Combining Weld Residual Stress Predictions and Measurement for Use in Probabilistic Structural Integrity Assessments

The use of residual stress information in probabilistic fracture assessments is hindered by difficulties in the quantification of uncertainty. At the same time, it is often necessary to consider residual stress data derived via two or more independent methods in an assessment: typically from a model of the process which introduced the stress, and from a direct physical measurement. The uncertainty in single weld process models is difficult to quantify and is strongly dependent on the process being modelled, the material constitutive behaviour assumed, and so on. Likewise, most experimental techniques for measuring deep residual stresses on welded metallic components, including relaxation methods such as Deep Hole Drilling and diffraction-based methods, also have multiple physical sources of uncertainty associated with them. This makes the uncertainty associated with single measurements difficult to estimate reliably. We explore the use of inverse-variance weighting to combine such datasets through “characteristic” uncertainties derived from prior round robin studies, and we use data from the NeT TG4 residual stress measurement and modelling round robin to illustrate this approach. Although it requires some significant simplifications, it allows convenient synthesis of residual stress data while gaining more realistic uncertainty estimates than are typically available from single measurements. This is significant because straightforward yet robust uncertainty estimates will be key for enabling future structural integrity assessment methodologies.