A fault diagnosis method of rolling bearings using empirical mode decomposition and hidden Markov model

This paper describes a new approach to detect localized rolling bearing defects based on Empirical Mode Decomposition (EMD) and Hidden Markov Model (HAIM). In view of the non-stationary characteristics of bearing fault vibration signals, using EMD method, the original non-stationary vibration signal can be decomposed into a finite number of stationary signals. The stationary signal adapts itself better to the conditions of fault characteristic parameter based on power spectrum analysis and also show bearing fault characteristics clearly. By setting envelope-singles fault-characteristic parameters of each main stationary signal to train HMM, this study also presents a method of pattern recognition for bearing fault diagnosis using HMM. Experimental results show that (1) the approach has successful bearing fault detection rates as high as 98% for every single fault; (2) although fault styles sometimes are confusing, the approach proves better at recognizing combinations of these faults.