Nonlinear vibration control based on an artificial neural network without learning

An Artificial Neural Network Method (ANNM) used for nonlinear vibration control systems is described. ANNM is a very efficient method in case that an analytical model of a plant is difficult to develop for their nonlinearity and measurement noise. A back propagation rule is usually adopted for their learning rule. Because iterations for learning become thousands or more, it is difficult to tune time-varying nonlinear systems in real-time. Thus, in this paper ANNM where there is no need of learning is proposed for real-time tuning. The key methodology is that the weights of the hidden and output neuron are set to be the input and output pattern, respectively. Thus, the iterations for learning are not necessary. Applying this approach to the estimation of damping ratio of second-order systems, its availability was verified.