Application of Automated Optimisation to Pump Impeller Design

Dominating factors in today’s pump design process are reducing time to market and increasing demands on product flexibility. With these requirements, an adaptation of the pump design process is crucial and the need for an optimisation environment is only one part. In this paper we present our semi-automated impeller design process and its integration into a fully automated optimisation environment. The setup of the workflow for both approaches including impeller design tool, mesh generation, CFD analysis, and the steering tool are described. The optimisation process starting from a set of basic designs (Designs of Experiments, DOE) over the proper optimisation to the design analysis is explained by means of a suction impeller with opposing objectives like maximising hydraulic efficiency and minimising NPSH_3% values yielding different geometry solutions all being a compromise. During this process, performance values for duty point, partload and overload are evaluated by CFD. An overall efficiency parameter is defined representing the averaged efficiency of these three operating points. Suction performance can be handled differently and evaluated only for different operating points and treated either as objectives or constraints. The approach developed in this paper is finally applied for two test cases. The first example describes the design of a suction impeller which has to fit into an existing casing and limits us to the modification of the blade shape only. Objectives are overall and duty point efficiency, NPSH_3% at duty point is handled as a constraint. The overall efficiency objective requires a high plateau-like efficiency from partload to overload operating condition. The classical way of an optimisation is starting with a DOE, followed by an optimisation, visualisation of the Pareto Frontier and selection of the final design. The second example requires the replacement of a suction impeller which is supposed to operate cavitation free at site conditions. Main objectives are the reduction of NPSH_Inc and increase of partload efficiency while maintaining NPSH_3% and overall efficiency. Limitations in development time forces us to go a different way in the design process. We start with a general exploration of the parameter range and study the parameter influence on the objectives and complete the process with an optimisation for a very limited number of blade parameters to obtain the objectives.