To provide an improved active vibration controlling system using a feedback and feedforward sensor input for solving the problem of the control of random and repeated active vibration and the deletion of the noise of the system.

An artificial neural network learns the dynamics of a structural body 570, and provides an output signal following up the state variable of the structural body 570. In one execution of the neural network, plural neurons obtain a biasing input derived from a sensor input and an input from the other neuron of the network. Moreover, each neuron obtains the feedback input from itself. Each input to the neuron is weighted by using a weight function derived in on-line. The neural network supplies a structural parameter and the state variable to an optimal controller 530, and the optimal controller 530 derives a control signal, and provides it to an actuator 560 so that vibration and/or a noise sensed in the system can be canceled.