The Uplift Capacity Prediction for Regular and Enlarged Piles in Sandy Soils Using Artificial Neural Networks

Abstract

The uplift capacity of piles is considered as a crucial aspect in practice for a geotechnical engineer. Nowadays, artificial machine learning technique has emerged as a powerful tool in engineering for prediction and estimation with reasonable accuracy. This paper investigates the uplift capacity of two types of single piles; regular and enlarged piles installed in sand using artificial neural network (ANN). Different activation functions have been used and the ANN results were compared with other algorithms. The results showed that one unified machine learning model has proven its efficiency to give reasonable and accurate estimates of the uplift capacity of regular and enlarged piles. The ANN algorithm had the best results compared with other algorithms (Random forests, XGBoost and Adaboost) with coefficient of determination R² equals to 0.970151 and 0.96924 for training and testing data respectively while other algorithms showed a sign of over-fitting. Finally, the ANN model was compared to well-known theoretical models and the ANN had better results.