Artificial neural network design for fault identification in a rotor-bearing system

Abstract A neural network simulator built for prediction of faults in rotating machinery is discussed. A back-propagation learning algorithm and a multi-layer network have been employed. The layers are constituted of nonlinear neurons and an input vector normalization scheme has been built into the simulator. Experiments are conducted on an existing laboratory rotor–rig to generate training and test data. Five different primary faults and their combinations are introduced in the experimental set-up. Statistical moments of the vibration signals of the rotor-bearing system are employed to train the network. Network training is carried out for a variety of inputs. The adaptability of different architectures is investigated. The networks are validated for test data with unknown faults. An overall success rate up to 90% is observed.