A fault detection strategy based on intelligent particle filter for nonlinear systems

The general particle filter (GPF) provides an effective tool of fault detection for non-linear systems with non-Gaussian cases. However, due to the particle impoverishment problem of the GPF, the estimation of hidden state in process may get misleading results and thus lead to undesired conclusion for fault detection purpose. To solve this problem, a fault detection strategy, which employs a novel kind of particle filter, i.e. intelligent particle filter (IPF), is proposed in this paper. The proposed IPF strategy could improve the impoverishment problem of GPF and increase the accuracy of hidden state estimation, thus offers desired results for fault detection. Two numerical examples show superior performance of the proposed IPF for fault detection of nonlinear system with high non-Gaussian noises compared with the GPF-based strategy.