Lateral Bearing Capacity of Pile Foundations in Sandy Soils using Neural Network Analyses

In the past decades, there have been a number of theoretical methods developed to calculate the lateral bearing capacity of deep foundations, and especially short and long piles. Most of these methods are based on the soil reaction models such as the coefficient of lateral pressure and soil passive pressure. The associated calculations would also usually follow the allowable strength design by applying appropriate factor of safety in regard to the bearing capacity of the surrounding soil or limited lateral deformation of pile and soil. However, with more applications of pile foundations in the structures in the recent years, and consequently more tests and investigations of the behavior of piles under lateral loads, it has become clearer that the behavior of a pile under lateral loads is more a problem of combined soil and structure reactions, where the deformation of the pile would heavily depend on the reaction of the soil, and vice versa. In the past few years, it was suggested that due to the complicated interaction of soil and pile, both behavior of pile and soil should be modeled simultaneously in a comprehensive analysis to have more realistic results. However, due to these complications, most of the analytical models have generally applied many simplifications to the models, which might sometimes lead to under designing (risky) or over designing (expensive) of the pile systems due to the inaccuracies of the models. Therefore, in this paper, a set of sensitivity analyses have been performed on a number of different short and long piles, using the neural network analyses. As the result, the effects of various parameters on the ultimate lateral bearing capacity of pile foundations were discussed to identify the most important ones. At the end, a new bearing capacity equation has been developed and proposed in the paper that would consider various weighted parameters of combined reactions of soil and pile that provides more realistic results.

Keywords:

Lateral bearing capacity , Neural network , Pile foundation , Sensitivity analyses