Abstract Scope |
Additive manufacturing (AM), particularly laser powder bed fusion (LPBF), streamlines intricate component 3D printing in one step, omitting the need for assembly, and minimizing material and energy waste while cutting production time. Crucial in the AM-LPBF manufacturing process is alloy printability which implies the ability to resist solidification cracking impacted by composition, process parameters, and thermophysical properties. Despite the optimized process parameters employed during AM-LPBF, some alloys remain prone to cracking. Current research uses kinetic, heat transfer, fluid flow models, simulations, and experiments to analyze the compositional impact on solidification cracking, which are computationally intensive and costly. To overcome these constraints, this study employs six traditional machine-learning models, with extreme gradient boost (XGB), rapidly predicting alloy printability, achieving high accuracy, recall, and precision scores of 0.9185, 0.914, and 0.9412 respectively. To further optimize and validate the chosen ML model, alloy composition generated from variational autoencoder (VAE) and thermo-calc software will be evaluated by the model to predict novel alloys with outstanding printability properties for AM-LBPF processes. |