This paper deals with a fault prognosis method based on physical and stochastic models. Health Indicators (HIs) are generated from a bond graph model using analytical redundancy principle, then, the generated HIs are used to assess the current state of the system and to detect the degradation beginning. To overcome the lack of information on the degradation dynamic, an updated stochastic model, associated with the available information about the Mean Time Before failure (MTBF) is used to estimate the Remaining Useful Life (RUL). An application example illustrates the implementation of the proposed approach and the obtained results show its effectiveness.