Prediction of Roadside Soil Slope Stability Using Artificial Neural Networks Based on Finite Element Method

The stability assessment of soil slopes remains a fundamental challenge in geotechnical engineering. This study employs Artificial Neural Networks (ANN) to predict the stability of roadside soil slopes. A representative slope configuration was defined using six key geotechnical parameters: unit weight, cohesion, friction angle, Poisson's ratio, Young's modulus, and slope angle, as input variables. A comprehensive parametric analysis was conducted using Finite Element Analysis (FEA) to compute the Factor of Safety (FS) across ۴,۰۰۰ distinct parameter combinations, resulting in a robust and high-fidelity dataset. Two distinct ANN architectures were developed: the first model is trained to predict the FS, while the second model estimates the slope angle. The results demonstrate that the models effectively capture the complex, nonlinear relationship between soil properties and slope stability with reasonable accuracy. The ANN models were rigorously validated using two empirical case studies, exhibiting strong concordance between model predictions and observed field performance. The findings substantiate the efficacy of ANN based methodologies as a reliable computational tool for slope stability evaluation, facilitating rapid, accurate, and providing rapid, accurate, and efficient support for geotechnical design optimization and risk mitigation.

کلیدواژه ها:

Artificial neural network ، Factor of safety ، Finite element method ، Roadside slope ، Soil slope stability