Open Collections

Abstract

Thermal modelling plays a crucial role in understanding the interaction between materials and process parameters in metal additive manufacturing. Semi-analytical models, in particular, offer a computationally efficient alternative to numerical simulations, enabling fast prediction of thermal histories in Powder Bed Fusion (PBF). However, the limiting assumptions used to derive such models hinder their accuracy and rely on empirical calibration to align with experimental results. This thesis addresses those limitations by introducing physically-based corrections that incorporate temperature-dependent material properties and radiative heat loss. These improvements significantly enhance model fidelity while preserving computational efficiency and eliminating the need for calibration. In parallel, a novel layer-truncation strategy is developed to overcome the computational bottlenecks associated with part-scale simulations. This method retains accuracy by modelling only the most recently printed layers while approximating the contribution of older layers with an analytical approximation. This approach yields faster thermal predictions, achieving more than an order-of-magnitude speed-up in computational efficiency with minimal errors compared to the full simulations. These improvements enable faster simulations of part-scale builds that would otherwise be infeasible. Building on these enhanced thermal predictions, a methodology is introduced to identify regions prone to lack of fusion or keyhole defects directly from the simulated thermal history. This enables quick screening of a given beam trajectory in order to make informed decisions. The model is then coupled with an optimization framework that dynamically adjusts the input beam power to mitigate overheating to improve build quality. Finally, the work is extended to integrate machine learning methods capable of classifying and predicting high-risk regions in complex geometries, further enabling scalable thermal risk assessment. The combined contributions of this work significantly advance the capability and practicality of semi-analytical thermal modelling for powder bed fusion. By balancing accuracy and speed, the developed framework supports fast and practical beam trajectory planning, and process control strategies for part-scale components.