Computational Intelligence Approaches to Ground Settlement Prediction in Tunneling: A Review of Recent Advances

Ground settlement induced by tunneling remains a critical challenge in geotechnical and structural engineering—especially in dense urban corridors where small deformations can compromise infrastructure safety. Over the past decade, machine learning (ML) has become a practical complement to analytical and empirical methods for settlement prediction and stability assessment, owing to its ability to capture nonlinear, multi-factor interactions in ground–structure systems. This chapter reviews ML-based approaches for ground settlement prediction and structural stability in tunneling projects, focusing on studies published between 2018 and 2025. We categorize methods by data sources (site investigation, construction logs, and monitoring), model families (support vector machines, artificial neural networks, ensemble learners, and deep architectures), validation strategies, and performance metrics. We then compare strengths, limitations, and typical use contexts. Recent developments are highlighted, including real-time integration with sensor networks, data-fusion schemes to improve reliability, and physics-informed machine learning to enhance robustness and physical consistency. Case studies from large-scale tunnel projects illustrate how ML has been used to mitigate settlement risk, optimize construction parameters, and support integrity management. We also discuss emerging directions—digital twins and explainable AI for transparency, operator trust, and adaptive decision-making. The chapter closes with key gaps—data heterogeneity, limited generalizability across sites, and non-standardized evaluation—and outlines practical steps for advancing safe, resilient, and sustainable tunneling operations.