Algorithms for Bayesian network modeling and reliability assessment of infrastructure systems

•Novel algorithms developed for Bayesian network modeling of infrastructure systems.•Algorithm presented to compress information in conditional probability tables.•Updating algorithm presented to perform inference on compressed matrices.•Algorithms applied to example systems to investigate their performance.•Orders of magnitude savings in memory storage requirement demonstrated.

Novel algorithms are developed to enable the modeling of large, complex infrastructure systems as Bayesian networks (BNs). These include a compression algorithm that significantly reduces the memory storage required to construct the BN model, and an updating algorithm that performs inference on compressed matrices. These algorithms address one of the major obstacles to widespread use of BNs for system reliability assessment, namely the exponentially increasing amount of information that needs to be stored as the number of components in the system increases. The proposed compression and inference algorithms are described and applied to example systems to investigate their performance compared to that of existing algorithms. Orders of magnitude savings in memory storage requirement are demonstrated using the new algorithms, enabling BN modeling and reliability analysis of larger infrastructure systems.