andbuses. However, such reductions in drag make itmore difficult to stop the vehicles, and place ahigher burden on the vehicle brake materials andsystem. Higher brake loads, combined with thedesire and legislation to reduce the stoppingdistance of trucks and buses, mean that theperformance of the braking systems must beimproved substantially over what is availabletoday. The desire to reduce vehicle weight andoperating costs also calls for new brake materialsthat are lightweight and last longer.

The trend is toward higher operating temperatures,beyond the safe envelope of the current generationof brakes that use gray cast-iron, and in particularthe phenolic binder friction materials (brake pads).The consequence of using such conventionalmaterials in heavy vehicles is poor and unreliableperformance, excessive wear, and more weardebris. These lead to increased environmentalimpact, higher maintenance costs, and, ultimately,lower safety and productivity.

For future vehicles, especially on-highway long-haul vehicles (Class 7 and 8 trucks), it is likelythat conventional propulsion platforms will beused; whereas for city and military operations,hybrid and/or electric propulsion systems will beprevalent. Both of these systems will likely requirewheel-mounted friction brake systems.

The main operational difference is that, for on-highway long-haul vehicles, most of the braking is

expected to be accomplished by the frictionbrakes; for hybrid systems, most of the brakingwill be done by regenerative braking using theelectric drive motors as generators, and thebalance will be done by a wheel friction brake.The latter could reduce the brake temperature forvehicles operated in stop-and-go mode, comparedto today’s brakes that run hotter and have a muchshorter lifespan between replacements than brakesfor on-highway long-haul vehicles.

Unfortunately, it appears that U.S. suppliers ofheavy vehicle brakes are not aggressively pursuingresearch on new materials or designs for discbrakes, potentially because of their vested interestin drum brake systems. This project will addressdisc brake system design and development needs,as well as friction pair and brake rotor materialneeds that can be applied to disc brake systems.

ApproachThis project addresses the broad goal of producinghigher-performance braking systems for heavyvehicles with a multifaceted approach. Aninvestigative and development approach aimed atdetermining the feasibility of new disc brakesystem designs and application methods will bepursued. The main project tasks include:(1) development of novel disc brake designs andthermal management approaches, (2) enhancedperformance and durable friction pairdevelopment, and (3) exploration of the conceptfeasibility of alternate or auxiliaryretarding/stopping methods.


178

Fiscal year 2004 was the first of a multi-yearprogram, and as such, a brief description of theproject scope is warranted.

Task 1 – Disc Brake Design and SystemModelingWith the need to improve braking performanceand durability beyond the 500,000-mile minimumexpected life of the vehicle, new friction materialswill be required. However, direct materialsubstitution from current drum brake systems toadvanced disc brake geometries is not likely to beeffective. As such, disc brake design and systempackaging are critical to achieving the projectobjectives. This task will focus on developing andevaluating innovative disc brake designs, with theemphasis on enhanced thermal management.

Modeling methods, such as computational fluiddynamic (CFD) approaches, will be used to assistin the design and evaluation of novel discgeometries and system components. The goal ofthe modeling will be to develop a functional solidmodel and use predictive tools to simulate theperformance (thermal and fluid profiles, as well asstress analysis) of the proposed disc brakesystems.

Some of the brake system design concepts thatwill be explored include, but are not limited to,floating single or twin rotors, full-phase pad/rotor,and mechanical/electric actuation.

Task 2 – High-Performance Friction PairsFor this task, we will acquire several candidaterotor and friction materials and evaluate them invarious friction pair combinations that simulatetypical braking scenarios for on-highway andvocational heavy-duty vehicles. The friction andwear will be measured and the failure mode(s)determined. The tribology assessment will be doneby Tribo Materials at Rockwell ScientificCompany (RSC). The relevant physical andmechanical properties of the most promisingmaterials will be measured by PNNL, which willalso complete an assessment of manufacturingtechnologies for metallic and ceramic compositematerials. These data will delineate the safeoperating envelope of the materials that is neededby the brake designers to determine where and

how the materials can be employed for current andnext-generation braking systems.

The outcome of this task will be to provide brakedesigners with relevant data on new materials,enabling them to explore new brake concepts that,until now, have not been possible with currentlyavailable gray iron and phenolic-binder frictionmaterials.

Task 3 – Concept Feasibility of AlternativeBraking TechnologiesThe third task is generally focused on analyzingspecific concepts that may be employable forbrake systems to either enhance brakingefficiency, improve thermal management offriction surfaces, or recover and store of wasteheat for use at a later time to drive an electricalmotor/pump or in the propulsion of the vehicle.

Task objectives will be achieved throughevaluation of emerging braking system concepts— such as the use of electromagnetic techniques— to scrub kinetic energy from a moving vehicle.

Thermal management is critical in brake systemdurability and maintaining adequate frictionperformance. Methods such as employing duel-phase materials will be investigated, in which thefrictional energy from braking is used to make aphase change within a rotor, instead of generatingheat that must be conducted or convected awayfrom the friction surface. Methods that showpromise will be presented to brake suppliers andOEMs and investigated further as appropriate.

ResultsFriction pair selection is a critical aspect ofdesigning a disc brake system. As such, frictionpair evaluation and selection for a high-performance, high-durability heavy-vehicle discbrake system was the predominate focus for initialproject tasks. In addition, a simplified disc brakerotor model was developed to (1) evaluate rotorgeometries, component sizing, and location and(2) assist in material selection.


179

Task 1 – Disc Brake Design and SystemModelingA model was developed for a simplified disc brakerotor geometry that consists of a solid disc andbrake pad. This model was created to predict thefluid distribution (airflow) around the brakeassembly and to allow thermal profile and stressanalysis. The model will be used to establishdesign limits (bounding loads) in potential discbrake designs. The thermally induced stresses areproportional to spatial and temporal gradients intemperature that develop during braking andcooling.

The analysis and CFD simulation predict disctemperatures and heat flux during a maximumbraking stop from highway speed. This analysisincludes essential features of the vehicle wheelwell airflow that determine convective coolingrates at the disc surface. Full-vehicle CFDsimulations from manufacturers and otherFreedomCAR projects will be used, as available,as guidance for the convective cooling portion ofthis model. Radiation heat transfer is also expectedto significantly affect disc cooling. Conductionfrom the rotating disc to the caliper, hub, andwheel components is also accounted for in thesimulation. The simulation employs a two-stepapproach in order to minimize time and computingrequirements. StarCD was selected for the CFDpart of this application. This software has beenused by Volvo for a similar disc brake coolingapplication. PNNL results of near-disc flow for apreliminary isothermal simulation of a disc brakeare shown in Figure 1.

Because the CFD analysis will be a time-consuming transient simulation, selected resultsfrom StarCD will be used to form inputs to anadditional thermal analysis tool that can be usedefficiently for preliminary/scoping sensitivityanalyses. The tool selected for this purpose isRadTherm, because it is widely used by industryfor brake thermal analysis.

RadTherm is typically used to predict componenttemperature distributions during repetitive brakingtest sequences. This package is capable ofsimulating simple radiation heat transfer fromsurfaces and also includes the capability of

Figure 1. Simplified geometry and flow-fields resultsfor preliminary disc brake simulation using StarCD.

simulating conduction through solids. Theefficiency of this tool is realized by not solving forcooling airflow, but by applying convective heattransfer as surface heat transfer coefficients. Theseheat transfer coefficients are obtained from thepreceding StarCD CFD simulation. Sensitivity toparameters such as disc thickness, thermalradiation emissivities, or thermal contactresistances can be investigated with this softwareto better define the bounding conditions for thestress analysis.

The time history of predicted temperatures in thedisc will be transferred to the stress analysissoftware to determine worst-case thermallyinduced stress loading. The mesh geometry andassociated results sets from the CFD analysis willbe used directly to perform the stress analysisusing the general-purpose finite-element analysis(FEA) software package ANSYS. Selected


180

conditions stemming from expected boundingcases will be analyzed. Disc material propertiessensitivity will be investigated by comparing stressstates to material degradation/failure limits. Thisevaluation cycle will be iteratively applied toarrive at an optimized disc design.

Task 2 – High-Performance Friction PairsA survey of was conducted of friction pairmaterials, which included commercial andemerging materials options, as well as those still inthe research and development phase by variousvendors. Over 20 friction material options andvendors were surveyed and, based on thetechnology readiness level and cursory economicconsiderations, four were down-selected. Thesefour friction pair and rotor material candidates willbe pursued further in order to evaluate their fullpotential as cost-competitive and technicallyviable options for disc brake systems.

A description of the down-selected candidatematerials is given below. The features they have incommon are that all have the potential to be low-cost cermets materials fabricated by liquid metalinfiltration, by reaction processing, or by frictionstir processing.

1. Excera Materials Group –50SiC/35Al2O3/W1461-XBr, in situcomposite.

2. Starfire Systems – Rotors – polymer-basedceramic reinforced with chopped PAN carbonfibers. Pads – conventional material withSiOC polymer substituted for phenolic.

3. MER Corp. – Composite of pitch-basedgraphite cloth in SiC matrix, in situ attached tohigh-temperature (1,000°C) MMC core.

4. PNNL – Steel with TiB2 particlesincorporated by friction stir processing. Castiron with TiC particles incorporated by in situfriction stir reaction processing.

Excera Materials GroupThe structure of ONNEX material is produced bya low-cost SiO2-Alliq in situ reaction method thatuses a net-shape ceramic preform. The precursormaterials are chemically transformed into theONNEX cermet material by using the company’spatented immersion process. This process creates

an interwoven ceramic and metallic network. Thealuminum matrix is then substituted with bronzefor high-temperature applications.

The Excera ONNEX materials feature a uniqueand tailorable array of physical and mechanicalproperties. While these properties can be modifiedto match the demands of specific applications,they can be generally classified by the following:

• Low density (half the density of steel)• High stiffness (slightly stiffer than steel)• Good strength (similar to aluminum castings)• High fracture toughness (similar to cast iron)• Low thermal expansion (30% less than steel)• High thermal conductivity (between twice that

of cast iron and aluminum alloys)• Exceptional wear resistance (better than most

ceramic materials)

The manufacturer claims that for automotiveapplications, ONNEX materials exhibit ten timesthe wear resistance of iron-based materials, andunder severe conditions, can operate attemperatures approximately 100°C cooler —which yields shorter stopping distance for brakeapplications.

Starfire SystemsStarfire materials utilize a polymer infiltration andpyrolysis (PIP) process that is the latest techniqueresulting from years of research to develop aprocess that enables the fabrication of advancedceramics more efficiently than conventionalprocesses. This technique involves soaking a fiberpreform or powder compact with a liquid polymerprecursor that converts to ceramic material uponpyrolysis. Pre-ceramic polymers are available thatform silicon carbide, silicon nitride, siliconoxycarbide, and silicon oxynitride. Advantages ofPIP include simpler, less costly equipment, lowerprocess temperatures, shorter cycle times, and thecapability to produce more complex parts. Also,polymers afford the potential to control materialschemistry at the molecular level.

Starfire’s material systems offer a wide range ofpotential solutions for transportation markets, suchas high-wear engine components, exhauststructures, diesel particulate filtration systems, and


181

(presumably) friction materials that are suitable forheavy-vehicle brakes. These materials can bemodified for high-wear applications. Starfirematerial claims for a silicon carbide materialsystem currently under development are asfollows:

• One-third the weight of steel and up to 20%lighter than carbon/carbon (C/C).

• Excellent static friction and stable basefriction (0.2) that can be tailored to exceed0.35; both static and dynamic friction.

• Significantly lower wear rates versus C/C.• Greater mechanical stability versus C/C.• Demonstrated 30–50% lower cost compared to

C/C.• Friction performance is unaffected by fluids or

temperature.• Improved oxidation resistance.• Suitable for automobiles due to low wear and

stable friction.• Can eliminate overheating or fade associated

with current heavy truck brakes under extremehighway braking conditions.

MER CorporationThe MER friction materials are still in adevelopment stage, but they show strong promiseas friction pairs for vehicle braking systems. Thesecomposites are made from pitch-based graphitecloth in a SiC matrix, which is created via anin situ process. The reacted C/SiC product isattached to a high-temperature (1,000°C) MMCcore. MER SiC/SiC composites are producedusing CVR SiC fibers, as well as commerciallyavailable SiC fibers (including Nicalon andothers). The SiC matrix is produced usingpreceramic polymers, chemical vapor infiltration(CVI), and hybrid combinations of these and otherprocessing methodologies. The C/C compositesare converted to SiC/SiC or C/SiC composites, aswell as to other carbides, including ZrC and HfC.It is possible to produce pure SiC/SiC compositeswithout carbon, BN, or other debonding layers andachieve two to five times the through-the-thickness thermal conductivities as state-of-the-artSiC/SiC composites. These C/SiC and SiC/SiCcomposites are thermally stable to at least1,800°C. Cost projections for C/SiC or SiC/SiCcomposites are typically high, but MER claims its

process is economical. These materials will beevaluated for their wear resistance and thermalstability. Furthermore, MER will provide updatedcost information for further evaluation.

Pacific Northwest National LaboratoryPNNL, in conjunction with South Dakota Schoolof Mines & Technology (SDSMT), have initiatedwork to develop friction stir reaction-processing(FSRP) to create an in situ surface composite insteel, cast iron, or other appropriate base-metalsystem. A schematic of the friction stir processingmethod is shown in Figure 2.

Material systems currently being investigatedinclude steel with imbedded TiB2 particles andcast iron with TiC particles. Figure 3 shows that

Figure 2. Friction stir processing is accomplished byplunging a spinning tool into a material and translatingthe tool across the surface. Ceramic powders, or othermaterials, can be captured under the shoulder of the tooland mixed downward into the processed region.

Figure 3. Friction stir processed cast iron(Ramasubramanian et. al. 2004).


182

cast iron can be plasticized by means of FSRP.Recent work at PNNL on FSRP wear-resistantsurfaces has demonstrated the incorporation of SiCparticulate into the surface of aluminum at depthsof 2 mm. These and other systems, including TiCor WC-reinforced high-strength steels, will also beinvestigated, as appropriate.

FSRP was invented by PNNL and SDSMT withinthe last two years, so the economics for theprocess are currently unknown. Samples of thesematerials will be produced and evaluated for theirmerit as brake friction pairs. If the initial technicalfeasibility study shows promise for this class ofmaterial, the economics of the process will bemodeled to determine whether the resultingmaterials are a feasible option for heavy-vehiclebrake systems.

Future DirectionThe scope of work for FY 2005 will focuspredominantly on brake rotor geometry design,thermal management, and model simulationactivities in support of system designconsiderations. Multi-disc brake rotor geometrieswill be investigated for thermal and brakingperformance. Other new design concepts will beinvestigated to determine (1) what benefits theymay provide and (2) the impact of each on brakingperformance. The modeling task will continue toinvestigate heat generation, fluid flow, and theresulting surface temperature of the brake systemcomponents.

The modeling task will require thermomechanicaldata on proposed rotor/pad material combinations,as well as material physical properties such asconductivity, diffusivity, density, wear resistance,and friction coefficients, and other mechanicalproperties. These properties will be determined,along with the evaluation of system response toNVH for the various friction pair combinations,and used as input to the model. The model will beupdated with more complex geometries asnecessary.

In addition, alternative energy absorption/conversion and heat management/rejectionmethods will be investigated for their potential asprimary or secondary methods to control vehiclespeed.

AcknowledgmentsThe principal investigators would like toacknowledge the valuable contributions of JamesFort and Harold Adkins of the Pacific NorthwestNational Laboratory for their work in developingthe brake model and simulations.

References1. Jerhamre and Bergstrom, “Numerical Study of

Brake Disc Cooling Accounting for bothAerodynamic Drag Force and CoolingEfficiency,” SAE Tech Paper 2001-01-0948.

2. Jansen, “Vehicle Disc Brake Cooling FactorAnalysis,” RadTherm Users Meeting.


183

XI. More Electric Truck

Parasitic Energy Loss Reduction and Enabling Technologies for Class 7/8 Trucks

Program Manager: William LaneCaterpillar, Inc., Technology and Solutions DivisionP.O. Box 1875, Peoria, IL 61656(309) 578-8643, fax: (309) 578-4722, e-mail: [email protected]

Technology Development Manager: Sid Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected]

Technical Program Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4798, e-mail: [email protected]

Contractor: U.S. Department of Energy, Albuquerque, New MexicoContract No.: DE-FC04-2000AL67017

Objectives• Reduce parasitic losses for fuel savings.

• Provide idle-reduction solutions.

• Reduce radiator heat load.

Approach• Electrify truck accessory functions and hotel-type loads.

• Decouple accessories from main truck engine.

• Match accessory power demand to real-time need.

• Enable use of alternative power sources.

• Develop and implement comprehensive systems-level perspective.

Accomplishments• Conducted idling, over-the-road, and dyno fuel economy testing on technology demonstration truck.

• Exhibited technology demonstration truck at Mid-America Truck Show, Louisville, KY.

• Developed virtual simulation model of demonstration truck for exploration of advanced power managementcontrol strategies.

Future Direction• Utilize existing technology demonstration truck as platform for “Advanced Electric Systems & Aerodynamics

for Efficiency Improvements in Heavy Duty Trucks” DOE Program Award.

• Evaluate additional fuel savings opportunities enabled by truck electrification.

• Upgrade modular HVAC component.


184

IntroductionAs fuel prices increase, there is more interest in thedevelopment of electric and hybrid electric vehiclesand in their higher operational efficiency. Themajority of developments have focused onpropulsion systems. Parasitic losses from truckauxiliary functions, (such as pumps, compressors,etc.) are another source of inefficiencies that haveremained unchallenged. This research programcovers developments associated with theelectrification of Class 8 on-highway trucks toreduce parasitic losses. Incremental gains in fueleconomy could be attained by using the present 12-Velectrical system, if the alternator efficiency wereimproved. However, elevating efficiency from lessthan 50% to over 80% would require the applicationof different alternator/generator technology. Such animprovement may come with a cost premium that isunacceptable to the trucking industry. This creates asignificant cost challenge for improving truck fuelefficiency by boosting the efficiency of individualcomponents.

An alternative to component-by-componentreplacement is to take an overall system perspectiveto truck electrification. In some cases, electrifying acomponent may only provide marginalimprovements in fuel savings when used in the samemanner as its conventional counterpart. However, anelectric component could also be the enabler to adifferent mode of operation that may realizesignificant fuel savings. For example, overnightidling of a truck engine is the common means ofproviding cabin heat or cooling. This practice usesfuel inefficiently, increases engine maintenance,creates additional exhaust emissions, and increasesoperating expenses. The electrification of theheating, ventilation, and air conditioning (HVAC)system provides the capability of conditioning the airto the cabin without having to run the truck engine.This is important because large quantities of fuel areburned just for that purpose.

With that precedent, this research program focusedon converting conventional belt- or gear-drivenaccessories to electric variable-speed motor-drivenones. This allows them to operate independently ofthe main engine, at only the actual speed, pressure,or flow required. By continuously adjusting to thedesired operating points, there are multiple

opportunities for parasitic loss reduction and controlfeatures. To demonstrate and test the benefits oftruck electrification, a prototype Class 8 heavy-dutytruck was built. This More Electric Truck (MET),with associated accessory components, is illustratedin Figure 1.

Figure 1. Demonstration vehicle.

The MET is based on a Kenworth T-2000 truckchassis with a Caterpillar® C-15 engine rated at 500HP. On it is implemented a comprehensiveelectrification of truck accessory functions as ameans to improve fuel economy, heat rejection,reliability, power management, and packagingflexibility. Several high-power accessories weredemonstrated to quantify the benefits of truckelectrification, during both normal “over-the-road”and “overnight idling” operating conditions. Amongthe electrical accessory systems demonstrated in thisprogram were an integrated starter-generator (ISG), amodular heating ventilation and air-conditioner(MHVAC), a shore power converter, an electronicbattery charger (down converter), a water pump, anoil pump, a service-brake air compressor, and anauxiliary power unit (APU). These components andtheir functions have been described in previousreports [1].

A comprehensive series of fuel economy tests wasconducted during the third and fourth quarter ofcalendar year 2003 to quantify the fuel savings madepossible by the MET electrification. These tests weredesigned to obtain real-world truck performance databy using the MET and a control vehicle over a rangeof terrain and conditions and conformed to SAE


185

Type-II (J1321) procedure [2]. The MET vehicle,before conversion to electrified accessories, wastested for baseline fuel economy in September andOctober of 2002. The MET vehicle’s engine,transmission, and drivetrain were consistent betweenbaseline (BL) and experimental (EX) tests.

Quantification of fuel savings for an average over-the-road cycle is challenging. The savings largelydepend on how the truck is used, weather conditions,and road characteristics, among other factors. It isdifficult to conceive of a single test cycle that wouldbe representative of the average truck. Truck fleetsuse their own load cycles based on their morecommon routes. Engine manufacturers useproprietary engine load profiles to baseline the fueleconomy of their engines. For quantification of METover-the-road fuel economy benefits, both a tracktest and interstate highway route test wereundertaken. Environmentally controlled truck idlingfuel consumption tests were also conducted with theMET.

Test Results: IdlingA major fuel savings opportunity for long-haultrucks is to curtail overnight idling of the mainengine. It has been estimated that long-haul trucksidle an average of 1,830 hours per year. Engineidling speeds range from 600 to 1,000 rpm,depending on the driver’s preferences and weatherconditions. Thus, truck idling fuel consumption canbe calculated, for defined climatic conditions, byusing measured engine fuel consumption atfractional loads.

As part of the MET truck electrification, a diesel-powered, variable-speed generator was designed,fabricated, and mounted on the truck chassis, toprovide DC power to the HVAC systems duringovernight idling conditions. Quantification of fuelsavings as a result of using the APU and modularHVAC (versus a baseline of using the main engineand conventional HVAC) to cool the cab over duringidling periods was sought as an objective of thisprogram. The MET was tested for APU fuelconsumption to maintain cab cooling in thecontrolled environmental chamber of the PaccarCooling Test Cell in Everett, Washington, October2003. Ambient temperature was maintained to 95°F,and the humidity was 45%. The results of both the

baseline (C15 at 1000 rpm) and the APU testing areshown in Table 1. The figures shown in the tablereflect the average temperatures and fuel rate over a4-h period of stable, steady-state cab temperatures.

Table 1. MET idling fuel testing results.

Avg. FuelRate

(gal/h)

Avg. CabTemp(°F)

Avg. SleeperTemp(°F)

Avg. APUPower(kW)

C15 at 1000 rpm 1.31 65 69 2.8APU 0.21 71 76 -

From the table, a reduction from 1.31 to 0.21 gal/h,or an 84% improvement, is evident from the baselineto the APU test. The fuel consumption during theAPU test may be slightly low because the cab andsleeper temperatures could not be maintained at thesame temperatures as those measured in the baselinetest, probably because there was inadequate air massflow from the MHVAC blower and ducting system.

Also of interest is the amount of fuel used by theC-15 in the baseline test. In separate engine-onlytesting of the C-15 at Caterpillar test facilities, theengine fuel consumption with a 4-kW output loadwas measured at 1.1 gal/h. With the assumption ofsimilar loading in the baseline idle test, a 0.21 gal/hincrease is seen. The likely cause for this increase isthe 50% duty cycle recorded for the time that thecooling fan was on in the baseline test. This providesadditional evidence for possible benefits provided byusing an APU for cab cooling.

The APU engine coolant system was integrated withthe main engine so that the main engine and radiatoracted to dissipate the heat energy produced by theAPU. It was desired to see if the APU heat could berejected in this manner to a high ambient airtemperature. Testing of the APU in high ambientconditions (85°F) also revealed the need forsupplemental cooling-fan-provided airflow acrossthe truck radiator. With only passive, naturalconvective airflow through the radiator, heat transferwas impeded, insufficient heat was rejected to theatmosphere, and coolant temperature rose aboveacceptable levels. The addition of two small, 12-in.-diameter, 12-V fans in front of the radiator resultedin enough forced air through the radiator to maintainadequate cooling for the APU.


186

The levels of cab interior acoustic noise during idleengine and APU operation were measured at PaccarTechnical Center. The baseline test, with the mainengine idling at 1000 rpm and AC blower fan set tomedium, resulted in a sound pressure level near thedriver’s head of 68.3 dB and 66.1 dB at the sleeperbunk. The test with the APU at 2400 rpm (typicalspeed for normal cab cooling power needs), andelectric HVAC at medium blower speed, resulted ina sound pressure level near the driver’s head of60.2 dB and 57.5 dB at the sleeper bunk. Thisamounts to a reduction of about 8 dB for both caband sleeper, which is well over a 50% reduction.This reduction allows for a more comfortable restenvironment for a truck operator, further increasingthe value of an onboard APU.

Test Results: Test TrackThe track testing was undertaken at Paccar TechnicalCenter. The test track was a 1.6-mi banked turn oval,where a test run consisted of 50 mi distance at60 mph average speed. This type of test providedconsistent, constant-speed, and level cruisingconditions for the truck.

The baseline track test was performed September 3,2002, with the MET (pre-electric accessories) as thetest truck (TT) combined with a loaded trailer for aGVW of 76,240 lb. The control truck (CT) used forboth the baseline and experimental test was aKenworth T-2000, with a Caterpillar C15 enginerated at 475 HP and a loaded trailer for a GVW of74,940 lb. A summary of the track testing results isshown in Table 2. The columns in the tablecorrespond to each of five separate test segments –each test segment consisted of three individual testruns. For a test segment to be valid, the test run’sTT/CT fuel used ratio must have been within a 1%spread. The fuel consumption for each truck in eachrun was measured by weighing of removable fueltanks before and after each run. All fuel use shownin the table has been normalized to the control truckfuel use from the baseline segment [3].

From Table 2, it can be seen that a significantdecrease (ranging from 2.7% to 5.8%) in the controltruck fuel use was measured between the baselineand experimental segments. Some of this decreasecan be attributed to a slight decrease in average windspeed, along with a small increase in average

Table 2. Baseline and MET test track results, normalizedto baseline control truck fuel use.

Date 9/3/02 9/22/03 9/23/03 9/24/03 9/25/03Test BL EX EX EX EX

Temperature 65.2 69.0 65.0 69.8 70.0Wind Speed 6.0 4.9 7.9 3.7 4.4CT Fuel Use 1.000 0.944 0.973 0.942 0.944TT Fuel Use 1.013 0.981 1.010 0.969 0.954TT/CT Ratio 1.013 1.039 1.038 1.029 1.010TT Fuel UseChange fromBL (%) 2.49 2.38 1.49 -0.32CT Changefrom BL (%) 0.0 -5.6 -2.7 -5.8 -5.6CT Comp.by 4 % 1.000 0.982 1.012 0.980 0.982CorrectedTT/CT Ratio 1.013 0.999 0.998 0.989 0.971Corrected TTFuel UseChange fromBL (%) -1.45 -1.56 -2.41 -4.16

OtherCond. BL GVW

BL GVW+1200 lb

BL GVW+1200 lb BL GVW

BL GVW,APU Pwd.

Acc.

ambient temperature. From Chu [4], a 5-mphheadwind or crosswind can account for a 5–10%increase in fuel use in long-haul trucks, and a 10°Fincrease in temperature may result in a 1–2%decrease in fuel use in long-haul trucks. Separateanalysis of the individual test runs with associatedtemperatures and wind speeds resulted in a factor of0.65% increase in fuel consumption for each 1-mphincrement in wind speed.

Comparing the 9-23-03 segment with the baselinesegment, the temperatures are virtually identical: a1.9-mph increase in wind speed is seen, and thecontrol truck used 2.7% less fuel. Canceling out thewind speed effect by 1.3% (1.9 mph × 0.65%/mph),and adding this to the 2.7% decrease in fuel use,gives a compensated fuel use decrease for the controltruck of 4.0%. This means that the overall fuelconsumption of the control truck used for the testsdecreased by 4% from the time the baseline segmentwas performed to the time the experimentalsegments were performed. This decrease in fuelconsumption is likely due to an additional 10,000 miof break-in time on the control truck’s engine andpossibly the tires, which, from to Chu, can result in a2–5% reduction in fuel use.


187

On the basis of the analysis above, the fuel use of thecontrol truck in the experimental segments wascorrected to compensate for the decrease in actualfuel used and to give more accurate results to thetesting. This is shown in the “CT Comp. By 4%” rowof the table, and results in the corrected TT/CT ratioand the corrected test truck fuel use change relativeto the baseline, as shown in the subsequent two rowsof the table.

The two experimental segments from 9-22-03 and 9-23-03 were performed with the test truck at itsbaseline GVW plus 1200 lb; the additional weightwas attributable to the increase from additionalelectric components and APU. The two experimentaltest segments from 9-24-03 and 9-25-03 wereperformed with the 1220 lb subtracted from thetrailer cargo weight, to give the same GVW for thetest truck in both the baseline and experimental tests.Also, the segment performed on 9-25-03 wasperformed with the electric accessories poweredsolely by the onboard APU, to give a betterindication of the contribution of electric componentsto fuel consumption.

The corrected results in Table 2 show that the METversion of the test truck had decreased fuel usage ofabout 1.5% (average for the 9-22-03 and 9-23-03segments, both with an additional 1200 lb of GVW)and 2.4% for the 9-24-03 segment with GVW equalto the baseline segment. This difference of 0.9% is inline with the findings of Chu, who reported anincrease in fuel use of about 1% for each 1000 lbincrease in vehicle weight.

Regarding the difference in weight, for the testvehicle MET of 1200 lb, approximately 400 lb canbe attributed to the APU for idle reduction. About150 lb might be saved by using a cast-aluminuminstead of a cast-iron flywheel/ISG housing. Also,more-efficient, production-intended packaging andsize reduction of power electronics may result in anadditional 200 lb weight reduction. The MET alsohad approximately 50 lb of additional weight in theform of an AC inverter. This is mentioned to assuageconcerns about the GVW as tested for the 9-24-03segment, in which the trailer ballast was removed tocounter the 1200-lb weight increase for the MET. Ofthis 1200 lb, one-third is attributable to the APU, andone-third would be eliminated in more-production-intent versions. Therefore, it was felt that the best

comparison for the MET fuel economy would be aGVW matched to the baseline.

It is important to note that the MET experimental testsegments were performed with a non-functioning airconditioning unit, whereas in the baseline segment,the test truck was tested with the air conditioningfunctioning and set to mid range. A componentfailure occurred in the MET HVAC prior to the tests,and because of tight test-track scheduling, the testswere continued before the module could be repaired.

The last experimental segment was performed withthe MET electric accessories powered exclusively bythe APU, eliminating all the electric parasitics on themain engine. This was done to quantify thecontribution of the combined electric accessory loadand resulted in an additional decrease in fuel use of1.75%, for a total decrease over the baseline of4.16%. The fuel consumption by the APU was notmeasured in this test segment. This segment alsohelped to carry out a component-by-componentbreakdown of accessory contribution percentage tooverall truck fuel use. This was made possible bysampling electric current and voltage sensor data toeach of the components and then calculating therequired fuel rate, by engine brake-specific fuelconsumption, needed to produce mechanical powerthat was then converted to electrical power by theISG. A similar breakdown of the mechanicalaccessories was performed earlier in this project andsubsequently enhanced from the experimentaltesting. The breakdown of mechanical accessorieswas accomplished by inference from componentpower curves, rather than actual measurement ofmechanical torque and speed. The results of thecomponent analysis are shown in Figure 2.

0.0%0.2%0.4%0.6%0.8%1.0%1.2%1.4%

AC Comp Oil Pump WaterPump

Alt./ Dn.Conv.

Fan Air Comp

Baseline Experimental

Figure 2. Test track component contribution to total fueluse.


188

The component analysis shows that the contributionof the mechanically driven, or belt-driven, ACcompressor accounted for about 1.1% of test truckfuel consumption in the baseline segment. Assumingthe contribution of the electrically driven HVACsystem to fuel use is comparable to that of a belt-driven AC compressor, an additional 1.1% can besubtracted from the 2.4% improvement shown inTable 2 for the 9-24-03 segment, to give an overallon road fuel economy benefit from under test trackconditions of 1.3%. This is probably the closestnumeric figure that can be arrived at for METelectrification fuel economy benefit, on the basis ofthe data produced as a result of the track test portionof the MET fuel economy testing.

The component breakdown of fuel use for thebaseline and experimental (MET), shown inFigure 2, reveals that the electric water pump was thelargest, single contributor to diminished fuel use forthe MET in the track tests. The control strategy forthe water pump allowed for minimal coolant flow(and, thus, low power draw) at the relatively lowengine torque demands (about 40%), as seen in thelevel cruising conditions of the track test. Thecombination of reduced flow mechanical and electricoil pump in parallel configuration provided a small(0.2%) benefit in power consumption. About a 0.3%benefit was seen from the electrically driven aircompressor because of the elimination of parasiticcompressor torque to the power train duringunloaded conditions in the gear-driven, non-MET aircompressor.

The summation of the fuel percentage from theelectrically powered oil pump, water pump, downconverter, air compressor, (0.32 + 0.11 + 0.80 + 0.35= 1.58%) matches well with the 1.59% fuelimprovement difference between the 9/24/03 and9/25/03 tests, where the latter test had the APUproviding power to the electrical accessories.

A small increase in power use was seen in the METdown converter, or 12-V power supply, versus theconventional alternator used in the baseline test. Thiswas due to a significant increase in 12-V loads fromthe addition of an AC power inverter, extra powerconverters to MET power electronics, and a faulty(shorted) relay solenoid to the air compressorcooling fan. Subsequent static testing of the downconverter revealed an operational efficiency of

85–90%, compared to the near 50% efficiency of abelt driven alternator. This would indicate that (1) anoverall fuel economy benefit could be expected byimplementing the MET down converter and (2) theresults seen in the track test were an anomaly.

Also, from Figure 2, equally low energyconsumption was seen from the cooling fan for bothbaseline and experimental tests because of lowengine loading and high ram air available to thecooling system during the level cruising conditionsof the track test. Also, as noted above, the METmodular HVAC system was temporarily non-functional during the track testing, and so acomparison of this component was not possible inthis test. Finally, a summation of the componentbreakdown reveals an overall fuel benefit of 1.2%from the MET components versus the baseline,which is close to the 1.3% derived from themeasured fuel use described above.

Test Results: Highway TestingThe over road test was conducted by Kenworth Labs,Renton, Washington, where one test segmentconsisted of three, 76.5-mi laps on Interstate 90 nearEllensburg Washington. The terrain wasapproximately one-third level or slightly slopinggrade and two-thirds long, steep, mountainous terrainacross the Columbia River Gorge. The topography ofthe route is shown in Figure 3.

0

500

1000

1500

2000

2500

3000

0 10 20 30 40 50 60 70 80

Distance (Miles)

Elev

atio

n (ft

)

Eastbound Westbound

Ellensbur

Ellensbur

Ryegrass Pass

Ryegrass Pass

Silica Road Exit

Vantage BridgeColumbia River

Vantage BridgeColumbia River

Figure 3. Vantage Grade I-90 route.

This route was chosen to provide extreme loadingconditions for the truck and its cooling system. Thewestbound segment of the route includes an 11-miuphill portion (the Vantage Grade), with an averageslope of 3.5%; this grade, in combination withtypically high ambient conditions, puts anextraordinary load on a truck’s cooling system. In


189

traveling up the Vantage Grade, a truck typicallyslows down to about 35 mph, decreasing ram airavailable to the aftercooler and radiator, which thennecessitates the engagement of the cooling fanclutch, putting an additional 10–20-kW fan load onthe engine.

The baseline Vantage Grade test was performed onSeptember 11 and 12, 2002; again, the test involvedcombining the truck with a loaded trailer for GVWof 76,240 lb. The same control truck and loadedtrailer were used as in the track tests. A summary ofthe Vantage testing results is shown in Table 3. Thefirst data column shows and average of two testsegments conducted on 9/11/02 and 9/12/02. Theexperimental test was performed on 10/14/03, againwith the electric accessories (MET convention).Only one experimental test was conducted during thescheduled testing, since inclement weather made thehighway wet on the remaining testing days. The fuelconsumption for each truck in each run wasmeasured by metered pumping to a demarcated levelin the fuel tank after a test segment. All fuel useshown in the table has been normalized to the controltruck fuel use from the Vantage baseline segment.

The corrected MET fuel use shows a decrease of4.32% over the baseline test truck. The Vantage datawere also analyzed for a component-by-componentbreakdown of fuel consumption contribution.

Table 3. Baseline and MET Vantage Grade test results,normalized to baseline control truck fuel use.

DateAvg 9/11/02and 9/12/02 10/14/03

Test Baseline ExperimentalTemperature 81 60Wind Speed 2.2 2.5CT Fuel Use 1.000 0.964TT Fuel Use 1.030 0.988TT/CT Ratio 1.030 1.025TT Fuel Use Changefrom BL (%) -0.49CT Change from BL (%) -3.6CT Comp. by 4 % 1.000 1.003TT/CT Comp. Rat. 1.030 0.985TT Fuel Use Changefrom Comp. BL (%) -4.32

Similar to the findings from the track test, the controltruck’s fuel use also decreased between the Vantagebaseline and experimental tests. A nominal decreaseof 3.6% was observed. About a 3% increase in fuelcould be attributed to the 21°F decrease intemperature, but this is offset by about 1.5% fromthe decreased load on the AC compressor andanother 1.2% from the decreased fan-on time. Thenet of these, +0.3%, which is offset by a 4%expected decrease in fuel use of the control truck dueto additional break-in miles, gives an expecteddecrease in fuel consumption for the control truckexperimental versus the baseline of 3.7%, which isvery close to the 3.6% decrease observed in the data.So, as in the test track results, the experimental datafrom the control truck were corrected by 4%, whichyields a valid comparison of the MET electrificationfuel benefit.

Correcting for the 4% decrease of the control truck, afuel consumption decrease of 4.3% was observed forthe MET versus the baseline. However, because ofthe 21°F decrease in ambient temperature for theexperimental segment, the AC compressor andcooling fan contribution must be voided to give avalid comparison between the segments. Therefore,subtracting the 1.8% due to the AC compressor and1.2% for the cooling fan leaves a 1.3% increaseattributable to the MET electrification.

The results of the component breakdown analysis forthe Vantage Grade test segment are shown in the barchart of Figure 4. A summation of increases for theoil and water pumps, alternator, and air compressorgives a total component increase of 0.7%, which issomewhat less than the 1.3% from the total fuelanalysis above. This difference may be attributableto the baseline breakdown of the mechanicalcomponents being understated, since the power inputto these components was not measured but onlyestimated on the basis of component characterizationcurves. Also, since it was only possible to performone experimental test segment, more variability maybe present in the Vantage grade tests than in the tracktests, which leads to lower confidence in theseresults.


190

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

Baseline Experimental

Figure 4. Vantage grade component contribution to totalfuel use.

Also note that the power consumption of the electricwater pump was more than in the track test becausethe topography forced the engine to operate underhigher loading conditions, which required morecoolant flow at higher pump speed. Again, the downconverter of the MET consumed more energy thanthe alternator of the baseline as a result of increased12-V loads, as mentioned in the track test above.Overall, the percentage of component energyconsumption was lower in the Vantage Gradesegment than in the track test, simply because ahigher percentage of fuel was spent powering thetruck over the steep terrain.

Advanced Power Management/SimulationThe Kenworth T-2000 Class-8 truck was modeledfor simulation purposes by using Caterpillar’sDynasty dynamic-analysis software combined withthe Simulink control-analysis package byMathworks. The truck’s inertial characteristics,power train, and cooling system were modeled withDynasty, and the MEI components and controls weremodeled with Simulink. The model was validatedagainst measured data from the truck traveling theVantage Grade I-90 route, which was describedpreviously and depicted in Figure 3. For a course runat 16°C average ambient temperature, the simulationcalculated that 54.51 L of fuel was required tocomplete the course, compared with the measuredamount of 54.75 L, which is a difference of less thana half of a percent.

The baseline (non-MEI) truck was also modeled andsimulated to compare performance with the MEItruck on the Vantage Grade route. The simulated fuelusage results with two different average ambient

temperatures are shown in Table 4. In addition, anMEI truck with a controlled cab air-conditioning unitis also listed. In this case, the cab a/c was shut downwhenever the engine load became greater than 80%of the maximum engine power of 350 kW. Thesimulation indicates that at higher temperatures, fuelsavings could possibly be doubled by using such acontrol; however, how hot the cab might become ifthis control were actually employed in the real truckis unknown at present.

Table 4. Results of power management simulation.

AmbientTemperature

(°C) Baseline MEIMEI w/ CabA/C Control

16 54.97 54.51(-0.84%)

54.43(-0.98%)

25 55.88 55.41(-0.84%)

55.01(-1.56%)

Investigation of further power-management controlschemes is ongoing and will carry over into theAdvanced Electric Systems program mentionedbelow in the Future Work section. Operating theMET in mild-hybrid mode, using regenerativebraking and energy storage, and motoring the ISGwhen the engine is under heavy load are beingexplored through simulation with the high-fidelitymodels mentioned above. Preliminary simulationresults of the truck operating in hybrid control modeover sinusoidal (hilly) terrain indicate a fuel savingspotential of 2–4%.

Future WorkThe MET, with electrified accessory components,will be maintained as a research platform foradditional study in reduction of parasitic engineloads. A further proposal, “Advanced ElectricSystems and Aerodynamics for EfficiencyImprovements in Heavy Duty Trucks,” has beenaccepted in response to DOE Solicitation DE-SC52-03NA68471. Four main areas of exploration for thisproject will include (1) a high-temperature-charge aircooler, for more heat rejection per unit area, and lesspumping restriction; (2) an electrically driven enginecooling fan; (3) a more compact and efficient direct-drive scroll-type air compressor; (4) an electroniccontrolled engine coolant radiator by-pass valve, forelevated and more accurate engine coolanttemperature regulation; and (5) advanced power


191

management and hybrid traction use of the integratedstarter generator.

The prototype modular HVAC component of theMET will be upgraded to a more production-intended second-generation design. This design willenable more compact, lightweight installation, aswell as provide a higher-powered and more efficientblower to provide additional airflow to both cab andsleeper compartments.

The simulation work described in the previoussection will continue as part of the AdvancedElectric Systems project. Advanced powermanagement strategies and hybrid configuration willbe explored and implemented on the MET platform.

Summary and ConclusionsA prototype test vehicle, the More Electric Truck,(MET) was designed, built, tested, and demonstratedas the principal effort of this program (Figure 5).This Class-8 heavy-duty long-haul truck capitalizeson using variable-speed electric motors as primemovers for accessory engine systems to improveefficiency, power management, reliability, packagingflexibility, and customer value. Among theelectrically driven accessory systems demonstratedin this program were the starter-generator, shorepower converter, heating ventilation and air-conditioner, electronic battery charger (downconverter), water pump, oil pump, service-brake aircompressor, and auxiliary power unit. This truckconfiguration was tested and showed significantopportunity for reducing fuel consumption on long-haul Class 8 trucks.

Figure 5. MET with APU and power electronics box.

Reducing the main engine idle time will save fuel.An onboard APU may provide an important initialstep toward reducing the fuel consumption of Class 8trucks. A 75–80% reduction in fuel use wasdemonstrated on the test vehicle, with the APUproviding cab-cooling power and main engine idlingproviding power to the belt-driven air conditioningcompressor. The sound pressure levels inside thetruck cab were also reduced by over 50% by usingthe APU instead of the main engine.

Another significant opportunity for reducing overallfuel consumption on long-haul trucks is to reduceover-the-road parasitic loads. SAE type II fuelconsumption test procedures were used to documentfuel savings of the MET test vehicle versus abaseline of the same vehicle configured withconventional engine accessories. For the MET withelectric accessories versus the vehicle configuredwith conventional accessories, a fuel savings of 1.3%was observed under both level-cruising track testconditions and mountainous interstate highwayconditions. The fuel savings were mostly attributableto the electric water pump and electric aircompressor, with a small percentage decrease fromthe electric oil pump.

Advanced power management control strategies forMET accessory systems have been developed andsimulated as part of the extension grant for thisprogram. An improvement of up to 1.5% in fueleconomy has been demonstrated in simulation.Simulation efforts aimed at using the MET startergenerator in a mild-hybrid mode have shown a 2–4%improvement in fuel economy, depending on thetype of terrain traveled, and the drive cycle.

This is the final annual report for this project;however, the MET test vehicle will be maintained asa research platform for additional study in reductionof parasitic engine loads. A further proposal,“Advanced Electric Systems and Aerodynamics forEfficiency Improvements in Heavy Duty Trucks,”has been accepted in response to DOE SolicitationDE-SC52-03NA68471. Also, the modular HVACprototype component will be upgraded to a more-production-oriented version with increased blowerairflow.


192

References1. Algrain, Marcelo, “Parasitic Energy Loss

Reduction and Enabling Technologies for Class7/8 Trucks,” FY 2003 Annual Report,Caterpillar, Inc., 2003.

2. SAE J1321, Joint TMC/SAE Fuel ConsumptionTest Procedure – Type II, Surface VehicleRecommended Practice, Society of AutomotiveEngineers, Inc., 1986.

3. Duffy, John, and Everett Chu, MEI-Engine FuelConsumption Project, KW02-150, KenworthTruck Company and Paccar, Inc., 2003.

4. Chu, Everett, “Effect of Factors AffectingAbsolute Fuel Economy, Based on TMCRP 1111,” Paccar, Inc., 2002.


193

XII. Ultralight Transit Bus System

Vehicle System Optimization of a Lightweight, Stainless-Steel Bus

Principal Investigator: J. Bruce EmmonsAutokinetics, Inc.1711 West Hamlin RoadRochester Hills, MI 48309-3368(248) 852-4450, fax: (248) 852-7182, e-mail: [email protected]

Technology Development Area Specialist: Sidney Diamond(202) 586-8032, fax: (202) 586-1600, e-mail: [email protected] Technical Manager: Jules Routbort(630) 252-5065, fax: (630) 252-4798, e-mail: [email protected]

Contractor: Argonne National LaboratoryContract No.: 4F-02161

Objectives• Perform the integration and optimization of a hybrid or battery/electric propulsion system and various vehicle

subsystems into a lightweight bus body. Autokinetics will use as much off-the-shelf technology as possible.Optimization of the propulsion and vehicle systems will primarily be through careful selection of appropriatelysized components. This project will result in a single proof-of-concept prototype bus, suitable for testing andevaluation of performance under controlled conditions.

• Identify one or more paths to rapid commercialization of the technology.

Approach• Conduct computer simulations of a number of different types of propulsion systems to predict performance and

energy efficiency.

• After identifying the most promising propulsion system architecture, use computer simulations to evaluate andselect the individual components with the best combination of performance and affordability.

• Purchase and install the integrated propulsion system.

• Design or select optimized vehicle subsystems, such as seats, glass, and air conditioning.

• Purchase or fabricate and install all vehicle subsystems.

• Evaluate the performance and make modifications, if necessary.

• Perform initial testing.

Accomplishments• Completed first project task — Propulsion System Modeling and Simulation.

• Initiated numerous ongoing discussions about commercialization and deployment in California, Tennessee,Michigan, and China.

Future Direction• Complete the design of the propulsion system, including traction motor controller with regenerative braking,

fuel converter and generator, vehicle controller, and energy storage system.


194

• Select specific propulsion components, including energy storage system, generator, low-voltage system, andvehicle controller.

• Design cooling system and hydraulic system.

• Purchase and/or fabricate the propulsion system.

• Design, fabricate, and install vehicle subsystems.

• Install and test the propulsion system.

• Perform FMEA (failure modes and effects analysis).

• Design, fabricate, and install the driver’s station and controls.

• Perform limited testing and development of the complete vehicle.

• Continue efforts aimed at rapid commercialization.

IntroductionThe intent of this project is to perform theintegration and optimization of a hybrid orbattery/electric propulsion system and variousvehicle subsystems into a lightweight bus body. Off-the-shelf technology will be used wherever possible.Optimization of the propulsion and vehicle systemswill primarily be through careful selection ofappropriately sized components.

This project will result in a single proof-of-conceptprototype bus, suitable for testing and evaluation ofperformance under controlled conditions. The designwill not necessarily be ready for mass production,nor will it include trim and appearance items.

The completed prototype will include the primarybody structure, suspension, glazing, driver’s station,electric wheel motors, inverters, and energy storagesystem. It will also include regenerative braking,limited lighting, bumpers, and full seating. If ahybrid propulsion system is selected, an internalcombustion engine with generator, fuel tank, andhybrid electronics or controls will be included.

Propulsion System SimulationsThe lightweight stainless-steel body and chassis thathas been developed by Autokinetics for theFreedomCar and Vehicle Technologies Program(under another project) is an important enabler foradvanced technology propulsion systems of severaltypes. The purpose of the initial task of this projectwas to use computer simulation tools to evaluate theperformance of various candidate propulsion

systems in order to select the most promising one forfurther development.

The ADVISOR (version 2002) program developedby NREL (National Renewable Energy Laboratory)for the development of hybrid and other fuel-efficient vehicles was used for this study.

ADVISOR includes a substantial database ofvehicles and components, which allows comparisonsto be made with certain existing vehicles. Theapproach taken by Autokinetics was to use existingADVISOR models of a conventional diesel bus anda series hybrid bus as comparisons (the NovaBusRTS Conventional Diesel Bus and the Orion VI LowFloor Hybrid Bus). In addition, Autokinetics createdfour new models representing four differentpropulsion system architectures. These includediesel hybrid, fuel cell, battery-electric, and plug-inhybrid versions of the Autokinetics lightweight bus.Fuel consumption was evaluated by using the CBD14 drive cycle. This represents a city businessdistrict bus route and is commonly used in theindustry. For comparison purposes, the passengerload was standardized at 27 passengers for allvehicles (approximately one-half of the full seatedcapacity). The power required to run the air-conditioning system was not included, but it will beevaluated in a later task.

All fuel consumption results were compared bycalculating the gasoline-equivalent miles per gallon.A comparison of the energy efficiencies of thevarious buses is presented in Figure 1. This chartdoes not include the energy required to produce anddistribute the different types of fuels. This


195

Figure 1. ADVISOR results showing comparativevehicle-only energy efficiency of six buses with 50%passenger load.

information is available from the GREET(Greenhouse Gases, Regulated Emissions, andEnergy Use in Transportation) analysis modelcreated by the Transportation Technology R&DCenter of Argonne National Laboratory. Thespecific efficiency factors for diesel, hydrogen, andelectricity were provided by Ye Wu by using version1.6 of the GREET model. Figure 2 shows acomparison of the complete “well to wheels” energyefficiency of the six bus configurations.

Figure 2. ADVISOR results showing comparative well-to-wheels energy efficiency of six buses with 50%passenger load.

ConclusionsThe superior performance of a lightweight busplatform with a range of optimized propulsionsystems is clearly shown in the results of thiscomparison study. The four versions of theAutokinetics bus outperform the more conventionalbuses by a wide margin. The best performance isachieved by the battery-electric version, with animprovement of over 300% compared to the Orionhybrid diesel. Since this configuration is relativelysimple and straightforward to implement with off-the-shelf technology, it has been selected as the bestchoice of propulsion system for the prototype.


196

This document highlights work sponsored by agencies of the U.S. Government. Neither the U.S. Government nor any agency

thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for

the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use

would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name,

trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by

the U.S. Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect

those of the U.S. Government or any agency thereof.

Printed on recycled paper

A Strong Energy Portfolio for a Strong AmericaEnergy efficiency and clean, renewable energy will mean a stronger economy, a cleaner environment,

and greater energy independence for America. Working with a wide array of state, community, industry, anduniversity partners, the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy

invests in a diverse portfolio of energy technologies.

For more information contact:EERE Information Center

1-877-EERE-INF (1-877-337-3463)www.eere.energy.gov

2 0 0 4A N N U A L

P R O G R E S S

R E P O R T

L e s s d e p e n d e n c e o n f o r e i g n o i l , a n d e v e n t u a l t r a n s i t i o n t o

a n e m i s s i o n s - f r e e , p e t r o l e u m - f r e e v e h i c l e

HEAVY VEHICLESYSTEMS

OPTIMIZATION

FreedomCAR and Vehicle

Technologies Program

heavy vehicle systems optimization · pdf filereport less dependence on ... a study of...

Documents