Shape optimization of a centrifugal compressor impeller

The multi-objective optimization algorithms become more suitable for general aerodynamic design problems, despite the number of objectives evaluated by CFD calculations. This paper presents an aerodynamic shape optimization (ASO) algorithm based on the idea of replacing some of the CFD evaluations by those provided by an approximation model known as "Radial Basis Functions" RBF. As a matter of fact that this latter requires high number of evaluations to perform an accurate approximation, herein is proposed a combination of the "Multidisciplinary Aero/Struc Shape Optimization Using Deformation" MASSOUD and the Bspline techniques based on the idea of parameterizing the shape perturbations instead of the shape itself. Ansys-CFX flow solver is used to evaluate the objectives for each element of the database. Based on MASSOUD and Bspline techniques, and the RBF used to approximate the objectives with respect to design variables, the NSGA-II algorithm is applied to produce the Pareto front. This procedure was applied for the NASA low speed centrifugal compressor LSCC to optimize its performance within a reasonable computational time.