关键词: 滚动轴承, 经验模态分解, 小波包去噪, 自适应神经模糊推理系统, 故障诊断

Abstract: In order to diagnose rolling bearing’s three fault types more effectively, such as inner race fault, outer race fault and balls fault, a method that Adaptive Neuro-Fuzzy Inference Systems（ANFIS） and wavelet packet de-noising based on Empirical Mode Decomposition（EMD） is proposed. As the signals are often corrupted by noise, so they are de-noised, and preprocessed signals are investigated using ANFIS analysis. The results show that the wavelet packet de-noising based on EMD can improve the Signal-to-Noise Ratio（SNR） effectively. After signals are preprocessed, the result of ANFIS analysis shows that average error is low. It can diagnose the three fault types above-mentioned better.

Key words: rolling bearing, Empirical Mode Decomposition（EMD）, wavelet packet de-noising, Adaptive Neuro-Fuzzy Inference Systems（ANFIS）, fault diagnosis