Development of models to predict swell potential based on soil index properties

Abstract

This paper suggests empirical relationships based on statistical analysis relating the swell potential to single and multiple soil index properties for the expansive soils in Kibaha, Tanzania. Correlation analysis identified a number of significant relationships between soil index properties and swelling potential that were used in model development. The analysis identifies strong relationships between swelling potential and a variety of site-specific soil index properties, particularly plasticity index (r=0.888), liquid limit (r=0.803), activity (r=0.881), bulk density (r=0.663), clay content (r=0.642) and shrinkage limit (r=-0.642). The regression found that the swell potential predicted by a univariate regression model involving plasticity index (PI), activity and liquid limit (LL) can reasonably evaluate the swell with the remarkable good coefficient of determination (r²) of 0.79, 0.78 and 0.65 respectively. However, the multivariate equations were found to be the best estimates of swell potential of expansive soils with the highest coefficient of determination. The best performing model had r² of 86.2% and the least performing model had r² of 83.6%. Finally, predictions show good agreement with actual swell potential measurements, thus developed models can be confidently used to predict swell potential from a set of reliable soil-swelling indices.

Keywords: Soil index properties, swell potential.