Accurate Recurrent Neural Network for Active Earth Pressure Coefficient Estimation

Compared to traditional civil engineering approaches, artificial intelligence (AI) applied to data-centric systems has shown extraordinary predictive ca-pabilities. As a result, Artificial Intelligence (AI) is frequently used to model the complexity of behavior of a large number of geotechnical engineering materials. This study proposes a new recurrent neural networks (RNN) model to predict accurately the lateral earth pressure coefficient Kaγ of a retaining wall for different geometric configurations on frictional soil. The performance of the RNN model has been assessed using the determination coefficient and mean square error. The results obtained are contrasted with those obtained by the analytical and numerical methods. It has been demonstrated that our model gives more accurate outcomes than analytical and numerical techniques with a precision of up to 10 ⁻⁴ .