Finite element analysis of localised rolling to reduce residual stress and distortion

Abstract

Fusion welding processes cause residual stress due to the uneven heat distribution produced by the moving welding torch. These residual stresses are characterised by a large tensile component in the welding direction. Due to the self-equilibrated nature of the residual stress, compressive ones are present in the far field next to the weld seam, which can cause different kind of distortion such as bending or buckling. Welding residual stress can be responsible of premature failure of the components, such as stress crack corrosion, buckling, and reduction of fatigue life. Localised rolling is a stress engineering technique that can be used to reduce the residual stress and distortion caused by welding. It induces plastic strain in the rolling direction, counteracting the plastic strain produced during welding. In this thesis three techniques were investigated, pre-weld rolling, post-weld rolling, and in situ rolling. These techniques have been seldom studied in the past, particularly pre-weld rolling; consequently the mechanisms are poorly understood. Finite element models allow stress and strain development during both welding and rolling processes to be better understood, providing an improved understanding of the mechanisms involved and aiding process development. A literature survey was done to find the state of the art of the computational welding mechanics simulations, stress management, and the residual stress measurement techniques, as well as the knowledge gaps such as, the thermal losses through the backing-bar in the thermal simulation, the frictional interaction in the rolling process, and the material properties of the steel used in the models. In the literature not many models that investigate the management of welding residual stress were found. After this, the general considerations and assumptions for the welding thermal mechanical models presented in this thesis were discussed. The effect of different backing-bar conditions, as well as different material properties where investigated. Both influenced the residual stress profile to varying degrees. In particular, temperature dependent heat loss to the backing-bar was necessary to capture the improved heat loss near the weld. The distortion predicted by the model was investigated to determine whether it was due to bending or buckling phenomena. Lastly, the temperature distribution and residual stress predictions were validated against thermocouple and neutron diffraction measurements conducted by Coules et al. [1–3]. Pre-weld rolling was the first of the three rolling methods considered, in which rolling is applied to the plates before performing GMA butt-welds. The principle behind this technique consisted in inducing tensile residual stress in the weld region before welding; therefore, it is similar to mechanically tensioning the weld, which can significantly reduce the residual stress and distortion. However, there was no significant change in the tensile residual stresses. On the other hand, it was possible to achieve a small reduction in the distortion, when the plates were rolled on the opposite surface to the weld; rolling in this way induced distortion in the opposite direction to the distortion induced by welding, reducing the magnitude of the latter. These results were compared with experiments conducted by Coules et al. [1,4]. A subsequent investigation combined pre-weld rolling with post-weld heating. With this additional process the residual stress and distortion were significantly reduced, and flatter residual stress profile was achieved. The post-weld rolling and in situ rolling techniques were discussed afterwards. In the post-weld rolling models, rolling was applied after the weldment was cooled to room temperature. In in situ rolling the roller was applied on top of the weld bead at some distance behind the torch, while it was still hot. The principle behind these techniques consisted in applying positive plastic strain to the weld bead region by a roller, counteracting the negative plastic strains produced in the welding process. Two roller profiles were investigated, namely, grooved, and double flat rollers. The post-weld rolling on top of the weld bead models, which used the grooved roller, showed good agreement against experimental results, producing a large reduction of the residual stress and distortion. Some discrepancies were present when the weld toes were rolled with the dual flat roller. The former roller was more efficient for reducing residual stress and distortion. The influence of different friction coefficients (between the roller and weldment, and between the backing-bar and the weldment), were investigated. It showed significant dependency on the residual stress distribution when high rolling loads were used. The frictional interaction constrained the contact area inducing more compressive stress in the core of the weld bead; therefore it produced more tensile residual stress in the surface of the weldment. Additionally, the influence of rolling parameters on the through-thickness residual stress variation was investigated. Low loads only influence the residual stress near the surface, while high loads affected the material through the entire thickness. When the dual flat roller was used to roll next to the weld bead, significant compressive residual stress was induce in the weld bead; however, the residual stress reduction was very sensitive to the contact of the roller to the weld toes; therefore, when rolling a weld bead that varies in shape along the weld, the residual stress reduction is not uniform and varies along the length. On the other hand, the in situ rolling did not produced significant residual stress or distortion reduction in all the cases analysed. The rolling occurred when the material was still hot and the residual stress was subsequently formed as the material cooled to room temperature. Numerical modelling was a very useful tool for understanding the development of stress and plastic strain during the welding and rolling processes.