Validation of Compression Index Approximations Using Soil Void Ratio

Abstract

Published equations for approximations of soil compression index in terms of in situ void ratio are varied and confusing. Many published expressions yield a positive compression index even for soils with zero void ratios. Furthermore, published expressions are normally based on a high correlation coefficient and limited test data. Significant measures such as standard errors are often omitted. Some published equations are limited to fine-grained or organic soils. Most available data exhibit a behavior in which significant scatter is observed for small values of void ratio. High void ratio values associated with organic soils produce high compression indices. The use of compression index in settlement calculation of organic soils deposits and especially peat is not justified. Over the past five years, significant amount of data was collected from published articles in reputable journals and conferences. A total of 1722 data points relating the compression index to soil void ratio were analyzed. Using MATLAB, the authors developed new regression models that relate the compression index of normally consolidated soils to its void ratio and including careful examination of data outliers. Using the methods of Robust Regression, the proposed regression models minimized the effects of data outliers and produced higher correlation coefficients with reduced associated standard errors. Although several artificial neural networks (ANN) regression models are available, such models were based on limited data without dealing with data outliers. The combined data set used is then largest ever assembled and offers clear insights into the behavior of mineral and organic soils. The proposed regression models differ from published empirical approximations by considering the impact of data outliers.