Through-Thickness Residual Stress Profiles in Austenitic Stainless Steel Welds: A Combined Experimental and Prediction Study

Jino Mathew, Richard Moat, Sanjoo Paddea, J. A. Francis, Michael Fitzpatrick, John Bouchard

Research output: Contribution to journalArticle

6 Citations (Scopus)
15 Downloads (Pure)

Abstract

Economic and safe management of nuclear plant components relies on accurate prediction of welding-induced residual stresses. In this study, the distribution of residual stress through the thickness of austenitic stainless steel welds has been measured using neutron diffraction and the contour method. The measured data are used to validate residual stress profiles predicted by an artificial neural network approach (ANN) as a function of welding heat input and geometry. Maximum tensile stresses with magnitude close to the yield strength of the material were observed near the weld cap in both axial and hoop direction of the welds. Significant scatter of more than 200 MPa was found within the residual stress measurements at the weld center line and are associated with the geometry and welding conditions of individual weld passes. The ANN prediction is developed in an attempt to effectively quantify this phenomenon of ‘innate scatter’ and to learn the non-linear patterns in the weld residual stress profiles. Furthermore, the efficacy of the ANN method for defining through-thickness residual stress profiles in welds for application in structural integrity assessments is evaluated.
Original languageEnglish
Pages (from-to)6178-6191
Number of pages14
JournalMetallurgical and Materials Transactions A
Volume48
Issue number12
Early online date5 Oct 2017
DOIs
Publication statusPublished - Dec 2017

Bibliographical note

The final publication is available at Springer via http://dx.doi.org/10.1007/s11661-017-4359-4

Fingerprint Dive into the research topics of 'Through-Thickness Residual Stress Profiles in Austenitic Stainless Steel Welds: A Combined Experimental and Prediction Study'. Together they form a unique fingerprint.

  • Cite this