Abstract: In this study, a condition monitoring system for fault diagnosis of ball bearings in rotating machines was developed. Features extraction is based on the relevant information calculated from the vibration signal by wavelet transform. The faults diagnosis procedure is achieved by Hidden Markov Models and uses the wavelet feature as inputs to the HMM. This procedure includes training of the HMM and faults recognition by choosing the model that gives maximum probability of the observation. The designed system was developed to be able to classify four types of pre-established faults in ball bearings and the normal condition. The system was trained and tested by experimental data collected from drive end ball bearing of an induction motor, operating under several shaft speeds and load conditions. The method was applied successfully. It permits the separation of different faults with high recognition rate, almost all fault samples of the database were assigned to the appropriate classes.
Salah Chenikher , Messaoud Ramdani and Bedda Mouldi , 2007. Diagnosis of Ball Bearing Faults Using Wavelet Analysis and Hidden Markov Models (HMM) . Asian Journal of Information Technology, 6: 342-347.