JBT Journal
Advanced

Raqamli Transformatsiya va Sun’iy Intellekt

PHYSICS-INFORMED HYBRID MACHINE LEARNING FOR RESIDUAL MOISTURE PREDICTION IN ROLLER PRESSING OF LEATHER SEMI-FINISHED MATERIALS

Hudayberdiyev Mirzaakbar Khakkulmirzayevich1, Khusainova Khusanidaxon Saloyidinova2, Ravshanov Anvar Asatilloyevich3

1 Tashkent University of Information Technologies Named After Muhammad Al-Khwarizmi

2 Namangan State Technical University

3 Department of Higher Mathematics

Keywords

Physics-Informed Information Model technological process intelligent control XGBoost physics-informed machine learning leather roller pressing residual moisture Machine Learning

Abstract

Roller pressing of leather semi-finished materials is a complex multiparametric technological process in which mechanical deformation, contact pressure, liquid filtration, roller kinematics, and material moisture interact in a nonlinear manner. Conventional mathematical and automatic control approaches are often developed around a limited number of process parameters and therefore have difficulties in representing the combined influence of variable material properties and operating conditions. At the same time, machine learning (ML) methods can identify nonlinear dependencies from experimental data but do not inherently preserve the physical constraints of the technological process. This study proposes a Physics-Informed Hybrid Machine Learning (PI-HML) approach for residual moisture prediction in leather roller pressing. The proposed framework integrates a physical model of contact deformation and moisture filtration with a data-driven ML model through a unified Physics-Informed Information Model (PIIM). The methodological framework was derived from the physical and technological relationships of a two-roll pressing system and implemented for a pilot computational evaluation using 21 experimental observations with roller load intensity, roller radius, roller speed, and residual moisture as the principal variables. Leave-One-Out Cross-Validation was applied to compare Linear Regression, Support Vector Regression, Random Forest, Gradient Boosting, and XGBoost models. In the pilot experiment, XGBoost achieved the best data-driven performance with an MAE of 1.355% and RMSE of 2.136%, while the proposed physics-informed fusion reduced these errors to 1.289% and 1.969%, respectively, with R^2=0.791. These results demonstrate the feasibility of integrating physical process knowledge with ML for more consistent technological prediction. The proposed framework provides a methodological basis for the subsequent development of sensor-driven adaptive intelligent control of roller pressing machines.

Views: 102 Published: 2026-10-13 Pages: 131-139 Volume: 4 Issue: 5

References

  1. Бахадиров Г.А. Выбор кривой поверхности при подаче листового материала в зону обработки. // Динамика систем, механизмов и машин. ‒2016. ‒ С.72-77.
  2. Khurramov Sh. R, Khalturaev F.S., Buriev E.S. Residual moisture concent in semi-finished leather under roller pressing. AIP Conference Proceedings, MIP Injenering -III, Krosnoyarsk, Russia, 2402(2021)030038 https://doi.org/10.1063 /5.0071265.
  3. Amanov, A., Khurramov, S.R., Bahadirov, G.A., Abdukarimov, A., Amanov, T.Y. Modeling of Strain and Filtration Properties of a Semi-Finished Leather Product. Journal of Leather Science and Engineering, 3(1), 14, 2021.
  4. Khudayberdiyev M.Kh., Tojiboyev B.M., Achilov B,S., Alimkulov N.M., Ravshanov A.A. Integrating fractal dimension and time series analysis for optimized hyperspectral image classification // Advanced Computing: An International Journal (ACIJ), Vol.16, No.4, July 2025. Pp. 1-20
  5. Khurramov, Sh.R. Some Questions of the Contact Interaction Theory in Two-Roll Modules. Journal of Physics: Conference Series, 1546 (2020), 012132.
  6. Kamilov Mirzoyan, Hudayberdiev Mirzaakbar, Khamroev Alisher. The Procedure for Defining the best Recognition Module of the Algorithms for Calculating Estimates // Techno-Societal 2018. Proceedings of the 2nd International Conference on Advanced Technologies for Societal Applications. Springer, Cham. - Pp 25-32.
  7. Turgunov, K.K. Investigation of Contact Interaction in Roller Pairs of Leather Machines for Improving Product Quality and Production Efficiency. PhD dissertation, Tashkent Architectural and Construction University, Tashkent, 2023. The dissertation develops the physical and experimental basis for two-roll pressing of leather semi-finished materials.
  8. Хуррамов Ш.Р. Теоретические основы контактного в взаимодействия в двухвалковых модулях и ее использование совершенствовании процессов механической обработки. Дис. … докт. техн. наук. Ташкент: ТТИЛП, 2022. – 192 с.
  9. Khurramov, Sh.R. Simulation of the Form of Contact Curves Rolls in Two-Roll Modules. IOP Conference Series: Earth and Environmental Science, 614 (2020), 012096.
  10. Khurramov Sh. R. Filtration rates in roller pressing of fibrous materials" AIP Conference Proceedings, MIP Injenering -III, Krosnoyarsk, Russia, 125 2402(2021)030042 https://doi.org/10.1063/5.0071266 .
  11. N. R. Yusupbekov, S. M. Gulyamov and M. Y. Doshchanova, "Neural identification of a dynamic model of a technological process," 2019 International Conference on Information Science and Communications Technologies (ICISCT), Tashkent, Uzbekistan, 2019, pp. 1-5, doi: 10.1109/ICISCT47635.2019.9011912. (bu adabiyotlar ro’yxatiga kirsin!)
  12. Лапин И.Р. Покрытия для валов прессовой части бумагоделательной машины. // Новые импульсы развития: вопросы научных исследований. Серия «Механика и машиностроения» – 2020.− С. 94-96.
  13. Khurramov, Sh.R., Khalturaev, F.S., Kurbanova, F.Z. Deformation and Filtration Characteristics of a Leather Semi-Finished Product. Studies in Systems, Decision and Control, Vol. 342, Springer, 2021.
  14. Hudayberdiyev M.Kh., Tojiboyev B.M., Ravshanov A.A. Classification Of Earth Remote Sensing Images Using Artificial Neural Networks And Algorithms for calculation of estimates // Bulletin of TUIT: Management and Communication Technologies № 2024 3(29).