Scheduling for time-constrained model-based diagnosis
Arantza Aldea · 1995
We see the need of a supervisor which directs the model based diagnosis (MBD) engine to be able to use model based diagnosis efficiently. We also consider the importance of a decomposition of the system to be diagnosed into easy to handle pieces when MBD is employed for industrial complex applications, and the necessity of a time constrained mechanism to guide the MBD engine and navigate through the suspect components when there is a deadline. We also develop the idea of the combination of economic and behavioural knowledge of the system to be diagnosed which lead to a time limited diagnosis, and the incorporation of scheduling techniques to MBD to perform the time constrained reasoning. A bicriterion scheduler algorithm has been developed which aims to minimise the total replacement cost of the final diagnosis. As a result, we propose a system which can adjust the quality of its solution to the time available, and therefore it guarantees the generation of a diagnosis in the available time. Naturally, a solution will be more accurate, when more time is available for diagnosis.