Introducing Non-Hierarchical RSM and MIGA for Performance Prediction and Optimization of a Centrifugal Pump under the Nominal Condition

In order to improve the operation performance of multi-stage double-suction centrifugal pump and reduce internal energy loss pump, this paper proposes a single-objective optimization design method based on non-hierarchical response surface methodology (RSM) multi-island genetic algorithm (MIGA). Nine parameters, such as blade outlet width wrap angle, were used variables, objective was efficiency under conditions. total, 149 sets valid data obtained Latin hypercube sampling (LHS), corresponding thresholds set for head, 99 obtained. A cross-validation analysis sieved carried out RSM, global using MIGA, numerical verification via CFD. The research results show that compared with hierarchical RSM can approximate nonlinear relationship between function variables higher accuracy, model fitting R2 value 0.919. improved by 3.717% after optimization. overall prewhirl impeller inlet decreased, speed volute significantly improved, large-area vortex at pipe eliminated, impact separating tongue disappeared, hydraulic improved. total entropy output optimized reduced 4.79 (W/K), mainly concentrated reduction in double volute, dissipation reduced.