Aimed at the non-linear and non-stationary characteristics of vibration signals of fault rolling bearings, a bearing fault diagnosis method was proposed based on time domain index, wavelet packet energy and least squares support vector machine (LSSVM).The original signal of the rolling bearing was analyzed and calculated in the time domain and decomposed by wavelet packet, and the time domain index with obvious state difference and the wavelet packet energy with great energy difference after wavelet packet decomposition were extracted as fault feature vectors.Data samples contained multiple feature vectors were divided into training samples and test samples and normalized.Training samples were used as the input of LSSVM to train the model, and the trained LSSVM model was used to classify and diagnose the test samples.The experimental results show that the bearing state overall recognition rate is 97.5%