Improved DCNN Based on Multi-Source Signals for Motor Compound Fault Diagnosis

Induction motors, the key equipment for rotating machinery, are prone to compound faults, such as a broken rotor bars and bearing defects. It is difficult extract fault features identify faults from single signal because multiple overlap interfere with each other in fault. Since current signals vibration have different sensitivities multi-channel deep convolutional neural network (MC-DCNN) diagnosis model based on multi-source proposed this paper, which integrates original of motor. Dynamic attenuation learning rate SELU activation function were used improve hyperparameters MC-DCNN. The dynamic attenuated can stability training avoid collapse effectively. problems gradient disappearance explosion during iteration due its configuration, thereby avoiding falling into local optima. Experiments showed that effectively solve problem motor identification, three comparative experiments verified improved method accuracy identification.