Exploiting fuzzy reasoning optimized by Particle Swarm Optimization and adaptive thresholding to diagnose multiple faults in dynamic hybrid systems

In this paper, a general methodology to diagnose multiple faults in hybrid dynamic systems is proposed. The considered dynamic hybrid systems exhibit continuous dynamics with discernable discrete functioning modes. The inputs of the proposed methodology are residuals representing the numerical evaluation of Analytical Redundancy Relations extended to hybrid systems. These residuals are generated due to the use of switched bond graph modeling. The evaluation of these residuals is based on the combination of adaptive thresholding and fuzzy logic reasoning optimized by Particle Swarm Optimization (PSO). The results of the detection module are displayed as a colored causal graph. This causal graph is exploited to correctly isolate multiple faults. The ongoing experiments focus on a simulation of the three-tank hydraulic system, a benchmark in the diagnosis domain.