Finite element study of deep excavation construction processes

Available online at www.sciencedirect.com

HOSTED BY

ScienceDirect Soils and Foundations 57 (2017) 965–979 www.elsevier.com/locate/sandf

Finite element study of deep excavation construction processes q Yuepeng Dong a,1,⇑, Harvey J. Burd b, Guy T. Houlsby b b

a Singapore-MIT Alliance for Research and Technology, Singapore Department of Engineering Science, University of Oxford, OX1 3PJ, UK

Received 23 August 2016; received in revised form 24 May 2017; accepted 3 August 2017 Available online 20 November 2017

Abstract Excavations for deep basements in urban areas present complex design questions, e.g. related to the selection of retaining systems and the specification of construction processes. Consideration must be given to the design of structural systems to support the excavation and the mitigation of any impact that the construction might have on any nearby infrastructure. The finite element analysis is routinely used (in design and research) for the analysis of this type of complex soil-structure interaction problem. Care is needed to ensure that representations of the construction processes, soil and structural behaviour are incorporated in the finite element model at an appropriate level of detail. This paper addresses the implementation of construction procedures in a finite element analysis of a deep excavation in Shanghai. Initially, an analysis is developed to model the construction processes actually employed in the project. The results of this model compare favourably with data measured during construction. The model is then developed to investigate the influence of certain aspects of the construction processes on the computed results. The results indicate the following: (i) the construction sequence for the floor slabs does not have a significant influence on the computed deformations in the retaining walls or the nearby ground at the end of construction; (ii) earth beams (used as temporary supports) are effective in reducing the computed wall and ground movements, and (iii) neglecting the presence of openings in the floor slabs may lead to unconservative calculations of the retaining wall and ground movements. Ó 2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BYNC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Deep excavations; Finite element analysis; Openings; Construction sequence; Floor slabs; Earth berms

1. Introduction Deep excavations are routinely employed in urban development projects. The performance of an excavation is affected by various factors, including the local ground conditions, the structural support systems, and the construction processes. The finite element (FE) analysis has been widely used to model deep excavation construction processes (e.g., Dong et al., 2016; Whittle et al., 1993; Zdravkovic et al., 2005). However, the development of

Peer review under responsibility of The Japanese Geotechnical Society. ⇑ Corresponding author. E-mail addresses: [email protected], [email protected] (Y. Dong). 1 Former DPhil Student at University of Oxford, UK.

FE procedures that provide reliable data for design purposes is challenging, particularly when the excavation is of irregular geometry or where innovative construction processes are proposed. Recent studies (Dong, 2014; Dong et al., 2016) demonstrate the importance of considering a range of issues in the analysis of excavations, including small-strain soil nonlinearity, construction joints in the retaining walls, post-cure thermal behaviour of concrete floor slabs and the detailed nature of the initial ground stresses. The current paper extends this previous work to a case study of a complex deep excavation, the North Square Centre, which forms part of the Shanghai South Railway Station, completed in 2005. Initially, an FE analysis is developed to model the construction processes that were actually employed in this project. This model is then

https://doi.org/10.1016/j.sandf.2017.08.024 0038-0806/Ó 2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

966

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

modified to investigate the influence of aspects of the construction procedure on the computed results. In projects involving top-down construction methods (such as the North Square Centre), the floor slabs are typically cast, sequentially, in independent sections. However, the performance of the excavation may depend on details of the construction sequence. Questions exist about the appropriate level of detail for modelling the construction operations to achieve results that are sufficiently reliable for design purposes. In this paper, these questions are investigated. Earth berms are often used within deep excavations to provide temporary lateral support to the retaining wall, maintaining stability during construction (Clough, 1977; Georgiadis and Anagnostopoulos, 1998; Gourvenec and Powrie, 2000; Daly and Powrie, 2001; Powrie and Daly, 2002; Smethurst and Powrie, 2008). However, earth berms inevitably add complexity to the construction. It is not clear whether berms need to be included in an FE model of the construction process. The analysis described here is focussed on the effectiveness of the berms used in the North Square Centre project in reducing deformation in the retaining wall and the neighbouring ground. In top-down construction, temporary openings in the floor slabs are used to provide access. Because these openings reduce the effectiveness of the propping action of the slabs, an increase in wall deflections and local ground movements may occur. The size and distribution of openings needs to be carefully considered when top-down construction is employed. In FE models of top-down excavations, the openings are typically included in the analysis, approximately, by artificially reducing the stiffness of the lateral support system (e.g., Simpson, 1992; St. John et al., 1993). This simplification may not capture the excavation behaviour well, especially when the geometry of the retaining system is complex. We describe an FE study on the influence of temporary openings on the deformations in the retaining wall and the surrounding soil.

2. The north square centre 2.1. General description The Shanghai South Railway Station was constructed to increase the capacity of existing passenger terminals, forming the basis of a major urban development project. The Station integrates subways, a light rail system, aboveground public transport, an elevated freeway and a passenger interchange. The project included an underground shopping centre, the North Square Centre (NSC). The excavation for the NSC, to the north of the Main Station (Fig. 1), forms the basis of the finite element modelling presented here. The NSC required a deep excavation, constructed using a top-down method. Site investigation data were reported by Xu (2007). Extensive field monitoring data were collected during construction (Xu, 2007; Hou et al., 2009). The excavation is 12.5 m deep, and covers an area of about 40,000 m2. The structure has two basement levels and is supported on a piled raft foundation. The soil retaining system uses a diaphragm wall, propped by the floor slabs. The geometry of the excavation is irregular: roughly 400 m long and 100 m wide. Construction of the NSC began in 2003 and was completed in 2005. The NSC excavation is close to several current and planned underground transportation links. To the east is the new LRT (Light Rail Transit) Line 1 (within 3 m of the diaphragm wall), and the Interchange Station for the new Metro Lines R1 and LRT Line L1. The Main Station and the relocated Metro Line R1 (within 2 m of the diaphragm wall) are to the south. To the north is the operational Metro Line R1 (with a minimum distance of 3 m to the diaphragm wall). The excavations for these neighbouring developments were completed before the start of the NSC excavation. The NSC excavation has several distinctive features: (i) it is close to adjacent infrastructure, so care was

Metro line R1 (in operation)

N

A

Bottom-up area IT10 IT9 3m I35 I36

I40

I28 I30 I29

IT 8 I34

I41

I37 I38

I39

C #6

#7

#9

#5

I11

I10 #1

I9

#2

I8

I12

Bottom slab of the Main Station

I7

I3

I: Inclinometers in wall IT: Inclinometers in soil #1 Number of excavation block

B

I5

I19

B

I18 I13 I14

I2 I4

I21 I20

#8 #10

I1

2.0m

3m

I22

#4

#3 I45

I23

North Square of Shanghai South Railway Station

Bored pile wall

I44

LRT line L1 I24

I32 I33

I43

I27I26 I25

I31

B

I42

Diaphragm wall I

I6

A

I 15

I16 I17

Interchange Station

Relocation of Mero line R1

Main Station of Shanghai South Railway Station

Fig. 1. The North Square Centre excavation plan and instrumentation (Xu, 2007). (The measurement locations used to compare with numerical analyses are indicated in red).

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

needed to ensure that local ground movements were minimised, (ii) the retaining structures used for the project are of irregular geometry, (iii) the floor slabs had large temporary openings, and (iv) complex construction sequences were employed. The complex nature of this excavation, together with the availability of a substantial amount of ground and wall movement data collected during and after construction, means that it provides a valuable case history to inform the development of finite element methods for the design and analysis of deep basements. 2.2. The NSC retaining system The soil retaining system for the NSC excavation forms part of the permanent structure. The main structure, shown in Fig. 2, consists of a diaphragm wall, two floor slabs providing horizontal support, steel lattice columns and reinforced concrete piles. The bottom slab (1 m thick) was cast in situ after the completion of the excavation. Soil cement columns and compaction grouting were used to improve the soil properties and to restrain the lateral displacement of the diaphragm wall. The Interchange Station shares the diaphragm wall with the NSC on the west side and employs a soil mixing wall on the other sides. The Interchange Station retaining struc-

967

ture, shown in Fig. 2, has three levels of temporary steel struts at
2.83 m,
6.43 m and
9.83 m respectively. The excavation is 9.1 m deep, and was constructed using a bottom-up technique. The NSC diaphragm wall is 0.8 m thick and up to 28 m deep (the diaphragm wall next to the Main Station is 3 m shorter (25 m) due to the slope in the ground level floor slab). The diaphragm wall was constructed in separate panel sections. Bored piles formed a temporary partition wall between the top-down and bottom-up excavation areas on the north side (Fig. 1). The bottom-up areas were excavated after the relocation of the Metro Line R1, after the completion of the NSC excavation. The vertical support system consists of bored piles (45 m embedded length, 700 mm diameter) and steel lattice columns (450 mm  450 mm), which were cast into the piles. The piles were grouted at the toe to improve the bearing capacity and reduce settlements. The floor slabs, Fig. 3, consist of a concrete floor slab (0.2 m thick) supported by a grid of concrete beams (500 mm  900 mm in section). The ground level slab was used as the working space during the excavation. A total of 22 large temporary openings in each floor slab were included to provide access. The largest opening is 700 m2; the others are between 400 m2 and 500 m2.

Fig. 2. Sectional view of the excavation (Xu, 2007).

Fig. 3. Plan view of the top floor slab (Xu, 2007).

968

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

2.3. Construction sequence The excavation was constructed using a top-down method, as summarised in Table 1. The excavation was divided into 10 zones, see Fig. 1, and constructed in sequence from zone #1 to zone #10 at each excavation level. To restrain the lateral wall deflections, temporary earth berms were employed on the north side of the excavation, Fig. 4. 2.4. Instrumentation The excavation was monitored, to assess its performance and ensure its safety. The measurements (Xu, 2007) are summarised below: (1) Lateral displacement of the diaphragm wall (from I-1 to I-45, see Fig. 1) using inclinometers; (2) Ground settlement along BC outside the excavation using level instruments, and soil lateral displacement at the positions IT7 to IT10 using inclinometers. These measurements were made to assess ground movements in the vicinity of Metro Line R1; (3) Vertical displacement of the diaphragm wall at positions I-1 to I-45 using level instruments; (4) Vertical displacement of piles and columns using level instruments.

2.5. Soil properties The site investigation (Xu, 2007) indicates that the site consists of thick layers of soft quaternary alluvial and marine deposits, typical of Shanghai Clay (e.g., see Dong et al., 2016). The ground surface level is approximately at elevation +4.430 m (using Shanghai Wusong elevation standard) and the ground water table is between 0.5 m and 1.0 m below the ground level. The geological profile and soil properties from Xu (2007) are shown in Fig. 5. The soil consists of several distinct layers. The natural water content of the clay and silty clay layers is close to, or higher than, the liquid limit, suggesting that the soil is normally consolidated or lightly overconsolidated. The undrained shear strength, su, from field vane tests, is significantly higher than values normally associated with clay at the liquid limit, suggesting that the clay is sensitive. The soil data provided in the site investigation are limited in scope and not sufficient (on their own) to calibrate an advanced constitutive model for the soil. However, a comprehensive study on the engineering properties of Shanghai Clay is described in Dong (2014) and Dong et al. (2016), including the undrained shear strength profile (depth up to 70 m) measured using shear box and field vane tests, the small-strain shear modulus profile (depth up to 100 m) determined from shear wave velocity tests, and

Table 1 Construction sequence (Xu, 2007). Stage no.

Stage description

0 1

Install the diaphragm wall and piles Excavate to
3.750 m inside and outside the excavation Excavate to
7.500 m on the south side close to the Main Station and to
5.000 m on the north side away from the Main Station with earth berms (10.0 m wide) Install the ground level floor slab (B0F) at
3.000 m Excavate to
13.000 m on the south side close to the Main Station and to
10.000 m on the north side away from the Main Station with earth berms (10.0 m wide) Install the lower level floor slab (B1F) at
8.450 m Excavate to
14.700 m on the south side close to the Main Station and to
13.000 m on the north side away from the Main Station with earth berms (10.0 m wide) Cast the bottom slab on the side close to the Main Station Remove the remaining earth berms and excavate to
14.700 m on the north side away from the Main Station Cast the remaining bottom slabs

2

3

4

Fig. 4. Earth berms at the north side of the excavation (Xu, 2007).

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

969

Fig. 5. Geotechnical profile and soil properties from site investigation (Xu, 2007).

the variation of tangent shear modulus with shear strain. These data are combined with the site investigation information (Fig. 5) to calibrate the soil constitutive model employed in the current analysis. 3. Finite element model 3.1. Geometry and mesh Any analysis of a complex excavation necessarily involves compromises, with judgement necessary about the level of detail for aspects of both the structure and the construction process. Considering the irregular geometry of the NSC excavation and the complex construction sequence, a 3D model is necessary. The finite element model (Figs. 6–8), was developed in Abaqus v6.11; it includes the retaining structure (i.e. retaining wall, piles, columns, beams, and floor slabs), detailed geometry of the floor slabs, the cast-in-situ basement raft, the zoned construction sequence, earth berms, the excavation for the adjacent

Interchange Station, and the bottom-up excavation area. The four vertical sides of the mesh are assigned roller boundary conditions, and the base of the mesh is fixed. The soil is modelled with linear displacement, 8-noded, hexahedral elements with reduced integration (C3D8R). Linear (rather than higher order) elements were adopted to limit the required computing resources. As discussed in Dong et al. (2016), linear elements typically exhibit an over-stiff response for failure analyses, and higher-order elements should be considered for more accurate analysis when failure issues are important. In this application, however, the performance of the excavation at relatively small soil strain is of primary importance, and linear elements were considered adequate. The mesh for the retaining walls includes the diaphragm wall for the main excavation, the soil mixing wall for the Interchange Station, and the bored pile wall for the partition of the bottom-up area (Fig. 7). The retaining walls were modelled using C3D8R elements with two elements within the wall thickness, providing balance between accu-

Fig. 6. Mesh and boundary conditions of the entire model.

970

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

Fig. 7. Mesh of the retaining wall.

Fig. 8. Mesh of the supporting system.

racy and computational efficiency in this complex case study. Parametric analyses (details given in Appendix B) based on an idealised deep excavation (model details provided in Dong (2014)) showed that computed wall deflections and ground settlements were almost identical when 1, 2, 4 and 6 elements were used within the wall thickness. The supporting system (Fig. 8) consists of two levels of concrete floor slabs and beams. The ground level slab is sloped (reflecting the changes in level across the site). Temporary openings are incorporated in the slabs, which are supported by piles and columns. Temporary struts are used to support the retaining structures in the bottom-up construction area and the Interchange Station. In the finite element mesh, the slabs were modelled with 4-noded quadrilateral shell elements with reduced integration (S4R). The horizontal beams, vertical columns, piles and temporary struts were modelled using 2-noded linear beam elements (B31). Tie constraints were applied at the connections of the various elements. The piles were embedded within the mesh used to model the ground beneath the excavation. Slip at the interface between the soil and the structural elements (diaphragm wall and piles) was not specifically accounted for in the analysis; full fixity is assumed between the structural elements and the neighbouring soil. Note, however, that the behaviour at the soil/structure interfaces may have influenced the results of computations for the performance of deep excavations (e.g., Dong (2014)). Allowing slip on the soil/wall interface increases the wall deflection and ground settlement outside the excavation; the amount of this additional movement is affected by the

soil/wall interface properties. Incorporating a model for the soil/structure interface in this complex case study is not practical: (i) it may cause numerical problems and significantly increase the computational time since contact is a complex nonlinear problem, and (ii) appropriate soil/structure interface properties are difficult to determine. Neglecting the details of the soil/structure interface in the analysis may cause a discrepancy between the numerical analysis and the field measurements. However, since this case study was focused on the influence of the construction process, neglecting the soil/structure interface did not affect the overall conclusions. The finite element model has 188,593 elements and 214,360 nodes. All analyses were conducted assuming undrained soil conditions (using a total stress approach), on the basis that the permeability of clay was low (typically of the order of 10
9 m/s) and the construction period was relatively short (around 1 year). The justification for the assumption of undrained soil behaviour in this context is discussed in Dong et al. (2016), considering that drainage effects were minimal and little excess pore water dissipation was likely to occur during the construction process of this relatively large scale excavation. The advantage of the total stress analysis is its numerical robustness (e.g., in comparison with effective stress modelling). A further advantage is that it allows the use of undrained shear strength (measured during the site investigation) to be used as a model parameter. Total stress analysis, however, cannot be used to investigate the influence of any pore pressure dissipation that might occur during construction. This was not a major

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

difficulty in this case study since computing the displacements was the primary focus of the work. The analysis steps follow closely the construction sequence specified in Table 1. However, the retaining wall installation was not modelled in detail; instead, the wall was assumed to be ‘wished-in-place’. Furthermore, dewatering processes were not included in the analysis, on the basis that a total stress soil model was used for the soil. Dewatering is normally conducted inside excavations to provide a dry working environment. If any unit weight change in the soil is neglected assuming that the soil remains saturated, pore pressure reduction at constant vertical stresses requires the effective stress to increase, and the principal mechanism of dewatering is consolidation (Powrie and Preene, 1994). Inward wall displacements and ground settlements outside the excavation have been observed during the dewatering process prior to excavation activities (Xu, 2007; Zheng et al., 2014); this is likely to be induced by decreased lateral total stress on the wall inside the excavation during the consolidation process. However, this movement is typically small compared to the effect of the bulk excavation, especially the final movement when the excavation is completed. Neglecting the wall installation and dewatering processes in the analyses may therefore have certain effects on the accuracy of the computed results, although the influence of these effects is likely to be minimal. 3.2. Soil constitutive model The soil is modelled using a multi-surface kinematic hardening model (Houlsby, 1999) developed to represent the non-linear behaviour of soil at small strains. The model, formulated in terms of total stresses, has been implemented in Abaqus via a UMAT subroutine (Dong, 2014). An application of this model in 3D finite element analysis of shallow tunnelling processes is described in Burd et al. (2000), and for a deep excavation in Shanghai Clay in Dong et al. (2016). The model uses an arbitrary number of kinematic hardening yield surfaces of the same shape as a fixed outer von

971

Mises failure surface, Fig. 9a. As a stress point moves and encounters a yield surface, the model causes the tangent stiffness to reduce (see Fig. 9b). The model is able to capture with reasonable accuracy the reduction of tangent stiffness with strain magnitude, the effects of immediate past stress history on tangent stiffness, and the hysteresis effects on cycling. The model is specified by a small strain shear modulus, G0, bulk modulus, K, undrained shear strength, su, and a set of non-dimensional parameters ci and gi (i = 1, n) that specify the size and work hardening characteristics of each inner surface. The size of each inner surface is ci su and the tangent shear modulus, Gt, when the ith surface is active, is giG0. Nine inner yield surfaces (i.e. n = 9) are used in the analyses, chosen as a balance between accuracy and computational efficiency. The model is calibrated by choosing values for ci and gi (i = 1, 9) to provide an acceptable fit to a stiffness degradation curve (which is defined by the relationship between Gt/ G0 and Irc for primary loading, where Gt is the tangent shear modulus at shear strain c, and Ir = G0/su). Following Dong et al. (2016), the assumed stiffness degradation curve, which is based on published data for Shanghai Clay, is Gt 1 ¼ G0 ð1 þ I r cÞ2

ð1Þ

Procedures to determine the parameters ci and gi to fit the stiffness degradation curve in Eq. (1) are described in Dong et al. (2016). The resulting parameters, which are identical to those used by Dong et al. (2016), are given in Table 2. The normalised stiffness degradation curve is assumed to be invariant with depth. Arbitrary functional forms can be employed for the variation of the undrained shear strength (su) and shear modulus (G0) with depth. In the current analysis, the undrained shear strength is assumed to vary linearly with depth, z, as su = 15 + 1.5z, where z is in units of metres and su is in units of kPa. The su profile in Fig. 5 indicates a shallow overconsolidated crust and the undrained shear strength su at shallow depth (up to 4 m below ground surface) may be slightly higher than the

Fig. 9. Multiple yield surface kinematic hardening model.

972

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

Table 2 Derived properties for the 9 inner yield surfaces. su = 15 + 1.5z, Ir = G0/su = 1000, cs = 18.5 kN/m3, Kt0 = 0.843 n gn cn

1 0.9 0.005

2 0.7 0.0952

3 0.5 0.2264

4 0.3 0.3537

5 0.15 0.5358

6 0.075 0.6769

7 0.03 0.7678

8 0.0075 0.8708

9 0.00058 0.942

Table 3 Input parameters for the retaining wall. Components

Material properties

The The The The The The

Ev = 30 GPa, m = 0.2, mhv = mvh = 0, b = 0.1, cc = 25 kN/m3 Ev = 3 GPa, m = 0.2, mhv = mvh = 0, b = 0.001, cc = 23 kN/m3 Ev = 3 GPa, m = 0.2, mhv = mvh = 0, b = 0.0001, cc = 23 kN/m3 E = 30 GPa, m = 0.2, a = 10
5/K, DT =
25 K, cc = 25 kN/m3 E = 210 GPa, m = 0.3, cm = 78 kN/m3 E = 30 GPa, m = 0.3, cc = 25 kN/m3

diaphragm wall soil mixing wall bore pile wall beams and floor slabs steel struts piles and columns

implied by the linear equation used in the analysis. This detail was not considered in the analysis, because the very limited data from the site investigation provided a simplified ground profile. A refined ground profile may provide improved numerical predictions, but this shallow layer is likely to influence the analyses only at a shallow depth. In the absence of detailed data on the shear modulus, the rigidity index is taken as a constant, Ir = 1000. This value was determined by Dong et al. (2016) for a nearby site in Shanghai and is adopted here for convenience. Following Dong et al. (2016), the soil unit weight is taken as cs = 18.5 kN/m3, on the basis of the average bulk unit weight of the soil from the site investigation (Fig. 5). This gives an initial linear vertical total stress distribution in the model, rv = 18.5 z kPa; the horizontal total stress is computed using a lateral earth pressure coefficient (Kt0, defined here in terms of the total stresses) as rh = Kt0rv. The lateral earth pressure coefficient is assumed to be Kt0 = 0.843, determined on the basis of the conventional Jaky formula for K0 with an angle of friction of 20° estimated from Fig. 5 to determine the effective stress ratio, followed by an adjustment to account for the influence of the pore pressures on the total stresses (Dong, 2014). 3.3. Structural components The retaining walls (diaphragm wall, soil mixing wall, and bored pile wall) are modelled using a crossanisotropic linear elastic model. This provides a means of modelling the influence of the construction joints in the retaining wall (Zdravkovic et al., 2005). This approach has been explored in previous parametric analyses and case studies (Dong, 2014). The assumed elastic properties of the concrete for the diaphragm wall are Ev = 30 GPa, where Ev is the Young’s modulus in the vertical direction, and Eh = bEv, where Eh is the Young’s modulus in the horizontal direction and b is a parameter specifying the degree of anisotropy. Values of Poisson’s ratio employed in the analysis

are mh = mv = 0.2, mhv = mvh = 0. The values employed for mhv and mvh are based on the assumption that no change in horizontal length will occur in the wall length direction due to gaps between wall panels. The value b = 0.1 was selected for the diaphragm wall on the basis of a previous analysis (Dong et al., 2016) of a diaphragm wall-supported excavation in Shanghai. Reduced values of anisotropy parameter (b = 0.001, 0.0001) were selected for the soil mixing wall and the bored pile wall respectively; the relatively flexible connections between the piles in these cases are thought to justify these lower values of b (also on the basis of a previous parametric study (Dong, 2014)). The horizontal beams and floor slabs were modelled as linear elastic. Prescribed thermal strains were used to model the shrinkage that occurs during the concrete curing process, and also other related effects (e.g. cracks, creep of the concrete, gaps between the wall and the support system) that act to reduce the propping action of the slabs. Thermal effects occur due to temperature changes during concrete curing and also ambient temperature variations (Whittle et al., 1993; Boone and Crawford, 2000; Hashash et al., 2003; Kim and Ahn, 2009). The coefficient of thermal expansion of concrete, a, was assumed to be 10
5/K; the prescribed temperature change was denoted DT. The steel struts in the bottom-up area and the exchange station area were represented by linear elastic materials. The piles and columns were modelled as a linear elastic material with concrete properties. The material parameters for the structural components are listed Table 3. 4. Analyses strategies An exploratory study has been conducted on the NSC excavation using the analysis strategy shown in Fig. 10. The calculations conducted are specified in Table 4. Initially a ‘central’ analysis was conducted; this was intended to provide a representation of the construction processes

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

Fig. 10. Analysis strategies (with nomenclature defined in Table 4).

actually employed in the project. The procedures used to develop the central analysis were based on those in Dong

973

et al. (2016). Appropriate values of b (the degree of anisotropy) for the diaphragm wall and DT (temperature change required to model post-cure shrinkage in the horizontal beams and slabs) need to be selected. The values adopted in all the analyses, b = 0.1 and DT =
25 K, were calibrated in the central analysis to provide a reasonable comparison with the field data (especially the inclinometer readings at I-6, I-25, and IT-10, and the settlement marker data along BC, as shown in Fig. 1). All other parameters adopted in the central analysis were based directly on the available geotechnical and structural data. Four sets of subsidiary calculations were then conducted to investigate the influence of key aspects of the model on the computed performance of the excavation. The primary purpose of these calculations was to investigate of the sensitivity of the results to the incorporation of alternative construction procedures in the analysis. In the central analysis, the actual construction sequence specified in Fig. 1 and Table 1 (in which the soil was excavated and the horizontal beams and slabs were installed in zones #1 to #10 in a sequential manner from number 1 to 10) was adopted; the modelled construction sequence was

Table 4 Specification of the subsidiary analyses. Procedures explored

Nomenclature

Analysis description

Un-zoned excavation

Central analysis

Remove the soil in one zone (from #1 to #10) in each layer, and install the horizontal beams and floor slab in this zone Remove the whole layer of soil at one time, and install the whole level of horizontal beams and floor slab

Z1 Construction sequence

Central analysis C1

C2

Follow the real construction sequence Follow the assumed alternate excavation (see Fig. 11a) which follows the construction sequence from zone #1 to zone #10. First, the sequence is used to construct the ground level slab. Next, the sequence is used to construct the lower level floor slab Follow the assumed double excavation (see Fig. 11b) which starts from the construction of zone 1 in both levels, and then moves to the next zone till the end

Earth berms

Central analysis B1

Excavation proceeds with earth berms (see Fig. 4) Excavation proceeds without earth berms

Openings

Central analysis O1

Openings in the floor slabs are modelled (see Fig. 3) Openings in the floor slabs are omitted

Fig. 11. Assumed construction sequence, (a) the alternate excavation sequence employed in analysis C1, (b) the double excavation process employed in analysis C2.

974

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

also indicated in Fig. 6. The subsidiary analysis Z1 investigated the sensitivity of the computed performance of the excavation to this aspect of the construction process. Analysis Z1 employs a simplified construction procedure in which all the zones #1 to #10 are constructed (by soil excavation followed by installation of the floor slab and beams) in a single analysis step. This allows a significant reduction in computation time in comparison with the central analysis. Analyses C1 and C2 were used to investigate the sensitivity of the results to the detailed order in which zones #1 to #10 are excavated and constructed. Two alternative construction schemes were investigated: ‘alternate excavation’, analysis C1 (in which the construction of zones #1 to #10 was conducted in the order indicated in Fig. 11a) and a ‘double excavation’, analysis C2 (in which the construction of zones #1 to #10 proceeded sequentially in Fig. 11b, but with the construction of each zone in turn to the full depth of the excavation). In Analysis B1, the earth berms (present in the central analysis) were excluded from the calculation. In analysis O1, the floor slabs were modelled as continuous (i.e. the openings in the slabs were not present in the model).

5. Results The finite element analyses produced a very substantial amount of data. For presentation purposes, interpretation of the results is focused on the retaining wall deflections at inclinometer positions I-6 and I-25, the ground settlement outside the excavation along Line BC, and the lateral ground displacement at IT-10 (Fig. 1). Inclinometer I-25 was selected because it is near the centre of the diaphragm

wall and near to several large openings in the floor slabs. Inclinometer I-6 is close to a corner, and is also near to several large openings in the slab. Ground settlements along BC and the lateral displacements at IT10 provide a representative indication of the ground movements outside the excavation. More comprehensive data on the computed results and field measurements are given in Dong (2014). In the finite element results presented here, the computed wall toe movement is subtracted from the computed wall deflection to provide a fair comparison with the field measurements, which were made relative to the wall toe. The ground movements induced during the wall installation and dewatering process are excluded from the field measurements for comparison with the numerical analyses, which did not include an analysis of these processes.

5.1. Central analysis and simplified unzoned excavation The central analysis captures the wall deflections and ground movements indicated by the field data (Fig. 12) reasonably well. There are, however, some discrepancies. For instance, although the results for the maximum value of wall deflection at I-6 and I-25 agree well with the field data, the computed wall deflections at the top and bottom parts of the wall at I-6, and bottom part of the wall at I-25 are lower than the corresponding values from the field. Moreover, the central analysis underestimates the ground settlement close to point B along line BC, and overestimates the lateral displacement at shallow depths at the location of inclinometer IT10. These discrepancies arise from assumptions and limitations in the finite element model (e.g. simplified ground profile (i.e., stratigraphy, initial stress distribution, stiffness and strength profile), ignoring the wall installation and dewatering processes, limitations in

Fig. 12. Wall deflections and ground movements (effect of unzoned excavation, analysis Z1).

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

the constitutive modelling procedures) in conjunction with uncertainties in the local ground conditions and possible errors in the field measurements. Sensitivity studies on the likely influence of these aspects of finite element models of excavation behaviour in Shanghai Clay are described in Dong et al. (2016) and Dong (2014). For instance, the overprediction of lateral movements at shallow depth is probably related to the underestimation of the undrained strength (Fig. 5) and stiffness profile, as indicated in the parametric analyses in Dong (2014). In general, however, the central analysis provides a reasonable representation of the observed performance of the excavation, giving some confidence in the results of the subsequent exploratory study. The analysis with a simplified unzoned excavation (Z1) gives results that are almost identical to the central analysis at the end of construction (Fig. 12). However, analysis Z1 employs fewer calculation steps and, as a consequence, is significantly faster. This suggests the possibility that, for design purposes, it may not be necessary to incorporate the full details of a sequential zoned construction process within a finite element model if estimates are only needed of the final movements. Because analysis Z1 cannot provide details at intermediate stages of construction, if performance during construction is of interest, the full sequence should be followed. Although this finding is applicable to this particular case, it has yet to be verified for other cases with different ground conditions and construction methods. 5.2. Influence of construction sequence The results of analyses C1 and C2 (in which alternative construction sequences are explored) are indicated in Fig. 13, which shows that the computed data from

975

C1 and C2 are not significantly different from the results of the central analysis. It appears that details of the construction sequence do not have a significant influence on the computed ground and wall movements at the final stage of the excavation (although differences will undoubtedly occur at intermediate stages in the construction process). In practice, it is necessary to adopt a sequence of construction that ensures that an appropriate level of structural stability is maintained during the construction process. This confirms the observation with regard to analysis Z1 that construction sequence may not be important in determining the final deformation of deep excavations. 5.3. Influence of earth berms The computed data (Fig. 14) indicate that when the earth berms inside the excavation are excluded (analysis B1) the computed displacements at the end of construction are systematically greater than those from the central analysis. If earth berms are excluded from the model, the largest wall deflection at I-25 and the ground settlement along BC are increased by about 25% and 40% respectively with respect to the central analysis. The wall deflection at I6 (Fig. 14a) is influenced less by the berms than the deflection at I-25 (Fig. 14b); this reflects the fact that the earth berms are mainly deployed on the north side of the excavation, and are therefore likely to have a stronger influence on the portion of the wall monitored by inclinometer I-25. The current results confirm previous research that indicates that earth berms can be effective in reducing wall deflections and ground movements (Clough, 1977; Georgiadis and Anagnostopoulos, 1998; Gourvenec and Powrie, 2000; Daly and Powrie, 2001; Powrie and Daly, 2002; Smethurst and Powrie, 2008). Earth berms provide

Fig. 13. Wall deflections and ground movements (effect of construction sequences, analyses C1 and C2).

976

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979 0 Field data Central analysis B1

-4

-2

(b) I-25

-2

-4

Field data Central analysis B1

-4

-6

-8

-10

-10

-10

-12

-12

-12

-16

Depth (m)

-8

-14

-14 -16

-16

-18

-18

-20

-20

-20

-22

-22

-22

-24

-24

-24

-26

-26 0

10

20

30

40

50

60

-10

-26 0

Lateral displacement (mm)

0

Field data Central analysis B1

-14

-18

-10

(c) IT-10

-6

-8

Depth (m)

Depth (m)

-6

Ground settlement (mm)

0

0 (a) I-6

-2

10

20

30

40

50

60

-10

0

10

20

30

40

50

60

Lateral displacement (mm)

Lateral displacement (mm)

(d) BC

-5

B

Field data Central analysis B1

-10

C

-15 -20 -25 0

50

100

150 Distance (m)

200

250

300

Fig. 14. Wall deflections and ground movements (effect of earth berms, analysis B1).

lateral support to the retaining wall and increase the effective embedment depth of the wall, increasing the magnitude of the passive earth pressures. The effectiveness of earth berms is determined by various factors (e.g., berm geometry and spacing, soil properties, excavation depth and wall stiffness). However, they add complexity to the construction process. Careful consideration needs to be given to the design of optimum berm configurations and construction processes for any particular application.

deformations in the retaining wall to increase. The wall deflections at I-6 and I-25 and the ground settlement along BC computed in the central analysis are of the order of 20% greater than those obtained from analysis O1, in which the slabs are continuous (i.e., without openings). It is noticeable that openings affect the detailed ground settlement pattern along BC in the central analysis, increasing the ground settlement close to the openings (Fig. 15d). These results suggest that any openings in the slabs should be included explicitly within a finite element model of the excavation process. This involves a more complex approach than has been adopted in some previous work (e.g. Simpson, 1992; St. John et al., 1993) in which the stiffness of the floor slab is modified by an assumed reduction

5.4. Influence of openings in floor slabs The computed results in Fig. 15 indicate that, as expected, temporary openings in the floor slabs cause the 0

0

(a) I-6

-2

Field data Central analysis O1

-4

-6

-6

-8

-8

-10

-10

-12

-12

-12

Depth (m)

-8

-16

-14

-14

-16

-16

-18

-18

-18

-20

-20

-20

-22

-22

-22

-24

-24

-24

-26

-26

-10

0

10

20

30

40

50

60

-10

-26 0

10

20

30

40

50

60

-10

0

10

20

30

40

50

60

Lateral displacement (mm)

Wall deflection (mm)

Wall deflection (mm)

Field data Central analysis O1

-4

-10

-14

(c) IT-10

-2

Field data Central analysis O1

-4

Depth (m)

Depth (m)

-6

0

(b) I-25

-2

Ground settlement (mm)

0

(d) BC -5

Field data Central analysis O1

B

-10

C

-15 -20 0

50

100

150 Distance (m)

200

250

Fig. 15. Wall deflections and ground movements (effect of openings, analysis O1).

300

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

factor to account for openings. The reduction factor is an empirical parameter which depends on the size and distribution of the openings, and for practical design an appropriate value may be difficult to determine. In addition, this method cannot account for the regional variations in the stiffness of the floor slabs caused by the openings. 6. Conclusions The central analysis (based on a detailed representation of the structural behaviour and geometry of the retaining system, openings in the floor slabs, post-cure shrinkage of concrete beams and floor slabs, construction sequence, temporary earth berms, and small-strain non-linearity of the soil) was shown to provide a good match with data collected during the construction of the NSC project. This provides some confidence that the central analysis can be used as the basis of a set of exploratory calculations to determine the influence of construction processes on the computed results. The following conclusions are drawn from the results of the exploratory study: (a) The computed deformations are relatively insensitive to the order in which the various floor slab/excavation zones are constructed. Moreover, the results obtained for the case where the construction procedure was modelled as a sequential zoned process are similar to those obtained when the construction of the floor slab (and associated excavation) was modelled as a single operation. This suggests that for design calculations (at least for preliminary design), approaches employing simplified models of the construction sequence may provide a convenient and efficient approach. It is noted, however, that details of the modelling procedures for the construction process influences the way in which the displacements develop during the construction process. It is also noted that the range of construction processes explored in this study is limited; the conclusions drawn from the present study should therefore be treated appropriately.

977

(b) The use of earth berms in deep excavations can be an effective means of reducing the wall deflections and ground movements. However, earth berms may also complicate the construction process. Careful design is needed to ensure that the system of earth berms employed in any particular project is effective, economical and efficient. (c) Any temporary openings in the floor slabs will act to reduce the overall stiffness of the retaining system. As a consequence, wall deflections and ground movements are increased. The sizes and distribution of the openings need to be considered carefully in the design process, to provide a balance between performance and efficiency. Any temporary openings in the floor slabs need to be included explicitly in the modelling process to obtain reliable results. Acknowledgement The first author was sponsored by the China Scholarship Council to study at Oxford University. The field measurements were conducted by Dr Z.H. Xu who also analysed the initial data. The calculations were conducted at the Oxford Supercomputing Centre. Appendix A. Soil model calibration The soil constitutive model used in this case study is a theoretical model specially developed to consider the small-strain stiffness nonlinearity for undrained behaviour of clays in a total stress analysis. Features of this model are described in the main text. Calibration of this soil model for Shanghai Clay is described in Dong (2014) and Dong et al. (2016). Simulation of triaxial tests using this soil model and derived parameters are presented in this Appendix, to show its performance in modelling the behaviour of Shanghai Clay in undrained conditions. Huang et al. (2011) presented data on CIU (isotropic consolidated undrained) and CAU (anisotropic consolidated undrained) triaxial tests of Shanghai Clay under constant axial strain rate, using undisturbed samples retrieved from 10 m below the ground level. Stress conditions for these tests are shown in Table A1. Single element tests have been conducted in

Table A1 Test conditions for undrained triaxial tests on Shanghai Clay. Test no.

Lab test conditions 0

CIU-1 CIU-2 CIU-3 CIU-4 CAU-1 CAU-2

0

Simulated element test conditions

r a (kPa)

r r (kPa)

u0 (kPa)

K0

ra (kPa)

rr (kPa)

Kt0

z (m)

su (kPa)

qu (kPa)

G0/su

50 100 150 200 68.6 136.4

50 100 150 200 41 81.8

100 100 100 100 100 100

1.0 1.0 1.0 1.0 0.6 0.6

150 200 250 300 168.6 236.4

150 200 250 300 141 181.8

1 1 1 1 0.835 0.769

5.75 11.51 17.26 23.01 7.89 15.70

23.63 32.26 40.89 49.52 26.84 38.54

20.46 27.94 35.41 42.89 23.25 33.38

1000 1000 1000 1000 1000 1000

Note: r0 a and r0 r are effective stresses of the sample in axial and radial directions; ra and rr are the corresponding total stresses in the simulated tests. z is depth below ground level estimated from the axial effective stress in the test r0 a, by z = r0 a/(18.5–9.81). The value of z is used to estimate the corresponding su for the element test, by su = 15 + 1.5z. su is the undrained shear strength in simple pffiffiffiffiffiffiffi shear mode and is used directly in the soil model, while qu is the undrained shear strength in triaxial compression, and qu is related to su by qu ¼ 3su =2.

978

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979

Fig. A1. Simulation conducted in Abaqus using the CIU and CAU tests on Shanghai Clay and comparison with laboratory data.

Abaqus to simulate these triaxial tests under the conditions employed in the tests, with test conditions also shown in Table A1. The undrained shear strength, su, for each test is selected based on the equivalent depth, z (units metres), of the corresponding axial effective stress, by su = 15 +1.5 z (units in kPa). For instance, the su for CIU-1 test (r0 a = 50 kPa), is estimated from the equivalent depth z = 5.88 m, calculated using the bulk unit weight of soil (18.5 kN/m3) and assumed hydrostatic pore water pressure distribution with the water table at the ground surface, by z = 50/ (18.5–10). The other parameters are the same as those employed in the case study, such as Ir = G0/su = 1000, ci, and gi. Comparisons of the simulated results and test data are shown in Fig. A1. Results indicate that the soil model can represent reasonably well the undrained shear strength at large strains in both CIU and CAU tests (except CIU-4), which is expected because su is an input parameter in the model. However, due to the nature of the model it cannot capture the strain softening behaviour as observed in the laboratory tests. The soil constitutive model represents well the pre-yield stiffness in CAU tests, but overestimates the pre-yield stiffness in CIU tests. Since the soil model cannot simulate the stress-strain data perfectly,

discrepancies between the computed results and field measurement are expected in this case study, as discussed in the main text. Appendix B. Influence of number of elements through the wall thickness Two elements were used through the wall thickness in the finite element meshes employed in the case study, providing a balance between accuracy and computational efficiency. Elements will become very thin when more elements are created through the wall thickness. Very fine meshes are required to maintain a reasonable element aspect ratio to avoid numerical problems, and this will significantly increase the total number of elements in the model. To investigate the influence of number of elements through the wall thickness, calculations have been conducted for the idealised excavation described in Dong (2014), with model details shown in Fig. B1. Parametric analyses (Fig. B2) showed that wall deflections and ground settlements are almost identical when one, two, four and six elements are employed through the wall thickness. This suggests that two elements through wall thickness, as employed in the current study, is sufficient.

Fig. B1. Mesh for the soil and retaining structures in the analysis (Dong, 2014).

Y. Dong et al. / Soils and Foundations 57 (2017) 965–979 2

0 (a) Wall deflection along AB

(b) Ground settlement along CD 0

Ground settlement (mm)

Wall Depth (m)

5 10 1 element 2 elements 4 elements 6 elements

15 20 25 30

979

0

2

4

6 8 10 12 14 Lateral displacement (mm)

16

18

20

-2 1 elements 2 elements 4 elements 6 elements

-4 -6 -8

-10

0

20

40 60 Distance from wall (m)

80

Fig. B2. Wall and ground movement at the final stage of the excavation.

References Boone, S.J., Crawford, A.M., 2000. Braced excavations: temperature, elastic modulus, and strut loads. J. Geotech. Geoenviron. Eng. 126 (10), 870–881. Burd, H.J. et al., 2000. Modelling tunnelling-induced settlement of masonry buildings. Proc. Inst. Civ. Eng.: Geotech. Eng. 143 (1), 17–29. Clough, G.W., 1977. Stabilizing berm design for temporary walls in clay. ASCE J. Geotech. Eng. Div. 103 (2), 75–90. Daly, M.P., Powrie, W., 2001. Undrained analysis of earth berms as temporary supports for embedded retaining walls. Proc. Inst. Civ. Eng.: Geotech. Eng. 149 (4), 237–248. Dong, Y., 2014. Advanced finite element analysis of deep excavation case histories. DPhil thesis, University of Oxford, Oxford, UK. Dong, Y., Burd, H.J., Houlsby, G.T., 2016. Finite element analysis of a deep excavation case history. Ge´otechnique 66 (1), 1–15. Georgiadis, M., Anagnostopoulos, C., 1998. Effect of berms on sheet-pile wall behaviour. Ge´otechnique 48 (4), 569–574. Gourvenec, S.M., Powrie, W., 2000. Three-dimensional finite element analyses of embedded retaining walls supported by discontinuous earth berms. Can. Geotech. J. 37 (5), 1062–1077. Hashash, Y.M.A. et al., 2003. Temperature correction and strut loads in Central Artery excavations. J. Geotech. Geoenviron. Eng. 129 (6), 495–505. Hou, Y.M., Wang, J.H., Zhang, L.L., 2009. Finite-element modeling of a complex deep excavation in Shanghai. Acta Geotech. 4 (1), 7–16. Houlsby, G.T., 1999. A model for the variable stiffness of undrained clay. In Proceedings of the International Symposium on Pre-Failure Deformations of Soil. Torino, pp. 443–450.

Huang, M., Liu, Y., Sheng, D., 2011. Simulation of yielding and stressstain behavior of shanghai soft clay. Comput. Geotech. 38 (3), 341– 353. St. John, H.D. et al., 1993. Prediction and performance of ground response due to construction of a deep basement at 60 Victoria Embankment. Predictive Soil Mechanics. Proc. of the Wroth Memorial Symposium, Oxford, 1992, 581–608. Kim, H., Ahn, H., 2009. Shrinkage stress analysis of concrete slabs in a multi-storey building considering internal and external restraints. Struct. Des. Tall Spec. Build. 18 (5), 525–537. Powrie, W., Daly, M.P., 2002. Centrifuge model tests on embedded retaining walls supported by earth berms. Ge´otechnique 52 (2), 89– 106. Powrie, W., Preene, M., 1994. Time-drawdown behaviour of construction dewatering systems in fine soils. Ge´otechnique 44 (1), 83–100. Simpson, B., 1992. Retaining structures: displacement and design. Ge´otechnique 42 (4), 541–576. Smethurst, J.A., Powrie, W., 2008. Effective-stress analysis of bermsupported retaining walls. Proc. Inst. Civ. Eng.: Geotech. Eng. 161 (1), 39–48. Whittle, A.J., Hashash, Y.M.A., Whitman, R.V., 1993. Analysis of deep excavation in Boston. J. Geotech. Eng. – ASCE 119 (1), 69–90. Xu, Z.H., 2007. Deformation Behaviour of Deep Excavations supported by Permanent Structure in Shanghai Soft Deposit. PhD thesis, Shanghai Jiao Tong University, Shanghai, China. Zdravkovic, L., Potts, D.M., St. John, H.D., 2005. Modelling of a 3D excavation in finite element analysis. Ge´otechnique 55 (7), 497–513. Zheng, G. et al., 2014. Test and numerical research on wall deflections induced by pre-excavation dewatering. Comput. Geotech. 62, 244–256.