A PSO ‐Optimized Multimodel Residual‐Correction Approach for MLR ‐Based In Situ Stress Inversion
Engineering Reports · 2026-11 · 8 (11)
DOI: 10.1002/eng2.71143 ↗Kurz gefasst
Fachartikel in der Zeitschrift Engineering Reports, online erschienen am 08.11.2026. Verfasst von Weibin Lin und 3 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Finite‐element‐based multivariate linear regression (MLR) can leave local discrepancies in an engineering in situ stress field. This study develops a nonlinear residual‐correction method that preserves an existing FE–MLR baseline. Seven loading‐case responses and six component identifiers form a 13‐feature input. XGBoost, support vector regression and a multilayer perceptron learn the differences between measured and baseline stress components. Particle swarm optimization selects model hyperparameters through grouped inner validation. Entropy–TOPSIS determines fusion weights from principal‐stress RMSE and MAE. The predicted residuals are added to the matching baseline components. The resulting three‐dimensional principal stresses are classified by direction as S H , S h , and S v . The case study contains 25 measuring points from four boreholes at a pumped‐storage power station. With the engineering baseline held fixed, 100 repeated point‐group splits reduce mean RMSE from 0.9185 to 0.7960 MPa. Mean MAE decreases from 0.7683 to 0.5980 MPa. The five‐point reference split yields reductions of 26.63% and 40.24%, respectively. Borehole‐specific depth profiles provide an engineering comparison with the site‐investigation report. The method enables local refinement of an established stress field through reuse of existing finite‐element outputs. These gains quantify residual correction conditional on the original site calibration.
Zitieren
APA
Lin, W., Cao, Z., Xiong, X., & Zhang, J. (2026). A PSO ‐Optimized Multimodel Residual‐Correction Approach for MLR ‐Based In Situ Stress Inversion. Engineering Reports, 8(11). https://doi.org/10.1002/eng2.71143
Vancouver
Lin W, Cao Z, Xiong X, Zhang J. A PSO ‐Optimized Multimodel Residual‐Correction Approach for MLR ‐Based In Situ Stress Inversion. Engineering Reports. 2026;8(11). doi:10.1002/eng2.71143
BibTeX
@article{Lin2026,
author = {Lin, Weibin and Cao, Ziyang and Xiong, Xiong and Zhang, Jinxin},
title = {A PSO ‐Optimized Multimodel Residual‐Correction Approach for MLR ‐Based In Situ Stress Inversion},
journal = {Engineering Reports},
year = {2026},
volume = {8},
number = {11},
doi = {10.1002/eng2.71143}
}
EUVIMED