Compression Algorithm for Bayesian Network Modeling of Binary Systems
Tien, I., and Der Kiureghian, A., “Compression Algorithm for Bayesian Network Modeling of Binary Systems,” In G. Deodatis, B. Ellingwood, and D. Frangopol, eds., Safety, Reliability, Risk and Life-Cycle Performance of Structures and Infrastructures, New York: CRC Press, pp. 3075-3081, June 2013
Abstract — A Bayesian network (BN) is a useful tool for analyzing the reliability of systems. The BN framework is limited, however, by the size and complexity of the system that can be tractably modeled. Each node in a BN graph is associated with a conditional probability table (CPT), the size of which grows exponentially with the number of connected nodes in the graph, presenting a memory storage challenge in constructing and analyzing the BN. In this paper, we look at binary systems, where components of the system are in either one of two states, survival or failure, and the component states deterministically define the system state. This analysis is particularly useful for studying the reliability of infrastructure systems, where, e.g., the states of individual gas pipelines or roads directly impact the state of the overall natural gas or transportation system. We present a compression algorithm for the CPTs of such systems so that they may be modeled on a larger scale as BNs. We apply our algorithm to an example system and evaluate its performance compared to an existing algorithm.
Leave a Reply