Diagnosis of mechanical fault signals using continuous hidden Markov model

Hidden Markov Model (HMM) has been actively studied in speech recognition since 1960s and increasingly used in many other fields. However, its application to mechanical engineering has been very limited. HMM is not only very accurate and robust in analyzing signals but also can be a very powerful method of predicting target system's condition change. In this paper, continuous HMM (CHMM) has been tuned to be used in mechanical signal analysis and applied to diagnose of various mechanical signals including rotor fault signals. The results show HMM's big potential as an intelligent condition monitoring tool based on its accuracy, robustness, and forecasting ability.