The complexity and sophistication of the new generation of aircraft, automobiles, satellites, chemical plants, and manufacturing lines, along with growing demands for their reliability and safety while keeping cost low, is being met by more automated control and monitoring systems, and the use of functional redundancy techniques for fault detection and isolation (FDI).
Typically system models capture relations between measured variables and system or component parameters. Simulation and reasoning methodologies generate system behavior from these models, and when combined with techniques for identifying and analyzing observed deviations can be used to isolate a large number of possible faulty situations [1, 2, 5, 19]. Some faults, such as a pipe blockage that completely isolates two parts of a system, change system structure and require a change in the system model itself. This renders these faults hard to diagnose unless failure mode models are explicitly incorporated into the analysis scheme [8]. In this paper we investigate FDI techniques that apply to complex dynamic systems. The focus is on applying dynamic models to exploit transients in dynamic behavior caused by discontinuous parameter changes (faults). The aim is to quickly identify the root-causes for discrepancies in system behavior [11, 15].