Искусственный интеллект в инженерной термодинамике: методика обучения студентов нейросетевым методам решения задач тепломассообмена

UDC 37.09:536-042.4:004.8
DOI 10.20339/AM.07-26.054

Leonid V. Bykov, PhD in Engineering, Docent, Moscow Aviation Institute (National Research University), e-mail: bykovlvl@mail.ru

Aleksandr M. Molchanov, D.Sc. in Engineering, Professor, Moscow Aviation Institute (National Research University), e-mail: alexmol_2000@mail.ru

The article presents a teaching methodology for instructing students in the application of neural network methods to solve heat and mass transfer problems within an engineering thermodynamics course. Emphasis is placed on integrating physical content with digital tools: students systematically progress through stages involving problem formulation, synthetic data generation, and neural network model development. A key instructional example is the inverse problem of spectral diagnostics — determining combustion parameters from the infrared radiation emitted by a gas flow. The architecture of the multilayer perceptron (MLP) is examined as an efficient tool for rapidly and accurately reconstructing temperature and concentrations of radiating species. The proposed methodology helps overcome the perception of artificial intelligence as a ‘black box’ and fosters in students a holistic understanding of the synergy between classical physical models and modern computational approaches. Implementation of this methodology in a master's-level course has demonstrated increased student motivation, enhanced digital competencies, and a deeper comprehension of inverse problems.

Keywords: artificial intelligence, machine learning, engineering thermodynamics, heat transfer, spectral diagnostics, multilayer perceptron, teaching methodology

References

1. Pope, S.B. Turbulent Flows. Cambridge University Press, 2000.
2. Wilcox, D.C. Turbulence Modeling for CFD. DCW Industries, 2006.
3. Lagaris, I.E., Likas, A., Fotiadis, D.I. Artificial neural networks for solving ordinary and partial differential equations. IEEE Transactions on Neural Networks. 1998. Vol. 9 (5). Pp. 987–1000.
4. Duraisamy, K., Zhang, Z.J., Singh, A.P. New approaches in turbulence and transition modeling using data-driven techniques. AIAA Paper. 2015. Vol. 1284.
5. Ling, J., Kurzawski, A., Templeton, J. Reynolds averaged turbulence modelling using deep neural networks with embedded invariance. Journal of Fluid Mechanics. 2016. Vol. 807. Pp. 155–166.
6. Petrunin, A.P., & Smirnov, N.N. The application of neural network models for solving transport equations in turbulent flows. Mathematical Modelling. 2020. Vol. 32 (5). Pp. 45–60.
7. François, G., et al. Neural network approach for solving radiative transfer problems. Numerical Heat Transfer, Part B: Fundamentals. 2008. Vol. 53 (3). Pp. 213–229.
8. Modest, M.F., et al. Machine learning in radiative heat transfer. Annual Review of Heat Transfer. 2020. Vol. 23. Pp. 345–387.
9. Blaythoff, A., et al. Deep neural network based chemical mechanism reduction. Proceedings of the Combustion Institute. 2019. Vol. 37 (1). Pp. 651–659.
10. Aristarkhov, V.M., et al. Neural network methods in the kinetic modelling of complex chemical reactions. Chemical Physics. 2021. Vol. 40 (2). Pp. 25–35.
11. Rothman, L.S., et al. The HITRAN molecular spectroscopic database. Journal of Quantitative Spectroscopy and Radiative Transfer. 2013. Vol. 130. Pp. 4–50.
12. Zwiefhofer, M., et al. Temperature and concentration measurements in combustion gases using FTIR spectroscopy and artificial neural networks. Applied Optics. 2005. Vol. 44 (5). Pp. 678–686.
13. Tan, S., et al. Deep neural networks for multi-parameter combustion diagnostics. Combustion and Flame. 2022. Vol. 235. Pp. 111–125.