use of advanced cfd for the design of snecma moteurs turbomachine

24TH INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES

1

Abstract

During the last two decades, CFD tools havebeen a significant source of improvements in thedesign process of turbomachinery components,leading to higher performances, cost and cyclesavings as well as lower associated risks. Suchmethods are the backbone of compressor,combustion chamber and turbine designmethodology at Snecma Moteurs and their useis a key factor in designing components withhigh performance level. Moreover, the first testseries of components designed with thismethodology generally show performancesclose to the objectives that were assigned.

In the late 90’s new, modern and powerfulmethods have been developed, led by a constantprogress in computation power. As a matter offact, methods such as steady 3D Navier-Stokesmultistage solvers, unsteady 3D Navier-Stokesstage solvers, coupled heat transfer andcombustion solvers, and even Large EddySimulation are now commonly used in thedesign process. As an example, severalthousands of 3D Navier-Stokes computationsare done yearly at Snecma Moteurs for thedesign of low and high pressure compressor andturbine blades.

In addition to their modeling capabilities,the efficient use of such methods through thedesign process has been enabled by a closeintegration in a global methodology and anadequate exploitation environment. Theirvalidation against experimental data, theircalibration, the correlation between differentlevels of modeling are of critical importance toan actual improvement in design know-how.

In the present paper, the integration ofdifferent CFD methods in the design process isdescribed. Several examples of applicationillustrate their practical use. Some comparisonsbetween computational results and experimentaldata show their capabilities as well as theirpresent limitations. The prospects linked todevelopments, currently under way, arediscussed.

1 Introduction

The design of advanced turbomachinerycomponents has to meet ever more requirementsthat are demanding. Higher performance mustbe reached within shorter design cycles and atlower cost. Moreover, ambitious objectives ofweight, complexity and manufacturing costreduction are assigned. For compressor,combustion chamber and turbine designers, thisimplies the capability to control the complexflow phenomena occurring in all the operatingrange of the component, early in the program.Besides performances, the aggressive design ofadvanced components also requires an earlyfocus on all the aspects related to engineintegrity.

Up to the mid 70’s, most of the design andoptimization process relied on an empiricalapproach, which meant a large number of tests.The all-experimental optimization strategy wastime and cost consuming for two reasons atleast. Each iteration implies all the phases fromdesign to manufacturing, instrumentation andtesting. Secondly, determining what must beimproved in the design required heavyinstrumentation on each component, in order toidentify the “hard spots”. For these reasons,

USE OF ADVANCED CFD FOR THE DESIGN OFSNECMA MOTEURS TURBO-MACHINERY

N. Liamis, F. Bastin, J. LépineSnecma Moteurs

Centre de Villaroche, 77550 Moissy Cramayel, France

Keywords : CFD, Turbo-machinery, Compressor, Combustion Chamber, Turbine

N. LIAMIS, F. BASTIN, J. LÉPINE

2

Snecma Moteurs started very early to takeadvantage of the fast growing computer powerand of the concurrent advances in CFD. The useof CFD tools in the design methodology hasbrought major improvements in the iterativeoptimization process of each component. Anattempt to summarize this contribution coulduse the following keywords : faster response,broader range of alternative solutions, betterdescription of flow complexity. Indeed, everycomputation node in a numerical simulation isalso a “measurement” node, which allows aneasy and comprehensive analysis of the flowprediction.

But unfortunately, CFD is far fromfaithfully reproducing reality yet. Even with thepower of the most recent super computers,simulation capabilities still depend on models orare limited to component parts. Thecomputation of a complete 3D component withunsteady and viscous small-scale phenomena isstill out of reach for a long time. So the majorchallenge for both the engine componentdesigner and the CFD method developerconsists in integrating new computationalmethods, with their capabilities and limitations,in the design process in a fast, safe and efficientway. Every new method brings new answers,but also raises new questions. The most obviousrisks in using a new, more powerful tool areeither misunderstanding or overconfidence inthe results. Therefore, a constant effort must bededicated to methods of comparison, validationand calibration. This means in particular thatheavily instrumented rigs, representative of realengine components must be used to produce avalidation database.

At Snecma Moteurs, a strong interactionbetween component designers and CFD toolsdevelopers allows an early use of advancedmethods in the design process. Most of the CFDdevelopments have been undertaken in closecooperation with Research Laboratories andespecially with Onera.

This paper will present some currentapplications of advanced CFD tools forcompressor, combustion chamber and turbinedesign. Some comparison with experimentalresults will show the prediction capabilities of

different methods. Finally, conclusions will bedrawn on the current use of CFD tools in thedesign methodology.

2 CFD applications in compressor and fandesign

2.1 Objectives of modern compressors andfans design

During the last 3 decades, the introduction ofever-improving CFD tools in the methodologyof compressors design has led to animprovement of global performance as well as aconstant reduction of design time cycle. Thanksto the implementation of potential methods inthe 70’s, followed by Euler methods in the early80’s and then by stationary Navier-Stokesmethods since 90’s, the aerodynamic efficiencyof fans has been improved by about 10 % in 25years, as illustrated by figure 1. High and lowpressure compressors have known the samekind of evolution with another step in the late90’s with the introduction of steady multi-stageand unsteady rotor-stator simulation.

Figure 1 : Impact of CFD on the improvement ofefficiency of Snecma Moteurs civil fans.

Nevertheless, this race towards higheraerodynamic efficiency is now slowing and theobjectives of modern compressors and fans areslightly different than it used to be in lastdecades. In the case of fans, the next generationis most likely to have the same efficiency thanthe previous one because it is close to a

3

USE OF ADVANCED CFD FOR THE DESIGN OF SNECMA MOTEURS TURBO-MACHINERY

theoretical optimum. Thus, the goal fordesigners will be to maintain this level ofperformance while focusing on noise reductionand cost savings. For multi-stage compressors,the main targets are a better operating marginand lower costs. In both cases, reducing cost canbe translated into two main objectives. On theone hand the number of blades and stages has tobe reduced, which leads to highly loaded bladesand very complicated flows, and on the otherhand, the design time cycle has to be shorten,which requires a fast and reliable designprocess. The only realistic way to achieve allthese objectives is rely on an intensive andappropriate use of CFD [1].

2.2 Use of CFD in compressor and fandesign

Thanks to modern super-computers, simulatingthe flow in a whole compressor module of acivil aircraft engine is today something quiteeasy to achieve. A stationary solution of theflow in a multi-stage compressor can beobtained within few hours with a well optimizedcode used on a multi-processor computer. As anexample, figure 2 displays the results of a fan-booster-OGV set and of a high pressurecompressor simulations, obtained by using themulti-stage 3D Navier-Stokes solver CANARI,developed by ONERA and based on the mixingplane method. This kind of multi-stagecomputations is very helpful to optimize thematching of high pressure and low pressurecompressors, but it can also be very hazardous ifthe use of CFD is not perfectly under control.

Most of the computations that are done aresteady because they require a reasonable timeand because not much more information isneeded for the aerodynamic design of a blade.However, although they are very expensive interms of computer time, unsteady simulationsare very useful for acoustic and aero-elasticanalysis. Therefore, this kind of tool is usedonly for final analysis, but there is no doubt thatwithin few years it will be intensively used,being part of a whole integrated multi-disciplinary design process.

a) Flow field of a fan-OGV-booster configuration(static pressure)

b) Flow field of a HP compressor(static pressure)

Figure 2 : 3D multistage steady Navier-Stokescompressor simulations.

In the next 2 parts, we focus on aspects ofCFD that are of main importance forcompressor designers.

2.2.1 Mesh and turbulence modelThe boundary layer effect in a compressor

is of main importance, especially in the laststages of high pressure compressors, where it isof the same order of size than the annulus. At ofoff-design conditions (i.e. close to surge point),secondary flows, chock/boundary layerinteraction and stall dominate the whole flow.For these raisons, mesh refinement andturbulence models have a strong impact on thecomputation results. A bad choice in theseparameters can result to an over-estimation ofthe surge margin and a poor operability. This isillustrated by figure 3, displaying for the samestator at 10% span four computations with fourdifferent turbulence models. This simulation isclose to the surge point and a large stall occurs,which is well correlated by measurements. Theresults have been obtained with a structured gridof about 1 million points using similar codes :CANARI from ONERA and Turbo3D from


4

UPMC/LEMFI. In the case of algebraic(CANARI/Michel) and 1 transport equation(CANARI/Spalart-Allmaras) models, the stall isnot predicted because the model is unable topredict the boundary layer thickening, whereasin the case of 2 transport equations models(CANARI/k-l and Turbo3D/k-ε) the predictionis very satisfactory.

a) Algebraic model (Michel)

b) Spalart-Allmaras model

c) k-ε model

d) k-l model

Figure 3 : Importance of the turbulence model in off-design computations.

The same computation with a 2 transportequations model of turbulence on a lighter mesh(400 000 points) is also unable to predict thestall. In fact, it appears that the parameter thathas to be controlled is the size of the first mesh

cell. A good rule is generally to have a y+ onhub, casing and blade of about 1. However,some models require thinner mesh as, forexample, k-ε model that gives better results witha value of y+ about 0.3. Although this rule iseasy to apply for structured grids, the control ofthe mesh in the boundary layer becomes a trickyproblem for unstructured grid.

For some cases, 2 transport equationsturbulence models are not precise enough for anaccurate flow analysis. For example, to carryout an acoustic analysis from a computed flow,it is needed to take into account the non-isotropic character of the turbulence, which isnot possible with such turbulence models. Agood way to overcome this difficulty is to lead afull resolution of the Reynolds stress tensor.This is the solution chosen by LEMFI [2], [3]for the code Turbo3D, used at Snecma Moteursfor computing flows with a Reynolds StressModel (RSM).

2.2.2 Technology modelingA great challenge for the compressor simulationis to take into account the technology impact onthe flow. Indeed, tip clearance, bleeding, hub orshroud misalignment can have a major effect onthe flow and the compressor performance [4],[5], [6]. The small size and the complexgeometrical definition of these technologicalelements make them difficult to implement,whereas their influence can be of first order.

Tip clearance and bleeding can be analyzedwith quite simple modeling, requiring no majormodification of the grid topology and of thecode. Therefore, they are commonly taken intoaccount and their impact is well understood.

Other technological effects require eithergrid topology modifications by adding blocks,either an implementation of a more complexmethodology. Therefore, to tackle with thesekind of configurations, it is a good compromiseto compute the compressor once, including allthe technology elements, in order to extractrules or models applicable to simplersimulations. One application of this kind, ispresented on figure 4, where the cavity betweenthe fixed and rotating parts of the hub has beencomputed. With an analysis of this result, it is

5


possible to understand the phenomena takingplace in this complex flow area and then toimplement the cavity effect in a lighter way,using only boundary conditions.

Blade

Hub Cavity

Radial velocity

Figure 4 : Modeling a cavity in front of rotor.

3 CFD applications on combustion chamberdesign

3.1 The 3D Navier-Stokes platform forcombustion: N3S-Natur

To the aerodynamic challenges that have beenthe target of CFD for a long time, such asturbulence, mixing, swirling flows or cross-flows interactions, the design of a combustoradds the complexity of specific physicalphenomena : essentially two-phase flow physicsand combustion. The CFD available experiencefor these last aspects is more recent. Therefore,combustion design generally relies more onexperimental investigations and empiric rules ofdesign.

Snecma Moteurs has decided very early toinvest into CFD, by developing a strongcooperation with Onera, CNRS (LaboratoriesEM2C, CORIA, LCSR, LCD…) and Cerfacs. Itwas decided to build a 3D Navier-Stokes solvertargeted on combustion applications. A fewconstraints were highlighted. The computationalgrid had to be easily generated, and was to copewith the general geometrical complexity anddiversity of combustion systems (single annular,double annular, axially staged combustors,reheat systems…). The reheat system, and tosome extent the main combustion chamber areregulated downstream by a sonic throat (in the

first case the main nozzle, in the second case thehigh pressure turbine distributor), which couldbe represented. Industrial turnaround times werefinally expected on easily accessible computerresources.

These objectives led to the development ofN3S-Natur, which is a compressible, multi-species 3D Navier-Stokes solver, usingunstructured tetrahedral meshes, coupled with avariety of turbulence, two-phase flow andcombustion models. The code is parallelizedand runs on platforms ranging from vector orscalar supercomputers to PC clusters. The codewas co-developed by the french electric powercompany EDF, the car manufacturer Renault,and Snecma Moteurs, with scientific input fromInria, CNRS/LMFA, and Simulog. Thisassociation, which allows to reduce theinvestment costs for each partner, also turnedout to be an efficient way of technical "cross-fertilizing".

The core industrial application for N3S-Natur is the definition of the air distribution inthe dilution zone to optimize the temperaturefield at the exit of the combustor, and thereforeensure the safety of the high pressure turbine.Dilution air control is an essential factor foravoiding excessively hot spots seen by theturbine blades. The detail of the modeling levelfor theses simulations may be found in [7] or[8]. The basic components are a lagrangiansolver for the liquid fuel break-up andvaporization, a k-ε type turbulence model, and apresumed pdf (probability density function)turbulent combustion model. The combustionmodel is suited to non-premixed combustion,and based on infinitely fast chemistry, where thereaction rate source term is controlled bymixing, with a characteristic time provided bythe turbulence model. It is limited by chemicalequilibrium, which is precomputed andtabulated using the detailed (complex)chemistry of kerosene oxidation.

3.2 N3S-Natur: current and futureapplications

Grid construction is essential for combustionchamber simulation. Snecma Moteurs was


6

previously relying on a code using structuredmeshes. Mesh construction is actuallyunstructured, automated and coupled to acombustion chamber database, and takesbetween 15 minutes and two days for the mostcomplex injection system. An example ofsurface mesh is shown in figure 5, where specialcare is taken to provide enough grid pointsacross each air passage, without impacting thegrid across the whole chamber, as in the casewith structured grids.

Inlet bowl Dilution holes

Figure 5 : Unstructured surface mesh for a singleannular flame tube.

The code is continuously validated versusthe experimental database of Snecma Moteurscombustors. Figure 6 shows results obtained ona single annular combustor of currentgeneration. The exit temperature field ispredicted at a level allowing optimization anddesign choices. It must be noticed that suchresults are due not only to the code, but also tothe boundary condition definition methodology.Regarding the injection system, a complex mixof numerical simulation, experimentalcharacterization, and extrapolation from test todesign point conditions is still necessary.

T =>

r

RTF Simulation

RTF measurements

LTF simulation

LTF specification

T=>

r

T simulation -plane 1

T simulationplane 3

measured(plane 1)

measured(plane 3)

Figure 6 : Exit temperature profiles. The RTF is theradial temperature factor, where a spanwise average is

performed at each radial location. The LTF is the lateraltemperature profile, indicating the maximum

temperature at each radial location. Plane 1 is the radialplane including the injection system axis. Plane 3 is theintermediary radial plane, between injection systems.

As mentioned above, correlations may beestablished for classical single annularcombustors with constant cooling or injectiontechnological level, but fail to cover thediversity of solutions developed by thecombustion designer. CFD is therefore usefulfor concepts such as double annular or axiallystaged combustors, which are one of the ways toresolve contradictory emission and stabilityrequirements, and therefore obtain low emissionengines. An illustration of the use of N3S-Naturon a chamber design is given in figure 7 for anaxially staged combustor. CFD is used here tooptimize the exit temperature field underconstraint of maintaining the efficiency(limiting over-rich regions within the chamber).

7


Fuel mixture fractions

Figure 7 : Simulation of an axially staged combustor.

Models regarding combustion andturbulence aspects are continuously developedfor N3S-Natur, but the present developmentplan is focused on finite rate chemistry effects.Infinitely fast chemistry, used on the presentapproach, is suited to the temperature predictionat high engine operating points (cruise or take-off), and to some extent of NOx production. Butit is not appropriate to the simulation of CO, orto the behavior of the chamber far from thosepoints (stability analysis). These performancesare presently estimated via correlations andsimplified approaches. Current developmentsaim to take into account finite rate chemistryeffects within the Navier-Stokes simulationitself. Advanced transported pdf methods, whichproduced excellent result [9], are thereforeincorporated in N3S-Natur. The actualcomputation is still not affordable because thechemistry of kerosene oxidation involves a fewhundred chemical species and more of athousand chemical reactions. Therefore, reducedchemistries must be developed, posing theproblem of reducing a high-dimensionality realdynamical system to an much lower-dimensionality approached system. Techniquessuch as neuron networks are investigated forthis aspect. A real combustion chambercomputation including detailed chemistryeffects, which was not even considered a fewyears ago, is finally considered as closeperspective, with turn-around times a few timeshigher than present industrial simulations.

3.3 Large Eddy Simulation for combustion

Large Eddy Simulation (LES) is a full ofpromise technique, in which Snecma Moteursdecided to invest for combustion applications,for two main reasons. The first is the predictionof combustion instabilities, a complex couplingbetween combustion and pressure oscillationswithin the chamber. An essential component ofsuch a process is the flame response to upstreamperturbations. This response and more generallythe combustion dynamics are not seen byclassical RANS approaches, while LES hasshows promising results. The second motive isrelated to combustion modeling prospects.Current modeling is applied on a statisticallyaveraged aerodynamic field, provided by RANSsimulations, while the combustion process takesplace on a mixture of fuel and oxidizer that islocal in space and time, and is generallydifferent from the statistical average. LES, by itsability to capture large scale intermittency,simplify the combustion modeling problem ; theturbulence model "closure" problem is simpler.

LES being an advanced technique, thechoice was made to rely on the expertise ofCerfacs Research Center. The code developedby Cerfacs and relying on the parallel libraryCOUPL (maintained at Oxford University) isAVBP [10]. The basic philosophy was to designa code perfectly suited for compressible LES,and therefore to pay special attention to theproperties of the numerical scheme, dissipation,and boundary conditions. An efficientparallelism and the use of multi-elements grids,able to represent arbitrarily complex geometries,ensure industrial turn-around times.

An interesting application of LES wasdone on LPP (Lean Premixed Prevaporized)injection systems. These systems presentattractive properties in terms of NOx reduction,but they are also prone to combustioninstabilities, especially to flashback, where theflame moves upstream into the LPP, leaving itsintended stabilization location in the chamber.Successful LES of flashback were reported in[8] in 2D, but computations are now performedin 3D with a good agreement with the measuredbehavior of the module (figure 8).


8

Velocity and temperature

Figure 8 : 3D Large Eddy Simulation of a premixinginjection system.

An interesting quantitative information thatLES can produce is the flame response,represented by a delay time τ and a sensitivityindex n between a pressure or velocityperturbation and the heat release response.Comparison of such quantities with experimentsconducted by the CNRS laboratory EM2C haveshown how LES is able to produce quantitieswhich are critical to dynamic problems such ascombustion instabilities, and difficult toevaluate via RANS methods. The perspectivefor LES is taking into account two-phase flowsand complex chemistry. Work is going on andLES is expected to gain ground in the followingyears.

4 CFD applications on turbine design

4.1 Aerodynamic analysis

The first step of the turbine aerodynamic designconsists to the definition of its operating point.Two ways are used to model its behavior :through-flow simulations or 3D multi-stagesimulations. If the first method can be judgedsatisfactory during the pre-design phase whencycles are not well defined, the second one ispreferable during the design phase for itscapacity to provide more accurately the velocitytriangles and the coolant injection effects.

The steady multi-stage 3D Navier-Stokescalculations are carried out with the CANARIsolver, developed by Onera. They are based on

a mixing plane model and all the technologicaleffects (tip clearance, filet radii, groove) andcoolant injection regions (cooling holes, trailingedge slots, platforms leakage, high pressurecavity leakage) can be taken into account [11],[12]. A result of a multi-stage dual spoolconfiguration simulation (cooled high pressureturbine + low pressure turbine) including alltechnological effects is given on figure 9 a).

Turbulence and transition phenomena canmodify significantly the turbine behavior. Fordesign practices, the boundary layer of filmcooled blades can be considered as fullyturbulent, while for blades not cooled with filmbut protected by internal circuits or/and thermalbarrier cooling, transition modeling can behelpful to predict blade heat transfer coefficient.In all cases the turbulence and the transitionmodeling have an impact on the turbineperformances prediction as shown in figure 9 b)for a low pressure turbine.

a) Flow field of a HPT + LPT configuration(static pressure)

b) Influence of turbulence model and transition on LPTperformances

Figure 9 : 3D multistage steady Navier-Stokes turbinesimulations.

9


Moreover, unsteady effects due to potentialinteractions, turbulence, shocks and wakes onblade leading edge, as well as cold / hot spotimpact on blade profile are important in turbineenvironment. So, even if their prediction is notyet entirely part of the design process, unsteadyphenomena are today studied and quantified onturbine configurations. Two differentapproaches, based on the 3D Navier-Stokessolver CANARI, are used : unsteady stagecalculations and steady stage calculations usingthe deterministic stresses model, which takesinto account through modelisation, for a givenrow, the effects of the neighboring rows [13].Figure 10 a) and b) presents, for an uncooledhigh pressure turbine stage, the results obtainedby a steady stage simulation using the mixingplane model, an unsteady stage simulation and asteady stage simulation using the deterministicstresses model, compared to measurements.

a) NGV 50% span isentropic Mach number

b) Rotor blade 50% span static pressure

Figure 10 : Influence of stage modeling for highpressure turbine simulations.

4.2 Aerothermal analysis

The aerothermal design of turbine blades is ofprimary importance for the high pressureturbines, where the inlet temperature, severalhundreds of Kelvin higher than the maximumtemperature admissible by the material, impliesthe use of cooling techniques. It is closelylinked with the aerodynamic design and themechanical design. Moreover, thismultidisciplinary iterative design has to takeinto account the manufacture constrains, duringeach project step. Several computationalprocedures are used for the design and analysisof turbine aerothermal phenomena : the first onefor the investigation of turbine external flows,the second one for internal flows in bladecooling cavities and the third one for theprediction of blade metal temperature field.

4.2.1 External aerothermal analysisThe critical environment, generated by the highinlet turbine temperature, has led to the use ofcooling flows, in order to increase blade life. Inparticular, film cooling is now a commontechnique used to prevent the blades fromoverheating and the prediction of itsperformance is a major issue in view of theblade efficiency and life. For a givenaerodynamical flow field, the number, thelocation, the pitch, the shape, the streamwiseand compound injection angles of holes andtrailing edge slots are important geometricalparameters for the design of a cooled blade.

3D Navier-Stokes calculations of filmcooling is a challenge, because the phenomenainvolved are very complex, due to the veryvortical flows and to their length scales, whichare very different from the common lengthscales of the flow around the blades. As for theaerodynamic analysis, the Navier-Stokessimulations for the film cooling prediction [14],[15] are performed with the CANARI solver,developed by Onera. Some results, concerninggrooved tip, leading and trailing edge cooling,are presented on figure 11 a), b) and c).


10

a) blade tip cooling(Mach number)

b) leading edge cooling(wall heat flux)

c) trailing edge cooling(Mach number)

Figure 11 : External aerothermal analysis of turbineblade.

4.2.2 Internal aerothermal analysisModern turbine airfoils are designed withinternal cooling passages, arranged with radialor serpentine patterns, through which thecoolant air travels to remove heat from theairfoil walls, in order to ensure blade life. Theinternal cooling performances of turbine bladesare highly dependant on the technologies used :ribs, pedestals, pins fins, impingement, bladeinternal circuits. An important part of thedesigner's work is to optimize the cooling flowsystems encountered inside the blade,minimizing the amount of cooling mass flowand optimizing the cooling efficiency. Thus, 3DNavier-Stokes calculations are carried out toanalyze, in details, flow and heat transfer onnew complex technological concept [16].

The Navier-Stokes simulations, performedwith the MSD solver developed by Onera,provide results close to experimental data, ascan been shown by the comparison for the Ubend duct case presented on figure 12 a) and b).So, their use, for the prediction of the flow field

into internal cooling circuits, is increasinglyencouraged. Actually, the whole internalcooling circuit can be considered, as shown onfigure 12 c) for a high pressure turbine rotorblade internal cooling circuit containing holes,ribs and pins fins.

a) experiment U bend duct case(wall heat flux)

b) calculation U bend duct case(wall heat flux)

c) rotor blade internal circuit(Mach number)

Figure 12 : Internal aerothermal analysis of turbineblade.

4.2.3 Thermal analysisThe prediction of the blade metal temperature isthe final achievement of the aerothermal design.Due to the fluid / solid interaction, the globalheat transfer analysis on blades is one of themost challenging works for turbine designers.

Generally, the Navier-Stokes simulationsand the conduction calculations are usedindependently, a manual iteration model beingused between the fluid and the solid media. Themethodology adopted is to prescribe the coupleheat transfer coefficient / reference temperaturefrom the Navier-Stokes code to the conductionsolver, and the wall temperature from the

11


conduction solver to the Navier-Stokes code. Aresult of the iterative (uncoupled) thermalanalysis process, provided by the commercialconduction solver ABAQUS, for a high pressureturbine rotor blade protected by thermal barriercoating, is presented on figure 13 a).

However, the manual iterationmethodology is time consuming and can befaulty in general 3D flows. Moreover, resultsdepend on the number of manual iterations doneby the designers. Therefore, to prevent theseproblems, an automatic general couplingprocedure between the Navier-Stokes solversand the conduction solver ABAQUS has beendeveloped [17]. A result of a conjugate heattransfer calculation, carried out on the NASAC3X cooled blade with internal cavities, ispresented on the figure 13 b).

a) uncoupled heat transfer(metal temperature)

b) conjugate heat transfer(static and metal temperature)

Figure 13 : Thermal analysis of turbine blade.

5. Conclusion

The role of CFD in the design process of aero-engine components is fast growing. Numericalsimulation significantly contributes to reachingindustrial objectives such as the reduction ofcycle time and cost in a project development. It

is constantly demonstrating its effectiveness onseveral applications, allowing an earlier andmore reliable component optimization.

The advances in computer technologiesand the combined development of CFDtechniques increasingly enlarge the field ofpossible applications. In this paper, a broadspectrum of examples in aero-engine componentdesign and analysis shows the capabilities oftoday’s CFD methods.

However, the increasing importance ofCFD requires special care to turn numericalmethods into integrated tools easy to handle bycomponent designers. Validation and calibrationare essential to an effective use of advanced, butstill limited, simulation techniques. Well-instrumented experimental vehicles aredefinitely needed for this purpose.

The integration of advanced CFD tools inthe design methodology of aero-enginecomponents have marked an importantmilestone and have led to significant changes inthe overall design process. The development ofnew applications makes the near future verypromising. The next generation of computerswill enable a new step in componentoptimization. However, much has to be done innumerical techniques and modeling of flowphysics, in order to turn this opportunity into anindustrial achievement.

Acknowledgements

The authors wish to thank Snecma Moteurs forthe authorization to publish the present paper.The results previously presented are issued fromseveral works, partly supported by the frenchorganizations DPAC and DGA. It must beemphasized that these results have beenobtained by many co-workers at SnecmaMoteurs, during the past years. Their help inreviewing and shaping up the manuscript is alsogratefully acknowledged.

References[1] J.F. Escuret, D. Nicoud, P. Veysseyre, Recent

advances in compressor aerodynamic design and


12

analysis, AGARD RTO Lecture series 211, RTO-EN-1, September 1998.

[2] G.A. Gerolymos and I. Vallet, Tip-clearance andsecondary flows in a transonic compressor rotor,ASME 98-GT-366.

[3] G.A. Gerolymos and I. Vallet, Wall-normal-freeReynolds-Stress Model for compressible rotatingflows applied to turbomachinery, AIAA Journal, vol.40, 2002.

[4] S.R. Wellborn, I. Tolchinsky and T.H. Okiishi,Modeling shrouded stator cavity flows in axial-flowcompressors, ASME 99-GT-75.

[5] A.A.J. Demargne and J.P. Longley, Cavity andprotrusion effects in a single-stage compressor,ASME 2001-GT-433.

[6] J.D. Denton, Lessons learned from Rotor 37,presented at the 3rd International Symposium onExperimental and ComputationalAerothermodynamics of Internal Flows (ISAIF),Beijing, China, September 1-6, 1996.

[7] F. Ravet, L. Vervisch, Modelling non-premixedturbulent combustion in aeronautical engines usingPDF generator, AIAA Paper 98-1027.

[8] O. Mahias, F. Bastin, F. Ravet, C. François, Leanpremixed combustor emissions performancemodeling using 3D CFD codes, AIAA Paper 2000-3199.

[9] J.C Larroya, C. François, M. Cazalens, L. Vervisch,Testing a new Monte Carlo method for solving pdfequation in turbulent combustion, Fifth InternationalConference on Technologies and Combustion for aclean environment, vol I, pp 177-183, 1999.

[10] T. Schönfeld, M.A. Rudgyard, Steady and unsteadyflow simulations using the hybrid flow solver AVBP.AIAA Journal, 37(11), November 1999, pp 1378-1385.

[11] Liamis, N., Duboué, JM., CFD Analysis of HighPressure Turbines, ASME Paper 98-GT-453.

[12] Liamis, N., Brisset, C., Lacorre, F., Duboué JM.,CFD Analysis of Dual Spool Turbine Configuration,AIAA Paper 99-2525.

[13] Kirtley, KR., Turner MG., Saeidi S., An AveragedPassage Closure Model for General Meshes, ASMEPaper 99-GT-77.

[14] Fougères, J.-M. & Heider, R. (1994) : Three-Dimensional Navier-Stokes Prediction of HeatTransfer with Film Cooling. ASME Paper 94-GT-14.

[15] Ginibre, P., Lefebvre, M., Liamis N., NumericalInvestigation of Heat Transfer on Film CooledTurbine Blades, Heat Transfer in Gas TurbineSystems, Annals of the New York Academy ofSciences, vol 934, pp 377-384, 2001.

[16] Paté, L., Duboué, JM., CFD : A Tool to Design JetEngine Internal Cooling Systems, AIAA Paper 98-3563.

[17] Montenay, A., Paté, L., Duboué, JM., Conjugate HeatTranfer Analysis of an Engine Internal Cavity,ASME Paper 2000-GT-282.

use of advanced cfd for the design of snecma moteurs turbomachine

Documents