Active vibration damping based on neural network theory

Abstract The possibility of utilizing neural networks (NN) theory for active vibration control in a longitudinal cantilevered-beam system is investigated by simulation and experiment. The feedback control system providing active vibration damping is constructed with a three-layer-type neural network controller, a strain gauge sensor, and an actuator generating electro-magnetically interactive control force. It is found that the active damping effect using the NN controller is obtained as Q max − 1 = 0.144 by the specific damping capacity criterion, which corresponds to about 6.2 times the maximum damping effect observed in the Fe-based ferromagnetic high damping metal called SIA, and that the NN vibration control system is also robust against some parameter variation.