BEARING CAPACITY PREDICTION OF SHALLOW STRIP FOUNDATIONS BASED ON COHESIVE LAYERED SUBSOIL USING ARTIFICIAL NEURAL NETWORKS

Bearing capacity prediction of shallow strip foundations is one of the most important issues in geotechnical engineering. There are some bearing capacity equations and design charts proposed by researchers for this purpose. Most of the methods are for homogeneous and up to two layered subsoil. Recently, predicting bearing capacity of shallow foundations on multi layered soils have been interested. In this study, an artificial neural network is developed for bearing capacity prediction of shallow strip foundation based on cohesive multi layered subsoil. Training process of ANN needs a wide range of input data. In this study, a FEM is used for generating data for ANN. ANN has been trained for both kind of layered subsoil. Finally a multiple regression analysis which can generate a relationship between input parameters and target data is defined. Validation of neural network shows that an AAN can predict the ultimate bearing capacity of shallow strip foundation on cohesive layered subsoil with an acceptable accuracy. Effect of each parameter on ultimate bearing capacity of foundation using ANN is investigated. The results are compared with FEM and indicate that an ANN can evaluate the effect of input parameters with an acceptable accuracy. Validation of equations proposed for predicting bearing capacity of shallow foundations using multiple regression analysis indicates that this method can fit our purposes.