Methods to choose the best Hidden Markov Model topology for improving maintenance policy

Prediction of physical particular phenomenon is based on partial knowledge of this phenomenon. Theses knowledges help us to conceptualize this phenomenon according to different models. Hidden Markov Models (HMM) can be used for modeling complex processes. We use this kind of models as tool for fault diagnosis systems. Nowadays, industrial robots living in stochastic environment need faults detection to prevent any breakdown. In this paper, we wish to find the best Hidden Markov Model topologies to be used in predictive maintenance system. To this end, we use a synthetic Hidden Markov Model in order to simulate a real industrial CMMS∗. In a stochastic way, we evaluate relevance of Hidden Markov Models parameters, without a priori knowledges. After a brief presentation of a Hidden Markov Model, we present the most used selection criteria of models in current literature. We support our study by an example of simulated industrial process by using our synthetic model. Therefore, we evaluate output parameters of the various tested models on this process: topologies, learning algorithms, observations distributions, epistemic uncertainties. Finally, we come up with the best model which will be used to improve maintenance policy and worker safety.