Abstract: Aiming at the difficulty of fault diagnosis of motor bearings, a state diagnosis method of motor bearings based on multi-feature fusion(MFF) and improved whale optimization algorithm (IWOA) optimized least squares support vector machine (LSSVM) was proposed. Firstly, the algorithm population was initialized by the Sobol sequence, the Levy flight strategy was added to the search process of the algorithm population, and the inertia weight was added to the position update formula of the WOA algorithm. Then, the wavelet packet energy features, average value and kurtosis of the motor bearing vibration signal were extracted, and the above motor bearing vibration signal features were used as the input of the algorithm. Finally, in order to verify the effectiveness of the motor bearing diagnosis methods based on MFF and IWOA-LSSVM, the wavelet packet energy feature alone was used as the algorithm input, and the wavelet packet energy feature and the time domain feature were used together as the input of the algorithm, two related motor bearing state recognition comparative experiments were made. The research results show that the multi-feature fusion can reflect the real operating state of the motor bearing more comprehensively than the single wavelet packet energy feature. Comparing with the PSO and GA algorithms, the basic WOA algorithm can effectively avoid local optima more. Comparing with the basic WOA algorithm, the improved WOA algorithm can effectively avoid local optimization. The IWOA-LSSVM algorithm has a recognition rate of 99.5% for the motor bearing state, which is higher than other motor bearing state recognition algorithms, and has better classification performance.

Key words:  motor bearing; fault diagnosis; multi-feature fusion(MFF); improved whale optimization algorithm (IWOA);least squares support vector machine (LSSVM)