This paper proposes an efficient online monitoring and diagnosis scheme for hybrid systems. The modeling framework is based on extended probabilistic hybrid I/O automata that represent the system behavior, coupling the continuous and discrete dynamics. The diagnosis architecture consists of two primary components: (1) a hybrid observer and (2) a diagnostics module. The hybrid observer employs a focused probabilistic estimation technique to track continuous behavior interspersed with discrete mode changes. The diagnostics module combines fault detection and isolation using a structural residual analysis method. Both discrete and parametric faults are considered in our framework. The working is illustrated on a hybrid two-tank system to demonstrate the effectiveness of this approach.
展开▼
