基于多目标遗传算法和反向传播神经网络的调节阀流道结构优化  

Optimization of control valve flow channel structure based on MOGA and BPNN

作　　者：吕家皓 吴欣[1] 何磊 LV Jiahao;WU Xin;HE Lei(College of Mechanical Engineering,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学机械工程学院,浙江杭州310018

出　　处：《机电工程》2023年第12期1880-1888,共9页Journal of Mechanical & Electrical Engineering

基　　金：国家重点研发计划重大专项(2022YFB3401900)。

摘　　要：以往的研究中,只针对调节阀迷宫流道结构和内部流场特性进行了分析,但对迷宫流道抗空化性能和流通性能的优化设计较欠缺。为了满足阀门实际工程中的设计需求,迷宫式调节阀需要具有流道抗空化性能和流通性能。为此,提出了一种基于多目标遗传算法(MOGA)和反向传播神经网络(BPNN)的方法,对调节阀迷宫流道进行了结构优化,提高了迷宫流道的抗空化性能和流通性能。首先,基于对冲耗能原理和多级降压原理,设计了弧形对冲式迷宫流道,并建立了流体力学仿真计算的数学模型;然后,利用计算流体动力学(CFD)仿真软件,对模型进行了空化仿真,根据仿真的数据构建了BPNN代理模型,通过结合Sobol敏感度分析方法与代理模型,分析了迷宫流道各参数对仿真结果的影响,采用多目标遗传算法,优化了迷宫流道的结构;最后,搭建了实验测试平台,测量了迷宫流道的阻塞流曲线,对比分析了测试结果与仿真结果。研究结果表明:采用优化算法得到的迷宫流道最大流量由0.0876 kg/s提高到0.1174 kg/s,提高了34%;线性压差由762.163 kPa提高到811.280 kPa,提高了6%;优化的迷宫流道实际最大流量为0.1159 kg/s,提高了33%;线性压差为819 kPa,提高了7%。迷宫流道抗空化性能和流通性能同时得到了提高,证明了仿真的有效性和该方法的可行性。In previous studies,only analysis was conducted on the structure and internal flow field characteristics of the control valve labyrinth channel,but the optimization design of the anti-cavitation performance and flow performance of the labyrinth channel was relatively lacking.In order to meet the design requirements of the valve in practical engineering,the labyrinth type regulating valve needs to have channel anti-cavitation performance and flow performance.Therefore,a method based on multi-objective genetic algorithm(MOGA)and back propagation neural network(BPNN)was proposed to optimize the structure of the regulating valve labyrinth flow channel,and improve its anti-cavitation performance and flow performance.Firstly,based on the principle of hedging energy dissipation and the principle of multi-stage bucking,the arc-shaped hedging maze flow channel was designed,and a mathematical model for fluid dynamics simulation calculation was established.Then,the model was simulated using computational fluid dynamics(CFD)simulation software.And based on the simulated data,the BPNN surrogate model was constructed.The Sobol sensitivity analysis method was combined with the surrogate model.The influence of the parameters of the maze flow channel on the simulation results was analyzed.The structure of the flow channel was optimized by multi-objective genetic algorithm.Finally,a test platform was built to measure the blocking flow curve of the flow channel.The results show that the maximum flow rate of the maze flow channel obtained by the optimization algorithm is increased from 0.0876 kg/s to 0.1174 kg/s,which is increased by 34%.And the linear pressure difference is increased from 762.163 kPa to 811.280 kPa,which is increased by 6%.The actual maximum flow rate of the optimized maze flow channel is 0.1159 kg/s,which is increased by 33%,and the linear differential pressure is 819 kPa,which is increased by 7%.The cavitation resistance and flow performance of the maze runner are improved at the same time,which proves the effect

关 键 词：液压控制阀 迷宫流道 抗空化性能 流通性能 反向传播神经网络 多目标遗传算法 计算流体动力学 

分 类 号：TH137.5[机械工程—机械制造及自动化]