A.O. Belousov1, G.Yu. Kim2, V.O. Gordeyeva3

1–3 Tomsk State University of Control Systems and Radioelectronics (Tomsk, Russia)

1 anton.o.belousov@tusur.ru; 2 georgii.i.kim@tusur.ru; 3 viktoriia.gordeeva@tusur.ru

Problem statement. Currently, modern radio-electronic equipment (REE) is subject to stringent requirements for ensuring electromagnetic compatibility, especially under deliberate electromagnetic interference such as ultra-wideband noise and ultra-short pulses. Traditional protective devices often prove inefficient under these conditions, necessitating the use of alternative approaches, in particular various designs of modal filters (MF). Optimizing MFs is a complex multi-parameter task, and the classical global optimization methods traditionally employed for this purpose require significant computational resources, especially when dealing with complex MF structures. Consequently, in order to accelerate the optimization process and improve its accuracy, it is advisable to develop a promising approach based on the use of machine learning (ML) methods.

Purpose. To develop an approach to optimizing strip and cable MFs by applying ML regression models that can significantly reduce the time required to identify optimal parameters while maintaining the accuracy of the obtained results.

Results. An MF optimization approach based on ML regression models is proposed, demonstrating high efficiency. It was established that, for two different MF designs, the LightGBM and CatBoost models yielded the best results (high-speed predictions of optimal MF parameters without compromising accuracy) with a coefficient of determination R2>0.99 and low root mean square error (RMSE) values between calculated and predicted outputs (RMSE<2.4% for LightGBM and RMSE<1.7% for CatBoost). The overfitting coefficient for all models remained within acceptable limits, indicating good generalization ability. Meanwhile, finding optimal parameters by global optimization methods takes about 16,000 s, whereas the trained ML models require less than 1 s.

Practical significance. The developed approach to MF optimization can be used to significantly speed up the design of various types of transmission lines (a strip structure with broad-side coupling and a flat cable) with specified protective characteristics. The application of this approach substantially reduces the time and computational resources required for MF optimization, while ensuring a high level of protection for REE against deliberate electromagnetic interference, without the need for the resource-intensive computations of global optimization methods.

Belousov A.O., Kim G.Yu., Gordeyeva V.O. Application of modern machine learning models to improve the optimization efficiency
of strip and cable protection structures. Radiotekhnika. 2025. V. 89. № 4. P. 61−77. DOI: https://doi.org/10.18127/j00338486-202504-06 (In Russian)

1. Leferink F. Conducted interference, challenges and interference cases. IEEE EMC Magazine. 2015. V. 4. № 1. P. 78–85.
2. Van Doom M. EMC demo box: conducted & radiated emission. Proceedings of IEEE International Symposium on Electromagnetic Compatibility (EMC '03). 2003. P. 521–522. DOI: 10.1109/ICSMC2.2003.1428307.
3. Radasky W.A., Hoad R. Recent developments in high power EM (HPEM) standards with emphasis on high altitude electromagnetic pulse (HEMP) and intentional electromagnetic interference (IEMI). IEEE Letters on Electromagnetic Compatibility Practice and Applications. 2020. V. 2. № 3. P. 62–66. DOI: 10.1109/LEMCPA.2020.3009236.
4. Giri D.V., Tesche F.M. Classification of intentional electromagnetic environments. IEEE Transactions on Electromagnetic Compatibility. 2004. V. 46. № 3. P. 322–328.
5. Giri D.V. High-power Electromagnetic Radiators: Nonlethal Weapons and Other Applications. Harvard University Press, 2004. 198 p.
6. Giri D.V. History of high-power electromagnetics (HPEM) from 1940s to 21st century. Proceedings of the 7th Asia-Pacific Conference on Environmental Electromagnetics (CEEM). 2015. P. 1–5. DOI: 10.1109/CEEM.2015.7368630.
7. Hamamah F., Ahmad W.W., Gomes C., Isa M.M., Homam M. Concerns on the risk of Malaysian Civil and defense systems due to intentional electromagnetic interference. Proceedings of IEEE Asia-Pacific Conference on Applied Electromagnetics (APACE). 2019. P. 1–6. DOI: 10.1109/APACE47377.2019.9021096.
8. Radasky W.A., Wik M.W. Overview of the threat of intentional electromagnetic interference (IEMI). Proceedings of IEEE International Symposium on Electromagnetic Compatibility (EMC '03). 2003. P. 1024–1027. DOI: 10.1109/ICSMC2.2003.1429088.
9. Nuriev M.G., Gizatullin R.M., Gizatullin Z.M. Physical Modeling of Electromagnetic Interferences in the Electronic Devices at Direct Impact of Lightning on Protection System of Building. Proceedings of the XIV International Scientific and Technical Conference Actual Problems of Electronic Instrument Engineering (APEIE). 2018. P. 355–358. DOI: 10.1109/APEIE.2018.8545950.
10. Gazizov A.T., Zabolotsky A.M., Gazizov T.R. UWB pulse decomposition in simple printed structures. IEEE Transactions on Electromagnetic Compatibility. 2016. V. 58. № 4. P. 1136–1142. DOI: 10.1109/TEMC.2016.2548783.
11. Belousov A.O., Gazizov T.R. Systematic approach to optimization for protection against intentional ultrashort pulses based on multiconductor modal filters. Complexity. 2018. V. 2018. P. 1–15. DOI: 10.1155/2018/5676504.
12. Belousov A.O., Gordeyeva V.O. Comparison of a Genetic Algorithm and Evolutionary Strategies in Optimization of Strip Modal Filters. Journal of Communications Technology and Electronics. 2023. V. 68. № 11. P. 1079–1089. DOI: 10.1134/S1064226923110037
    (in Russian).
13. Kuksenko S.P., Gazizov T.R., Kvasnikov A.A. i dr. Razrabotka programmnogo obespechenija dlja modelirovanija radiojelektronnyh sredstv s uchetom jelektromagnitnoj sovmestimosti v TUSUR. Nanoindustrija. 2023. T. 16. № S9-1(119). S. 170–178 (in Russian).
14. Belousov A.O., Gordeeva V.O. Optimizacija pomehozashhitnyh struktur s modal'nym razlozheniem: monografija. Tomsk: Izd-vo Tomskogo gosudarstvennogo universiteta sistem upravlenija i radiojelektroniki. 2023. 136 s.
15. Hansen N. Python: modulebarecmaes2. URL: http://www.cmap.polytechnique.fr /~nikolaus.hansen/barecmaes2.html (дата обра-щения: 10.10.2024).
16. Pereira V. Python: Random Search algorithm. URL: https://github.com/Valdecy/pyMetaheuristic /blob/main/pyMetaheuristic/algo-rithm/random_s.py (дата обращения: 18.10.2024).
17. Goodfellow I., Bengio Y., Courville A. Deep Learning. Cambridge. MA: MIT Press. 2016. Chapter 6.2.1.
18. Ke G., et al. LightGBM: A highly efficient gradient boosting decision tree. Proceedings of Advances in Neural Information Processing Systems (NeurIPS). 2017. P. 3146–3154.
19. Prokhorenkova L., Gusev G., Vorobev A., Dorogush A.V., Gulin A. CatBoost: unbiased boosting with categorical features. Proceedings of Advances in Neural Information Processing Systems (NeurIPS). 2018. P. 6638–6648.
20. Hastie T., Tibshirani R., Friedman J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed. New York. NY: Springer. 2009. Chapter 3.4.

​​​​​​​