You are on page 1of 15

Cite this article Research Article Keywords: computational mechanics/

Ieronymaki ES, Maniscalco JD and Corral G (2021) Paper 2000113 excavation/retaining walls
Excavation case study in Manhattan, NY with practical approach to predict ground Received 02/06/2020;
movements. Accepted 02/12/2020;
Proceedings of the Institution of Civil Engineers – Geotechnical Engineering 174(4): 446–460, Published online 10/02/2021
https://doi.org/10.1680/jgeen.20.00113
ICE Publishing: All rights reserved

Geotechnical Engineering

Excavation case study in Manhattan,


NY with practical approach to predict
ground movements
Evangelia S. Ieronymaki PhD Gonzalo Corral ScD
Assistant Professor, Department of Civil and Environmental Engineering, CEO & Founder, Inteligencia Geotécnica SpA, Santiago, Chile
Manhattan College, Riverdale, NY, USA (Orcid:0000-0003-3059-3639)
(corresponding author: ieronymaki@manhattan.edu)
James D. Maniscalco MSc
Project Engineer, Underpinning & Foundation Skanska, Maspeth, NY, USA

This paper presents the case study of the staged excavation for the New York University College of Nursing building
in Manhattan, NY, supported by corner-braced secant pile walls. A commercial finite-element software was used to
simulate the complex supported excavation and predict the retaining wall deflections using simple soil models
calibrated from limited soil data. The results of the two-dimensional (2D) simulations were compared with actual wall
deflections measured by inclinometers during the excavation process. The aim of this research was to determine
whether the complex 3D characteristics of the problem could be captured and accurately simulated by equivalent 2D
models, using soil properties estimated from limited field data, which is typically the case in engineering practice.
Equivalent axial and bending stiffnesses were also estimated for simulating the square bracing system in plane-strain
conditions. The numerical analyses yielded results that correlated with the measured data which indicates that – for
this type of excavation problem – simplified plane-strain analyses can predict sufficiently small wall movements, given
that appropriate stiffness magnitudes are assigned to the earth-retaining system.

Notation is still the most reliable numerical method used by engineers


A cross-sectional area for modelling and predicting soil and structural deformations
c′ soil cohesion in deep excavations. However, the accuracy of these methods
d secant pile wall width depends on the selection of the soil properties and the appro-
E′ elastic stiffness modulus priate constitutive model, as well as proper modelling of the
Eu,r elastic stiffness modulus under unloading/reloading structural elements and selection of the material properties.
I moment of inertia
K0 lateral earth pressure coefficient at rest Many constitutive models have been developed claiming to
k spring stiffness accurately predict soil behaviour (e.g. Pestana and Whittle,
keq perpendicular axial stiffness per node 1999; Schanz et al., 1999; Whittle, 1993; Yuan and Whittle,
L length 2018) and significant research has been conducted on appro-
Lstrut length of strut priately applying these models in numerical simulations for
Lt tributary length deep excavations (Corral and Whittle, 2010; Hashash and
P point load Whittle, 1996; Khoiri and Ou, 2013; Orazalin et al., 2015;
patm atmospheric pressure Whittle et al., 2015). However, their capability relies on the
su undrained shear strength appropriate selection of the necessary input parameters, which
w distributed load are often calibrated against data from laboratory or in situ soil
z depth tests (Ieronymaki et al., 2017; Pestana et al., 2005).
γ soil unit weight The quality of soil data therefore governs the performance of
δ deflection constitutive models and thus the accuracy of the prediction.
δ° interface friction angle The most complex of these soil models requires the calibration
ν′ Poisson’s ratio of parameters that do not have immediate physical meaning
ϕ′ friction angle (e.g. parameters that describe the curvature of failure
envelopes, the small-strain non-linearity in shear, the evolution
1. Introduction of anisotropy, etc.) and are not as easily recognisable as the
Due to the high density and close proximity of buildings in well-known Young’s modulus, Poisson’s ratio, cohesion and
overpopulated urban areas such as New York City, deep friction angle (Lade, 2005). In addition, the calibration of
excavations for building foundations require rigid earth- these parameters is based on high-quality laboratory data,
support systems to limit lateral ground movements and prevent which is a time-consuming procedure and often impractical for
settlements of the surrounding structures. Numerical analysis practising engineers, especially if only field data is available.

446
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

To deal with this challenge, several researchers have proposed predicted wall deflections (type C1 predictions) (Lambe, 1973).
inverse modelling for back-calculating soil properties obtained The results were then compared against the actual measured
from monitoring in the construction phase (Calvello and inclinometer data obtained during construction and against
Finno, 2004; Corral Jofré, 2013; Poeter and Hill, 1997). This the results of limit equilibrium basic elasto-plasticity analysis
method requires monitoring of ground deformations and uses undertaken using the commercial software Deepex (version
the data measured at an early stage of excavation to optimise a 2017).
numerical model automatically and therefore predict the
ground deformations for the following stages of construction
2. Project overview and site characteristics
(i.e. application of the observational method (Peck, 1969)).
This procedure can become very complicated, depending on 2.1 General project characteristics
the type of system to be modelled, the number and type of The case study selected for analysis is the secant pile wall
instrumentation and measured data, the selected constitutive supported excavation for the construction of the 11-storey
model, the number of parameters to be optimised and the New York University College of Nursing building in
interdependence of these constitutive model parameters. Manhattan (Figure 1). Secant pile walls are an attractive
option for earth-retaining systems in dense urban environments
Another challenge is that, among practitioners, two-dimensional as they provide both a rigid support of the excavation and a
(2D) modelling (usually plane-strain) is still preferred to 3D means of facilitating water cut-off of the excavated area when
analyses as it is less complex and requires much less compu- socketed into a relatively impermeable layer (e.g. rock).
tational time. Thus, engineers have to define the appropriate
equivalent properties for the structural elements in order to Secant pile walls consist of drilled cast-in-place overlapping
accurately simulate 3D problems in 2D plane-strain conditions. circular concrete shafts that form a continuous wall along the
For the majority of typical construction projects involving deep excavation perimeter. Wall construction is sequenced by
excavations, geotechnical engineers use (a) in situ soil testing initially installing unreinforced concrete ‘primary’ piles, allow-
data to estimate soil properties, (b) simplified 2D numerical ing the concrete to harden and then subsequently drilling over-
analyses and (c) simple constitutive soil models to predict the lapping ‘secondary’ piles between the primary ones. Typically,
ground response. The aim of this paper is to suggest a direct a steel reinforcing cage or steel wide-flange section is encased
method for predicting lateral wall movements of complex in the secondary piles to provide flexural rigidity to the wall
supported deep excavations using simple 2D FE analyses with and limit lateral deflections during excavation. The continuous
soil properties obtained from limited geotechnical information concrete overlap provides wall continuity, which is crucial to
(in situ test data). the impermeability of the wall, acts as lagging between the
steel reinforcements and adds to the overall rigidity of the
The method presented in this paper is compared against a case system.
study on the excavation for the construction of the New York
University College of Nursing in Manhattan (Maniscalco and An existing 25-storey reinforced concrete residential building,
Ieronymaki, 2018). In this project, a secant pile wall socketed supported on steel H-piles (Figure 2), is located directly
into rock was installed to serve as the support and water adjacent to the western side of the excavation, while an existing
sealing of the excavation for the construction of the 11-storey six-storey brick building is located about 6.1 m south of the
building. The secant pile wall served not only as the temporary site. Due to the close proximity of the adjacent high-rise build-
support-of-excavation system, but also as the permanent ing, inclinometers were installed in two secondary secant piles
foundation walls of the basement of the building. During of the west wall to monitor lateral deflections during construc-
construction, the walls were supported by three levels of steel tion and ensure that ground movements remained within
corner bracing and the lateral deflections were monitored by tolerable limits (Figure 3).
inclinometers installed in two secondary piles of the wall.
The excavation depth was approximately 13.1 m (square plan
This case provides various levels of complexity. Its 3D geo- view of 37 m  37 m) to accommodate for the construction of
metry comprises a four-stage square excavation with an the two basement levels of the new building. To provide exca-
exterior secant pile wall supported by three levels of internal vation support and a relatively impermeable dewatering
diagonal bracing and a non-symmetrical external surface system, both primary and secondary secant piles (1 m diameter
loading resulting from a pile-supported high-rise building on and 1.543 m spacing from centre to centre) were socketed into
one side of the excavation only. The system was modelled rock at a minimum of 0.61 m, or they were extended 0.61 m
using 2D plane-strain numerical analyses and the simple linear below the bottom of the excavation if that was deeper than the
elastic–perfectly plastic Mohr–Coulomb (M–C) soil model, bedrock interface (Figure 4). The depth to rock generally
calibrated against standard penetration test (SPT) data. varied across the site; for the west wall specifically, the depth
The commercial finite-element (FE) program Plaxis 2D to rock was 12–20 m, with an average depth of 16 m.
version 2016 (Brinkgreve et al., 2007) was used to obtain Inclinometer 1 was installed in pile 164 and inclinometer 2

447
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

E.
26
th
Str
eet

Approximate site limts


and outline of
secant pile wall

Adjacent 25-storey
building

e
enu
Av
1st
E.
25
th
Str
eet Adjacent six-storey
50 ft
brick building
10 m

Figure 1. Site location plan view (Map Data: Google Maps)

N
5.8 m
8.2 m

(a) (b)

Figure 2. 25-storey building adjacent to the excavation: (a) view of the structure; (b) partial plan view of the steel H-pile foundation and
cross-section of the west secant pile wall

was installed in pile 184, with depths to rock of 16.5 m and steel wide-flange members diagonally bracing walers that
14 m, respectively (Figure 3(b)). spanned the full length of the wall. For all three levels, the
bracing system layout was the same, but the member sizing
As the excavation proceeded, the secant pile walls were sup- varied at each level. Figure 5(a) shows a view of the first-
ported by three levels of internal bracing, with diagonal corner level bracing system, while Figure 5(b) shows a schematic
struts and continuous walers. The bracing system consisted of representation and dimensions of all three levels of the system.

448
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Inclinometer 2
186
pile 184
185
184
183
182
181
180
25-storey building

179

25-storey building
178
177
176
175
174
173
172
171
170
169
168
167
166
165
164
163
162 Inclinometer 1
pile 164
(a) (b)

Figure 3. (a) Location of the 25-storey building relative to the excavation. (b) Location of the inclinometers in the west secant pile wall

25-storey building
Inclinometer 1 Inclinometer 2

A A

(a)

6
Approximate ground surface
4 elevation 5.2 m

2
0
–2
Elevation: m

–4
–6
Bottom of excavation
–8 elevation –7.8 m

–10 Actual top of rock


–12
–14
–16
–18 (b)

Figure 4. (a) Plan view and (b) elevation view (elevation A–A, looking west) of the west secant pile wall

2.2 Soil stratigraphy was determined by on-site engineers. Based on information


The ground investigation comprised a total of 12 borings. SPT obtained from the borehole logs, the groundwater table sits
N60 values were recorded and visual classification of the soil about 3.5 m below the existing ground surface and the site

449
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Waler

Lace

Excavation level Bracing system

ce
a
Br
Walers W21 × 111
First level at depth 2.7 m
and Braces W24 × 76
Second level at depth 5.2 m
Laces W18 × 50
Walers W24 × 146
Third level at depth 9.1 m Braces W27 × 114
Laces W18 × 86
(a) (b)

Figure 5. (a) First-level corner bracing at 2.7 m depth. (b) Schematic representation and dimensions of the three levels of the bracing
system

stratigraphy consists of (a) manmade fill (typically encountered conditions. It should be noted that, given the alternate layers
in Manhattan) consisting of granular soil with varying of sand and clay encountered in this middle section, a uniform
amounts of construction debris, (b) alternating layers of silty sandy clay layer was assumed, as shown in Figure 7.
sand, sandy/silty clay and sandy/clayey silt, (c) decomposed
mica schist rock with varying amounts of clay and silt and The soil properties, shown in Table 1, were determined
(d) Manhattan schist bedrock with an average total core exclusively from the measured SPT data, with average values
recovery value of 83% and an average rock quality designation estimated from empirical correlations and typical values
of 63%. Of the 12 borings at the site, the four closest to the suggested in the literature. The unit weights (γ) for the top
west wall were used to determine the stratigraphy at the west three layers (fill, clay and decomposed rock) were estimated
side of the excavation, as shown in Figure 6. based on average values suggested by Hough (1969). More
specifically, the fill unit weight was taken as the average dry
The manmade fill layer gave SPT N60 values in the range of unit weight of a well-graded gravel, sand, silt and clay soil
4–20, with an average of 10, indicating a loose to medium (γ = 18.8 kN/m3). The unit weight of the clay layer (represent-
dense condition (with the higher N values most likely from ing the varved deposit of sand, silt and clay) was taken as the
construction debris). A varved layer is present directly below average saturated unit weight of silty sand, sandy clay and
the fill and was found to consist of alternating layers of clayey silt (γ = 18 kN/m3). The decomposed rock layer was
silty/clayey sand and silty/sandy clay with SPT N60 values of also assumed to be equivalent to well-graded gravel, with
5–22, with an average of 12, indicating a stiff clay condition. an average saturated unit weight of γ = 22 kN/m3. Finally,
Below the varved soils, there is a decomposed rock layer with γ = 27.5 kN/m3 for mica schist (Zhang et al., 2011) was
N60 values increasing with depth, z, starting from 9 to selected for the bedrock.
‘refusal’, with an average of 30. Mica schist bedrock was
encountered below the decomposed rock in three of the bore- The drained elastic Young’s moduli (E′) were selected for all
holes. The actual rock head encountered in the field at each soil layers, based on empirical correlations with the SPT values
secant pile location is shown in Figure 6 and was used in the available in the literature. A constant value of E′ was assigned
stratification considered in the analysis. to the fill layer, based on Equation 1a (Mitchell and Gardner,
1975) assuming a constant, C = 12 (suggested for gravelly sand)
and an average N60 = 10. The Young’s modulus of the clay
3. Numerical modelling
layer was estimated from Equations 1a and 1b for N60 values
3.1 Soil properties varying with depth, z, between 4 and 23 (Figure 6) and
Due to the changing stratigraphy along the west wall, a assuming C = 3, as suggested for clayey/silty soils. Similarly, for
generalised soil stratigraphy was considered in the 2D analysis; the decomposed rock with N60 varying with depth between 9
this was taken at the centre of the adjacent 25-storey building, and 84, E′ was estimated by averaging the values computed
between the two inclinometers, to represent average soil from Equations 1 and 2 (Bowles, 1996) for gravelly sand (with

450
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Inclinometer 1 Stratification Inclinometer 2


location mid-building location
LB-5 B-1 LB-7 B-2
6
15 12
4 6
20 7 8
4 4 16 16 6
8 N60 value
5 4
11
5
(typ.)
11
2
7 8 18 Ref.

0
13 15 22 9

–2 13 11 9 15

14 14 13
–4 11
Elevation: m

15 23 28 23
–6
25 24 35 Ref.

–8 29 29 41

–10 40 28

46
–12
84
Actual top
of rock
–14

–16

–18

Decomposed
Fill Sand Clay Rock
rock

Figure 6. Soil stratigraphy and location of inclinometers on the west secant pile wall

C = 12) and overconsolidated sand soils, respectively. Finally, a average N60 = 10. For the decomposed rock with
constant value of E′ = 11.4 GPa, representative for mica schist, average N60 = 30, an average value of ϕ′ = 38° was selected, as
was assigned to the bedrock, as suggested by Zhang et al. suggested by Peck et al. (1974). Undrained strength properties
(2011). These results are also shown in Table 1. were assigned to the clay layer, with ϕ′ = 0° and undrained
shear strength (su) computed using Equation 4 (Hara et al.,
1a: E 0 ðkPaÞ ¼ 100 CðN60 þ 6Þ for N60  15 1974), for N60 values between 4 and 23. Finally, typical values
of Poisson’s ratios were assigned for each soil layer (Bowles,
1996), while K0 values (lateral earth pressure coefficient at rest)
were computed using the classic Jaky (1944) equation
1b: E 0 ðkPaÞ ¼ 4000 þ 100 CðN60  6Þ for N60 . 15
(Equation 5). These values are also listed in Table 1.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3: ϕ¼ 20 ðN60 Þ þ 20
2: E 0 ðkPaÞ ¼ 2900N60

4: su =patm ¼ 0:29ðN60 Þ0:72


The friction angles (ϕ′) were also defined from empirical corre-
lations using average SPT values. More specifically, the friction
angle for the fill layer was estimated as ϕ′ = 32°, which is
the average of 30° suggested by Peck et al. (1974) and 34° com- 5: K0 ¼ 1  sin ϕ0
puted using Equation 3 (Hatanaka and Uchida, 1996) for an

451
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

et al., 1999; Duc et al., 2019). Researchers have shown that the
3.4 m
3m

Fill unloading/reloading stiffness (Eu,r) can be between two and


five times the loading stiffness (E50). However, the proposed
soil model assumes inherent isotropy. To impose induced
mechanical anisotropy, increased stiffnesses were manually
4.7 m

Clay
assigned: Eu,r = 3E ′ to the soil layers that entered unloading/
reloading conditions as the excavation proceeded (i.e. fill, clay
and decomposed Rock; Table 1). More specifically, increased
stiffness was assigned in stage 1 only to the fill layer, in stages
2 and 3 to both the fill and clay layers, and in stage 4 to all
6.9 m

Decomposed top three layers. It should be noted that increased stiffness was
rock
assigned to the entire soil layer with the condition that the
excavation level was greater in depth than half the height of
the corresponding layer.

Rock 3.2 Modelling of structural elements


One of the biggest challenges was to simulate the various struc-
tural components of the square 3D system to a plane-strain
2D model. These structures included the secant pile wall
Figure 7. Average soil stratigraphy between the two around the excavation, the three levels of corner-braced struts
inclinometers and the adjacent 25-storey building, supported on H-piles.

3.2.1 Secant pile wall


In the numerical analyses, the bedrock was modelled as linear Plate elements were used to model the secant pile wall in
elastic, while the linear elastic–perfectly plastic M–C soil Plaxis, using an equivalent elastic stiffness of the composite
model was used for the top three soil layers, as is widely used wall (i.e. steel core beam in secondary piles and concrete
in engineering practice. The fill and decomposed rock were ‘lagging’ in between). The overall wall moment of inertia was
assumed drained, while the clay layer was modelled as calculated per unit length based on the geometry of the secant
undrained (undrained B). The input parameters selected for pile wall. A composite modulus of elasticity was computed to
the linear elastic and M–C soil models are summarised in account for the presence of the steel beam by taking a
Table 1. One of the limitations of the M–C model is that it weighted average of each material’s contribution to the stiffness
cannot capture the small-strain non-linear stiffness that more (i.e. concrete and steel). Thus, the elastic modulus used was
advanced soil models can (hardening soil model, Modified increased from the elastic modulus of concrete alone. The
CamClay, MIT-S1, etc.), but the latter require many more equivalent cross-section of the wall is shown in Figure 8 and
input parameters that are obtained from project-specific lab- the structural properties of the wall are summarised in Table 2.
oratory testing. Finally, interface elements were placed on both sides of the
wall with the interface friction angle (δ°) between the soil and
It has been shown that soil stiffness is higher in the the concrete wall equal to two-thirds the friction angle of the
unload/reload path than in loading (Brinkgreve, 2005; Schanz soil (δ° = 2/3 ϕ′).

Table 1. Soil properties and input parameters for the linear elastic and M–C soil models
Fill Decomposed rock Clay Rock

M-C (drained) soil model M-C (undrained B) soil model Linear elastic soil model

Range of N60 4–20 9–84 4–23 —


Average N60 10 30 13 —
γ: kN/m3 18.8 22 18 27.5
E0 : MPa 19 74 + 6.3z 7.6 + 0.5z 11 400
Eu,r: MPa 57 222 + 19z 22.8 + 1.5z —
v0 0.3 0.2 0.4 0.2
c0 : MPa 0.0 0.0 — —
ϕ0 : degrees 32 38 — —
K0 1 − sinϕ 1 1
su: kPa — — 42 + 3z —

452
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Ixx_total = 9 433 844 cm4 EC = 34 GPa Ixx_total = 9 433 844 cm4


Atotal = 13 716 cm2 Atotal = 13 716 cm2

W24 × 131
Ixx = 167 325 cm4
ES = 200 GPa

Ecomp = 37 GPa

Figure 8. Equivalent composite cross-section of the secant pile wall as modelled in Plaxis 2D

Table 2. Structural properties of secant pile wall the middle of the member. Using the full strut length in
Structural type Plate this calculation would decrease the member’s perpendicular
Axial stiffness, EA : MN/m 33 405 stiffness contribution on the wall by half.
Bending stiffness, EI: MN·m2/m 2 260 148
Width, d: m 0.9 Another important factor to consider when determining
Poisson’s ratio, ν0 0.15
the overall perpendicular stiffness at a node is the axial
flexibility of the waler supporting the diagonal struts. As
shown by the red hatching (representing axial load) in
3.2.2 Bracing system Figure 9(c), the walers were axially compressed by the axial
The bracing system consisted of three levels of internal corner loading of the diagonal struts. To determine the perpendicular
bracing continuous walers supported by diagonal wide-flange stiffness provided by the diagonal struts accurately, the spring
corner struts (Figure 5). The member layout was the same for of the waler support has to be incorporated at each node as
each bracing level, but the member sizing varied per level. In an additional spring in series (Figure 9(a) – springs at nodes B,
order to model the support system in 2D FE analyses, each C and D). Considering the additional waler support spring
member was resolved into an equivalent spring acting perpen- accurately reduces the overall stiffness provided by each strut
dicular to the wall (Figure 9(a)) using Plaxis elastic fixed-end and takes into account the bracing system compressing in on
anchor elements. The continuous nature of the bracing support itself. Ignoring this axial waler spring results in doubling the
along the wall was incorporated in the model by inputting the calculated stiffness of the bracing for each level.
tributary spacing of each strut (Lspacing) to provide an average
strut stiffness per unit length of the secant wall. It is noted that A representative hand calculation of the spring stiffness
beam elements could alternatively be used for representing the (equivalent axial strut stiffness) at node B′ (keq-B′) is explained
support system. below. The spring stiffness keq-B′ (Figure 9(a)) was computed
using Equation 6a, where PB′ is the point load and δB′ is the
The average strut stiffness was calculated by considering an corresponding axial deflection at node B′ . The point load PB
external distributed unit load (w) acting along the perimeter of ′ = wB′Lt, where wB′ is the unit distributed load corresponding
the frame (Figure 9(b)) for each bracing level. This distributed to node B′ and Lt is its tributary length. The axial deflection
load was resolved into perpendicular point loads at each node δB′ was computed using Equation 6b, in which δBB′ is the
(P), where the diagonal struts connect to the wall. Hence, perpendicular deflection component of strut BB′ at node B′,
the resulting perpendicular axial stiffness per node (keq) was calculated using half the length of the strut (Lstrut,BB′) as it is
estimated by the corresponding perpendicular deflection (δ) symmetrically loaded from both sides, and δBC, δCD and δDD
caused by the point load at each node. The frame was loaded are the axial waler deflections of spans BC, CD and DD,
equally on all sides and thus the bracing system was under respectively (Equation 6c). It is noted that the axial waler
compression throughout (Figure 9(c)). This is important to deflection of span DD was calculated using half the length of
consider when computing the axial strut stiffness, as the the span because of load symmetry. The same procedure was
required length has to be half the actual length of the strut followed for estimating the spring equivalent axial stiffnesses
(Lstrut). Thus, half the axial deflection was computed at each of nodes C′ and D′ (keq-C′ and keq-D′, respectively). Finally,
end of the strut as the point of zero axial deflection occurs in the overall equivalent axial stiffness of each strut level in the

453
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

25-storey building Y
Lt P P P P P P
s
re
P

w
A B' C' D' D' C' B' A

X
B'
tB
ru

keq-B'
st
L

B Z
P B
LBC

s4

w
co
P/

C C
=

LCD
s
re
P

D D
LDD
2

(a) (b)

(c)

Figure 9. (a) Schematic representation of symmetric bracing layout showing connection types at each node and the input parameters for
estimating the strut axial stiffness on one node (node B0 representatively). (b) 3D modelling of support system in Staad software. (c)
Bracing system axial load magnitude graph (Staad output) and visualisation of spring supports for one side of secant wall

2D numerical model (EA in Table 3) is the average of keq-B′, P= cos 45° P


δB0 ¼ þ
keq-C′ and keq-D′. ½AEstrut;BB0 cos ð45°Þ=ðLstrut;BB0 =2Þ ðAEÞBC =LBC
6c:
2P 3P
þ þ
PB0 ðAEÞCD =LCD ðAEÞDD =ðLDD =2Þ
6a: keqB0 ¼
δB0

To confirm the stiffness computed based on classical structural


analysis equations, the commercial structural engineering soft-
6b: δB0 ¼ δBB0 þ δBC þ δCD þ δDD ware Staad was used to model the bracing system and estimate
its stiffness. The secant pile wall corners were considered to be
rigidly connected to their corresponding perpendicular sections
and did not apply any axial load to the walers (Figure 9(c)).
Additionally, the corner nodes in the Staad analysis allowed
Table 3. Structural properties of the bracing system the waler sections to deform axially but prevented perpendicu-
Strut: levels 1 and 2 Strut: level 3 lar movement in the direction of loading. These methods of
calculation (Staad analysis and hand calculations) yielded
Structure type FE anchor
overall bracing stiffness results at each node within 5% of each
Axial stiffness, EA: MN/m 120 168.8
Spacing, Lspacing: m 4.85 4.85 other. The structural properties of the bracing system are
summarised in Table 3.

454
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

N
5.8 m
8.2 m

Figure 10. Pile group layout from project contract drawings and representation in FE analysis as embedded beam row

3.2.3 Adjacent 25-storey pile-supported building The length of the building perpendicular to the secant pile
Limited information on the adjacent 25-storey building and its wall was estimated from Google Maps images as 45.4 m and
foundations was available. Test pits excavated adjacent to the the out-of-plane spacing of the pile group was assumed to be
building during the site investigation revealed that the building constant. The pile groups were spaced every 6.4 m and were
was supported by groups of H-piles (the bottom of the base- connected to a rigid plate element by a hinged top connection
ment being 4 m below the existing ground surface). The piles and extended to bedrock. It was assumed that all pile loads
were described as HP10  57 wide-flange beams with a 26 cm were transferred to the bedrock. Finally, the building load was
web and a flange width of 26 cm. H-piles are typically used in calculated assuming a 10 kPa areal load per floor, for 25
New York City as end bearing piles; they were thus assumed floors plus roof and basement.
to be bearing on rock.
4. Interpretation of results
Based on the available drawings (Figure 10), the piles were The excavation sequence was divided into four stages: a 3.4 m
installed in groups of eight, in a square formation. To incor- cut with the wall behaving as a cantilever (stage 1), a 5.8 m cut
porate their effect in the numerical analyses, the pile groups with braces at 2.7 m depth (stage 2), a 9.8 m cut with braces at
were modelled in Plaxis 2D using embedded beam row 2.7 m and 5.2 m (stage 3) and, finally, a 13.1 m cut with
elements that allow the soil to flow around them (Figure 10). bracing at 2.7 m, 5.2 m and 9.1 m depth (stage 4). Figure 11
The axial and bending stiffness of the pile group was estimated shows the excavation sequence in stages, as it was modelled in
using the properties of an HP10  57 beam and the geometry the two commercial software programs, Plaxis 2D and Deepex.
was determined from the foundation layout in the contract The excavation process was completed in 75 days, while there
drawings. The structural properties of the pile foundation were monitoring data of horizontal movements available for
system are shown in Table 4. another 45 days after completion of the excavation (total avail-
able monitoring data for 120 days).

The 25-storey building, with its basement and piles embedded


Table 4. Structural properties of H-pile foundation system of to bedrock are shown on the left of the retaining wall, while the
adjacent building locations of the bracing levels are shown with arrows on the
Structure type FE embedded beam row right side of the wall. It has to be mentioned that the building
Modulus of elasticity, E : MPa 7464 was modelled only in Plaxis 2D. The excavation levels were con-
Unit weight, γ: kN/m3 21.4 sidered to be 0.61 m below each bracing level. It is important to
Area, A: m2 2.3
note that, for excavation stages below the water level, the water
Moment of inertia, I: cm4 5007
Spacing, Lspacing: m 4.3 table was drawn down 0.61 m below the excavation level to
account for the sequential dewatering of the excavation.

455
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Stage 0: Stage 1: Stage 2: Stage 3: Stage 4:


secant wall installed 3.4 m cantilever 5.8 m excavation 9.8 m excavation 13.1 m excavation

Figure 11. Excavation sequence in stages as modelled in numerical simulations

4.1 Plaxis analyses behaviour, to account for potential drainage during the 45 days
Figure 12 shows the geometric model of the full-scale after completion of the excavation. Figure 14(a)) compares the
excavation at the final stage 13.1 m cut (with three bracing wall deflections immediately after the excavation completion
levels): the high-rise building is adjacent to the west wall with its with those 45 days later (Figure 14(b)). In both cases the calcu-
4 m basement and the assumed pile foundation layout. The lated deflections matched the inclinometer data both in terms of
lateral movement data available from the inclinometers were shape and magnitude.
compared with the Plaxis predictions at three excavation levels:
5.8 m cut with one bracing system (Figure 13(a)), 9.8 m cut Comparing the west wall deflections with the east side of the
with two bracing systems (Figure13(b)) and 13.1 m cut with box (where there is no building surcharge), the east-side move-
three bracing systems (Figure13(c)). Figure 13 shows the inclin- ments were much larger (Figure 15). This can be considered
ometer data up to completion of the last stage of the excavation, counter-intuitive given that the building essentially applies
which lasted for 75 days. Given the short duration of the exca- additional surcharge on the horizontal stresses; however, the
vation process, the numerical analyses were performed assuming building vertical load is transferred down to bedrock through
undrained conditions. Although long-term horizontal movement its pile foundation system. In addition, the pile groups serve as
data were not available, the last stage of the excavation (stage 4 soil reinforcement, preventing lateral ground movements
in Figure 11) was further simulated, ignoring the undrained through their passive resistance. Capturing this reduction in

Superstructure load

Fill
Clay

Decomposed rock

Piles
Rock

Figure 12. Plaxis 2D FE geometric model for the excavation at 13.1 m

456
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

Plaxis prediction Inclinometer 1 Inclinometer 2


0
Fill Fill Fill
2

4
Clay Clay Clay
6
Depth: m

10 Decomposed Decomposed
rock rock
12
Decomposed
14 rock

16 Rock Rock Rock

18
0 0.4 0.8 1.2 1.6 0.4 0.8 1.2 1.6 0.4 0.8 1.2 1.6
Horizontal displacement: cm Horizontal displacement: cm Horizontal displacement: cm
(a) (b) (c)

Figure 13. Comparison of measured and predicted horizontal wall movements (west-side wall) using Plaxis 2D FE analyses for excavations
of: (a) 5.8 m; (b) 9.8 m; (c) 13.1 m

Inclinometer 1 Inclinometer 1
Inclinometer 2 Inclinometer 2
Plaxis Plaxis (undrained behaviour ignored)

0
Fill Fill
2

4
Clay Clay
6
Depth: m

10

12
Decomposed Decomposed
14 rock rock

16 Rock Rock

18
0 0.4 0.8 1.2 1.6 0.4 0.8 1.2 1.6
Horizontal displacement: cm Horizontal displacement: cm
(a) (b)

Figure 14. Comparison of measured and predicted wall deflections using Plaxis 2D FE analyses: (a) immediately after excavation
completion (undrained behaviour); (b) 45 days excavation completion (ignoring undrained behaviour)

horizontal loading on the west wall is essential in generating 4.2 Deepex analyses
realistic deflections, which were also verified by the inclin- In order to make a fast and efficient comparison of the wall
ometer data. movements predicted by the two commercial software

457
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

East side (no adjacent building)


West side (with adjacent building)
0

2 Fill Fill Fill

4
Clay Clay Clay
6
Depth: m

10
Decomposed Decomposed
rock rock
12
Decomposed
14 rock

16 Rock Rock Rock

18
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Horizontal displacement: cm Horizontal displacement: cm Horizontal displacement: cm
(a) (b) (c)

Figure 15. Comparison of predicted horizontal wall movements using Plaxis 2D FE analyses for the west-side wall (with building) and the
east-side wall (no building) for excavations of: (a) 5.8 m; (b) 9.8 m; (c) 13.1 m

programs, a full 2D section was modelled using Deepex soft- the same construction sequence was also used. Figure 17 shows
ware (version 2017), as shown the representation in Figure 16. a comparison of the east-side wall deflections obtained using
Deepex is based on limit equilibrium and basic elasto- Plaxis and Deepex, neglecting the neighbouring 25-storey build-
plasticity, while the computational time for the analyses is ing. The figures show that both wall movements were almost
minimal (a few minutes). The elasto-plastic analyses are con- identical for the three excavation stages. To some extent, this is
ducted using spring models but with the flexibility of varying an important outcome as the limit equilibrium results indepen-
the stiffness of the soil layers in different directions (i.e. loading, dently validate the suggested FE analysis procedure.
unloading, pressure-dependent etc.).
5. Conclusions
All the soil and structure properties applied in the Deepex ana- This article has presented a procedure to simulate complex 3D
lyses were equivalent to those assumed in the Plaxis 2D analyses; deep supported excavations, using simple numerical models

Fill

Clay

Decomposed
rock

Rock

Figure 16. Deepex geometric model for the excavation at 13.1 m

458
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

East side - no building


Plaxis results
Deepex results
0

2 Fill Fill Fill

4
Clay Clay Clay
6
Depth: m

10 Decomposed Decomposed
rock rock
12
Decomposed
14 rock

16 Rock Rock Rock

18
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Horizontal displacement: cm Horizontal displacement: cm Horizontal displacement: cm
(a) (b) (c)

Figure 17. East-side wall deflections (no building in place) predicted using Plaxis 2D and compared with Deepex predictions for
excavations of: (a) 5.8 m, (b) 9.8 m and (c) 13.1 m

with plane-strain conditions, and accurately predict retaining Another key element that had to be carefully modelled was the
wall deflections. The case of the staged excavation of the foundation of the high-rise building. It was found from
New York University College of Nursing building in the numerical analyses that the fact that the piles were founded
Manhattan was used in the current analyses. This problem on rock minimised the induced horizontal stress that the
involved several levels of complexity that needed to be care- building applied on the retaining wall, thus reducing its lateral
fully modelled in plane-strain conditions. These complexities movement. Finally, the last important element that had to be
included the square geometry of the excavation, the secant pile correctly modelled in the numerical analyses was the bracing
walls with three levels of corner bracing and the asymmetry of system. More specifically, the bracing system was simulated
the project, with a high-rise building on the west side sup- as fixed-end anchors or cross-lots, with equivalent stiffness
ported on piles founded on bedrock. to that of the corner braces. Assigning the appropriate
connection type at the nodes of each strut of the bracing
An additional difficulty encountered was the lack of infor- system was a key component for accurate estimation of the
mation about the soil properties, with the only available bracing stiffness. A structural analysis with unit loading was
data from in situ SPT tests. The soil properties in the current conducted to confirm the equivalent stiffness for each level of
study were thus defined based on empirical correlations with support.
SPT data and typical values used in past projects in the
vicinity. When arching effects, anisotropic and non-homogeneous
properties are present, and one of the dimensions is compar-
For the soil properties estimated from the SPT data, the simple able in magnitude to the others (e.g. box excavation), it is
(and widely used in practice) linear elastic M–C soil model was expected that the ground behaviour of a complex 3D exca-
adopted in the numerical analyses. While the M–C model is vation problem cannot be fully captured in a 2D plane-strain
generally adequate for this type of project, more advanced soil condition. Nevertheless, this study suggested a practical
models can account for the stiffer behaviour of soils under approach to simplify that complexity, assuming proper stiff-
unloading/reloading conditions, which is essential for this nesses for the support system and the soil layers, especially
problem. Given that the M–C model does not allow for when there is limited soil testing data available.
varying stiffness, it was manually induced in these analyses by
assigning to the soil layers entering the unloading/reloading The results from the plane-strain FE analyses using Plaxis 2D
condition a Young’s modulus three times higher than the were in good agreement with the measured wall deflections.
initial value as excavation proceeded. This indicates that complex 3D problems of braced deep

459
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.
Geotechnical Engineering Excavation case study in Manhattan, NY
Volume 174 Issue 4 with practical approach to predict ground
movements
Ieronymaki, Maniscalco and Corral

excavations can be reasonably portrayed using simple 2D (Yamamuro JA and Kaliakin VN (eds). ASCE, Reston, VA, USA,
models with soil property correlations readily available to GSP 139, pp. 1–34.
Lambe TW (1973) Predictions in soil engineering. Géotechnique 23(2):
engineers in practice and the well-known and widely used M–
151–202, https://doi.org/10.1680/geot.1973.23.2.151.
C soil model. However, to create an adequately representative Maniscalco JD and Ieronymaki ES (2018) Performance and analysis of a
numerical model that will produce accurate predictions it is braced secant pile wall for a multistory building in Manhattan,
important to understand the physical effect of each component NY. In Proceedings of DFI-EFFC International Conference on
of the system, avoid over-simplifying the problem and accu- Deep Foundations and Ground Improvement, Rome, Italy. Deep
Foundation Institute, Hawthorne, NJ, USA, article no. 3013.
rately model the involved structural systems in two dimensions.
Mitchell JK and Gardner WS (1975) In situ measurement of volume
change characteristics. In Proceedings of the ASCE Geotechnical
Acknowledgements Specialty Conference on In Situ Measurement of Soil Properties.
The authors would like to thank Underpinning and ASCE, Reston, VA, USA, p. 279–345.
Foundation Skanska and Laquila Group, Inc. for providing Orazalin ZY, Whittle AJ and Olsen MB (2015) Three-dimensional
analyses of excavation support system for the Stata Center
construction information and Langan Engineering for provid-
basement on the MIT campus. Journal of Geotechnical and
ing soil investigation data. Geoenvironmental Engineering 141(7): 05015001.
Peck RB (1969) Advantages and limitations of the observational
method in applied soil mechanics. Géotechnique 19(2): 171–187,
REFERENCES https://doi.org/10.1680/geot.1969.19.2.171.
Bowles LE (1996) Foundation Analysis and Design. McGraw-Hill, Peck RB, Hanson WE and Thornburn TH (1974) Foundation Engineering.
New York, NY, USA. Wiley, New York, NY, USA.
Brinkgreve RB (2005) Selection of soil models and parameters for Pestana JM and Whittle AJ (1999) Formulation of a unified constitutive
geotechnical engineering application. In Soil Constitutive Models: model for clays and sands. International Journal for Numerical and
Evaluation, Selection, and Calibration (Yamamuro JA and Kaliakin Analytical Methods in Geomechanics 23(12): 1215–1243.
VN (eds). ASCE, Reston, VA, USA, GSP 128, pp. 69–98. Pestana JM, Nikolinakou M and Whittle AJ (2005) Selection of material
Brinkgreve RBJ, Broere W and Waterman D (2007) Plaxis 2D, V8. Delft parameters for sands using the MIT-S1 model. In Soil Constitutive
University of Technology & Plaxis BV, Delft, the Netherlands. Models: Evaluation, Selection, and Calibration (Yamamuro JA and
Calvello M and Finno RJ (2004) Selecting parameters to optimize in Kaliakin VN (eds). ASCE, Reston, VA, USA, GSP 128,
model calibration by inverse analysis. Computers and Geotechnics pp. 425–439.
31(5): 410–424. Poeter EP and Hill MC (1997) Inverse models: a necessary next step in
Corral G and Whittle AJ (2010) Re-analysis of deep excavation collapse ground-water modeling. Groundwater 35(2): 250–260.
using a generalized effective stress soil model. In Proceedings of Schanz T, Vermeer PA and Bonnier PG (1999) The hardening soil model:
Earth Retention Conference. American Society of Civil Engineers, formulation and verification. In Beyond 2000 in Computational
Reston, VA, USA, vol. 3, pp. 720–731. Geotechnics (Brinkgreve RB (ed.)). Balkema, Rotterdam, the
Corral Jofré GA (2013) Methodology for Updating Numerical Netherlands. pp. 281–296.
Predictions of Excavation Performance. Doctoral dissertation, Whittle AJ (1993) Evaluation of a constitutive model for
Massachusetts Institute of Technology, Cambridge, MA, USA. overconsolidated clays. Géotechnique 43(2): 289–313, https://doi.
Duc TN, Vo P and Thi TT (2019) Determination of unloading–reloading org/10.1680/geot.1993.43.2.289.
modulus and exponent parameters (m) for hardening soil model of Whittle AJ, Corral G, Jen LC and Rawnsley RP (2015) Prediction and
soft soil in Ho Chi Minh city. In Proceedings of International performance of deep excavations for courthouse station, Boston.
Conference on Sustainable Civil Engineering and Architecture Journal of Geotechnical and Geoenvironmental Engineering
(Reddy JN, Wang CM, Luong VH and Le AT (eds)), Springer 141(4): 23.
Nature, Singapore, pp. 677–689 Yuan Y and Whittle AJ (2018) A novel elasto-viscoplastic formulation
Hara A, Ohta T, Niwa M, Tanaka S and Banno T (1974) Shear modulus for compression behaviour of clays. Géotechnique 68(12): 1044–1055.
and shear strength of cohesive soils. Soils and Foundations 14(3): Zhang XP, Wong LNY, Wang SJ and Han GY (2011) Engineering properties
1–12. of quartz mica schist. Engineering Geology 121(3): 135–149.
Hashash YM and Whittle AJ (1996) Ground movement prediction for
deep excavations in soft clay. Journal of Geotechnical Engineering
122(6): 474–486.
Hatanaka M and Uchida A (1996) Empirical correlation between
penetration resistance and internal friction angle of sandy soils. How can you contribute?
Soils and Foundations 36(4): 1–9.
Hough BK (1969) Basic Soils Engineering, 2nd edn. Ronald Press, To discuss this paper, please email up to 500 words to the
New York, NY, USA. editor at journals@ice.org.uk. Your contribution will be
Ieronymaki ES, Whittle AJ and Sureda DS (2017) Interpretation of forwarded to the author(s) for a reply and, if considered
free-field ground movements caused by mechanized tunnel
appropriate by the editorial board, it will be published as
construction. Journal of Geotechnical and Geoenvironmental
Engineering 143(4): 04016114. discussion in a future issue of the journal.
Jaky J (1944) The coefficient of earth pressure at rest. Journal of the Proceedings journals rely entirely on contributions from the
Society of Hungarian Architects and Engineers 78(22): 355–358. civil engineering profession (and allied disciplines).
Khoiri M and Ou CY (2013) Evaluation of deformation parameter for
deep excavation in sand through case histories. Computers and
Information about how to submit your paper online
Geotechnics 47: 57–67. is available at www.icevirtuallibrary.com/page/authors,
Lade PV (2005) Overview of constitutive models for soils. In where you will also find detailed author guidelines.
GeoFrontiers 2005: Calibration of Constitutive Models

460
Downloaded by [ University of Hong Kong] on [06/07/23]. Copyright © ICE Publishing, all rights reserved.

You might also like