Directed energy deposition combining high-throughput technology and machine learning to investigate the composition-microstructure-mechanical property relationships in titanium alloys

Traditional approaches to alloy design, such as “trial-and-error” experiments, are costly and time-consuming in developing titanium alloys (and other alloys as well) for various requirements. Herein, we present a high-throughput technology combining the Directed energy deposition (DED) process and machine learning to elucidate the composition-microstructure-mechanical property relationships of DED new Ti-Al-V alloys. A total of 144 sets of ternary Ti-xAl-yV (0 ≤ x ≤ 11, 0 ≤ y ≤ 11, all in wt%) alloys were synthesized by DED, and the microstructure, microhardness, and yield strength of the alloys were rapidly characterized through image processing methods and instrumented micro-indentation. Backpropagation (BP) neural network models were developed to determine the microstructure parameters (average width of α-laths, W_α, and volume fraction of α-phase, V_α), microhardness, and yield strength as a function of the composition of DED Ti-Al-V alloys. The results showed that the V_α increases linearly with increasing Al content and decreases with increasing V content. However, a nonlinear relationship between W_α and contents of Al and V was found, which is mainly responsible for the nonlinear relationship between mechanical properties and composition. The approach established in this work can shed insight into developing alloys suitable for additive manufacturing.