基于神经网络的磁流变阻尼器逆向模型辨识研究

Study of Inverse Model Identification for a Magneto-Rheological Damper Based on Neural Network

磁流变阻尼器应用于振动控制系统时,阻尼器逆向模型的精度直接关系到整个振动控制系统的优劣。针对以上问题,首先,通过磁流变阻尼器的力学性能试验获得试验数据,如磁流变阻尼器的阻尼力、阻尼器两端的相对位移、相对速度和输入电流。然后,分别采用BP神经网络和RBF神经网络对磁流变阻尼器进行逆向模型辨识,并对逆向模型辨识的效果进行对比分析。结果表明:RBF建立的神经网络逆向模型能够更好地预测磁流变阻尼器所需的控制电流,应用于振动控制的效果更好。

When Magneto- Rheological （MR） damper is applied in a vibration control system, the precision of its inverse model directly influences control performances of the resulted closed - loop system. To deal with the prob- lem, mechanical properties of the MR damper will be firstly tested to obtain the experimental data such as the MR damping force, relative displacement, velocity of the piston rod, and the input current applied to the coil. Then, the inverse model identification for the MR damper is carried out by using BP neural network and RBF neural network, respectively. A comparative analysis of the model identification results is also presented. It is shown that the inverse model based on the RBF neural network can predict the required control current for MR damper precisely, and can be well applied in vibration control systems.