Detection of NARMA-sequence order using recurrent artificiai neural networks

Recently artificial neural networks have been advocated as a possible technique for fault detection. Neural networks are noise tolerant and their ability to generalise the knowledge as well as to adapt during use are extremely interesting properties. The approach to solution of properties changes detection problem for stochastic sequences circumscribed by the nonlinear equations of an autoregressive-moving average is proposed in the paper. The architecture of multilayer recurrent neural network and algorithms of neurone parameters tuning ensuring maximum speed of learning process are proposed. The virtue of the proposed approach is the possibility of stochastic sequences of any structure diagnosing, high speed and computing simplicity.