6.8 Discussions

6.8.1 Stress path and time dependent behaviour

The stress path of the soil at P1 in test Ts5 simulated by CRISP is shown in Figure 6.21. Comparing to Figure 6.11, the stress path at the same position for test To1 is different. As mentioned in Section 4.2.2, the draining rate of zinc chloride solution in To1 is 50 mm/min. which is equal to a rate of 0.72 m / day in prototype scale. Such relatively fast excavation rate make the soil behaviour essentially undrained. The subsequent soil expansion causes the negative excess pore water pressure to increase significantly. In the mean time, the dissipation of such excess pore pressure is slow due to the low permeability of clay. Hence, the stress path goes up vertically, yielding and finally fail as shown in Figure 6.11. The undrained condition of test To1 is indicated by stress path shown by the failure pattern in Figures 4.8 and 4.12.

In contrast, for the CRISP analysis on test Ts5, the excavation rate is about 9 mm/min which is 0.13 m/day in prototype scale. It is a typical excavation process which is much slower than that of test To1. The corresponding magnitude of deformation of soil is also much smaller than those of test To1. In this case, the built-up of negative excess pore pressure is not as significant when compared to test To1. For this reason, the stress path in the analysis of test Ts5 shows a different trend ( Figure 6.21) compared to test To1(Figure 6.11). A partially drained condition is obtained by the reduction of mean normal effective stress due to the dissipation of negative excess pore pressure in a slow excavation work. In conventional engineering practices, typical fully undrained or fully drained condition are normally idealised in most simplified analyses. The undrained condition generally under estimates the wall deflection and ground settlement as the dissipation of negative pore pressure is ignored. In contrast, drained condition is over estimating the soil movements, and is more commonly used especially for the long term trend. The real situation is between these two cases and a better prediction should be modelled by considering excavation rate and the permeability of the soil.

The CRISP program, using coupled consolidation analysis, are able to handle such time dependent problem. The preliminary comparison between the experiment and FEM results shows good agreement and confirms this.

6.8.2 Comparison between the FEM and centrifuge tests

By comparing the results between the FEM and centrifuge tests, it can be found that the surface ground settlement measured in centrifuge tests decreases sharply with distance away from excavation which is in agreement with the behaviour in the field. Typical ground settlement profiles from centrifuge tests, field observation and from FEM are illustrated in Figures 6.23 and 6.24. At a distance of 2 times the excavation depth away from the wall, settlement from FEM is about 10-30% larger than that from centrifuge and field measurements. This can be considered as within an acceptable range in design. At a distance further away from the excavation, difference in settlement between the FEM and centrifuge test results becomes larger.

6.8.3 Effects of small strain behaviour

Recently, there is an increasing awareness that the soil stiffness at very small strain is actually much higher than the stiffness obtained from conventional triaxial testing (Simpson et al. (1979), Jardine et al. (1986), Burland (1989) and Simpson (1992). As FEM adopted the soil parameters obtained from triaxial tests, which is a large strain testing, the stiffness used in the analysis for soil far away from the wall may be too small. Due to this, the deformation from FEM is considerably larger then the experimental results. A fact recognised by many engineers now. This aspect looks increasingly important.

6.8.4 Effects of boundary condition, shear slip surface and interface

Some differences are indicated as shown in Figure 6.23 by comparing the surface settlement profile from centrifuge to that from FEM, which influence the accuracy of the estimation on the surface settlement. First, the small strain effect as discussed in the last section is not properly modelled in FEM, as shown in the Figure 6.23. Second, the existence of a shear slip surface observed in the centrifuge model tests obviously influenced the soil movement field near to the retaining wall. The shear band reflecting the discontinuity of the soil cannot be simulated perfectly by the continuum model in FEM. Third, the condition of the right boundary of the mesh will also affect the settlement profile especially the profile near to the boundary. In Figure 6.20, the right boundary was fixed in vertical direction. In contrast, Figure 6.22 gives the surface settlement profiles when this right boundary is allowed to move freely in the vertical direction. The difference of such effect between the free and fixed conditions is mainly near the boundary and can be reduced effectively by extending the mesh far enough towards the right.

The incorrect simulation of the interface condition between soil and diaphragm can be clearly seen in the analysis as explained in Section 6.7.4. Due to a lack of information and understanding of the behaviour of interface element, such effect is frequently ignored in engineering practice. The previous analyses conducted without considering the interface condition, highlighted such an effect. To verify the improvement in trend if an interface element is included, another analysis was conducted using the slip element provided in CRISP. Slip elements are provided along the entire retaining wall on the active side, but not the passive side. The use of interface elements also on the passive side and wall toe produce numerical instability which is beyond the scope of the present study to solve. Figure 6.24 shows an improved simulation of the ground surface settlement profile when such soil-structure interface elements are used along the model diaphragm wall. Figures 6.25 and 6.26 give some detail comparisons to show the effect of interface and boundary condition on surface settlement profile at excavation depth 50 mm and long term after excavation, respectively. Soil-structure interface elements used along the model diaphragm wall effectively increase the settlement near the retaining wall and reduce the over-estimation of surface settlements further away from the wall. This resulted in a significantly improved trend of the predicted ground movement. The parameters of the interface elements used in the analysis are shown in Table 6.3. This preliminary simulation is just to indicate this effect and further study on the interface behaviour is still needed.