The bidirectional associative memory neural network based on fault tree and its application to inverter´s fault diagnosis

With study on fault tree analysis (FTA) and bidirectional associative memory (BAM) neural network, a new method of intelligent fault diagnosis is proposed. All the knowledge on the happening of top events is stored in fault tree, in which the whole fault modes are obtained. The priori knowledge and experience of system diagnosis are introduced to FTA. The learning sample of BAM neural network is deduced by the corresponding relations between the fault modes and the fault analysis. The diagnosis results are associated parallel by the associative memory matrix; also the general ability of fault diagnosis is being expanding. With experiments and application to inverter´s fault diagnosis, results show that this method has better performance for real-time and effectivity.