基于NARX神经网络的磁流变阻尼器模型研究

为了能够较精确地拟合磁流变阻尼器在低速区的非线性及滞回特性,提出一种基于NARX神经网络的磁流变阻尼器动力学模型。以力学性能实验数据为前提,利用LMBP算法训练NARX神经网络的串-并行结构,然后将串-并行结构转化为并行结构,使得NARX神经网络模型能够在不同电流值下得到与实验结果相吻合的输出值。最后将实验结果和模型结果进行对比,通过比较两者所得的结果发现NARX神经网络模型的计算结果与实验结果的最大相对误差仅为3.77%,而且该模型能够表征磁流变阻尼器在低速区的非线性及滞回特性,证明NARX神经网络模型在处理磁流变阻尼器的非线性及滞回特性力学行为具有较高的拟合精度。

Research on Magnetorheological Damper Model Based on NARX Neural Network

In order to more accurately fit the non-linear and hysteretic characteristics of the magnetorheological dampers in the low-speed region,a dynamic model of magnetorheological dampers based on NARX neural network was proposed.Based on the experimental data of mechanical properties,the LMBP algorithm was used to train the serial-parallel structure of the NARX neural network,and then the serial-parallel structure was converted into a parallel structure,so that the NARX neural network model can obtain output values consistent with the experimental results at different current values.Finally,the experimental results were compared with the model results,the comparison results show that the maximum relative error between the calculated results of the NARX neural network model and the experimental results is only 3.77%,and the model can be characterized nonlinearity and hysteresis characteristic of the magnetorheological dampers in the low speed region,which proves that the NARX neural network model has a high fitting accuracy in handling the nonlinear and hysteresis characteristic of magnetorheological dampers.