4.2 Excavation Supported with Unstrutted Diaphragm Wall

Two centrifuge tests with test number To1 and To2 had been carried out to study the behaviour of excavations supported by unpropped diaphragm wall.

4.2.1 Preparation and instrumentation

The soil model was prepared in accordance to the method stated in Chapter 3. The properties of the soils used for these two tests and other tests are shown Table 3.4. As shown in Figures 3.4 and 3.7, there were two nearly uniform layers of soils comprising a soft marine clay of 200 mm thick overlying a stiff soil layer of 170 mm in thickness. For test To1, the undrained shear strength, c_u, ranged from 12.5 kPa to 15.5 kPa for the upper soft clay and 152 kPa for the lower stiff soil. For test To2, c_u of the upper layer was 25.2-27.5 kPa and the c_u of the lower layer was 156 kPa.

In this series of tests, the instrumented model diaphragm wall and the rubber bag containing zinc chloride solution were placed in the cut prior to the experiment. The depth of the cut is 150 mm and the embedded depth of wall into stiff soil is 25 mm, corresponding to a prototype scale of 15 m excavation with the wall embeded 2.5 m into the stiff layer. These tests were deliberately designed to ensure that the retaining wall would fail during excavation. The instrumentation such as the strain gauges for measuring the bending moment of the wall, the PPT for measuring pore pressure in the soil and the LVDTs for measuring the ground surface of these tests are shown in Figure 4.1. The stages of excavation are also given in the same figure.

4.2.2 In-flight Excavation

In the test, in-flight excavation was simulated by draining zinc chloride solution from the rubber bag. The density of zinc chloride solution is chosen to be 17 kN/m³, the same as that of the marine clay used. To simulate the excavation sequence, the draining was divided into 8 stages and the draining rate was controlled at 50 mm/min. This is equal to a rate of 0.72 m/day in prototype scale which is compatible to a medium scale excavation work or may be a little bit faster than the normal practical construction speed in the whole area of a site. A faster rate allows the observation of the change of pore water pressure caused purely by the release of lateral stress. The swelling of soil will cause a drop of pore water pressure. However, the suction and seepage effect over a longer time can conceal this drop.

4.2.3 Changes in pore water pressure

The changes in pore water pressures as obtained from PPT1, 2, 3 and PPTzncl during the reconsolidation before the excavation are shown in Figure 4.2. Figure 4.3 shows the changes in pore water pressure during the excavation process. The drop in pore water pressures in terms of the excess negative pore pressures are plotted in Figure 4.4. The readings of PPTzncl in Figure 4.4 indicate the elevation of the zinc chloride which represents the excavation depth. The readings of PPT1, PPT2 and PPT3 follow a similar trend which shows the negative pore pressure increases progressively with deeper excavation until stage 4. When excavation started, the wall began to move towards the excavated side, and the retained soil would follow and expand. As a result, negative pore water pressure was expected to be developed in the retained soil and this was indeed so. The draining rate of zinc chloride in this test was relatively fast and at such unloading rate, the test could be considered as representing an undrained condition during the excavation as the permeability of the marine clay is low.

4.2.4 Surface settlements

The variation of soil settlements versus time are plotted in Figure 4.5. Due to large soil displacement, the LVDTs soon run out of range with the one closest to the wall being the first one to reach the maximum travel of 10 mm. The settlement profiles at stages 1 and 2 are shown in Figure 4.6. It can be seen that the magnitude of settlement near the wall was very large, but rapidly decreased with distance away from the wall.

4.2.5 Visual observations

The development of the deformation and failure pattern of the soil around the excavation during the test are recorded by a video camera. The profiles of surface settlements after failure, as measured from video pictures using an image processing software as described in Chapter 3, are plotted in Figure 4.7. Photographs obtained from the video pictures for the different excavation stages are as shown in Figure 4.8. Stages 1 and 2 display the soil deformation in the initial stage. Between stage 2 and 3, the shear band appears gradually and then the retained soil fails dramatically. When the excavation reached stage 3, a shear failure surface was clearly visible and the wall has failed. Draining of a heavy solution, just as in a real excavation, causes an unbalanced stress profile at the two sides of the wall leading the wall to move towards the excavation. In the case of no lateral support, the near rigid wall was observed to rotate around the pivot which is near the wall toe, as shown in Figure 4.8. Between stage 2 and stage 3, the settlements of LVDT2 and LVDT3 increase considerably more than the earlier stages, indicating the occurrence of the shear failure. During this stage, the critical excavation depth measured by the pore pressure transducer within the latex bag, in prototype scale was 5 m.

From the photograph of stage 3 shown in Figure 4.8, it can be found that the slip line starts from the rotation point and extends up to the ground level at an angle of about 45 ^o. This angle shows good agreement with the classical theory that the angle of the failure surface is (45^o-f/2) in which f should be zero under an undrained condition. This suggests that during the simulated excavation, the soil behaviour is essentially undrained. Part of the reason for this is the fact that the marine clay used has a low permeability.

4.2.6 Bending moments

The bending moments in the wall for test To1 was measured from the strain gauges placed along the wall as described in Chapter 3. Figure 4.9 shows a change of bending moment versus time. The profile of bending moment along the wall are given in Figure 4.10. The cantilever effect of the wall was exhibited in the initial stages from stage 1 to stage 4. The maximum value of bending moment is a nominal 135 kNm which is located near strain gauge W5. At the later stages from stage 5 to stage 8 as shown in Figure 4.8, the wall has failed and fallen down progressively. The wall was restrained at the wall toe and partially confined by rubber bag at the top of the wall and the weight of soil acts on the wall. The profile of bending moments after failure are shown with curves noted by 180 sec and 240 sec. The changes in the bending moment reading throughout the whole process of centrifuge test illustrate that all the strain gauges were in working order.

4.2.7 Effects of soil stiffness

In order to investigate the effect of soil stiffness on the behaviour of an unpropped excavation, test To2, was performed. Unfortunately, only partial data of the early settlement of soil can be provided because most of the data for this test were corrupted. In this test, a stiffer upper soil of c_u, from 22.5 to 27.3 kPa was constructed compared to a c_u of 12.5 to 15.2 kPa in test To1. Figure 4.11 shows the surface settlement profile before failure. As compared to Figure 4.6, the magnitude of settlements is considerably reduced when the soil is stiffer. At the same excavation depth of 50 mm, which is 5m in prototype scale, the reading from LVDT2 was 15 mm for test To1 and 8.5 mm for test To2. The excavation level when failure occurred was about 35 mm for test To2 as compared to 30 mm for test To1. Figures 4.12 and 4.13 show the photographs of the failure patterns of To1 and To2. The angle of the failure plane for the two tests was similar indicating that the behaviour of soil in test To2 is also almost undrained. Within the wedge failure zone of To2, there were multiple failure planes, this is typical of shear failure pattern in a medium stiff clay. In test To1, only some cracks appeared on the surface. Another difference was that there was a second failure plane during the later stages of test To1, while there was no such secondary failure line for test To2. Unfortunately, due to the data corruption, the changes in pore water pressure and bending moment cannot be presented here.

4.2.8 Equilibrium analyses for stability

Based on the classical theory of earth pressure on retaining wall, the active and passive earth pressures and water pressures acting on the two sides of the retained wall are illustrated in the figures of Table 4.1. Based on the observations from the video pictures, the wall rotated around a pivot point located between the top of the bottom stiff soil and the wall toe. For the analysis presented in Table 4.1, the soil parameters under undrained condition are given and all the dimensions are converted to the prototype scale. From the observation of centrifuge test, the critical depth of excavation was around 3 m for test To1. In order to indicate the effect of Zncl below the excavated surface, the material property below the excavated surface to 15 m deep was modelled as a heavy liquid with density of 17 kN/m³. The critical depth is calculated to be 2.24 m. By referring to Figure 4.8, it can be seen that the failure happened gradually with the excavation, instead of a sudden collapse as in a stiffer soil. The observed critical depth is between 2 m to 3 m which is close to the results of the computations given in Table 4.1. In Table 4.2, another set of calculations are also presented by simulating the stiffer soil of test To2. The effect of Zncl in the centrifuge test is also modelled in the same way as in the calculations for test To1 in Table 4.1. The calculated critical depth of To2 is 3.0 m and the corresponding critical depth from test is around 3 to 4 m, again indicating resonable matching.

This prelimilary set of tests indicates that Rankine theory works reasonably well for a nearly uniform soil in an undrained condition. However, the failure surface cannot be clearly captured at the precise moment when excavation reaches its critical depth until the next excavation stage. This would partially explain the discrepancy between the calculated critical depth and the critical depth estimated from observation of video images.