Multi Objective Aerodynamic Optimization Design and Data Mining of a High Pressure Ratio Centrifugal Impeller

Abstract

An automated threedimensional multiobjective optimization and data mining method is presented by integrating a selfadaptive multiobjective differential evolution algorithm (SMODE), 3D parameterization method for blade profile and meridional channel, Reynoldsaveraged Navier–Stokes (RANS) solver technique and data mining technique of selforganizing map (SOM). Using this method, redesign of a high pressure ratio centrifugal impeller is conducted. After optimization, 16 optimal Pareto solutions are obtained. Detailed aerodynamic analysis indicates that the aerodynamic performance of the optimal Pareto solutions is greatly improved. By SOMbased data mining on optimized solutions, the interactions among objective functions and significant design variables are analyzed. The mechanism behind parameter interactions is also analyzed by comparing the data mining results with the performance of typical designs.