Multi-Objective Optimization of Deep-Sea Mining Pump Based on CFD, GABP Neural Network and NSGA-III Algorithm

In order to improve the hydraulic performance of a deep-sea mining pump, this research proposed multi-objective optimization strategy based on computational fluid dynamics (CFD) numerical simulation, genetic algorithm back propagation (GABP) neural network, and non-dominated sorting algorithm-III (NSGA-III). Significance analysis impeller diffuser parameters was conducted using Plackett–Burman experiment filter out design variables. The optimum Latin hypercube sampling method used produce sixty sample cases. GABP network then utilized establish an approximate model between pump’s Finally, NSGA-III solve approximation determine for diffuser. results demonstrate that can accurately forecast performance, global is effective. On rated clear water conditions, optimized pump has 14.65% decrease in shaft power 6.04% increase efficiency while still meeting requirements head. Under solid-liquid two-phase flow head meets requirements, decreased by 15.64%, increased 6.00%.