Aero-Mechanical Multidisciplinary Optimization and Knowledge Discovery of High Pressure Ratio Centrifugal Impeller

Integrating data mining techniques, RANS solver technique, finite element method (FEM), self-adaptive multi-objective differential evolution (SMODE) algorithm and B-spline surface parameterization method, a multidisciplinary and multi-objective design optimization strategy for centrifugal impeller is proposed to obtain the optimal properties in given design space on an optimization platform. This strategy includes modules optimization algorithm, parameteriz-ation method, constraint handling and multidisciplinary assessment. The first three modules are established by self-coding program, and the last module is implemented by the software NUMECA and ANSYS. Data mining techniques, like total variation analysis, are introduced to discover the significant variables in design space, the design optimization is carried out for maximizing isentropic efficiency and minimizing the maximum stress in impeller surface of a typical high pressure ratio centrifugal impeller SRV2-O. After optimization, the efficiency increases by 2.0% and the maximum stress lowers by 2.08%. Knowledge discovery of design space is carried out, and the effects of design variables on aerodynamic and mechanical performances of impeller are also analyzed. It is indicated that for centrifugal impeller with similar high pressure ratio, the shroud of diffuser section ought to be mainly controlled in design to weaken tip leakage flow, and flow passage area ought to be reduced to decrease leakage flow loss.