Page 18 - IJAMD-2-2

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

Table 4. Summary of AI applications for metal materials for AM
AM Material Optimization AI method Target Model References
process type performance
DED Fe–Ni–Ti–Al Design RF Composition optimization R2=0.998 Tan et al. 58
MAE=0.292
PBF SS316L Performance Adaptive Neuro-Fuzzy Fatigue life prediction RMS=14.66% Zhang et al. 59
Inference System
(ANFIS)
PBF Zr52.5Cu17.9Ni14.6Al10Ti Performance HGP Materials characteristics RMSE=2.58% Chernyavsky
prediction et al. 60
PBF AlSi10Mg Performance GPR Tensile property optimization - He et al. 61
(YS and elongation)
PBF AlSi10Mg Performance GPR Density variations and - Liu et al. 62
microstructural characteristics
prediction
PBF Ti-6Al-4V Performance MML Fatigue strength design - Awd et al. 63
PBF Ti-6Al-4V Performance ANN Tensile property optimization R2: YS=0.9887, Maleki et al. 64
(YS, UTS, and elongation) UTS=0.9921,
elongation=0.9917
PBF Ti-6Al-4V Performance RSM+GA Energy absorption optimization R2=0.9431 Meng et al. 65
DED Ti–Mn alloy Performance GPR YS and modulus optimization MAPE: YS=6.26%, Gong et al. 66
E=2.02%
Abbreviations: ANN: Artificial neural networks; DED: Directed energy deposition; E: Elastic modulus; GA: Genetic algorithms; GPR: Gaussian process
regression; HGP: Heteroscedastic Gaussian process; MAE: Mean absolute error; MML: Mechanistic machine learning; PBF: Powder bed fusion;
RF: Random forest; RMS: Root mean square; RMSE: Root mean square error; RSM: Response surface methodology; UTS; Ultimate tensile strength;
YS: Yield strength; AI: Artificial intelligence; AM: Additive manufacturing.

A B C

F E D

Figure 7. The schematic of ML-assisted composition design of Fe–Ni–Ti–Al NMS. (A) Feature selections in the design of NMS. (B) Data collections from
Thermo-Calc software and the correlation matrix of the input composition (Ni, Ti, and Al) and output (Ni Ti precipitate and Laves phase weight fractions)
3
in the surrogate models. (C) ML by various algorithms (random forest is the most accurate one). (D) Composition optimization for the allowable range of
alloying elements. (E) Time-dependent dynamic precipitation behaviors of different compositions at 490°C (the balance is Fe). (F) Optimal composition
Fe-20.8Ni-6.2Ti-1.7Al (wt%) along with the morphology and elemental mapping of the produced powder. Reproduced from Tan et al. 58
Abbreviations: NMS: Novel maraging steel; ML: Machine learning.

5.1.2. Performance optimization in metal AM optimization. Microstructural optimization focuses
Alloy performance optimization includes both on the regulation of grain size, phase distribution,
microstructural optimization and mechanical property precipitate morphology, and porosity to achieve improved

Volume 2 Issue 2 (2025) 12 doi: 10.36922/IJAMD025100007