A new method of bearing fault diagnostics in complex rotating machines using multi-sensor mixtured hidden Markov models

Vibration signals from complex rotating machines are often non-Gaussian and non-stationary, so it is difficult to accurately detect faults of a bearing inside using a single sensor. This paper introduces a new bearing fault diagnostics scheme in complex rotating machines using multi-sensor mixtured hidden Markov model (MSMHMM) of vibration signals. Vibration signals of each sensor will be considered as the mixture of nonGaussian sources, which can depict non-Gaussian observation sequences well. Then its parameter learning procedure is given in detail based on EM algorithm. In the end the new method was tested with experimental data collected from a helicopter gearbox and the results are very exciting.