Abstract:

Condition monitoring of machinery can provides real time information regarding machine status on-line,thus avoiding the production losses and minimizing the chances of catastrophic machine failures.In this paper,we present a fault classification system with an on-line model training and estimating algorithm.It is based on the wavelet transform and Hidden Markov Models (HMMs).The machinery condition is identified by selecting the HMM which maximizes the probability of a given observation sequence.The observation sequence is based on the wavelet modulus maxima distribution,which was proved to be effective in fault detection in previous research.Using observation sequences obtained from real vibration signals the developed classification system is validated.

Key words: condition monitoring, wavelet modulus maxima distribution, Lipschitz exponent, hidden Markov model