379 Unsaturated Soils: Theory and Pracice 2011 Joisankasa, Sawangsuriya, Soralump and Mairaing (Editors) Kasetsart University, Thailand, ISBN 978-616-7522-77-7 1 INTRODUCTION Describing the mechanical behaviour of unsaturated soils based on an effective stress equation, like the one proposed by Bishop and Donald (1961), and its limitations has led to different approaches for mod- elling the behaviour of these soils. Regardless of the approach adopted for modelling the soil behaviour (effective stress-based or based on independent state variables), soil suction plays a significant role in de- scribing the soil behaviour. Soil–water characteristic curve (SWCC) is one of the most important parts of any model for describing unsaturated soil behaviour because it describes the variation of soil suction with changes in water content. SWCC can be viewed as a continuous function describing the water storage ca- pacity of a soil as it is subjected to various soil suc- tions. SWCC contains important information regard- ing the amount of water contained in the pores at a given soil suction and the pore size distribution cor- responding to the stress state in the soil (Fredlund et al., 2002). Different aspects of unsaturated soil be- havior such as shear strength, volume change, diffu- sivity, and adsorption are related to the SWCC (Fredlund and Rahardjo, 1993). There are several methods available for obtaining the SWCC for a particular soil. SWCC may be de- termined directly or indirectly in the laboratory. Di- rect methods include pressure plate, Buchner funnel, tensiometers, and pressure membranes. These meth- ods measure the pore-water pressure in the soil or impose a known air pressure to soil and allow the water content to come to equilibrium with the im- posed air pressure. Among these methods, conven- tional pressure plate test (ASTM D 6836) is the most common method. Indirect methods include filter pa- per and heat dissipation sensors. These methods use measurements or indicators of water content or a physical property that is sensitive to changes in wa- ter content. However, these experiments are costly and time consuming. Therefore, several empirical methods have been proposed in the literature to de- termine SWCC for soils. These methods can be clas- sified into four major groups: (i)Water contents at different suctions are correlated to specific soil properties such as D10 (sieve size for 10% passing) and porosity. This process generally requires a re- gression analysis followed by a curve fitting proce- dure. Among important contributions are the works of Hutson and Cass (1987), and Aubertin et al. (1998); (ii) Correlating parameters of an analytical equation with basic soil properties such as grain size distribution and dry density using a regression analysis. Among important contributions are the ABSTRACT: An evolutionary data mining technique that generates a transparent and structured representa- tion of the behaviour of a system directly from data is used to model soil-water characteristic curve in unsatu- rated soils. The evolutionary polynomial regression (EPR) is a method that can operate on large quantities of data in order to capture nonlinear and complex interactions between variables of the system. It also has the additional advantage that it allows the user to gain insight into the behaviour of the system. Results from pres- sure plate tests carried out on clay, silty clay, sandy loam, and loam are used for developing and validating the EPR model. The model inputs are the initial void ratio, initial gravimetric water content, logarithm of suction normalized with respect to atmospheric air pressure, clay content, and silt content. The model output is the gravimetric water content corresponding to the assigned input suction. The EPR model predictions are com- pared with the experimental results as well as the models proposed by previous researches. The results show that the proposed approach is very effective and robust in modelling the soil-water characteristic curve in un- saturated soils. KEYWORDS: unsaturated soil, SWCC, pressure plate test, data mining Modeling soil-water characteristic curve using EPR A. Ahangar-Asr & A.A. Javadi A. Johari College of Engineering Mathematics and Physical Sciences, University of Exeter, UK aa375@ex.ac.uk; a.a.javadi@ex.ac.uk Department of Civil and Environmental Engineering, Shiraz University of Technology, Iran johari@sutech.ac.ir