An Engineering Model of Hysteresis for Soil-Water Characteristic Curves

Abstract

Soil-water characteristic curves play an important role in understanding the behaviour of unsaturated soils. Soil-water characteristic curves have been used to estimate the hydraulic conductivity and shear strength functions. The hysteretic nature of soil-water characteristic curves has been known for a long time but in most engineering and agriculture applications the soil-water characteristic curves have been assumed to be unique. It is not possible to measure all possible hysteretic soil-water characteristic curves for a soil. A simple model is required to predict hysteretic soil-water characteristic curves based on a few measured data points in order to apply the effect of hysteresis in the engineering practice. A summary of the development of the various models that have been proposed in the literature to predict soil-water hysteresis is presented in this thesis. Some modifications to the existing soil-water hysteresis models are described. Seven soil water hysteresis models are selected and compared based on datasets collected from the research literature and obtained from a laboratory test program. In addition, an investigation of the hysteretic characteristics of the soil-water characteristic curve was implemented. Six soil-water hysteresis models found in the research literature; namely, 1) Mualem (1974) model-II, 2) Mualem and Miller (1979) model-IIIexp, 3) Mualem (1984) independent model, 4) Mualem (1984) dependent model, Hogarth et al. (1988) model and M. Feng and M. Fredlund (1999) model were selected for comparison. One more hysteresis model; namely, the Simplified version of the Feng and M. Fredlund (1999) model were added for comparison. The comparison of the soil-water hysteresis models was based on three criteria; namely, 1) Percentage deviation, 2) Absolute percentage deviation and 3) R squared. The results showed that the Man Feng and Fredlund (1999) model using two measured data points is the most appropriate model for predicting the main wetting curve. The Mualem (1974) model-II is the best model for predicting scanning curves. The R squared criterion is not a reliable criterion for comparing soil-water hysteresis models. The investigation of the hysteretic nature of the soil-water characteristic curve found that there are some consistent relationships between the initial drying curve, the main drying curve and the main wetting curve. These relationships provide a simple method to estimate the initial drying curve from the main wetting curve or the main drying curve. The study also indicated that some modifications to the measured data points are needed in order to obtain reasonable best-fit hysteretic soil-water characteristic curves. In practice, it is hard to know the hysteretic history of the soil-water characteristic curve (Le., wetting and drying cycles); therefore it is not applicable to apply a soil-water hysteresis model requiring the hysteretic history of the soil-water characteristic curve. A method of predicting scanning curves passing through any given point inside the main hysteresis loop does not require that the hysteretic history of the soil sample be known is presented in this thesis. This method is simple and applicable in engineering practice.