Title

Plaxis Bulletin
Issue 35 / Spring 2014

Ed
Comparison of Structural Elements Response in PLAXIS 3D and SAP2000

Reliability of Quay Walls Using Finite Element Analysis

3D Finite Element Analysis of a Complex Excavation

Table of contents
The Plaxis Bulletin is the combined
magazine of Plaxis bv and the Plaxis users
association (NL). The bulletin focuses on the use
03
04
Editorial
New Developments

Page 4
of the finite element method in geotechnical
engineering practise and includes articles on the
practical application of the PLAXIS programs, 05 PLAXIS Expert Services:
case studies and backgrounds on the models 3D Modelling of a Building
implemented in PLAXIS.
Subjected to Earthquake
The bulletin offers a platform where users of
PLAXIS can share ideas and experiences with
Loading
each other. The editors welcome submission of
papers for the Plaxis bulletin that fall in any of 06 Comparison of Structural
Page 6

these categories.
Elements Response in PLAXIS
The manuscript should preferably be submitted 3D and SAP2000
in an electronic format, formatted as plain text
without formatting. It should include the title
of the paper, the name(s) of the authors and 12 Reliability of Quay Walls
contact information (preferably e-mail) for the
corresponding author(s). The main body of
Using Finite Element Analysis
the article should be divided into appropriate
sections and, if necessary, subsections. If any 16 3D Finite Element Analysis of
Page 12

references are used, they should be listed at the

end of the article. a Complex Excavation
The author should ensure that the article is
written clearly for ease of reading.
22 Recent Activities
In case figures are used in the text, it should 24 Upcoming Events
be indicated where they should be placed
approximately in the text. The figures themselves
have to be supplied separately from the text in
Page 16

a vector based format (eps,ai). If photographs

or scanned figures are used the author should
ensure that they have a resolution of at least 300
dpi or a minimum of 3 mega pixels. The use of
colour in figures and photographs is encouraged,
as the Plaxis Bulletin is printed in full-colour.
Page 22

Colophon
The Plaxis Bulletin is a publication of Any correspondence regarding the Plaxis For information about PLAXIS software contact
Plaxis bv and is distributed worldwide among Bulletin can be sent by e-mail to: your local agent or Plaxis main office:
Plaxis subscribers
bulletin@plaxis.nl Plaxis bv
Editorial board: P.O. Box 572
Erwin Beernink or by regular mail to: 2600 AN Delft
Ronald Brinkgreve The Netherlands
Martin de Kant Plaxis Bulletin
Arny Lengkeek c/o Annelies Vogelezang info@plaxis.nl
PO Box 572 www.plaxis.nl
Design: 2600 AN Delft Tel: +31 (0)15 251 7720
Jori van den Munckhof The Netherlands Fax: +31 (0)15 257 3107

Cover photo courtesy of CUR 211 (2013). Quay Walls, 2nd edition. Gouda
Editorial
20 years of Plaxis bv has shown countless
new features and developments throughout
the years. This has also been the case with the
Plaxis Bulletin. Again we have had some changes,
namely in the editing staff. We would like to firstly
thank Wout Broere for al his years of hard work and
help with the Plaxis Bulletin. We are also happy
to introduce the newest editor for this magazine,
Martin de Kant. Welcome and good luck!
The begining of 2014 was also the begining of
quite some new developments at Plaxis. We
released the completely restyled PLAXIS 2D AE.
Furthermore we are expanding our offices and are
happy to announce the opening of the new branch
office in Houston TX, U.S.A., Plaxis Americas LLC.
You can read all about these and other news in the
recent activities column.
Furthermore, in this 35th issue of the Plaxis
bulletin, we have again tried to compile interesting
articles and useful information for you. In the
New Developments column we will discuss some
different developed features for applications in
rock in PLAXIS software over the last years, with
a specific focus on the recent release of the new
Swelling Rock Model.
The first users article discusses the analysis of
the response of a number of structural models
subjected to different loading conditions. The
goal of such a comparison is the assessment of
the structural elements performance in PLAXIS
3D as compared to that obtained by the well-
known SAP2000, a widely used code for structural
analysis. An overall good match was obtained,
as such highlighting the possibility to use the
code PLAXIS 3D to perform both structural
and geotechnical calculations in soil-structure
interaction problems.
The second users article takes a more detailed
look into the combination of FEM analysis in
combination with the Eurocode. Two quay walls
were examined to check the applicability of
the existing FEM design method of the Dutch
Handbook Sheet Pile structures (CUR 166) on quay
walls with relieving floor. Differences in partial
safety factors are proposed to reach the required
reliability index.
The third users article describes the development
of a comprehensive three-dimensional finite
element model for the Stata Center basement
excavation (Cambridge, USA) using PLAXIS 3D
2012. The analyses highlight the effects of the
3D excavation and support geometry on wall
deflections and show a good agreement with
the measured response assuming undrained
conditions using the Mohr-Coulomb soil model.
In addition there is a joint presentation about a
project where Deltares and Plaxis worked together
to set up a 3D non linear dynamic model of a
building interacting with the subsoil through its
pile foundation, for the assessment of the impact
of man-induced earthquakes on infrastructures.
Plaxis provided assistance in setting-up the 3D
finite element model to evaluate the buildings
seismic performance under possible moderate
seismic activity.
We wish you an enjoyable reading experience and
look forward to receiving your comments on this
spring 2014 issue of the Plaxis bulletin.
The Editors
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 3
New Developments
Ronald Brinkgreve, Plaxis bv

Starting as a finite element software package for geo-engineering applications in soft soil, PLAXIS has meanwhile evolved
to cover most geo-materials, ranging from soft soil to rock. This makes PLAXIS a convenient tool to analyse not only
embankments, foundations and excavations at shallow depths, but also applications in the deeper underground, like deep
tunnels and underground openings in hard soils and soft rocks as well as deep mining applications.

Specific features that have been developed

in the last decade include:
Hoek-Brown isotropic material model for
weathered rock
Jointed Rock anisotropic material model, also
available with Mohr-Coulomb failure criterion
Swelling Rock model for time-dependent
anisotropic swelling of clay- and siltstones
Embedded piles and embedded pile rows,
useable as rock bolts and forepooling elements
Biots pore pressure coefficient, for a proper
effective stress calculation considering
compressible solid material at high pressure
New tunnel designer and parametric geometry
definition
The latter two features will become available in the
upcoming PLAXIS 3D 2014 release.
In the remainder of this article I would like to
focus on the Swelling Rock model. This model
is available as a user-defined model that can be
used to simulate the time-dependent anisotropic
swelling of anhydrite and argillaceous rocks.
The model was originally implemented by Prof.
Thomas Benz of NTNU and further developed
by Dr. Bert Schdlich of TUGraz for PLAXIS. The
model is based on previous work by Wittke-
Gattermann & Wittke (2004), Anagnostou (1993)
and Heidkamp & Katz (2002). It accounts for
the stress- and time dependency of swelling
deformations. The following features are included:
Transverse isotropic elastic stiffness
Elastic-perfectly plastic material behaviour with
Mohr-Coulomb failure criterion.
Three swelling laws, all going back to the idea
formulated by Grob (1972): The Wittke model,
the Anagnostou model, and a mixed variant.
The swelling parameters kq, sq0 and hq can be
obtained from a so-called Huder-Amberg test
(Huder & Amberg, 1970). Details of the model
References:
Anagnostou, G. (1993). A model for swelling
rock in tunnelling. Rock Mechanics and Rock
and its parameters are described in a document Engineering 26 (4), 307-331.
that is available for PLAXIS users. The model and Grob, H. (1972). Schwelldruck im Belchentunnel.
the document are free of additional charges for Proc. Int. Symp. fr Untertagebau, Luzern, 99-119.
VIP users. Interested users may contact our sales Heidkamp, H. & Katz, C. (2002). Soils with
department. The model has been validated and swelling potential - Proposal of a final state
applied in a case study of the Pfaendertunnel formulation within an implicit integration scheme
near Bregrenz, Austria (Schdlich et al, 2013). and illustrative FE-calculations. Proc. of the 5th
The results of the case study turned out to give World Congress on Comp. Mec., Vienna, Austria.
a good match with the measured displacement Huder, J. & Amberg, G. (1970). Quellung
profile below the tunnel invert. The model can also in Mergel, Opalinuston und Anhydrit.
be used to evaluate the increase in tunnel lining Schweizerische Bauzeitung 43, 975980.
forces as a result of swelling rock and to analyse Schdlich B., Schweiger H.F., Marcher T. (2013).
appropriate mitigating measures. Numerical Analysis of Swelling Deformations in
Tunnelling A Case Study. Proc. EURO:TUN 2013,
With the Swelling Rock model and the other 429-436.
features as mentioned above, we trust that you Wittke-Gattermann, P. & Wittke, M. (2004)
will confidently use PLAXIS in deep underground Computation of Strains and Pressures for Tunnels
applications. We welcome your feedback on these in Swelling Rocks. Proc. ITA 2004 E14, 1-9.
features as well as your article on such applications
in future Plaxis Bulletins.
Grobs swelling law (left) and the influence of the swelling parameter hq on the
time-dependency of the swelling process for a maximum swelling strain of 2% (right)

4 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl

PLAXIS Expert Services: 3D Modelling of a
Building Subjected to Earthquake Loading
Jaap Bijnagte and Mandy Korff, Senior Geotechnical Engineers, Deltares, The Netherlands

For the assessment of the impact of man-induced earthquakes on infrastructures, Deltares and Plaxis worked together to
set up a 3D non linear dynamic model of a building interacting with the subsoil through its pile foundation. Plaxis provided
assistance in setting-up the 3D finite element model to evaluate the buildings seismic performance under possible moderate
seismic activity. The main challenge of this project was the rather short time frame within which such FE analyses needed to be
carried out. Thanks to PLAXIS Expert Services, Deltares managed to deliver FEA results in an efficient and timely manner.

The results of this study were used by Deltares to

evaluate the influence of the soil and foundation
stiffness on the transfer of the earthquake signal
from the subsoil into the structural elements of the
building. The impact of the presence of the
building was evaluated, as well as the structural
integrity of the pile and the connection to the
structure. These results will be applied as a starting
point for the development of guidelines for the
vulnerability assessment of piled buildings in typical

soft soil conditions under earthquake loading.

The dimension of the FE model is 60m long,
30m wide and 30m deep for a rather simple
5m*5.5m building resting on four 15 m long piles.
Only half of the geometry has been modelled due
to symmetry conditions. The model contains a
total of roughly 36,800 elements and 55,000 nodes
(i.e 165,000 dof). Each finite element has 10 nodes.
The soil stratigraphy consists of a relatively soft
sand layer up to -1 m, a sandy clay layer from -1 m
to -2 m, a medium dense sand layer from -2 m to
-3.5 m, a clay layer from -3.5 m to -14 m resting of
a rather dense soil layer all the way down to the
model base. For the analyses, different stiffness
values have been adopted for respectively the
static analysis and the dynamic one (the static
analysis aiming at modelling the construction
sequence of the building under drained situation
in order to obtain a realistic state of stresses
before running the dynamic analysis for the
earthquake loading and for which considerably
larger material stiffnesses need to be taken into
account).
PLAXIS Expert Services added value:
The building has been entirely modelled by Quick start on the job
means of plate elements (walls, floor, ceiling and Set-up of fully optimized & ready to run models
supporting beams each with different properties). Regular model review
The piles have been modelled has volume Next business-day advanced technical assistance
elements surrounded by interface elements for
optimum soil-structure interaction modelling. About Deltares
Along their neutral axis, very soft beams have been Deltares is an independent institute for applied
introduced for the purpose of easing results post- research in the field of water, subsurface and
processing in terms of structural forces. infrastructure. Throughout the world, Deltares works
Plaxis experts helped us to quickly set up the model and optimize it to run
the complex analyses very efficiently. Having his first-hand experience present
in house while discussing the best approach supported our vulnerability
assessment team to deliver these state of the art results to our client in time.
The input motion applied at the model base has on smart solutions, innovations and applications for
been derived from a so-called deconvolution people, environment and society focussing mainly
analysis carried out using a companion PLAXIS 1D on deltas, coastal regions and river basins. Managing
model for which a target peak ground acceleration these densely populated and vulnerable areas is
along with relevant frequency ranges were already complex, which is why Deltares works closely with
provided and used to generate the input motion governments, businesses, other research institutes
at the model base under free-field assumption. and universities at home and abroad.

www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 5

Comparison of Structural Elements Response in
PLAXIS 3D and SAP2000
Gragnano C. G., Fargnoli V., Boldini D. (Corresponding Author), University of Bologna, Italy
Amorosi A., Technical University of Bari, Italy

In this paper the response of a number of structural models subjected to different loading conditions is analysed with the
codes PLAXIS 3D and SAP2000. The goal of such a comparison is the assessment of the structural elements performance in
PLAXIS 3D as compared to that obtained by the well-known SAP2000, a widely used code for structural analysis. An overall
good match was obtained, as such highlighting the possibility to use the code PLAXIS 3D to perform both structural and
geotechnical calculations in soil-structure interaction problems.

The impulse in software technology and

computational power of personal computers
has recently offered the possibility to perform
Modelling a Spatial Frame with Beams and
Columns: Model M1
The reference structure of the case study
of 30 cm x 30 cm (Fig. 1). The figure illustrates the
dimension of the structural elements, the right-
handed global reference system (x, y, z) and the
fully-3D finite element analyses of complex described in this section is a single-bay spatial local coordinate (s), this latter represented only for
engineering projects. In particular, in the field of frame fixed at the base and consisting only of beam 2-6 for sake of simplicity.
civil engineering it is nowadays feasible to perform beams and columns all characterised by a section
the analysis of a soil-structure interaction problem
by a unique model, accounting at the same time Unit weight (kN/m3) 24
for both geotechnical and structural issues.
Young's modulus E (GPa) 25
The three-dimensional version of the finite
Poisson's ratio (-) 0.2
element code PLAXIS includes a wider choice of
structural elements (such as beams, plates and
Table 1: Material properties of beams and columns
node-to-node anchors), enhancing its modelling
capability at the cost of a deeper structural
Unit weight (kN/m3) 32.36
competence required to the user. This paper
provides a contribution on this specific topic, Young's modulus E (GPa) 10
illustrating a number of structural models, where
the different structural elements were employed, Poisson's ratio (-) 0.2
to investigate and clarify their response under Figure 1: Spatial frame with beams and columns and global
Table 2: Material properties of the isotropic floor slab coordinate system
different loading conditions. These models range
from simple single-bay spatial frame to multi-
storey frame with cross-bracings simulating the
presence of infilled panels. All the models are
assumed fixed at base, i.e. no foundation systems
were considered, in order to focus the attention on
the structural response only.

The observed behaviour was compared with that

obtained analysing the same structure by the finite
element code SAP2000, a widely used software
for structural analysis. This assessment was useful
to highlight some differences in the formulation of
the corresponding structural elements in the two
codes. Figure 2: Three-dimensional view of model M1 under loading conditions C1 (a), C2 (b) and C3 (c)

6 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl


Figure 3: Model M1: response of beam 6-7 under loading
condition C1 in PLAXIS 3D and in SAP2000
In this example, defined model M1, as in the
following ones, beams and columns are modelled
as one-dimensional elements of frame-type in
SAP2000 and beam-type in PLAXIS 3D. This latter
element, differently from the frame type, is not
able to react to torsional actions. Both elements
allow for deflections due to shearing as well as
bending.
Figure 4: Model M1: response of beam 3-7 under loading
condition C2 in PLAXIS 3D and in SAP2000
A linear-elastic constitutive law was adopted for
these elements, whose parameters were selected
consistently with the assumed reinforced concrete
material (Table 1).
All the six displacement components were
restrained at the base of the model in SAP2000. In
an interaction problem, this condition simulates
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin
a rigid contact at the soil-structure interface,
thus being appropriate for the modelling of a
soil-foundation system much stiffer than the
superstructure.
On the contrary, a foundation plinth 1 m high and
characterised by a square section (1 m x 1 m) was
assumed at the base of each column in the PLAXIS
3D analysis, modelled by a two-dimensional
plate element. As this code does not allow to
perform numerical analyses without including
soil elements, a soil volume (12 m x 15 m x 15 m)
was defined at the frame base, assuming for it a
very rigid behaviour, characterised by a Youngs
modulus of 750 GPa and a Poissons ratio equal
to zero.
The response of the model was analysed
considering the following loading conditions:
C1 = gravity loads + uniformly distributed
vertical loads equal to 10 kN/m acting on the
beams (Fig. 2 a);
C2 = gravity loads + concentrated vertical loads
of 50 kN acting at nodes 3 and 6 (Fig. 2 b);
C3 = gravity loads + concentrated horizontal
loads of 50 kN acting at nodes 3 and 6 (Fig. 2 c).
Numerical analyses were carried out using a finite
element mesh of medium density in PLAXIS 3D
(i.e. the average size of the finite element is equal
to 1.3 m), while adopting the default option in
SAP2000.
Distributions of shear, bending moment and
inflection for beams 6-7 (relative to loading
conditions C1 and C3) and 3-7 (for loading
condition C2) as calculated by the two codes are
shown in Figures 3, 4 and 5. This latter figure also
reports the horizontal displacements along x
direction of column 1-2 under loading condition
C3.
It is possible to note that the results calculated by
SAP2000 and PLAXIS 3D are fairly coincident in
terms of shear, bending moment and inflection,
7
Comparison of Structural Elements Response in PLAXIS 3D and SAP2000

while the horizontal displacements evaluated for

column 1-2 differ in a non-negligible way. Such
difference is due to the characteristics of the beam
element in PLAXIS 3D which, as anticipated, does
not sustain the torsional action induced by loading
condition C3 (Fig. 5). This is confirmed by the
results of a further analysis, illustrated in Figure
6, identical to the previous one except for the
torsional constraint at the column head which was
removed in the SAP2000 model: this modification
leads to an almost coincident response as
obtained by the two codes.

Modelling a Floor Slab in a Simple Spatial Frame:

Model M2
Figure 7 shows a single-bay spatial frame differing
from the simple structure of model M1 (Fig. 1)
for the presence of a floor slab at the top. A
brick-reinforced concrete floor slab is a structural
element having a heterogeneous composition
(i.e. reinforced concrete and brick) and a different
stiffness in the two plane directions (i.e. higher
stiffness in the warping direction). It is subjected
to a plane stress condition and it is mainly loaded
in its out-of-plane direction.

Figure 5: Model M1: response of beam 6-7 and column 1-2 under loading condition C3 in PLAXIS 3D and in SAP2000
The numerical model of this structure (model
M2) is coincident to model M1 in terms of beams,
columns and constraint conditions at the base.

Concerning the floor slab, two different

mechanical hypotheses were considered, namely
isotropic and anisotropic. This latter allows to
reproduce the main characteristic of a floor slab,
that is a structural element rigid in its own plane
and capable of differentiating the load transferred
to the main beams as compared to the secondary
ones.
The isotropic behaviour was obtained in PLAXIS
3D using a two-dimensional linear-elastic plate
element of thickness equal to 25 cm with the
material properties listed in Table 2.

A two-dimensional shell element with the same

geometrical and material properties was selected
to model the isotropic floor slab in SAP2000.

The presence a floor slab with anisotropic

behaviour was represented in SAP2000 without
simulating the structural element itself, but just
applying the constraint diaphragm to the nodes 2,
3, 6 and 7 (Fig. 7). This constraint, generally used
to model structural components which have very
high in-plane stiffness, forces the nodes belonging
Figure 6: Model M1: response of beam 6-7 and column 1-2 under loading condition C3 in PLAXIS 3D and in SAP2000 without to the plane of the slab to move together in a rigid
torsional constraints at column heads. way. Assuming the warping direction of the floor
slab along x-axis and according to the current
design practice, the weight of the floor slab was
tw (m) 0.3 accounted for applying vertical forces to the main
beams (in y direction) and to the secondary ones
hw (m) 4 (in x direction) with reference to the influence
areas: a load equal to 64.1 kN and 16.8 kN was
Ew (GPa) 3
attributed to the main and secondary beams,
Ec (GPa) 25 respectively. In particular, the first load is equal
to half of the floor slab weight (80.9 kN, being the
Ic (m4) 0.000675
total weight equal to 161.8 kN), reduced of the load
() 45 (16.8 kN) transferred to the adjacent secondary
beams by a floor slab slice 50 cm wide.
w (1/m) 1.351
When modelling the same slab in PLAXIS 3D, an
dw (m) 5.657
anisotropic elastic model was employed. More
bw (m) 0.504 specifically, according to the warping direction
Figure 7: Spatial frame with beams, columns and a floor
along x-axis, the Youngs modulus, Ey, and the
slab
Table 3: Values of the parameters for estimating the shear modulus, Gyz, were reduced as compared to
equivalent diagonal width, bw those adopted in the isotropic case. The amount

8 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl


Figure 8: Model M2: response of beam 3-7 under loading
condition C1 in PLAXIS 3D and in SAP2000
Figure 9: Model M2: response of beam 3-7 under loading
condition C2 in PLAXIS 3D and in SAP2000
Comparison of Structural Elements Response in PLAXIS 3D and SAP2000
of the necessary reduction of the moduli to match
the reference results obtained by SAP2000 is equal
to 10%, as such the adopted parameters are Ey = 1
GPa; Gyz= 416.7 MPa.
The same loading conditions previously analysed
for model M1were considered, namely C1 (taking
also into account the floor slab weight), C2 and C3.
The finite element mesh used for this model in
PLAXIS 3D is similar to that defined in model
M1; in SAP2000, on the contrary, the mesh of the
model with isotropic slab was modified to make
it roughly equivalent to that defined in PLAXIS
3D. This expedient is related to the fact that in
SAP2000 the load of the floor slab is transferred to
the beams in correspondence of the mesh nodes,
therefore a similar finite element discretisation is
required in order to obtain consistent results by
the two different codes.
Figures 8, 9 and 10 show the comparison between
models M1 and M2 in terms of shear, bending
moment and inflection for beam 3-7 under loading
conditions C1, C2 and C3, respectively. Figure 10
also shows the horizontal displacements of column
1-2 along x-axis.
Results demonstrate the good agreement
between the structural responses obtained by
the two different numerical codes. In general, it
is possible to observe an equivalent response
of beam 3-7 under loading conditions C1 and
C2 for model M2 too. As expected, the different
assumption concerning the behaviour of the floor
slab (i.e. isotropic or anisotropic) plays an essential
role in the intensity and distribution of shear,
bending moment and inflection.
In the anisotropic case, the structural element 3-7
is one of two main beams of the floor slab and it
results to be more heavily loaded as compared to
what observed in the isotropic model, where all
the beams were equally loaded per unit of length.
On the contrary, the different mechanical
hypotheses seem to have a barely relevant
influence on the horizontal displacement of the
column: this should be due to the fact that in both
isotropic and anisotropic cases the relevant shear
stiffness Gxy assumes the same value, leading
to a similar head restrain acting on the column,
therefore resulting in a correspondingly similar
displacement pattern.
Modelling a 2D-Frame with Diagonal Elements:
Model M3
The simple structure shown in Figure 11 is a
single-bay plane frame with cross bracings.
These elements are commonly adopted in
numerical studies to account for infill panels (e.g.:
Panagiotakos and Fardis, 1996). Those latter,
although being non-structural components,
significantly contribute to the overall structural
response in the in-plane horizontal direction,
leading to a generally stiffer behaviour as
compared to open-frame ones.
In the corresponding numerical model, defined as
model M3, the structural elements (i.e. beam and
columns) are represented by frames and beams
in SAP2000 and PLAXIS 3D, respectively, and are
characterised by the material properties listed in
Table 1. The base of the frame is constrained as in
all the other models.
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin
The diagonal elements of the frame were
modelled in order to make them equivalent
to a building infill panel, adopting a simplified
version of a formulation proposed in the literature
(Panagiotakos and Fardis, 1996; Fardis, 1997). The
width of the cross bracings, bw, was defined with
reference to the expression of Mainstone (1971):
bw = 0.175 ( h hw ) 0.4 dw (1)
where: d w is the diagonal length of the panel, hw is
the panel height and the parameter h is equal to:
Ew t w sin( 2 )
h = 4
(2)
4 Ec Ic hw
where Ew and Ec are the Youngs moduli of the infill
panel and of the reinforced concrete structural
elements surrounding the panel, respectively;
is the angle formed by the diagonal of the infill
panel with respect to the horizontal axis; t w is the
panel thickness; Ic is the moment of inertia of the
columns adjacent to the infill panel. The values of
these parameters are summarised in Table 3.
The cross bracings were modelled as weightless
one-dimensional elements reacting only to axial
stress (denoted as truss elements in SAP2000 and
node-to-node anchor elements in PLAXIS 3D),
characterised by an axial stiffness equal to
K = Ew * bw * t w = 450000 kN. An elastic-plastic
constitutive law was selected for the elements to
introduce a limit value of the tensile strength equal
to zero, aimed at neglecting tensile stresses for
the cross bracings.
The response of model M3 was assessed by
considering the structural elements weight (beams
and columns) and a force of 20 kN applied at
node 2 along x-axis (loading condition C4). Figure
12 shows a perfect match among the results of
the two models in terms of normal stress acting
in column 3-4 and diagonal element 2-4; shear,
bending moment and inflection in beam 2-3;
horizontal displacement in column 3-4.
Modelling a Spatial 3-Storey Frame with and
without Cross Bracings: Models M4(I) and M4(II)
In this section the responses of two 3-storey
frame structures subjected to horizontal loads
are compared, the structures differing only for
the presence of cross bracings (Fig. 13). The inter-
storey height is 4 m and the beams length is equal
to 4 m in x direction and 5 m in y direction.
The numerical models of the open-frame structure
and that of the structure with diagonal elements
are denoted as M4(I) and M4(II). In the models
beams and columns are represented by one-
dimensional elements (frames and beams in the
two codes) and, for sake of simplicity, the floor
slabs are modelled as linear-elastic-isotropic
elements of shell-type in SAP2000 and plate-type
in PLAXIS 3D. For both models the mechanical
properties of columns, beams and floor slabs
are those listed in Tables 1 and 2; the usual rigid
constraint conditions are assumed at the base of
the frames.
The equivalent width dw of the cross bracings,
modelled as node-to-node anchor and truss
elements in PLAXIS 3D and SAP2000 respectively,
was defined using Eq. (1) and the same elastic-
plastic constitutive law assumed for model M3 was
selected in this case.
Both models were analysed under gravity loading
9
Comparison of Structural Elements Response in PLAXIS 3D and SAP2000

and horizontal ones acting along x-axis, those

latter equal to 20 kN, 40 kN and 60 kN at the
first, second and third frame level respectively
(loading condition C5) (Fig. 13). A control point
position was selected at the top level (node 3.4) as
representative of the horizontal displacement of
the structure.

The horizontal displacement distributions in

columns 0.4-1.4, 1.4-2.4, 2.4-3.4 are reported in
Figure 14 for the two models. It is worth noting
that both codes provide the same results: the
maximum horizontal displacement is equal to 8 cm
for model M4(I) and about 0.8 cm for model M4(II).
The outcome of the analyses clearly highlights
the effect of claddings on the overall structural
stiffness, although simply accounted for by means
of equivalent diagonal elements: in fact, the
presence of cross bracings produces a horizontal
displacement reduction of an order of magnitude
as compared to the reference case where they are
not included.

Conclusions
In the paper the response of a number of
structural models subjected to different loading
conditions was analysed by the finite element
codes PLAXIS 3D and SAP2000. The main Figure 10: Model M2: response of beam 3-7 and column 1-2 under loading condition C3 in PLAXIS 3D and in SAP2000
outcomes resulting from the comparison, carried
out in terms of stress and displacements, can be
summarised as follows:
beams and columns can be modelled with
frame elements in SAP2000 and beam elements
in PLAXIS 3D. The main difference in the ele-
ment formulations resides in the inability of
beam elements to react to torsional actions.
In fact, the release of torsional constrains in
SAP2000 produces perfectly matching results;
the floor slab can be modelled in SAP2000 by a
shell element or using a diaphragm constraint
combined with some additional vertical forces
at the top of the columns to simulate the effect
of the slab weight. In the first case an isotropic
behaviour is obtained, while in the latter a more
realistic response is reproduced, as it allows to
account for the higher stiffness observed in the
warping direction. A plate element is instead
available in PLAXIS 3D. The use of an isotropic
formulation allows to nicely reproduce the
response of the shell element, while an aniso- Figure 11: 2D frame with cross bracings
tropic model should be selected to fit, after
a careful calibration of its elastic parameters,
the response of the more advanced scheme of
SAP2000;
infill panels can be modelled in a simplified
manner as cross bracings, whose characteristics
were obtained using the formulation proposed References
by Mainstone (1971). Truss and node-to-node 1. M.N. Fardis, 1997. Experimental and numeri-
anchor elements were used respectively in cal investigations on the seismic response of
SAP2000 and PLAXIS 3D, leading to perfectly RC infilled frames and recommendations for
consistent structural responses. code provisions. Report ECOEST-PREC8 No. 6.
Prenormative research in support of Eurocode
This study should be considered as a preliminary 8.
step towards more complex soil-structure 2. R.J. Mainstone, 1971. On the stiffnesses and
interaction problems, which indeed require a strengths of infilles frames. Proc. Inst. Civil.
good level of confidence in the use of structural Engineers, iv 7360s: 59-70.
elements in 3D analyses with PLAXIS. 3. T.B. Panagiotakos and M.N. Fardis, 1996.
Seismic response of infilled RC frames struc-
Acknowledgements tures. 11th World Conference on Earthquake
Special thanks to Ph.D. Eng. Francesco Tucci for Engineering, Acapulco, Mxico, June 23-28.
his helpful support during this research activity. Paper No. 225.

10 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl

 Comparison of Structural Elements Response in PLAXIS 3D and SAP2000

Figure 12: Model M3: responses of column 3-4, beam 2-3, and diagonal element 2-4 under C4 load condition in PLAXIS 3D and
in SAP2000

Figure 13: Three-dimensional view of the structures and

loading distributions with (top) and without (bottom) cross
bracings. Each node of the frame is defined through a
double number: the first indicates the level it belongs to,
while the second is a sequential number.

Figure 14: Models M4(I) and M4(II): comparison between horizontal displacements obtained in PLAXIS 3D and in SAP2000 with
(on the right) and without (on the left) cross bracings.

www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 11

Reliability of Quay Walls Using Finite Element Analysis

Authors: H.J. Wolters, K.J. Bakker and J.G. de Gijt, Delft University of Technology. The Netherlands

During the last years the Finite Element Method (FEM) is increasingly applied in the design of quay walls. Especially in
case of quay walls with relieving floors and bulk-storage as surcharge load, sub-grade reaction models are limited in their
accuracy of modelling the situation. The Finite Element Method often is the only option to more detailed design calculations
of quay walls. In the recent years the introduction of Eurocode and the increasing use of Finite Element analysis for design
calculations has triggered the update of the CUR Quay walls handbook CUR 211. The latter second edition has recently been
published. In advance of this second edition it was decided to look into more detail into the combination of FEM analysis in
combination with the Eurocode which lead to the study that is described in this article. In order to infer a more fundamental
base for the design method with FEM, two quay walls were examined to check the applicability of the existing FEM design
method of the Dutch Handbook Sheet Pile structures (CUR 166) on quay walls with relieving floor. Furthermore, it was checked
whether the current partial safety factors needed to be adapted. This research is done by performing probabilistic FEM
calculations. The First Order Reliability Method is incorporated in the software Prob2B (Courage & Steenbergen, 2007) to
perform the calculations. It appeared that using the design method of CUR 166 for quay walls with relieving floor leads to an
underestimation of the reliability of the structure. Therefore it is advised to adapt the design method. Furthermore, differences
in partial safety factors are proposed to reach the required reliability index.

During the last two years, CUR committee

183 has worked on the upgrade of the
Dutch Quay Walls handbook (CUR 211), which was
published in November 2013. Two of the main
elements that are considered in this new edition
are the addition of FEM analysis as a method
for design, comparable to the description in
the Dutch Handbook Sheet Pile Structures (CUR
166), and the calibration of partial safety factors
design with FEM. With respect to the actuality
of this update it must be remembered that with
the new 2nd Maasvlakte and other changes in the
Rotterdam harbour area, several quay walls are
under construction, such as illustrated in Fig. 1,
which configuration was a reference site for the
CUR analyses.
One of the arguments for the further introduction
of FEM analysis for quay wall design is that for the
larger quay walls, relieve platforms are often used
under heavy loaded conditions. In that situation
one must consider that the piles under the
relieve platform may bear a part of the horizontal
load that is normally taken by the retaining wall.
Furthermore, if the load itself is bulk-storage
there are limitations to the accuracy of modelling
surcharge load with sub-grade reaction models
that are normally applied for soil retaining wall
design; horizontal components of bulk storage
may be difficult to model; see Fig. 2.
The influence of relieve platforms and high
surcharge loads on greenfield is one of the reasons
to evaluate the design procedures in CUR 166 in
the combination with Finite Element analysis, for
different failure mechanisms. The method applied
in this study is to put Finite Element analysis in a
framework of probabilistic analysis. Within this
framework, partial safety factors are inferred from
the reliability indices and the influence coefficients
of the variables of the FEM model, obtained with
FORM analysis. This paper presents the method Fig. 1: A quay wall under construction at Maasvlakte 2

12 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl

 Photo courtesy of CUR 211 (2013). Quay Walls, 2nd edition. Gouda

Fig. 2: Horizontal shear forces due to bulk-storage, that will increase the anchor force, are
difficult to model with sub-grade reaction model Fig. 3: Quay wall with relieve platform, modelled in PLAXIS

that is used to obtain the reliability indices and
partial safety factors and compares the results
from these calculations with the 2003 edition of
CUR 211.
Quay Walls and Failure Mechanism
The research is done for two different types of
quay wall, modelled in Plaxis. To begin with, an
anchored sheet-pile with two different sheet-pile
lengths (21m and 23m) was analysed. The sheet-
pile is anchored 2 m below ground level (ground
level is NAP) and has an AZ36-700N profile. The
wall is excavated till NAP -12 m and a surcharge
load of 30 kN/m2 is present. The upper sand layer
reaches till NAP -10 m. Below there is a clay layer
till NAP -15 m followed by another sand layer. The
soil and structural parameters can be found in
the report of Wolters (2012, pp. 83-84). This first
configuration was used to check the method.
Secondly, a heavier quay wall with relieving
floor was modelled; see Fig. 3. In this article
the main focus is on this second analysis. The
model is based on the quay wall from Fig. 1. The
quay wall is anchored by a double anchor that
must guarantee a top displacement of less than
50 mm. The combi-wall consists of tubular piles
with 1420 mm diameter and 18 mm wall thickness.
In between the piles there are three sheet-piles
with profile AU20. The wall is excavated till NAP -19
m which implies 24 m retaining height. A bollard
force of 70 kN/m and a surcharge load of 40 kN/m2
behind the quay wall are taken into account. The
level of the top of the quay wall is NAP +5,0 m.
The soil configuration is based on Maasvlakte
conditions. The upper sand layer reaches till
NAP -8,50 m. Below there is a clay layer till
NAP -11,0 m, a sand layer till NAP -19,0 m, another
clay layer till NAP -22,0 m and Pleistocene sand.
The soil and structural parameters can be found in
the report of Wolters (2012, pp. 137-138). For each failure mechanism a reliability function is
The PLAXIS Hardening Soil model was used to
model the soil, because this enables a better
description of the unloading behaviour of the soil
behind the wall and gives a description of the
soil deformations under the relieving platform.
In contrast to the normal design procedure, here
for the probabilistic analysis, mean values of the
parameters where used. Normally, characteristic
values would need to be applied, according to
the Eurocode. The difference between mean
values and characteristic values was discounted
for afterwards when partial safety factors where
derived.
As a starting point for the analysis the FEM
procedures, as described in CUR 166, were
taken. The analysed failure mechanisms of these
structures are anchor failure in tension (ULS), wall
failure in bending (ULS), soil mechanical failure
(ULS) and excessive deformations (SLS).
defined of the form:
Z = Resistance(R) Solicitation (S),
which implies that failure is assumed for Z<0. The

www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 13

Reliability of Quay Walls Using Finite Element Analysis

output that is searched for are the reliability index

, linked to the probability of failure (pf = (-)) Number of calculations (FORM): 78
and the influence coefficients per variable (i).
: 2.646
As tool for the probabilistic analysis the PLAXIS
kernel was linked in with software (i.e. Prob2B Pf: 4.068*10-3
(Courage & Steenbergen, 2007)), which controlled
the variations that are necessary to derive the Parameter (X) V=/ X* (design point) Unit
influence coefficients ai.
Gsilty moderately packed sand 0.3 -0.12 43310 [kPa]
First Order Reliability Method (FORM) & Prob2B
The used probabilistic method FORM is an Gpleistocene sand 0.3 0.48 106400 [kPa]
iterative level II approach, entailing linearization in sin()silty moderately packed sand 0.18 0.67 0.48 [-]
the design point to derive influence coefficients.
Using this method the uncertainty in the output of sin()pleistocene sand 0.18 0.46 0.38 [-]
the analysis can be objectively weighted with the
sat,silty moderately packed sand 0.05 -0.30 20.15 [kN/m3]
uncertainty in the input parameters discounting
these objectively with the variation coefficient calc. Z-value
of a parameter that is uncertain. The variation
coefficient V i being defined as: 1 170400

Vi = i 78 8110
i
The approach that is applied in Prob2B can be Table 1: Example of the output of a FORM calculation
described in different steps:
The calculation starts with a normal PLAXIS
calculation using mean parameter values.
The parameter values are transformed into
standard-normal u-space by the transformation:
x i
ui = i
x ,i
For each stochastic variable an additional
calculation is made by increasing the variable
by a fraction of the standard deviation. This
variation is made in order to check the variables
influence on the Z-value. This influence is
expressed in the derivative dZ/dx, in which x is a
parameter value.
This derivative is important in the FORM
procedure to determine the standard deviation
and mean value of Z, which define the reliability
index:
dZ
ui
z dui
= =
z dZ
2

dx
i
From the derivatives for each variable an
influence coefficient, i, can be determined by:
dZ
dui
i = 2 The sum of i2 is equal to 1.
dZ
dxi
For each parameter a possible design point
value can be determined by:
X i = i i x ,i
This procedure is repeated by iterating with
the new obtained parameter values instead of
the initial mean values. After some iterations,
convergence may be found and the design
point and corresponding and i can be
determined. For the failure mechanisms the Z
function should approach zero.
This FORM procedure is incorporated in Prob2B.
However, not all stochastic parameters can be
included. Variations in surcharge loads, water
levels and bottom level (dredging depth) need to
be made manually. With these manual variations
dZ/dx is calculated and from there the uncertainty
in those parameters can be incorporated in the
reliability index and influence factors can be
derived.
Fig. 4: Example of a plastic point plot in the design point of wall failure in bending (Red: Mohr-Coulomb point, Green: Hardening
point, Blue: Cap and Hardening point)
From the output of Prob2B partial safety factors
can be derived using the formulas:
by the software. Table 1 is an example of the
1 + 1.64Vi
k ,S = (load) and
1 i Vi
1 1.64Vi
k ,R = (resistance),
1 i Vi
The coefficients of variation for the soil variables
and their correlations with other variables are
derived from a database of Rotterdam Public
Works, containing 3000 tri-axial tests and added
by information from the Dutch national annex to
NEN-EN 1997 (former NEN 6740). For the structural
variables the prescriptions of the Joint Committee
on Structural Safety (2002) are followed.
Output
For each FORM calculation output is generated
output for the limit state function of wall failure
in bending. The table includes the variation
coefficients (input) the influence factors, design
point values and the failure probability (reliability
index). When implementing the design values
a plastic point plot can visualize the failure
mechanism.
An example of such a plot is given in Fig. 4. In this
plot it can be seen that wall failure occurs due to
failure of the soil elements at the lower part of
the piles. This soil failure leads to a reduced fixing
moment which causes an increase in the maximum

14 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl

 Reliability of Quay Walls Using Finite Element Analysis
relatively high prescribed target reliability in CUR
calculated CUR 211 211 (Wolters, 2012 pp. 170-174).
Anchor failure 4.09 3.828
For the quay wall with relieving floor the
Wall failure 3.96 3.872 differences are shown in Table 3. The target
reliability indices for the lower anchor failure, wall
Soil mechanical failure 3.38 4.396 failure and soil mechanical failure are not reached.
Deformations 2.69 1.800
Results Partial Safety Factors
Table 2: Reliability calculated for the simple Quay wall Compared to the method as described in CUR 166
for Sheet Pile walls, for Quay walls with relieving
floor the partial safety coefficients need to be
adapted. A summary as proposed for CUR 211 for
calculated CUR 211 Consequence class 2, i.e. b= 3.8 is indicated in
Table 4.
Upper Anchor failure 4.40 3.828
With respect to the analyses as performed,
Lower Anchor failure 2.99 3.828
compared to the results based on point variation
Wall failure 2.65 3.872 values of parameters, it appeared to be necessary
to include spatial correlation in soil layers in order
Soil mechanical failure 2.78 4.396 to derive realistic values of the partial safety
Deformations 2.44 1.800
factors.
Table 3: Reliability calculated for the Quay wall with relieve The main consequences with respect to the partial
platform safety factors are:
Theoretically the partial safety factors should
be subdivided in factors per failure mechanism
to be checked. For practical purposes one set
Reliability /
Parameter of safety factors is proposed for all mechanisms
Consequence Class
that cover the safety that needs to be realized.
RC2 In deviation from the structural mechanisms
for the design of walls and anchors, where
Angle of internal friction gj 1,25
a factor of 1.25 for soil strength suffices,
Effective cohesion g c 1,25 for soil mechanical stability a higher partial
safety factor on the internal angle of friction
Density gg 1,0 is necessary, should be increased for this
mechanism up to 1.35.
Surcharge load gF 1,05
For practical purposes, in addition to the
Soil Stiffness gE 1,0 structural design, stability needs to be
checked for a higher value of MSF (in a f c
Table 4: Partial safety factors with respect to the characteristic reduction procedure). All procedures further
values of parameters in agreement with Eurocode being related to
characteristic values of the parameters
With respect to stiffness, for the structural
design, characteristic low values of the soil
stiffness need to be applied, however for
anchor failure in addition to that a check needs
to be done with a partial safety coefficient of
0.4.
The latter means that for anchor design the
results of an analysis with a high characteristic
value of soil stiffness needs to be checked.
The analysis indicates that for the geometrical
parameters an additional retaining height and
Fig. 5: Distribution of variation in soil stiffness, (CUR 166) water level difference of 15 cm needs to be
taken into account for both variables.
In engineering practice other uncertainties such
as dredging depth are more uncertain and it is
moment in the piles. advised to account for these small differences
by including them in the definition of the design
Results reliability indices shape of the geometry before doing a Finite
The calculated reliability indices for the different Element analysis.
failure mechanisms are compared to the
prescribed target reliability indices in the first Discussion
edition of CUR 211 to see whether the FEM design The influence factors combined with failure plots
method of CUR 166 leads to sufficient reliable quay (e.g. see Fig. 4) explain how a failure mechanism
walls. Table 2 shows that this target reliability index occurs. Anchor failure in tension mostly occurs
is reached for the failure modes anchor failure due to exceedance of the soil shear resistance
and wall failure. Furthermore the reliability level is in the layer where the anchor bonding element
by far reached for the deformations requirement is positioned, combined with reduction of the
(maximum displacement of 50 mm). However, the passive shear resistance in the lowest sand layer.
obtained reliability for soil mechanical failure is too Sometimes this passive shear resistance is not
low. This can be explained to some extent by the fully developed due to the higher soil stiffness,
applied calculation method, but is also due to the which gives the elastic parameters more influence.
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin
Finally, the uncertainty in anchor parameters can
contribute significantly to the anchor failure.
Wall failure is mainly induced due to reduction
of the internal angle of friction and therewith the
passive shear resistance in the Pleistocene sand
layer. The fixing moment is reduced, which causes
the bending moment in the wall to increase. The
bending moment gets even larger by failure of
soil elements in upper layers, which act as an
additional load on the wall.
Soil mechanical failure (Fig. 5) is also induced
by reduction in passive shear resistance, which
implies that this mechanism is correlated with the
wall failure mechanism.
Conclusions
For the anchored sheet-pile the target reliability
is well approached for the mechanisms wall
failure in bending and anchor failure in tension,
when designing according to the FEM design
prescriptions of CUR 166. The obtained reliability
index for soil mechanical failure is too low, but this
is mainly due to the used calculation method.
In case of quay walls with relieving floor, the
obtained reliability indices are too low. It is
therefore necessary to adapt the FEM design
procedure for this more complicated type of quay
wall and subsequently to modify the partial safety
factors, in order to design the structure with an
acceptable failure probability.
Note that the results are obtained for structures
in the Maasvlakte area of the Port of Rotterdam.
Therefore the conclusions are only valid for those
specific circumstances. The applied method of
analysing the quay walls by probabilistic FEM
calculations, however, can be used for quay walls
in different situations.
References
1. Courage, W.M.G., Steenbergen, H.M.G.M.
(2007). Prob2B: variables, expressions and Excel
Installation and Getting Started. Delft: TNO
Built Environment and Geosciences
2. CUR 166 (2008). Handbook Sheet pile
structures, 5th edition (in Dutch). Gouda
3. CUR 211 (2003). Handbook Quay Walls, 1st
edition. Gouda
4. CUR 211 (2013). Quay Walls, 2nd edition. Gouda
5. Joint Committee on Structural Safety (2002).
Probabilistic Model Code. Retrieved on 22
February 2012 from http://www.jcss.byg.dtu.dk
6. Wolters, H.J. (2012). Reliability of Quay Walls.
MSc thesis on probabilistic Finite Element
calculations of quay walls. Delft: Delft University
of Technology
15
Title

3D Finite Element Analysis of a Complex Excavation

Zhandos Y. Orazalin, Ph.D. Candidate, Department of Civil & Environmental Engineering, Massachusetts Institute of Technology, Cambridge, USA
Andrew J. Whittle, Professor of Civil & Environmental Engineering, Massachusetts Institute of Technology, Cambridge, USA

This article describes the development of a comprehensive three-dimensional finite element model for the Stata Center
basement excavation (Cambridge, USA) using PLAXIS 3D 2012. The project involved a complex sequence of berms, access
ramps and phased construction of the concrete mat foundation. Lateral wall movements and building settlements were closely
monitored throughout construction, while photos from a network of webcams located around the open-plan site provided
a detailed time history of the construction processes. The analyses highlight the effects of the 3D excavation and support
geometry on wall deflections and show a good agreement with the measured response assuming undrained conditions using
the Mohr-Coulomb soil model.

The Ray and Maria Stata Center building at

MIT (Massachusetts Institute of Technology)
was designed with a basement for underground
parking requiring a 12.8m deep excavation.
The excavation was supported by a perimeter
diaphragm wall that formed part of the permanent
structure and extended 14 m into a deep layer
of underlying Boston Blue clay. The diaphragm
wall was braced by a combination of prestressed
tieback anchors, preloaded raker and corner
bracing support elements. The control of ground
movements was a critical aspect of the subsurface
design due to the close proximity of the
excavation to the historical MIT Alumni swimming
pool building (a meter away from the edge of the
excavation).
that ranged from 51mm to 89mm and maximum
settlements exceeding 50mm. Fortunately, these
movements did not cause any noticeable damage
to adjacent structures and were eventually
deemed acceptable. Nonetheless, the magnitude
of these unforeseen movements could have
potentially caused more problems.
More comprehensive three-dimensional finite
element analyses of the Stata Center basement
excavation have been enabled by the recent
advances in PLAXIS 3D software including the
efficient multicore iterative solving capabilities
and geometric data import from CAD files.
These analyses showed a good agreement with
the measured response assuming undrained
conditions in clay and highlighted the effects of
the 3D excavation and support geometry on wall
deflections.
Project Description
The site for the MIT Stata Center has a very large
rectangular plan area (approx. 100m x 119m),
which abuts an existing building along its southern
edge (Figure 2). A floating mat foundation system
was designed so that the weight of the building is
balanced against the weight of the soil extracted
from the site. The excavation support system
comprises a reinforced 76 cm thick concrete
diaphragm wall that is supported through the use
of:
1. three levels of tiebacks on the west, south, and

The complexity of the excavation process and

structural supports presented a significant
modeling challenge that exceeded the Layer Soil Model Top Elevation, m , kN/m3 Su, kPa 'ps G/'vo ' K0
computational capabilities of finite element codes
available at the time of construction (in 2001). Fill MC (D)* 6.4 18.9 - 35 75 0.3 0.5
The predictions of performance were limited to
simplified 2D finite element models (plane strain, Organics MC (UD) 3.0 15.7 48 - 150 0.3 0.5
half sections) that were assumed to generate
Sand MC (D) 1.2 20.4 - 37 230 0.3 0.5
worst-case scenarios for wall deflections and
ground deformations. BBC (Upper) MC (UD) -3.0 18.4 68 - 61 - 75 0.3 0.8

The measured performance during the actual BBC (Lower) MC (UD) -17.0 19.3 61 - 93 - 75 0.3 0.6
excavation exceeded the allowable wall Glacial Till MC (D) -29.0 22.0 - 43 385 0.3 1.0
deformations (38mm) prescribed at the start of
the project with maximum lateral movements Table 1: Soil properties (*D = drained, UD = undrained)

16 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl

 Title

Photo courtesy of http://philip.greenspun.com

Figure 1: Excavation support plan and site view

Figure 2: Subsurface conditions

www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 17

3D Finite Element Analysis of a Complex Excavation
Figure 3: Design of soil clusters to be excavated in several PLAXIS phases
east sides,
2. two levels of corner bracing, and
3. two levels of raker supports on the north side.
Tiebacks were installed at an angle of 20 degrees
from the horizontal at El. 3 m, El. 0.3 m, and El. 3
m and preloaded from 498 to 569 kN. Two levels of
corner bracing consisting of 91-cm-diameter pipe
struts were installed at El. 3 and El. 3 m.
The City of Cambridge prohibited the installation
of tiebacks beneath the Vassar Street such that
two levels of inclined raker bracing (at El. 3 and
El. 3 m) were used along the North side of the
excavation. The raker bracing consisted of 91 cm
in diameter pipe struts extending from embedded
plates in the diaphragm wall to kicker blocks
embedded in the concrete mat foundation (Figure
1).
Site Characterization and Soil Properties
The site is located at the eastern part of the
MIT campus in Cambridge, Massachusetts. The
ground surface is approximately level at El. +6 m,
and the groundwater table is in the overlying fill
18 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl
at the depth of 1.8 m. A typical subsurface profile
underlying the Stata Center in the middle of the
site would consist of 3.4 m of fill, 1.8 m of organics,
4.3 m of sand, 26 m of clay, and 4.6 m of glacial
till (Figure 2). The principal stratum is the marine
clay (Boston Blue Clay), which can be sub-divided
into an upper overconsolidated clay crust and a
lower lighly-overconsolidated unit. The clay has
low hydraulic conductivity and is modeled as an
Undrained Elastic - Perfectly Plastic (EPP) material
with the undrained shear strength that ranges
from a minimum value, su = 60kPa at El. -16m to
a maximum, su 90kPa at the base of the clay.
The other layers are also represented by the EPP
(Mohr-Coulomb) model. Table 1 illustrates soil
properties based on the subsurface exploration
program.
Model Description
The excavation for the Stata Center has a complex
geometry and variety of structural support
systems which makes the project challenging to
model. However, the uniqueness of this project
is that the excavation process was very well
documented - it was constantly photographed,
monitored, and described in daily field logs, as
well as recorded on webcams located around the
construction site. Using these data, it was possible
to create a full three-dimensional numerical model
of the actual excavation with respect to the time
frame of construction sequence.
Olsen (2001) developed a series of 3D geometric
models to represent the construction process by
reconciling daily field reports, photographs and
time-lapse of the project. Figure 4 illustrates the
process of converting the geometric information
into Phases used in the development of the 3D
finite element model. Project data from project
drawings were initially used to construct a CAD
model of the support system. This information
was then used to create a base case model within
PLAXIS 3D. The excavated surface geometry is
obtained from the models reported by Olsen
(2001) that are converted into a set of tetrahedral
elements using mesh tessellation operators in the
CAD program. These are then imported into Plaxis
3D as soil clusters that represent the excavation

process as a series of 36 staged construction
steps in the finite element model. Figure 5 shows
that the resulting finite element model represents
a close approximation to the original geometric
model.
The overall finite element model extends
laterally beyond the footprint of the excavation
to a distance of 150-170m in all directions and
vertically to the base of the glacial till. The model
represents the soil mass using approximately
11200 tetrahedral elements (10-noded) with the
second order interpolation of displacements.
The calculation time lasted less than 20 hours
on an Intel Core i7-3960X Extreme Edition CPU
overclocked to 4.0 GHz with 16 GB RAM on a SSD
hard drive.
The diaphragm wall is represented by three-
dimensional elastic plate elements. The toe of
the diaphragm wall does not extend into the
underlying rock; therefore, it is free to move
within the soil mass. The mat foundation was
also modeled using elastic plate elements;
tiebacks and corner bracers were represented
Figure 4: Staged construction mode and output example
Figure 5: Output of the phase April 25, 2001 in PLAXIS3D (vertical displacements) and corresponding photo of the construction site
3D Finite Element Analysis of a Complex Excavation
as prestressed node-to-node anchors. The
free lengths of tiebacks were modeled using
node-to-node anchors, and the embedded pile
elements represented the grouted part. Each
raker comprises a 91 cm diameter steel pipe strut
(prestressed node-to-node anchors) that supports
the wall and is inclined downward to a kicker block
cast into the foundation slab.
The Boston Blue Clay has been modeled assuming
undrained conditions using the Mohr-Coulomb
model. The layer is subdivided into two units, each
with undrained shear strength varying linearly
with elevation (i.e., matching the undrained
strength profile shown in Figure 2). Since the site
surface is assumed to be mainly horizontal, the
initial stresses are defined by K0 conditions, and
hydrostatic pore pressures, with groundwater
table at El. 4.6m. The boundary conditions used
in the model are the Plaxis standard boundary
conditions with fixity in the horizontal plane at
the basal boundary and zero prescribed lateral
displacements along the corresponding axes at
the borders.
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin
Results
A general pattern of measured movements at the
center of a wall typically correspond to an initial
cantilever movement of approximately 10-20 mm
during the excavation to the first tieback support
level, as well as 32 mm before the first level of
raker support. The movement was fully recovered
(except the North Wall) and the wall moved back
during pre-stressing the first level of bracing.
After the installation of the first level support, the
wall rotated at the brace during the excavation
progress. At the subsequent bracing, the wall kept
moving laterally below the brace location. The
maximum movements (June, 2001) measured in
the inclinometers range from 51-64 mm at SC-02
(North), SC-04 (East), and SC-10 (West) to about
82 mm at SC-07 (South). The greatest movements
were observed within the middle of a tieback
supported wall while the smallest movements
were recorded by the inclinometers located
closer to the corners. In contrast, the North wall,
supported by the raker support, showed an
opposite pattern with the smallest movements
occurring at the center due to the fact that the
plan geometry of the North wall consisted of two
19
3D Finite Element Analysis of a Complex Excavation
planes that intersect at the wall center (Figure 7).
Figure 6 presents comparisons between the
computed and measured wall deflections for the
11 inclinometers located around the perimeter
(Figure 1) at 4 stages of construction (spanning
the period from mid-January to June 2001). The
inclinometers can be sub-divided into sections
where the wall is supported by tieback anchors
(SC-10, SC-08, SC-07, and SC-04), corner bracing
(SC-11, SC-09, SC-06, SC-05 and SC-03) and raker
supports (SC-02, SC-01).
In general, the patterns of measured wall
deflections are very well described by the base
case finite element model. The results are within
5-10mm of the measured maximum and toe
deflections of the diaphragm wall at the end of
construction (Phase 35), with the noted exception
of conditions at the NW and NE corners of the
site (SC-11, SC-01 and SC-03), where measured
maximum wall deflections are 20mm higher
than the numerical predictions. The most likely
causes influencing the results are the ground
loss during construction of the wall panels and
lack of preloading of the 2nd level corner bracing
prior to final excavation of berms in the NE
Figure 6: Lateral wall deflections results
20 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl
corner. Nevertheless, the base case results are in
particularly good agreement with wall deflections
along the tieback-supported South wall and in
close agreement with maximum deflections at the
center of the raker-supported North wall.
Conclusions
The application of a full 3D analysis in PLAXIS
3D 2012 to the Stata Center excavation project
has been demonstrated. In order to capture the
3D effects of soil and support system responses
from a non-uniform excavation process, complex
shapes of soil volumes were extruded based on
the photographs and excavation plans using
CAD. The non-uniform soil excavation resulted
in the three-dimensional effects which were
well-captured by the 3D model predictions. The
analysis results show a good agreement with the
measured data and provide keys to explain many
features of the observed performance including
the differences in diaphragm wall deformations
associated with sections supported by tieback
anchors, raker beams and corner bracing. The
usage of a relatively simple constitutive soil model
(within the undrained conditions) was sufficient
due to the overconsolidated state of the marine
clay. The study has shown that the full 3D finite
element analysis can be effectively used for such
complex excavation projects.
References
1. Hewitt, R. D., Haley, M. X., Kinner, E. B. (2003).
Case History of Deep Excavation on an Urban
Campus. Proc., 12th Panamerican Conf. on Soil
Mechanics and Geotechnical Engineering, Soil
Rock America. Boston.
2. Olsen, M. B. (2001). Measured Performance of
a Large Excavation on the MIT Campus. SM
Thesis, Department of Civil and Environmental
Engineering, Massachusetts Institute of
Technology, Cambridge, MA.
3. Orazalin, Z. Y. (2012). Three-Dimensional Finite
Element Analysis of a Complex Excavation
on the MIT Campus. SM Thesis, Department
of Civil and Environmental Engineering,
Massachusetts Institute of Technology,
Cambridge, MA.
4. Orazalin, Z. Y., Whittle, A. J., & Olsen, M.
B. (2014) Three-dimensional analyses of
excavation support system for Stata Center
basement on MIT campus, in review ASCE
Journal of Geotechnical and Geoenvironmental
Engineering.
 3D Finite Element Analysis of a Complex Excavation

Phase 3: Phase 13:

Excavation EL. +2.4m Excavation EL. -1.2m

Phase 3: Phase 3:
Excavation EL. -4.3m Excavation EL. -6.4m

Figure 7: Three-dimensional diaphragm wall displacements

www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin 21

Recent activities
Product updates: PLAXIS 2D AE
Early 2014 the PLAXIS 2D Anniversary Edition was
released. The user interface of PLAXIS 2D AE has
been restyled to follow the flexible and easy to use
workflow of the PLAXIS 3D program. In this new
graphical user interface the geometric modelling
and staged construction is integrated allowing
for quick and easy switching between input
(geometry) and calculation phases.
Other new and improved features include;
Command Line and Commands Runner, borehole
wizard for soil modelling, CPT import, Remote
Scripting API with python wrapper, Phases
Explorer and Phases Window, and much more.
Furthermore we also deliver a conversion tool,
which proved to be very versatile. In case you run
into a conversion issue, please let us know and we
will update this conversion tool accordingly.
There are also two new user defined soil models
available. The first is an update for the Visco-
Elastic Perfectly Plastic Model. It is a simple and
robust model which can be used to model time-
dependent behaviour (creep and relaxation) of
various materials. The second is the new Swelling
Rock Model, which can be used to simulate the
22 Plaxis Bulletin l Spring issue 2014 l www.plaxis.nl
time-dependent anisotropic swelling of rocks. In
the new developments column at the beginning
of this bulletin you can read more about this new
model.
A look back: European Plaxis Users Meeting
The European Plaxis Users Meeting 2013 was
held in Karlsruhe, Germany. Not only was it the
20th edition of the meeting, Plaxis bv was also
celebrating its 20th Anniversary. For this reason
we had extra social activities including some food
and drinks to commemorate the occasion. We
also announced the winners of the master thesis
competition. The committee members were
Erjona Engin, MSc. (Plaxis bv) Dr. Claire Heaney
(Cardiff University), Dr. Bert Schdlich (TU Graz),
Dr. Nallathamby Sivasithamparam (NGI), and Dr.
Phil Vardon (TU Delft).
The winner was Zhandos Orazalin with his
research on the use of 3D FE analyses in PLAXIS
3D to simulate ground deformations, pore
pressures, and diaphragm wall deflections. In
this issue of the Plaxis Bulletin you can read his
article entiteled 3D Finite Element Analysis
of a Complex Excavation. The Runner-up was
Christian Elescano with his study on soil ground
improvements by grout injection, comparing
solutions obtained with FE and analytical analyses.
The Notable prize went to Jasper Sluis for his
validation of the 2D embedded pile row feature.
His article was already available in the previous
34th edition of the Plaxis Bulletin. You can look
back at all our issues of the Plaxis Bulletins over
the years on our site at www.plaxis.com/bulletin.
The last article by Christian Elescano will be
included in an upcoming bulletin, so keep an eye
out for it.
All in all it was a great moment to look back at 20
years of Plaxis bv and an even longer history of our
software. You can also check out the video on the
evolution of PLAXIS software on our website.
Plaxis Americas LLC
A big step for our North American operations was
the formation of a US based company by Plaxis,
Plaxis Americas LLC. This new company makes
doing business with Plaxis easier and faster, and
underlines the commitment of Plaxis to the North
American geotechnical community. Sales and
backoffice activities for all US and Canadian clients
are now coming from this new Houston based
company.

The past fall and winter saw Plaxis exhibit at a
variety of national and regional geotechnical
events this illustrates the diverse geotechnical
sectors in which Plaxis is being used (e.g. DFI, Dam
Safety). At CGSs GeoMontreal a lot of interest in
PLAXIS came from existing and potential users
from all across Canada. And more recently, at
ASCEs Geo-Congress, the new PLAXIS 2D AE
version was demonstrated to numerous engineers.
Well continue to visit and exhibit at events
across North America in 2014, so check the list of
upcoming events at www.plaxis.com/events to see
when and where you can meet us in person. We
look forward to meet you!
The upcoming months will see two educational
opportunities: a standard course in New York
in June, and an advanced course in Houston in
October. These well-balanced courses are led
by a course leader from Plaxis and have in-depth
contributions from American professors. And for
the first time we are offering early bird discounts
for our North American courses (which stacks with
VIP discount) so make sure to register early to get
the best price!
News from Plaxis Asia-Pacific
Plaxis AsiaPac started the year by conducting a
series of technical seminars and workshops in Asia.
India
In association with our Indian agent, Ramcaddsys,
we conducted a series of lectures on the
modelling of Geotechnical problems.
In collaboration with the Civil and Ocean
Engineering Departments from IIT-Madras,
two seminars were conducted on the 11th and
12 th February 2014. The seminars were on the
application of PLAXIS 2D and 3D programs.
The seminars centered on the modelling
of excavations, foundations, tunnelling and
earthquake analysis.
www.plaxis.nl l Spring issue 2014 l Plaxis Bulletin
In collaboration with Cochin University of Science
and Technology (CUSAT) a workshop on the
application of PLAXIS 3D was conducted on 14th
February 2014.
Finally the 5th Indian PLAXIS Users meeting was
held in the city of Kolkata on the 18th of February
2014. The users meeting was well received and was
attended by users from all academia and industry.
We look forward to be back for the 6th edition of
Indian PUM in New Delhi in 2015.
Plaxis AsiaPac and Ramcaddsys will be back in
New Delhi for the 4th Indian Advanced Course.
It will be held from the 29 th to 31st of October
2014 in collaboration with the Indian Institute of
Technology, New Delhi.
Hong Kong
PLAXIS AsiaPac in association with our Hong
Kong agent, Solution Research Centre, conducted
a series of Entry and training workshops in
March 2014. This series of modular training
workshops offered greater flexibility and allowed
our valued users to subscribe to a structured
training program. Module HKG 1 on the topic of
Introduction to PLAXIS 2D was held on 23 rd of
February. This was followed by Module HKG 3
which is on the topic of Geotechnical Modelling.
We will be back in the second quarter of the year
to conduct these modules again with the addition
of advanced application modules.
Singapore
PLAXIS AsiaPac in association with Norwegian
Geotechnical Institute conducted a one-day
seminar on Offshore Geotechnical Modelling
using FEM on the 1st of April 2014. The seminar
was well received and attended by users from the
Offshore industry. This is the first of our annual
seminar on the subject of Offshore Geotechnics.
We look forward to conducting this event in 2015.
23
Title

Upcoming Events 2014

April 9, 2014 June 22 - 25, 2014 September 23 - 26, 2014
Lancement de la nouvelle interface graphique de 2014 North American Tunneling Conference 33. Baugrundtagung
PLAXIS 2D Los Angeles CA, U.S.A. Berlin, Germany
Paris, France
June 23 - 26, 2014 September 28 - October 1, 2014
April 10, 2014 Standard Course on Computational Geotechnics GeoRegina 2014
Workshop: Introduction PLAXIS 2D AE Manchester, United Kingdom Regina, Canada
Paris, France
June 24 - 25, 2014 29 September - 2 October, 2014
April 23 - 25, 2014 Russian Plaxis Users Meeting Advanced Course on Computational Geotechnics
27th Central Pennsylvania Geotechnical St. Petersburg, Russia Trondheim, Norway
Conference
Hershey PA, U.S.A. 30 June - 3 July, 2014 30 September - 3 October, 2014
Curso de Geotecnia Computacional Standard Course on Computational Geotechnics
May 5 - 8, 2014 Santiago de Quertaro, Mexico Brisbane, Australia
Offshore Technology Conference
Houston TX, U.S.A. July 8 - 10, 2014 6 - 10 October, 2014
Journees nationales de geotechnique et de Advanced Course on Computational Geotechnics
June 1 - 4, 2014 geologie de lingenieur & Dynamics
48th ARMA Rock Mechanics Symposium Beauvais, France Wellington, New Zealand
Minneapolis MN, U.S.A.
September 8 - 11, 2014 October 7 - 10, 2014
June 3, 2014 Standard Course on Computational Geotechnics Advanced Course on Computational Geotechnics
Workshop: Introduction to PLAXIS 3D Zurich, Switzerland Houston TX, U.S.A.
Delft, The Netherlands
September 10 - 13, 2014 October 13 - 15, 2014
June 4, 2014 COBRAMSEG 2014 AFTES 14th International Congress
Workshop: Dynamics in PLAXIS Goinia, Brasil Lyon, France
Delft, The Netherlands
September 12, 2014 October 21 - 24, 2014
June 17 - 20, 2014 Workshop on Foundations and 3D Modelling DFI 39th Annual Conference on Deep
Standard Course on Computational Geotechnics Zurich, Switzerland Foundations
& 3D Modelling Atlanta GA, U.S.A.
New York, U.S.A. September 21 - 25, 2014
Dam Safety 2014
June 18 - 20, 2014 San Diego CA, U.S.A.
Numerical Methods in Geotechnical Engineering
Delft, The Netherlands

www.plaxis.com
Plaxis bv Computerlaan 14 Plaxis AsiaPac Pte Ltd 16 Jalan Kilang Timor Plaxis Americas LLC 2500 Wilcrest Drive, St. 300
Tel: +31 (0)15 2517 720 2628 XK Delft Tel: +65 6325 4191 #05-08 Redhill Forum Tel: +1 (650) 804 4729 Houston TX 77042
info@plaxis.com The Netherlands asiapac@plaxis.com 159308 Singapore americas@plaxis.com U.S.A.