Summary: | Titanium alloys fabricated by laser powder bed fusion (LPBF) often suffer from limited ductility because of the inherent acicular α′ martensite embedded in the columnar parent phase grains (prior-β grains). The post-built heat treatment at a relatively high temperature (∼1075 K) necessary for decomposing martensite results in improved ductility at the cost of strength. It, however, remains difficult to achieve balances between strength and ductility in as-printed conditions due to the huge range of possible compositions of printing process variables. Herein, using LPBF-processed Ti-6Al-4V (Ti64) alloy as an example, we demonstrate that machine learning (ML) is capable of accelerating the discovery of the proper sets of processing parameters resulting in a superior synergy of strength and ductility (i.e., yield strength, Ys0.2 = 1044 ± 10 MPa, uniform elongation, UEL = 10.5 ± 1.2 % and total elongation = 15 ± 1.5 %). Such property improvement is found to be enabled by an unique refined prior-β grains decorated by confined α′-colony precipitates. In particular, the uniform deformation ability of α′ martensite is improved due to the enhanced microstructure uniformity achieved by weakening variant selection. ML-based processing parameter optimization approach is thus well-positioned to accelerate the qualification of a wide range of l-PBF manufactured alloys beyond Ti-alloys.
|