Abstract

For the past few decades, the subject of pile group action has been of interest to many researchers in the area of foundation engineering. Closely placed piles interact with each other through the surrounding soil upon loading and block failures are more likely to occur in this case. Therefore, the objective of this research is twofold: first, to evaluate the reliability of existing design theories; and second, to develop a new model that eliminates the shortcomings of the existing theories. To fulfill the first objective, the results of several laboratory and field tests were obtained from the literature and compared with the pile Group efficiency calculated using the existing design theories. This comparison revealed the inadequate accuracy of these theories in addition to their contradictory predictions. To fulfill the second objective, artificial neural networks (ANN), one of the artificial intelligence techniques, was used to develop a computer model that predicts pile group efficiencies. This model benefits from the actual data that are available in the literature to link the pile group efficiency variable with several governing parameters, such as the method of pile installation, soil condition, cap condition, type of loading, pile cross section, pile length/diameter ratio, pile spacing/diameter ratio, and pile arrangement. Validating the ANN model using a set of data that is different from the one used in model development has indicated that the ANN model has better performance characteristics (i.e. efficiency, consistency, and accuracy) than existing design theories. In addition, the developed ANN model can be easily updated when new data becomes available and further extended to accommodate new design parameters. (Abstract shortened by UMI.)