An optimization process is employed to improve the performance of an industrial radial flow pump impeller. A hybrid optimization scheme is coupled to a cost effective potential flow solver that computes the flow through the blade channel. The impeller geometry is parameterized to reproduce variations over a wide design space and the objective function is evaluated at each iteration to account for the performance of each candidate blade. The concept of the hybrid approach is to employ a global stochastic optimization method for the diversification of the design space and a deterministic local method for efficiently intensifying the search towards the optimum. The trend of computer industry to multi-core processors is a promising platform for population based optimization methods and even on a quad core the timescales needed to solve the problem with the proposed methodology are reasonable. Results present an optimised impeller with improved performance but also a clear trade-off among contradictory design objectives.

This content is only available via PDF.
You do not currently have access to this content.