摘要：AbstractThis paper presents a two-phase hybrid prognostics approach; in the first phase, the model’s parameters are estimated using available training data in the least squares sense using the Levenberg-Marquardt algorithm. The second phase consists of using a particle filter to update the knowledge acquired so far and to predict future states of the system using in the Bayesian sense. The approach is used for an accelerated ball bearing data set, the PRONOSTIA platform, where a general fractional polynomial model is proposed as degradation model. The results of the Remaining Useful Life estimation are compared with another work in the literature, indicating its suitability and competitiveness for prognostics in this data set.
