Analysis of Laser Melting and Resolidification Related to Additive Manufacturing

Metal additive manufacturing (AM) has witnessed rapid development and ever-increasing interest from various industrial sectors over the last two decades. However, insufficient research into the process control, and resultant microstructure and defects is inhibiting the widespread adoption of AM technologies for critical structural components. The complexities associated with AM, such as the interplay between process parameters, thermal conditions, vapor depression dynamics, material properties, necessitate comprehensive research to optimize the manufacturing process and ensure the production of high-quality components. 

Acknowledging that the main hurdles in laser fabrication of single crystal components revolve around tackling the formation of stray grains and controlling grain orientation during the resolidification process, an extensive exploration of the impact of laser processing parameters on stray grain formation is undertaken. The objective of this endeavor is to establish a suitable processing window and validate the solidification models through the analysis of experimental data. 

In laser-based AM, high-irradiance laser interacting with metal surfaces initiates complex processes causing rapid shifts in the substrate’s optical behavior. Accurate quantification of absorbed light is crucial for understanding laser-material interactions, minimizing defects, and supporting solidification simulations. Keyhole geometry in laser melting relates to energy absorption, observed through advanced synchrotron x-ray and radiometry techniques. We propose two predictive models: an end-to-end deep learning approach using x-ray images, and a two-stage method involving geometric features and regression. This study establishes a connection between keyhole geometry and energy absorption, resulting in a validated absorption model for Ti-6Al-4V. 

Furthermore, in response to the challenges of analyzing large volumes of synchrotron imaging data, automated pipelines were developed to process the x-ray images of keyholes. These pipelines encompass segmentation and the extraction of feature attributes. In order to keep pace with the expanding amount of research papers, we showcased the viability of fine-tuning a large language model customized for the materials science field. This was achieved for tasks such as abstract classification, as well as the extraction of tables and metadata from scientific papers.