Bayesian network discretization of reliability problems
A MATLAB based software prototype performing efficient static discretization for structural reliability problems, such that they can be updated in a discrete Bayesian network framework. The user defines his/her reliability problem through a limit state function and a statistical model of the corresponding basic random variables. Furthermore a couple of parameters corresponding to the discretization and the BN structure have to be provided.
The software:
- Comes up with an efficient discretization scheme, applying the heuristics developed in (Zwirglmaier & Straub, 2015)
- Establishes the CPTs of the reliability Bayesian network.
- Outputs the BN model in the form of a xdsl-file, which can be read and modified e.g. using the (free) BN software Genie (https://dslpitt.org/genie/)
Download
Requirements
Matlab, incl. Statistical toolbox and Optimization toolbox
The Bayesian network is outputed in the format of the Genie BN software, which can be downloaded for free here: https://dslpitt.org/genie/
Documentation & background
Zwirglmaier K., Straub D.: A discretization procedure for rare events in Bayesian networks. Reliability Engineering & System Safety, under review.