AI-Driven Prediction of Tunneling Squeezing: Comparing Rock Classification Systems

Understanding tunneling-induced ground deformation, particularly squeezing behavior, is crucial for safe and efficient underground construction. This study employs Artificial Neural Networks (ANN) and Multivariate Adaptive Regression Splines (MARS) models to predict tunneling squeezing behavior usin...

Full description

Saved in:
Bibliographic Details
Published in:Geotechnical and geological engineering 2024-05, Vol.42 (3), p.2127-2149
Main Authors: Al-Sadoon, Zaid A., Alotaibi, Emran, Omar, Maher, Arab, Mohamed G., Tahmaz, Ali
Format: Article
Language:English
Subjects:
Accuracy
Artificial neural networks
Civil Engineering
Classification
Classification systems
Compressing
Construction
Critical point
Deformation
Earth and Environmental Science
Earth Sciences
Geotechnical Engineering & Applied Earth Sciences
Hydrogeology
Mathematical models
Model accuracy
Modulus of deformation
Neural networks
Original Paper
Parameter sensitivity
Parameters
Performance evaluation
Predictions
Regression analysis
Rock
Rock mass rating
Rocks
Sensitivity analysis
Splines
Terrestrial Pollution
Tunneling
Tunnels
Underground construction
Waste Management/Waste Technology
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Understanding tunneling-induced ground deformation, particularly squeezing behavior, is crucial for safe and efficient underground construction. This study employs Artificial Neural Networks (ANN) and Multivariate Adaptive Regression Splines (MARS) models to predict tunneling squeezing behavior using various rock classification systems, namely rock quality index (Q), rock mass rating (RMR), and geological strength index (GSI). The objective is to assess model performance, evaluate the influence of classification systems, and conduct sensitivity analyses on key parameters. The investigation reveals that both ANN and MARS models exhibit enhanced accuracy as model complexity increases, up to a critical point where overfitting occurs. Comparing model performance, ANN outperforms MARS, and the most accurate ANN model is identified as ANN50-RMR with an R 2 of 0.978. This confirms the ANN’s capability to capture non-linear relationships inherent in tunneling-induced ground deformation. Choosing a rock classification system as an input parameter significantly impacts model accuracy. RMR and GSI classification systems exhibit improved performance over the conventional Q-system. In particular, GSI-based models offer more consistent and accurate predictions, emphasizing GSI’s suitability for modeling tunneling squeezing behavior. Variables’ importance analysis elucidates the dependence of parameter relevance on the chosen classification system. Sensitivity analyses on tunnel depth, diameter, and rock mass deformation modulus reveal logical correlations between these parameters and tunnel squeezing behavior, further validating model predictions. By enhancing our understanding of tunneling-induced ground deformation, these models contribute to safer and more efficient underground construction practices.
ISSN:0960-3182
1573-1529
DOI:10.1007/s10706-023-02665-5