Page 17 - IJAMD-2-2

International Journal of AI for
Materials and Design AI-driven material development for AM

Figure 6. Link between additive manufacturing process parameters and material microstructure and properties. Adapted from Jin et al. 10

5.1. Metallic materials for AM algorithms to optimize the composition of Fe–Ni–Ti–Al
Table 4 presents a summary of AI applications related to maraging steel, tailoring it for AM. Specifically, this study
metal materials in AM. The following sections will explore utilized thermodynamic simulations to generate a dataset,
representative examples to elucidate how AI facilitates which was subsequently integrated with RF, decision trees,
material design and performance optimization. AdaBoost, and k-nearest neighbor to predict the formation
of Ni₃Ti precipitates and Laves phases during AM. Based
5.1.1. Alloy design for metal AM on these predictions, the concentrations of Ni, Ti, and Al
A prominent success of AI-driven alloy design is its were adjusted to enhance precipitation strengthening while
application in phase prediction for high-entropy alloys. mitigating the formation of detrimental phases, thereby
Due to their multi-principal element nature, high-entropy improving microstructural stability and mechanical
alloys exhibit a vast range of possible phase structures, properties. Finally, the newly developed maraging steel
which directly influence their mechanical, thermal, and exhibits exceptional mechanical properties, achieving a
chemical properties. A variety of ML models have been tensile strength of 1,538 MPa and a uniform elongation of
67
employed to classify high-entropy alloy phases, employing 8.1%, validating the effectiveness of the AI-driven material
diverse feature sets to improve prediction accuracy. development approach.
Commonly used models include ANNs, 68-70 RF, 71,72 support Overall, while AI-driven methodologies have
vector machine, 73,74 logistic regression, gradient-boosted demonstrated significant potential in alloy design, the
75
trees, and Gaussian process classification. These models direct application of AI in AM-specific alloy design is
77
76
utilize input features, such as atomic size difference (δ), still rare. Most studies still focus on general material
valence electron concentration, mixing enthalpy (H ), and design, where AI is primarily employed for composition
mix
entropy parameters, to predict phase formation, including optimization and phase prediction, without explicitly
78
79
face-centered cubic, body-centered cubic, hexagonal close- considering the constraints imposed by AM processing.
packed, and multiphase structures. To enhance the applicability of AI-driven alloy design
Beyond high-entropy alloys, AI has also been employed for AM, it is essential to further integrate it with process-
to optimize the composition of green maraging steels aware models that account for rapid solidification, thermal
for AM. As seen in Figure 7, Tan et al. employed ML cycling, and process-induced defects.
58

Volume 2 Issue 2 (2025) 11 doi: 10.36922/IJAMD025100007