Prediction of ultimate bearing capacity of circular foundation on sand layer of limited thickness using artificial neural network

Abstract

ABSTRACT The present study focuses on the development of an Artificial Neural Network (ANN) model equation to estimate the average ultimate load of shallow circular foundation on a sand layer of limited thickness underlain by a rigid rough base subjected to an eccentrically inclined load. The model is developed using the 260 number of data points obtained from extensive laboratory model tests. The input parameters considered are depth of embedment (Df /B) of the foundation, thickness of the sand layer to diameter of foundation ratio (H/B), eccentricity ratio (e/B), and load inclination to friction angle ratio (α/ϕ) to estimate the reduction factor RF as output. The RF is the ratio of the ultimate eccentrically inclined load on a sand layer of limited thickness to the ultimate eccentrically inclined load on a sand layer, which is extending to a great depth. Importance of input parameters, which reflect the output, is studied using Pearson’s correlation and Spearman’s rank correlation. The sensitivity analyses are carried out using Variable Perturbation methods and Weight methods. It has been professing that H/B ratio is the most important input parameter.