基于前馈神经网络的磁流变阀参数优化及性能研究  

Parameter Optimization and Performance Analysis of Magnetorheological Valve Based on Feedforward Neural Network

作　　者：胡国良 方冰 杨肖 周锋 喻理梵 HU Guoliang;FANG Bing;YANG Xiao;ZHOU Feng;YU Lifan(School of Mechatronics and Vehicle Engineering,East China Jiaotong University,Nanchang,Jiangxi 330013,China)

机构地区：[1]华东交通大学机电与车辆工程学院,江西南昌330013

出　　处：《北京理工大学学报》2024年第12期1263-1276,共14页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金资助项目(52165004);江西省国际科技合作重点项目(20232BBH80010)。

摘　　要：提出一种径向流磁流变阀,阐述了其工作原理,同时分析了磁路.基于磁流变液本构模型推导阀的输出压降与各尺寸参数之间的数学关系式,建立线圈功耗、磁流变液响应和电磁回路响应的计算模型.在COMSOL仿真软件中进行多物理场仿真分析,得到磁流变阀内部磁场特性及磁性流体流动的动态历程.利用COMSOL软件FEM生成磁场数据,提出单隐层前馈神经网络(feedforward neural network,FNN)以预测磁流变阀磁感应强度的经验模型.采用实验设计(design of experiment,DoE)技术进行相关性分析,筛选出待优化的结构参数.搭建结构参数优化平台,开展基于FNN模型和多目标遗传算法(multi-objective genetic algorithm,MOGA)的多目标优化求解,得出最优结构参数集合,并对优化前后磁流变阀4个随电流变化的性能指标进行数值仿真.以磁流变阀压降和动态响应时间作为评价指标,搭建磁流变阀性能测试平台.结果表明,当施加电流为2.0 A时,优化前后磁流变阀压降提升了近2倍,压降可调系数提升24.40%,对应的上升响应时间效率提升了6.12%,下降响应时间效率提升了5.61%.In this paper,a radial flow magnetorheological(MR)valve was proposed,analyzing the working principle and magnetic circuit of the MR valve,deriving the mathematical relationship between the output pressure drop and the structural parameters of the MR valve based on MR fluid response basic mode,and establishing the calculation models for valve power consumption,MR fluid response and electromagnetic loop response.And a multi-physical field simulation analysis was carried out in COMSOL software to obtain the internal magnetic field characteristics and the dynamic process of magnetic fluid flow,generating magnetic field data based on the FEM model,and proposing an empirical model of feedforward neural network(FNN)to predict the magnetic flux density of MR valve.Adopting the design of experiment(DoE)technology for correlation analysis,two works were arranged,selecting the structural parameters to be optimized and developing an optimization platform for structural parameters.And then,the multi-objective optimization calculation was carried out based on the FNN model and multi-objective genetic algorithm(MOGA),and the optimal structural parameter set was obtained.Finally,four performance indexes of the initial and optimal MR valve with current change were simulated numerically.Taking the pressure drop and dynamic response time of the MR valve as the evaluation indicators,a test platform was setup to analyze the MR valve performance.The results show that when taking 2.0 A current,the pressure drop of the MR valve can increase up to nearly 2 times,the adjustable coefficient of pres-sure drop can increase by 24.40%,the corresponding rising response time efficiency can increase by 6.12%,and the falling response time efficiency can increase by 5.61%.

关 键 词：磁流变阀 前馈神经网络(FNN) 参数优化 压降 

分 类 号：TH134[机械工程—机械制造及自动化]