Performance optimization of liquid ring pumps based on Gappy POD surrogate model

Abstract

To solve the problems of high calculation cost and difficult optimization caused by the complex gas–liquid two-phase flow in liquid ring pumps, an optimization method of a liquid ring pump impeller based on the proper orthogonal decomposition (POD) surrogate model was proposed. The impeller blade shape was parameterized by using the quartic Bezier curve. The sample was obtained from the uniform experimental design of blade shape design parameters. The POD surrogate model was constructed by the parameters of sample blade shape and its corresponding flow field data, which was used instead of CFD numerical simulation to predict the flow field variables under the perturbations of blade design parameters in the optimization process. And then, the gradient vector of the objective function to design variables was predicted quickly and accurately. The blade shape was updated continuously along the negative direction of the gradient vector, and then, the optimal design of the impeller was acquired. The calculation case results show that the relative error between the pressure field in the impeller predicted by the POD method and that of CFD simulation is less than 5%, and the calculation cost is only 1/8460 of CFD simulation. Compared with the original model, the flow field distribution in the optimized model is significantly improved, and the efficiency is promoted by 3.8%.