400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org

SAE TECHNICALPAPER SERIES 2006-01-0197

The Effect of Swirl Ratio and Fuel InjectionParameters on CO Emission and Fuel Conversion

Efficiency for High-Dilution, Low-TemperatureCombustion in an Automotive Diesel Engine

Sanghoon Kook and Choongsik BaeKorea Advanced Institute of Science and Technology

Paul C. Miles and Dae ChoiSandia National Laboratories

Michael Bergin and Rolf D. ReitzUniversity of Wisconsin-Madison

Reprinted From: Compression Ignition Combustion Process 2006(SP-2012)

2006 SAE World CongressDetroit, Michigan

April 3-6, 2006

The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE's peer review process under the supervision of the session organizer. This process requires a minimum of three (3) reviews by industry experts.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE.

For permission and licensing requests contact:

SAE Permissions400 Commonwealth DriveWarrendale, PA 15096-0001-USAEmail: permissions@sae.orgTel: 724-772-4028Fax: 724-776-3036

For multiple print copies contact:

SAE Customer ServiceTel: 877-606-7323 (inside USA and Canada)Tel: 724-776-4970 (outside USA)Fax: 724-776-0790Email: CustomerService@sae.org

ISSN 0148-7191Copyright © 2006 SAE InternationalPositions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in SAE Transactions.

Persons wishing to submit papers to be considered for presentation or publication by SAE should send the manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE.

Printed in USA

Copyright © 2006 SAE International

ABSTRACT

Engine-out CO emission and fuel conversion efficiency were measured in a highly-dilute, low-temperature diesel combustion regime over a swirl ratio range of 1.44–7.12 and a wide range of injection timing. At fixed injection timing, an optimal swirl ratio for minimum CO emission and fuel consumption was found. At fixed swirl ratio, CO emission and fuel consumption generally decreased as injection timing was advanced. Moreover, a sudden decrease in CO emission was observed at early injection timings. Multi-dimensional numerical simulations, pressure-based measurements of ignition delay and apparent heat release, estimates of peak flame temperature, imaging of natural combustion luminosity and spray/wall interactions, and Laser Doppler Velocimeter (LDV) measurements of in-cylinder turbulence levels are employed to clarify the sources of the observed behavior. Mixing processes occurring after the pre-mixed burn are found to be the likely source of the optimal swirl ratio, while enhanced pre-combustion mixing dominates the reduction in CO with earlier injection. Liquid fuel films formed on the bowl lip are not found to significantly impact CO emissions, and increased injection pressure typically reduces CO emissions at this high dilution rate. Fuel conversion efficiency is examined in terms of component efficiencies related to combustion phasing, heat loss, and combustion efficiency. The influence of swirl and injection timing on each of these efficiencies is discussed.

INTRODUCTION

Diesel engines are an attractive option for automotive powerplants due to their high fuel conversion efficiency

(often termed the thermal efficiency), which can exceed the fuel conversion efficiency of modern SI engines by as much as 40% [1]. However, for conventional diesel combustion regimes, emissions of NOx and particulates from these engines are excessive, and costly exhaust-gas after-treatment systems are required to meet current emission regulations.

Recently, diesel combustion systems employing low combustion temperatures and enhanced fuel-air pre-mixing have been proposed to drastically reduce NOxand particulate emissions (e.g. [2]-[6]). In these systems, low combustion temperatures are achieved by employing high levels of exhaust gas recirculation (EGR) or retarded injection timing, while pre-mixing is enhanced by the use of split injections, high-injection pressures, or measures that increase the ignition delay. This work focuses on those systems that employ high EGR rates to maintain low combustion temperatures, with no special measures taken to increase pre-mixing other than the increased ignition delay associated with dilute mixtures.

Traditionally, low combustion temperatures are thought to reduce NOx emissions but increase particulate emissions due to a reduction in particulate oxidation rates. However, as EGR rates are increased to very high levels (approximately 60% or greater) simultaneous soot and NOx reduction is observed [3, 5]. At these high dilution levels, characterized by O2 concentrations of about 11%, the flame temperatures are limited to levels at which the soot and NOx formation rates are low. Unfortunately, the region of applicability of these low soot and NOx combustion regimes within the load-speed operating map is often limited by high CO and UHC emissions, with an accompanying fuel economy penalty.

2006-01-0197

The Effect of Swirl Ratio and Fuel Injection Parameters on CO Emission and Fuel Conversion Efficiency for

High-Dilution, Low-Temperature Combustion in an Automotive Diesel Engine

Sanghoon Kook and Choongsik Bae Korea Advanced Institute of Science and Technology

Paul C. Miles and Dae Choi Sandia National Laboratories

Michael Bergin and Rolf D. Reitz University of Wisconsin-Madison

CO emission generally stems from two primary sources. The first source, “under-mixing” of fuel, is the CO formed in the products of rich pre-mixed combustion. If mixing rates are too low, this CO will fail to mix with sufficient additional O2 to complete the oxidation process before expansion quenches chemical reactions. The second source, “over-mixing,” is associated with fuel that is mixed to very lean equivalence ratios during the ignition delay period. After ignition, these overly lean mixtures do not reach sufficiently high temperatures to complete the oxidation process on engine time scales.

For thoroughly pre-mixed, very lean combustion systems, the over-mixed fuel and subsequent low peak combustion temperature is the source of engine CO emission [7]. Similarly, the results of numerical simulations point to over-mixed fuel as the principal source of CO emission from conventional diesel combustion systems [8]. However, as dilution levels increase, the correlation between CO emission and ignition delay observed in dilute, low-temperature diesel combustion systems becomes negative [9]—that is, increased ignition delay is found to correspond to reduced CO emissions for highly-dilute mixtures. Additionally, the average equivalence ratio at ignition is estimated to exceed 2, and to increase with increasing dilution level. Jointly, these observations indicate that for highly-dilute systems the dominant CO emission source more likely stems from under-mixed fuel.

More importantly, significant changes in CO emission and in fuel economy can be obtained by changing the injection timing, as is shown in Fig. 1. With early injection, i.e. SOI=-26 CAD ATDC, CO emission is decreased 3-fold over emission levels measured with injection timing at -10 CAD ATDC. Clearly, mixing processes can be significantly influenced by injection timing, and potentially by other factors as well.

In this paper, we focus on clarifying the causes of the CO emission and fuel economy trends observed in Fig. 1, through application of a combined approach employing conventional pressure-based diagnostics, optical diagnostics, and numerical simulation. In addition to injection timing, we have also varied both the injection pressure and the swirl ratio, in an effort to further clarify both the dominant source of CO emission as well as the particular aspects of the mixing processes that influence CO emission and fuel economy. Increased injection pressure can be expected to primarily influence mixing processes during the ignition delay period, while increased swirl is anticipated to influence the later mixing processes to a greater extent than the injection-dominated early mixing processes. Additionally, swirl is a parameter which has been previously cited as important for reducing emissions [2] in low-temperature combustion systems, and is generally found to be beneficial for soot emission and fuel economy in conventional diesel combustion systems (e.g. [10]-[12]).

The dilution level employed is characteristic of the Toyota “smokeless” combustion regime [3] or the AVL HCLI regime [5], and represents a very low-NOx, low-soot operating regime where CO emissions and fuel economy have already begun to suffer [9]. In clarifying the source of CO emission and changes in fuel economy, our objective is to identify potential pathways for improvement of the performance of these combustion systems.

0

500

1000

1500

2000

220

230

240

250

260

270

280

290

300

-30 -25 -20 -15 -10

SOI [CAD]

Sk

ip-f

ire

d C

O E

mis

sio

n [

pp

m]

ISF

C [g

/kW

-hr]

ISFC 4%

CO

Figure 1 The influence of injection timing on CO emissions and fuel economy for a highly-dilute (O2=10.1%) intake charge measured in the same engine of this study. 1500 RPM, 3 bar IMEP, Pinj=800bar, Rs=3.77. Reproduced from Ref. [9] Figure 2 Optical engine geometry and experimental set-

up for velocity measurements. High-speed video imaging of combustion luminosity and spray penetration was also performed, observing the combustion chamber from below (via the 45° mirror) and from the side (via the liner windows)

Table 1 Engine specifications

Basic Geometry Bore: 79.5 [mm] Stroke: 85.0 [mm]

Disp. Vol.: 422 [cm3] Comp. Ratio: 18.7

Valve Events IVO: -360 CAD ATC EVO: 145 CAD ATC IVC: -140 CAD ATC EVC: -350 CAD ATC

Fuel Injection Equipment Bosch Flow No.: 320 Included Angle: 145˚Number of Holes: 6 Hydro-erosion: 10 %

Nozzle Style: Cylindrical Minisac (Bosch DLLA)

EXPERIMENT

RESEARCH ENGINE

A single-cylinder optically-accessible diesel research engine was employed to study the effect of swirl and injection timing on CO emission and fuel conversion efficiency. The engine has characteristics typical of small-bore engines intended for automotive applications, i.e.: four valves; a central, vertical injector; a 6-hole; minisac nozzle; and a concentric, re-entrant bowl. A schematic of the engine is shown in Fig. 2, and its main specifications are shown in Table 1. The floor of the piston bowl was comprised of a quartz insert that permitted in-cylinder fluid velocity measurements to be obtained from below via laser Doppler velocimetry.

The engine is equipped with a common-rail fuel injection system capable of a maximum injection pressure of 1350 bar. This system allows the injection pressure, injected mass (duration of injection) and start-of-injection (SOI) to be set independently. The injector is instrumented with a needle-lift sensor that permits the actual start of injection to be accurately determined.

The approximate fueling rate was determined by measuring the mass injected into a small chamber which seals against the engine head. The calibration was performed with the head/injector at operating temperature. Fuel was injected into the chamber for 575 injection events, and the average injected quantity determined by dividing the total mass by the number of injection events. This averaged value was used to calculate the fuel conversion efficiency. Note that this procedure accounts for thermal changes in the injector performance, but does not account for the higher ambient pressures encountered when the engine is operating. Accordingly, the average injected mass is likely to be overestimated, resulting in low values for the fuel conversion efficiency. Neglect of the influence of cylinder pressure will also introduce a bias into the estimated fueling rate, wherein the injected fuel amount at later injection timing will be less due to the higher cylinder pressures.

The magnitude of the overestimation/bias in the injected fuel mass can be estimated by assuming that, at the baseline injection pressure of 800 bar, a mean effective injection pressure of 400 bar is achieved. At the earliest injection timing employed, the average cylinder pressure during injection is 22 bar, resulting in an overestimation of the injected fuel by 3%. At the latest injection timing, the 47 bar cylinder pressure results in a 6% overestimation. As will be seen below, the magnitude of this bias is small as compared to the magnitude of the changes in fuel conversion efficiency observed.

In-cylinder pressure measurements were acquired using a water-cooled (KISTLER 6043A60) piezoelectric pressure transducer. Measurements obtained over 50 engine cycles were averaged to calculate the indicated

mean effective pressure (IMEP) and the apparent heat release rate. Care was exercised to ensure that pressure-record filtering did not introduce artificial side-lobes in the calculated apparent heat release rates. Additional details of the calculation methodology are presented in our earlier work [9]. In-cylinder pressure data, and the velocity data described below, were acquired with a resolution of 0.25 crank angle degrees (CAD). Unless otherwise specified, all crank angles presented in this paper are reported as CAD after TDC compression.

Engine-out CO emission was measured using a California Analytical Instruments NDIR (Non-Dispersive Infra-Red) CO analyzer (model 300). Samples were taken from the exhaust plenum and transferred to the analyzer through a heated sample line. Moisture and condensable hydrocarbons were removed from the sample by a condenser prior to the analyzer.

OPTICAL DIAGNOSTICS

Spray and Luminosity Imaging

To provide information on spray-targeting, an expanded 514.5 nm Ar+ ion laser beam was directed into the combustion chamber through one of the side liner windows shown in Fig. 2. A high speed digital video camera (Integrated Design Tools Inc., X-Stream Vision) was employed to view the spray penetration and wall-film formation from below, via the extended piston and 45° mirror. The frame rate was set to 21,593 frames per second, corresponding to 0.4168 CAD between images at 1,500 rpm. The image resolution is 100 x 120 pixels and an exposure time of 30 μs was employed.

High speed videos were also obtained of the natural luminosity from soot emissions, viewing the combustion chamber through a side window. In this case, the framing rate employed was 9,000 frames per second, corresponding to 1 image per crank degree.

Table 2 Operating Parameters

Basic Geometry Speed [rpm] 1500

Load 3 bar gross IMEP TCoolant [°C] 88

Simulated EGR [Mass %] 65

Intake charge composition 10.1% O2, 80.6% N2,9.3% CO2

Typical In-Cylinder 0.57 PInjection [bar] 600, 800, 1000, 1200

Swirl Ratio (Rs) 1.44, 2.59, 3.77, 4.94, 7.12PIntake [bar] 1.2 TIntake [°C] 90

PExhaust [bar] 1.3 Start-of-Injection (SOI)

[CAD ATDC] -30.25 to -7.75

Flow Velocity

Measurements of the tangential component of velocity were acquired with a custom-designed, fiber-coupled LDV. A laser power of approximately 260 mW per beam was used at a wavelength of 514.5 nm. A 5.0 MHz differential frequency shift removed directional ambiguity. ZrO2 seed particles were introduced with a fluidized bed seeder. The seed particles have a mean diameter of 0.48 m and a standard deviation of 0.15 m. Additional details can be found in Ref. 21.

ENGINE OPERATION

The engine operating parameters are described in Table 2. All data were acquired at a fixed engine speed of 1500 rpm.

Recirculated exhaust gas was simulated by diluting the intake air stream with N2 and CO2. The relative proportions of CO2 and N2 were chosen to match the mixture molar specific heat of real engine exhaust gas at the selected load and O2 concentration. This matching was performed with gas properties evaluated at 600 K. The 10.1% O2 concentration selected corresponded to a 65% exhaust gas recirculation (EGR) rate. At Rs = 3.77 the typical in-cylinder equivalence ratio ( ), was 0.57 ( = 1.75), corresponding to an equivalence ratio based on intake air of 0.79 ( = 1.27).

Prior to obtaining in-cylinder pressure, exhaust emissions and flow velocity, the engine was motored for a minimum of 90 s in order to pre-heat the combustion chamber walls and to allow the intake plenum pressure to stabilize. For in-cylinder pressure and exhaust emission data, the engine was skip-fired for 70 s, with

fuel injection occurring on only one of every four engine cycles. This skip-firing period allowed the species concentrations in the exhaust system and the emission analyzers to stabilize. Finally, data were acquired over 50 additional skip-fired cycles. For velocity measurements, data were acquired for 120 s after the commencement of skip-fired operation. After data acquisition, a fixed cool-down/cleaning period was employed to ensure that the gross thermal state of the engine was the same for all measurements.

The swirl ratio is varied by throttling the low-swirl intake ports. To achieve the large range of swirl ratios reported here, the intake valve in the high-swirl port was fitted with a 180° shroud, oriented to enhance the component of the intake flow tangential to the cylinder wall. The swirl ratios reported in Table 2 are the numeric values resulting from flow-bench measurements performed by Ricardo, Inc. Radial profiles of the tangential velocity at three axial locations, measured with the LDV through the liner windows, provide a more direct measure of the in-cylinder angular momentum. At -55 CAD, integration of these profiles resulted in swirl ratios of 1.10, 2.75, 3.86, 5.05, and 6.53. With the exception of the lowest swirl ratio, these measurements are within 8% of the flow-bench swirl ratios. The swirl ratios reported throughout the remainder of this document are those determined via flow-bench.

The fueling rate was controlled to provide a constant 3 bar gross IMEP (indicated mean effective pressure) for each swirl ratio, at the injection timing providing best fuel economy. Thereafter, as SOI was varied, the fueling rate was held fixed. The most advanced injection timing considered at each swirl ratio was determined by the achievable IMEP. When the IMEP dropped significantly from the desired load no earlier timings were considered. As discussed below, excessive spray penetration due to low in-cylinder gas density is thought to limit the earliest practical injection timing. Injection timing was retarded to the crank angle at which impaired fuel economy was observed at the lower swirl ratios, typically -10.25 CAD. The approximate fueling rates employed for each swirl ratio are listed in Table 3. The fuel employed was a 2007 emission certification diesel with a cetane number of 47.1 and a lower heating value of 42.98 MJ/kg.

Table 3 Variation in fueling rate with swirl ratio

Swirl Ratio Inj. Duration [ sec]

Fueling Rate [mg/stroke]

1.44 409 8.325 2.59 409 8.325 3.77 421 8.908 4.94 425 9.108 7.12 456 10.258

NUMERICAL SIMULATION

In previous studies, the three-dimensional intake flow field was calculated with STAR-CD. At IVC, those results were mapped to a sector grid and used to initialize the KIVA-3V [14] simulations of the combustion process. However, a comparison of the temporal history of the tangential velocity and turbulent kinetic energy for various initialization schemes revealed that a simpler solid-body-rotation swirl initialization performed well compared to the mapped results [15]. Due to the computational expense of computing the intake flow for the range of swirl velocities considered here (Rs = 1.44–7.12), solid-body-rotation was used for initialization of the tangential velocity in this study. The usual practice of initializing the axial velocities by linearly interpolating between the piston surface and the head, and initializing the radial velocities to zero was followed.

The spray and combustion sub-models employed are described more fully elsewhere ([15],[16]). The “characteristic time” combustion sub-model [17], is of greatest relevance to the present study. In this model, the instantaneous production and oxidation rates of the important combustion species are not calculated directly via either simplified or detailed chemical mechanisms. Rather, the rate of change of each species is calculated as proportional to the difference between the actual species mass fraction and its instantaneous thermo-dynamic equilibrium value. The constant of propor-tionality (the inverse of the characteristic time) is a weighted combination of a chemical timescale and a turbulent timescale, with progressively greater weighting toward the turbulent scale as the combustion proceeds.

RESULTS AND DISCUSSION

The results of this study are presented in four main parts. In the first part, the variation in CO emissions with swirl ratio and injection timing is reported. Second, the source of the influence of swirl ratio on CO emissions is clarified through examination of the heat release characteristics, numerical simulations and in-cylinder imaging of combustion luminosity. Third, the factors potentially responsible for the observed dependency of CO emissions on injection timing are examined. Finally, we examine the influence of both swirl ratio and injection timing on fuel conversion efficiency.

CO EMISSION TRENDS

Figure 3 shows the engine-out CO emissions as a function of both swirl ratio and injection timing (SOI). Two features are particularly notable:

First, at fixed injection timing, there is generally an optimal swirl ratio that gives the lowest CO emission. The minimum CO emission is typically measured near the swirl ratio of 2.59. As noted in the Introduction, at this high dilution rate the dominant source of CO emission is thought to be under-mixed fuel. Thus, the

observed optimal swirl ratio is somewhat unexpected. Higher swirl ratios are typically thought to enhance turbulence production and mixing processes, resulting in more rapid, complete combustion [18]. Consequently, if under-mixed fuel is the dominant source of CO emissions, we would anticipate a continuous improvement in CO emissions as the swirl level is increased.

The existence of an optimal swirl level suggests that a mechanism by which the highest swirl levels impede mixing and combustion is present. However, the traditional mechanism by which excessive swirl is thought to impede mixing, by limiting jet penetration (e.g. [19],[20]) is not believed to be significant for the small bore diameters and high injection pressures that characterize modern automotive diesel engines. This belief is supported by Fig. 4, which shows the fuel mass distributions predicted for various swirl ratios at -17 CAD ATDC. In these simulations, SOI was fixed at -22.25 CAD and the crank angle shown occurs shortly after the end of injection. A high fuel mass fraction is observed near the bowl lip at all swirl ratios, indicating penetration of the fuel jet to the maximum possible extent. At the higher swirl ratios, more fuel is observed in the cut-plane shown (22° down-swirl of the fuel jet) due to greater convection by the swirl velocity.

The second dominant feature of Fig. 3 is the rapid decrease in CO emission observed at fixed swirl ratio when the injection timing is advanced. This behavior is also apparent in Fig. 1. The decrease is larger at the higher swirl ratios. For a swirl ratio of 7.12, CO emission was reduced by over 83% as the injection timing was advanced from -22.25 to -30.25 CAD ATDC.

In the following sections we investigate the causes of these two behaviors.

OPTIMAL SWIRL RATIO

1000

1500

2000

2500

3000

3500

-30 -25 -20 -15 -10

2

3

4

5

6

7

SOI [CAD ATDC]

Sw

irl R

atio

[ppm]

4 Skip-Fired

CO Emission

Figure 3 Effect of swirl ratio and injection timing on CO emission

As a first step toward examining the cause of the optimal swirl ratio exhibited in Fig. 3, the variation in the measured apparent heat release rate and in the integrated, cumulative heat release rate with flow swirl level are shown in Fig. 5 for fixed SOI = -22.25 CAD. Note that, as observed previously for conventional diesel combustion over a more limited range of swirl ratios [12], increased swirl is found to reduce the ignition delay and to advance the phasing of the early portion of the heat release.

A second, more important observation can be made from the cumulative heat release behavior shown in the lower part of Fig. 5. By about -4 CAD, the heat release rate observed for Rs = 7.12 slows considerably, followed by slowing of the heat release observed at Rs = 4.94 and Rs = 3.77.

Beyond TDC the cumulative heat release curves continue to diverge, but by this time a clear ordering with swirl ratio in the amount of heat released has been established. This ordering persists until EVO. The behavior of the cumulative heat release curves suggests that the mixing processes responsible for this ordering commence between -5 CAD and TDC, and endure for several crank degrees thereafter.

The ordering in the cumulative heat release is also found to correlate inversely with the engine-out CO emissions. That is, lower CO emissions are observed for swirl ratios providing the highest cumulative heat release. Accordingly, insight into the mechanisms responsible for the CO emission behavior are likely to be found by examining the mixing processes occurring during the latter portion of the combustion process, beyond -5 CAD.

From Fig. 5, it is also apparent that, in addition to high CO emissions, the high swirl cases will suffer from reduced fuel economy—a topic that is discussed in greater detail below.

To help clarify the cause of the behavior seen in Figs. 3 and 5, the flow structures and spatial distributions of CO (or partially-burned fuel) and O2 computed in the

Figure 5 Variation in the apparent heat release rate and the cumulative heat release rate as the swirl ratio is varied. SOI = -22.25 CAD

Figure 4 Predicted fuel mass distribution for various swirl ratios at -17 CAD ATDC: SOI = -22.25 CAD, Pinj = 800 bar. The r-z plane velocity vectors are also shown superimposed

0

10

20

30

40

50

60

70

80

1.442.593.774.947.12

Rs

0

0.1

0.2

0.3

0.4

0.5

0.6

-15 -10 -5 0 5 10 15 20

Crank Angle [CAD]

Increasing Swirl

App. H

eat R

ele

ase R

ate

[J/C

AD

]C

um

ula

tive

AH

RR

/ m

fQ

LH

V

numerical simulation were examined. In our past work, we have shown that the mean (bulk) flow structures predicted by the simulation capture the measured mean flow structures remarkably well [21]. However, this comparison was performed for a single, moderate swirl ratio. To extend this validation to the higher swirl ratios employed here, we have measured axial profiles of the tangential velocity in the central region of the bowl at each swirl ratio, and compare these to the simulation results in the Appendix. Although some differences exist, overall the comparison shows that the axial distribution of the tangential velocity is similar to the computed results, and evolves in a like manner.

Figure 6 shows the variation in the r – z plane flow structures and in the spatial distribution of CO mass fraction predicted for various swirl ratios at 3 CAD. This

crank angle is during the mixing-controlled portion of combustion, just after the secondary peak in the apparent heat release rates seen in Fig. 5. At this time, the cumulative heat release rates seen in the lower portion of Fig. 5 continue to diverge.

At the lowest swirl ratio, Rs = 1.44, the bulk flow structure is dominated by a single clockwise-rotating structure that fills the outer portion of the bowl. In the central region of the bowl, above the pip, a smaller secondary vortex rotating in the opposite direction is seen. From Fig. 7, which shows the distribution of the remaining O2 in the combustion chamber, it is seen that this secondary vortex transports O2 to the same region where the dominant vortex transports CO. In this interfacial region between the two vortices, steep gradients in swirl velocity (Fig. 8), as well as high rates

Figure 7 Changes in the spatial distribution of O2 at 3 CAD for various swirl ratios, as predicted by numerical simulation. SOI = -22.25 CAD, cut-plane 22° down-swirl from the fuel jet

Figure 6 The mean r – z plane flow structure and the spatial distribution of CO at 3 CAD for various swirl ratios, as predicted by numerical simulation. SOI = -22.25 CAD, cut-plane 22° down-swirl from the fuel jet

of mean flow deformation in the r – z plane ([22],[23]) lead to enhanced turbulence production. Thus, the observed pair of counter-rotating flow structures forms a mixing system, wherein large-scale structures transport the partially-burned fuel and O2 to a common interface, while simultaneously providing mechanisms to produce the turbulence needed to achieve rapid small-scale mixing.

As swirl ratio is further increased to 2.59, the same dual-vortex flow structure is again apparent. In this case, the centrifugal forces acting on the high angular momentum fluid transported into the inner regions of the bowl are larger, and the lower vortex is smaller due to the tendency of this high angular momentum fluid to return to the bowl periphery. Consequently, the upper vortex is larger and encompasses a greater fraction of the

remaining O2. The mixing system created is thus anticipated to be more effective than is seen for Rs = 1.44, an expectation which is supported by the heat release characteristics seen in Fig. 5. The dual-vortex structures formed at these swirl ratios are long-lived; the simulation results suggest that they continue to provide effective mixing for at least another 5 CAD beyond the crank angle illustrated in Figs. 6–8.

The formation of these structures correlates well with the secondary peak in the apparent heat release rate, seen near TDC in Fig. 5. Similar dual-vortex structures, whose formation coincides with increased heat release rates, have also been observed under late-injection operating conditions [23].

Secondary peaks in the experimental heat release rate near TDC are also seen in Fig. 5 for the higher swirl ratios. Although small, these peaks are both reproducible and of a magnitude that easily exceeds the precision of the measured heat release during this portion of the cycle, approximately ±1 J/CAD [12]. Corresponding to the appearance of these peaks, the numerical simulations also predict the formation of counter-rotating vortex structures in the higher swirl cases, though at slightly earlier crank angles. The important distinction appears to be the lifetime of these structures, and the manner in which they evolve and transport the unburned fuel and oxidant later in the cycle. Already by 3 CAD, Figures 6 and 7 indicate that for Rs = 3.77 a third vortex structure is forming in the central portion of the cylinder. The original upper vortex is displaced outward, and rather than acting to bring the partially-burned fuel (CO) and O2 to the same interface, it now predominantly acts to transport the partially-burned fuel away from the O2 in the center of the bowl. This effect becomes even more pronounced as the swirl level increases further. By Rs = 7.12 yet a fourth vortical structure is apparent, and the original bowl vortex— dominant at Rs = 1.44—is now barely visible in the reentrant portion of the bowl.

Comparison of the CO spatial distributions in Fig. 6 with the swirl velocity distributions in Fig. 8 reinforces the above discussion. In general, the partially-burned fuel is spatially co-located with the high-swirl fluid. As the swirl ratio increases, this co-location results in greater average radial locations of the CO, due to the return of the high-swirl fluid to the combustion chamber periphery. At moderate swirl ratios (e.g. Rs = 2.59) this is advanta-geous: not only is the dual-vortex structure formed, but the partially-burned fuel is positioned to exit the bowl into the squish volume, where additional turbulence generated by the exiting flow is anticipated to facilitate mixing with the remaining O2 in the squish volume.

In contrast, at the highest swirl ratios, co-location with high-swirl fluid both impedes mixing with the O2 in the center of the bowl and interferes with the exit of the partially-burned fuel from the bowl with the squish flow. Fundamentally, the centrifugal forces acting on the high-swirl fluid prevent the inward displacement required to

Figure 8 Changes in the spatial distribution of the swirl velocity at 3 CAD for various swirl ratios, as predicted by numerical simulation. SOI = -22.25 CAD, cut-plane 22° down-swirl from the fuel jet

move the fluid over the bowl lip. The co-location of the CO and the high-swirl fluid therefore persists until well into the expansion stroke, effectively trapping the CO in the re-entrant portion of the bowl. This is illustrated in Fig. 9 for the highest swirl case, Rs = 7.12.

To provide additional experimental support for this picture, the spatial distribution of soot luminosity, and its temporal evolution, was captured using the high speed

digital video camera. Figure 10 compares the side-view images of the natural soot luminosity observed at Rs = 1.44 with that observed at higher swirl ratios, Rs = 3.77 and Rs = 7.12. Although the intensity of the soot luminosity is less, images obtained at the higher swirl ratios clearly show that the luminous soot (which is anticipated to be co-located with partially-burned fuel) is found deeper in the bowl and more uniformly fills the bowl than at the lower swirl ratios. This condition

Figure 9 Simulation results illustrating the co-location of CO (partially-burned fuel) and high-swirl fluid during the expan-sion stroke. Rs = 7.12

Figure 10 Images of natural luminosity from soot emissions at three swirl ratios. At higher swirl, the visible soot is lower and more uniformly fills the bowl. Clockwise from the upper-left corner of each figure are superimposed the transmissivity of the neutral density filter employed, the exposure time, the crank angle relative to SOI, and the true crank angle at which the image was obtained

persists over all later crank angles during which the bowl is visible.

In summary, the numerically predicted bulk flow structures provide a mechanism that explains the observed optimal swirl ratio. At the lowest swirl ratios, a dual-vortex structure forms that enhances the mixing of partially-burned fuel and air. For moderately increased swirl ratios, this dual-vortex system appears to be more effective in enhancing the mixing processes. However, as swirl is increased further, the bulk structures created actually impede the late-cycle mixing, even though early mixing and heat release appears to be enhanced. Ultimately, too high a swirl ratio leads to trapping of partially-burned fuel within the bowl.

RAPID DECREASE IN CO EMISSION WITH ADVANCED SOI

The potential causes of the second major observation made from Fig. 3, a decrease in CO emissions as SOI is advanced, are investigated in this section from several perspectives. Once again we rely heavily on the results of numerical simulations, which suggest that the initial degree of fuel-air mixing achieved is a critical factor. To support this finding, we examine the measured influence on CO emission of injection pressure and ignition delay. The correlation of CO emissions with peak flame temperature is also examined for evidence that early CO oxidation rates could also be an important factor influencing CO emission. Finally, we scrutinize other processes that influence the early mixture preparation process—including wall-wetting, spray-targeting, and flow turbulence.

Simulation Results

Figure 11 shows simulation results for the total, in-cylinder mass of CO for two different injection timings at the highest swirl ratio, Rs = 7.12. Also shown are

experimentally-determined apparent heat release rates for each case. There is a noticeable difference in the amount of CO present in the cylinder, in agreement with the experimental results shown in Fig. 3. For both cases, CO formation started in the middle of the premixed burn phase and the peak CO mass is found near the start of the mixing-controlled burn phase. Examining the evolution of the CO mass, it is apparent that the latter oxidation rates are similar and that the lower CO mass observed for the earlier injection timing is principally due to lower initial CO formation or more rapid early oxidation of CO.

The nature of the characteristic-time combustion sub-model is such that CO must be formed from rich mixtures. For lean mixtures, instantaneous equilibrium CO concentrations are small and the model will not predict large CO production rates; i.e., CO emissions stemming from incomplete combustion of “over-mixed” fuel will not be captured. Thus, the higher CO mass formed at the later injection timing must be caused by the presence of a greater amount of rich mixture. The fraction of rich mixture found in the cylinder for both injection timings, shown in Fig. 12, supports this conclusion. At the earlier injection timing, the peak fraction of the in-cylinder mass characterized by fuel-rich conditions is smaller, and the fuel-rich mass fraction is already decreasing at the onset of combustion near -12 CAD. For the later injection timing, however, the peak mass of rich in-cylinder mixture is not reached until after TDC—well into the mixing-controlled portion of the combustion process.

The better mixture preparation characteristics observed at the earlier injection timing in Fig. 12 are not the result of an increased ignition delay. While increased ignition delay is undoubtedly beneficial, the mixing process itself is enhanced with the earlier injection timing, such that for equal times after the start of injection, a smaller fraction of the in-cylinder mass is fuel-rich for the earlier injection

0.0

0.5

1.0

1.5

0

20

40

60

80

100

-30 -20 -10 0 10 20Crank Angle [CAD ATDC]

In-Cylinder CO mass

Apparent Heat Release Rate

SOI: -30.25 CAD ATDC

SOI: -22.25 CAD ATDC

In-C

ylin

de

r

CO

Mass [m

g]

Appara

nt H

eat R

ele

ase R

ate

[J/C

AD

]

Figure 11 Spatially-integrated in-cylinder CO mass from the numerical simulations, and corresponding exper-imental apparent heat release for different injection timings. Rs=7.12

SOI: -30.25 CAD ATDC

SOI: -22.25 CAD ATDC

0

0.1

0.2

0.3

-40 0 40 80 120Crank Angle [CAD ATDC]

Ric

hMass Fraction

( o

> 1

)

Figure 12 Numerical simulation of the mass fraction ofthe cylinder contents characterized by a (pre-combustion) equivalence ratio greater than 1. Rs=7.12

timing. This enhanced mixture preparation is reflected in the higher peak heat release rates seen in Fig. 11 for SOI=-30.25, behavior which is well-captured in the simulation results, as are the ignition delay periods observed for both injection timings.

Examination of the spatial distributions of fuel predicted numerically suggest that the enhanced mixing observed with earlier injection is not due to a significant difference in the flow structures existing during the ignition delay period, but rather to initial deposition of a significant fraction of fuel vapor within the squish volume. As the piston rises, this fuel is returned to the bowl with the squish flow and results in a better mixed, broader spatial distribution of fuel vapor than is seen with later injection. Peak fuel vapor concentrations (and subsequent CO concentrations), for both injection timings, are found deep in the bowl, as can be seen in the lower portion of Fig. 6. In both cases, the CO remains trapped in the lower regions of the bowl during expansion. For the earlier injection timing, however, there is far less of it.

Thus, the simulation results suggest that the dominant cause of the reduced CO emissions with advanced injection timing is associated with reduced peak CO mass during the early combustion process. Early mixing processes, occurring prior to or during the first portion of the premixed burn, are at least partly responsible for the reduced CO mass.

The Influence of Injection Pressure and Ignition Delay

The numerical simulations of the previous section suggest that more rapid initial mixing is an effective method of reducing CO emission—provided that CO emission stems predominantly from under-mixed fuel. A straight-forward test of this prediction can be made through increasing injection pressure, a measure that will enhance initial mixing rates but which is less likely to significantly impact the late-cycle mixing. As demonstrated by Fig. 13, increased injection pressure is found to generally reduce CO emissions, in agreement

with the numerical predictions. At the earliest injection timings, however, increased injection pressure may be detrimental.

Ignition delay is well-known to increase as injection timing is advanced, due to the lower ambient temperature and pressure at earlier SOI [24]. Here, the ignition delay is defined as the time from SOI until the time at which 10% of the total heat release has occurred. The 10% heat release timing correlates well with the beginning of rapid, high-temperature heat release. As noted in the introduction, we have previously observed an inverse correlation at high dilution rates between CO emission and ignition delay. In general— provided spray targeting, wall impingement, or spray/flow interactions do not vary significantly with SOI— increased ignition delay should correlate well with a higher degree of fuel-air premixing†. In turn, more premixing will correlate inversely with CO emission, provided CO emission is dominated by the combustion of rich mixtures. Conversely, CO emissions resulting from over-mixed fuel would be expected to increase with increasing ignition delay.

The correlation between ignition delay and CO emission measured for Rs=3.77 is shown in Fig. 14. This intermediate swirl ratio was selected for detailed examination because a minimum in CO emission is observed near SOI =-26 CAD in both Figs. 1 and 3. In keeping with the predictions of the numerical simulations, the variable injection pressure tests, and our previous results, Fig. 14 shows a strong negative correlation of CO emission with increasing ignition delay, indicating the benefits of increased early mixing for reducing CO emission.

At the earliest injection timing shown in Fig. 14, increased CO emission is observed despite an increase † This view must be adopted with caution. As shown in the previous section, under some circumstances mixing-rates can be affected significantly by changes in SOI.

600

1000

1400

1800

1.4 1.6 1.8 2 2.2

τid

[ms]

Rs=3.77

SOI=-26.25 CAD ATDC

SOI=-22.25 CAD ATDC

SOI=-28.25 CAD ATDC

4 S

kip

-Fire

d

CO

Em

issio

n [

pp

m]

Figure 14 Correlation of CO emission with the ignition delay at swirl ratio of 3.77

500

1500

2500

3500

4500

-30 -25 -20 -15 -10 -5

600 bar800 bar1000 bar1200 bar

Pinj

:

SOI [CAD ATDC]

4 S

kip

-Fire

d

CO

Em

issio

n [

pp

m]

Figure 13 Effect of injection pressure and injection timing on CO emission at swirl ratio of 3.77

in ignition delay. This behavior is both repeatable and reflected in the CO emission behavior seen at other swirl ratios (see Fig. 3). Furthermore, it is consistent with the behavior seen at the earliest injection timings of Fig. 13, where increased injection pressure was no longer beneficial. It would appear that at very early injection timing, factors other than ignition delay are influencing pre-mixing and subsequent CO emission—behavior that will be examined more closely below.

Flame Temperature

The evidence presented thus far suggests that decreased peak CO mass early in the combustion process is primarily responsible for the decrease in CO emissions observed with advanced injection timing. Simulation results suggest that the decreased CO mass is due to enhanced mixing. However, earlier injection results in advanced combustion phasing which results in higher peak cylinder pressures and higher peak combustion temperatures. Accordingly, it seems plausible that the reduction in peak CO mass may be partly associated with the more rapid early oxidation of CO due to these higher temperatures.

We have selected the peak adiabatic flame temperature to characterize the CO oxidation rates early in the combustion process (i.e., during the premixed burn period). The peak flame temperature is the adiabatic flame temperature achieved by combustion of a stoichiometric mixture at the peak core charge temperature, which is estimated assuming adiabatic compression of the in-cylinder charge. Although CO oxidation will always occur at temperatures lower than this, the relative magnitude of the peak flame temperature is likely a good indicator of the relative magnitude of CO oxidation temperatures.

The correlation of CO emission with peak flame temperature is shown in Fig. 15, where the CO emissions measured at Rs = 3.77 are plotted on a logarithmic axis. As anticipated, the peak flame

temperature increases monotonically as injection is advanced. Notably, as SOI is advanced to -22.25 CAD, there is little change in CO emission despite a near 30 K change in peak flame temperature. As SOI is further advanced to -26.25 CAD, however, a significant re-duction in CO emission is observed, although the flame temperature only increases by 5 K. Overall, the observed correlation between peak flame temperature and CO emission is not consistent with the proposition that early CO oxidation significantly influences the emission level. The subsequent increase in CO emission as SOI is further advanced to -28.25 CAD, while the peak flame temperature continues to increase, reinforces this view.

Spray Penetration and Targeting

An additional important factor related to early mixture formation processes in small-bore diesels is spray penetration and targeting. Excessive penetration will clearly lead to piston wall-wetting, and in extreme cases can lead to liquid fuel impacting the cylinder liner (e.g.[25]). Spray targeting is expected to influence the initial mixture formation process by affecting the initial “splitting” of the spray upon striking the bowl lip, and to further influence the size and location of liquid films that may be formed on the piston surface.

Figures 16 and 17 illustrate the liquid spray penetration and wall-film formation, as determined from high-speed imaging of the liquid sprays. Also shown, below the experimental images, is the evolution of the gas-phase fuel distribution predicted in the numerical simulations. The computed fuel distributions are shown for a vertical cut-plane 22° downstream of the spray axis. The center of the piston and location of the bowl lip are scaled to the spray image and are designated with dashed lines. The time after the start of injection and the current crank angle are shown at the right-hand side of each gas-phase fuel mass image.

For the injection timing of -26.25 CAD seen in Fig. 16, where the lowest CO emission was observed at the Rs = 3.77 swirl ratio, the spray tip clearly impacts the bowl lip and creates a liquid film. In contrast, with SOI = -22.25 CAD, Fig. 17 shows that the spray tip reaches to the bowl lip, but little or no liquid film can be seen. Formation of liquid films is generally thought to increase UHC emissions (e.g. [26]), and would likely increase CO emissions as well due to an overall retarded fuel preparation process. The lower CO emission observed at the earlier injection timing suggests, therefore, that formation of wall films is not a dominant process influencing CO emission here.

A second interesting observation can be made from Figs. 16 and 17. The predicted gas-phase fuel mass distributions show that for the earlier, -26.25 CAD injection timing, some of the fuel penetrates into the squish area. However, the squish flow induced by the rising piston pushes this fuel back into the bowl. By the end of the sequence shown, the fuel distribution is

4 S

kip

-Fire

d

CO

Em

issio

n [

pp

m]

600

700

800900

1000

2000

2000 2010 2020 2030 2040

Peak Flame Temperature [K]

SOI=-26.25 CAD ATDC

SOI=-22.25 CAD ATDC

SOI=-28.25 CAD ATDC

Rs=3.77

Figure 15 Correlation of CO emission with peak flame temperature at a swirl ratio of 3.77

almost identical for both injection timings. Accordingly, the initial distribution of the vapor phase fuel does not appear to be a dominant factor in determining the CO formation and ultimately the CO emission level. This stands in contrast to our previous discussion of the fuel vapor distributions observed for Rs = 7.12. Recall that, in that case, significantly more advanced injection timing was considered.

At these more advanced injection timings, liquid films can be expected to form not only on the bowl rim, but also on the piston top. Figure 18 shows an idealized view of the spray targeting on the piston for various crank angles. With a start of injection at -26 CAD and later, we predominantly expect wetting of the bowl rim, as was seen in Fig. 16. Minor wetting of the piston top may occur with SOI = -26 CAD, but only from the earliest portion of the injection event. For earlier injection timings,

Figure 17 Visualized spray impingement on the bowl lip and predicted fuel mass distribution: SOI = -22.25 CAD, Rs = 3.77

Figure 16 Visualized spray impingement on the bowl lip and predicted fuel mass distribution: SOI = -26.25 CAD, Rs = 3.77

wetting of the piston top is expected to become increasingly important, due to both increased penetration and the spray targeting. At high swirl ratios, we anticipate that vaporization of these films will be more rapid, and their potential influence on emissions diminished. Piston top wall films are thus a possible explanation for the increased CO emissions seen in Fig. 3 for the earliest injection timings and lower swirl ratios.

Turbulence Enhancement

A final factor that may influence the early mixture preparation process and increase mixing rates during the ignition delay period is enhanced flow turbulence. In earlier work we have shown that, over the course of the latter compression stroke, production of turbulence by bulk compression acting on the isotropic normal stresses can often dominate [16]. For a uniform density flow, this production can be expressed as:

dt

dV

VkPiso

13

2 (1)

Here, the production of turbulence is directly proportional to the level of pre-existing turbulent kinetic energy k, and the negative of the logarithmic rate-of-change of cylinder volume - dtdVV1 —a quantity which is large between about -30 and -5 CAD. Thus, turbulent energy “injected” earlier is amplified by the compression process over a

greater period of time, and could significantly enhance the early mixture preparation process.

To examine this potential source of enhanced mixing, tangential velocity data were acquired at a radius of r = 13.6 mm and axial locations of z = 2 – 20 mm below the cylinder head. The data discussed below were obtained at the locations z = 4 – 10 mm, which are shown in Fig. 19, along with superimposed bowl outlines which show the piston position relative to these measurement locations at various crank angles. At the early injection timings considered here, these measurement locations were approximately centered along the path of a fuel spray during the injection event (cf. Fig. 18). Later, these locations are characteristic of locations within the central region of the bowl—where much of the later mixing occurs. All measurements were obtained in a plane 22° down-swirl of a fuel jet.

The measured temporal evolution of the mean flow velocity and RMS fluctuations are shown in Figs. 20 and 21, respectively. Velocities measured under both motored and fired operation are presented in each figure. The swirl ratio shown is 7.12. Recall that, as the injection timing is advanced, the maximum reduction in CO emission was observed at Rs = 7.12 (see Fig. 3). Measurements obtained at injection timings of -22.25 and -30.25 CAD are presented.

The mean flow evolution, shown in Fig. 20, is consistent with both physical expectations and with the results of numerical simulations. For motored operation, Fig. 20 shows that the tangential velocity increases in the latter part of the compression stroke as high angular momentum is transferred into the bowl by the squish flow. With fuel injection, little influence of the injection event is observed until after injection has ended. After the fuel injection event, a distinct perturbation in the tangential velocity is observed at all injection timings and measurement locations, associated with a displacement of the high angular momentum fluid near the bowl rim by the fuel spray. This perturbation rapidly subsides, however, and beyond about -10 CAD there is little

z[m

m]

-20

-15

-10

-5

0

-25 CAAD ATDC

0

CA

TDCD

r = 13.6 mm3r

Figure 19 Velocity measurement location and piston position at -25 CAD ATDC: r = 13.6 and various axial (z)locations

-5 CAD

-15 CAD

-22 CAD

-26 CAD

-30 CAD

Figure 18 Idealized depiction of the un-deflected spray targeting for various crank angles

Tangential V

elo

city [

Sp]

Crank Angle [CAD ATDC]

0

2

4

6

8

10

-30 -20 -10 0 10

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 10 mm

Fuel Injection Event

2

4

6

8

10

Motored-22.25 SOI-30.25 SOI

z = 8 mm

Fuel Injection Event

2

4

6

8

10

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 6 mm

Fuel Injection Event

2

4

6

8

10

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 4 mm

Fuel Injection Event

Rs = 7.12

Pinj = 800 bar

r = 13.6 mm

Figure 20 Measured mean velocity in tangential direction at r = 13.6 and various axial (z) locations for Rs =7.12

Tangential R

MS

Velo

city [

Sp]

Crank Angle [CAD ATDC]

0

1

2

3

4

5

-30 -20 -10 0 10

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 10 mm

Fuel Injection Event

0

1

2

3

4

5

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 8 mm

Fuel Injection Event

0

1

2

3

4

5

MotoredFired, -22.25 SOIFired, -30.25 SOI

z = 6 mm

Fuel Injection Event

0

1

2

3

4

5

6

MotoredFired, -22.25 SOIFired, -30.25 SOI

Rs = 7.12

Pinj = 800 bar

r = 13.6 mm

z = 4 mm

Fuel Injection Event

Figure 21 Measured tangential velocity RMS fluctuations at r = 13.6 and various axial (z) locations for Rs =7.12

evidence that the mean flow evolves in a significantly different manner than the motored flow. For SOI = -30.25 CAD combustion begins at about -11 CAD, while for SOI = -22.25 CAD combustion begins near -8 CAD (see Fig. 11). Hence, at these locations, we see little effect of combustion on the mean flow evolution.

The corresponding RMS fluctuations are shown in Figure 21. Changes in RMS velocity fluctuations could be due to cycle-to-cycle fluctuations in the mean flow structures or to variability of the fuel injection event, rather than additional turbulence. However, previous cycle-resolved analysis has shown that enhanced RMS fluctuations observed after fuel injection correlate well with enhanced small-scale, high-frequency fluctuations characteristic of turbulent motions [27]. Consequently, we believe that the increased RMS fluctuations seen in Fig. 21 represent, at least in part, increased flow turbulence.

In general, increased RMS fluctuations are observed coincident with the period in which the largest perturbation in the mean flow structure is seen. At the highest axial location, z = 4 mm, these fluctuations are particularly large, and are both larger in magnitude and of a longer duration for the earlier injection timing. Later in the cycle, a more moderate increase in fluctuations over the motored fluctuations is seen, which typically persists for at least 10 CAD. Considering the -11 and -8 CAD start-of-combustion for the -30.25 and -22.25 CAD injection timings, respectively, we see that this more moderate increased turbulence, which appears earlier for the earlier injection timing, may also increase pre-combustion mixing. Overall, the measurements indicate that with earlier injection, turbulence levels are higher and act on the pre-combustion mixture for a longer period. Increased turbulent mixing may therefore assist in the early mixture formation process, complementing the longer ignition delay. However, it must be recognized that the spatial locations sampled with LDV are sparse, and increased turbulence will not increase mixing rates if it occurs in regions with a homogeneous mixture composition.

FUEL CONVERSION EFFICIENCY

In the previous sections we have examined how swirl ratio and combustion phasing (injection timing) influence CO emission, which is an indication of complete, efficient combustion. Complete combustion is a pre-requisite for obtaining a high fuel conversion efficiency

fc . However, other factors also influence fc . In this section our goal is to clarify the broader impact of variations in swirl ratio and injection timing on fc . In pursuing this goal, we consider only the gross indicated fuel conversion efficiency. The net fuel conversion efficiency may be influenced through other factors (e.g.increased pumping work associated with high-swirl port designs) that are not addressed here.

The gross indicated fuel conversion efficiency is the ratio of the gross indicated work W to the chemical energy in the injected fuel, calculated from the product of the fuel mass and the lower-heating value LHVQ of the fuel

LHVffc Qm

W (2)

As an aid to understanding the influence of various factors on fc (including CO emission and combustion efficiency), we have previously proposed [9] that fc be decomposed as a product of three separate efficiencies

chlwc

LHVf

chem

chem

hlchem

hlchemfc Qm

Q

Q

QQ

QQ

W (3)

-30 -25 -20 -15 -10

2

3

4

5

6

7

24

26

28

30

32

34

SOI [CAD ATDC]

Sw

irl R

atio

[%]

Fuel Conversion Efficiency

Figure 22 Effect of swirl ratio and injection timing on fuel conversion efficiency

82

84

86

88

90

92

94

-30 -25 -20 -15 -10

2

3

4

5

6

7

SOI [CAD ATDC]

Sw

irl R

atio

[%]

Combustion Efficiency

Figure 23 Effect of swirl ratio and injection timing on combustion efficiency

Because the difference between the chemical heat release chemQ and the energy losses due to heat transfer and crevice flows hlQ is the apparent net heat release, the first term on the right-hand-side (RHS) represents the fraction of the apparent heat release that is converted into useful work—the work conversion efficiency wc . Changes in this efficiency are primarily due to changes in the combustion phasing. The center term on the RHS of Eq.(3) is a straight-forward representation of the fraction of the chemical heat release that is manifested as apparent heat release—a heat-loss efficiency hl , and the last term on the RHS is the combustion efficiency c .

The fuel conversion efficiency, calculated from Eq.(2), is shown in Fig. 22. For comparison, an estimate of the combustion efficiency is calculated from the CO emissions (see Fig. 3) using the expression

R

COfCOfc H

hhf

0,

0, 21 (4)

In Eq.(4), f represents the fraction of fuel carbon found in the CO emissions, RH is the negative of the fuel lower heating value, and 0

fh is the heat of formation of the sub-scripted species. Additional details can be found in Ref. 9. Combustion efficiencies estimated from Eq.(4) are shown in Fig. 23.

Although many similarities exist between Figs. 22 and 23, there are also distinct differences that bear noting:

1) While the combustion efficiency remains high in the upper-left quadrant of Fig. 23, fc is decreasing at the higher swirl ratios.

2) The best fc is observed at somewhat lower swirl ratios and more retarded injection timing than the best c

3) In the lower-right quadrant of Fig. 22, a rapid decrease in fc is observed. No similar decrease can be seen in c .

4) In the lower-left quadrant, a local minimum in fc is seen that is not present for c

To assist in the explanation of these differences, the work conversion efficiency is presented in Fig. 24. All of the differences between fc and c identified above can be seen to correlate well with the variation of wc with SOI and with swirl ratio. First, the work conversion efficiency can be seen to decrease gradually in the upper-left quadrant of Fig. 24. Thus, decreased fc at early SOI and high swirl ratio can be, at least partially, explained by decreased wc . Second, peak work conversion efficiencies are observed at lower swirl ratios and at more retarded injection timing than peak c , an observation that helps explain the second difference between Figs. 22 and 23 enumerated above. Third, the behavior of fc in both the lower-left and the lower-right quadrants is mirrored in wc .

In addition to explaining the differences between fc and c , the trends in wc seen in Fig. 24 also serve to

illustrate the mechanisms by which changes in injection timing and swirl affect fc . For example, a general characteristic of Fig. 24 is a decrease in wc for both early and late injection timings. In the former case, early heat release results in negative work as the combustion-induced pressure rise occurs before compression has ended. Although some of this work is recovered during the subsequent expansion, due to heat transfer and

70

75

80

85

90

95

-30 -25 -20 -15 -10

2

3

4

5

6

7

SOI [CAD ATDC]

Sw

irl R

atio

[%]

Heat Loss Efficiency

Figure 25 Effect of swirl ratio and injection timing on heat-loss efficiency

32

34

36

38

40

-30 -25 -20 -15 -10

2

3

4

5

6

7

SOI [CAD ATDC]

Sw

irl R

atio

[%]

Work Conversion Efficiency

Figure 24 Effect of swirl ratio and injection timing on work conversion efficiency

blow-by some of this work is inevitably lost. In the latter case, heat release is sufficiently retarded that much of the volume expansion has already occurred.

A second observation is the low wc in the upper-central region of Fig. 24, which coincides with a minimum in the combustion efficiency. As discussed in the section describing the optimal swirl ratio, the cause of this minimum in c is thought to be trapping of CO in the bowl. Apparently, this trapping also retards the heat release thereby influencing wc .

Finally, wc also exhibits a general decreasing trend as the swirl level increases. As discussed elsewhere [28], the work conversion efficiency is influenced by heat transfer losses that occur in a non-combusting cycle. The higher heat transfer losses incurred in high swirl flows result in a decreased wc . Consequently, the influence of heat loss is not fully-confined to the heat-loss efficiency.

Although chemQ and hlQ have not been measured directly, the heat-loss efficiency hl can be estimated from the remaining efficiencies via Eq.(3). The results must be employed cautiously, however, as the combustion efficiencies calculated from Eq.(4) will likely be overestimated, due to the neglect of the energy content in the unburned hydrocarbons and other unmeasured species. Provided that these species follow the same trends as the CO emissions, their neglect implies that the changes observed in c are underestimated.

Fig. 25 illustrates the resulting estimates of hl . Note that the overall variation in hl is larger than is seen in either wc or c . However, because the variation in c is underestimated, the variation seen in hl is likely excessive. Nevertheless, the magnitude of the true variation is likely to be significant, and the trends observed are physically plausible and bear consideration. First, a general trend of decreasing heat-loss efficiency is seen as swirl is increased—a result that is not unexpected. More interestingly, the heat loss efficiency shows a greater variation with injection timing. This variation might be expected based on the simple observation that there is more time available for heat transfer at the earlier injection timing. At least three other factors also contribute to this behavior, though. First, with earlier combustion phasing the average surface area-to-volume ratio of the hot, in-cylinder gases is increased. Second, mixture that burns first and then is compressed reaches a higher temperature than mixture that burns after compression. Third, convective heat transfer coefficients are expected to be larger at the higher peak cylinder pressures observed with early injection. Each of these factors is expected to increase heat transfer losses as injection is advanced.

SUMMARY AND CONCLUSION

CO emission and fuel conversion efficiency are examined in a highly-dilute, low-temperature diesel combustion regime as the injection timing and swirl ratio are varied. To clarify the dominant factors influencing the emission and efficiency behavior, the cylinder pressure, engine-out emissions, and flow velocities are measured. High speed video imaging and multi-dimensional numer-ical simulations of the combustion process are also performed.

The CO emission behavior exhibits two interesting trends:

First, an optimal swirl ratio exists at which the lowest CO emission and best fuel conversion efficiency are observed. Heat release analysis, the results of the numerical simulations, and imaging of combustion luminosity indicate that this behavior is likely due to mixing processes occurring after the premixed burn period. At the lowest swirl ratios, a dual-vortex vertical plane structure forms that enhances the mixing of partially-burned fuel and air. This structure is most effective at a moderate swirl ratio ( 2.5). At higher swirl ratios, the flow structures formed impede mixing, and can lead to trapping of partially-burned fuel within the bowl.

Second, CO emission generally exhibits a rapid decrease from the maximum as SOI is advanced, particularly at the highest swirl ratios. The numerical simulations indicate that, at a fixed swirl ratio, earlier injection timing leads to enhanced pre-combustion mixing and hence lower peak in-cylinder CO mass. The enhanced mixing is due not only to increased ignition delay, but also to increased mixing rates under high-swirl conditions. A reduction in CO emission with increased pre-mixing CO implies that CO emission stems predominantly from under-mixed fuel (rich mixtures), a finding which is supported experimentally by a strong tendency toward reduced CO emission with increased ignition delay and with increased injection pressure. Correlation of CO emission with peak adiabatic flame temperature suggests that temperature related changes to the early CO oxidation rates are not a significant factor influencing peak in-cylinder CO mass and subsequent engine-out emission. Spray penetration and spray targeting, with concomitant liquid film formation was also examined as a potential source of modified premixing. Significant liquid film formation along the bowl lip was observed at early injection timings. However, low CO emission was observed under these conditions, signifying that wall-film formation in this region is not a dominant factor influencing CO emission. At the earliest injection timings, the likely existence of wall films on the top of the piston is identified as a potential source of increasing CO. Finally, the possibility that enhanced pre-combustion turbulence was promoting mixing under early-injection conditions was examined. Increased turbulence was observed with

earlier injection timing, and likely assists in the initial mixing taking place in the ignition delay period.

The impact of swirl level and injection timing on the fuel conversion efficiency is also assessed. Qualitative differences are observed between the fuel conversion efficiency and the combustion efficiency derived from the CO emissions. These differences can be accounted for by changes in the work conversion efficiency (associated primarily with combustion phasing) and by heat transfer losses.

ACKNOWLEDGMENTS

Support for this research was provided by the U.S. Department of Energy, Office of FreedomCAR and Vehicle Technologies. The research was performed at the Combustion Research Facility, Sandia National Laboratories, Livermore, California. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000. The BK21 and Future Vehicle Technology Development Corps. of Korea supported Sanghoon Kook's visiting research. The authors express their appreciation to Mark Musculus and Lyle Pickett for providing the high speed camera and the Matlab source code to calculate the adiabatic flame temperature.

REFERENCES

1. Cowland, C., Gutmann, P. and Herzog, P.L., “Passenger Vehicle Diesel Engines for the US,” SAE Paper No. 2004-01-1452, 2004.

2. Kimura, S., Aoki, O., Kitahara, Y. and Aiyoshizawa E., “Ultra-Clean Combustion Technology Combining a Low-Temperature and Premixed Combustion Concept for Meeting Future Emission Standards,” SAE Paper No. 2001-01-0200, 2001.

3. Akihama, K., Takatori, Y., Inagaki, K., Sasaki, S., and Dean, A.M., “Mechanism of the Smokeless Rich Diesel Combustion by Reducing Temperature,” SAE Paper No. 2001-01-0655, 2001.

4. Walter, B. and Gatellier, B. “Development of the High Power NADI Concept Using Dual Mode Diesel Combustion to Achieve Zero NOx and Particulate Emissions,” SAE Paper No. 2002-01-1744, 2002.

5. Weißbäck, M., Csató, J., Glensvig, M., Sams, T. and Herzog, P., “Alternative Brennverfahren – ein Ansatz für den Zukünftigen Pkw-Dieselmotor,” MTZ: vol 64: pp 718-727, 2003.

6. Hasegawa R. and Yanagihara, H., “HCCI Combustion in DI Diesel Engine,” SAE Paper No. 2003-01-0745, 2003.

7. Sjöberg, M., and Dec, J. E., “An Investigation into the Lowest Acceptable Combustion Temperatures for Hydrocarbon Fuels in HCCI Engines”, Proceedings of the Combustion Institute, Vol. 30, pp.2719-2726, 2005.

8. Adomeit, P., Pischinger, S., Becker, M., Rohs, H., and Greis, A., “Laser Optical Diagnostics and Numerical Analysis of HSDI Combustion,” THIESEL 2004: Conference on Thermo- and Fluid Dynamic Processes in Diesel Engines, Sept. 8-10, Valencia, Spain, 2004.

9. Kook, S., Bae, C., Miles, P. C., Choi, D, and Pickett, L. M., “The Influence of Charge Dilution and Injection Timing on Low-Temperature Diesel Combustion and Emissions”, SAE Paper 2005-01-3837, 2005.

10. Timoney, D.L., “Smoke and Fuel Consumption Measurements in a Direct Injection Diesel Engine with Variable Swirl,” SAE Paper No. 851542, 1985.

11. Van Gerpen, J.H., Hwang, C.-W., and Borman, G.L., “The Effects of Swirl and Injection Parameters on Diesel Combustion and Heat Transfer,” SAE Paper No. 850265, 1985.

12. Miles, P. C., “The Influence of Swirl on HSDI Diesel Combustion at Moderate Speed and Load”, SAE Paper 2000-01-1829, 2000.

13. Miles, P.C., Choi, D., Pickett, L.M., Singh, I.P., Henein, N., RempelEwert, B.H., Yun, H., and Reitz, R. D., “Rate-Limiting Processes in Late-Injection, Low-Temperature Diesel Combustion Regimes,” THIESEL 2004: Conference on Thermo- and Fluid Dynamic Processes in Diesel Engines, Sept. 8-10, Valencia, Spain, 2004.

14. Amsden, A. A., “KIVA-3V: A Block Structured KIVA Program for Engines with Vertical or Canted Valves”, Los Alamos National Laboratory Report No. LA-13313-MS, 1997.

15. RempelEwert, Bret H., “The Influence of Swirl on Flow, Fuel Injection, and Emissions in an HSDI Diesel Engine,” Master’s Thesis, Dept. of Mechanical Engineering, University of Wisconsin – Madison, 2004.

16. Miles, P., Choi, D., Megerle, M., RempelEwert, B., Reitz, R. D., Lai, M. D., Sick, V., “The Influence of Swirl Ratio on Turbulent Flow Structure in a Motored HSDI Diesel Engine – A Combined Experimental and Numerical Study”, SAE Paper 2004-01-1678, 2004.

17. Kong, S.-C., Han, Z., and Reitz, R.D., ”The Development and Application of a Diesel Ignition and Combustion Model for Multidimensional Engine Simulation,” SAE Paper 950278, 1995.

18. Ikegami, M., “Role of Flows and Turbulent Mixing in Combustion and Pollutant Formation in Diesel Engines,” COMODIA 90: Proceedings of the International Symposium on Diagnostics and Modeling of Combustion in Internal Combustion Engines, 3–5 September, Kyoto, Japan, 1990.

19. Khan, I.M., Wang, C.H.T., and Langridge, B.E., “Effect of Air Swirl on Smoke and Gaseous Emissions from Direct-Injection Diesel Engines,” SAE Paper No. 720102, 1972.

20. Binder, K. and Hilburger, W., “Influence of the Relative Motions of Air and Fuel Vapor on the

Mixture Formation Processes of the Direct Injection Diesel Engine,” SAE Paper No. 810831, 1981.

21. Miles, P., Megerle, M., Sick, V., Richards, K., Nagel, Z., and Reitz. R. D., “The Evolution of Flow Struc-tures and Turbulence in a Fired HSDI Diesel Engine”, SAE Paper 2001-01-3501, 2001.

22. Miles, P.C., RempelEwert, B.H., and Reitz, R.D., “Squish-Swirl and Injection-Swirl Interaction in Direct Injection Diesel Engines,” ICE 2003: 6th International Conference on Engines for Automobiles, Capri, Naples, Italy, Sept. 14-19, 2003

23. Miles, P.C., “In-Cylinder Turbulent Flow Structure in Direct-Injection, Swirl-Supported Diesel Engines,” Ch. 1 in Flow and Comb. in Automotive Engines, C. Arcoumanis, ed. Springer-Verlag, in press.

24. Heywood, J. B., “Internal Combustion Engine Fundamentals”, International Edition, McGraw-Hill, Inc., 1988.

25. Akagawa, H., Miyamoto, T., Harada, A., Sasaki, S., Shimazaki, N., Hashizume, T., and Tsujimura, K., “Approaches to Solve Problems of the Premixed Lean Diesel Combustion,” SAE Paper 1999-01-0183, 1999.

26. Henein, N. “Analysis of Polllutant Formation and Control and Fuel Economy in Diesel Engines,’ Prog. Energy Combust. Sci., Vol. 1, pp.165-207, 1978.

27. Miles, P., Megerle, M., Hammer, J., Nagel, Z., Reitz, R. D., Sick, V., “Late-Cycle Turbulence Generation in Swirl-Supported, Direct-Injection Diesel Engines”, SAE Paper 2002-01-0891, 2002.

28. Choi, D. and Miles, P.C., “A Parametric Study of Low-Temperature, Late-Injection Combustion in an HSDI Diesel Engine,” COMODIA 2004: 6th Intl. Symp. On Diagnostics and Modeling of Combustion in IC Engines, Aug 2-5, Yokohama, 2004.

NOMENCLATURE

ATDC After Top Dead Center

CAD Crank Angle Degrees

EGR Exhaust Gas Recirculation

EVC Exhaust Valve Closing

EVO Exhaust Valve Opening

IMEP Indicated Mean Effective Pressure

IVC Intake Valve Closure

IVO Intake Valve Opening

SOI Start of Injection

Sp Mean Piston Speed

APPENDIX: COMPARISON OF MEASURED AND

SIMULATED TANGENTIAL VELOCITIES

Figures A1 ~ A5 compare the evolution of the measured mean tangential velocities with the numerical predictions for Rs = 1.44 ~ 7.12. The velocity data are normalized with the piston speed Sp: 4.25 m/s at 1500 rpm. At each crank angle in the figures, axial profiles of the tangential mean velocities are depicted on a schematic showing the relative position of the piston. The velocities were measured at a fixed radial location of r = 13.6 mm, which is represented by the vertical, dashed lines. The crank angle range investigated commences well before injection occurs at -22 CAD ATDC and continues to the end of the premixed combustion at around 10 CAD ATDC.

The overall development of the measured and predicted mean velocity fields is very similar, though several discrepancies in the detailed shape of the profiles exist. For example, the measured velocities are lower than the computation at many locations lower in the bowl. These data are currently being re-assessed for noise contributions arising from scattering from the bowl floor.

A second discrepancy is under-predicted velocities at -15 CAD ATDC for Rs = 3.77 ~ 7.12. At this crank angle, the calculations show a marked influence of the fuel injection event (SOI=-22.25 CAD ATDC) near z = -6 mm, resulting in a minimum in the tangential velocities. This

Figure A1 Comparison of the axial profile of the measured mean tangential velocities with the results of the numerical simulation Rs = 1 44

Tangential Velocity [Sp]

z [m

m]

Experiment

Computation

Tangential Velocity [Sp]

z [m

m]

Rs = 1.44, SOI = -22.25 CAD ATDC

-20

-15

-10

-5

0

-25 CAD

-20

-15

-10

-5

0

-20 CAD

-20

-10

0

-10 CAD

-15

-5

-20

-15

-10

-5

0

-15 CAD

-20

-15

-10

-5

0

-5 CAD

-20

-15

-10

-5

0

TDC

-20

-15

-10

-5

0

5 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

10 CAD

0 2 4 6 8 10 12

pe

m

Measurement Locatione Measurement Locatione

Tangential Velocity [Sp]

z [m

m]

Experiment

Computation

Tangential Velocity [Sp]

z [m

m]

Rs = 2.59, SOI = -22.25 CAD ATDC

pe

m

Measurement Locatione Measurement Locatione

-20

-15

-10

-5

0

-25 CAD

-20

-15

-10

-5

0

-20 CAD

-20

-15

-10

-5

0

-15 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

-10 CAD

-20

-15

-10

-5

0

-5 CAD

-20

-15

-10

-5

0

TDC

-20

-15

-10

-5

0

5 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

10 CAD

Figure A2 Comparison of the axial profile of the measured mean tangential velocities with the results of the numerical simulation. Rs = 2.59.

Tangential Velocity [Sp]

z [m

m]

Experiment

Computation

Tangential Velocity [Sp]

z [m

m]

Rs = 3.77, SOI = -22.25 CAD ATDC

pe

m

Measurement Locatione Measurement Locatione

-20

-15

-10

-5

0

-25 CAD

-20

-15

-10

-5

0

-20 CAD

-20

-15

-10

-5

0

-15 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

-10 CAD

-20

-15

-10

-5

0

-5 CAD

-20

-15

-10

-5

0

TDC

-20

-15

-10

-5

0

5 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

10 CAD

Figure A3 Comparison of the axial profile of the measured mean tangential velocities with the results of the numerical simulation. Rs = 3.77.

mm, resulting in a minimum in the tangential velocities. This behavior is not seen in the measurements. Finally, the overall trends in the axial velocity profile prior to injection differs for Rs = 7.12—the measured tangential velocity has a maximum near z = -12 mm, while the simulation predicts a maximum near the cylinder head. This behavior was repeatable. Additional experiment and analysis are needed to clarify this discrepancy.

Despite the discrepancies mentioned above, the important point to note is that, overall, the measured velocities and their evolution correlate well with the numerical simulations.

CONTACT

Sanghoon Kook, kooks@kaist.edu

Tangential Velocity [Sp]

z [

mm

]

Experiment

Computation

Tangential Velocity [Sp]

z [

mm

]

Rs = 4.94, SOI = -22.25 CAD ATDC

pe

m

Measurement Locatione Measurement Locatione

-20

-15

-10

-5

0

-25 CAD

-20

-15

-10

-5

0

-20 CAD

-20

-15

-10

-5

0

-15 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

-10 CAD

-20

-15

-10

-5

0

-5 CAD

-20

-15

-10

-5

0

TDC

-20

-15

-10

-5

0

5 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

10 CAD

Figure A4 Comparison of the axial profile of the measured mean tangential velocities with the results of the numerical simulation. Rs = 4.94

Tangential Velocity [Sp]

z [

mm

]

Experiment

Computation

Tangential Velocity [Sp]

z [

mm

]

Rs = 7.12, SOI = -22.25 CAD ATDC

pe

m

Measurement Locatione Measurement Locatione

-20

-15

-10

-5

0

-25 CAD

-20

-15

-10

-5

0

-20 CAD

-20

-15

-10

-5

0

-15 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

-10 CAD

-20

-15

-10

-5

0

TDC

-20

-15

-10

-5

0

-5 CAD

-20

-15

-10

-5

0

5 CAD

-20

-15

-10

-5

0

0 2 4 6 8 10 12

10 CAD

Figure A5 Comparison of the axial profile of the measured mean tangential velocities with the results of the numerical simulation. Rs = 7.12.