ANN Model for Estimating Ultimate Bearing Capacity of Piles Socketed in Dubai Limestone

1. Home
2. Publications
3. Online Library
4. ANN Model for Estimating Ultimate Bearing Capacity of Piles Socketed in Dubai Limestone

An accurate prediction of the ultimate bearing capacity of piles is crucial for a reliable deep foundations design. Several approaches are used to estimate the ultimate capacity of piles including numerical modelling based on soil properties and soil/pile interaction, yet the most accurate way remains the in-situ tests results. Static and dynamic pile load tests are the most common ones and they are a mandatory requirement in any deep foundations project. Recently, Artificial Neural Networks (ANN_S) are applied effectively in a wide range of complex problems in geotechnical engineering. The objective of this study is to create a model using ANN to predict the ultimate bearing capacity (Q_ult) of piles socketed in rock. A database of 82 High Strain Dynamic Pile Load Tests (HSDT), collected from different sites in Dubai yet all socketed in Dubai limestone formation, is used to develop the ANN model. The developed model accounts for the pile geometry presented in pile socket length to its diameter, the rock characteristics as in rock Uniaxial Compressive Strength (UCS), in addition to the dynamic pile test key parameters in particular hammer weight and hammer drop height. The results show that the ANN provides a reliable prediction tool for piles ultimate capacity with coefficient of determination values (R²) of 0.94 and 0.965 in the training phase and testing phase respectively and a normalized Root Mean Squared Error (RMSE) less than 5% of the data range in both phases. A sensitivity study is also conducted to identify which input parameters most significantly impact the predicted ultimate capacity. Based on the analysis, it is detected that both the rock UCS and the HSDT impact characteristics have the most effect on the ultimate pile capacity prediction.

4th Asia-Pacific Conference on Physical Modelling in Geotechnics (ACPMG2024)

Other