The investigation on compressor's end wall and suction surface profiling using the surrogate model assisted optimization design method

The design of non-axisymmetric end wall shapes today primarily relies on optimization methods based on surrogate models. In recent years, this approach has increasingly faced the “curse of dimensionality” due to the large number of design parameters. The key to addressing this issue lies in the choice of shaping methods. To explore more effective modeling techniques, this paper employs two different shaping methods: the lofted surface method and the surface-mapping method, to optimize the modeling of the end wall and blade. This study is based on a high-load blade cascade, where a multi-objective optimization design is conducted using the same optimization method. The study finds that the surface-mapping method exhibits a higher convergence rate during the optimization process and achieves better flow control under different incidences. Therefore, the comparison in this paper demonstrates that the surface-mapping method is more effective in addressing the “curse of dimensionality” problem.