This study presents a comprehensive review of artificial intelligence (AI) and machine learning (ML) techniques for back analysis of geomechanical parameters in tunnel engineering. The increasing complexity of underground construction and the growing availability of monitoring data make AI-based inverse modeling a valuable tool for estimating key rock mass properties, including elastic modulus ( E ), cohesion ( c ), friction angle ( φ ), and Poisson’s ratio ( ν ). Recent advances have been examined across neural networks, support vector machines, fuzzy logic systems, ensemble learning, and hybrid models that integrate evolutionary algorithms with data-driven approaches. The results show that hybrid AI methods, such as genetic algorithm (GA)–enhanced neural networks or particle swarm-based regression, outperform conventional approaches in terms of prediction accuracy, convergence speed, and robustness under incomplete or noisy data. Surrogate modeling and probabilistic frameworks further reduce computational cost while addressing parameter uncertainty. Practical applications demonstrate that AI-enhanced models can reproduce measured tunnel displacements and adapt to varying geological conditions. The integration of AI with numerical modeling improves reliability, efficiency, and automation in parameter identification. The paper concludes with key challenges and future directions, such as real-time updating, model generalization, and the creation of standardized benchmark datasets.
Saadati et al. (Fri,) studied this question.