Knowledge Mining of Low Specific Speed Centrifugal Pump Impeller Based on Proper Orthogonal Decomposition Method

To clarify the complex relation between the pump blade shape and its corresponding hydraulic performance, the knowledge mining method of centrifugal pump impeller based on proper orthogonal decomposition (POD) was proposed. The pump blade shape was parameterized by cubic Bezier curve. The Latin hypercube design method was employed to supply the necessary samples for producing the perturbations of blade wrap angle, and blade angle at inlet and outlet. The hydraulic efficiency and head were optimized by NSGA-II and RBF hybrid algorithm. The Pareto-optimal solutions were obtained. In order to further illustrate the relationship between the centrifugal pump blade shape and its hydraulic performance, the POD method was used to discover the effects of optimized blade shape to the flow solutions. For the optimization of centrifugal pump MH48-12.5, blade shape and relative velocity field in impeller from Pareto-optimal solutions were analyzed. The results demonstrate that larger blade angle and smaller wrap angle increase the average kinetic energy in impeller, resulting in higher pump head design. Smaller blade angle and larger wrap angle decrease the velocity gradient from the pressure side to suction side, resulting in smaller hydraulic loss and higher efficiency design.