Probabilistic Evaluation of Welded Structures

Some deficiencies in structual welding, particularly when thousands of welds are installed, are expected. This deficiency may or may not compromise the strength of the joint or the structural element. A deterministic evaluation of this deficiency may not be possible. A probabilistic procedure is suggested. Limited sampling information from the site is used to develop this methodology. Sampling information obtained from a project should be checked before the information is used for any statistical prediction. The χ² test of homogeneity could be used to check the homogeneity of the samples. The hyperbinomial distribution may be the most appropriate distribution to study the characteristics of a group of welds. It considers the effect of sample size on the prediction. A large sample size is always beneficial, but the cost-benefit ratio for the collection of samples should be studied before any decisions are made. The method suggested here would help to study this area. Repairing the welds, whenever necessary, is the final step. This method is a safe, economical and conservative solution to this complex problem.