| Aerodynamic optimization design is a new research field based on numerical optimizationmethods which combines nonlinear optimization theory and computational fluid simulation method.Adjoint method, as a kind of gradient algorithm, when applied to get gradient information, is notrelated with the number of design variables in computational cost, making the multi-variable andcontrol of complex surfaces of aerodynamic optimization design possible.For a typical large curvaturediffuser channel in turbine engine internal flow with complex flow, difficult to control the surface andhigh design sensitivity, a study was carried out by using adjoint method based on free-formdeformation techniques.Grid rebuilding is the most critical aspect in aerodynamic optimization design process.Free–form deformation (FFD) method was studied and program development was performed firstly. Griddeformation examples showed FFD is applicable for arbitrary topology and grid form, with largefreedom of surface control and can achieve local deformation, but in3D grid deformation, it istime-consuming to compute the local grid coordinates in volume based deformation, for this problem,a surface based grid deformation method was obtained, in which surface grid deformation using FFDmethod and the internal grid adjustment using radial basis function (RBF) method, this furtherimproved the grid deformation efficiency with90%in maximum case. This research found basisvolume points used as middle points of RBF can maintain the relative position of grids which ishelpful to maintaining grid quality and greedy method used to reducing the middle points can improvethe efficiency of the grid deformation with grid quality guaranteed.Based on the modular idea, the development of adjoint optimization system based on FLUENTsoftware with FFD techniques was accomplished, the FFD code above used as mesh deformationmodular which can output the file format of FLUENT, Flow and adjoint equation solving usingsecondary developed FLUENT solver, in this system, a variety of unconstrained gradient optimizationalgorithm was adopted with strong extensibility, FFD deformation control points used as designvariables, optimization target can be customized according to the need.Secondly, the typical largecurvature diffuser channel flow field computation methods were verified and the flow mechanism wasstudied, found SST k turbulence model used to predict the flow field best, which guarantee theaccuracy of optimization target.Then three kind of unconstrained gradient optimization algorithm with inexact line search method was compared, found Quasi-Newton method was of highest efficiency andhad lowest reliance on the accuracy of line search.This aerodynamic optimization design system was used in total pressure loss minimizationoptimization, compared to finite difference method (FDM), the gradients achieved by adjoint methodof this paper was larger, but the trend of each design variable gradients was same. Optimizationresults show that Quasi-Newton algorithm was of highest efficiency, channel total pressure loss isreduced by0.15%.Lastly, a three-dimensional S-shaped channel with large offset was optimized with target of totalpressure recovery coefficient, total pressure distortion index of channel outlet and multi-objective ofboth above. Total pressure loss was reduced by0.33%. DC60, DC90, and DC120were reduced by8.19%,11.17%and14.8%. Optimization results showed similarity that a dimple structure wasachieved before the separation region, which delayed the separation and suppressed the interaction ofsecondary flow and separation zone, the size of the streamwise vortex was reduced, thereby the totalpressure loss and ununiformity were reduced, this further proved the design capacity of aerodynamicoptimization design system developed by this paper when applied in large curvature diffuser channelsoptimization. |