Surrogate-Based Shape Optimization of the ERCOFTAC Centrifugal Pump Impeller

Abstract

Centrifugal pumps are largely used in several fields and for different applications. Despite their wide diffusion, they are often not optimized for working at the design conditions. The aim of this paper is to investigate the potentialities offered by surrogate-based optimization techniques in centrifugal pump impeller shape optimization, to obtain a robust and fast algorithm for performance improvement. The geometry chosen for validating the proposed method is the ERCOFTAC centrifugal pump where accurate measurements and simulations are available in the literature. The three-dimensional geometry of the impeller is parametrized by means of parametric Bezier surfaces with an in-house Scilab script, which allows to export the dictionary used by the utility blockMesh to create the mesh for the CFD simulation. The surrogate-based optimization method here described maximizes the pump hydraulic efficiency, while keeping the total pressure rise prescribed to the design condition, in order to find the optimal impeller design. The whole optimization chain is designed for running in HPC environment with open-source software, i.e. OpenFOAM for CFD simulation, Dakota for the optimization and Scilab for the geometry parametrization.