Optimization of a peristaltic blood pump using PSO-BPNN and NSGA-III: balancing pumping efficiency and hemocompatibility

Peristaltic pumps are promising for wearable dialysis, but optimizing flow and hemocompatibility under roller-tube contact is limited by costly 3D two-way fluid-structure interaction (FSI). We propose a surrogate-assisted three-objective workflow that couples 3D two-way FSI, a PSO-BPNN surrogate, NSGA-III search, and entropy-weighted TOPSIS decision-making. Sixty samples were used to train the surrogate for average flow rate fave, time-averaged shear stress τave, and shear-stress pulsation τp, achieving R2 values of 0.97/0.91/0.94. Compared with the baseline pump, the selected design delivers 208.79 mL/min, reduces τave and τp by 6.30% and 40.76%, lowers the predicted hemolysis index by 6.71%.