Representation and Partitioning of Welding Residual Stress Distributions for Use in Structural Integrity Assessments

In order to obtain good estimates of stress intensity factors in a structural integrity assessment, the accuracy with which a residual stress distribution is represented should be commensurate with the importance of localised peaks in residual stress, in particular where such peaks lie within the region under assessment. This paper describes work undertaken to investigate the importance of accurately representing residual stress distributions in structural integrity assessments. This has been carried out by comparing regular polynomial representations of residual stress distributions, combined with available weight function stress intensity factor solutions (as provided in the R6 procedures) with alternative polynomial representations of residual stress distributions, which provide a more accurate fit in the region of the crack. Such improvements in representation of residual stress profiles provide an indication as to how stress intensity factor solutions could, in future, be modified in order to result in improved accuracy of calculated stress intensity factors. Representation by partitioning residual stress profiles into membrane, bending and self-balancing components, in terms of providing a more straight-forward route for curve-fitting of residual stress profiles is considered. The investigation considers several transverse, through-thickness residual stress distributions. Stress intensity factors are calculated for a variety of crack sizes. Representation of the residual stress profiles in the stress intensity factor solutions are compared, as are the results of the stress intensity factor calculations. The conclusions arising provide guidance as to how current methods of curve fitting a residual stress distribution may be improved in cases where current methods may not be accurate. Advice is also provided as to the relative merits of representing residual stress distributions as a set of partitioned components or as a single distribution.