A Multi-Objective Optimization for a Centrifugal Fan Impeller

Abstract

A reduced-order flow analysis method is developed to predict impeller flow field and impeller performance characteristics for a high-specific-speed centrifugal fan attached with a double-discharge volute. The method is validated by comparing its performance prediction for a baseline fan with measurements. The flow analysis calculation not only predicts impeller flow field, but also generates impeller configurations based on given impeller design parameters. A multi-objective genetic algorithm is also developed and coupled with the reduced-order flow analysis to perform a direct online search of the nominal (deterministic) optimum impeller design envelope. An existing fan serving as the baseline fan was reverse engineered to regenerate its geometry using the current method. The optimization calculations were used to explore design variables including blade inlet and outlet angles, shroud angle and blade leading-edge span as 2, 3 and 4-variable design calculations. The 2-variable prediction concludes that the fan is limited in its performance improvement if only the blade shape is changed, which agrees with early CFD prediction results of the baseline fan. Comparing 3 and 4-variable predictions with the 2-variable predictions, larger ranges of improvement in fan performance are feasible by only considering fan aerodynamics. Furthermore the conflicting nature of design variables with regard to the fan performance improvement is also demonstrated by the prediction results.