5.5 Discussions

5.5.1 Effect of upper strut

To highlight the behaviour of excavation with and without upper struts, test Ts6 was conducted to compare with the results of test Ts5. The soil condition and set-up for tests Ts5 and Ts6 were the same except that for Ts5, there was an upper strut to prop the wall while for Ts6, there was no strut. In test Ts6, although there was no upper strut to support the wall, the test result showed that there was no failure up to 50 mm excavation and after 6 hours post-excavation high-g spinning. The surface settlement measured by LVDT2 is shown in Figure 5.17. At the same excavation depth of 50 mm, by comparing the settlement at the same location, it was found that the settlement for Ts5 was 0.15 mm while for Ts6, it was 0.5 mm. Although no failure of the retaining structure occurred for tests Ts5 and Ts6, it was observed that there was a difference of more than 3 times in the magnitude of settlement near the wall. Also, at LVDT3 and LVDT4 which are far away from the wall, there was a difference of about two times in the magnitude of settlement, as shown in Figure 5.18.

5.5.2 Effect of embedment depth of wall in stiff layer

Refering to Figures 5.9, 5.10, 5.13 and 5.14, it is noted that the surface settlements nearly double as the embedment depth reduces from 50 mm (Ts5) to 25 mm (Ts4). The curves with the labels Ts4 and Ts5 in Figure 5.17 also give a direct comparison of settlement at position LVDT2. Figure 5.18 gives the comparison for such effect in the form of settlement profiles at 50 mm excavated depth.

5.5.3 Long term behaviour of post-excavation

As we know, after an excavation has been completed, the pore pressure and ground settlement will still be changing. Such time-dependent behaviour is due to three reasons, namely seepage of water toward the excavation, progressive movement of the soil and wall due to soil creep and the further consolidation due to high-g field,, since the soil was not fully consolidated in the present tests.

From the pore pressure data, it is observed that all the pore pressures in the retained soil drop during excavation. As the wall moves towards the cut, the lateral stress is reduced and the soil tends to swell. Due to the low permeability of clay, the negative excess pore pressure dissipates slowly. When the excavation stops, the wall movement slows down and the PPTs readings begin to rebound. This phenomenon is caused by the dissipation of negative pore pressure. With this dissipation, the effective stress of soil would decrease and result in further movement of the wall towards the cut. This is a time-dependent behaviour in deep excavation. In fact, the PPTs readings were the combined effects of stress relief, dissipation of pore pressure and seepage towards the excavation side.

Refering to the pore pressure data (Figures 5.8 and 5.12), it is noted that the negative pore pressure rebounds sharply immediately after the excavation, then slows down and gradually becomes stable. The rebound extends for about 600 seconds, which is equivalent to 70 days in prototype scale after the excavation. This means that, within 70 days after an excavation, the effective stress of soil decrease with the dissipation of negative pore pressure. The corresponding surface settlement at this period is shown in Figure 5.16. During this period, there is further ground movement and the risk of collapse of the retaining system is higher. Hence, this period is critical for the safety of excavation and the effect on nearby structures is larger. The duration of this period depends on the permeability of retained soil and others unfavourable factors such as cracks on the wall which may lead to faster dissipation of the negative excess pore pressure.

After the critical period, the pore pressure stabilises and begins to decrease with time at a very slow rate. As shown in Figure 5.15, from the 3600th second to the 21600th second which is about 5.7 years in prototype scale, the pore pressure drops by only about 5 kPa. This drop is mainly caused by the seepage towards the excavation. The corresponding increases of the surface settlements at LVDT1 and LVDT2 during the same period are about 0.3 mm which is 30 mm in prototype scale. These settlements indicate the existence of further consolidation due to the seepage towards the excavation. Another component of long term settlement is due to the fact that soil sample is not fully consolidated with limited centrifuge working time. The post-excavation settlement is about 1/3 of the total settlements as shown in Figure 5.16.

5.5.4 Centrifuge model and prototype

Centrifuge model test is an effective approach to model an excavation in the laboratory. The progress from initial soil movement to collapse of an excavation system can be simulated in the centrifuge tests. The behaviour of soil and support system can be observed such as the time dependent behaviour due to the built-up of negative excess pore pressure and its subsequent dissipation. With an appropriate model excavation rate based on the principle of the centrifuge model scaling, a nearly undrained condition during the excavation and a drained condition in a long term post-excavation period have been observed. Some typical shear failure surfaces with an angle of about 45^o can be seen in Figures 4.11 and 4.12 which exhibit a pattern of undrained failure of soil. Also, the surface settlement, which decreases more sharply with distance away from the excavation as observed in the model tests, is closer to that seen in excavation site than most the results from numerical analyses. The details on this will be discussed in the next chapter.

The ground surface settlement profile simulated in tests Ts4, Ts5 and Ts6 are plotted on the O¡¯Rourke¡¯s empirical chart as shown in Figure 5.19. The surface settlements from centrifuge test at LVDT1 compared reasonable well with the data from site monitoring. But at LVDT2, the values of centrifuge test results are larger than the data in the empirical chart. The reason of such difference is that a relative shallow excavation depth of 50 mm and a deep embedment of rigid diaphragm wall were used in the study which may not be identical with the situation of the empirical chart. Actually, the influence range of surface settlement is not only related to the excavation depth, but also related to the embedment and rigidity of retaining wall.

In the present study, such results are still not accurate enough for calibration of numerical modelling, but it indicates that accurate simulation is possible. In further research, more work simulating realistic stress history as that of prototype should be carried out.