On the Use of Particle Flow to Enhance the Computational Performance of Particle-Filtering-based Prognostics

Prognostic approaches based on particle filtering employ physical models in order to estimate the remaining useful life (RUL) of systems. To this aim a set of particles is used to first estimate the degradation state of the system and then to predict the distribution of the RUL through simulation. The computational complexity of this approach is a function of the number of particles used in the state estimation and of the time each particle needs to simulate the RUL. It is therefore clear that enhancing the computational performance of this approach requires reducing the number of particles. In this paper we investigate the applicability and suitability of the particle flow particle filter for particle-filtering-based prognostics. The estimation of the remaining driving range (RDR) of an electric vehicle is used as the case study to illustrate the improvement in computational performance of the proposed approach in comparison to the standard particle filter.