Adaptive Anomaly Detection Using a Hidden Markov Model

Online condition monitoring and diagnosis systems are very important in the modern manufacturing industry. We present a new method to assess the degradation processes of multiple failure modes using the Hidden Markov Model (HMM). The HMM is combined with statistical process control (SPC) to detect the occurrence of unknown faults. This method allows an HMM to adjust and update the state space with the identification of new states. Hence, the online degradation assessment and adaptive fault diagnosis can be simultaneously obtained. The turning process are used to illustrate that previously unknown tool wear processes can be successfully detected at the early stages using the HMM.