A graph model-based fault detection framework for structural analysis of complex systems

摘要

This paper deals with a new causal graph-based fault detection and isolation framework. The Causal Ordering Graph is herein used to overcome difficulties of modeling dynamic large-scale systems due to both qualitative and quantitative reasoning. For the fault monitoring tasks, the generic model in natural causality is enable to infer systematically redundancy relationships. Compared with classical graphical tools-based residual signals, the developed scheme allows to get the better of problems related to sensor dualizations, aspects of derivative causality, the presence of algebraic loops and non-invertible constraints. At last, the RLC electrical system is presented as an illustrative example.