GE Power Generation

Advances in Steam PathTechnology

John I. Cofer, IVJohn K. Reinker

William J. SumnerGE Power Systems

Schenectady, NY

GER-3713E

29

GE Power Generation

Advances in Steam PathTechnology

John I. Cofer, IVJohn K. ReinkerWilliam J. SumnerGE Power SystemsSchenectady, NY

GER-3713E

John K. ReinkerJohn joined GE’s Power Generation Business in 1984 as a Steam

Turbine Engineer. He has worked in several design offices culminatingin his assignment of Manager New Unit Design for Small and LargeSteam Turbines. He was named to his current position of Product LineLeader - Large Steam Turbine in June 1996.

John I. Cofer, IVJack began his career at GE in 1974 as a marine turbine requisition

and development engineer in the Medium Steam Turbine Departmentin Lynn, MA. In 1979 he received the GE Power Systems SectorEngineering Award for Young Engineer in the Industrial and MarineSteam Turbine Division.

For the past 18 years, he has provided technical leadership in thedevelopment and use of advanced numerical and analytical methods forthe design of turbomachinery, with special emphasis on aerodynamics,fluid mechanics and numerical optimization.

Jack has held several management positions in which he was responsi-ble for the development of new, advanced aerodynamic designs forsteam turbines for Navy surface ship and submarine propulsion, forcommercial power generation, and for mechanical-drive applications.

He is Manager of Aerodynamics Engineering. He is responsible forleading a team of engineers in the development of new high-efficiencyturbine and exhaust hood design concepts. .

He holds a BSAE from the University of Virginia, an MS in aeronau-tics and astronautics from the Massachusetts Institute of Technology,and an ME degree from Northeastern University.

A list of figures and tables appears at the end of this paper

William J. SumnerBill is a Senior Engineer in GE’s Steam Turbine Design and

Development Engineering Department. He is currently working in theSteam Turbine Advance Design Department.

Since joining GE in 1971, Bill has been responsible for both experi-mental and analytical emission, moisture erosion, flow measurement,and solid particle erosion.

Bill has co-authored numerous technical papers and twice receivedthe ASME-EEI Prime Mover Award. He has three patents.

ABSTRACTFor many years, GE has been conducting

research to better understand the loss mecha-nisms that degrade the aerodynamic perfor-mance of steam turbine stages, and to developnew computational fluid dynamics (CFD) com-puter programs to predict these losses accurate-ly. This paper describes a number of new steampath design features introduced to the GE steamturbine product line to improve turbine perfor-mance and reliability. These features includediaphragms with contoured sidewalls; AdvancedVortex blading with compound tangential lean,new continuously-coupled last-stage buckets withimproved aerodynamic efficiency and reliability,improved downward and axial flow exhausthoods; and better steam leakage control devices.The benefits of these new features for new unitsand retrofits of existing units are discussed.

In addition, this paper discusses the new gen-eration of three dimensional viscous CFD analy-sis codes being used to develop new design con-cepts, including codes developed by GE as wellas those obtained externally. Also described arethe extensive laboratory test programs beingconducted to validate the CFD codes and verifythe predicted efficiency gains for new designfeatures.

Lastly, new and unique state-of-the-art steampath design automation and optimization toolsthat have been developed by GE to dramaticallyreduce the design cycle time for new advancedaerodynamic designs are discussed.

INTRODUCTION

Turbine Efficiency LossesThe thermodynamic performance of a steam

power plant depends on many factors, includingcycle arrangement and inlet and exhaust steamconditions. However, the dominant contributorto the overall power plant efficiency is the steamturbine itself. The thermodynamic and aerody-namic performance of the steam turbine is pri-marily determined by steam path components,including the valves, inlet, nozzles, buckets,

steam leakage control devices and the exhaust.To maximize power plant efficiency, aerodynam-ic and steam leakage losses in the turbine steampath must be minimized in both the rotatingand stationary components. Figure 1 shows thetypes of efficiency losses that occur in a typicalturbine stage and the approximate percentagethat each type contributes to the total stage loss.Nozzle and bucket aerodynamic profile losses,secondar y flow losses, and leakage lossesaccount for roughly 80% to 90% of the totalstage losses.

Nozzle and bucket profile losses can be signif-icant if the blade shapes are not optimized forthe local operating conditions. Profile losses aredriven by surface finish, total blade surface area,airfoil shape and surface velocity distributions,and proper matching between nozzle and buck-ets to minimize incidence losses.

Equally significant losses can be caused by thecomplex secondary flows (Figure 2, adaptedfrom Reference 1) generated as the viscousboundary layers along the inner and outer side-walls of the steam path are turned through theblade rows. These complex, three dimensionflows must be thoroughly understood beforeeffective methods can be developed to reducethe associated losses.

Steam leakage through the seals between sta-tionary and rotating components of the turbine

1

GER-3713E

ADVANCES IN STEAM PATHTECHNOLOGY

J.I. Cofer, IV, J.K. Reinker and W.J. SumnerGE Power SystemsSchenectady, NY

GT19893A

Figure 1. Typical HP turbine stage efficiencylosses

constitutes an efficiency loss because the leakageflow does not contribute to the work output ofthe stage. This loss can be quite significant, par-ticularly at bucket tips. In short high pressure(HP) stages, the tip leakage loss is driven by thehigh-stage pressure levels and the relatively larg-er amount of radial clearance area compared tothe nozzle flow area. In taller intermediate (IP)and low pressure (LP) stages, the tip leakageloss is driven by the higher reaction levels at thebucket tips, which increases the pressure dropacross the bucket tip. Steam leakage throughthe diaphragm shaft packing and shaft endpacking also causes losses, but these are general-ly of lesser magnitude than tip leakage losses.

Equal attention must be paid to the losses instationary flow path components, such as inlets,valves, and exhausts. Any pressure drop occur-ring in these components constitutes a loss inavailable energy and a reduction in work output.

Key Elements of GE AerodynamicDevelopment Programs

Development efforts have been underway atGE for many years to better understand andreduce the various losses listed above. Theobjective of this long term program is to developspecific design features, for both new turbinesand retrofits of existing units, that maximizeoverall turbine efficiency while maintaining ahigh degree of reliability and cost effectiveness.Most aspects of this multifaceted developmentprogram are being conducted in cooperationwith other company components, such asAircraft Engines, Gas Turbine and CorporateResearch and Development (CRD).

The four key elements of GE’s overall aerody-

namic development program are:• Development of better computational fluid

dynamics computer programs, which allowmore accurate predictions of the complexbehavior of the steam flow in the turbine

• Development of new design concepts toimprove both baseline and sustained effi-ciency, making full use of the new CFDcodes

• An extensive laboratory test program tovalidate the CFD codes and verify the pre-dicted efficiency gains

• Development of a suite of powerful designautomation and optimization tools to allowimplementation of advanced aerodynamicdesign features on a custom basis to maxi-mize efficiency for each specific applica-tion in short design cycle times

This paper reviews recent advances in GEsteam path technology resulting from thesedevelopment initiatives and the impact of theseadvances on current GE steam turbine designs.

CFD METHODSIn order to fully utilize the available energy of

the steam for maximum efficiency, the complexthree dimensional flow field throughout the tur-bine must be accurately predicted. In particular,a better understanding is needed of the effectsof wet steam, viscosity, and unsteady rotor-statorinteractions if the remaining losses are to bereduced. Recent advances in the developmentof CFD computer programs have provided GEwith the tools needed to achieve better designswith reduced cost and shorter design cycle time.

Viscous Euler CFD CodeIn the early 1980s, the GE steam turbine, gas

turbine and aircraft engine businesses began along-term joint research program with CRD tostudy secondary flow phenomena using a com-bined analytical and experimental approach.Analytical investigations included the develop-ment of complex computational fluid dynamicscomputer codes based on three dimensional for-mulations of the inviscid Euler equations andthe viscous Navier-Stokes equations. The initialinviscid code was developed by CRD and calledthe EULER3D program (Reference 2). A viscouscode that included turbulence models was laterdeveloped by GE Aircraft Engines and called theViscous Euler program (References 3 and 4).

Comprehensive experimental investigationswere conducted to test the validity of the predic-tions made by the EULER3D code. These tests

2

GER-3713E

GT19894A

Figure 2. Secondary flows in a turbine nozzlecascade

utilized CRD’s large-scale wind tunnel test facili-ty (Figure 3), and the results are described indetail in Reference 5. In this study, EULER3Dpredictions of the flow in a subsonic turbinenozzle cascade were compared to cascade testdata, including lampblack and oil flow visualiza-tion studies. Comparisons were made for nozzleswith both parallel and diverging sidewalls, andthe code was able to predict inviscid flows in tur-bine blade passages quite well.

A unique flow visualization method was devel-oped for the wind tunnel facility to aid in under-standing the physics of secondary flows. Theflow particle trajectories shown in Figure 4 wereobtained by shining a strobe light on helium-filled zero-buoyancy soap bubbles injected intothe flow. Figure 4 clearly shows one leg of the

“horseshoe” vortex spiraling around the nose ofa nozzle cascade with parallel sidewalls.

In the last few years, the Viscous Euler codehas replaced EULER3D as the analysis code ofchoice for a wide variety of turbomachinerydesign problems. Improvements in computertechnology have made it practical to run viscouscodes such as Viscous Euler on a network ofhigh-powered workstations, without the need forsupercomputers.

Figure 5 shows the Viscous Euler calculationgrid on a blade-to-blade plane for a high pres-sure bucket vane. A similar calculation grid isgenerated for the nozzle passage, and a com-plete stage solution is obtained by automaticallyiterating between the nozzle and bucket solu-tions as they run in parallel on separate worksta-tions. For a complete 3D solution, up to 250,000grid points might be needed in each blade pas-sage to accurately resolve secondary flow details.Figure 6 shows Mach number contours for thissame vane in subsonic flow, and the trailing

3

GER-3713E

GT19895A

Figure 3. Nozzle cascade wind tunnel

GT19897

Figure 4. Helium bubble traces in nozzlecascade

GT25784

Figure 5. HP bucket - Viscous Euler calculation grid

GT25785

Figure 6. HP bucket – Viscous Euler Machnumber controls

GT25993

Figure 7. IP nozzle Viscous Euler results andcomparison to test data

GT24583

Figure 8. Comparison of structured andunstructured grids in CFD codes.

currently under development for incorporationinto NOVAK3D. This model will includedetailed 3D wet steam effects, such as homoge-neous moisture droplet formation and growth,droplet trajectories and condensation shocks.The code will then be used to develop improvedmoisture removal devices for wet steam turbinestages. Better moisture removal devices will pro-vide two significant benefits: improved efficien-cy, since stage efficiency improves about 1% forevery additional 1% of moisture that is removed;and reduced bucket erosion. An extensiveexperimental program is planned to validate thewet steam CFD model to be incorporated in

NOVAK3D.

Multi stage and Unsteady CFD CodesA better understanding of unsteady flow

effects and rotor-stator interactions is needed ifthe losses associated with them are to bereduced. The influence of these effects on per-formance has been implicitly accounted for inGE’s traditional turbine performance predictionmethods through the use of empirical correla-tions based on laboratory and field test data.However, it has been difficult to develop specificdesign features that reduce these losses due tothe high expense and difficulty of the requiredtesting. For this reason, renewed effort has beenplaced on developing the analytical capabilitiesto predict unsteady flows and rotor-stator inter-actions.

In early 1994, GE Power Generation and GEAircraft Engines used the unsteady flow codedeveloped by Rai and Madavan (Reference 9) toanalyze a steam turbine stage with a compoundlean nozzle. Because the computing timerequired to achieve a solution with unsteadycodes is still excessive, more recent efforts havefocused on the calculation of steady state multi-stage solutions using the passage-averageapproach of Adamczyk (Reference 10).

Losses due to shock waves, wakes, hub and tipleakage flows, and secondary flows propagatedownstream and influence the performance ofsuccessive stages. Figure 10 shows some initialresults from the Adamczyk code obtained forthe third stage in a group of Advanced VortexHP stages. The figure on the left shows vorticitycontours at the nozzle exit plane, while the fig-

5

GER-3713E

GT22182

Figure 9. NOVAK3D refined calculation gridfor transonic nozzle

GT25787

Figure 10. Multi-stage CFD analysis — vorticity at stage 3 nozzle and bucket exit

NOZZLE BUCKET

ure on the right shows vorticity contours at thebucket exit plane. Comparisons of the coderesults with laboratory test data showed goodagreement.

NEW DESIGN FEATURES FOR IMPROVED BASELINE

EFFICIENCY

Advanced Vortex BladingThe flow inside a turbine stage must satisfy

the requirements of radial equilibrium. Theradial equilibrium equations are derived fromthe equations of motion and basic thermody-namic relations, and express the balancebetween pressure forces and inertial forces act-ing on the fluid. Before the advent of powerfulcomputers and modern CFD codes, the com-plete, 3D form of these equations could not beused for practical design purposes. Designerswere forced to make simplifying assumptionsabout the flow to reduce the radial equilibriumequations to a manageable form.

Dating back to the 1950s, GE steam turbinestages traditionally have been designed using atrue free vortex radial flow distribution. In thistype of design, the radial components of velocityare neglected and the assumption is made thatthe product of radius and tangential velocity isconstant. These assumptions greatly simplify theradial equilibrium equations and result in a flowcharacterized by constant axial velocity and con-stant work at all radii in the blade row. For short-er stages, free vortex designs with tapered andtwisted nozzles and buckets can be very effi-cient, and considerably more efficient thanpurely cylindrical stages. For longer stages, how-ever, the free vortex design often results inextremely twisted buckets that are difficult todesign mechanically, and it produces high tipreactions that lead to high tip leakage losses. GEhas for many years designed non-free-vortexstages in LP turbines to achieve better longbucket designs.

The CFD codes available today allow accuratesolutions of the fully 3D viscous Navier-Stokesflow equations, and designers are no longerbound to the free vortex concept. In 1984, GEinitiated a development program to investigatevarious “controlled vortex” design concepts, inwhich the radial flow distribution is tailored tomaximize the efficiency of individual stages.This program was accelerated in 1990 andexpanded to include the design and testing of a

number of “advanced” vortex stage designs thatbegan with the controlled vortex concept andincorporated variable tangential lean into thenozzles. Many of these concepts have beenderived from advanced GE aircraft enginedesign technology, and were developed jointlywith GE Aircraft Engines using the ViscousEuler CFD code. An extensive ongoing test pro-gram has been underway since 1990 to verify thepredicted efficiency gains for these new con-cepts, utilizing GE’s single-stage subsonic air tur-bine and the multi-stage low speed research tur-bine (LSRT) at GE Aircraft Engines (Figures 58and 59, described later in this paper). All ofthese concepts share the following characteris-tics:

• The radial flow distribution is tailored tomaximize efficiency based on the individu-al stage geometry and operating conditions

• The total nozzle surface area has beenreduced relative to a conventional designby using fewer nozzles per 3608 with betteraerodynamic profile shapes that retain thesame total mechanical strength as a con-ventional design. In many cases, the num-ber of nozzles in the diaphragm has beenreduced by as much as 50%. This providesa net reduction in profile loss with no lossin mechanical integrity

• Variable tangential or compound lean isused in the nozzles when it produces a sig-nificant gain in overall stage efficiency

• Root reaction is increased in var yingdegrees to improve bucket root perfor-mance, and tip reaction is generallydecreased relative to a free vortex design toreduce tip leakage

Due to the quite different operating condi-tions in HP, IP and LP stages, different conceptsare used in each type of stage. In the HP section,stage pressure ratios are about 1.2. This pressureratio is not limited by aerodynamic considera-tions, but rather by the pressure loading acrossthe diaphragm at the high pressure levels occur-ring in the HP section. Consequently, nozzlecross-sections must be relatively larger in sizethan bucket cross-sections to carry the mechani-cal load. Figure 11 shows a typical AdvancedVortex HP stage. Figure 12 compares the radialdistribution of reaction and flow angle for thisdesign to a conventional free vortex design. Inthe Advanced Vortex design, overall stage reac-tion has been increased somewhat, althoughmore at the root than at the tip, and root reac-tion is limited to about 10% to minimize thruston the rotor. The flow angle relative to the axial

6

GER-3713E

direction has been decreased at the root andincreased at the tip to shift flow radially inwardtoward the root. The flow angle shift at the rootallows a more efficient, less cambered bucketroot section to be designed, as shown in Figure5. The nozzle and bucket for each stage arecarefully designed as a matched set to eliminatebucket incidence angle losses. Due to the rela-tively low velocities in the stage (see Figure 6),the small size and strength of the secondary flowvortices, and the short blade height, nozzle leanhas not proven to be an effective loss reductionmethod in this type of stage.

In the IP section, stage pressure ratios aregenerally in the range of 1.2 to 1.3. Velocitiesare somewhat higher than in HP stages, andblades are longer and more twisted. Figure 13shows a typical Advanced Vortex IP stage. Figure14 compares the radial distribution of reactionand flow angle for this design to a conventionalfree vortex design. In these stages, the flow isbiased away from the sidewalls toward the moreefficient mid-section of the blade by increasingthe throat at the pitchline, producing a nozzle

7

GER-3713E

GT25788

Figure 11. Advanced Vortex HP nozzle andbucket

GT25789

Figure 12. HP stage nozzle exit angle and reaction distributions

GT25790

Figure 13. Advanced Vortex IP nozzle and bucket

GT25791

Figure 14. IP stage nozzle exit angle andreaction distribution

that is slightly overcambered at the root and tiprelative to the pitchline shape. Compound leanhas been introduced in the nozzle to reduce thesecondary flow losses, and root reaction hasbeen increased to allow a more efficient, lesscambered bucket root section to be designed. Asin the HP stage, the nozzle and bucket for eachIP stage are carefully designed as a matched setto eliminate bucket incidence angle losses. Theradial flow distribution in the nozzle is deter-mined using numerical optimization techniquesdescribed later in this paper. Stage test resultsobtained in the subsonic air turbine havedemonstrated that Advanced Vortex stages ofthis type are the most efficient GE has ever built.

GE’s analytical studies have shown that redis-tributing the flow radially and leaning the noz-zle have only a weak effect on overall nozzle effi-ciency - the losses are simply redistributed.However, if the vortexing and lean are done inthe proper combination, they can reduce thesize and strength of the secondary flow vorticesentering the bucket, particularly at the root,which leads to substantial improvements inbucket efficiency. The nozzle and bucket mustbe designed as a matched set to take full advan-tage of the benefits of blade lean.

In the LP section, stage pressure ratios arehigh enough to produce transonic velocities inthe last few nozzles and buckets. Buckets arelong and highly twisted, and diaphragm outersidewall slant angles can approach 508 . TheAdvanced Vortex design concept used in theearly LP stages is similar to that used in IPstages, as depicted in Figure 13.

In taller LP stages, some modifications aremade to the IP design concept to produce thenozzle shown in Figure 15. First, the point ofmaximum bow in the blade is moved closer tothe root to put more lean at the root, where ithas the most benefit, and less at the tip. Thisalso makes the blade easier to assemble indiaphragms with high outer sidewall slants.Second, a small amount of axial sweep is addedto the nozzle leading edge to minimize the noz-zle axial width over most of the length of theblade, thus reducing profile losses. The last-stage nozzle, and in some cases the next to thelast nozzle, typically has straight tangential lean,with no bowing. In transonic nozzles, the tran-sonic portion near the root is designed with aconverging-diverging cross-section to minimizeshock losses.

Each individual Advanced Vortex stage in acustomer’s specific application is custom-designed to optimize its efficiency in conjunc-

tion with the other stages in the turbine section.Off-the-shelf blades are seldom used, becausethis necessarily involves compromising the effi-ciency of the design to fit the existing blading.The design automation and optimization toolsthat allow GE to do this in a very short designcycle time are described later in this paper.Advanced Vortex nozzles and buckets aredesigned to meet the same strict requirementsfor mechanical integrity and reliability that havegiven GE steam turbines a record of reliabilityunparalleled in the industry. Concurrent withthe development of new aerodynamic designconcepts, modern manufacturing technologiessuch as five-axis NC machining, precision forg-ing and precision casting are being utilized toallow advanced vortex blade shapes to be madewith reduced cost and in shorter manufacturingcycle times.

Overall stage efficiency gains of 2% to 2.5%have been achieved on Advanced Vortex stagesrelative to GE’s conventional free vortex designs,which are already very efficient. This includesthe combined effects of reduced nozzle count,improved profile designs, optimized radial flowdistribution, nozzle lean and improved buckettip seals. Advanced Vortex stage designs areoffered as a standard feature on all new utility-type fossil reheat steam turbines rated 150 MWand above. They can also be applied to fossilreheat turbines rated below 150 MW and tocombined-cycle reheat turbines, and they areavailable for retrofit in HP, IP and LP stages.

8

GER-3713E

GT25792

Figure 15. Advanced Vortex LP nozzle

Contoured SidewallsIn the early 1980s, experimental investiga-

tions were conducted by CRD and the steam tur-bine business to determine the effects of outersidewall contours on secondary flow losses. Priorresearch documented in the literature suggestedthat outer sidewall contours other than straight-parallel could have a beneficial effect on stageefficiency (see Reference 11). A properlydesigned sidewall contour reduces secondaryflow losses by reducing the cross-channel pres-sure gradients, thereby reducing the strength ofthe secondary flows. The contour also reducesthe size of the loss region near the inner side-wall, and reduces nozzle profile losses due to thelower velocity at nozzle inlet. Contoured side-walls are typically used in the first stage of a tur-bine section.

GE’s first running stage test of a steam tur-bine using contoured sidewalls was performedin 1983. This test was performed on a small low-energy, two-stage commercial mechanical-driveturbine that included a converging sidewall con-tour on the first stage and a diverging-converg-ing contour on the second stage. The contourswere designed using guidelines presented inReference 11. An overall efficiency gain of 0.7%was realized. Wind tunnel cascade tests conduct-ed in 1984 by CRD produced a diverging-con-verging sidewall contour that increased stageefficiency by 1%. An improved method for opti-mizing the shape of the contour for a givenstage was also developed by CRD during thistime period based on experimental results andextensive CFD analysis. In a parallel effort, arunning stage test of a contoured sidewall wasconducted in 1985 in the subsonic air turbine.

The stage efficiency with the sidewall contourproved to be 1.5% greater than that for theequivalent stage with cylindrical sidewalls, asshown in Figure 16. Subsequent tests onimproved contoured sidewalls yielded gains ofup to 2%.

Contoured sidewalls are routinely used inmechanical-drive turbines, ships service turbine-generator sets, and in the HP, IP and LP sectionsof large fossil and nuclear power generation tur-bines, for both new units and uprate applica-tions. A typical application might include con-toured sidewalls in up to eight stages in themachine, depending on the turbine configura-tion, for an overall turbine efficiency gain rang-ing between 0.25% and 0.50% due to the con-touring alone.

A family of sidewall contours has also beendeveloped for the new solid particle erosion-resistant nozzle designs described in Reference16. The contours improve efficiency and, at thesame time, increase the resistance of the nozzleto SPE damage. These contours are available fornew and retrofit nozzle box designs for controlstage and first reheat stage applications.

New Last Stage BucketsThe last stage bucket is one of the most

important contributors to the performance andreliability of the steam turbine. GE has devel-oped complete families of last stage buckets(Figure 17) that provide sustained levels of highperformance with virtually trouble- and mainte-nance-free operation. These last stage bucketsall share a common continuously-coupleddesign technology, and they range in size from a20-inch (508 mm) 3600 RPM bucket to a 48-inch

9

GER-3713E

GT22187A

Figure 16. Contoured sidewall test data

(1219 mm) 3000 RPM titanium bucket. Designfeatures of these families include:

• Improved vane profiles, including a con-vergent-divergent supersonic passagedesign with high solidity (Figure 18)

• Improved radial mass flow distribution tooptimize efficiency

• Improved tip leakage control that permitsmoisture removal (Figure 24)

• Continuously coupled covers, either sideentr y or over-and-under (Figure 19).These covers provide structural couplingand damping, and also control vane tipflow passage geometry for the reduction ofleakage, secondary flow and shock losses

• Optimum location of midspan connectionfor structural damping and vibration con-trol, either a loose tie wire or a nub-sleeveconnection

• Self-shielded erosion protection. No sepa-rate attachment of shielding is necessarydue to the use of high strength erosion-resistant materials with excellent tough-ness characteristics and a proven servicerecord.

With the last stage buckets typically producing10% of the total unit output, and up to 15% insome combined-cycle applications, improve-ments in last stage efficiency can significantlyimpact the output of the total unit. Retrofittingan older last stage design with a moderndiaphragm and last stage bucket can typicallyimprove heat rate and output by up to 1%.Typical economic payback periods for the newlast stage buckets are two to three years.References 12 and 13 describe the family ofmodern last stage buckets in more detail. Themodern buckets can be retrofit into most exist-ing units. Table 1 lists the overall heat rate

10

GER-3713E

GT22198

Figure 17. 3600 rpm family of continuouslycoupled buckets

GT21881

Figure 18. Convergent-divergent supersonicbucket tip profile design

GT22200A

Figure 19. Continuously coupled bucket tipdesign

Over/Under Cover Design of 20” (508mm)

and 23” (584mm) LSB Redesign

Side Entry Cover Design of 26” (660mm)

and 33.5” (851mm) LSB Redesign

improvements that can be expected byretrofitting the new designs in older machines.

In general, complete new modern LP turbinesteam paths have been designed to go alongwith all of the new last stage buckets.Replacement of an older LP turbine steam pathwith a modern steam path design can improveheat rate by an additional 1.0% to 1.5% beyondthe gain achieved by replacing the last stagealone.

In the 1960s, GE first developed the continu-ously coupled last stage bucket design for 3600RPM operation with the 30-inch (762 mm) and33.5-inch (851 mm) buckets (Reference 14).These buckets had been developed to supportthe increasing unit size requirements of the timeand had been applied to units as large as 840MW. With the shift of energy demands to small-er units, GE began a series of development pro-

grams to apply the latest design technology,along with the proven concepts of the 30-inch(762 mm) and 33.5-inch (851 mm) last-stagebuckets, to the smaller last-stage buckets. Theseefforts resulted in the development of the familyof modern continuously coupled last stage buck-ets for 3600 RPM operation shown in Figure 17.This family includes high efficiency last-stagebuckets in 20-inch (508 mm), 23-inch (584mm), 26-inch (660 mm), 30-inch (762 mm) and33.5-inch (851 mm) sizes for both new units andthe upgrade market.

A recent addition to the 3600 RPM family is a40-inch (1016 mm) titanium last stage bucket,described in Reference 15. A series of rigorousmechanical tests of the new bucket were com-pleted on the test rotor shown in Figure 20, andall design criteria were met. This bucket hasbeen shipped for units in a STAG™ application.New bucket and rotor materials are being devel-oped to allow new, even longer titanium last-stage buckets to be designed over the next sever-al years. With longer last-stage buckets andlarger annulus sizes, more compact and cost-effective turbine designs are achievable. For amore detailed discussion of GE’s line of com-pact High-Power-Density™ turbine-generators,see Reference 16.

GE has also applied the successful continu-ously coupled technology to the 50-Hz productline. The new 3000 RPM last-stage buckets con-tain all of the modern design features listedabove for the 3600 RPM family. The first of the50-Hz product line to be modernized was the33.5-inch (851 mm) last-stage bucket, developedin the 1980s. GE has produced several largesteam turbines that utilize this bucket. Designshave also been completed that use the 33.5-inch(851 mm) last-stage bucket on combined-cycleSTAG™ applications.

11

GER-3713E

GT25793

Figure 20. 40-in. (1016 mm) titaniumlast stage bucket

Table 1PERFORMANCE GAINS WITH RETROFITTED NEW LAST STAGE BUCKETS

OVERALL PERFORMANCE IMPROVEMENTRPM BUCKET BUCKET ONLY BUCKET & DIAPHRAGM

3600 20 inches (508 mm) * 0.25% ro 1.0%23 inches (584 mm) 0.25% to 0.75% 0.5% to 1.0%26 inches (660 mm) * 1.0%30 inches (762 mm) 0.6% 1.0%

3000 26 inches (660 mm) * 1.0%

1800 43 inches (1092 mm) 0.35% 0.5%

*Bucket & diaphragm replacement only

The latest developments in the 50-Hz productline include new 26-inch (660 mm), 42-inch(1067 mm) and 48-inch (1219 mm) last-stagebuckets and LP turbine steam paths. The new26-inch (660 mm) low-pressure turbine, which isa modernization of GE’s older but very success-ful 26-inch (660 mm) design, has been devel-oped specifically for combined-cycle STAG™applications. The 42-inch (1067 mm) low pres-sure turbine was developed to provide a morecompact single-flow alternative to the double-flow 26-inch (660 mm) design for STAG™ appli-cations. As shown in the top of Figure 21, the LPcasing in the single-flow design is considerablysmaller and the large crossover pipe is eliminat-ed, compared to the double-flow design at thebottom of the figure. The 42-inch (1067 mm)last stage bucket is also used in a compact two-casing 500 MW utility unit. A 48-inch (1219mm) titanium bucket is also available for appli-cations requiring more annulus area.

GE continues to advance the state-of-the-art inthe development of half-speed turbines for largecross-compound and nuclear units. The sameadvanced aerodynamic and mechanical featuresdescribed above for the 3000- and 3600-RPMfamilies have been applied to the 1500- and1800-RPM families. These buckets have beendesigned with two loose tie wires, and the bucketleading edges are protected from erosion byflame hardening.

Recent developments include a 52-inch (1321

mm) 1500 RPM LP turbine for a large nuclearapplication. The low-pressure turbine was com-pletely redesigned using the latest aerodynamicand mechanical design methods, resulting inimproved thermodynamic performance andreliability. In addition, the 43-inch (1092 mm)last-stage bucket for 60-Hz applications has beenmodernized to improve its performance andreliability. Retrofitting the new bucket anddiaphragm in a typical nuclear unit results in aperformance improvement of 0.5%.

Bucket Tip and Shaft Packing SealsAn extended effort, until recently mostly

experimental in nature, has been directedtoward finding ways to minimize leakagethrough bucket tip and shaft packing clear-ances. Bucket tip leakage controls have typicallyutilized a single spill strip mounted on theupstream side of the bucket shroud, or two spillstrips mounted on either side of the bucketcover tenon. Leakage along the shaft is con-trolled by packings containing multiple teethlocated between the diaphragms and the shaftand at the ends of the shaft. Leakage can bereduced by either reducing the radial clearancebetween stationary and rotating components, orby designing the seal geometry to create a tor-turous path that induces turbulence andrestricts the flow.

Experimental data on leakage control hasbeen obtained largely through two-dimensionalstationary air tests (such as the rig shown inFigure 57) rather than more expensive 3D rotat-ing steam tests. Stationary air tests allow a muchlarger variety of leakage control configurationsto be tested at relatively low cost. In the past,rotating steam tests were expensive and usuallyrun to determine the overall effects of designchanges, thus making it difficult to separate outthe leakage effects. However, GE has developeda unique, cost-effective method for testing newbucket tip seal concepts in our subsonic air tur-bine (Figure 50), by shrink-fitting a thin ringcontaining the tip seal geometry onto the outerrim of the blisk test wheel (described later inthis paper). By testing the wheel with and with-out the new seal, the effects of the seal can beisolated.

Over the years, GE has tested a great varietyof bucket tip seal configurations. Figure 22shows a sampling of these configurations. Thetest results show that leakage flow is substantially

12

GER-3713E

GT19363/GT19644

Figure 21. Comparison of single-flow 42-inch(1067 mm) and double-flow 26-inch(660 mm) units

reduced when a stepped or high-low spill strip isused. The configuration currently used on themajority of HP and IP bucket tip seals includessteps with double teeth on both the upstreamand downstream sides of a recessed tenon. Theapplication of these improved radial tip leakagecontrols improves sustained efficiency as well,because any damage to the improved controlsdue to a rub results in less increase in leakagethan an equivalent rub to a single radial tip spillstrip. Slant teeth also help to reduce leakageflow, as long as the proper tooth spacing hasbeen selected. Figure 23 shows test data

obtained in our seal test rig that illustrates theimportance of choosing the optimum ratio oftooth pitch to radial clearance to minimize leak-age flow.

In recent years, particular attention has beenfocused on last stage bucket tip leakage control.In older industrial and mechanical drive turbinedesigns, the only leakage control provided at thetip of the last stage bucket is the axial clearancebetween the downstream face of the diaphragmand the leading edge of the bucket band. Thisclearance generally is larger on the last stagethan on any other, and the resultant leakagelosses can be significant. If a radial spillband canbe placed on the last-stage bucket shroud and atight radial clearance maintained, stage efficien-cy can be improved by as much as 3% (depend-

13

GER-3713E

GT22123A/GT22208

Figure 22. Bucket tip leakage controls

GT25794

Figure 23. Slant tooth bucket tip seal testresults – optimum tooth spacing

GT21884

Figure 24. Side entry last stage bucket coverdesign tip leakage control

GT25798

Figure 25. CFD analysis of bucket tip sealpassage – velocity vectors

ing on bucket length), which is equivalent toabout 0.3% in overall turbine efficiency. Allmodern GE 20-inch (508 mm) last stage bucketsand larger now include a radial spillband on thebucket cover to maintain a tight radial clearanceand minimize tip leakage flow. Figure 24 showsthe radial spillband configuration used on the26-inch (660 mm) and 33.5-inch (851 mm) last-stage buckets.

To complement GE’s experimental program,the company is using a commercially availableviscous 3D CFD code called STAR-CD to per-form detailed analyses of new bucket tip leakagecontrol and shaft packing configurations. Thiscode has a user-friendly graphical user interfaceand excellent post-processing capabilities, and isideally suited to performing parametric studies.Figure 25 shows velocity vectors calculated bySTAR-CD at the leading edge of a flat bucketcover with slant teeth mounted on the casing.The code clearly shows the vortex structures inthe leakage path. The code is used as a numeri-cal test vehicle to identify promising seal config-urations that warrant testing in either the sta-tionary seal test rig or the subsonic air turbine.

A new area GE is investigating in more detailis the effect of the tip seal leakage flow on thedownstream blade rows as the leakage flow reen-ters the steam path. The flow leaking over thetop of the bucket cover enters the tip seal pas-sage with a high tangential velocity component.While the torturous path through the seal geom-etry reduces the axial component of the leakage

flow, it does relatively little to affect the tangen-tial component. As the leakage flow reenters thesteam path on the downstream side of the buck-et, the still largely tangential flow must mix withthe largely axial flow leaving the bucket. Thisgenerates mixing losses and can lead to inci-dence angle problems in the downstream noz-zle. Traverse measurements of stage exit swirlangle obtained in the subsonic air turbine andin the low speed research turbine have con-firmed this behavior, as shown in Figure 26. GEhas refined the tip leakage models in its quasi-3D flow solver to reflect this behavior (alsoshown in Figure 26). This allows designers tomore effectively account for this effect in thepreliminary design of multi-stage turbines.

Exhaust HoodsThe steam leaving the last row of buckets in

the turbine is directed into the condenser inletby the exhaust casing, also known as the exhausthood. Poor aerodynamic design of the exhausthood can produce a large pressure drop fromthe last-stage bucket exit to the hood exit. Thispressure drop constitutes an efficiency lossbecause it represents energy that cannot be usedby the turbine to produce power. A well-designed hood can diffuse the flow, causing thestatic pressure to increase from the last stagebucket exit to the hood exit flange to match thefixed condenser pressure. This pressure recov-ery gives the turbine an effective back pressurethat is lower than the condenser pressure, sothat more energy is available for producingpower. In this manner, a good diffusing hoodcan produce a significant overall gain in effi-ciency. A diffusing hood also decelerates theflow as it approaches the exhaust flange, thusreducing the likelihood of flow-generated noiseand vibration in the condenser tubes.

In downward flow configurations (Figure 27),the exhaust hood must turn the flow downward90° before it enters the condenser inlet. In thepatented GE hood design concept, structuralsupport plates inside the hood are used to parti-tion the total flow area into smaller passages inwhich flow diffusion can be carefully controlled.Streamlined steam guides and flow guide vanesare used to further improve the aerodynamicperformance of the flow passages so that majorflow separations are avoided. Internal flanges,support struts, pipes and other protrusions intothe flow passages that cause efficiency losses andgenerate noise are removed or minimized.

GE’s approach to exhaust hood design in the

14

GER-3713E

GT25795

Figure 26. Influence of tip seal re-entry flow onstage exit flow angle — comparisonof quasi-3D analysis to test data

past was purely experimental. To optimize a newdesign, the internal flow passages of the hoodwere duplicated in an exact scale model (Figure28) which was tested in an air flow test facility atGE’s Aerodynamic Development Laboratory.Pressure measurements were taken at variouspoints inside the hood model to determine theamount of flow diffusion taking place.Interchangeable parts in the model allowedmany different configurations to be tested sothat optimum flow passage configurations couldbe achieved. Since the early 1960s, literallydozens of different exhaust hood models inhundreds of different configurations have beentested, and a performance prediction anddesign method based on various hood geometryparameters has been developed from this exten-sive test data base.

Axial flow exhausts (Figure 29) have signifi-cant size and efficiency advantages over down-ward flow exhausts. If the requirements of theoverall power plant dictate that the best locationfor the condenser is on the same level as the tur-bine exhaust, then an axial exhaust is the pre-ferred choice. Axial flow exhausts are simpler todesign and build, and can be much more effi-cient diffusers than downward flow exhaustsbecause the flow does not have to be turned 90°.The velocity distribution at the exhaust flangecan also be made quite uniform, resulting inimproved condenser performance. Single-casingaxial exhaust designs can be conveniently base-mounted for easier shipment and installation, asshown in Figure 29.

Axial flow hoods lend themselves to a moreanalytical design approach because of their sim-pler geometry compared to downward flowhoods. Numerous papers published in the openliterature by a group of researchers at StanfordUniversity (Kline, Johnston, et al.) have beenparticularly useful. However, the principal refer-ence used as a guideline for the design of axialflow hoods is a paper by Sovran and Klomp(Reference 17). GE has designed many axialflow hoods for commercial applications basedon this paper, and laboratory tests on scale mod-els have confirmed the excellent performanceof these designs.

GE’s approach to exhaust hood design hasimproved dramatically in the last few years, withthe advent of CFD codes that are capable of

15

GER-3713E

GT22194

Figure 27. Downward-flow exhaust hood

GT22195

Figure 28. Downward-flow exhaust hood testmodel

accurately modeling the complex geometry andflow phenomena inside exhaust hoods. Figure30 shows a computer-generated solid model ofthe internal components of a downward flow dif-fusing exhaust hood that had been tested in thelaboratory using lampblack and oil to visualizethe flow. Figure 31 shows the corresponding cal-culation grid required for a detailed CFD analy-sis of this hood. Figure 32 shows sample results

from an analysis using the STAR-CD code, inthis case velocity vectors and magnitudes. Thecode was able to predict the visualized flow fea-tures and measured pressure recovery very well.Because STAR-CD uses a structured gridapproach, generating the solid model and thecalculation grid took several man-months.However, this was less time than it took to buildthe original air test scale model, and the analysis

16

GER-3713E

GT23020A

Figure 29. Axial-flow exhaust hood

GT24590

Figure 30. Solid model of downward-flowexhaust hood for CFD analysis

GT24591

Figure 31. Calculation grid for exhaust hoodCFD analysis

yielded far more detailed information about theflow inside the hood than was obtained in thelaboratory model tests. Numerous design refine-ments have already been incorporated into GE'sdownward flow hood designs as a result of usingCFD analysis.

The NOVAK3D CFD code has proven to bean extremely useful tool for the analysis of com-plex stationary flow components such as exhausthoods. Because it eliminates the need for time-consuming manual grid generation and refine-

ment, solutions to complex problems can beobtained in a fraction of the time required withstructured grid codes. Figure 33 shows an analy-sis of an axial flow hood that contains a numberof struts, pylons and pipes. The conical outerwall has been removed for clarity, and only thesurface grid is shown. Once the surface modelof the hood was created using a CAD program,NOVAK3D was able to create a grid and con-verge to a solution in only five days on a singleworkstation. Figure 33 also shows “flow ribbons”that follow streamtubes through the flow, whichindicate that the flow is generally well-behavedthroughout the hood.

The NOVAK3D code was used to redesign thelarge pylons supporting the hood to eliminatelarge regions of separated flow downstream ofthe pylons. On the top portion of Figure 34,velocity vectors around the original pylon areshown. A packing seal pipe is located down-stream of the pylon in a region of massive sepa-ration. On the bottom part of Figure 34, theredesigned, more streamlined pylon is shown.The packing seal pipe has been relocated insidethe new pylon, and the region of separation hasbeen eliminated.

Valves and InletsA new development program has been initiat-

ed to develop more efficient and cost-effectiveinlet and valve designs using the NOVAK3Dcode to analyze the flow in the complicatedgeometries of these components. Figures 35through 38 show examples of the use of thecode to better understand the behavior of exist-ing valves. Figure 35 shows the surface grid gen-erated by NOVAK3D for a steam turbine bypassvalve that previously had been air tested in GE’sAerodynamic Development Laboratory. Figure36 compares the valve flow coefficient calculatedby NOVAK3D at two different valve lifts to theair test data, and it shows that NOVAK3D is ableto predict the flow coefficient quite well.

Figure 37 shows the complex surface grid cre-ated by NOVAK3D for an older control valvedesign. This detailed model includes the holesin the strainer located in the upper part of thevalve body. The “flow ribbon” plot (Figure 38)generated from the analysis of this valve shows atightly-wound vortex structure in the lower valvebody, which could be a source of pressure lossand noise generation. NOVAK3D solutions suchas this have provided many valuable insights intothe behavior of the flow in valves, and manyimprovements are being implemented to reducevalve pressure losses by up to 30%.

17

GER-3713E

GT24592

Figure 32. Velocity vectors in downward-flowexhaust hood

GT25796

Figure 33. NOVAK3D analysis of axial-flowexhaust with “flow ribbons”

NEW DESIGN FEATURES FOR IMPROVED SUSTAINED

EFFICIENCY

SPE-Resistant Steam Path DesignsA primary cause of steam path efficiency

degradation in units with high temperature inletstages (> 900 F/482 C) is solid particle erosion(SPE) damage. The steam path degrades as the

nozzles and buckets are worn away by the pas-sage of steam contaminated with iron oxide par-ticles exfoliated from the inside surfaces of boil-er tubes and main and reheat steam piping. Thedamage results in reduced steam path efficiency,lost power generation, shortened inspectionintervals and the costly repair and replacementof damaged components. Industry estimatesmade in the early 1980s pegged the annualindustry-wide cost of SPE damage at $150 mil-lion.

In 1980, GE launched a multi-faceted devel-opment program dedicated to finding cost effec-tive solutions to this nagging industry problem.Early in this program it was recognized that thedevelopment of a sound, fundamental under-standing of the various erosion mechanisms atwork in the turbine steam path would be essen-tial to the success of any program dedicated toreducing SPE. It was further recognized that thebasic elements of this understanding would haveto come from analytical studies of solid particlebehavior as it occurs in the steam path. The ana-lytical capability required to accomplish thiscomplex task was quickly put in place, becomingoperational in mid-1981.

The heart of the technology is the ability tomodel the 3D trajectories of the particles as theypass through the steam path. This allows theidentification of the locations of particle impactand the attendant critical parameters of impactvelocity and impact angle which altogetherdefine the erosion potential for a specificdesign. The influence of key factors such asbasic stage design variables, geometry, steamconditions and mode of turbine operation can

18

GER-3713E

GT25797

Figure 34. NOVAK3D analysis of flow near struts in axial-flow exhaust

GT25803

Figure 35. NOVAK3D grid for bypass valve

be investigated to determine how they con-tribute to the occurrence of SPE damage. Theability to predict the occurrence of SPE on noz-zles and buckets in HP and IP section steampaths has been clearly demonstrated with thistechnology. The analysis has identified the exis-tence of completely different erosion mecha-nisms at work in the HP and IP turbine sectionsrequiring different design modifications to miti-gate the SPE damage.

Service experience has consistently shown

that the first stage of the high pressure turbine(i.e. the control stage) and the first stage of thereheat turbine generally receive the predomi-nant share of SPE damage in their respectivesections. The following paragraphs briefly sum-marize the erosion mechanisms at work in thesetypes of stages and describe field-proven designmodifications that vastly reduce the susceptibili-ty of these components to SPE damage.

Particle trajectory analysis of control stagenozzles (see Reference 18) has demonstratedthat the erosion is caused by the high velocity,low angle impact of particles on the trailingedge pressure surface of the partition. The cal-culated range of impact angles coincides withthat which produces the maximum erosion ratein nozzle partition material. One commonlyused means of reducing the damage is the appli-cation of an erosion-resistant coating. The vari-ous coating alternatives available for controlstage nozzle assemblies are described in detail inReference 18. This component requires a non-line-of-sight coating process to provide protec-tion on the pressure side trailing edge surface ofthe partitions. Service experience has demon-strated that the use of a protective coating alonewill only provide about one to two years of SPEreduction when applied to conventional controlstage nozzle designs. Thus the limited life of theavailable coating only produces a “time delay” inthe erosion process. This limitation was recentlyovercome by the development of a new steampath design that significantly improves the ero-sion resistance of control stage nozzles. The newdesign (shown in Figure 39) uses slanted nozzlepartitions which significantly reduce the num-ber of particle impacts on the nozzle trailingedge pressure surface. Those impacts that dooccur are at a reduced velocity and extremelyshallow impact angles. A significant reduction inthe rate of SPE damage is therefore obtained.The combination of this new nozzle shape witha protective diffusion coating produces a controlstage nozzle design with three times the servicelife of conventional designs. This new design hasbeen achieved with no sacrifice in the “as new”performance while providing tremendous bene-fits in long-term sustained efficiency. The heatrate benefit achievable is shown in Figure 40 fora typical unit on a five-year maintenance inter-val.

The SPE-resistant control stage design hasbecome standard equipment on new units andcan be accommodated in either a plate or noz-zle box control stage design. It is retrofittable toexisting nozzle box designs as well. Retrofitting a

19

GER-3713E

GT25799

Figure 36. NOVAK3D analysis of bypass valve –comparison of calculated flow coeffi-cient with test data

GT25804

Figure 37. NOVAK3D grid for control valve

new SPE-design steam path restores the controlstage efficiency to the “as new” level eliminatingany nozzle profile losses that may have beencaused from weld repairing eroded nozzles.

The new SPE-resistant control stage nozzledesign has been in service since 1987. The ini-tial application was made to two 650-MW super-critical double-reheat units with well-establishedhistories of severe control stage SPE. In fact, theerosion was severe enough to require major con-trol stage nozzle repairs on an 18- to 24-monthcycle throughout the 1980s. The SPE design wasrun for three years (twice the normal time)before it was inspected in October 1990 duringa planned maintenance outage. Figure 41 showsa side-by-side comparison of the condition of thenew design after three years of service with theseverely eroded conventional design after onlyone and a half years of service. The SPE designis virtually erosion-free. The utility demonstrateda 0.4% heat rate improvement after the first twoyears of service with the new design. They rein-stalled the nozzle box with no repairs requiredand have extended the date of the next plannedmaintenance inspection based on the perfor-mance of the new SPE design.

Additional field confirmation of the SPEdesign has been obtained on two supercriticalunits with well-established histories of severeSPE. A major repair of the control stage nozzlewas traditionally required on these machinesevery four to five years. The new SPE-resistantdesign was run for four years in one machineand four and a half in the other. Both machineswere recently opened for inspection revealingonly minimal SPE damage. No repairs wererequired and the nozzle boxes were returned toservice.

Particle trajectory analysis has also been usedto determine the cause of erosion to the firstreheat stage nozzles and buckets. (For a detailedreview of the analysis and its findings, seeReference 18.) The basic erosion mechanism atwork in a reheat stage is a complex particleentrapment phenomena that results in the mul-tiple rebounding of captured particles betweenthe nozzles and buckets. As Figure 42 shows, thetrajector y analysis demonstrated that firstreheat-stage nozzle partition erosion resultsfrom the rebounding of particles from the lead-ing edges of the buckets to the suction-side trail-ing edge region of the nozzles. The high tangen-tial velocity acquired by the particles from theirimpact with the rotating buckets basicallybecomes the impact velocity of the particles asthey strike the nozzle convex surfaces at a low

20

GER-3713E

GT25811

Figure 38. NOVAK3D analysis of control valve

GT22201

Figure 39. Modified control stage to minimizeSPE damage

GT16903A

Figure 40.Control stage heat rate loss due tosevere SPE damage

impact angle — a very erosive condition for thenozzles. Figure 43 shows the typical SPE damageexperienced by a reheat diaphragm with thenozzle convex surface damage clearly evident.

As discussed in Reference 18, the most benefi-cial action that can be taken to combat thereheat erosion mechanism is to substantiallyincrease the edge-to-edge axial clearancebetween the nozzles and buckets. Increasing thisclearance profoundly affects the particle entrap-ment/rebounding mechanism to the pointwhere the rebounding is nearly eliminated.Since some rebounding by very large particlesmay still persist even with an increased clear-ance, the nozzle trailing edge suction surfaceand the diaphragm outer sidewall are also pro-tected by GE’s newly developed Diamond Tuff™HVOF (high velocity oxygen fuel) erosion-resis-tant coating. Figure 44 shows the coating beingapplied to a reheat diaphragm by a robot. (Amore detailed description of the Diamond Tuffcoating can be found in Reference 19.)

Figure 45 compares the performance deterio-ration produced by SPE of a conventional

21

GER-3713E

GT23105A

Figure 41. Field performance of new SPE resistant design

GT21959/GT22203

Figure 42. First reheat stage suction surfacedamage caused by particlerebounding

Old Design New Design

GT17231

Figure 44. Automated process coats reheatdiaphragm

GT17232

Figure 43. Severe erosion on suction side offirst reheat stage diaphragm

reheat stage design with that expected fromGE’s new SPE-resistant design. A significant heatrate benefit is provided by the sustained efficien-cy available from the new design.

The new SPE-resistant design features havebeen incorporated into GE’s new unit designssince 1985. They can also be retrofitted to alarge selection of existing reheat diaphragmdesigns in service today, and many applicationshave been implemented since 1985. Figure 46shows an SPE-resistant diaphragm that was inservice for three and a half years in the firstreheat stage of a super-critical reheat machine.The normal condition of the conventional firstreheat stage diaphragm from that unit afterthree to four years of service was comparable tothat shown in Figure 43. Negligible damage hasoccurred on the new design after the same timeperiod. No repairs were required and the com-ponent was reinstalled in the unit.

The results of a more demanding applicationare shown in Figure 47. This diaphragm wasinstalled in the first stage of the second reheatsection of a super-critical double reheatmachine with a previously demonstrated severecase of SPE. The relatively low design pressureof this stage aggravates the rebounding damagemechanism, providing a potent challenge forany SPE-resistant design. After three and a halfyears of service, Figure 47 shows the design tobe up to the challenge. It was reinstalled in theunit without repairs to run until the next sched-uled maintenance outage.

Positive Pressure Variable ClearancePacking

A recent innovation that improves the clear-ance control at shaft packings is the GE positivepressure variable clearance packing. Labyrinthseal packings close to the midspan of a high-temperature steam turbine rotor are susceptibleto rubbing. Operation below the first criticalrotor speed, acceleration through criticalspeeds, and boiler temperature variations alloccur at startup, making the packing most vul-nerable during this period. Excess clearancecaused by rubbing during the startup of the unitresults in increased fuel costs and a reduction inunit capacity. In addition, vibration problemsassociated with packing rubs can prevent theturbine from getting through its critical speeds,prolonging the startup of the unit. GE positivepressure variable clearance packing provides alarge clearance during startup, and reducedclearance after the unit has synchronized. This

22

GER-3713E

GT22204

Figure 46. SPE-resistant diaphragm after3 1/2 years of service in the firstreheat stage of a supercriticalreheat turbine

GT16904

Figure 45. First reheat stage heat rate loss dueto severe SPE damage

GT22205

Figure 47. SPE-resistant diaphragm after3 1/2 years of service in the firststage of second reheat stage of asupercritical reheat turbine

arrangement minimizes rubs associated with tur-bine startups while providing optimum sealingwhen the unit is loaded.

The GE positive pressure variable clearancepacking, shown in Figure 48, utilizes a combina-tion of the pressure drop across the packing andan additional pressure force, when required, toclose the packing rings after synchronization.Also, an external control of the packing rings ina mid-span packing of an opposed flow unit canbe provided. This allows the unit to be pre-warmed by pressurization on turning gear aftera prolonged outage without the unit rolling offturning gear. A more detailed description of thepositive pressure variable clearance packing isgiven in Reference 20.

For an example of the application of positivepressure packing, consider a 500 MW reheatunit that has opposed-flow HP and IP sections in

a single casing. The internal mid-span packing(N2 packing) and the first three HP and IPstage diaphragm packings are susceptible torubs during the startup of the unit. The normaldesign clearance for these packings is 0.015 inch(0.381 mm), but after five years of operation,the packing clearance is typically found to beopened up to 0.06 inch (1.524 mm). The impactof this increased radial packing clearance onheat rate can be calculated using Reference 21.Compared to design, the 0.06 inch (1.524 mm)radial clearance causes the unit to experience aheat-rate loss of about 0.35%. The packings arenormally replaced every five years and the clear-ances restored. Figure 49 shows this perfor-mance loss with time for the present standardpackings, assuming that the clearance increaseslinearly with time. Application of the positivepressure packings will eliminate the major shaftpacking rubs since these rubs usually occur atstartup, when the positive pressure shaft pack-ings have a large radial clearance. The positivepressure packings will hold design clearancewith time, since they will not be subject to thestartup transients seen by the standard packings.Therefore, the heat rate advantage of the posi-tive pressure packing would be the area underthe curve shown on Figure 49.

In another example, positive pressure variableclearance packings were installed on the inter-nal packing between the HP and IP sections (N2packing) and the first three HP and IP stages ofthe 525 MW Jim Bridger #2 unit of Pacific Power& Light. Although no efficiency tests were con-ducted to identify the improvement resultingfrom the HP and IP diaphragm packings, pres-sure measurements were recorded at the startupof the unit that clearly demonstrated the closureof each of the packing rings in the N2 packing.Similar tests were conducted after six months ofoperation that demonstrated the continued suc-cessful operation of the positive pressure pack-ings.

Typical economic payback periods forinstalling positive pressure variable clearancepacking have been less than a year.

LABORATORY TESTPROGRAMS

Subsonic Air TurbineThe principal test vehicle used for the devel-

opment of new stage design concepts is the sub-sonic air turbine located in GE Power

23

GER-3713E

GT22026

Figure 48. GE positive pressure packing

GT22207

Figure 49. Heat rate improvement for positivepressure packing in 500 MW units

Generation’s Aerodynamic DevelopmentLaboratory in Schenectady, New York. This rig,shown in Figures 50 and 51, permits the quickand accurate evaluation of turbine test stages atactual operating speeds and pressure ratiosusing compressed air as the working fluid. Inaddition to traditional performance testing, noz-zle and stage exit traversing as well as laser-assist-ed data acquisition procedures are employed tocollect more detailed flow field information.Numerous new configurations have been tested

in this facility over the past several years to con-firm that significant efficiency gains have beenachieved, and to validate the CFD codes.

The facility is arranged as a closed loop. A4810 HP centrifugal compressor is used to circu-late the air around the loop and a smaller make-up air compressor permits the desired stageinlet pressure to be maintained. Prior to reach-ing the test section, the air supply passesthrough a water-cooled heat exchanger to per-mit the stage inlet temperature to be accuratelymaintained. A hydraulic actuation system slidesthe inlet plenum apart from the exhaust scroll,permitting easy access to the test stage. Testwheels are about 38 inches (965 mm) in diame-ter overall and have active blade lengths in the2-inch (50.8 mm) to 5-inch (127 mm) range.Shaft speeds vary between 2000 and 4800 RPM.A typical intermediate-pressure test stage pro-duces about 650 HP (485 kW) near its designpoint. Figure 52 shows a typical air turbine testdiaphragm with compound lean nozzlesdesigned for an intermediate pressure stage,while Figure 53 shows the test wheel for thesame stage.

A water-cooled inductor type dynamometerthat is supported on hydrostatic oil pads is usedto measure the shaft torque developed by thewheel. An enhanced accuracy rotary torquemeter provides an additional torque reading forcomparison and redundancy. A coupling that

24

GER-3713E

GT24585

Figure 50. Subsonic air turbine test facility

GT25805

Figure 51. Cross-section of subsonic air turbine test facility

permits axial movement is used for the gear-driven positioning of the wheel relative to thediaphragm. This feature is particularly useful foraccurately setting the nozzle-to-bucket axial gapand it provides a way to easily study the effect ofchanging the edge-to-edge distance. A bearingload loss cell system floats the entire stub shaftand bearing assembly on a hydrostatic oil film.This permits the torque loss due to bearing fric-tion to be accurately determined during all run-ning conditions.

Traversing at nozzle and bucket exit planes isdone using thermocouple probes for tempera-ture measurement, and three-hole wedge and

cobra probes with auto-nulling features for pres-sure and flow angle measurements. A recentinnovation in the facility allows laser Dopplervelocimetry and laser light plane flow visualiza-tion techniques to be used to further GE’sknowledge of the flow mechanisms responsiblefor improved stage designs. An instrumentationtunnel that includes a five-axis traverse mecha-nism (shown in Figure 54) permits the accuratepositioning of laser probes and a mirror fordirecting the laser light into the interior of anozzle passage. The laser light passes through aspecially designed fused silica window installedin the outer wall of the diaphragm. Figure 55shows a vortex structure at the exit plane of aleaned nozzle visualized with the laser lightplane technique. The laser light plane dataacquisition system is currently undergoing devel-opment to improve its ability to capture videoimages of flow features with greater clarity. Thistechnique, along with the laser Doppler

25

GER-3713E

RDC27193-03-06

Figure 53. Advanced Vortex “blisk” test wheel

GT25812

Figure 54. Laser tunnel and optics in subsonicair turbine

GT25806

Figure 55. Laser light plane flow visualization

RDC27193-03-06

Figure 52. Advanced Vortex test diaphragm

velocimetry, is providing additional data for vali-dating the CFD codes used to design and evalu-ate alternative aerodynamic design concepts.

Air turbine test wheels, such as the one shownin Figure 53, are produced as a “blisk” (short for“bladed disk”). The wheel, buckets and integralbucket covers are machined out of a single diskof metal with a high degree of precision usingstate-of-the-art five-axis NC machining technolo-gy. The blisk construction eliminates the time-consuming operations associated with assem-bling individual buckets on the wheels, andreduces the manufacturing time by more than50%. Before a new design concept is committedto test hardware, a computer-generated model iscreated, such as the one shown in Figure 56 fora diaphragm segment with compound lean noz-zles. Full size plastic parts are made from thecomputer-generated model using a stereolithog-raphy process to visualize the new components,and to perform manufacturing producibilitystudies. By using these rapid prototyping meth-

ods, we have dramatically reduced our develop-ment cycle time for new aerodynamic designconcepts, and have accelerated the introductionof new high efficiency features into our productline.

Tip Seal Test RigNew non-contacting leakage control devices

are tested in the rig shown in Figure 57 usingcompressed air. The rig accommodates station-ary two-dimensional models of seal geometriesthat are machined into six-inch long (152 mm)flat plates. These seal strips feature straight,slant, stepped, high-low and other tootharrangements in combination with smooth andcontoured bucket cover or shaft surfaces. Forbucket tip seal tests, the test section in the rigcan simulate the leakage flow that goes over thebucket cover and through the bucket tip seal, aswell as the main flow that goes under the coverand contributes to the stage output. These flowsare kept separate and can be independentlycontrolled and measured. The leakage flow ismeasured using an ASME orifice plate, while themain flow is measured using an ASME flow noz-zle. The pressure level in the main flow and thepressure ratio across the seal can also be variedseparately. The maximum pressure ratio thatcan be achieved across the seal is 2.5.

CFD analyses are used along with this rig toscreen potential seal design candidates. Figure23 shows typical data obtained in the rig in arecent test series designed to determine theoptimum tooth spacing for a slant tooth buckettip seal, while Figure 25 shows the equivalentCFD analysis. The more promising candidatesare then tested on a running stage in the air tur-bine.

GEAE Low Speed Research TurbineTest facilities at the GE Aircraft Engines

Aerodynamics Research Laborator y inCincinnati, Ohio, are also used to develop newaerodynamic design concepts for steam tur-bines. The large-scale, multi-stage low speedresearch turbine (LSRT) shown in Figure 58 isused to test large-scale models containing up tothree stages in an environment where ver ydetailed measurements of the flowfield can bemade. A cross-section of the LSRT is shown inFigure 59. The LSRT has a constant casing diam-eter of 60 inches (1.524 m), a vertical axis ofrotation, a calibrated intake system, andPlexiglas casing windows which mount into largesteel casings. Ambient air is pulled through the

26

GER-3713E

GT25783

Figure 56. Computer-generated model ofAdvanced Vortex diaphragmsegment

GT25807

Figure 57. Tip seal test rig

turbine blading from the inlet at the top into alarge plenum beneath the turbine. This air,which is drawn through the blading by a largecentrifugal blower located next to the plenum,causes the turbine to rotate. The power generat-ed by the test turbine is used to assist in drivingthe blower, saving roughly one-third of thepower requirements for the facility. The designtip speed for the turbine is 161 ft/sec (49.1m/sec) at 616 RPM.

The rate of mass flow through the turbine iscontrolled by a set of variable inlet guide vaneson the blower and by variable air-bleed valves inthe outer cylinder. The blower exhausts the airfrom the turbine into an air-handling/mixingsystem outside the laboratory. This system mixesthe exhaust air with outside ambient air in vary-ing proportions before it is returned to the labbuilding, thereby maintaining a constant tem-perature and pressure environment in the lab.

Individual nozzles and buckets are precision-molded from high-strength plastic and assem-bled into rings and discs, thus allowing a widevariety of turbine geometries to be modeled.The airfoils are large enough to be instrument-ed with a matrix of static pressure taps. A rotat-ing scanivalve and slipring allow measurementsto be made in the rotating frame. A variety of

measurement techniques, including laserDoppler velocimetry, are used to gather detailedinformation about the flowfield.

FIELD TEST PROGRAMSGE recognizes that the true measure of the

performance of a new design feature is its per-formance in the field. Analytical predictions andlaboratory tests are extremely valuable, but it isthe field test that once and for all demonstratesthe value and integrity of a new feature.Consequently, extensive precision field tests areperformed on all of the new steam path designfeatures described in this paper as they areintroduced in actual operating units. To date, allunits incorporating advanced aerodynamicdesign features that have been field tested havemet their performance guarantees. The dataobtained from these tests is used to calibrate theCFD codes to ensure that these increasinglycomplex codes are firmly grounded in reality,and are giving the designers the right answers.As these CFD methods are validated by test data,the need for expensive prototype testing will

27

GER-3713E

GT24586

Figure 58. Low speed research turbine testfacility at GE Aircraft engines

GT25800

Figure 59. Cross-section of low speed researchturbine test facility

diminish, leading to substantial reductions indevelopment cycle time and cost and more effi-cient, reliable turbine designs.

DESIGN AUTOMATION ANDOPTIMIZATION

For new turbine designs, it is often necessaryto perform numerous labor-intensive and time-consuming design iterations before all designrequirements are met. The turbine stages in agiven section must be optimized to producemaximum efficiency, while at the same timenumerous mechanical and producibility require-ments must be met. To achieve dramatic reduc-tions in design cycle time, GE has developed asuite of powerful computer-based steam pathdesign automation and optimization tools.

CRD has developed a unique and innovativeworkstation-based tool for the design of bladecross-sections. In this tool, called Bezblade,blade cross-section shapes are described in para-metric form as smooth Bezier curves. Through auser-friendly graphical interface, shown inFigure 60, the aerodynamic designer has directcontrol over familiar geometric design parame-ters such as blade stagger, passage throat, bladethickness, cross-section area, and leading andtrailing edge wedge angles. The blade shape canbe changed by either using a mouse to drag theBezier curve control points, by dragging slidingcontrols representing each design parameter, or

by directly typing in the parameter values. TheBezier curve representation allows individualparameters to be varied without requiring finetuning of the other parameters to achieve rea-sonable shapes. The program is coupled to a fast2D flow solver that allows the designer to manip-ulate the blade shape and receive quick feed-back on the effect of the shape change on theblade sur face Mach number distribution.Individual blade cross-sections can be radiallystacked to form a complete blade, and curves

28

GER-3713E

GT25808

Figure 60. Bezblade screen

GT22214B

Figure 61. Engineous design optimizationprocess

are automatically plotted of the various bladegeometry parameters as functions of radius. Bydirectly modifying these curves using the mouse,the stackup can be quickly smoothed out to pro-duce the final blade shape.

A significant improvement was made toBezblade by coupling it to a 2D version ofNOVAK3D and using a linear perturbationmethod to compute sensitivities of the flow solu-tion to the positions of each of the Bezier con-trol points. Once the sensitivities have beencomputed, the designer can manipulate theblade shape and receive immediate, real-timeupdates on the effect of the shape change. Theblade sur face Mach number distributionchanges literally as fast as the designer canchange the shape with his mouse.

Bezblade has been used with great success todesign advanced vortex stages for productionunits. While this design tool has dramaticallyreduced design cycle time, the design engineermust still use Bezblade in a manual, iterativefashion until he is satisfied that he has achievedan optimum design, and he must visually evalu-ate the quality of the blade design. To achievefurther reductions in cycle time in the optimiza-tion process, GE has made use of a design opti-mization computer program developed by CRD,called Engineous (References 22 and 23). Asshown in Figure 61, Engineous is a user-friendlygeneric software shell that automatically iteratesdesign programs until constraints and designgoals are met. It captures human design knowl-

edge in a “knowledge base” by querying theengineer for the required rules. It does notattempt to replace the designer, but retains theknowledge of different specialists in multipledisciplines. When the design knowledge isincomplete, which is often the case with rapidlychanging technology or very large knowledgedomains, Engineous combines its own intelli-gent searching heuristics with the existingknowledge to solve the problem. It performs thetedious work required to iterate design and anal-ysis codes, such as manipulating and editinginput data, running programs, scanning outputfiles, and it applies design knowledge to modifydesign parameters. By delegating this work toEngineous, engineers can realize a 10:1improvement in the time required to develop anew design, can minimize human error and canachieve improved designs.

Because Engineous is automated, it canexplore many more design options and parame-ter trade-offs in a given period of time.Additionally, the program’s own searchingheuristics and numerical optimization tech-niques often turn up new design solutions thatdesigners had not thought of before. Engineouscan also couple design codes from many disci-plines, such as aerodynamics and mechanicaldesign, to effectively balance interrelated andoften conflicting goals. The program dynamical-ly displays its results as they unfold, allowingdesigners to analyze solutions during and afterthe run, applying their own judgment to the

29

GER-3713E

Select TenonTenon

DatabaseBlade

Database

Best Tenon

Optimize TipRestack

ThermodynamicLayout

Blade ShapeOptimization

AerodynamicTip Section

Optimization

TLP

Initial Design

Section 1 Section 2 Section 3 Section 4 Section 5

Optimize Stack

GT25809

Figure 62. TLP/TADS blade design optimization process

problem. The program has been used success-fully by many GE components for the design ofsuch products as aircraft engine centrifugalcompressors and turbines, superconductinggenerators, cooling fans, and DC motors. GEPower Generation is also using it to optimizepower plant cycles.

GE Aircraft Engines has used Engineous forpreliminary turbine stage layout (References 24and 25) and achieved a better than five to oneimprovement in design cycle time. Encouragedby this success, GE Power Generation coupled aquasi-3D throughflow program to Engineous, sothat the radial flow distribution in advanced vor-tex stages could be optimized. Engineous auto-matically runs a sequence of six design codes tooptimize the radial flow distribution for a givenstage in only a few hours running on a worksta-tion, a procedure that used to take an engineerseveral weeks to do manually.

Based on the success of an early attempt byGEAE to use Engineous to design airfoil sec-

tions (Reference 26), GE Power Generation andCRD developed an automated workstation-basedsystem that uses Engineous to optimize three-dimensional blade profiles to maximize efficien-cy while meeting all mechanical design require-ments. The system was created by couplingEngineous to Bezblade, a 2D flow solver, aknowledge base containing aerodynamic andmechanical design rules and constraints, and anumber of other blade design utilities. Thissuite of programs is known as the TurbineAir foil Design System (TADS), and it isdescribed in Reference 27.

The overall blade design process is illustratedin Figure 62. The initial layout of the turbinesection is done by the Thermodynamic LayoutProgram (TLP) module, which generates ther-modynamic data including entrance and exitflow angles for all of the blade rows based on aquasi-3D streamline calculation. This layoutincorporates the effect of blade lean and otheradvanced vortex features. An engineer sets up

30

GER-3713E

GT25810

Figure 63. TADS/Engineous screen

an initial baseline blade design file based onTLP output, and launches TADS. For each buck-et design, five cross-sections equally spaced fromroot to tip are designed simultaneously on fiveseparate workstations to reduce the design cycletime. For each of these cross-sections, Engineouschanges the blade cross-section shape usingBezblade, runs the flow solver, evaluates the out-put and makes new changes to the shape basedon the results of its last change. It continues to

iterate on the design until it achieves maximumblade “quality,” which is defined as a combina-tion of factors that ensure that the optimumblade surface Mach number distribution isachieved, and that the blade shape is radiallysmooth. The most difficult element in the sys-tem to develop was the translation of the design-er’s visual subjective perception of the designinto an analytic evaluation that the computercould perform.

The designer can monitor the optimizationprocess for each cross-section by calling up theworkstation screen shown in Figure 63. The cur-rent blade shape is displayed in the lower leftwindow, and the corresponding blade surfaceMach number plot is displayed in the lowerright window. The blade quality parameter is dis-played in the upper right window. Figure 64illustrates the type of shape changes made bythe system during the optimization process.These windows are updated every few seconds asthe optimization proceeds. Typical cross-sectiondesigns require one to four hours of executiontime, depending on the quality of the initialbaseline design.

Once the five cross-sections have beendesigned, TADS brings them together and usesEngineous to smooth out the blade stackup. Thesystem then enters the tenon database andselects a tenon that meets mechanical designrequirements and fits on the bucket tip sectionas closely as possible. If the tenon cross-section islarger than the vane tip section, Engineousredesigns the tip section to fit around the tenon,and then modifies the vane below the tip section

31

GER-3713E

GT25801

Figure 64. Results of TADS blade shape optimization

GT25802

Figure 65. Solid model of typical bucketdesigned by TADS

to achieve a smooth blend. Figure 65 shows atypical intermediate-pressure bucket designedby the system.

This system is now used to design alladvanced vortex buckets for production steamturbines. It is also being used to design airfoilsfor aircraft engines and advanced gas turbines.A fifteen-to-one reduction in design cycle timehas already been achieved, and further gainscontinue to be made as the system is refined. Byincorporating mechanical design constraintsinto the system, TADS allows us to design fullycustomized turbines that exactly meet the needsof each individual customer in considerably lesstime than it took in the recent past, and with nocompromise in mechanical integrity and relia-bility. This remarkable design tool also allows usto respond to retrofit opportunities more quick-ly, thus giving our customers more options forimproving existing designs.

CONCLUSIONIn this paper, recent advances in GE steam

path technology have been described. Advancedcomputational fluid dynamics codes are beingused to develop new aerodynamic design con-cepts to dramatically improve efficiency.Extensive laboratory test programs are beingperformed to verify the predicted efficiencygains of new design features, and to validate theCFD codes. Unique design automation and opti-mization tools have been developed that allowfully customized and cost-effective turbines to bedesigned in short cycle times to meet the needsof individual customers.

REFERENCES1. Langston, L.S., “Crossflows in a Turbine

Cascade Passage,” ASME Journal ofEngineering for Power, Vol. 102, No. 4,October, 1980, pp. 866-874.

2. Holmes, D.G. and Tong, S.S., “A Three-Dimensional Euler Solver forTurbomachiner y Blade Rows,” ASMEJournal of Engineering for Gas Turbinesand Power, Vol. 107, April, 1985, pp. 258-264.

3. Turner, M.G. and Jennions, I.K., “AnInvestigation of Turbulence Modeling inTransonic Fans Including a NovelImplementation of an Implicit k-eTurbulence Model,” ASME Paper 92-GT-308, 1992.

4. Turner, M.G., Liang, T., Beauchamp, P.P.and Jennions, I.K., “The Use of Orthogonal

Grids in Turbine CFD Computations,”ASME Paper 93-GT-38, 1993.

5. Holmes, D.G. and Warren, R.E., “DetailedStudies of Inviscid Secondary Flows,” GECRD Report No. 85CRD133, July, 1985.Presented at the 1985 ASME Winter AnnualMeeting, 11/18/85.

6. Connell, S.D., Holmes, D.G. and Braaten,M.E., “Adaptive Unstructured 2D Navier-Stokes Solutions on Mixed Quadrilateral/Triangular Meshes,” ASME Paper 93-GT-99,1993.

7. Connell, S.D. and Holmes, D.G., “A 3DUnstructured Adaptive Multigrid Schemefor the Euler Equations,” AIAA Paper 93-3339, 1993.

8. Braaten, M.E. and Connell, S.D., “A 3DUnstructured Adaptive Multigrid Schemefor the Navier-Stokes Equations,” GE CRDReport No. 94CRD146, August, 1994.

9. Rai, M. and Madavan, N., “Multi-AirfoilNavier-Stokes Simulations of Turbine Rotor-Stator Interaction,” Journal ofTurbomachinery, Vol. 112, July, 1990, pp.377-384.

10. Adamczyk, J.J., Celestino, M.L., Beach, T.A.and Barnett, M., “Simulation of Three-Dimensional Viscous Flow Within aMultistage Turbine,” ASME Paper 89-GT-152, 1989.

11. Dejch, M.E., et al., “Method of Raising theEfficiency of Turbine Stages with ShortBlades,” Teploenergetika, February, 1960,pp. 18-24.

12. O’Connor, M.F., Robbins, K.E. andWilliams, J.C., “Redesigned 26-Inch LastStage for Improved Turbine Reliability andEfficiency,” ASME Paper No. 84-JPGC-GT-17, 1984. (Also published as GE PowerGeneration Turbine Technology ReferenceLibrary Paper No. GER-3399, 1984.)

13. O’Connor, M.F., Williams, J.C., Dinh, C.V.,Ruggles, S.G. and Kellyhouse, W.W., “AnUpdate on Steam Turbine Redesigns forImproved Efficiency and Availability,” GEPower Generation Turbine TechnologyReference Library Paper No. GER-3577,1988.

14. Morson, A., “Steam Turbine Long BucketDevelopments,” GE Power GenerationTurbine Technology Reference LibraryPaper GER-3647, 1990.

15. Morson, A.M., Williams, J.C., Pedersen, J.R.and Ruggles, S.G., “Continuously Coupled40-Inch Titanium Last Stage BucketDevelopment,” GE Power Generation

32

GER-3713E

Turbine Technology Reference LibraryPaper No. GER-3590, 1988.

16. Moore, J.H., “High-Power-Density™ SteamTurbine Design Evolution,” GE PowerGeneration Turbine Technology ReferenceLibrary Paper No. GER-3804, 1994.

17. Sovran, G. and Klomp, E.D.,“Experimentally Determined OptimumGeometries for Rectilinear Diffusers withRectangular, Conical or Annular Cross-Section,” from Fluid Mechanics of InternalFlow, edited by G. Sovran, symposium heldat General Motors Research Laboratories inWarren, MI in 1967, published by ElsevierPublishing Co.

18. Sumner, W.J., Vogan, J.H. and Lindinger,R.J., “Reducing Solid Particle Damage inLarge Steam Turbines,” Proceedings of theAmerican Power Conference, Vol. 47, 1985,pp. 196-212. (Also published as GE PowerGeneration Turbine Technology ReferenceLibrary Paper No. GER-3478A, 1985.)

19. Shalvoy, R.S., et al, “An Improved Coatingfor the Protection of Steam TurbineBuckets from SPE,” presented at the EPRISteam Turbine/Generator Workshop, July20-23, 1993, Albany, NY.

20. Morrison, B.L., Booth, J.A. and Schofield,P., “Positive Pressure Variable ClearancePacking,” 1989 EPRI Heat RateImprovement Conference, Knoxville,Tenn.

21. Schofield, P., “Maintaining OptimumSteam Turbine-Generator ThermalPerformance,” Missouri Valley ElectricAssociation 1981 Engineering Conference,Kansas City, Mo.

22. Powell, D.J, Skolnick, M.M. and Tong, S.S.,“Engineous: A Unified Approach to DesignOptimization,” Applications of ArtificialIntelligence in Engineering, Vol. 1: Design,pp. 137-157, Computational MechanicsPublications, Southampton, UK, 1990.

23. Ashley, S., "Engineous Explores the DesignSpace," Mechanical Engineering, Vol. 114,No. 2, February 1992, pp. 49-52.

24. Lee, H., Goel, S. and Tong, S.S., “TowardModeling the Concurrent Design ofAircraft Engine Turbines,” ASME Paper 93-GT-193, 1993.

25. Goel, S., Gregory, B.A. and Cherry, D.G.,“Knowledge-Based System for thePreliminar y Aerodynamic Design ofAircraft Engine Turbines,” Society of Photo-optical Instrumentation Engineers,Applications of Artificial Intelligence,

Knowledge-Based Systems in Aerospace andIndustr y, April 13-15, 1993, Orlando,Florida.

26. Shelton, M.L., Gregory, B.A. and Lamson,S.H., “Optimization of a Transonic TurbineAirfoil Using Artificial Intelligence, CFDand Cascade Testing,” ASME Paper 93-GT-161, 1993.

27. Goel, S., Singh, H. and Cofer, IV, J.I.,“Turbine Airfoil Design Optimization,”ASME Paper 96-GT-158, 1996, presented atthe ASME Turbo Expo ‘96 conference inBirmingham, UK, June 11, 1996.

Note: Substantial portions of this paperappeared in ASME Paper 95-CTP-2, “Advancesin Steam Path Technology,” published in theJournal of Engineering for Gas Turbines andPower, Vol. 118, April 1996, pp. 337-352.

33

GER-3713E

© 1996 GE Company

LIST OF FIGURES

Figure 1. Typical HP turbine stage efficiency lossesFigure 2. Secondary flows in turbine nozzle cascadeFigure 3. Nozzle cascade wind tunnelFigure 4. Helium bubble traces in nozzle cascadeFigure 5. HP bucket — Viscous Euler calculation gridFigure 6. HP bucket — Viscous Euler Mach number contoursFigure 7. IP nozzle Viscous Euler results and comparison to test dataFigure 8. Comparison of structured and unstructured grids in CFD codesFigure 9. NOVAK3D refined calculation grid for transonic nozzleFigure 10. Multistage CFD analysis — vorticity at stage 3 nozzle and bucket exitFigure 11. Advanced Vortex HP nozzle and bucketFigure 12. HP stage nozzle exit angle and reaction distributionsFigure 13. Advanced Vortex IP nozzle and bucketFigure 14. IP stage nozzle exit angle and reaction distributionsFigure 15. Advanced Vortex LP nozzleFigure 16. Contoured sidewall test dataFigure 17. 3600 RPM family of continuously-coupled bucketsFigure 18. Convergent-divergent supersonic bucket tip profile designFigure 19. Continuously-coupled bucket tip designsFigure 20. 40-in. (1016 mm) titanium last stage bucketFigure 21. Comparison of single-flow 42-inch (1067 mm) and double-flow 26-inch (660 mm) unitsFigure 22. Bucket tip leakage controlsFigure 23. Slant tooth bucket tip seal test results — optimum tooth spacingFigure 24. Side entry last stage bucket cover design tip leakage controlFigure 25. CFD analysis of bucket tip seal passage — velocity vectorsFigure 26. Influence of tip seal reentry flow on stage exit flow angle -comparison of quasi-3D analysis test

dataFigure 27. Downward-flow exhaust hoodFigure 28. Downward-flow exhaust hood test modelFigure 29. Axial-flow exhaust hoodFigure 30. Solid model of downward-flow exhaust hood for CFD analysisFigure 31. Calculation grid for exhaust hood CFD analysisFigure 32. Velocity vectors in downward-flow exhaust hoodFigure 33. NOVAK3D analysis of axial-flow exhaust with “flow ribbons”Figure 34. NOVAK3D analysis of flow near struts in axial-flow exhaustFigure 35. NOVAK3D grid for bypass valveFigure 36. NOVAK3D analysis of bypass valve — comparison of calculated flow coefficient with test dataFigure 37. NOVAK3D grid for control valveFigure 38. NOVAK3D analysis of control valveFigure 39. Modified control stage to minimize SPE damageFigure 40. Control stage heat rate loss due to severe SPE damageFigure 41. Field performance of new SPE-resistant designFigure 42. First reheat stage suction surface damage caused by particle reboundingFigure 43. Severe erosion on suction side of first reheat stage diaphragmFigure 44. Automated process coats reheat diaphragmFigure 45. First reheat stage heat rate loss due to severe SPE damageFigure 46. SPE-resistant diaphragm after 3 1/2 years of service in the first reheat stage of a supercritical

reheat turbineFigure 47. SPE-resistant diaphragm after 3 1/2 years of service in the first stage of second reheat section

of a supercritical double reheat turbineFigure 48. GE positive pressure packingFigure 49. Heat rate improvement for positive pressure packing in 500 MW unitsFigure 50. Subsonic air turbine test facilityFigure 51. Cross-section of subsonic air turbine test facility

GER-3713E

Figure 52. Advanced Vortex test diaphragmFigure 53. Advanced Vortex “blisk” test wheelFigure 54. Laser tunnel and optics in subsonic air turbineFigure 55. Laser light plane flow visualizationFigure 56. Computer-generated model of Advanced Vortex diaphragm segmentFigure 57. Tip seal test rigFigure 58. Low speed research turbine test facility at GE Aircraft EnginesFigure 59. Cross-section of low speed research turbine test facilityFigure 60. Bezblade screenFigure 61. Engineous design optimization processFigure 62. TLP/TADS blade design processFigure 63. TADS/Engineous screenFigure 64. Results of TADS blade shape optimizationFigure 65. Solid model of typical bucket designed by TADS

GER-3713E

For further information, contact your GE Field SalesRepresentative or write to GE Power Systems Marketing

GE Power Systems

General Electric CompanyBuilding 2, Room 115BOne River RoadSchenectady, NY 12345

8/96

40

GER-3713E