基于神经网络的主动隔振控制技术研究

针对主动隔振系统存在非线性的情况,提出一种神经网络控制方法.控制器采用CMAC(cerebellar model articulation control)网络,参数调整基于梯度下降法.为消除系统次通道对控制器参数调整的影响,利用BP(back propagation)网络离线辨识得到次通道模型.系统的输出误差信号与次通道模型参数相结合,共同调整控制器参数.仿真结果表明,该控制方法对于存在非线性的主动隔振系统具有良好的控制效果,隔振能力超过常用的滤波LMS(least mean square)方法.

STUDY ON AN ACTIVE VIBRATION ISOLATION METHOD BASED ON NEURAL NETWORK

Aimed at the nonlinearities in active vibration isolation system(AVIS),an adaptive control method based on neural network was proposed.The CMAC(cerebellar model articulation control)neural network was used as controller,whose parameters were updated based on stochastic gradient descent algorithm.To reduce the influence of secondary path on controller parameters,a BP(back propagation)network was used to identify the secondary path offline.The observable error signal was combined with model parameters of the secondary path to jointly update controller parameters.Simulation results showed the proposed method has preferable effectiveness for AVIS with nonlinearities,and its capacity of vibration isolation is much better than that of FLMS(filtered least mean square).