Diagnosis of dynamic systems based on explicit and implicit behavioral models: An application to gas turbines in esprit project tiger

We are interested in the monitoring and diagnosis of dynamic systems. We combine explicit temporal models of the behavior of a dynamic system with implicit behavioral models supporting model-based approaches. This work is driven by the needs of and is applied to two gas turbines of very different size and power. We describe the problems of building systems for these domains and illustrate how we have developed a system where these two approaches complement each other to provide a comprehensive fault detection and diagnosis system. We also explore the strengths and weaknesses of each approach. The work described here is currently working continuously, on-line to a gas turbine in a major chemical plant.