journal article·2026-06-04

Machine learning paradigms with multi-strategy data augmentation for predicting anisotropic impact strength of FDM-printed composites

Shengming Huang, Qibin Fang, Jing Yu, Bowen Shi, Yanchuan Liu

Publisher page

Abstract

This study addresses two challenges in machine learning prediction of anisotropic impact strength for fused deposition modeling (FDM) printed composites, namely handling categorical variables and overcoming small sample sizes (43 samples per orientation). Two modeling strategies were evaluated and compared. The unified modeling treated build orientation as an input feature, while the orientation-specific modeling used separate models for each orientation. To address the issue of small samples in the orientation-specific models, three data augmentation techniques, including generative adversarial network (GAN), synthetic minority oversampling technique for regression (SMOTER) and gaussian process regression (GPR), are introduced. Multi-scale characterizations elucidate the physical origins of orientation-dependent failure. Results indicated that GAN-based data augmentation substantially enhanced the predictive capability of the orientation-specific model, with the most pronounced improvement observed in the on-edge orientation. Although the unified model demonstrated superior external generalization by integrating multi-orientation data, SHapley Additive exPlanations (SHAP) analysis revealed that the model predominantly relied on build orientation (81.6% contribution), potentially obscuring the contribution of other input features. In contrast, orientation-specific modeling with GAN augmentation effectively captured unique feature dependencies across directions, which is consistent with the statistical analysis. This work establishes a framework for selecting modeling paradigms when handling categorical variables in anisotropic property prediction and validates generative data augmentation for small-sample modeling in composite additive manufacturing.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Additive Manufacturing and 3D Printing Technologies · Epoxy Resin Curing Processes · Additive Manufacturing Materials and Processes

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Shantou University

Key Laboratory of Intelligent Manufacturing Technology (Shantou University), Ministry of Education, Shantou, China

Liaoning University of Technology

Dalian University of Technology

Funding

STU Scientific Research Initiation Grant

References (65)

Show source-provided references
1. 10.1007/978-1-4939-2113-310.1007/978-1-4939-2113-3 ↗
2. 10.1177/0892705722110285710.1177/08927057221102857 ↗
3. 10.1002/pc.2749810.1002/pc.27498 ↗
4. 10.3390/ma1524872210.3390/ma15248722 ↗
5. 10.1016/j.compositesa.2024.10841910.1016/j.compositesa.2024.108419 ↗
6. 10.1007/s40964-024-00731-910.1007/s40964-024-00731-9 ↗
7. 10.1007/s11665-024-10150-010.1007/s11665-024-10150-0 ↗
8. 10.1007/s11665-020-05422-410.1007/s11665-020-05422-4 ↗
9. 10.1007/s11665-025-10928-w10.1007/s11665-025-10928-w ↗
10. Rath D, Kumar S, Duvedi RK, et al. Characterization and optimization for mechanical and wear performance of reinforced Nylon-12/carbon fiber matrix. J Reinf Plast Compos 2024; 45: 1–12. https://doi.org/10.1177/07316844241307079
11. 10.1002/pc.2728510.1002/pc.27285 ↗
12. 10.1007/s00170-025-17127-w10.1007/s00170-025-17127-w ↗
13. Gültekin MŞ. Prediction of tensile strength of PLA parts produced with FDM using Taguchi L27, ANOVA, and large-scale machine learning models and optimal parameter design. J Thermoplast Compos Mater 2026; 08927057261429023. https://doi.org/10.1177/08927057261429023
14. Maurya V, Borah J, Chandrasekaran M, et al. ML-enabled prediction and optimization of tensile strength and surface characteristics in additively manufactured PETG components. J Thermoplast Compos Mater 2026; 08927057261439032. https://doi.org/10.1177/08927057261439032
15. 10.1016/j.rineng.2025.10494110.1016/j.rineng.2025.104941 ↗
16. 10.1007/s11665-024-10629-w10.1007/s11665-024-10629-w ↗
17. 10.1088/1402-4896/ad651410.1088/1402-4896/ad6514 ↗
18. El IMAY, Laaouina L, Jeghal A, et al. Modeling of energy consumption and print time for FDM 3D printing using multilayer perceptron network. J Manuf Mater Process 2023; 7(4): 128. https://doi.org/10.3390/jmmp7040128
19. 10.1007/s00170-024-14044-210.1007/s00170-024-14044-2 ↗
20. 10.1016/j.commatsci.2023.11214010.1016/j.commatsci.2023.112140 ↗
21. 10.1002/app.5753310.1002/app.57533 ↗
22. 10.1007/s11837-026-08302-910.1007/s11837-026-08302-9 ↗
23. 10.3390/technologies1207011210.3390/technologies12070112 ↗
24. 10.1016/j.ese.2022.10017210.1016/j.ese.2022.100172 ↗
25. 10.3390/app1302122710.3390/app13021227 ↗
26. 10.1038/s41598-023-38527-610.1038/s41598-023-38527-6 ↗
27. 10.1115/1.406236810.1115/1.4062368 ↗
28. Fang Q, Yu J, Liu Y, et al. Machine learning and SHAP for predicting impact strength of FDM-printed short carbon fiber-reinforced nylon. J Reinf Plast Compos 2026; 07316844261421791. https://doi.org/10.1177/07316844261421791
29. Royston P. Polynomial regression. In: Balakrishnan N, Colton T, Everitt B, et al. (eds). Encyclopedia of Biostatistics. 2nd ed. Chichester: John Wiley & Sons, Ltd., 2014. https://doi.org/10.1002/9781118445112.stat05771
30. 10.3390/app1416700910.3390/app14167009 ↗
31. 10.1111/j.1467-9868.2005.00503.x10.1111/j.1467-9868.2005.00503.x ↗
32. 10.22214/ijraset.2025.7313710.22214/ijraset.2025.73137 ↗
33. 10.1080/00207543.2018.150505810.1080/00207543.2018.1505058 ↗
34. 10.3390/technologies1306022810.3390/technologies13060228 ↗
35. Shanto TA. Predicting mechanical strength in FDM printed ABS parts with in-process annealing: a machine learning approach. Research Square 2025, [preprint]. https://doi.org/10.21203/rs.3.rs-7437589/v1
36. 10.1038/s41524-020-00392-610.1038/s41524-020-00392-6 ↗
37. Rasmussen CE, Williams CKI. Gaussian processes for machine learning. Cambridge, MA: MIT Press, 2006. https://www.gaussianprocess.org/gpml/chapters
38. 10.1007/s40033-024-00735-310.1007/s40033-024-00735-3 ↗
39. 10.1016/j.matdes.2018.07.00110.1016/j.matdes.2018.07.001 ↗
40. 10.1007/s00170-023-11364-710.1007/s00170-023-11364-7 ↗
41. 10.1039/C7SM00950J10.1039/C7SM00950J ↗
42. 10.3390/polym1316267710.3390/polym13162677 ↗
43. 10.1016/j.addma.2020.10165810.1016/j.addma.2020.101658 ↗
44. 10.1007/s13726-024-01315-810.1007/s13726-024-01315-8 ↗
45. 10.1016/j.polymer.2017.06.05110.1016/j.polymer.2017.06.051 ↗
46. 10.3390/polym1523451810.3390/polym15234518 ↗
47. 10.1016/j.jmatprotec.2017.02.02610.1016/j.jmatprotec.2017.02.026 ↗
48. 10.3390/thermo103002110.3390/thermo1030021 ↗
49. 10.1016/j.addma.2021.10209710.1016/j.addma.2021.102097 ↗
50. 10.1007/s11665-021-05624-410.1007/s11665-021-05624-4 ↗
51. 10.1108/1355254081086202810.1108/13552540810862028 ↗
52. 10.1002/pen.2559010.1002/pen.25590 ↗
53. 10.1002/app.4878610.1002/app.48786 ↗
54. 10.1007/s00170-025-16232-010.1007/s00170-025-16232-0 ↗
55. 10.1016/j.compstruct.2022.11637010.1016/j.compstruct.2022.116370 ↗
56. 10.1016/j.compositesb.2022.11037010.1016/j.compositesb.2022.110370 ↗
57. 10.1038/s41524-019-0153-810.1038/s41524-019-0153-8 ↗
58. 10.1016/j.array.2022.10025810.1016/j.array.2022.100258 ↗
59. 10.1016/j.mtcomm.2026.11502110.1016/j.mtcomm.2026.115021 ↗
60. 10.1038/s41524-023-01000-z10.1038/s41524-023-01000-z ↗
61. 10.1007/s44244-024-00021-x10.1007/s44244-024-00021-x ↗
62. 10.3389/fmtec.2024.132016610.3389/fmtec.2024.1320166 ↗
63. 10.1016/j.cma.2024.11690710.1016/j.cma.2024.116907 ↗
64. 10.1016/j.jcp.2018.10.04510.1016/j.jcp.2018.10.045 ↗
65. Elumalai B, Kumar AP, Sahu SK, et al. Physics-informed machine learning for predicting specific energy absorption in additively manufactured tetra-chiral auxetic lattice structures. J Thermoplast Compos Mater. 2026; 08927057261448114. https://doi.org/10.1177/08927057261448114