|
|
|
| Optimization design of centrifugal pump impeller based on dynamic RBF surrogate model and NSGA-Ⅱ genetic algorithm |
| ZHANG Renhui1,2*, LIU Feng1, CHEN Xuebing1, LI Rennian1,2 |
| 1. School of Energy and Power Engineering, Lanzhou University of Technology, Lanzhou, Gansu 730050, China; 2. Key Laboratory of Fluid Machinery and Systems, Gansu Province, Lanzhou, Gansu 730050, China |
|
|
|
| Guide |
|
|
Abstract In the multi-objective optimization design of traditional centrifugal pumps, the prediction accuracy of the surrogate model will gradually decrease as the Pareto frontier advances. In order to improve the effect of multi-objective optimization results of centrifugal pump impeller, the optimization method of centrifugal pump based on the dynamic RBF surrogate model and the NSGA-Ⅱ algorithm was proposed. Some optimal samples from the generated Pareto frontier solution were added to the RBF sample set and the RBF surrogate model was retrained and reconstructed. The objective function value of each sample of offspring samples were predicted by using the dynamic surrogate model. The MH48-12.5 centrifugal pump was selected as the research object, and the blade inlet angle, blade outlet angle and blade wrap angle were selected as the optimization variables. Latin Hypercube Sampling(LHS)was used to construct the initial sample space of the surrogate model, and multi-objective optimization analysis was carried out with optimization objectives of head and efficiency. The results show that the Pareto frontier obtained by the multi-objective optimization method of dynamic RBF surrogate model is much better than that obtained by the static surrogate model method. The Pareto front points obtained by the static surrogate model method are all dominated by that of the dynamic model method. The prediction accuracy of Pareto front solutions for the dynamic surrogate model is higher than that of static surrogate model. The optimal maximum head obtained by the dynamic surrogate model is 2.86% higher than the original design head and 1.03% higher than the static model. The optimal maximum efficiency obtained by the dynamic surrogate model is 4.36% higher than the original design efficiency and 1.32% higher than the static model.
|
|
Received: 08 September 2020
|
| Fund: |
|
|
|
|
|
|
