多级离心泵叶轮和蜗壳协同优化研究  

Matching optimization of impeller and volute in a multistage centrifugal pump

作　　者：赵建涛 裴吉[1] 袁建平[1] 王文杰[1,2] ZHAO Jiantao;PEI Ji;YUAN Jianping;WANG Wenjie(National Research Center of Pumps,Jiangsu University,Zhenjiang 212013,China;Key Laboratory of Fluid and Power Machinery(Xihua University),Ministry of Education,Chengdu 610039,China)

机构地区：[1]江苏大学国家水泵及系统工程技术研究中心,江苏镇江212013 [2]流体及动力机械教育部重点实验室(西华大学),四川成都610039

出　　处：《哈尔滨工程大学学报》2024年第9期1670-1678,共9页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(51879121);江苏省自然科学基金项目(BK20190851);流体及动力机械教育部重点实验室(西华大学)开放课题(szjj2019-007).

摘　　要：为了解决多级离心泵高效运行区窄,整体能效偏低的问题,本文对比转速为64.3的双进口多级离心泵叶轮和蜗壳进行了优化设计研究。对比了不同代理模型在多级离心泵水力性能优化中的适用性,选择了GA-BP神经网络作为最优代理模型,以叶轮和蜗壳的9个主要设计参数作为优化变量,0.6Qd和1.0Qd工况泵的效率为优化目标,通过拉丁超立方抽样方法和自动数值分析程序构建样本库,使用NSGA-II算法获得多目标优化问题的帕累托最优解,并根据实际工程需求选取了适当的参数组合。结果表明:模型泵在小流量工况和设计工况点效率分别提高了2.49%和3.09%,大流量工况扬程陡降问题得到缓解。该方法可以为多级离心泵的正向设计提供参考。To solve the problem of the narrow high-efficiency operation zone and low overall energy efficiency of a multistage centrifugal pump,an optimal design study of the impeller and volute of a multistage double-suction cen-trifugal pump with a specific speed of 64 was conducted.The applicability of different surrogate models in the opti-mization of the hydraulic performance of the multistage centrifugal pump was investigated and compared,and a GA-BP neural network was selected as the optimal surrogate model.Nine main design parameters were chosen as opti-mization variables,and the efficiency of the pump under partial-load and nominal conditions 0.6Qd and 1.0Qd were set as the optimization objectives.The Pareto-optimal solution of the multiobjective optimization problem was ob-tained using the NSGA-II algorithm with the Latin hypercubic sampling method and an automatic numerical analysis program to construct a sample database,and appropriate combinations of parameters were selected according to ac-tual engineering requirements.The analysis results showed that the efficiency of the model pump was increased by 2.49%and 3.09%under partial-load and nominal conditions,respectively,and the issue of the steep drop in the head under overload conditions was alleviated.This method can be a reference for the positive design of multistage centrifugal pumps.

关 键 词：多级离心泵 匹配优化 能效优化 多目标优化 反向传播神经网络 遗传算法 数值模拟 代理模型 

分 类 号：TK311[动力工程及工程热物理—热能工程]