Intelligent Particle Filter and Its Application on Fault Detection of Nonlinear System

TLDR

The particle filter is a popular method for estimating hidden states in nonlinear or non‑Gaussian systems, but it often suffers from particle impoverishment due to limited diversity. This paper proposes an intelligent particle filter (IPF) to address the particle impoverishment problem. IPF incorporates genetic‑operator–based strategies to enhance particle diversity, effectively generalizing the conventional particle filter. Experimental results demonstrate that IPF reduces impoverishment, yields more accurate state estimates than the standard filter, and achieves satisfactory real‑time fault detection on a three‑tank system.

Abstract

The particle filter (PF) provides a kind of novel technique for estimating the hidden states of the nonlinear and/or non-Gaussian systems. However, the general PF always suffers from the particle impoverishment problem, which can lead to the misleading state estimation results. To cope with this problem, a modified particle filter, i.e., intelligent particle filter (IPF), is proposed in this paper. It is inspired from the genetic algorithm. The particle impoverishment in general PF mainly results from the poverty of particle diversity. In IPF, the genetic-operators-based strategy is designed to further improve the particle diversity. It should be pointed out that the general PF is a special case of the proposed IPF with the specified parameters. Two experiment examples show that IPF mitigates particle impoverishment and provides more accurate state estimation results compared with the general PF. Finally, the proposed IPF is implemented for real-time fault detection on a three-tank system, and the results are satisfactory.