The paper discusses theoretical foundations and practical aspects of applying model-based diagnosis (particularly consistency-based diagnosis) to dynamic systems. Many approaches to this task take it for granted that it requires simulation of the system being diagnosed. We present conditions for avoiding the often prohibitively expensive step of simulation, which are stated as properties of the model and the predictive algorithm and the observability of the system. The results provide design criteria for models and diagnostic systems and a foundation for tackling new significant types of applications. This fact is illustrated by a case study on diagnosis of the hydraulic circuit of an anti-lock braking system.
展开▼