Active learning framework to optimize process parameters for additive-manufactured Ti-6Al-4V with high strength and ductility
Jeong Ah Lee, Jaejung Park, Man Jae SaGong, Soung Yeoul Ahn, Jung‐Wook Cho, Seung‐Chul Lee, Hyoung Seop Kim
Abstract
Optimizing process and heat-treatment parameters of laser powder bed fusion for producing Ti-6Al-4V alloys with high strength and ductility is crucial to meet performance demands in various applications. Nevertheless, inherent trade-offs between strength and ductility render traditional trial-and-error methods inefficient. Herein, we present Pareto active learning framework with targeted experimental validation to efficiently explore vast parameter space of 296 candidates, pinpointing optimal parameters to augment both strength and ductility. All Ti-6Al-4V alloys produced with the pinpointed parameters exhibit higher ductility at similar strength levels and greater strength at similar ductility levels compared to those in previous studies. By improving one property without significantly compromising the other, the framework demonstrates efficiency in overcoming the inherent trade-offs. Ultimately, Ti-6Al-4V alloys with ultimate tensile strength and total elongation of 1190 MPa and 16.5%, respectively, are produced. The proposed framework streamlines discovery of optimal processing parameters and promises accelerated development of high-performance alloys. Process and heat treatment parameters are critical in laser powder bed fusion but difficult to optimize. Here, the authors achieve high tensile strengths and elongation in Ti-6Al-4V alloys using Pareto active learning with few experimental trials.