3.6 Model Preparation

Based on the previous experience in preliminary centrifuge model tests on deep excavation conducted at NUS by Ooi (1994) and Lee (1995) and in trial tests conducted in this study, a standard model preparation procedure was developed and is now described.

3.6.1 Soil preparation

In Singapore, the hard strata underlying the Kallang formation are typically dense clayey silt, siltstone or sandstone which belong to the Jurong formation or the fluvial clay of the Old Alluvium formation. In order to simulate this situation, disturbed residual soil samples were collected from the site and compressed in layers each of 10-20 mm thick using hydraulic pressure to a total thickness of 170 mm.

As mentioned earlier, the soft clay stratum had been simplified by ignoring the intermediate stiff clay layer of the Kallang formation shown in Figure 2.6. The marine clay acquired for this study was from an excavation site near the City Hall area of the Singapore city centre. The clay is a kaolinite-rich, bluish-grey soft clay with shell fragments. The properties of the marine clay are summarised in Table 3.3. In the laboratory, the clay was remoulded and mixed with additional amount of water to a water content around 120% which is 1.5 times the liquid limit of the soil. The slurry was then poured into the model container above the compressed stiff clay layer. Before filling up the model container with soils, two piece of fabric drain 1.5 mm thick with filter paper was placed along front and back walls which were used to aid the drainage of water during consolidation. Also, a fibre mat was placed on top of the clay slurry to improve the drainage.

The soil sample was placed under a hydraulic pressure and gradually compressed one-dimensionally in stages to a vertical effective stress of 55 kN/m², as illustrated in Figure 3.5. The soil settlement was monitored until primary consolidation had been completed. Then the model was removed from the press and mounted onto the centrifuge where it was further subjected to consolidation with surcharge of 55 kN/m² under 100g for a few hours for the first series of tests and 24 hours for the second series of tests. Immediately after consolidation and test, the soil strength at various depths was measured using the laboratory vane shear device, as shown in Figure 3.6. The undrained strength of soils consolidated using this approach is shown in Figure 3.7. The soils for the first series of tests exhibited a nearly uniform strength profile with slightly softer strength in the middle part of the soil layer. In the second test series, the effect of the two consolidation processes is to produce a layer of marine clay that is nearly normally consolidated at depth. A slightly overly consolidated condition near the ground surface was obtained by retaining a sand surcharge of 20 kN/m² during test as compared to a surcharge of 55 kN/m² used during the high-g consolidation. Figure 3.7 shows the shear strength of the soil model of Test Ts4 is 15.2 kPa at 20 mm, to 23 kPa at 100 mm depth and increasing to 35 kPa at 200 mm depth. The average bulk unit weights of the upper soft clay and lower stiff clay were measured to be about 17 kN/m³ and 20 kN/m³, respectively. The average water content of the upper soft clay was determined to be 65-75 % before the centrifuge test.

3.6.2 Model diaphragm wall and slabs

To simulate a prototype diaphragm wall of 1m thick and 25 m high at 1/100 th scale, a 8 mm thick and 250 mm high model aluminium alloy with elastic modulus E= 70 x 10⁶ kN/m² is used. The thickness of the model diaphragm wall is chosen based on the scaling relations given in Table 3.2. The flexural rigidity of the wall, EI, is 2.99 kNm²/m which is 1/N³ the bending stiffness of the prototype’s one meter thick concrete wall where N is the scaling factor. As shown in Figure 3.8(a), six sets of strain gauges were attached to each side of the wall to measure the bending moment of the wall.

The model floor slab is made of aluminium alloy as shown in Figure 3.8(b). A 3.5 mm thick aluminium alloy plate in model scale is equivalent to a 0.5 m thick concrete slab in prototype. Two sets of strain gauges were attached to each slab to measure its axial force and bending moment.

On completion of high-g consolidation, the model container was removed from the centrifuge. The front window and the back wall of the model container were removed and a clay block of 150 mm or 50 mm deep and 150 mm wide was cut and removed to form the excavation area. A thin wire was used to cut the soil to reduce soil disturbance and the wall was then inserted into this slot cut. The installation was carried out quickly and a rigid form work was placed to prevent swelling of clay.

The model wall and the slabs assembly are shown in Figure 3.9. The upper slab was installed in place before the excavation to simulate the cast slab near the ground level in a typical top-down construction. Before excavation, the lower slab was placed in position in the zinc chloride solution without contact with the wall. When the excavation reached the middle level near the lower slab, the hydraulic system was then activated to push the lower slab to support the wall, see Figure 3.10.

3.6.3 Zinc chloride solution and latex bag

In the centrifuge model tests, the in-flight excavation sequence was simulated by the drainage of a heavy liquid, zinc chloride solution. After the pre-consolidation and before the re-consolidation, the excavated area was cut and replaced by zinc chloride solution with the same density as the removed clay. The solution was contained in a flexible latex bag placed within the trench as shown in Figure 3.11. The bag was fabricated using cotton cloth on which latex was applied to make it water proof. A drainage pipe was connected to the base of the bag by a connector and a solenoid valve at the other end outside the container which would be used to activate and control the drainage of zinc chloride during the in-flight excavation.

The model was assembled after the rubber bag was put in. Zinc chloride solution was poured into the bag.

The height of zinc chloride solution in the rubber bag was 150 mm for the test series 1 to simulate a 15 m excavation. Under 100g, the maximum pressure in the bag was 255 kPa. The bag used was designed to be impermeable and strong enough to prevent leakage of heavy solution under such high pressure, but also be as flexible as possible to ensure full contact with the wall.

3.6.4 In-flight excavation and slab push-in system

In the present study, the upper slab was placed at 1g and the lower slab was installed in-flight within the excavation area, which was initially filled with zinc chloride solution. The details of the technique used in the test is shown in Figure 3.11. The upper strut was installed in place before the test. Before the excavation, the diaphragm wall was maintained in equilibrium under the earth pressure of the retained soil and the hydrostatic pressure of the zinc chloride solution. The process of excavation was simulated by releasing the zinc chloride solution. In the design, some special coating was applied to prevent corrosion of the slab, strain gauges and the connecting wires. Rubber seals were used to prevent leakage of the zinc chloride solution. As the excavation depth reached the level of the lower slab, the lower slab was pushed in to support the diaphragm wall by using the hydraulic push-in system. The travel of the slab was controlled by a stopper to prevent over-pushing towards the retaining soil.

3.6.5 Instrumentation

Surface settlements of retained soil were monitored by linear voltage displacement transducers (LVDT). The LVDTs were positioned along the soil surface as shown in Figure 3.4 at 50 mm, 150 mm, 250 mm and 350 mm from the wall. Light aluminium plate 25 mm in diameter and 1 mm in thickness were fixed at the end of the probe to act as footing thus reducing settlement due to the self-weight of the probe.

Three to six numbers of pore pressure transducers were embedded in the soft clay and a typical arrangement is shown in Figure 3.4. Two other pore pressure transducers were placed in the zinc chloride solution to measure the changes in pressure during the release of the zinc chloride solution from the rubber bag. From the readings obtained, the height of the zinc chloride solution in the rubber bag could be calculated and so the depth and rate of excavation could be monitored.

Druck miniature pressure transducers (Type PDCR 81) were used for the purpose mentioned above. This type of transducer consists of a single crystal silicon diaphragm with a fully active strain gauge bridge diffused onto the surface. It has a porous filter stone placed in front of its diaphragm such that only water is in contact with the diaphragm. The transducer has a rugged construction and possesses a high overload capacity. It has excellent linearity, negligible hysteresis and fast response in a high-g environment. The calibration constants obtained is very close to the constants provided by the manufacturer. It is small enough to be easily placed in soil. Before being placed in the soil, the porous filter stone was saturated with de-aired water by means of a vacuum pump. The transducer was inserted into a pre-drilled hole filled with soil slurry and then the hole was sealed with soil.

As shown in Figures 3.4 and 3.8, six sets of strain gauges attached on the model diaphragm wall were used to measure the bending moment of the wall during the excavation. Each strain gauge bridge consisted of two active gauges on opposite sides of the wall. With this arrangement, the bending strain can be obtained from the output voltage of the circuits. The bending stress and subsequently the bending moment can be calculated by considering the section properties such as elastic modulus and moment of inertia of the wall.

3.6.6 Surface grid on soil sample

Traditionally, the soil is labelled with markers on the front surface and photographs of the surface are taken during the test. Displacements of soil were then obtained from the photographs or by the image processing technique (Allersma, 1991). Metal slots, pine tips, Japanese noodles as well as rubber film with grids can be used to label points on the front surface of clayey soil. At high-g level, the self-weight effect of the slots and reinforcement of the pin can be quite significant. Through several trial tests, an effective approach to mark the clay surface with grids was developed in the present study. In this technique, a stiff paper grille was placed on top of the clay surface. White colour was then sprayed onto the clay surface. As shown in Figure 3.12, the entire grids can be formed by spraying horizontal and vertical lines through the grille. Using this method, a well-defined grid pattern is obtained. Before the perspex wall was placed back onto the model container, silicon oil was applied on the perspex wall to reduce the friction between the soil and the wall.