Investigation Of Active Vibration Control Methods Based On Neural Networks

As a hotspot of intellective control method, neural network has got adequately development and application in system identification and control. The active vibration control technique based on neural network is studied in theory and simulation in this paper, in which the technique of neural network and that of active vibration control are combined together tightly, and a diesel engine two-stage isolating system is studied as the primary object.As far as system identification is concerned, the back-propagation network is used in the paper. Considering the dynamic characteristic of the system itself, the outer input feedback network is used in the identifying network.Two control methods are adopted in the paper. The first method is the adaptive control by using neural networks based on offline secondary path modeling. The method uses two neural networks; one of them is used as a controller to control the vibration, and the other one which is the result of the system identification is used to update the controller's weights. The results of the simulation show the effectiveness of this method.The second method is the adaptive control by using neural networks based on online secondary path modeling. The method uses three cross-update neural networks, one of them is used as a controller to reduce the vibration and the others are identifying the secondary path at real time . It can reduce thevibration effectively in the two-stage isolation system.