Multiple failure diagnosis

Abstract

Model based diagnostic systems have generally avoided the issue of multiple failure diagnosis due to the computational complexity of covering all possible multiple faults and still providing an efficient diagnostic strategy. Optimization of decision trees is already known to be NP-complete, and the number of combinations of multiple faults just serves to exacerbate the problem. Nevertheless, model based diagnosis is becoming popular, and the need for multiple failure diagnosis is real. In this paper, we will provide a formal analysis of the multiple failure problem in the context of one model based approach. Specifically, we will discuss algorithms and their complexity for diagnosing multiple failures using the information flow model.<>