Prediction of Maximum Reinforcement Load of Reinforced Soil Retaining Walls Based on Machine Learning

In the design of reinforced soil structures, it is very important to estimate the loads and strains of reinforcement accurately for the stability of retaining walls. However, most of current theoretical methods can only consider limited factors due to no proper approach for considering more complex condition. In this study, by means of machine learning, 12 related factors affecting the reinforcement loads were chosen, and these factors are closely related to the wall geometry, overlying loads, facing, reinforcement and backfill. Meanwhile, three machine learning algorithms, namely, Support Vector Regression, Artificial Neural Network and XGBoost, were adopted to propose the corresponding prediction model for predicting the maximum reinforcement loads from the wall external features and internal material parameters, and a database of 196 sets of data were employed to train and validate the models. It was found that there is a good agreement between the predicted values and the measured values of the validation sets, especially the XGBoost model with stable and reliable performance. Therefore, machine learning can be regarded as a powerful tool to quickly evaluate the performance of reinforced soil retaining walls in engineering design.