Enhancing the Prediction Capabilities for Barrette Wall Displacement Using the Finite Element Method and Artificial Neural Networks

This study aimed to evaluate the lateral displacement of barrette walls in underground construction projects using the Finite Element Method (FEM) and an Artificial Neural Network (ANN). Lateral displacement analysis is crucial to ensure the safety and stability of structures, especially under complex geological conditions. FEM is used to simulate the displacement of barrette walls with varying thicknesses (400, 600, 800, and 1000 mm), while the ANN is applied to predict displacement based on FEM data, enhancing accuracy and analysis efficiency. The FEM results were compared with real-world data obtained from an inclinometer for an 800 mm thick barrette wall, showing a high correlation that confirms the model's reliability. The ANN model achieved high R² values in predicting displacements, with 0.9998 for the Uy direction and 0.9887 for the Ux direction, demonstrating its ability to accurately replicate lateral displacements. This study confirms that the combination of FEM and ANN can improve displacement prediction capabilities, laying the foundation for optimizing design and ensuring safety in underground structures. By utilizing FEM and ANN, this study not only enhances the accuracy of analyses but also contributes to the development of more effective prediction tools, supporting the design and construction processes of underground infrastructure in complex urban environments.