Integrated optimization of a high-lift low-pressure turbine cascade based on dynamic support vector regression

Abstract

With the increase of the lift of turbine components, the performance of the turbine is significantly affected by the secondary flow. The combined optimization of the root modification and the profiled endwall (PEW) is investigated for more efficient secondary flow control in the endwall region of a high-lift low-pressure turbine cascade. The NURBS surface is used in this paper to perform parametric modelling. Support vector machines (SVM) and particle swarm optimization (PSO) are used to complete the optimal design of the combination of cascade root and endwall. To ensure the accuracy of the regression model, dynamic parameter adjustment and dynamic point addition strategies are introduced in the optimization process to update the regression model. The study found that when the separate profiled endwall was optimized for the cascade, the total pressure loss of the cascade decreased by 10.39%; when the cascade was optimized by the combined shape of the cascade root and the endwall, the total pressure loss of the cascade decreased by 12.88%. The detailed flow field analysis shows that both the profiled endwall and integrated optimization can effectively weaken the cross-passage pressure gradient in the endwall region of the cascade passage. Compared with only profiled endwall, the integrated optimization restrains the passage vortex and the suction side leg of horseshoe vortex from developing towards the middle of the cascade, and the secondary flow loss in the endwall region is obviously reduced.