Trajectory-based observer for hybrid automata fault diagnosis

Temporal Logic Inference for Hybrid System Observation with Spatial and Temporal Uncertainties

“…Once the optimization procedure obtains an optimal formula φ, we can use the obtained φ to refine the basic observer constructed as in [37]. The refinement procedure is to shrink the observer's state (i.e., the state estimate for H) and the subsequent transitions based on satisfaction or violation of a MTL formula.…”

“…To that end, [36] presents a characterization of observability over an interval. In [37], the authors of the present paper propose a framework for hybrid system state estimation from the perspective of bisimulation theory [38]. The framework is based on the robust test idea [39], which extracts the spatial and temporal properties of infinitely many trajectories by a finite number of simulations.…”

“…In inferring a temporal logic formula that classifies different system behaviors, we can further design an observer for determining the location of the hybrid system at any time. Our previous work [37] results in a hybrid observer that estimates both the discrete and continuous states constantly. The observer only uses the discrete outputs generated by the hybrid system's observable events and their timing information as its input, and thus is referred to as the basic observer in this paper.…”

“…The observer only uses the discrete outputs generated by the hybrid system's observable events and their timing information as its input, and thus is referred to as the basic observer in this paper. Based on [37], we utilize the inferred MTL formula from the MTL classifier to refine the basic observer and the obtained observer is referred to as the refined observer. We illustrate the idea with Figure 1.…”

A Neural-Fuzzy Approach for Fault Diagnosis of Hybrid Dynamical Systems: Demonstration on Three-Tank System

“…As for diagnosis methodologies based on hybrid automaton (HA), the literature in that field is abundant and different solutions have been proposed in Derbel (2009) (Cocquempot et al, 2004), HA and state observers (Deng et al, 2015) and HA and bond graph (Abdallah et al, 2016)] combining the advantages of the both approaches for best performance of the diagnosis method.…”