Work

Optimized prediction and closed-form solutions for dual rectangular tunnel stability using rigorous limit analysis and advanced data-driven modeling

Year: 2026

Type: article

Abstract: The stability of dual rectangular tunnels in cohesive–frictional soils under surcharge loading is a critical yet understudied problem, particularly given the strong interaction effects between adjacent tunnels. Existing research focuses mainly on circular or square geometries, leaving the behaviour of dual rectangular tunnels insufficiently explored. This study addresses this gap by integrating Finite Element Limit Analysis (FELA) with advanced soft computing techniques to develop reliable and interpretable stability predictions. FELA, implemented using Optum CE, provides rigorous lower- and upper-bound solutions for ultimate surcharge loading while capturing complex interaction mechanisms. To enhance predictive capability and support practical design, a hybrid ensemble model combining Extreme Gradient Boosting, Categorical Boosting, Support Vector Regression, and Feedforward Neural Networks is optimized using the Tree-structured Parzen Estimator. In addition, symbolic regression is employed to derive closed-form expressions, offering transparent and computationally efficient alternatives to black-box models. Results show that stability is strongly governed by geometric ratios ( A / D, S / D, H / D ) and soil parameters (cohesion, friction angle, unit weight). The hybrid model achieves excellent accuracy (R² > 0.99), and the symbolic expressions align closely with FELA predictions. These outcomes provide a reliable and interpretable framework for rapid stability assessment and support informed decision-making in the design of dual rectangular tunnel systems.

Source: KSCE Journal of Civil Engineering

DOI: 10.1016/j.kscej.2026.100691

Authors: Tung Le-Thanh, Hung La, Duy-Khuong Ly, Tan Nguyen

Institutions: Ton Duc Thang University, Van Lang University

Language: English

Topic: Geotechnical Engineering and Analysis

Subfield: Safety, Risk, Reliability and Quality

Field: Engineering

Domain: Physical Sciences

SDGs: Industry, innovation and infrastructure, Sustainable cities and communities

Open Access status: gold

Estimated APC paid: $1,800