Abstract: Aiming at difficult problems of judging the running states of rolling bearings from the time domain waveforms and amplitude spectrums of rolling bearing vibration signals in various states for their little difference, a bearing status classifying method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) permutation entropy and least square support vector machine (LS-SVM) was presented. First, the collected bearing vibration signals were divided into a number of training samples and test samples. Then, each sample signal was decomposed by CEEMDAN to obtain several intrinsic mode functions (IMFs), and the permutation entropy of the first few IMFs of each sample signal was calculated. Finally, the bearing states were classified and identified by using the LS-SVM classifier. The method was applied to the classification of the four different states of bearings, and it was also compared with the bearing condition classification based on the permutation entropy of the original vibration signals and LS-SVM classifier. Results show that the overall classification precision of the proposed method reaches 98.75%, significantly increasing from 62.5% of the latter, which effectively verifies the accuracy and superiority of the proposed method.