Predicting the Displacement of Single Battered Pile in Sandy Soil under Pullout Loading Using Artificial Neural Network

The displacement of battered piles is one of the most critical parameters in the design of a pile foundation. In this study, an Artificial Neural Network (ANN) algorithm was utilized to predict the displacement of piles in sandy soils subjected to pullout loading. A finite element analysis (FEA) in three dimensions, performed with the PLAXIS 3D program, was utilized to gather 2380 databases, including the length/ diameter of pile, pullout, batter angle, Poisson ratio, friction angle, dilatancy angle, relative density, and Young’s modulus as input variables, whereas the displacement of battered piles was considered an output variable. The dataset was divided into three parts: training (80%), validation (10%), and testing (10%). The performance of the Artificial Neural Network (ANN) algorithm was evaluated using the Mean Squared Error (MSE) and the Coefficient of Determination (R^2). This study applied a procedure known as a backpropagation neural network. According to the analysis of relative significance, the pullout load (Pu) and the pile length to its diameter (L/D) were the most effective characteristics among the other inputs. The R-values of the ANN model for the displacement of the battered piles dataset were 0.99 across all three phases of testing, validation, and training. The findings substantiated the viability of employing Artificial Neural Networks as a successful method for obtaining the displacement values of a single battered pile in sandy soil when subjected to pullout loading.