A surrogate modeling approach for laser polishing of quartz glass

Marchanko L.N., Nikityuk Y.V., Bruttan I.V., Prokhorenko V.A., Zhang H., Zhang A., Sokolov S.I.
 pdf (1090K)

The study addresses surrogate modeling of thermomechanical responses in quartz-glass laser polishing using finiteelement simulation data. The relationships between thermal and mechanical responses and the main process parameters — laser power, source travel speed, and laser-spot radius — are investigated. Classical and regularized regression models, supportvector regression, tree-based ensembles, physics-structured models, tensor-product B-splines, and evolutionarily synthesized neural networks are compared using a common training, validation, and independent test split. Stacking ensembles based on Random Forest, Gradient Boosting, and XGBoost provided the lowest prediction errors for all investigated responses, with test coefficients of determination above 0.998. Physics-structured models were less accurate in terms of minimum prediction error but provided smooth and interpretable relationships between laser-processing parameters and thermomechanical responses. Evolutionary architecture search produced a two-output neural-network model for simultaneous prediction of the maximum temperature and maximum equivalent stress. Within the investigated dataset, source travel speed has the highest permutation importance for the predictive accuracy of the selected models. The results demonstrate that different classes of surrogate models are suitable for different modeling objectives: ensemble models are preferable when high predictive accuracy is required, physics-structured models are useful for response analysis and physical interpretation, and neural models provide simultaneous approximation of coupled thermal and mechanical responses. The developed surrogate models can therefore be used as a computationally efficient basis for analyzing laser-polishing regimes and for subsequent multicriteria optimization with simultaneous consideration of temperature and stress constraints.

Keywords: laser polishing, quartz glass, finite-element modeling, thermoviscoelasticity, surrogate model, physics-structured features, tensor B-splines, ensemble learning, neuroevolution
Citation in English: Marchanko L.N., Nikityuk Y.V., Bruttan I.V., Prokhorenko V.A., Zhang H., Zhang A., Sokolov S.I. A surrogate modeling approach for laser polishing of quartz glass // Computer Research and Modeling, 2026, vol. 18, no. 5, pp. 1223-1247
Citation in English: Marchanko L.N., Nikityuk Y.V., Bruttan I.V., Prokhorenko V.A., Zhang H., Zhang A., Sokolov S.I. A surrogate modeling approach for laser polishing of quartz glass // Computer Research and Modeling, 2026, vol. 18, no. 5, pp. 1223-1247
DOI: 10.20537/2076-7633-2026-18-5-1223-1247
URL: http://crm-en.ics.org.ru/journal/article/3778/
Creative Commons License This work is licensed under a Creative Commons Attribution-NoDerivs 3.0 Unported License.

Copyright © 2026 Marchanko L.N., Nikityuk Y.V., Bruttan I.V., Prokhorenko V.A., Zhang H., Zhang A., Sokolov S.I.

Indexed in Scopus

Full-text version of the journal is also available on the web site of the scientific electronic library eLIBRARY.RU

The journal is included in the Russian Science Citation Index

The journal is included in the RSCI

International Interdisciplinary Conference "Mathematics. Computing. Education"

Website Copyright © 2009–2026 Institute of Computer Science