Prediction of compression index of fine-grained soils using statistical and artificial intelligence methods

Abstract

Compression index is the slope of the void ratio-effective stres (log) curve obtained in the odometer test. It is an important parameter used to predict consolidation settlement of fine-grained soils. In this study, fuzzy logic and artificial neural Networks methods that rapidly evolved and widely used in many disciplines in recent years, have been employed to estimate the compression index values of fine-grained soils using their index properties. Laboratory data from 285 samples were collected from the literature. 200 of this data were used in the training phase and 85 data were used in testing phase. Multiple regression analysis was conducted to determine the effect of the independent variable on the dependent variable of compression index. The results suggest that liquid limit, natural water content, plasticity index, natural unit weight, void ratio and effective stress variables are the significant parameters that affect the compression index. The results indicate that compression index can best be estimated by the use of fuzzy logic methods. Articial neural networks method is the most suitable method model to estimate predicting C-c from index properties.