COUPLING EXPERIMENTAL DATA WITH MONTE CARLO SIMULATION AND AI FOR REALISTIC PREDICTION OF SLOPE INSTABILITY

Abstract

Excavation-induced slope instability threatens infrastructure safety and civil engineering operations, often leading to substantial economic losses and environmental degradation. Traditional predictive methods, such as the Limit Equilibrium Method (LEM) and Finite Element Method (FEM), are limited in their ability to address the nonlinear, dynamic, and uncertain conditions typically present in excavation scenarios. This study presents an integrated prediction framework that couples experimental observations with Monte Carlo simulation and artificial intelligence (AI) techniques to enhance the accuracy and robustness of slope instability prediction. The Monte Carlo simulation component estimates failure probabilities under various geotechnical conditions, including moisture content, vibration amplitude, slope angle, and pore water pressure. Simultaneously, machine learning algorithms—specifically random forests and neural networks—are employed to capture complex interactions among variables and improve classification performance. The proposed model demonstrates superior predictive accuracy, with machine learning classifiers achieving over 98% accuracy, and effectively identifies critical thresholds and dominant risk factors through statistical and visual analysis. The integration of experimental data further validates the model’s reliability. This hybrid approach provides a scalable, adaptive, and data-driven tool for real-time slope risk assessment and early warning applications, supporting safer and more resilient infrastructure development in geotechnically complex environments.