Bearing capacity and settlement prediction of multi-edge skirted footings resting on sand

Tammineni Gnananandarao,Vishwas Nandkishor Khatri,Rakesh Kumar Dutta

doi:10.15446/ing.investig.v40n3.83170

Abstract

This paper presents the application of artificial neural networks (ANN) and multivariable regression analysis (MRA) to predict the bearing capacity and the settlement of multi-edge skirted footings on sand. Respectively, these parameters are defined in terms of the bearing capacity ratio (BCR) of skirted to unskirted footing and the settlement reduction factor (SRF), the ratio of the difference in settlement of unskirted and skirted footing to the settlement of unskirted footing at a given pressure. The model equations for the prediction of the BCR and the SRF of the regular shaped footing were first developed using the available data collected from the literature. These equations were later modified to predict the BCR and the SRF of the multi-edge skirted footing, for which the data were generated by conducting a small scale laboratory test. The input parameters chosen to develop ANN models were the angle of internal friction (ϕ) and skirt depth (Ds) to the width of the footing (B) ratio for the prediction of the BCR; as for the SRF one additional input parameter was considered: normal stress (𝛔). The architecture for the developed ANN models was 2-2-1 and 3-2-1 for the BCR and the SRF, respectively. The R2 for the multi-edge skirted footings was in the range of 0,940-0,977 for the ANN model and 0,827-0,934 for the regression analysis. Similarly, the R2 for the SRF prediction might have been 0,913-0,985 for the ANN model and 0,739-0,932 for the regression analysis. It was revealed that the predicted BCR and SRF for the multi-edge skirted footings with the use of ANN is superior to MRA. Furthermore, the results of the sensitivity analysis indicate that both the BCR and the SRF of the multi-edge skirted footings are mostly affected by skirt depth, followed by the friction angle of the sand.

Highlights

The prediction of the bearing capacity and the footing settlement with a reasonable accuracy is required in the field of foundation design and affects the overall economy of a project
The bearing capacity of the footing on sand was predicted by Kalinli et al (2011) using artificial neural networks (ANN) based on 97 datasets by varying the footing width, embedment depth, geometry, unit weight of sand, and the friction angle of the cohesionless soil
The study carried out in this work showed the feasibility of using a simple ANN and multivariable regression analysis to predict the bearing capacity ratio and settlement reduction

Summary

Introduction

The prediction of the bearing capacity and the footing settlement with a reasonable accuracy is required in the field of foundation design and affects the overall economy of a project. On similar lines with Kalinli et al (2011), other researchers have explored the application of ANN on the prediction of the bearing capacity of footings resting on sand and rock The data for regular shaped (square and circular) skirted footings were collected from the published literature, whereas the data for the multi-edge (T, Plus, Double box) skirted footings were generated through experimentation in the laboratory. To model the settlement reduction factors, the considered input variables were skirt depth to width ratio of the footing, friction angle of the sand, and normal stress. The outputs for these ANN models were bearing capacity ratio and the settlement reduction factor, respectively

Objectives

Results

Conclusion