Aiming at the problem of bearing fault diagnosis

the relevance vector machine（ RVM） is used for diagnosis analysis. In order to improve the diagnostic accuracy rate

how to select the RVM input and nuclear parameters is studied. The original signal energy eigenvalue is extracted using wavelet packet energy entropy. It is found by study that the energy feature values extracted by each state have obvious differences

which can be used as the input of the RVM. Study the optimal choice of nuclear parameters

a bat optimization algorithm based on chaotic search is proposed. The improved bat algorithm solves the drawbacks of being trapped in local optimum.The chaos-bat algorithm（ CBA） is used to optimize the kernel parameters of the RVM

which effectively improves the diagnostic precision of the RVM. Through the comparison of simulation experiments

the optimized RVM has better diagnostic effect than the traditional model.