基于RBF神经网络与DECIMO算法的螺旋离心泵多目标优化设计

针对螺旋离心泵运行时性能普遍偏低的问题,以某一典型螺旋离心泵为研究对象,应用计算流体动力学软件CFX对螺旋离心泵进行数值计算得到其内部流动规律.以在设计流量工况下的扬程和效率作为优化目标,采用P-B试验与多因素方差分析筛选出优化变量,采用径向基(RBF)神经网络建立优化目标与优化变量之间的预测模型,并结合差分进化(DECIMO)算法在样本空间内全局寻优.取扬程最优、效率最优和初始个体进行数值计算,对比分析泵输送不同介质(清水与固液两相流体)时的流场及其外特性差异,并进行试验验证.研究结果表明:叶片轮毂进口角β_1b、叶轮出口宽度b₂、叶轮出口直径D₂和叶片包角φ是影响螺旋离心泵扬程和效率的显著因素;由RBF神经网络建立的预测模型精度较高;输送清水时,设计流量下扬程最优个体扬程为9.4 m,增长了13.5%;效率最优个体效率比初始个体提高了9.8%,优化效果显著.

Multi-objective optimization design of screw centrifugal pump based on RBF neural network and differential evolution algorithm

In order to solve the problem of low performance of the spiral centrifugal pump, a typical spiral centrifugal pump was taken as the research object, and the flow field in spiral centrifugal pump was solved by using computational fluid dynamics software CFX.Taking the head and efficiency under the design flow condition as the optimization objective, the optimization variables were screened out by P-B test and multivariate analysis of variance.The radial basis function(RBF)neural network was used to establish the prediction model between the optimization objective and the optimization variables, and the differential evolution(DECIMO)algorithm was used to find the global optimization in the sample space.The optimal head, the optimal efficiency and the initial individual were calculated numerically.The flow field and its characteristics of different medium(clean water and solid-liquid two-phase fluid)were compared and analyzed, and the experimental verification was carried out.The results show that the hub inlet angle β_1b, the impeller outlet width b₂, the impeller outlet diameter D₂ and the blade envelope angle φ are the significant factors affecting the head and efficiency of spiral centrifugal pump. The prediction model established by RBF neural network has high precision. When conveying clean water, the optimal individual head under the design flow is 9.4 m, increasing by 13.5%. The efficiency of the optimal indivi-dual is 9.8% higher than that of the initial individual, and the optimization effect is significant.