Machine Learning-Based Analysis and Prediction of Rainfall-Induced Slope Stability Using Numerical Coupling Models

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

This study investigates the slope stability of a photovoltaic power generation project in southern Shaanxi Province under rainfall infiltration conditions. Based on saturated-unsaturated seepage theory and strength reduction methods, combined with a three-dimensional geological model, a fluid-solid coupling analysis was conducted. A three-dimensional numerical model was established to simulate pore water pressure distribution, stress-strain response, and safety factor evolution under varying rainfall intensities (0.00375–0.03 m/h) and durations. Furthermore, the study introduced the Random Forest machine learning algorithm to train and predict numerical simulation results across multiple scenarios. This established a nonlinear mapping relationship between rainfall intensity, duration, and slope safety factor, enabling rapid intelligent assessment of slope stability. Results indicate: - Pore water pressure increases gradually with infiltration during rainfall onset, with the most significant changes occurring in the crown area; Under high-intensity rainfall, pore water pressure surges sharply within a short timeframe, causing the maximum compressive stress of the third principal stress to increase by approximately 178% and intensifying localized shear strain phenomena. The safety factor exhibits a three-stage characteristic: “slow decline—accelerated decrease—dynamic stability.” After prolonged rainfall, the stability value approaches the critical state, indicating slope instability. The random forest prediction model demonstrated high accuracy on the test set (R²>0.92), validating its effectiveness for rapid assessment of rainfall stability in such slopes. This study integrates traditional numerical simulation with machine learning methods, providing new insights and technical support for intelligent stability evaluation and disaster prevention in similar slopes under rainfall conditions.

Article activity feed