Physical and numerical modelling of the soft soil ground improved by deep cement mixing method

Abstract

The research described in this thesis focuses on the consolidation behaviour and vertical bearing capacity of soft soil ground improved by Deep Cement Mixing (DCM) method. The soft soil ground modelled in this study may support relatively light structures, reclaimed fills or road embankments. Firstly, an axisymmetric physical model (Model 1) test was carried out to investigate the consolidation behaviour of soft soil ground installed with a single DCM column. The surface settlement, excess pore water pressures at different locations in the soil, and pressures carried by the soil and the DCM column were all measured throughout testing. This model test revealed that the DCM column behaved as a vertical drain partially, which suggests that the DCM column might be regarded as a partial or full vertical drainage, somewhat similar to a Prefabricated Vertical Drain (PVD) or a sand drain in the DCM improved ground. Under the approximate rigid loading, the pressure on the surrounding soil was progressively transferred to the DCM column, which caused an increase in stress concentration ratio. The stress concentration ratio was also found to be dependent on both the external pressure and the degree of consolidation of the surrounding soil. On the basis of the findings of Model 1 test, another axisymmetric model (Model 2) test on the soft soil installed with a PVD strip was performed. This model test was attempted to make a comparative study on the dissipation of excess pore water pressure mechanism between the soil ground installed with DCM column and the soil ground installed with PVD strip. Compared to Model 1 test, the excess pore water pressure in the soil of Model 2 test was observed to dissipate slowly under the same load increment. This difference was mainly due to the different changing patterns of the surcharge carried by soft soil in the two model tests. The surcharge carried by soil was nearly constant for the soil with PVD strip in Model 2 test, but reducing gradually for the soil ground with DCM column in Model 1 test. After Model 1 test and Model 2 test, Model 3 test was carried out. Model 3 was a plane strain model consisting of a soft soil ground installed with nine mini DCM columns. The DCM columns were covered by a thin sand layer and rigid plate, which was then subjected to a vertical load. Model 3 test was used to investigate the vertical bearing capacity, and the failure mode of the soil ground improved by a DCM column group. Establishment of the model ground and the arrangement of transducers are both presented in detail. Throughout the test, surface settlement, pore pressures and the stress carried by soil were all measured. The recorded load versus displacement curve represents pronounced softening, which implies the progressive failure of the column group in the model ground. It was observed that responses of pore pressure were dependent on not only the drainage condition but also the local failure of DCM columns and the progressive failure of the column group. However, effective stress paths of soil elements in the ground could be useful tool to interpret the pore pressure responses reasonably. Moreover, post-test study verifies a wedge-shaped failure mode in the model test. It is possible that this type of failure pattern is explored experimentally for the first time. Numerical modelling was carried out in an attempt to improve understanding of the consolidation process and the failure mechanism of the improved soil ground. In order to take account of the time-dependent behaviour of soft soil, an Elastic Visco-Plastic (EVP) model is needed. For this purpose, a three-dimensional (3-D) EVP constitutive model originally proposed by Yin and Graham (1999) has been incorporated into a Finite Element (FE) package, ABAQUS, by means of a User MATerial (UMAT) FORTRAN subroutine. This model is formulated as incremental stress-strain relationship for use in FE analysis. For the purpose of verification, a number of single element numerical experiments were carried out to evaluate the overall performance and the efficiency of the UMAT subroutine. Finite element package associated with UMAT developed in this study was employed to conduct FE analysis on two model tests (Model 1 and Model 3). The softening behaviour of the DCM column has been incorporated by employing Mohr-Coulomb constitutive model with the progressive reduction of the cohesion with the deviatoric plastic strain. In terms of Model 1 test, fairly good agreement between measurements and computations is obtained. In particular, the sudden changes of the excess pore pressure resulted from the DCM column failure has been well captured. Based on a series of parametric studies, the permeability of the DCM column and the viscosity of the surrounding soil were found to have an influence on the dissipation of excess pore pressure in the soil, but little influence on the stress concentration ratio. Finite element analysis was also performed on Model 3 test. The results reveal that there is good agreement between computed and measured load versus displacement curve. The vertical bearing capacity of the DCM treated soil ground has been well predicted. Besides, the degradation process of the vertical bearing capacity has been reproduced successfully with the introduction of the softening behaviour of the DCM columns. Two field cases involving DCM method have been simulated using FE models with the UMAT. The permeability of DCM columns has been set to be much higher than that of surrounding soil. Computed results are in good agreement with field measurements. It is demonstrated that the permeability of the DCM column has a negligible influence on the deformation of the improved soil ground. This also indicates that a successful prediction of the performance of soil ground with DCM columns shall take account of the permeability of the DCM column. In order to identify the influence of soil viscosity on the current problems, one more elastic plastic (EP) FE analysis has been made, in which the Modified Cam-Clay (MCC) model is used to model the soft soil. A comparison between them reveals that EVP modelling is capable of providing better results than EP modelling, particularly for the long-term deformation. At the end of this thesis, a summary of the major conclusions and some recommendations for further studies are presented.