2.5 Centrifuge Model Studies of Deep Excavation

Study of deep excavation using centrifuge modelling technique has several advantages over field instrumentation techniques. Centrifuge modelling is relatively inexpensive because it can simulate actual size excavation in a small scale model. As the model tests can be conducted repeatedly under controlled and well-instrumented conditions, the behaviour of the retaining wall and associated ground movements including pre-failure situation can be examined thoroughly. Some early works in centrifuge laboratories involved excavation stability studies (Leung et al., 1991). Excellent examples on surface and underground excavation stability studies were reported by Schofield (1980) in his Rankine Lecture. Bolton and Powrie (1987) carried out centrifuge tests to explore both collapse mechanisms and soil and wall deformation patterns of diaphragm wall constructed in clay.

2.5.1 Methods of simulating excavation

Three methods have been used to simulate excavation in centrifuge model tests. In the first and obviously elementary method, the model excavation is done at 1g and the model is then subjected to increasing centrifugal acceleration until failure is achieved (Lyndon and Schofield, 1970).

The second method is to simulate the progress of excavation by releasing heavy liquid of equal unit weight of soil from the model during centrifuge flight. This technique was first performed by Lade et al. in 1981, subsequently by Kusakabe in 1982 and Philips et al. (1984), Bolton and Powrie, (1987, 1988), Ooi (1994), and Lee (1995). Although the second technique is useful in reproducing the excavation process more accurately, problems still remain as the lateral pressure coefficient of soil and heavy liquid is different.

In the third method, the excavation is implemented by means of an in-flight centrifuge excavator developed by Kimura et al. (1993). This method provides a better simulation of the excavation process.

Due to the high-g acceleration environment and limited space on the centrifuge platform, the design of centrifuge model needs special techniques. A comparison of the design of centrifuge models for the study on excavation by various researchers is given in Table 2.2.

2.5.2 Soil condition and retaining structures

Philips et al. (1984) conducted centrifuge model tests to investigate the overall stability of unsupported excavations in soft clay as well as the magnitude and extent of ground deformations. Lyndon and Pearson (1984) carried out centrifuge model tests of a rigid retaining wall in cohesionless material. The behaviour of both un-propped and propped retaining walls embedded in clay has been studied by Bolton and Powrie (1987, 1988) as shown in Figure 2.7. Unsupported excavation and excavation supported by wall with or without ties, were carried out.

Most of the research work on excavation dealt with stiff soil. Only a few studies have been carried out on excavation in extremely soft clay (Kimura et al., 1993). Generally, there were more centrifuge tests on un-propped retaining walls than on propped ones. In particular, there were very few studies on multi-level propped wall. In addition, far more attention has been paid to the failure mechanism of excavation than the pre-failure behaviours which is more useful under working condition in the field.

Bolton and Powrie (1987) and Kimura et al. (1993) have made significant advances in using centrifuge modelling to study the behaviour of unpropped and propped excavations in clay. Their results are reviewed in detail in the next section.

2.5.3 Studies by Bolton and Powrie

The collapse of both un-propped and propped retaining walls embedded in over-consolidated clay were studied by Bolton and Powrie (1987) using centrifuge models. In their experiments, zinc chloride having the same unit weight as the model clay was used to replace the clay in the excavation component. Their studies revealed the following observations and findings:

1. Unpropped walls are unlikely to present an economical method of retaining large heights of clay as the entire wall would collapse within a short time of excavation and large depth of wall penetration is needed for stability.
2.  
3. In stiff clay, unpropped wall of shallow penetration fails in a flooded tension crack mode (Figure 2.7). Tension cracks occur between the retained soil and the wall. The water flows into the cracks and causes the wall to fail within a short duration. The flow rate of water into tension cracks determines the rate of wall movement. The passive resistance of the clay will be insufficient to hold a full-height flooded cracks as the wall rotate about a point above the toe.
4.  
5. If the soil remains in contact with the wall, short term equilibrium of the wall is maintained by the development of large pore water suction in the soil immediately adjacent to the wall. Then the failure mechanism is governed by the rate at which the soil can shear, change volume or slide along rupture surfaces. For un-propped walls of deep penetration of about 15 m to 20 m, the failure mode is rotation failure about a point near the toe.

d) The rupture surfaces may develop along a plane of maximum stress obliquity, a zero-extension line, or some intermediate direction. The boundaries of significant soil movement and ruptures however lie in a 45^o triangle drawn from the base of the wall.

5. The critical penetration and pivot position match the values obtained by an analysis using undrained soil strength parameters. For walls propped at the crest, the kinematics restraint imposed by the prop results in a conservative simple stress field calculation. The kinematics restraint produces a rupture surface on the active side which is much steeper than those observed in unpropped walls.
6.  
1. The peak value of the mobilised soil friction observed in soil test is consistent with those generated by a back analysis of the observed failure mechanism. This value also gives a close match for the bending moments and propping forces measured in the test.

The behaviour of diaphragm walls in clay prior to collapse was also studied in detail using centrifuge model tests by Bolton and Powrie (1988). Based on the measurements of soil displacement vectors during excavation, Kinematically admissible strain fields for rigid retaining wall were derived which idealise soil behaviour in terms of uniformly deforming triangles. Assumed mobilised shear strain was measured and displacements were calculated using an appropriate idealised strain field. From their study, it has been demonstrated that centrifuge model tests can form the basis of research into soil-structure interaction. However, experimental techniques do need to be refined. Methods of grouting and excavation in-flight will have to be developed if more difficult problems are to be researched.

2.5.4 Studies by Kimura et al.

Kimura et al. (1993) developed an in-flight centrifuge excavator blade as shown in Figure 2.8 which gave a better simulation of an actual excavation process. This excavator also eliminates the problem of incompatability of the K_o condition where zinc chloride is used. Unsupported excavation and excavation with a supporting wall with or without ties, were carried out for two types of soft clay, namely normally consolidated and lightly over-consolidated clays. Earth pressures on sheet pile wall and deformations of the wall and clay were measured during the experiment. The results of their study are as follows:

(a) It is confirmed that support by sheet pile wall, tieing of the wall and pre-loading are effective in increasing the stability of excavation in soft clay.

(b) The mechanism of failure in excavation with sheet pile wall without ties is different for clay with uniform strength and for soil with strength increasing with depth. For the clay with uniform strength, the wall shows translation movement with well-defined linear slip surface both in the active and passive side. For clay with strength increasing with depth, the wall tilts around the base with linear slip surface only observed on the active side.

(c) In excavation with sheet pile wall without ties, the embedment depth of the wall does not effect the stability in clay with uniform strength. On other hand, the stability of the excavation increases significantly with depth of embedment for clay with strength increasing with depth.

(d) The characteristics of clay in the active state have a dominant effect on the deformations and strains of sheet pile wall. The wall in clay with higher initial stiffness in triaxial conditions show smaller strains until they reach failure.

(e) The importance of considering strength anisotropy in analysing the stability of an excavation is confirmed. This is particularly important for normally consolidated clay as it has an initial low value of earth pressure at rest, K_o.