Research On Induction Motor Fault Monitoring Based On Neural Network

This paper discusses the characteristics of induction motor fault diagnosis and its achievements in recent academic research, which is based on a large collection of the abroad and domestic scientific and technical documents. The diagnostic device is designed with the help of neural network to monitor whether an induction motor performs under normal conditions by detecting its stator current, which can examine the typical faults including electrical faults and mechanical faults.In the process of research, the classical stator fault, rotor fault, eccentricity fault, and bearing fault are firstly analyzed at length in the case of induction motor with squirrel cage based on its principle of operation. The typical frequencies and their relevant amplitude are extracted from the motor and the relation between motor faults and typical frequencies is presented.Secondly, considering the application of neural network to fault diagnosis field, a sort of modified form of weight learning algorithm, MABPM algorithm, based on the standard algorithm is applied to train the feedforward network, which generally accelerates the convergent speed and gains better generalization ability than usual. Fuzzy theory is also introduced to describe the fault symptom which can realize the diagnosis of fault degree and probability of fault occurring.Finally, according to different kinds of motor faults, their fault samples are collected to train the neural network that act as the final diagnostic device. A great deal of simulation work has been carried out for induction motor fault diagnosis and this new algorithm obtains more excellent results compared to the standard BP algorithm and modified BP algorithm, which proves that this method is precise, fault tolerant and effective.