Fault prognosis based on Hidden Markov Models

Monitoring electrical motors in critical or sensitive environments is one of the major challenges of our era. This can be applied both in electric vehicles, hybrid and avionics. Therefore the development of tools, which are able to ensure continuity of service by identifying and predicting faults, is crucial to provide a reliable monitoring. This paper presents two methods based on Hidden Markov Models for the prediction of impending faults. They are based on pattern recognition, that is a data-driven approach widely used in the field of faults detection and diagnostic. This paper aims to show that methods such as Hidden Markov Models, commonly used in the diagnosis, can also be used in the field of prognosis. The first method, based on the recognition of degradation processes, allows predicting the imminent appearance of the fault and the second is based on modeling the state of degradation of the studied system. An example of application is given to demonstrate their applicability. The results show their effectiveness to predict the imminent appearance of a fault.