Maximum Kinematic Pile Moment in Layered Soil Profile: Artificial Neural Network Approach

Abstract

This paper deals with the application of Artificial Neural Networks (ANN) to estimate maximum steady state kinematic pile bending moments in floating piles at the interface of two soil layers of vastly different shear modului. The data generated for training and testing the ANN model is based on Beam on Dynamic Winkler Formulation. The inputs to ANN model are the ratio of the elastic modulus of pile to that of the soil, the length to diameter ratio of pile, shear wave velocity ratio of the two soil layers, and the ratio of the thickness of the top soil to the length of the pile. The output of the ANN model is the normalized maximum steady state pile bending moment at interface of the two layers. Feed forward Levenberg-Marquardt backpropagation algorithm is used to train the ANN model and its performance model is evaluated using test data that forms 50% of the total data. The results show that ANN can be used to estimate the maximum steady state pile bending moment and the corresponding maximum pile moment in time domain using reduction factors.