Prediction of welding-induced residual stresses using a neural network approach

Jino Mathew

    Research output: Contribution to journalArticlepeer-review

    3 Citations (Scopus)


    Management of operating nuclear power plants greatly relies on the structural integrity assessments for safety critical pressure vessels and piping components. This work presents a novel application of artificial neural networks (ANN) to predict through-thickness residual stress profiles in austenitic stainless steel pipe girth welds for simplified fracture assessments. The model is validated by comparing predictions with experimental measurements using neutron diffraction and the Contour method undertaken on newly fabricated pipe girth welds within a wide range of wall-thickness, heat input and groove geometries. The performance and efficacy of the ANN approach is critically reviewed by comparison with stress profiles recommended in defect assessment procedures. Moreover, an assessment is carried out to evaluate whether the use of neural network bounding profiles can lead to non-conservative estimates of stress intensity factor in fracture assessments. The results demonstrate that the neural network approach can effectively predict through-thickness stress profiles and can be used in fracture assessment of welded components.
    Original languageEnglish
    Pages (from-to)232-233
    JournalWelding and Cutting
    Issue number4
    Publication statusPublished - 2016


    Dive into the research topics of 'Prediction of welding-induced residual stresses using a neural network approach'. Together they form a unique fingerprint.

    Cite this