Multi-objective optimization and design rule mining for an aerodynamically efficient and stable centrifugal impeller with a vaned diffuser

Abstract

This article presents a combined use of multi-objective optimization and quantitative design rule mining methods to improve the aerodynamic efficiency and stability of a centrifugal impeller with a vaned diffuser. A time-averaged but spatially distributed flow is considered at the mixing plane to evaluate the flow uniformity, which affects aerodynamic stability. First, the impeller's shape has been optimized using a multi-objective genetic algorithm to improve efficiency and flow uniformity. It was found that the trade-off among non-dominated solutions can be controlled by the vane-less diffuser's dimensions and the aerodynamic load distribution. A compromise solution has been experimentally shown to improve both objectives. Second, decision tree analysis and rough set theory have been applied to extract design rules for improving each objective. Although the design rules derived from both methods are consistent with regard to the main effects of design variables, some differences are found regarding the interaction effects.