A data-efficient machine learning approach for amorphous Fe-based bulk metallic glass fabrication in powder bed fusion

Abstract

The widespread adoption of bulk metallic glasses (BMGs) in aerospace and biomedical industries requires topology-optimized architectures that conventional manufacturing cannot achieve. In response, BMGs have been investigated for powder bed fusion (PBF), but the process remains challenging due to narrow thermal windows, expensive feedstock, and limited data. This study introduces a constrained multi-objective Bayesian optimization framework to optimize key PBF printing parameters, including laser power and scan speed, to maximize hardness while preserving the amorphous state of the printed BMG. Hardness is optimized as the primary objective with density incorporated in the scalarization to regularize the search space, and amorphous retention is enforced through a feasibility probability learned by a logistic classifier. Surrogate models are compared, including Gaussian process, Bayesian additive regression trees, Bayesian multivariate adaptive regression splines (BMARS), and a Bayesian attention neural network. Acquisition scores are computed with constrained expected improvement and are maximized on a uniform grid over power and velocity. Superior predictive accuracy is obtained with BMARS, and 95% credible intervals are calibrated to the measurements. A high-hardness region at high power and low velocity is localized by the surrogates. A fully amorphous sample at 60 W and 1300 mm/s is produced, and a hardness of 1010.4 HV is measured in agreement with the predicted high-hardness band. In conclusion, the study establishes a data-efficient process-window discovery method for BMG PBF, produces an interpretable process map, and demonstrates a screening framework suitable for constrained experimental budgets.

Keywords

Additive manufacturing / Bayesian optimization / Bulk metallic glass / Powder bed fusion

Acknowledgments

This research is supported by the Singapore Centre for 3D Printing (SC3DP) and the Ministry of Trade, Industry and Energy (MOTIE, Korea) through the 2025 R&D HRD Program for Problem-Solving Based on Future Mobility Innovation Technology Development (Project No. P0023726).

Funding

This research is conducted by the Industrial Technology Innovation Program (KEIT project no. 20024344, Development of AI-based high carbon steel alloy design and sintering-based additive manufacturing technology for 7.0 L/Hr-level high-speed production of powertrain components with tensile strength over 1.0 GPa in the next-generation mobility) funded by the Ministry of Trade, Industry & Energy of the Republic of Korea.

Conflict of interest

Seung Ki Moon is an Editorial Board Member of this journal, but was not in any way involved in the editorial and peer-review process conducted for this paper, directly or indirectly. Separately, other authors declared that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper.

Author contributions

Conceptualization: Jungyeon Kim

Formal analysis: Jungyeon Kim, Sangjun Jeon

Funding acquisition: Seung Ki Moon

Investigation: Jungyeon Kim, Seong Je Park

Methodology: Jungyeon Kim, Sangjun Jeon, Seong Je Park

Software: Jungyeon Kim

Writing-original draft: Jungyeon Kim, Sangjun Jeon, Seong Je Park

Writing-review & editing: Seong Je Park, Seung Ki Moon

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data

The code and datasets supporting the findings of this study are available from the corresponding author upon reasonable request.

References

[1] Tofail SA, Koumoulos EP, Bandyopadhyay A, Bose S, O’Donoghue L, Charitidis C. Additive manufacturing: Scientific and technological challenges, market uptake and opportunities. Mat Today. 2018; 21(1):22-37. doi: 10.1016/j.mattod.2017.07.001

[2] Kumar R, Kumar M, Chohan JS. The role of additive manufacturing for biomedical applications: A critical review. J Manuf Process. 2021; 64:828-850. doi: 10.1016/j.jmapro.2021.02.022

[3] Blakey-Milner B, Gradl P, Snedden G, et al. Metal additive manufacturing in aerospace: A review. Mater Des. 2021; 209:110008. doi: 10.1016/j.matdes.2021.110008

[4] Rehman M, Yanen W, Mushtaq RT, et al. Additive manufacturing for biomedical applications: A review on classification, energy consumption, and its appreciable role since COVID-19 pandemic. Prog Addit Manuf. 2023; 8(5):1007-1041. doi: 10.1007/s40964-022-00373-9

[5] ISO/ASTM 52900:2021(en), Additive Manufacturing -General Principles — Fundamentals and vocabulary.Available from: https://www.iso.org/obp/ui/#iso:std:iso.astm:52900:ed-2:v1:en [Last accessed on 2025 Aug 10].

[6] Wang J, Jeong SG, Kim ES, Kim HS, Lee BJ. Material-agnostic machine learning approach enables high relative density in powder bed fusion products. Nat Commun. 2023; 14(1):6557. doi: 10.1038/s41467-023-42319-x

[7] Park SJ, Heogh W, Yang J, et al. Meta-structure of amorphous-inspired 65.1Co28.2Cr5.3Mo lattices augmented by artificial intelligence. Adv Compos Hybrid Mater. 2024; 7(6):1-22. doi: 10.1007/s42114-024-01039-6

[8] Park SJ, Lee JH, Yang J, et al. Lightweight injection mold using additively manufactured Ti-6Al-4V lattice structures. J. Manuf Process. 2022; 79:759-766. doi: 10.1016/j.jmapro.2022.05.022

[9] Howard J, Carlson K, Chidambaram D. High-temperature metallic glasses: Status, needs, and opportunities. Phys Rev Mater. 2021; 5(4):040301. doi: 10.1103/physrevmaterials.5.040301

[10] Scully JR, Gebert A, Payer JH. Corrosion and related mechanical properties of bulk metallic glasses. J Mater Res. 2007; 22(2):302-313. doi: 10.1557/jmr.2007.0051

[11] Xu T, Pang S, Li H, Zhang T. Corrosion resistant Cr-based bulk metallic glasses with high strength and hardness. J Non Cryst Solids. 2015; 410:20-25. doi: 10.1016/j.jnoncrysol.2014.12.006

[12] Chen M. A brief overview of bulk metallic glasses. NPG Asia Mater. 2011; 3(9):82-90. doi: 10.1038/asiamat.2011.30

[13] Mi XL, Hu L, Wan ZX, Wu BW, Wei B. Liquid state properties and amorphous solidification kinetics of multicomponent Fe50-𝑥Co𝑥Cr14Mo14C9B8Tm 5 alloys investigated under containerless processing conditions. Phys Rev E. 2024; 110(3):034612. doi: 10.1103/physreve.110.034612

[14] Mahbooba Z, Thorsson L, Unosson M, et al. Additive manufacturing of an iron-based bulk metallic glass larger than the critical casting thickness. Appl Mater Today. 2018; 11:264-269. doi: 10.1016/j.apmt.2018.02.011

[15] Hooper PA. Melt pool temperature and cooling rates in laser powder bed fusion. Addit Manuf. 2018; 22:548-559. doi: 10.1016/j.addma.2018.05.032

[16] Yang Z, Markl M, Körner C. Comprehensive numerical investigation of laser powder bed fusion process conditions for bulk metallic glasses. Addit Manuf. 2024; 81:104026. doi: 10.1016/j.addma.2024.104026

[17] Li B, Yakubov V, Nomoto K, et al. Superior mechanical properties of a Zr-based bulk metallic glass via laser powder bed fusion process control. Acta Mater. 2024; 266:119685. doi: 10.1016/j.actamat.2024.119685

[18] Chen C, Fan Y, Zhang W, et al. Tailoring Nano-crystallization in Zr50Ti4Y1Al10Cu25Ni7Co2Fe1 complex multicomponent bulk metallic glass by O doping. J Non Cryst Solids. 2021; 553:120474. doi: 10.1016/j.jnoncrysol.2020.120474

[19] Chen Y, Zhang D, O’Toole P, et al. In situ observation and reduction of hot-cracks in laser additive manufacturing. Commun Mater. 2024; 5(1):84. doi: 10.1038/s43246-024-00522-3

[20] Ansari MA, Crampton A, Garrard R, Cai B, Attallah M. A convolutional neural network (CNN) classification to identify the presence of pores in powder bed fusion images. Int J Adv Manuf Technol. 2022; 120(7):5133-5150. doi: 10.1007/s00170-022-08995-7

[21] Westphal E, Seitz H. A machine learning method for defect detection and visualization in selective laser sintering based on convolutional neural networks. Addit Manuf. 2021; 41:101965. doi: 10.1016/j.addma.2021.101965

[22] Mozaffar M, Paul A, Al-Bahrani R, et al. Data-driven prediction of the high-dimensional thermal history in directed energy deposition processes via recurrent neural networks. Manuf Letters. 2018; 18:35-39. doi: 10.1016/j.mfglet.2018.10.002

[23] Chen Z, Mak S, Wu CF. A hierarchical expected improvement method for Bayesian optimization. J Am Stat Assoc. 2024; 119(546):1619-1632. doi: 10.1080/01621459.2023.2210803

[24] Lei B, Kirk TQ, Bhattacharya A, et al. Bayesian optimization with adaptive surrogate models for automated experimental design. NPJ Comput Mater. 2021; 7(1):194. doi: 10.1038/s41524-021-00662-x

[25] Zhang H, Chen WW, Iyer A, Apley DW, Chen W. Uncertainty-aware mixed-variable machine learning for materials design. Sci Rep. 2022; 12(1):19760. doi: 10.1038/s41598-022-23431-2

[26] Ament S, Daulton S, Eriksson D, Balandat M, Bakshy E. Unexpected Improvements to Expected Improvement for Bayesian Optimization. [arXiv Preprint]; 2025. doi: 10.48550/arXiv.2310.20708

[27] Johnson JE, Jamil IR, Pan L, Lin G, Xu X. Bayesian optimization with Gaussian-process-based active machine learning for improvement of geometric accuracy in projection multi-photon 3D printing. Light Sci Appl. 2025; 14(1):56. doi: 10.1038/s41377-024-01707-8

[28] Karkaria V, Goeckner A, Zha R, et al. Towards a digital twin framework in additive manufacturing: Machine learning and Bayesian optimization for time series process optimization. J Manuf Syst. 2024; 75:322-332. doi: 10.1016/j.jmsy.2024.04.023

[29] Frazier PI. A Tutorial on Bayesian Optimization. [arXiv Preprint]; 2018. doi: 10.48550/arXiv.1807.02811

[30] Laghi V, Palermo M, Bruggi M, Gasparini G, Trombetti T. Blended structural optimization for wire-and-arc additively manufactured beams. Prog Addit Manuf. 2023; 8(3):381-392. doi: 10.1007/s40964-022-00335-1

[31] Kavas B, Balta EC, Tucker MR, et al. In-situ Controller Autotuning by Bayesian Optimization for Closed-loop Feedback Control of Laser Powder Bed Fusion Process. Available from: https://arxiv.org/html/2406.19096v1 [Last accessed on 2025 Aug 11].

[32] Squires L, Roberts E, Bandyopadhyay A. Radial bimetallic structures via wire arc directed energy deposition-based additive manufacturing. Nat Commun. 2023; 14(1):3544. doi: 10.1038/s41467-023-39230-w

[33] Yang Z, Markl M, Körner C. Predictive simulation of bulk metallic glass crystallization during laser powder bed fusion. Addit Manuf. 2022; 59:103121. doi: 10.1016/j.addma.2022.103121

[34] Laws KJ, Miracle DB, Ferry M. A predictive structural model for bulk metallic glasses. Nat Commun. 2015; 6(1):8123. doi: 10.1038/ncomms9123

[35] Ding J, Ma E. Computational modeling sheds light on structural evolution in metallic glasses and supercooled liquids. NPJ Comput Mater. 2017; 3(1):9. doi: 10.1038/s41524-017-0007-1

[36] Xu D, Johnson WL. Crystallization kinetics and glass-forming ability of bulk metallic glasses Pd40Cu30Ni10P20 and Zr41.2Ti13.8Cu12.5Ni10Be22.5 from classical theory. Phys Rev B. 2006; 74(2):024207. doi: 10.1103/physrevb.74.024207

[37] Bordeenithikasem P, Liu J, Kube SA, et al. Determination of critical cooling rates in metallic glass forming alloy libraries through laser spike annealing. Sci Rep. 2017; 7(1):7155. doi: 10.1038/s41598-017-07719-2

[38] Lázaro-Gredilla M, Titsias MK. Variational Heteroscedastic Gaussian process Regression. In: Proceedings of the 28^thInternational Conference on International Conference on Machine Learning. ICML’11. Washington, DC: Omnipress; 2011. p. 841-848.

[39] Heinonen M, Mannerström H, Rousu J, Kaski S, Lähdesmäki H. Non-Stationary Gaussian Process Regression with Hamiltonian Monte Carlo. [arXiv Preprint]; 2015. doi: 10.48550/arXiv.1508.04319

[40] Jospin LV, Buntine W, Boussaid F, Laga H, Bennamoun M. Hands-on Bayesian Neural Networks -- a Tutorial for Deep Learning Users. IEEE Comput Intell Mag. 2022; 17(2):29-48. doi: 10.1109/mci.2022.3155327

[41] Gal Y, Ghahramani Z. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. [arXiv Preprint]; 2016. doi: 10.48550/arXiv.1506.02142

[42] Knowles J. ParEGO: A hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization problems. IEEE Trans Evol Comput. 2006; 10(1):50-66. doi: 10.1109/tevc.2005.851274

[43] Hernández-Lobato JM, Gelbart MA, Adams RP, Hoffman MW, Ghahramani Z. A General Framework for Constrained Bayesian Optimization using Information-Based Search. [arXiv Preprint]; 2016. doi: 10.48550/arXiv.1511.09422

[44] Wong TT. Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation. Pattern Recognit. 2015; 48(9):2839-2846. doi: 10.1016/j.patcog.2015.03.009

[45] Vehtari A, Gelman A, Gabry J. Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Stat Comput. 2017; 27(5):1413-1432. doi: 10.1007/s11222-016-9696-4