Using a Recurrent Neural Network for Solving Inverse Heat Conduction Problem with Application to Calculating the Temperature of Equipment of VVER-Based NPPs

12/20/2024 2024 - #04 Modelling processes at nuclear facilities

Deryabin I.A. Korolev V.V. Sorokin G.S.

https://doi.org/10.26583/npe.2024.4.12

The need for solving transient boundary inverse heat conduction problems (IHCP), as a tool that allows one to obtain missing information on the object under consideration, can arise both when processing experimental results, and in the process of thermal measurements, and, in some cases, in the process of design. Therefore, as applied to VVER-based NPPs, the use of the IHCP solution results may be associated with processing of data in the process of commissioning measurements or equipment life monitoring in the course of the NPP operation. This paper proposes and describes a method for calculating boundary inverse heat conduction problems using a NARX-type recurrent neural network. Unknown weights and biases are calculated using the Levenberg – Marquardt gradient descent algorithm in the process of training based on a dedicated data set obtained by calculating a direct heat conduction problem using a series of randomly generated thermal loads. Several examples are provided to demonstrate the applicability of the methodology for solving linear and nonlinear heat conduction problems. Issues have been considered concerning the neural network parameters, such as time lags and the number of previous time steps under consideration. A two-dimensional problem has been discussed individually, which shows that introducing additional information into the input data makes it possible to reduce substantially the error in the obtained results. The results show minor deviations of the predicted temperature from the actual one under different inner surface boundary conditions. The developed methodology offers a major potential with respect to the VVER equipment temperature measurement issues, as well as to processing thermal experiment results.

References

1. Alifanov O.M. Inverse heat conduction problems. Moscow, Mashinostroenie Publ., 1988. – 280 p. ISBN 5-217-00134-8 (in Russian).
2. Bogachev A.V., Dranchenko B.N., Semishkin V.P., Berkovich V.Ya. Development of the SAKOR-M system. Voprosy atomnoy nauki I tehniki. Serya: Obespechenie bezopasnosty AES. Nauchno-tehnicheskiy sbornik: No. 15. Reactornie ustanovki VVER. Podolsk, 2006 (in Russian).
3. Korolev V.V. Methodology for processing the results of monitoring temperatures and pressures in equipment elements of reactor plants. Voprosy atomnoy nauki I tehniki. Serya: Obespechenie bezopasnosty AES. Nauchno-tehnicheskiy sbornik: No. 17. Reactornie ustanovki VVER. Podolsk, 2007. (in Russian).
4. Deng S., Hwang Y. Applying neural networks to the solution of forward and inverse heat conduction problems. International Journal of Heat and Mass Transfer. 2006, vol. 49, iss. 25–26, pp. 4732 – 4750. DOI: https://doi.org/10.1016/j.ijheatmasstransfer.2006.06.009
5. Vasiliev A.N., Porubaev F.V., Tarkhov D.A. Neural network approaches to solving ill-posed heat transfer problems. Nauchno-tehnicheskie vedomosty SPBGPU. 2011, iss. 1, pp. 133 – 142 (in Russian).
6. Zhili He, Futao Ni, Weiguo Wang, Jian Zhang. A physics-informed deep learning method for solving direct and inverse heat conduction problems of materials. Materials today communications. 2021, vol. 28, 102719. DOI: https://doi.org/10.1016/j.mtcomm.2021.102719
7. Ruiz L.G.B., Cuéllar M.P., Calvo-Flores M.D., Jiménez M.D.C.P. An Application of Non-Linear Autoregressive Neural Networks to Predict Energy Consumption in Public Buildings. Energies. 2016, vol. 9, 684. DOI: https://doi.org/10.3390/en9090684
8. Jose Maria P.Menezes Jr, Guilherme A. Barreto. Long-term time series prediction with NARX network: An empirical evalution. Neurocomputing. 2008, vol. 71, iss. 16 – 18, pp. 3335 – 3343. DOI: https://doi.org/10.1016/j.neucom.2008.01.030
9. Zubchenko A.S., Koloskov M.M., Kashirsky Yu.V. Brand of steels and alloys. Moscow, Mashinostroenie Publ., 2003, 784 p. (in Russian).

temperature inverse heat conduction problem recurrent neural network VVER-based NPPs

Link for citing the article: Deryabin I.A., Korolev V.V., Sorokin G.S. Using a Recurrent Neural Network for Solving Inverse Heat Conduction Problem with Application to Calculating the Temperature of Equipment of VVER-Based NPPs. Izvestiya vuzov. Yadernaya Energetika. 2024, no. 4, pp. 144-154; DOI: https://doi.org/10.26583/npe.2024.4.12 (in Russian).