Contribution to the estimation of the residual lifetime of systems in the presence of uncertainties.

Summary The implementation of a predictive maintenance policy is a major challenge in the industry which tries to reduce as much as possible the costs related to maintenance. Indeed, the systems are more and more complex and require a more and more thorough follow-up in order to remain operational and secure. Predictive maintenance requires on the one hand to evaluate the state of degradation of the system components, and on the other hand to predict the future occurrence of a failure. More precisely, it is a question of estimating the time remaining before the arrival of a failure, also called Remaining Useful Life or RUL in English. The estimation of a RUL is a real challenge because the relevance and effectiveness of maintenance actions depend on the accuracy of the results obtained. There are many methods allowing to realize a residual life prognosis, each one with its specificities, its advantages and its disadvantages. The work presented in this manuscript focuses on a general methodology to estimate the RUL of a component. The objective is to propose a method applicable to a large number of different cases and situations without requiring major modifications. Moreover, we also seek to treat several types of uncertainties in order to improve the accuracy of the prognostic results. In the end, the developed methodology constitutes a decision aid for the planning of maintenance operations. The estimated RUL allows to decide on the optimal moment of the necessary interventions, and the treatment of uncertainties brings an additional level of confidence in the obtained values.