Influence Diagrams for Optimal Maintenance Planning

Over the last two decades Bayesian networks and influence diagrams have received notable attention within the field of artificial intelligence and expert systems. During the last few years the technology has been further developed for problem solving within other engineering fields. The objective of this study is to present a conceptual bayesian network model for probabilistic prediction of fatigue crack growth in welded steel tubes.

It is shown that despite discretization of the variable domain, the prediction is in good agreement with results obtained by the well-established structural reliability methods FORM/SORM. The Bayesian network model is argumented by decision and utility nodes, thus forming a full decision model for inspection planning.

With the applied program package the optimal inspection plan is easily obtained. Moreover, the updating facilities allow for fast changes of the inspection plan when new knowledge becomes availabe.