| The knowledge of hydraulic conductivity k in fine-grained soils is essential in the design of civil engineering structures. This parameter can be determined in situ or by laboratory experiments. In this latter case, a lot of experiments and methods exist that can define hydraulic properties of tested materials. This study is based on variable-head permeability tests conducted between the loading steps of oedometer tests. With those tests, there are two ways to determine permeability: indirect methods (using the coefficient of consolidation with Terzaghi theory) or direct methods (semi-log plots representing the head difference against time and the velocity graph method).;The principal objective of this research project lies in comparing these different methods of interpretation. This comparison has been conducted on numerous laboratory tests performed on Champlain Clays from Lachenaie and sand-bentonite mixes. Part of this objective is also to attempt to explain the different k-values. This study is divided into two parts, a detailed analysis of the velocity graph method (Chapuis et al., 1981) and the development of a numerical model able to reproduce laboratory tests.;During the permeability test, the application of an important hydraulic head produces a sample deformation. This deformation tends to bend the initial part the velocity graph and over evaluates the permeability and the hydraulic error h0. Therefore, to obtain a good estimation of k with the velocity graph method, permeability tests must be extended so that the data reaches the linear part of the graph. When the h0 value is larger than 20 cm, k is overestimated by half an order of magnitude. Nevertheless, an error remains even if tests are extended. This error could be caused by chemical gradients which also produce flow by osmosis. Some observations that were not noticeable in the laboratory were enabled by the numerical model. Thus, non-dissipated pore pressures resulting from the loading steps that precede the variable-head permeability tests may impact permeability test results by modifying flow gradients. Also, secondary settlement seems to be active throughout the consolidation. However, the variation of Calpha value is not accounted in the model.;Finally, this model could be improved by increasing the number of simulated tests and by integrating new compartmental laws that describe, for example, large strain deformations or variation of creep. |
