Developing a novel deep learning optimization method for forecasting soil friction gradient

In this study, the crucial factor in geotechnical design, soil friction, is predicted using artificial neural networks (ANN). Additionally, the Stochastic Henry gas solubility (SHGSO) optimization approach was used to determine the best number of hidden layers and neurons in each hidden layer, that is, the optimal structural ANN parameters to enhance the performance of ANN. We suggested an innovative way for forecasting Soil Friction Gradient (SFG) using a Stochastic Henry gas solubility optimization-based improved artificial neural network (SHGSO-IANN). A database with 316 laboratory test results was collected for forecasting the SHGSO-IANN model. This research aims to improve comprehension of internal SFG determination by investigating experimental and theoretical approaches. The main goal is to evaluate the viability of a theoretical approach that has been offered as a substitute for conventional exploratory testing to estimate soil friction angle. The SHGSO-IANN method is assessed based on its RMSE (5.74), R2 (96.7%), MAE (5.21) and MAPE (5.95), demonstrating robust performance metrics. This work requires an integrated approach that combines theoretical and experimental techniques to improve the effectiveness of internal friction angle calculation in soil mechanics.