Influence of welding sequences on induced residual stress and distortion in pipes. (1st August 2022)

• Record Type: Journal Article
• Authors: Moslemi, Navid; Abdi, Behzad; Gohari, Soheil; Sudin, Izman; Redzuan, Norizah; Ayob, Amran; Ahmed, Mizan; Rhee, Sehun; Burvill, Colin
• Is Part Of: Construction & building materials. Volume 342:Part A(2022)
• Publication Date: 2022-08-01
• Subjects: Welding sequence -- Residual stress -- Distortion -- Pipe welding -- Heat source parameters -- Artificial neural network -- AISI 316L stainless steel -- SYSWELD®; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2022.127995
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: A three-dimensional FE model was used to determine welding pool characteristics. Welding pool characteristics and Goldak's parameters were used to train a network. The trained network was employed to determine specified Goldak's parameters. Determined Goldak's parameters were validated with experimental tests. Welding sequences effects on residual stress and radial distortion was studied. A proposed welding sequence was compared with three conventional welding sequences. The FE residual stress results of the model were validated by experimental tests. Abstract: A three-dimensional (3D) finite element (FE) model based on thermal-elastic-plastic approach in SYSWELD® was employed to investigate the effects of welding sequences on induced residual stresses and radial distortion in AISI 316L stainless steel welded pipes. To mesh the model, the HYPERMESH® processor was utilized and the model was then exported to SYSWELD® for further thermo-mechanical analysis. The Goldak's double ellipsoidal moving heat source model was employed in SYSWELD® to obtain the volumetric heat flux density in the Gas Metal Arc Welding (GMAW) simulation process. A neural-network program was suitably trained and used to predict the parameters of Goldak's model. The obtained numerical models using predicted Goldak's parameters were then successfully validated by experimental results in terms of size of welding pool and the temperature distribution. Four different welding sequences (concepts) were investigated and compared: three quarters and one full circumferential welding. These welding sequences were anticipated to create the least residual stress and distortion in pipe welding. The results demonstrate that the welding sequences significantly influence induced residual stresses and distortions in magnitude and in distribution. For the full circumferential welding concept, which was proposed in this study, the residual stresses and distortion present more uniform distribution along the circumferential direction than those in quarter welding sequences. The magnitude of welding distortion was also significantly lower in the full circumferential concept when compared to the quarter welding concepts. The FE results of residual stresses in the full circumferential welding sequences were accurately validated by experimental tests.