Research On Aerodynamic Configuration Design Methods For High Performance Aircraft

The aerodynamic design technique based on Computational Fluid Dynamics (CFD)simulation has become the main research direction of the modern design aerodynamic.Withthe diversified aircraft design requirements and improved aircraft performance standards,aircraft configuration optimization methods are supposed to be more sophisticated andefficient. Therefore, researches including aerodynamic simulation technique, parameterizationmethod, grid deformation technique, surrogate model and optimization algorithms should beconducted to build an efficient and reliable aerodynamic optimization system.The main research issues and achievements in this paper are as follows:1. The RANS (Reynolds Averaged Navier-Stokes) equations simulation platform is foundedwith Roe spatial discretization scheme and Menter’s k-SST turbulence model. Thereliability of the platform is verified through numerical simulations of2D transonic flowsaround an airfoil and3D flows around complex geometries, providing trustworthyaerodynamic data for object characteristic analysis of aerodynamic optimizations. Besides, aboundary layer transition simulation platform is developed based on local Re ttransitionmodel. It provides reliable aerodynamic analyze tools for aircraft configuration optimizations.2. To meet the requirement of refined aerodynamic design, researches of parameterizations fordifferent kinds of aerodynamic shapes are conducted. For airfoil parameterization, a CSTmethod based on dispersion is developed from class/shape function (CST) method. Themethod features improved capability of refined geometric description and manipulation,ensuring the design space with limited design variables. For complicated3-dimensionalshapes, an arbitrary space-shape free form deformation (FFD) method is built. This methodhas a good arbitrary deforming description capability, whereas lacking the ability of accurategeometric control for refined design of complex3D geometries. Therefore, a directmanipulating FFD method is established, improving the capability of accurate geometriccontrol for complex3D geometries. A Delaunay graph based grid deformation method isestablished with comparison to the deformation method based on cubic spline and infiniteinterpolation technique on their ability handling large scale spinning and twistingdeformations. A properly background grid reformation is proceeded to avoid grid crossing problems of the raw background grid, making the grid capable of bearing larger scaledeformations. Therefore the deformation method is enhanced with more robust and powerfulautomatic grid deformation capability.3. To improve the prediction accuracy of surrogate model under limited samples, an ExpectedImprovement (EI) criterion based self-adapting Latin Hypercube method is established. Themethod takes fully advantage of the model surface information, with respect to the samplespace’s density and nonlinearity, thus providing more reasonable sampling distributions. TheKriging surrogate based on the established sample space is applied to take the place of timeconsuming N-S equation simulations to improve the searching efficiency. Coping with thelocal convergence problem of surrogate optimization, the surrogate is modified withself-adapting capability to improve global convergence.4. Coping with the problems that standard PSO tends to premature and lack of variety, threealgorithms are proposed by modifying the PSO parameters, population topologies andboundary restrains. The first one is BreedPSO with reference from Genetic Algorithm. Thesecond one is Select Dynamic Neighborhood PSO (SDNPSO) inspired from nature selectionbased comprehensive learning strategies. The last one is the Multi-swarm Collaboration PSO(MCPSO) method by introducing multi-swarm collaboration strategy. Their convergences andvarieties are verified by functional and aerodynamic optimizations, indicating their optimumsearching capabilities and efficiencies. For multi-object optimization problems, a multi-objectPSO based on crowding distance sorting is established by elitism strategies utilizing externalarchives from crowding distance sorting. The proposed algorithm can converge at the Paretofront edge at lower computation complexity, with good improvement on the variety anddistribution of solutions.5. Coping with problems that the traditional model is sensitive to disturbances from the flyingcircumstances, design methods based on responding surface model and6model are bothestablished. Robust design of nature laminar flow airfoil with uncertainties of lift coefficientand Mach number is carried out, improving the stability of laminar flow region and dragcoefficient. For supercritical airfoils, a single point design is conducted, improving the dragdivergence Mach number and lowering the mean value and variance of the drag coefficientwithin uncertain area. With the problem of lift coefficient’s over-sensitivity against attackangle for the single point design, a multi-point design is carried out, with improvement on both the drag divergence performances and stall performances.6. Coping with the low accuracy of traditional Gappy POD(Proper Orthogonal Decomposition)invert design method, an invert design technique utilizing redefined sampling scheme basedon characteristic normal decomposition POD method is established. A modified Gappy PODmethod is proposed by filling missing data on order-reduced models. The redefined samplingsolution structure can improve the accuracy of data filling. The missing objective airfoil datais restructured by the surface pressure distribution of known airfoils. The objective pressuredistribution can be utilized for highly accurate airfoil invert design.7. Based on the aforementioned optimization tools, an aerodynamic design platform withsingle/multiple objective optimization capability and an aerodynamic shape invert designscheme are established, providing corresponding method for different optimization object anddifferent requirements. For subsonic laminar flow airfoils, robust design against liftcoefficient and Mach number uncertainties is carried out, improving the drag divergenceperformances and the stability of laminar region. For laminar supercritical airfoils, an invertdesign is carried out, with two different pressure distributions proposed. The influences thatthe length of favorable region, pressure gradient of favorable region and pressure recoverlocation cast on the transition location are analyzed, providing good concepts for the highperformance natural laminar flow airfoil design. For transonic natural laminar flowwing-body robust designs, favorable region length and shock wave location are consideredcomprehensively, ensuring the robustness of shock wave development under largest laminarflow length. For DLR-F6wing-body-pylon-nacelle, the interference drag is diminished byoptimizing the position of nacelle.