Академический Документы
Профессиональный Документы
Культура Документы
FLOWCOMPRESSORS
AravindGP1,NileshPSalunke2,SalimA.Channiwala3
1,2,3,
SardarVallabhbhaiNationalInstituteofTechnology,
aredeterminedusingtheinformationofthe D. OptimizationofBaseCDAusingGA[11]
pointsineachsection.Thesubscriptkchanges Total pressure loss as objective function for
from1to2inordertoconsiderthelengthofthe optimization since it is more significant in the
upperandlowersurfaces,respectively. compressor blade efficiency. And the
optimization is carried out for compressor
C. BezierPARSECParameterization[4] cascade at high subsonic velocities. The
DerksenandRogalsky[4]haveintroducedthe optimization is meant for finding a profile
BezierPARSECparameterization.Thisapproach section with minimal loss for the compressor
willusetheadvantagesofboththeBezierand blade. In this investigation we selected a CDA
PARSEC parameterization and avoid the cascade, third stage of a compressor for the
disadvantages of both to represent the airfoil optimization. Before starting the optimization
andprovideenoughflexibilityovergeometrical processweusedtoanalyzethebasecascadeto
andaerodynamicparameters.Theirapproachis predicttheperformance.Theanalysisiscarried
out numerically in CFD softwares, Gambit for
further subdivided into two parameterization
methods viz. BP3333 and BP3434. In both the modeling and meshingandFluentforanalysis.
methods,Beziercontrolpointsaredetermined The optimization of cascade has mainly five
intermsofthePARSECparametersofanairfoil. stepsasshownontheoptimizationflowchart.
ThecamberthicknessformulationoftheBezier AlltheprocessiscarriedoutusingMatlabcode.
curvesismoredirectlyrelatedtotheflowthan The design parameters are selected from the
istheuppercurvelowercurveformulationfor parametersobtainedfromtheBP334.Thisnew
PARSEC,whilethePARSECparametersaremore parameters are generated at each iteration by
aerodynamically oriented than the Bezier the GA based on the constraints and the
objectivefunction.Weselected15parameters
parameters. The BP parameterization uses the
PARSEC variablesas parameters,whichinturn ofBP3434foroptimizingthecascade.Thefirst
definefourseparateBeziercurves.Thesecurves step is terminated with the generation of the
describetheleadingandtrailingportionsofthe new parameters by GA. The next step is to
camberline,andtheleadingandtrailingportion generate the airfoil section from these
ofthethicknessdistributions.WhiletheBezier parameters. In the third step CFD software
parameterization joins the leading and trailing GambitiscalledinMatlabinbatchmodeforthe
curveswithfirstordercontinuity. cascade modeling and meshing using reading
The BP parameterization uses secondorder the Gambit journal file in Matlab. After the
completionandgenerationofthemeshfileasa
continuity.Theparametersare:
Leadingedgeradiusrle, fourthstepFluentiscalledinMatlabusingthe
Trailingcamberlineangleate, systemcommandandreadstheFluentjournal
Trailingwedgeanglebte, file, which includes all the commands for the
Trailingedgeverticaldisplacementzte, analysis.BytheexecutionoftheFluentwewill
Leadingedgedirectiongle, getalltheinletandoutletparameterssuchas
Locationofthecambercrestxc,yc, total pressures, total temperatures, static
Curvatureofthecambercrestkc, pressures,Machnumbers,etc.Alsowewillget
Positionofthethicknesscrestxt,yt, theflowparametersoverthecascadei.e.Mach
Curvatureofthethicknesscrestkt, number, static pressure, etc. The objective
thehalfthicknessofthetrailingedgedzte,and function is selected as total pressure loss
severalBeziervariables,b0,b2,b8,b15andb17. coefficient for this optimization which
This type of parameterization improves the calculatedfromtheresultsofFluentanalysis.At
robustness and convergence speed for each iteration GA checks the value of the loss
aerodynamicoptimization,whichmakesitmore coefficient for the next generation of next
suitable for optimization using Genetic population of parameters. The process ends
algorithms. whenthelosscoefficientisminimized.Genetic
Algorithm (GA) is used as an optimization
algorithm because of its global optimization
natureandspeedofconvergence.Theobjective
function used for GA is total pressure loss Modified turbulent viscosity: Second order
coefficient.WeselectedtheconstraintasChord upwind
lengthoftheCascadeandisfixedas46.46mm InletTotalPressurePO1=338000Pa
andsetthenumberofgenerationas100witha InletTotalTemperatureTO1=426K
crossover fraction of 0.8. After calculating and
Boundaryconditions:
checking the value of loss coefficient GA
generates the new population based on the Table2:BoundaryConditionsforHigherPR
crossover, selection and mutation with a [16][17]
constraintfixedchordlength. Thetable3showsthetotalpressureloss
After the convergence of the optimization
Pre Inlet
algorithmforagenerationof100weobtained ssu Mach
the airfoil section which has minimized the P1 T02 P2
re No.
objectivefunction.Thetable1showsthenewly (Pa) (K) (Pa)
Rat
generated profile has optimal total pressure io
losscoefficientcomparedtothebaseprofile. 1.5 0.75 232737.6 478.3231 349106.5
1.6 0.75 232737.6 487.225 372380.2
Table1:ComparisonofBaseandOptimized 1.7 0.75 232737.6 495.7379 395654
Airfoil[11] 1.8 0.75 232737.6 503.9003 418927.8
1.9 0.75 232737.6 511.7449 442201.5
Airfoil Total Pressure 2.0 0.75 232737.6 519.2998 465475.3
Sections pressure loss 2.2 1.2 139383.4 533.6354 306643.62
Inlet Outle coefficient 2.4 1.4 106213.4 547.0681 254912.62
t coefficientobtainedupto2.4pressureratios
BaseAirfoil 3380 3351 0.0427
0 50 Table3:Pressurelosscoefficientcomparison
Optimized 3380 3353 0.0394 upto2.4PR[17]
Airfoil 0 70 Pressure CDABaseAirfoil OptimizedAirfoils
Thisprocesswasrepeatedforvariouspressure Ratio PressureLoss PressureLoss
ratiosrangingfrom1.1tohigherpressureratios Coefficient Coefficient
1.5 0.007903
anditwasfoundthattherewasadragreversal
1.6 0.006527
afterapressureratioof1.4.Thenegativedrag
1.7 0.009531
indicatesthereversalofflowhencewederived
1.8 0.005681
aconclusionthatapressureratiogreaterthan
1.9 0.004332
1.4 cannot be achieved from the above airfoil 2.0 0.017324
for the given set of conditions and there is a 2.2 0.008038
need to optimize the airfoil further to gain 2.4 0.015610
higherpressureratios.
E. OptimizationofCDAforHigherPressure
Ratiosupto2.4[16][17]
The following boundary conditions were
applied:
Solver: Green Gauss node based, 2d, steady,
implicit,densitybased
Model:SpalartAllmarus
ConvergenceCriteria:0.001
Fluid:AirwithidealgasdensityandSutherland
viscosity
Discretization: Flow: Second order
upwind
IV. SIMULATIONRESULTSOFHIGHER
PRESSURERATIOSMORETHAN2.4
Furtheroptimizationoftheairfoilanduptohow
muchpressureratiowillbepossibleisfoundout
inthiswork.Weobtainedapressureratioof3.0
withoutanyflowseparationforaMachnumber
of 1.4. Beyond that further optimization is not
possiblewiththismethod.
Table4:BoundaryconditionsforHigherPRup
to3.1
Inlet
Pressur P1 T02 P2
Mac
eRatio (Pa) (K) (Pa) Thenewlyoptimizedairfoilscanperformbetter
hNo
1.4 106213 559.72 276154
2.6 Figure 3: Mach number pot for optimized 3.0 PR airfoil
.4 33 .9
1.4 106213 571.70 297397 at higher pressure ratios up to 3.0. The
2.8 optimized blades have shown perfect velocity
.4 11 .6
1.4 106213 583.08 318640 and pressure distribution as of CDA. In the
3.0 aboveplotswecanseetheexitMachnumberis
.4 24 .2
Theboundaryconditionsforpressureratiosup reducingasthepressureratioincreases.
to3.0isasshownintable4.
Table5:PressureLossCoefficientComparison
DistributionofMachnumberofOptimized
forPRupto3.0
airfoilsaregivenbelow.
Pressure CDABase Optimized
Ratio Airfoil Airfoils
PressureLoss PressureLoss
Coefficient Coefficient
2.6 0.006009
2.8 0.001717
3.0 0.007102
V. VALIDATIONOFSIMULATIONRESULTS
The optimized airfoil showed close CDA
characteristicswhichconfirmitsgoodbehaviour
athigherpressureratios.Thesuctionpeakislow
at higher pressure ratios as expected and
Figure 1: Mach number plot for optimized 2.6 PR airfoil
uniform diffusion is there till the trailingedge.
The pressure on the lower surface increases
uniformly till trailing edge. It is also observed
that as the pressure ratio increases the peak
Mach number decreases.For given pressure
ratios,onuppersurface,theMachnumberfirst
increasestoavalueofpeakMachnumberand
thereafter it reduces continuously. On lower
surface, Mach number first reduces and then
increasesgradually.Itsignifiesthatoverupper
surface, fluid is accelerated first and then it
Figure 2: Mach number plot for optimized 2.8 PR airfoil decelerates constantly to match the flow
conditionsatthetrailingedge.Theseallarethe
typical characteristics of a controlled diffusion
airfoil.Henceouroptimizedairfoilsexhibitthe inEngineeringSoftware41(2010)923
characteristicsofaCDAairfoil. 930.
5. Athar Kharal, Ayman Saleem, Neural
VI. CONCLUSION networks based airfoil generation for a
The parameterization and GA given Cp using BezierPARSEC
optimization method is capable of Parameterization, Aerospace Science
findingefficientandoptimumairfoilsin andTechnology23(2012)330344.
fewernumberofgenerations. 6. Pierluigi Della Vecchia, Elia Daniele,
ThedevelopmentofacombinedBezier Egidio DAmato, An airfoil shape
PARSEC(BP)parameterizationutilizethe optimizationtechniquecouplingPARSEC
advantages of both the Bezier and parameterization and evolutionary
PARSECparameterizations. algorithm, Aerospace Science and
Coupling of BezierPARSEC Technology32(2014)103110.
parameterization with GA and CFD 7. Lars Sommer, Dieter Bestle, Curvature
together, offers an optimal cascade driven twodimensional multiobjective
profilewithareasonabletotalpressure optimization of compressor blade
losscoefficientreductionwithefficient sections, Aerospace Science and
flowpatternoverthecascade. Technology15(2011)334342.
ThebaseCDAairfoilcanoffermaximum 8. Yongsheng Lian, Akira Oyama, Meng
pressure ratio of 1.4, beyond which a Sing Liou, Progress in design
converged solution is not obtained optimization using evolutionary
indicating that it cannot gain pressure algorithms for aerodynamic problems,
ratioshigherthan1.4. Progress in Aerospace Sciences 46
The blade optimization with Bezier (2010)199223
PARSEC Parameterization has offered 9. Sergey Peigin, Boris Epstein, Robust
most optimized results. The newly optimizationof2Dairfoilsdrivenbyfull
optimizedbladescanperformbetterat NavierStokes computations,
higher pressure ratios up to 3.0. The Computers & Fluids 33 (2004) 1175
optimized blades have shown perfect 1200.
velocity and pressure distribution as of 10. Naixing Chen, Hongwu Zhang, Yanji Xu,
CDA.Upto3.0PRandMach1.4wecan Weiguang Huang, Blade
usetheseoptimizedairfoilswithoutany Parameterization and Aerodynamic
flowseparation. Design Optimization for a 3D Transonic
Compressor Rotor, Proceedings of the
REFERENCES 8th International Symposium on
Experimental and Computational
1. Naixing Chen., Aerothermodynamics of Aerothermodynamics of Internal Flows
TurbomachineryAnalysisandDesign,1st Lyon,July2007,Paperreference:ISAIF8
Edition, John Wiley & 0021
Sons,Singapore,2010,chap.14. 11. Syam,ChanniwalaSA,Optimizationof
2. Saravanamuttoo H.I.H, Rogers G. F. C, CDAcascadeusingParameterizationand
Henry Cohen., Gas Turbine Theory, 6th Genetic Algorithm coupled with CFD,
Edition,PrenticeHall,USA,2008.chap.5 2nd International Conference on
3. Ava Shahrokhi, Alireza Jahangirian, Mechanical,AutomotiveandAerospace
Airfoil shape parameterization for Engineering,ICMAAE2013.
optimum NavierStokes design with 12. T Sonoda, Y Yamaguchi, T Arima, M
genetic algorithm, Aerospace science Olhofer, B Senghoff and H A Schreiber,
andtechnology,11(2007)443450 Advanced High Turning Compressor
4. R.W. Derksen, Tim Rogalsky, Bezier Airfoils for Low Reynolds Number
PARSEC: An optimized aerofoil Conditions, Part 1: Design and
parameterization for design, Advances