基于神经网络的磁流变减振器力学模型研究

Research on Mechanical Model of Magnetorheological Damper Based on Neural Network

针对磁流变效应的非线性特性导致的磁流变减振器(MRD)力学模型难以精确描述的问题,设计了一种基于神经网络的磁流变减振器力学模型.利用径向基(RBF)神经网络较强的模拟非线性函数的能力,建立了减振器力学模型,并依据减振器的测试数据信息确定了RBF神经网络各节点的数量;利用遗传算法的全局优化能力和多参数并行优化能力,辨识神经网络的参数.通过RBF神经网络力学模型计算得到的阻尼力与试验数据的相对误差的平均值为2.41%,能够满足模型的实用需求.

A new mechanical model of magnetorheological damper（MRD）based on neural network was designed in considering of strong non-linear trait of magnetorheological effect,which made it hard to describe MRD mechanical model precisely.The radial basis（RBF）neural network had the ability of imitating non-linear functions,and it could be used to establish the mechanical model of the damper.On the basis of damper testing data,the nodes number of the RBF neural network was determined.Genetic algorithm was used to identify the parameters of the neural network for its ability of global optimization and multi-parameter optimization.The damping force was calculated by a mechanical model of RBF neural network.The average value of the computational and test data relative error is 2.41%,and it can meet the demand of applicability.