Examining AI-Enhanced Regression Models for Predicting Slope Stability in Earthen Dams During Extreme Weather Events

The engineering significance of earth dams emphasizes the importance of slope stability. This study underscores the importance of considering precipitation events and rapid drawdown in stability analyses, to ensure the maintenance and conservation of hydraulic infrastructure in the face of nonstationary extreme weather events associated with climate change. In this research, the combination of hybrid numerical modeling and regression models with artificial intelligence is used to analyze the stability of homogeneous earth dam slopes. Three case studies are established considering transitory water flow in all scenarios. Two separate external loading analyses are included: rapid drawdown, which affects the upstream slope, and precipitation, which affects the downstream slope. Considering changes in flow and precipitation patterns under climate change scenarios, three rapid drawdown speeds are established: 0.1; 0.15; and 0.3 m/day. In addition, three precipitation intensities are defined, considering their increase in future scenarios: 3; 50; and 150 mm/day for 24 h duration. A methodology is proposed for the analysis of regression with AI. After applying the regression models, it was obtained that the polynomial regression of degree two presents consistent results (adjusted R² greater than 0.84) for the analysis of the stability of the upstream slope considering rapid drawdown. However, none of the evaluated models obtained satisfactory results for predicting stability on the downstream slope subjected to precipitation (adjusted R² less than 0.65), so new conditions for this type of event must be explored. Based on these results, it is recommended to consider analysis of the stability of slopes subjected to rapid drawdown, and successive precipitation events.