An adaptive HMM based classification system in machinery condition monitoring

Machinery condition monitoring is a classical problem in industry. In this paper, an adaptive condition classification system based on hidden Markov model is proposed and validated. This research is based on the vibration analysis of gearbox vibration signals but its application can be expanded to other type of machinery by choosing appropriate signal processing techniques. Singularity analysis with wavelet is applied and wavelet modulus maxima are extracted as the input of classification system. Experimental validation demonstrates the proposed system with excellent performance and flexible extensibility.