A physics-data-driven method for predicting surface and building settlement induced by tunnel construction

Surface and surrounding building settlement is frequently caused by soil disturbance during subway tunnel construction, significantly impacting construction safety and structural stability. Traditional machine learning models have shown some effectiveness in settlement prediction but often fail to c...

Full description

Saved in:

Bibliographic Details
Published in:	Computers and geotechnics Vol. 179; p. 107020
Main Authors:	Wang, You, Fan, Qianjun, Dai, Fang, Wang, Rui, Ding, Bosong
Format:	Journal Article
Language:	English
Published:	Elsevier Ltd 01.03.2025
Subjects:	Metaheuristic optimization algorithms Pasternak foundation model Physics informed optimized extreme learning machine (PIOELM) Physics-data-driven Shield tunnelling Metaheuristic optimization algorithms Pasternak foundation model Shield tunnelling Physics informed optimized extreme learning machine (PIOELM) Physics-data-driven
ISSN:	0266-352X
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Surface and surrounding building settlement is frequently caused by soil disturbance during subway tunnel construction, significantly impacting construction safety and structural stability. Traditional machine learning models have shown some effectiveness in settlement prediction but often fail to capture the underlying physical mechanisms. This study proposed a novel physics-informed optimized extreme learning machine (PIOELM) to enhance prediction accuracy and physical interpretability. Based on the extreme learning machine (ELM), the model integrated the chaos adaptive sparrow search algorithm (CASSA) for parameter optimization and incorporated the Pasternak foundation model using automatic differentiation. The model’s accuracy was validated using precise engineering data and compared against the physics-informed neural network (PINN), physics-informed extreme learning machine (PIELM), and traditional data-driven models. The results show that the PIOELM model outperforms others in handling extreme values and maintains high accuracy across various scales of data prediction. Prediction accuracy improved by up to 85.29%, with a minimum improvement of 30.68%, demonstrating strong stability and generalization capabilities.
ISSN:	0266-352X
DOI:	10.1016/j.compgeo.2024.107020