The characterization of unsaturated soil behaviour from penetrometer performance and the critical state concept

Abstract

The proportions of the Critical State surfaces of a given soil depend on its stress and moisture history, its current moisture status and soil type. The influence of these factors on the geometry of the state surfaces, and their bearing on the mechanical behaviour of the soil itself, cannot be readily perceived because of the very large number of interacting effects present in the model. The use of sophisticated computer graphics for the three-dimensional visualization of critical state surfaces would therefore provide a ready means for obtaining an insight into the complex changes taking place in critical state space. The initial part of the thesis deals with the formulation of mathematical models and computer software for integrating numerical computations, data reduction and visualization techniques for analysing critical state surfaces. The programmes developed were used for interpreting the influence of moisture status on the key behavioural patterns of three different British Soils. Systematic changes to state space due to the combined influence of moisture content variations and soil type were readily traced. It is well known that collating data for the above analysis is difficult and requires advanced measuring techniques. An attempt was therefore made to establish a connection between the data obtained from a simple field measuring device, such as a cone penetrometer, with the volume-change behaviour of soil, as modelled by its critical state surfaces. This is attempted in two stages. The first stage presented in the thesis assumes the soil to be a rigid-plastic Mohr-Coulomb material and deals with the formulation of a mathematical model to predict cone index as a function of cone geometry, penetration depth, c, cp and soil-to-metal parameters cQ and S. This model is based on the extension of the basic two-dimensional Sokolovski solution to the three-dimensional slip-line field developed during the deep penetration phase of a cone. Shallow penetration depths, at which the standard Sokolovski rupture surface interacts with the soil surface, cannot be dealt with by this approach. The second stage of the investigation attempts to connect cone index with the stress (p, q) and pore space (v) parameters of the soil on the cone surface. The model developed in the thesis is based on identifying a state parameter yr (defined by Been and Jefferies for dry sands). This establishes the position, within critical state space, of the cone surface stress and pore space parameter (p, q and v) relative to the critical state wall. The state parameter yr is then associated with soil type and moisture 2 content by a two-parameter linear function. Once these two parameters are found experimentally, cone index can be readily translated into pore-space estimates. The thesis presents the mathematical analysis which provides the basis for this correlation. The thesis describes the experimental investigations carried out to verify the performance of the theoretical models developed. The validation of the state parameter concept required the design and development of a special calibration chamber which could apply controlled boundary stresses to a cylindrical soil sample into which the penetrometer is advanced. Ideally very large sample diameters are required to minimise boundary interference, but a compromise had to be made by using miniature penetrometers and a realistic sample diameter of 100 mm. The cone penetrometer performance model was tested under laboratory conditions in an indoor soil tank. Both these investigations required tedious back-up laboratory experimentation to establish the basic Mohr-Coulomb and critical state parameters of the test soil over a wide range of moisture contents. All the soils were dealt with in a remoulded state as consistently reproducible stress and moisture histories for this case can be easily maintained in each of the very large number of samples required in the experimental programmes. The experimental work shows very clearly that the state parameter concept is applicable to partly saturated c-Sp soils over a wide range of moisture contents and that it is possible to quantify the systematic changes in the state parameter y' with soil moisture content. The predictive performance of the cone penetrometer model, within the specified penetration range, was also good. Data reduction charts for interlinking these two models are presented and the use of these charts for the derivation of pore space particulars from cone index data predicted satisfactory trends. However, this procedure appears to over-predict dry bulk density by a considerable margin. The validation presented in this study is for a single sandy loam soil. Even though the overall predictive performance of the mathematical models in this particular soil is most encouraging, it should be borne in mind that the models developed are bound to be influenced by the drastic simplifications required to interlink two disparate models, one which ignores volume change with one which does not. Further work is required to remove any detrimental consequences of these compromises and to introduce confidence in extending the findings to other soil types.