The measurement of the deformation properties of Cowden Till at small strains
The work described in this thesis was firstly concerned with developing and evaluating automated soil testing equipment and associated instrumentation. The equipment consists principally of a triaxial stress path cell of the Bishop-Wesley type, a microcomputer and two pressure controllers. Inductive displacement transducers have been mounted inside the cell to measure axial and radial strains locally on the specimen boundary and axial strains between the end caps. The local axial strain measurements have proved superior to the end cap measurements which can be adversely affected by bedding errors and misalignment of the transducers relative to the loading axis. Following the development, the system was used to investigate the stress-strain behaviour of Cowden Till, particularly at small strains (0.01 - 0.10%). Cylindrical blocks of 250mm diameter were retrieved from the site and stored under isotropic stress. Eight specimens of 100mm diameter were trimmed from these blocks and subjected to either a drained or undrained compression test under load-controlled conditions. Cowden Till has been shown to exhibit strongly non-linear stress strain behaviour, even at small strains, and most of the shear strain is irreversible. The stress-strain characteristics were in acceptable agreement with those derived from a 865mm diameter plate loading test with under-plate instrumentation. Although the interpretation of the plate test is still being investigated, it is concluded that plate tests provide no better information about the stiffness of the material than triaxial tests of the type described in this thesis. The experimental stress-strain behaviour during compressive loading has been compared with the predictions of some mathematical models. The nonlinear elastic model of Atkinson (1973) appears to be applicable to Cowden Till, for which the behaviour is approximately isotropic. Simple stiffness predictions on the basis of critical state soil mechanics are inadequate at small strains. However, the model of Pender (1978) predicts the behaviour reasonably well. (ii) An attempt has been made to analyse the compression (bedding error) which occurs at the end of a triaxial specimen as the axial strain is increased. A quantification of the compression is hindered by the random nature of surface variations and by the limitations of present theories.