融合物理信息神经网络的钢渗碳过程建模与性能预测

doi:10.11896/cldb.25080137

Abstract
Figure/Table
References
Related Citation (12)

Download:

Full Text ( PDF ) (8581KB)
Export: BibTeX | EndNote (RIS)

Abstract Carburizing heat treatment is widely employed to enhance the surface hardness and wear resistance of steel components, but accurate control and prediction of the final performance usually rely on extensive experimental testing, which increases cost and reduces efficiency. In this study, a physics-informed neural network (PINN) is employed to solve the carbon diffusion problem in carburized steel, and the hardness gra-dient is subsequently derived from the predicted carbon concentration profile. Fick’s second law is explicitly embedded into the network architecture together with the corresponding initial and boundary conditions, enabling physically consistent simulation of the time-dependent evolution of the carbon concentration field. The carbon-to-hardness mapping is performed through an experimentally calibrated nonlinear relationship. Applied to a gas-carburized 20CrMnTi steel, the proposed model achieves a maximum error in carbon concentration prediction below 0.057wt% and a maximum hardness error not exceeding 0.93%, clearly outperforming conventional numerical solvers. The integration of physics-driven constraints and data-driven learning markedly reduces the reliance on large experimental datasets, enhances generalization to varying process conditions, and maintains a relatively low computational cost. The presented framework provides a practical tool for optimizing carburizing heat treatment and demonstrates strong potential for extension to other alloys and carburizing schedules.

Key words： physics-informed neural network carburizing carbon diffusion hardness prediction Fick’s second law

Published: 25 July 2026 Online: 2026-08-12

CLC number:

TG156

	Service
	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	ZHENG Yihao
	JIANG Di
	WANG Xuefei
	LUO Chunyang
	WANG Zhaodong

Cite this article:

ZHENG Yihao,JIANG Di,WANG Xuefei, et al. Physics-informed Neural Network-based Modeling and Performance Prediction of Steel Carburizing Process[J]. Materials Reports, 2026, 40(14): 25080137-6.

URL:

https://www.mater-rep.com/EN/10.11896/cldb.25080137 OR https://www.mater-rep.com/EN/Y2026/V40/I14/25080137

1 Qin S W, Di L Y, Wang L X, et al. Materials Reports, 2024, 38(2), 204 (in Chinese).
秦盛伟, 邸黎寅, 王连翔, 等. 材料导报, 2024, 38(2), 204.
2 Bartosz I, Artur W, Maciej M. Metals, 2023, 13(2), 371.
3 Zhang Y, Li Y, Zheng L C, et al. Metallurgical Analysis, 2022, 42(6), 18 (in Chinese).
张庸, 李瑶, 郑立春, 等. 冶金分析, 2022, 42(6), 18.
4 Chen W, He X F, Yu W C, et al. Metals, 2021, 11(2), 300.
5 Li D C, Yan B, Gao T Y, et al. ACS Omega, 2024, 9(3), 3846.
6 Yamamoto R, Nishimoto A. Materials Transactions, DOI:10. 2320/matertrans. MT-M2025025.
7 Ma W B, Sheng J J, Wang Y H, et al. Coatings, 2022, 12(12), 1886.
8 Semenov M Y, Smirnov A E, Fomina L P, et al. Metal Science and Heat Treatment, 2024, 66(1), 5.
9 Hiremath P, Gurumurthy B M, Shivaprakash Y M, et al. Cogent Engineering, 2024, 11(1), 2345847.
10 Liu B X. Numerical simulation of carbon concentration field and structure field in carburizing-quenching process. Master’s Thesis, Shanghai Jiao Tong University, China, 2016 (in Chinese).
刘博勋. 渗碳淬火过程浓度场与组织场的模拟研究. 硕士学位论文, 上海交通大学, 2016.
11 Uranaka S, Hirashima I, Maeda T, et al. ISIJ International, 2023, 63(3), 569.
12 Raissi M, Perdikaris P, Karniadakis G E, et al. Journal of Computational Physics, 2019, 378, 686.
13 Wang S F, Yu X L, Perdikaris P, et al. Journal of Computational Physics, 2022, 449, 110768.
14 Yu J, Lu L, Meng X, et al. Computer Methods in Applied Mechanics and Engineering, 2022, 393, 114823.
15 Sukumar N, Srivastava A. Computer Methods in Applied Mechanics and Engineering, 2022, 389, 114333.
16 Sitzmann V, Martel J, Bergman A, et al. Advances in Neural Information Processing Systems, 2020, 33, 7462.
17 Luo K, Zhao J S, Wang Y P, et al. Artificial Intelligence Review, 2025, 58(10), 1.
18 Wang Y F, Zhong L L. Journal of Computational Physics, 2024, 496, 112603.
19 Chen H, Constante F G E, Li C. Computers & Chemical Engineering, 2024, 189, 108764.
20 Wang C W, Li S D, He D, et al. In:Advances in Neural Information Processing Systems 35. New Orleans, 2022, pp. 8278.
21 Wu C X, Zhu M, Tan Q Y, et al. Computer Methods in Applied Mechanics and Engineering, 2023, 403, 115671.
22 Cuomo S, Schiano D C V, Giampaolo F, et al. Journal of Scientific Computing, 2022, 92(3), 88.
23 Hanna J M, Aguado J V, Comas-Cardona S, et al. Computer Methods in Applied Mechanics and Engineering, 2022, 396, 115100.
24 Zhang S H, Zhang C, Han X, et al. Computational Mechanics, 2025, 75(3), 1137.
25 Willmott C J, Matsuura K. Climate Research, 2005, 30(1), 79.