Bearing Capacity Prediction of Strip and Ring footings Embedded in Layered Sand

Abstract

In the present study, a prediction model for the bearing capacity estimation of strip and ring footing embedded in layered sand is proposed using soft computing approaches, i.e., Artificial Neural Network (ANN) and Random Forest Regression (RFR). In this regard, the required data for the model preparation was generated by performing a lower and upper bound finite elements limit analysis by varying the properties of the top and bottom layers. Two types of layered sand conditions are considered in the study; namely, (i) dense on loose sand and (ii) loose sand on dense sand. The investigation for strip footing was carried out by varying the thickness of the top layer, embedment depth of the foundation, and friction angles of the top and bottom layers. In the case of ring footing, the internal to external diameter ratio forms an additional variable. The bearing capacity was taken as an average of lower and upper bound values. A total of 1222 and 4204 data sets were generated for strip and ring footings, respectively. The models were trained and tested using 70 % and 30 % of the selected data, and the bearing capacity was predicted in normalized form. The performance measures obtained during the training and testing phase suggest that the Random Forest Regression model outperforms the ANN model. Additionally, following the literature, an analytical model was developed to predict the bearing capacity of strip footing on layered sand. The ANN and the generated analytical model predictions agreed with the published experimental data in the literature.