Analysis Methods of Multi-state Systems Partially Having Dependent Components Using Multiple-Valued Decision Diagrams
Shinobu Nagayama, Tsutomu Sasao, Jon T. Butler, Mitchell Aaron Thornton, Theodore W. Manikas · 2014
In a large system, such as a water, gas, or electrical distribution system, degraded performance due to failures of components can be modeled as a set of discrete states interconnected by edges with weights that represent conditional probabilities. To establish such a model, we compute the conditional probabilities with multi-valued decision diagrams (MDDs). Since a typical decision diagram is large, the computation time is also large. In this paper, we propose an edge-valued MDD (EVMDD) based method to avoid unnecessary path traversals. The proposed method is a hybrid method of a path traversal method and a bottom-up method that visits each node only once. By effectively combining both methods, we achieve a speed-up of the analysis by about 2:3 times for large systems compared to an existing method.