Motor Fault Diagnosis Based on MMAS-Optimized Integration Neural Network

In order to diagnose early faults of motor and improve motor running efficiency, an integration neural network method Based on Max-Min ant system algorithm (MMAS) is proposed. As input signals of diagnosis system, signals of stator current and rotor oscillation are corresponding to two diagnosis sub-networks. Thresholds and weights in sub-networks are optimized with MMAS, and then each of sub-networks diagnoses local faults. To reduce the complexity of mapping function and improve the reliability of fault diagnosis system, the final conclusion is obtained through the decision-making information fusion for the local results from sub-networks. Simulation results show that the proposed diagnosis method has fast convergence rate, small prediction error and strong generalization ability. It is a good reference to motor fault diagnosis.