Abstract: The author proposed a bearing fault detection algorithm based on Morlet wavelet and multimode kernel optimized by genetic algorithm. Firstly, the algorithm extracted the characteristics of Morlet wavelet transform under conditions of multiscale and multiple displacement for the fault signal of original
 bearing. Secondly, the author designed a multimode kernel method, including linear kernel and radical basis function (RBF) kernel. Finally, the genetic algorithm (GA) was used to optimize parameters of multimode kernel in the support vector machine (SVM) training process, and the  optimizing multimode kernel for bearing fault detection was carried out. The ball fault, inner ring crack fault and outer ring crack fault were tested on the data set of UoCn intelligent maintenance center, and the error rate and efficiency of singlemode kernel and multimode kernel were compared. The experimental results show that the improved algorithm can get the robust features for bearing fault detection, and the multimode kernel can converge quickly and get optimal results under the optimization of GA, the accuracy of bearing fault detection is greatly improved by sacrificing a small amount of time efficiency.

Key words: genetic algorithm （GA）, support vector machine （SVM）, multi-mode kernel method, bearing fault detection, Morlet wavelet transform