Fault diagnosis of induction motors with dynamical neural networks

The paper studies the fault diagnosis of induction motors using neural network time-series models. The problem has been widely discussed in the literature and neural networks have been used in the fault diagnosis of induction motors. However, the neural network models have been mostly static - dynamical neural networks have been overlooked and have not received enough attention in this context. Here neural network time-series models are created for the normal and faulty motor. A filter bank of the models is formed and a Bayesian classifier is used to determine the correct classification of the motor condition, when tested with different types of FEM simulated data for different degrees of load.