Hydrogen Diffusion in a Steel Pipe Wall as a Test Case for Physics-Informed Neural Networks
Main Article Content
Abstract
Hydrogen transport through existing gas pipelines causes hydrogen embrittlement of steel and reduces its fracture toughness, so a fast and reliable prediction of the hydrogen distribution in the pipe wall is necessary for safe operation. In this work, a physics-informed neural network (PINN) is developed for the hydrogen diffusion equation in a thick-walled cylinder. The network with 18,881 parameters was trained simultaneously on three sources of information: 20 nodes of the analytical solution (out of a 10×10 grid), the boundary and initial conditions, and the residual of Fick’s equation at 10,000 random collocation points. The mean absolute error over the entire domain is 0.029 ppm (1.1% of C₁, the hydrogen concentration at the inner surface), and the relative L₂ error on a dense grid is 5.4%. The error is shown to be concentrated in a single corner of the domain (r = r₁, t = 0), where r is the radial coordinate, r₁ is the inner radius of the pipe, and t is time; there the initial and boundary conditions contradict each other. The “engineering” metric, i.e., the time for the concentration to reach 1 ppm, is reproduced with an error of less than 1.5% for r ≥ 602 mm. It is also shown that, for the pipe under study (18 mm thick at a radius of 598 mm), the cylindrical solution differs from the planar one by no more than 0.2% of C₁; thus, the value of the PINN lies not in bypassing the Bessel-function series but in the universality of the approach for problems without a closed-form solution.
Article Details
Issue
Section

This work is licensed under a Creative Commons Attribution 4.0 International License.
How to Cite
References
[1] A. Laureys, R. Depraetere, M. Cauwels, T. Depover, S. Hertelé, K. Verbeken, Use of existing steel pipeline infrastructure for gaseous hydrogen storage and transport: A review of factors affecting hydrogen induced degradation, Journal of Natural Gas Science and Engineering 101 (2022) 104534.
[2] A. Campari, F. Ustolin, A. Alvaro, N. Paltrinieri, A review on hydrogen embrittlement and risk-based inspection of hydrogen technologies, International Journal of Hydrogen Energy 48(90) (2023) 35316–35346.
[3] O. Barrera, et al., Understanding and mitigating hydrogen embrittlement of steels: a review of experimental, modelling and design progress from atomistic to continuum, Journal of Materials Science 53 (2018) 6251–6290.
[4] O.I. Zvirko, O.T. Tsyrulnyk, O.I. Venhrynyuk, H.M. Nykyforchyn, Sensitivity of the J-integral method for estimating the hydrogen embrittlement of ferritic-pearlitic pipe steel, Strength of Materials 56(5) (2024) 928–935.
[5] H. Nykyforchyn, O. Tsyrulnyk, O. Venhryniuk, O. Zvirko, Techniques for investigation of hydrogen influence on fracture toughness and embrittlement of pipeline steels network, Procedia Structural Integrity 59 (2024) 125–130.
[6] O.T. Tsyrulnyk, O.Z. Student, O.I. Zvirko, D.O. Demianchuk, O.I. Venhryniuk, Assessment of hydrogen embrittlement of operated pipe steel using the J-integral method, Materials Science 59(6) (2024) 694–701.
[7] V. Jagota, A. Sethi, K. Kumar, Finite Element Method: An overview, Walailak Journal of Science & Technology 10(1) (2013) 1–8.
[8] O.I. Zvirko, S. Lipec, O.I. Vengreniuk, I. Dzioba, Evaluation of the stress-strain state at the crack tip in casing pipes based on numerical simulation, Materials Science 58(4) (2023) 460–465.
[9] T. Łodygowski, W. Sumelka, Limitations in application of Finite Element Method in acoustic numerical simulation, Journal of Theoretical and Applied Mechanics 44 (2006) 849–865.
[10] H.-T. Thai, Machine learning for structural engineering: A state-of-the-art review, Structures 38 (2022) 448–491.
[11] J. Behler, M. Parrinello, Generalized neural-network representation of high-dimensional potential-energy surfaces, Physical Review Letters 98 (2007) 146401. https://doi.org/10.1103/PhysRevLett.98.146401
[12] E. Malitckii, E. Fangnon, P. Vilaça, Evaluation of steels susceptibility to hydrogen embrittlement: A thermal desorption spectroscopy-based approach coupled with artificial neural network, Materials 13(23) (2020) 5500.
[13] K. Guo, Z. Yang, C.H. Yu, M.J. Buehler, Artificial intelligence and machine learning in design of mechanical materials, Materials Horizons 8(4) (2021) 1153–1172.
[14] S.-G. Kim, S.-H. Shin, B. Hwang, Machine learning approach for prediction of hydrogen environment embrittlement in austenitic steels, Journal of Materials Research and Technology 19 (2022) 2794–2798.
[15] I. Didych, A. Mykytyshyn, A. Stanko, M. Mytnyk, Application of machine learning methods to the prediction of NO₂ concentration in the air environment, CEUR Workshop Proceedings 3896 (2024) 569–577.
[16] F. Bünning, B. Huber, A. Schalbetter, A. Aboudonia, M. Hudoba de Badyn, P. Heer, Physics-informed linear regression is competitive with two Machine Learning methods in residential building MPC, Applied Energy 310 (2022) 118491.
[17] D. Cerbarano, L. Tieghi, G. Delibra, S. Minotti, A. Corsini, Modeling high-pressure hydrogen gas leakages with graph neural networks, Journal of Energy Resources Technology, Part A: Sustainable and Renewable Energy 1(3) (2025).
[18] X. He, D. Kong, X. Yu, P. Ping, G. Wang, R. Peng, Y. Zhang, X. Dai, Prediction model for the evolution of hydrogen concentration under leakage in hydrogen refueling station using deep neural networks, International Journal of Hydrogen Energy 51D (2024) 702–712.
[19] X. Zhang, J. Shi, J. Li, X. Huang, F. Xiao, Q. Wang, A. Usmani, G. Chen, Hydrogen jet and diffusion modeling by physics-informed graph neural network, Renewable and Sustainable Energy Reviews 207 (2025) 114898.
[20] M. Raissi, P. Perdikaris, G.E. Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational Physics 378 (2019) 686–707. https://doi.org/10.1016/j.jcp.2018.10.045
[21] A.I. Raichenko, Mathematical Theory of Diffusion in Applications, Naukova Dumka, Kyiv, 1981 (in Russian).
[22] N. Rahaman, et al., On the spectral bias of neural networks, in: Proceedings of the 36th International Conference on Machine Learning (ICML 2019), PMLR 97, 2019, pp. 5301–5310.
[23] M. Tancik, et al., Fourier features let networks learn high frequency functions in low dimensional domains, Advances in Neural Information Processing Systems 33 (2020).
[24] S. Wang, Y. Teng, P. Perdikaris, Understanding and mitigating gradient flow pathologies in physics-informed neural networks, SIAM Journal on Scientific Computing 43(5) (2021) A3055–A3081. https://doi.org/10.1137/20M1318043