Predicting the strut forces of the steel supporting structure of deep excavation considering various factors by machine learning methods

The application of steel strut force servo systems in deep excavation engineering is not widespread, and there is a notable scarcity of in-situ measured datasets. This presents a significant research gap in the field. Addressing this, our study introduces a valuable dataset and application scenarios, serving as a reference point for future research. The main objective of this study is to use machine learning (ML) methods for accurately predicting strut forces in steel supporting structures, a crucial aspect for the safety and stability of deep excavation projects. We employed five different ML methods: radial basis function neural network (RBFNN), back propagation neural network (BPNN), K-Nearest Neighbor (KNN), support vector machine (SVM), and random forest (RF), utilizing a dataset of 2208 measured points. These points included one output parameter (strut forces) and seven input parameters (vertical position of strut, plane position of strut, time, temperature, unit weight, cohesion, and internal frictional angle). The effectiveness of these methods was assessed using root mean square error (RMSE), correlation coefficient (R), and mean absolute error (MAE). Our findings indicate that the BPNN method outperforms others, with RMSE, R, and MAE values of 72.1 kN, 0.9931, and 57.4 kN, respectively, on the testing dataset. This study underscores the potential of ML methods in precisely predicting strut forces in deep excavation engineering, contributing to enhanced safety measures and project planning.