Development of a numerical modelling approach to predict residual stresses in Ti-6Al-4V linear fraction welds.

Abstract

Linear friction welding (LFW) is a solid-state joining process which has been successfully implemented to manufacture bladed-disks, chains and near-net shape components. During welding, large residual stresses are created as a consequence of a non-uniform heating of the component which can severely affect the integrity of the structure. Experimental measurement of residual stresses and temperatures on linear friction welds is difficult, so researchers have used modelling to provide a better understanding of these important characteristics. Models developed in the literature, replicate the welding process by including the oscillation of the workpieces, resulting in long computational times. Therefore, numerical models are mostly confined to 2D geometry and complex geometry cases such as keystone or bladed-disk welds are rarely considered. This thesis focuses on the development and validation of computational models capable of predicting the residual stress field developed in Ti-6Al-4V LFW without modelling the complex mechanical mixing occurring at the weld interface. Using a sequentially coupled thermo-mechanical analysis on a 3D model defined in ABAQUS, the heat was applied at the weld interface using the average heat flux post-processed from the machine data obtained during welding trials, for all the phases. The material deformation was ignored and the material expulsion is accounted for by sequentially removing rows of elements. The models were validated against thermocouples, neutron diffraction and contour method measurements. The shearing occurring at the interface while welding was found to have little effect on the final residual stress field and therefore can be omitted. The residual stress field was found to be driven by the temperature profile obtained at the end of welding, prior to cooling and by the weld interface dimensions. A low weld interface temperature, shallow thermal gradient across the weld and small weld interface dimensions should be sought to minimise the residual stress magnitude. Therefore, a low burn-off rate obtained with reduced welding frequency, amplitude and applied force should be used; however the impact of using these parameters on the microstructure and material properties may need to be considered. The modelling approach was successfully implemented on a blisk LFW and its peculiar geometry was found to have little effect on the residual stress field as the peak magnitude is driven by the overall length of the part and the thermal profile prior to cooling. Several cycles of post-weld heat treatment were also investigated for the blisk weld. The results showed that all post-weld heat treatments reduced the residual stresses, however the differences between the heat treatments on the resulting stress field was minimal. In conclusion, the thesis presents an innovative computationally efficient modelling approach capable of predicting the residual stresses within standard and complex geometry LFW.