Modelling and Analysis of Solidification Shrinkage and Defect Prediction in Metals

The dendritic structure is one of the most complex forms of crystallisation in nature and technology. In constrained crystal growth conditions, under uniform stress, packing and microsegregation, dendrites grow aligned as antiparallel as possible to the heat flow. Such directionally solidified single crystal alloys find a wide range of applications, from semi-conductors, laser optics to applications in aerospace engines. The purpose of single crystal solidification is to eliminate detrimental grain boundaries that limit the creep ductility of the component, whilst orienting the crystal structure as parallel as possible to the maximum loading direction. However, the manufacture of single crystals of high quality and of practical size is no trivial task. This is especially the case for multicomponent alloys with many constituent elements where large convective forces, non-uniform stresses and macrosegregation are present within the melt, which results in inherent casting defects. To predict and control the final microstructure, the influence of multiple thermo-physical mechanisms on both local and bulk scale must be well understood.

The present thesis is devoted to developing novel multiphysics models for the prediction and thus prevention of casting defects associated with single crystal growth. Research has shown that dendrites preferentially grow in the [001] direction, however, their growth direction may vary when influenced by non-uniform stress distribution (bending) and isotherm curvature. Currently, in computational solidification science, models do not take these into account thus fail to predict the complex variation in microstructural patterns. To solve this, a novel Fluid-Solid Interaction (FSI) formulation designed for multiphase flows is proposed and validated aiming to simulate stresses and strains in complex geometries evolving through the Phase Field (PF) method. The FSI-PF method is successfully applied to investigate stress during dendritic interface evolution revealing the potential mechanism behind defect formation and mosaicity.

To further elucidate the mechanism behind thermal contraction induced bending of the (001) plane, pattern recognition, characterisation and machine learning algorithms are built. In this work, DenMap is introduced - an automatic pattern recognition tool developed for feature extraction and optimisation of single crystal analysis. Furthermore, a Shape-Limited Primary Spacing (SLPS) algorithm is established for rapid and accurate determination of packing patterns, defects, and local primary spacing distribution within bulk single crystal microstructures. Incorporated in this work is also a foundation for 3D crystal reconstruction methodologies, advanced 3D misalignment and time-of-flight energy-resolved neutron imaging techniques. These new standardised methods facilitate the bulk investigation of large-scale deformation mechanisms such as mosaicity, low angle grain boundaries (LABs), enabling further improvement in computational approaches.