Multi-objective performance optimization of a solid–liquid two-phase annular jet pump

To enhance the performance of a solid–liquid two-phase annular jet pump, efficiency and energy consumption are selected as objective functions. Using the design of experiments method, area ratio, annular nozzle length, and diffuser angle are identified as variables. Optimal Latin hypercube sampling is employed to generate samples, and multi-objective optimization is performed using a particle swarm optimization-back propagation neural network combined with the non-dominated sorting genetic algorithm II. The results indicate that the established model demonstrates good fitting and generalization capabilities, achieving the expected predictive outcomes. After structural optimization, the overall length of the model is reduced. In terms of performance, efficiency increases 3.1%, and energy consumption reduces 34.3%. The area and distribution of low-pressure zones within the pump are reduced, resulting in smoother pressure changes. The entropy generation region extends along the flow direction to the outlet, but the entropy generation rate decreases, significantly reducing energy losses within the main flow channel. The total number of particle collisions decreases, significantly improving the wear characteristics of the pump internal walls.