Property optimization of TRIP Ti alloys based on artificial neural network

Abstract Transformation-induced plasticity (TRIP) Ti alloys are promising structural materials that offer high strength and ductility. However, these alloys often include heavy, expensive, and high-melting-point β-stabilizing elements such as V, Nb, Mo, and W. Herein, an artificial neural network (ANN) was used to develop a Ti–Al–Fe–Mn-based TRIP alloy comprising lighter and/or cheaper elements. The ANN model was trained with 30 experimental tensile datasets for heat-treated (830–920 °C) Ti–4Al–2Fe–xMn (x = 0–4 wt%) alloys, and used to generate 400 tensile datasets with more finely tuned composition and temperature intervals. Based on the predicted data, an 883 °C-heat-treated Ti–4Al–2Fe–1.4Mn alloy was produced (conditions not used in the training datasets), which exhibited ultra-high specific strength (289 MPa·cm3/g) and high elongation (34%). Thus, the ANN approach successfully led to the development of a new alloy while minimizing the number of labor-intensive and time-consuming experiments.