5.3 Simulation of earthworks and retaining system for a large excavation (Geiser et al., 2002)

5.3.1 Model

In the neighbourhood of Geneva, the construction of a watch production centre has been planned. The project involves the execution of a large excavation in soft and saturated clays. It concerns a 145 x 165 m soil surface with a maximum excavation level of about – 20 m. The designed retaining system is composed of a slurry wall braced at its top. The bracing leans on a 130 m diameter circular reinforced concrete beam supported by piles linked with a buried circular internal slurry wall located at the excavation’s bottom (see Figure 30).

Figure 30: Excavation during the earthworks.

A 3D numerical simulation has been conducted in order to control and optimise all the components interacting in the project. A similar case (large dimensions, similar soil conditions and retaining system) constructed in the 1970’s has been used as a real-scale test in order to precise the soil parameters and the hydro-mechanical behaviour with the help of a back analysis (Fig. 31).

Figure 31: Methodology comparison.

In the actual project, the soils consist mainly of soft and compressible silty clay and silty clay loam, over a thick compact Wurmian moraine. Based on the geotechnical study, six principal layers were schematically defined. It was immediately observed, that an "advanced" constitutive law (here Cap model) was essential to describe correctly the fined-grained soils. The parametric study emphasized also the influence of the compressibility parameter l on the observed displacements.

The change in the pore-water pressure was observed to be the main factor influencing the general behaviour in this project. As the soil permeabilities are low, the hydraulic conditions remain transient during the construction. For a year long excavation, the pore-water pressure looses about 25 to 30 % of its initial value. After reproducing these time-effects on a 2D model, a "pseudo-transient" model was developed for the 3D approach, in order to avoid days-long calculations with a time dependent problem.

The soil is modelled with about 10’000 8-nodes brick elements. The EAS (enhanced assumed strains) finite element technology is selected in order to prevent these elements to lock volumetrically. Structural elements (see Figure 32) can be divided into three sub-categories: slurry walls and mat foundations are modelled with thin shells (Mindlin-Reissner hypothesis), while 2-nodes trusses are used to introduce supporting piles and bracing. Finally, the circular reinforced concrete beam and the external slurry wall stiffener are introduced as Timoshenko beam elements.

An initial state analysis is conducted first in order to start with a non-zero stress field in equilibrium associated with a zero displacement field. After that, twelve construction and excavation steps take place as follows: first, the superficial soil layer is removed (3 meters deep), followed by the construction of the slurry walls and their supporting structure (circular beam, stiffener, prestressed bracing). The actual excavation can then begin, divided in four main zones. In each of the zones about half of the soil is removed, then the mat foundation is placed, and then the other part of the soil is excavated along with the construction of technical galleries (see Figure 33).

Figure 32: Structural elements static system.

Figure 33: Excavation stages.

The external slurry wall is reinforced by counterforts in the execution project. Introducing each counterfort into the global 3D mesh would have been too tedious. An auxiliary analysis has therefore been conducted on a smaller part of the wall in order to estimate the influence of the absence of the counterforts. Results show that settlements are overestimated by 20 to 30 % when modelling the wall with thin shells; the general behaviour of the retaining system is however correctly reproduced.

5.3.2 Results

The vertical displacements after the first excavation step are depicted in Figure 34. The maximal settlement at this time is located near the excavation and reaches 4 cm. Figure 35 illustrates the settlements around the excavation (and also the swelling of the subgrade) at the end of the earthworks. A maximal settlement of about 7 cm is predicted 30 m behind the external slurry wall.

Figure 34: Vertical displacements after the first excavation step

Figure 35: Settlements at the end of the earthworks

A cut parallel to the northern wall crossing the excavation at the middle of the side walls shows the predicted deformation of the system at the end of the earthworks (Figure 36). There is a 5 cm horizontal displacement at the bottom of the external slurry wall.

Figure 36: Horizontal colour maps and deformed mesh

Figure 37 shows the predicted repartition of the pore-water pressures behind the external slurry wall with looses of about 25 to 30 % of the initial hydrostatic pressures. As in situ measurements are now available, they are also represented, highlighting a good correlation with the predicted values.

Figure 37: Pore-water pressure behind the external slurry wall.

An integrated parametric analysis, backed with the experience of the constructors, helps the project engineers optimize the costs of the structure supporting the opening, in the sense that it gives them a qualitative analysis on the effective participation of any structural element to the excavation stability. It can be noticed in particular that the influence of the two supporting slurry walls linking the external and the circular walls on forces and displacements is little, as shown in Figure 38.

Figure 38: Comparison of the slurry walls deformation at the end of the excavation, with or without the supporting walls. Horizontal cut at the galleries’ bottom level.

Another parametric study on the circular buried slurry wall has been conducted in order to check the influence of the concrete quality (see Figure 39).

This excavation is currently under construction, and the first set of in situ measures (inclinometers, pore-pressure cells, optical fibers) have just been analyzed. Of course, modifications have occurred; in particular the excavation steps have been changed. A new calculation incorporating better the reality would be necessary, to allow a rigorous comparison. However, in Figure 40, the predicted deformations of the external slurry wall are compared with the actual observations. A rather good agreement can be found between the two curves, in particular in the order of magnitude of the displacements.

Figure 39: Horizontal membrane force in the circular wall.

Comparison between E = 2e7 (up) and E = 1e7 kN/m² (down).

Figure 40: Horizontal displacement of the wall, prediction vs. field measure at two depths: -3.1 m (up), -19.3 m (down).

The main discrepancies between prediction and measure can be explained in the following way:

- on the one hand at y = -3.1 m, the predicted upper displacement is too small at the middle of the wall. This is due to the fact that in the numerical simulation, the foundation mat was activated before the last excavation phase. However, in the reality, this area is less stiff than initially planned.

- on the other hand, at y = -19.3 m, the predicted displacement is larger than the field measure (bottom of Figure 40). But this part of the excavation has not been finished yet, and additional deformations are expected.

The pore water pressures have also been measured on the field and the “pseudo-transient” computation has shown to be efficient (see Figure 37).

To conclude, this example shows the importance of having complete initial data at hand for a 3D numerical simulation. The use of a real-scale test is also shown to be very useful in order to calibrate the parameters influencing most of the simulations, in particular the pore-water pressure decrease and the soil compressibilities leading to the necessity to choose an adapted constitutive law (Cap model). The time-consuming aspect of 3D numerical simulations can be reduced in conducting different parallel studies (influence of the counterforts, pseudo-transient calculation, no interface elements).

The comparison with in situ measures has validated the a priori predictions.