Probabilistic Updating of Flaw Information

The prediction and control of flaws in welds occupy an important role in design against fatigue and fracture failure. However, the present NDT devices can detect only a fraction of the flaws and they do not give the actual size of flaws detected. Using Bayes Theorem, a framework is proposed whereby distributions of flaw size and density are updated from NDT inspection data and the level of repair. The concept is equivalent to a filtering process where the detectability function of the NDT device acts as the filter. The information derived will help planning of NDT inspection programs and consistent code specifications.