Intelligent diagnosis of discrete-event systems with preprocessing of critical scenarios

Knowledge compilation is no novelty in model-based diagnosis of discrete-event systems. The system is preprocessed in order to generate a data structure that allows for the efficient explanation of any symptom online, while the system is being operated. Unfortunately, this technique requires the diagnosability of the system. Even worse, it comes with a prohibitive cost in terms of computational complexity, owing to the explosion of the state space even for systems of moderate size, which makes the whole approach impractical for real applications. To overcome these two obstacles, a novel technique based on scenarios is proposed. Scenarios are compiled into a flexible data structure called an open dictionary, which allows for the efficient explanation of symptoms. The dictionary is open inasmuch it can be expanded by new scenarios and symptoms.