journal article·2026-07

Optimizing 3D printed continuous CF/PEEK composites: A machine learning approach to strength prediction

Pengfei Liu, Baoning Chang, Chongxiao Xu, Wuxiang Zhang, Tong Yang, Tao Wu

Publisher page

Abstract

This study employs machine learning (ML) algorithms to predict the tensile strength of 3D printed continuous carbon fiber reinforced polyether ether ketone (CF/PEEK) thermoplastic composites, addressing the critical need for rapid and cost-effective material qualification in additive manufacturing. A total of 120 samples were fabricated using fused deposition modeling (FDM), with systematic variations in key printing parameters, including printing path design (unidirectional and orthogonal printing), printing speed (270–600 mm/min), and nozzle temperature (375–410°C). The experimental tensile strength of the composites ranged from 310 MPa to 976 MPa. A Gradient Boosted Decision Tree (GBDT) algorithm was utilized to establish correlations between printing parameters and tensile properties, achieving a comprehensive determination coefficient exceeding 0.8, demonstrating high predictive accuracy. Cross-validation and grid search methods were employed to optimize model parameters, resulting in robust generalization capabilities across training, testing, and validation datasets. Through prediction and validation using ML models, the optimal parameter combination was identified as follows: unidirectional printing, a printing speed of 330 mm/min, and a nozzle temperature of 380°C. The study reveals that the printing path has the most significant impact on tensile strength (relative importance: 0.8), followed by printing speed (0.15) and nozzle temperature (0.05). This research provides an efficient data-driven approach to predict and optimize the mechanical properties of 3D printed composites, aligning with industrial demands for sustainable, lightweight materials in aerospace, automotive, and energy sectors, and offering a scalable solution to accelerate the development and application of 3D printed thermoplastic composites.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Additive Manufacturing and 3D Printing Technologies · Advanced Sensor and Energy Harvesting Materials · Manufacturing Process and Optimization

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

Affiliations

Zhejiang University

Beihang University

Dalian University of Technology

Technische Universität Dresden

References (37)

Show source-provided references
1. Zhai Y Chen Y Zhao Q et al. A review on process parameter influence and optimization for 3D printing of fiber-reinforced thermoplastic composites. J Reinforc Plast Compos 2024; 1–19 (In Press).10.1177/07316844241273038 ↗
2. 10.1016/j.compscitech.2018.05.03210.1016/j.compscitech.2018.05.032 ↗
3. 10.1016/j.jmapro.2025.04.03510.1016/j.jmapro.2025.04.035 ↗
4. 10.1016/j.tws.2021.10771710.1016/j.tws.2021.107717 ↗
5. 10.1016/j.compositesa.2023.10743410.1016/j.compositesa.2023.107434 ↗
6. 10.1016/j.compositesa.2021.10653210.1016/j.compositesa.2021.106532 ↗
7. 10.1016/j.compositesa.2019.03.02010.1016/j.compositesa.2019.03.020 ↗
8. 10.1016/j.jmrt.2022.09.00510.1016/j.jmrt.2022.09.005 ↗
9. 10.1016/j.addma.2024.10405410.1016/j.addma.2024.104054 ↗
10. 10.1016/j.compscitech.2022.10961810.1016/j.compscitech.2022.109618 ↗
11. 10.1016/j.matdes.2022.11150510.1016/j.matdes.2022.111505 ↗
12. 10.1016/j.compscitech.2019.10790510.1016/j.compscitech.2019.107905 ↗
13. 10.1016/j.compscitech.2024.11072310.1016/j.compscitech.2024.110723 ↗
14. 10.1016/j.compstruct.2023.11691410.1016/j.compstruct.2023.116914 ↗
15. 10.1016/j.compositesa.2023.10745410.1016/j.compositesa.2023.107454 ↗
16. 10.1016/j.compscitech.2022.10972710.1016/j.compscitech.2022.109727 ↗
17. 10.1016/j.compositesa.2022.10716210.1016/j.compositesa.2022.107162 ↗
18. 10.1016/j.compstruct.2022.11539110.1016/j.compstruct.2022.115391 ↗
19. 10.1016/j.compstruct.2022.11524110.1016/j.compstruct.2022.115241 ↗
20. 10.1016/j.compositesa.2024.10833310.1016/j.compositesa.2024.108333 ↗
21. Shen XW. Simulation analysis and state recognition method research on warpage deformation in 3D printing of fused deposition molding. PhD Thesis. Hangzhou, China: Zhejiang University, 2018. [In Chinese].
22. 10.1016/j.compositesb.2020.10789310.1016/j.compositesb.2020.107893 ↗
23. 10.1016/j.compositesb.2020.10837310.1016/j.compositesb.2020.108373 ↗
24. 10.1016/j.compscitech.2021.10896210.1016/j.compscitech.2021.108962 ↗
25. 10.1016/j.jmrt.2021.02.03010.1016/j.jmrt.2021.02.030 ↗
26. 10.1016/j.compositesb.2018.06.01710.1016/j.compositesb.2018.06.017 ↗
27. 10.1007/s10462-020-09876-910.1007/s10462-020-09876-9 ↗
28. 10.1093/jcde/qwaa09410.1093/jcde/qwaa094 ↗
29. 10.1016/j.mtcomm.2022.10398510.1016/j.mtcomm.2022.103985 ↗
30. 10.1016/j.compositesb.2023.11095810.1016/j.compositesb.2023.110958 ↗
31. Polyzos E, Nikolaou C, Polyzos D, et al. Direct modeling of the elastic properties of single 3D printed composite filaments using X-ray computed tomography images segmented by neural networks. Addit Manuf 2023; 76: 103786.
32. 10.1177/0731684421101609110.1177/07316844211016091 ↗
33. 10.18421/TEM111-3910.18421/TEM111-39 ↗
34. Bates S, Hastie T, Tibshirani R. Cross-validation: what does it estimate and how well does it do it? Stat Sci 2021; 36: 506–525.
35. Arlot S, Celisse A. Segmentation and cross-validation: theory and new methodology. J Mach Learn Res 2021; 22: 1–62.
36. Loh WY. Regression by parts: fitting visually interpretable models with GUIDE. J Comput Graph Stat 2021; 30: 976–990.
37. Loh WY, Zhou Y. Mean squared error and R-squared: which to use for variable importance? Stat Anal Data Min: The ASA Data Science Journal 2022; 15: 98–120.