Physics-informed gradient boosting for particle size measurement in turbid media
Physica Scripta · 19.06.2026 · 101 (24) · 246010
DOI: 10.1088/1402-4896/ae79ab ↗Kurz gefasst
Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 17.06.2026. Verfasst von Jun Hu und 4 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Gefördert von: Natural Science Research Project of Colleges and Universities in Anhui Province, Collaborative Innovation Project of Colleges and Universities in Anhui Province.
Abstract (Originalsprache)
Particle size monitoring in complex turbid systems is frequently hindered by strongly nonlinear scattering and the limited interpretability of purely data-driven models. To tackle these issues, this study introduces a rapid detection method for suspensions utilizing a physics-informed gradient boosting regression tree (PI-GBRT) framework driven by optical feature engineering. This approach synergistically integrates multi-spectral sensing with machine learning by embedding Mie scattering theory within an ensemble learning framework. Through extinction space mapping, distinct physical fingerprints—including the generalized extinction index and cross-band scattering ratio—are extracted and fused nonlinearly with multi-spectral and colorimetric data. Experimental validation reveals that the PI-GBRT model achieves exceptional inversion precision across a 0.1–1.8 μ m range, yielding an R 2 of 0.9945 and an RMSE of 0.0369 μ m. Feature importance analysis further confirms the dominant role of physical priors in the decision path. Ultimately, this work underscores the critical value of physics-guided knowledge in boosting analytical accuracy, presenting a robust, non-invasive solution for the online monitoring of industrial fluids.
Förderung
- Natural Science Research Project of Colleges and Universities in Anhui Province
- Collaborative Innovation Project of Colleges and Universities in Anhui Province
Zitieren
APA
Hu, J., Wang, Y., Men, X., Yuan, L., & Liu, S. (2026). Physics-informed gradient boosting for particle size measurement in turbid media. Physica Scripta, 101(24), 246010. https://doi.org/10.1088/1402-4896/ae79ab
Vancouver
Hu J, Wang Y, Men X, Yuan L, Liu S. Physics-informed gradient boosting for particle size measurement in turbid media. Physica Scripta. 2026;101(24):246010. doi:10.1088/1402-4896/ae79ab
BibTeX
@article{Hu2026,
author = {Hu, Jun and Wang, Yanyan and Men, Xin-ao and Yuan, Linmeng and Liu, Sheng},
title = {Physics-informed gradient boosting for particle size measurement in turbid media},
journal = {Physica Scripta},
year = {2026},
volume = {101},
number = {24},
pages = {246010},
doi = {10.1088/1402-4896/ae79ab}
}
EUVIMED