EVMDD-Based Analysis and Diagnosis Methods of Multi-State Systems with Multi-State Components.

摘要

A multi-state system with multi-state components is a model of systems, where performance, capacity, or reliability levels of the systems are represented as states. It usually has more than two states, and thus can be considered as a multi-valued function, called a structure function. Since many structure functions are monotone increasing, their multi-state systems can be represented compactly by edge-valued multi- valued decision diagrams (EVMDDs). This paper presents an analysis method of multi-state systems with multi-state components using EVMDDs. Experimental results show that, by using EVMDDs, structure functions can be represented more compactly than existing methods using ordinary MDDs. Further, EVMDDs yield comparable computation time for system analysis. This paper also proposes a new diagnosis method using EVMDDs, and shows that the proposed method can infer the most probable causes for system failures more efficiently than conventional methods based on Bayesian networks.