应用神经网络技术建立了磁流变阻尼器的逆向模型,该模型含有4个输入神经元、1个输出神经元和15个隐层神经元.利用Bouc—Wen修正模型数值仿真生成数据,然后采用Levenberg—Marquardt法和OBS策略对逆向模型的结构进行训练和修剪.最后,将所建的磁流变阻尼器逆向模型应用于1/4车悬挂模型中,进行半主动控制的仿真分析.结果表明,所建立和优化的逆向模型可以较好地预测所需电流指令,应用于半主动控制中的效果明显.
An inverse model of magnetorheological damper with four input neurons, one output neurons and fifteen neurons in the hidden layer is built in this paper. Training and pruning of the model is done by a Levenberg-Marquardt method and Optimal Brain Surgeon method using data generated from the numerical simulation of the modified Bauc-Wen model. Finally, the MR damper inverse model is used in semi-active control simulation of 1/4 suspension model. The results show that the inverse model can predict needed current accurately and the performance of semi-active control is good.