Finite element modelling for component-level additive manufacturing applied to residual stress-deformation mitigation and fatigue crack initiation in welded connections

Additive manufacturing (AM) has attracted significant attention in many applications due to its capability of fabricating complex and customized metal parts. However, the potential for high inherent residual stresses that produce distortion in AM components and have detrimental effects on fatigue life, prevents more widespread application of the AM technique. Efficient and accurate prediction of residual stress and distortion at component level (macro-scale) is a complex task. Nowadays, the petrochemical and energy industries are evolving towards higher temperature and pressure operating conditions to improve efficiency and thereby reduce emissions, and thus help to reduce the “greenhouse effect”, as well as increased operational load cycling, to facilitate increased renewable uptake. Such increased temperature-pressure and load cycling conditions inevitably poses significant new challenges for safe design and life analysis of key high temperature components, requiring materials and structures, such as welded connections, which are resistant to thermal fatigue, creep-fatigue and thermo-mechanical fatigue. The heat affected zone (HAZ) of welded connections is particularly susceptible to fatigue crack initiation (FCI) due to the increased thermo-mechanical fatigue. This thesis presents the development of computational methods for addressing two specific challenges in relation to additive manufacturing and welding of metals, respectively. As a first step towards an efficient three-dimensional finite element (FE) methodology for thermo-mechanical simulation of additive manufacturing processes for realistic full-scale engineering components, the directed energy deposition (DED) manufacture of a realistic Ti-6Al-4V component is investigated using a recently developed AM capability of the nonlinear FE code, Abaqus. The method essentially combines the ‘element birth’ method with a layer-scaling approach for highly efficient simulation of AM processes. It is shown that the method can be implemented to achieve ...