- Copyright: © 2010 This is an Open Access article: verbatim copying and redistribution of this article are permitted in all media for any purpose, provided this notice is preserved along with the article's original URL.
This article addresses one approach that could be used for automating a soil stabilization system. Instead of manually guiding the injection of a binding agent, a scheme is introduced where pre-determined subsurface conditions of the target volume will control the injection. Many factors influence soil stabilization, but the single most important factor is water content. Several methods to measure and quantify water content were considered. Of these methods, geoelectrical resistivity sounding was deemed best suited because it offers bulk information of water content distribution in the soil, whereas other methods tend to give more localized sampling. Water content information is input for multivariable statistical models to estimate the amount of binding agent needed for target shear strength. Statistical models were created based on a large soils database that was established during the study from a multitude of stabilized soil samples of various types.
While resistivity and water content are known to be inversely related to each other, transforming resistivities to water contents is always a site specific issue that must be approached with care. In a soil stabilization context, this means specifically that the surface conductivity of clay particles should always be considered. Stabilized areas often contain clay, which prohibits the use of Archie's law for transforming resistivities to water contents because this law assumes a resistive surrounding. In our study, we tried a model that is an extension of Archie's law with an extra term depicting the surface conduction effect. Comparing modeled water contents with those measured in the laboratory, we found an agreement within 20%-units (weight).
Statistical models for soil shear strength included three different mixing agents: pure cement, 1:1 mixture of lime + cement and pure lime. Each model contained several variables such as temperature, pH, water content and humus content. Assuming that values for these variables can be measured or estimated, the amount of the binding agent can be determined from the formulas. This procedure, coupled with a precise positioning system, forms the basis of an automated stabilization system where the amount of binding agent is calculated for each location in the soil separately by a machine that holds statistical models in its memory.
Also during the project, an index method to expedite laboratory tests for stabilized soil samples was developed for means to verify estimated shear strengths given by the statistical models, and to offer an independent way to calibrate these models.