Abstract:Aiming at the problem that the stator current signal is not stationary when the induction motor fails under variable frequency environment, a fault diagnosis method for induction motor based on complementary set empirical mode decomposition (CEEMD) and convolutional neural network (CNN) was proposed.Firstly, the simulation current data were obtained by modeling the motor in frequency conversion environment with ANSYS, and the stator current signal was decomposed into a series of intrinsic mode functions (IMF) using CEEMD;secondly, by calculating permutation entropy and sample entropy, the IMF component with small complexity was selected and its average value was calculated to extract fault features; then, the feature data set was input into convolutional neural network (CNN) for training and verification; finally, the experimental platform was built to collect the current signal, and the signal was filtered and decomposed and reconstructed by CEEMD, which was put into the CNN trained model for testing, and the recognition rate reached 9556%.It is proved that this method is a feasible fault diagnosis method for induction motor, which can accurately identify the normal state, broken rotor bar and air gap eccentricity of induction motor.