An Optimization Technique for Radial Compressor Impellers

Abstract

A software tool has been created to aid in automated impeller design within an integrated design system for radial flow impellers. The design tool takes the results from the 1D preliminary design process and uses these to define a parameterized blade geometry, which incorporates features that are required for low mechanical stresses and simple manufacturing. This geometry is then adjusted to minimize a global objective function using a throughflow computation. The adjustment is based on selection with a breeder genetic algorithm. The initial population includes “elite” designs from a database of earlier well-proven experience, and the final design is honed to perfection with a hill-climbing method. With the help of a suitable global objective function incorporating mechanical and aerodynamic criteria, and taking into account wide experience with the design of impellers, the tool provides a fast screening of various design possibilities to produce a geometrical input for more advanced computational fluid dynamic and mechanical analysis. This is demonstrated through the redesign of an impeller previously designed by conventional methods. Comparisons of the results of the CFD analysis of the new impeller with that of the earlier design demonstrate that the tool can rapidly produce nearly optimal designs as an excellent basis for further refinement by the more complex analysis methods.