Geotextiles and Geomembranes: S.J.M. Van Eekelen, A. Bezuijen, A.F. Van Tol

Geotextiles and Geomembranes 39 (2013) 78e102
Contents lists available at SciVerse ScienceDirect
Geotextiles and Geomembranes

journal homepage: www.elsevier.com/locate/geotexmem
An analytical model for arching in piled embankments

S.J.M. van Eekelen a, b, *, A. Bezuijen a, c, A.F. van Tol a, b
a
Deltares, Unit Geo-Engineering, P.O. Box 177, 2600 MH Delft, The Netherlands
b
Delft University of Technology, The Netherlands
c
Ghent University, Belgium
a r t i c l e i n f o a b s t r a c t
Article history: Most analytical models for the design of piled embankments or load transfer platforms with geosynthetic
Received 7 February 2012 reinforcement (GR) include two calculation steps. Step 1 calculates the arching behaviour in the ll and
Received in revised form step 2 the load-deection behaviour of the GR. A calculation method for step 2 based on the results of
22 May 2013
model tests has been published by Van Eekelen et al. (2012a,b). The present paper analyses and presents
Accepted 4 July 2013
Available online 16 August 2013
a new model for step 1, which is the arching step. Additional tests, which are also presented in this paper,
were conducted for this purpose.
The new model is a limit-state equilibrium model with concentric arches. It is an extension of the
Keywords:
Load transfer platforms
models of Hewlett and Randolph (1988) and Zaeske (2001). The new model results in a better repre-
Arching sentation of the arching measured in the experiments than the other models mentioned, especially for
Piled embankments relatively thin lls.
Soil reinforcement Introducing GR in a piled embankment results in a more efcient transfer of load to the piles in the
Concentric arches model form of an arching mechanism. The load is then exerted mainly on the piles and the GR strips between
Analytical models the piles, on which the load is approximately distributed as an inverse triangle. The new model presented
in this paper describes this behaviour and is therefore meant to describe the situation with GR. The new
model provides a physical explanation for observations of the arching mechanism, especially the load
distribution on the GR. Other observations with which this model concurs are the dependency on ll
height and friction angle. The amount of arching increases with increasing subsoil consolidation and GR
deection. The paper describes how the new model relates to the development of arching as a result of
subsoil consolidation.
2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
1. Introduction load part B which passes through the GR to the piles, and a part C
resting on the subsoil, as indicated in Fig. 1.
Many analytical design models for the design of piled em- Van Eekelen et al. (2012b) analysed and made proposals for
bankments include two calculation steps. The rst step calculates calculation step 2. The present paper analyses and puts forward a
the arching behaviour in the ll. This step divides the total vertical new model for step 1, the arching step. Both papers compare the
load into two parts: load part A, and the residual load (B C in results with measurements from a model test series presented in
Fig. 1). Load part A, called arching A in the present paper, is the part the rst part (Van Eekelen et al., 2012a) of this three-part study.
of the load that is transferred to the piles directly. These tests are particularly suitable for the validation of calculation
The second calculation step describes the load-deection steps 1 and 2 separately because A, B and C were measured sepa-
behaviour of the geosynthetic reinforcement (GR, see Fig. 1). In rately. For the present paper, a number of additional tests were
this calculation step, the residual load is applied to the GR strip carried out with the same test set-up.
between each pair of adjacent piles and the GR strain is calculated. Several families of analytical models describing step 1 (arching)
An implicit result of step 2 is that the residual load is divided into a are available in the literature. Terzaghi (1943) listed a number of
them. Current arching models comprise:
Rigid arch models, such as several Scandinavian models

(Carlsson, 1987; Rogbeck et al., 1998, modied by Van Eekelen
et al., 2003; Svan et al., 2000) and the Enhanced Arching
* Corresponding author. Deltares, Unit Geo-Engineering, P.O. Box 177, 2600 MH
Delft, The Netherlands. Tel.: 31 88 335 72 87. model (also called the BusheJenner model or the Collin, 2004
E-mail address: [email protected] (S.J.M. van Eekelen). model) and the present design method of the Public Work
0266-1144 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
http://dx.doi.org/10.1016/j.geotexmem.2013.07.005
S.J.M. van Eekelen et al. / Geotextiles and Geomembranes 39 (2013) 78e102 79
Nomenclature Lx2D part of the GR strip that is oriented along the x-axis
(perpendicular to the road axis) and on which the 2D
A load part transferred directly to the pile (arching A in arches exert a force, see Fig. 23 and Eq. (12), m
this paper) expressed as kN/pile kN/unit cell, kN/pile Ly2D part of the GR strip that is oriented along the y-axis
A% arching A presented as a percentage of the total load, A (parallel to the road axis) and on which the 2D arches
% is the same as the pile efcacy (E) as used by several exert a force, see Fig. 23 and Eq. (12), m
authors: A% E 1 ABC BC or Lx3D width of square on which the 3D hemispheres exert a
A A load, see Fig. 22 and Eq. (8), m
A% E ABC gHp$sx $sy , %
P2D calculation parameter given by Eq. (1). Px2D refers to a
a width of square pile cap. Bers a, m 2D arch that is oriented along the x-axis, as indicated in
B load part that passes through the geosynthetic Fig. 12 and Eq. (14). Py2D refers to a 2D arch that is
reinforcement (GR) to the pile expressed as kN/pile oriented along the y-axis, kPa/mKp1
kN/unit cell, kN/pile calculation parameter given by Eq. (7), kPa/m2Kp2
p P3D
Bers equivalent size of circular pile cap, Bers 1=2$d$ p or p uniformly distributed surcharge on top of the ll (top
the width of a square pile cap, m load), kN/m2
C load part that is carried by the soft soil between the Q2D calculation parameter given by Eq. (1), kN/m3
piles (this soft soil foundation is called subsoil in this Q3D calculation parameter given by Eq. (7), kN/m3
paper) expressed as kN/pile kN/unit cell, kN/pile r radius of a 2D arch, m
C a constant to be calculated with boundary conditions R radius of a hemisphere (in this paper a hemisphere is a
(Eqs. (29)e(34) and (47)e(50) in the appendix) 3D arch), m
d diameter circular pile (cap), m Rb total friction between ll/box walls and foam
E pile efcacy, the same as A%, e (kN/kN) cushion/box walls and piles, see Van Eekelen et al.
F force, kN (2012a,b), kN/pile
GR geosynthetic reinforcement sd the diagonal centre-to-centre distance between piles
h or H height of the ll above bottom layer of GR, m q
Hg2D height of the largest of the 2D arches of the new sd s2x s2y , m
concentric arches model, see Eqs. (2) and (13) and sx, sy pile spacing perpendicular to the road axis (x) or
Figs. 10 and 12. Hxg2D refers to the height of a 2D arch parallel to the road axis (y), m
that is oriented along the x-axis (perpendicular to the Wn net load ( Ws C Rb), kN/pile
road axis), as indicated in Fig. 12. Hyg2D refers to the Ws total surcharge load on a unit area Ws p$sx$sy, kN/pile
height of a 2D arch that is oriented along the y-axis, m z distance along the vertical axis as indicated in, for
Hg3D height of the largest 3D hemisphere of the new example, Fig. 3, m
concentric arches model, see Eq. (4) and Fig. 10, m 4 internal friction angle,
hg arch height in EBGEO, hg sd/2 for h sd/2 or hg h for g ll unit weight, kN/m3
h < sd/2, m sr radial stress in a 2D arch, kPa
J2% tensile stiffness of the GR at a GR strain of 2%, kN/m sR radial stress in a 3D hemisphere, kPa
k subgrade reaction, kN/m3 sq tangential stress in 2D arch or 3D hemisphere, kPa
Kp passive or critical earth pressure coefcient, e PET polyester
PP polypropylene
PVA polyvinyl alcohol
Research Center in Japan (2000, discussed in Eskisar et al. 2012). (2010) as an alternative for the original empirical model in
In this class of models, it is assumed that an arch is formed that BS8006. The other frequently used equilibrium model is Zaeskes
has a xed shape. The shape of the arch is usually 2D or 3D model (2001, and also described in Kempfert et al., 2004), which
triangular. It is assumed that the entire load above the arch, is explained in Fig. 3. This model was adopted in the German
including the soil weight and the trafc load, is transferred EBGEO (2010) and the Dutch CUR226 (2010, described in Van
directly to the piles (load part A, or arching A, see Fig. 1). The Eekelen et al., 2010), and we refer to it here as EBGEO.
weight of the soil wedge is carried by the GR subsoil (B C). Another family of arching models is the family of frictional
These models do not consider the mechanical properties of the models. Several authors have adopted the frictional model
ll, such as the friction angle, in their equations and they are proposed by Terzaghi (1943), who in turn based his model on
therefore not discussed further in the present paper. previous work from other authors such as Cain (1916) and
In equilibrium models, an imaginary limit-state stress-arch is Vllmy (1937). McKelvey (1994) extended Terzaghi by assuming
assumed to appear above the GR soft subsoil between the stiff that a plane of equal settlement exists and combined this with
elements. In the 3D situation, these stiff elements are piles; in a tensioned membrane theory.
the 2D situation, they are beams or walls. The pressure on the Russell and Pierpoint (1997) extended the Terzaghi model to
GR subsoil (B C) is calculated by considering the equilibrium include a third dimension by assuming the presence of friction
of the arch. In most models, the arch has a certain thickness. in the vertical planes along the edges of the square pile caps.
Two limit-state equilibrium models are frequently used in piled McGuire et al. (2012) also adopted the idea of a plane of equal
embankment design today. One of them is the Hewlett and settlement, which they described as the critical height. They
Randolph model (1988), explained in Fig. 2, which was adopted conducted numerous tests and collected eld data to determine
in the French ASIRI guideline (2012) and suggested in BS8006 and validate their equation for the critical height. This critical
80 S.J.M. van Eekelen et al. / Geotextiles and Geomembranes 39 (2013) 78e102
geometry
step 1 step 2
properties load part B+C strain
arching membrane
load
load part A B+C
A A
GR strip
B+C B
soft C C
subsoil support from subsoil (C)
Fig. 1. Calculating the geosynthetic reinforcement (GR) strain comprises two calculation steps.
height should be used in combination with Russell and The nal type of model that should be mentioned is the family of
Pierpoints (1997) version of Terzaghi (1943). hammock models, such as the path of minor principal stress
Naughton (2007) determined the critical height with log spiral model described by Handy (1985).
shear planes. Britton and Naughton (2008) presented 3D ex-
periments validating the critical height of this model. Most step 1 models calculate one average pressure on the GR.
Although the ideas underlying the frictional models are EBGEO uses, in its step 2, a triangular pressure distribution. This is
extremely important, these models are not generally used in an assumption following from Fig. 9.15 of EBGEO (2010), not the
Europe for piled embankment design and they will not be dis- result of a calculation. Van Eekelen et al. (2012a,b) showed that the
cussed further. A possible reason for the infrequent use of these measured pressure distribution on the GR strip between the piles
models in Europe is that the results depend to a large extent on can be better approximated with an inverse triangle.
the value of K0 (the ratio between horizontal and vertical The present paper presents a new equilibrium model for step 1
pressure) and the fact that it is difcult to determine an accurate that is a better match for several experimental, numerical and eld
value for K0. observations, particularly the measured inverse triangle in the
Other examples of models considered in the literature are the pressure distribution. The starting point for this study consisted of
models using mechanical elements, like the load displacement the design models in general use in Europe. The new model is an
compatibility method of Filz et al. (2012), the one-dimensional extension of the Hewlett and Randolph (1988) and EBGEO (2010)
model of Chen et al. (2008) and the plane strain models of models.
Deb (2010), Deb and Mohapatra (2012) and Zhang et al. (2012).
Filz et al. (2012) model the ll, the GR and the piles subsoil as 2. Additional laboratory tests
separate elements. The boundary condition for each of the ele-
ments is that deformation must match neighbouring elements. 2.1. Measurements of steps 1 and step 2 separately
This determines the load transferred to the piles directly
(arching A, see Fig. 1). Filz et al. (2012) limit arching A with 3D Van Eekelen et al. (2012a) presented and analysed a series of
Terzaghi (cross-shaped, according to Russell and Pierpoint, twelve model laboratory model tests on piled embankments. Since
1997), plus critical height. then, eight additional model tests have been carried out with the
A familiar empirical model is the modied Marston and same set-up. Six of them were carried out specically to validate
Anderson model (1913) that was modied by Jones et al. variations in calculation step 1. These tests are presented here. One
(1990) and adopted in BS8006 and the Finnish design guide- of the strengths of the test set-up was that it was possible to vali-
line (Liikennevirasto, 2012). Marston and Anderson (1913) car- date calculation steps 1 and 2 (Fig. 1) separately because the values
ried out numerous experiments to determine arching above a of A, B and C are measured separately.
pipe in soil. They found a 2D equation that was modied by
Jones et al. (1990) for the 3D piled embankments, as explained 2.2. Description tests
and further modied in Van Eekelen et al. (2011). This model is
very important because of the widespread application of Fig. 4 shows the test set-up for both the rst and the second
BS8006. series, except for two additional total pressure cells (TPCs), which
Fig. 2. Hewlett and Randolph (1988) consider the crown element of the diagonal arch and the foot element (just above the pile cap) of the plane strain arch as indicated in this
gure.
Fig. 3. Zaeske (2001) considers the equilibrium of the crown elements of the diagonal arches.
were used in the second series only. The insides of the box walls of 16.4 kN/m3). The friction angle 4 49 of the granular ll was
were located on the lines of symmetry. A foam cushion modelled measured in large diameter ( 0.3 m) triaxial tests. The average
the soft soil around the four piles. This cushion was a saturated and relative density in the triaxial tests and scale model tests was nearly
watertight sealed foam rubber cushion. A tap allowed drainage of the same at 61.0% and 62.9% respectively. The top load was applied
the cushion during the test to model the soft-soil consolidation with a water cushion that made it possible to apply stresses com-
process. A stiff steel frame was installed to which the GR was parable with eld stresses.
attached. The steel frame could move freely in a vertical direction. The area replacement ratio a2/(sx$sy) in the test series
Since it tted precisely in the container, no horizontal movement described was 2.6%, while the area replacement ratio in Dutch
was possible. Differential settlement along the frame bars was not practice is between 4.5 and 12%. This was a deliberate decision
possible. It is assumed that this has a negligible inuence on dif- with the aim of generating enough GR tensile forces at this
ferential settlement between the piles. smaller scale.
The embankment in most tests was granular ll (crushed After the introduction of the ll, each test was carried out as
recycled construction material 1e16 mm, 4 49 at a unit weight follows: (1) one drainage step foam cushion (subsoil consolidation),
(2) rst top load increment, (3) one or more drainage steps (4)
second top load increment, (5) one or more drainage steps and so
on, up to the maximum top load (varying between 50 and 100 kPa)
water cushion for surcharge and the subsequent drainage steps. The test concluded with the
complete removal of the subsoil support by applying vacuum to the
0.10 m foam cushion.
TPC arching fill H Table 1 lists a selection of the tests in the rst series, and all the
load part A x tests in the new second series presented in the present paper. The
tests from the second series are all variations on test K2, except that
each test included one variation, which is indicated in bold in
load
Table 1. Furthermore, additional total pressure cells were added to
pile
pile
P part C TPC load

parts A+B measure the load distribution on the GR strips.
The GR in these additional tests consisted of two layers of woven
PVA grid, except for test K7, where the bottom layer was replaced
0.225 m 0.45 m 0.225 m
by a similar PVA geotextile. The two uniaxial reinforcement layers
0.10 m 0.10 m were placed directly upon each other on one frame. The strength
direction of one geogrid was perpendicular to the other. There was
0.225 m
frame for no distance between these two GR layers. The two layers are
reinforcement GR therefore considered to be a single GR layer that is completely
0.10 m
biaxial.
pile 3
GR stiffness depends on GR strain and the duration of loading,
pile 1 with as well as other factors. The stiffness values of the weak direction
2 TPCs of one GR layer and the stiff direction of the other layer were
0.45 m
x aggregated, resulting in a total GR stiffness J2% 2269 kN/m in

extra TPCs each direction. J2% is the GR stiffness for a GR strain of 2%, and is
determined in accordance with ISO 10319. These ISO tests are
much faster than the piled embankment model tests (that took
0.10 m
pile 4
3 days each). This means that the GR in the model tests is loaded
pile 2 with longer and will behave differently (less stify) from the
0.225 m
2 TPCs behaviour suggested by this J2%. However, GR stiffness is not

x
a parameter in analytical step 1: the arching calculations
considered in the present paper. Van Eekelen et al. (2012a)
conrmed that using GR has a major impact on the arching
Fig. 4. Test set-up for piled embankment model experiments. The stress distribution is
measured with total pressure cells (TPCs). See also Van Eekelen et al. (2012a), which
mechanism, but GR stiffness (as long as the stiffness complies
includes an extensive presentation of the test series, including measurements of with a good-quality design model) does not have a measurable
strains and deformations. inuence.
Table 1
Specication of the additional series 2 scaled model tests, and a selection of the series 1 tests. The series 2 tests were a variation on test K2 presented in Van Eekelen et al.
(2012a).
Sequence Code GR J2% 2269 kN/m except tests Fill height granular Fill unit Top load pa
of tests T2 and T3 ll 4 49.0 weight
except test T2
m kN/m3 kPa and kN/pile
Selection of tests of series 1 (Van Eekelen et al., 2012a)

6 K2 2 PVA grid 0.420 16.59 p 0e25e50e75e100 kPa ( 0e7.6e15.1e22.7e30.2 kN/pile)
peR 1.2e22.0e41.1e59.8e74.9 kPa
5 T2 13 PP geotextile J2% [ 970/1810 kN/m 0.420 sand 16.55 p 0e25e50e75 kPa ( 0e7.6e15.1e22.7 kN/pile)
4 40.9 peR 1.8e18.3e37.4e50.9 kPa
7 T3 13 PP geotextile J2% [ 970/1810 kN/m 0.420 16.70 p 0e25e50e75 kPa ( 0e7.6e15.1e22.7 kN/pile)
peR 0.2e19.7e39.0e55.2 kPa
Series 2
13 K4 2 PVA grid 0.655 16.80 p 0e25e50e75 kPa ( 0e7.6e15.1e22.7 kN/pile)
peR 0.9e13.9e28.5e44.7 kPa
peR 2.7e14.3e31.9e49.9e67.2 kPa
15 K6 2 PVA grid 0.429 16.35 p [ 0e50e100 kPa ([ 0e15.1e30.2 kN/pile)
peR 1.3e31.7e66.1 kPa
16 K7 Bottom layer PVA geotextile, 0.426 16.42 p 0e25e50e75e100 kPa ( 0e7.6e15.1e22.7e30.2 kN/pile)
top layer PVA grid peR 0.0e17.3e33.9e51.4e68.0 kPa
peR 0.1e18.0e35.6e57.7e73.8 kPa
peR 4.8e12.4e28.6e43.5e63.8 kPa
a
After each top load increase, controlled drainage of the foam cushion (subsoil) follows in one or more steps until the subsoil support C is nearly gone. The calculations
were carried out with surcharge load peR, where R is the measured friction between ll and box walls. The listed peR is given for the moments just before top load increase
(and therefore for minimal subsoil support C).
2.3. Results of the additional tests 3. Observations of arching in experiments, eld tests and
numerical calculations
Fig. 5 compares the measurements of arching A in the new
model tests with several old model tests. The numbers between 3.1. Measuring arching
brackets refer to the sequence of tests.
The granular ll was re-used for each test. It was observed that This paper focuses on the description of the arching mechanism,
large numbers of grains were crushed during the successive tests. This and therefore on calculation step 1 in Fig. 1. Arching divides the
will probably have caused a reduction of the friction angle. A lower vertical load into two parts, as shown in Fig. 1. One part is load part
friction angle results in less arching, as shown by the gure. Arching A (called arching A in this paper), which is the load that is trans-
was also relatively low in test T2, in which a sand ll was used. ferred to the pile caps directly. The remaining load part is B C.
The friction angle 4 49 of the granular ll was measured with Direct validation of arching in a GR reinforced piled platform
large triaxial tests carried out between experiments 12 and 13. This with measurements is only possible when A is measured sepa-
effect of reducing friction is neglected in the present paper. In the rately, as shown in Fig. 4. This gure shows how load A was
calculations, friction angle 4 49 was applied. measured in the tests, using total pressure cells (TPCs) with a
More results from the additional tests are presented in Figs. 7, diameter equal to the pile diameter. They were located on top of the
15, 17, 19 and 21 of this paper. piles and on top of the GR. In addition, two total pressure cells
measured A B. They were located below the GR on top of the piles.
16 This paper gives arching A in kN/pile. In several gures, arching is
presented as a percentage of the total load, which is referred to as
14
A%. A% corresponds to efcacy (E) as used by several authors,
measured load part A (kN/pile)
where E A% A/(A B C), with A, B and C given in kN/pile, A% in

12
% and E as a ratio (e).
10 Many researchers (such as Zaeske, 2001; Blanc et al., 2013) have
measured A B or measured A in a piled platform without GR
8 (Hewlett and Randolph, 1988; Ellis and Aslam, 2009, for example).
As far as we are aware, only the Dutch research programme has
6 reported measurements of A in experiments with GR (Van Eekelen
T2 (5) T3 (7)
K2 (6) K4 (15) et al., 2012a,b) and three eld tests with GR (Van Duijnen et al.,
4
K5 (16) K6 (17) 2010; Van Eekelen et al., 2010; Van Eekelen and Bezuijen,
2 K7 (18) K8 (19) 2012, 2013).
K9 (20)
0 3.2. Comparison of predictions and measurements of arching A
0 5 10 15 20 25
net load Wn (kN/pile) Van Eekelen et al. (2012b) measured A in their experiment series
Fig. 5. Measurements of arching A in a selection of the experiments in series 1 (Van and showed that EBGEO generally under-predicts the measured A.
Eekelen et al., 2012a) and the additional tests in series 2. The numbers between the This paper presents additional tests of the same type. Van Duijnen
brackets refer to the sequence of tests. et al. (2010) measured A in eld tests under a railway in Houten, the
Netherlands. They showed that EBGEO considerably under-predicts the piles was much larger with, respectively, a 2D analytical model,
A, as also shown in Fig. 16a. an axial-symmetric analytical model and in 2D experiments.
More recently, Van Eekelen and Bezuijen (2013) showed that The localisation of the load on the GR strips has been shown by,
EBGEO only slightly under-predicts A as a result of the permanent for example, Zaekse (2001) by measuring the pressure at three
load in eld measurements in Woerden, the Netherlands, as shown locations on the GR square and the GR strips (Fig. 6). Note that, in
in Fig. 16b. this paper, the square between four piles is referred to as the GR
It should be noted that the results of the predictions are highly square, even when no GR is in place, as indicated in Fig. 11. Fig. 6
dependent on the friction angle of the ll and that this friction shows that introducing GR clearly transfers the load towards the
angle is difcult to determine in the eld. The presented EBGEO GR strips. This results in a load distribution that is concentrated
prediction for Woerden, for example, is calculated with a best- mainly on the GR strips (and probably the piles, but Zaeske did not
guess friction angle 4 43 , resulting in A 113 kN/pile, as indi- measure A). As a result, it is expected that the strains in the GR
cated in Fig. 16b. However, this EBGEO-prediction of A falls to occur mainly in the GR strips between the piles. This was indeed
96 kN/pile for 4 37.5 and rises to 132 kN/pile for 4 49 . found in both Zaeske (2001) and Van Eekelen et al. (2012b).
It is also not certain that this friction angle or cohesion remains The pressure on the GR strips is not equally distributed; it rises
constant in all circumstances, such as heavy rain or long dry, hot towards the piles. In the additional tests presented in this paper, the
periods: the arching in Fig. 16b increased in the spring of 2011 load distribution on the GR strip was measured with additional
during a dry, hot period. total pressure cells on the GR strip (Fig. 4). These measurements
(Fig. 7) do indeed show that the load on the GR rises towards the
piles. This load distribution can be approximated by a model with
3.3. Impact of ll height
an inverse triangular load distribution. The inverse triangular
model has advantages since it is a relatively simple analytical
Several researchers have reported that the efciency of arching
model. Van Eekelen et al. (2012a,b) and Van Eekelen and Bezuijen
increases with increasing ll height. Examples are Chen et al. (2008)
(2013) showed that this simplied analytical model provides a
in 2D experiments with GR and Zaeske (2001) with 3D experiments
good match with measurements of deformation in laboratory ex-
(although the latter only showed the differences in his experiments
periments. Furthermore, the inverse triangle (or at least the con-
without GR), Han and Gabr (2002) with numerical analysis with GR,
centration of load close to the piles, and the minimum load in the
Le Hello and Villard (2009) with numerical analysis with GR (see
centre between piles) was also found in, for example, nite element
Fig. 18), Jenck et al. (2009) with 2D experiments without GR and
calculations on a soldier pile wall by Vermeer et al. (2001), discrete
numerical analysis, Ellis and Aslam (2009) with centrifuge tests
element calculations on a heap of grains on a deecting subsurface
without GR, and Deb and Mohapatra (2012) with 2D analytical cal-
(Nadukuru and Michalowski, 2012), numerical calculations by Han
culations. All these researchers showed that a higher ll results in
et al. (2012), with a inversed triangle in their Fig. 9, and by Den
relatively more load being transferred to the piles, either directly or
Boogert et al. (2012), settlement measurements in a eld test
via the GR. A higher ll therefore results in a relative reduction in the
(Van Eekelen and Bezuijen, 2012, 2013) and the large-scale model
load exerted on the GR and/or the subsoil between the piles.
tests of Filz and Sloan (2013).
The experiments presented in this paper also show that ll
GR also has a major effect on ground pressure in the ll above
height has an impact, as seen in Fig. 17. A% increases with ll height
the GR between the piles. Zaeske (2001, pages 55 and 63) showed
and seems to stabilise with increasing embankment height, as
that this ground pressure declines with increasing ll depth. When
shown in Figs. 17 and 18. When A% stabilises, the absolute values of
GR is applied, the fall in ground pressure with depth is much larger
A and B C (kN/pile or kPa) will increase with increasing ll height.
than without GR, as shown by the comparison of Zaeskes mea-
This tendency towards increasing arching is followed by many
surements in the situations with and without GR in Fig. 20.
design models, such as Carlsson (1987), Guido et al. (1987), Hewlett
Zaeskes ndings (2001) showed that there is an interaction
and Randolph (1988), Russell and Pierpoint (1997), Sintef (2002)
between the GR and the ll. Without GR, the arch is much less
and Kempfert et al. (2004), which was adopted in EBGEO (2010),
efcient than with GR. The GR attracts the load to the GR strips
as Le Hello and Villard (2009) have shown.
between the pile caps and then further to the pile caps, approxi-
The critical height decreases with increasing ll height. The
mately resulting in the inverse triangular load distribution on the
critical height is the height at which the shear forces in the
GR strips. This ultimately results in larger vertical loads on the pile
embankment ll are reduced to zero (Naughton, 2007), which is the
caps and on the GR close to the pile caps. GR therefore makes
case at the plane of equal settlement as dened by McKelvey
arching much more efcient.
(1994). Lally and Naughton (2012) carried out a series of 2D GR-
The current analytical models do not give the localisation of the
reinforced centrifuge tests. They found close agreement between
load on the GR strips. Nor do they result in a concentration of load
the observed critical heights and the critical height suggested by
on the GR in the area around the pile cap or, therefore, lead to an
Hewlett and Randolphs model (1988).
inverse triangular load distribution. This paper describes a new
equilibrium model that is a better match for load distribution ob-
3.4. Load distribution on the GR; inuence of using GR servations. The resulting model is only applicable to load transfer
platforms with GR.
There is a considerable difference between piled embankments
with or without GR. GR, when stiff enough, leads to (1) more ef- 3.5. Inuence of subsoil consolidation or GR deection and ll
cient arching and therefore a higher A, (2) a concentration of load properties
on the GR strips (3) an inverse triangular load distribution on the
GR strips and (4) a larger fall in stress between the piles with depth 1. Consolidation or compaction of the subsoil results in an in-
in the embankment above the GR. Each of these features will be crease of arching A, as shown by Van Eekelen et al. (2012a), and
explained in this section. in Fig. 8. Most analytical models, like EBGEO, do not calculate
Chen et al. (2008), Abusharar et al. (2009) and Deb and with increased arching due to subsoil consolidation. An
Mohapatra (2012) showed that the efciency of the piled embank- exception is the model of Deb (2010), which agrees with this
ment improves greatly when GR is used. They found that the load on inuence of consolidation.
p=20 kPa
no GR; p=20 kPa
p=54 kPa
60 no GR; p=54 kPa
p=104 kPa TPC3
measured ground pressure (kPa)

110
50 no GR; p=104 kPa
22 16 34 16 22
40
30 TPC2
TPC2 TPC1
25
TPC1 TPC 3 20
25
10
0
0 0.1 0.2 0.3
30 25 12.5 12.5 distance from centre between 2 piles (m)
Fig. 6. Comparison of results of the Zaeske test series (2001, page 60, GR is a woven grid PET 60/60-20). Top view. Zaeske measured the load distribution at 5 cm above the GR, at the
locations TPC1, TPC2 and TPC3. See also Fig. 20. Sizes in cm. H 0.7 m, 4ll 38 .
2. A lower friction angle of the ll gives less arching during the GR strips. They do not give an explanation for the approxi-
consolidation. This results in load part B C on the GR being mately inverse triangular load distribution on the GR strips. How-
some 39% larger for sand than for granular ll at the end of the ever, they do give decreasing ground pressure with depth in the ll
tests presented by Van Eekelen et al. (2012a) and Fig. 8b. above the GR square, and they do give increasing efciency in
arching with increasing ll height.
Most of the existing arching models assume that a slight
deection of the geotextile is sufcient to create a full arch. 4. A new equilibrium model: the concentric arches model
However, the measurements showed that A increases with
increasing GR deection (due to subsoil consolidation). The inu- 4.1. Introduction
ence of deformation cannot be incorporated in rigid-plastic models
such as the equilibrium models or frictional models. A new class of With equilibrium models, the pressure on the GR is calculated
models would be needed. This would, however, conict with the by considering the equilibrium of the arch. The models of Hewlett
initial principle of keeping as closely as possible to existing design and Randolph (1988) and Zaeske (2001), which are in widespread
models, and it is beyond the scope of this paper. A work-around is use, are shown in Figs. 2 and 3. These two models give quite
presented in section 4.2: the development of concentric arches. satisfactory results when used in a design (the predicted loading on
the GR is reasonable and on the safe side), but (1) do not explain the
3.6. Summation of section 3 concentration of load on the GR strip, (2) do not explain or derive an
inverse triangular load distribution on the GR strips and (3) do not
Comparing the existing models with measurements, it can be give increasing arching during subsoil consolidation. Furthermore,
concluded that none of the analytical models considered (equilib- the Hewlett and Randolph model is not meant for arching with GR,
rium, frictional, empirical) can explain the measurements. In and not particularly suitable for partial arching situations, which
several cases, they under-predict the arching A measured in the are situations where the ll or embankment is thinner than the full
eld. They do not describe the load and strain localisation on and in arch height, in other words when H < sd/2. The Zaeske model can
work with these low embankments.
normalised measured total B+C Q= 2 (B+C)average (kPa) This section introduces a new model, the concentric hemi-
local measured pressure
on GR / Q (kPa / kPa)
(distributed in a inverse
normalised spheres model. This model accounts for increased arching with
1.0 triangular shape, kPa)
measured B+C, subsoil consolidation and nds load localisation on the GR strips.
0.8
expressed as Furthermore, it gives a physical explanation for the inverse trian-
0.6 average pressure gular load distribution and is therefore a better match for the ob-
0.4 on GR strips servations in section 3.
0.2 (B+C)average
However, in practical applications, a limit-state version of the
(kPa)
model will be applied. In that case, the concentric arches model
K4; Q=161 kPa K7; Q=260 kPa behaves in a rigid-plastic way and will no longer describe the in-
K5; Q=265 kPa K8; Q=240 kPa uence of subsoil consolidation or deformation.
K6; Q=239 kPa K9; Q=265 kPa It should be noted that in this paper a hemisphere is a 3D arch as
indicated in Fig. 13, and an arch is a 2D arch, as indicated in Fig. 12.
Fig. 7. Measured load distribution on GR in newly presented tests (Table 1) compared
with the total measured load B C on the GR, presented as an equally distributed load 4.2. Development of concentric arches
and with the inverse triangular load distribution of Van Eekelen et al. (2012a,b). The
total measured B C (kN/pile) was measured with the transducers that measure A,
A B and C indicated in Fig. 4. These values were divided by the area of a GR strip to
Fig. 9 introduces a 2D picture of the new model that describes
get the equally distributed load (B C)average in kPa. The local pressures were the development of arching during subsoil consolidation, accom-
measured in kPa with the extra TPCs indicated in Fig. 4. panied by an increasing GR deection.
20 measured 16
consolidation
top load step
top load step

consolidation
consolidation
top load step
top load step
top load st.

consoli-
dation
consolidation
consolidation
consolidation
EBGEO
measured arching A (kN/pile)

18
14
K2, arching A (kN/pile)

16
12
14
12 10
10 8
8
6
consolidation
top load step
6
top load st.

consolid.
4
4
granular T3
2 2
sand T2
0 0
-5 0 5 10 15 20 25 -5 0 5 10 15 20
net load Wn (kN/pile) net load Wn (kN/pile)
Fig. 8. Measurements showing that arching increases with subsoil consolidation (a): Comparison of measured and EBGEO calculations for test K2. (b): The increase in arching is
larger for the stronger material granular (test T3) than for sand (Test T2, source: Van Eekelen et al., 2012a, see Table 1 for specications of tests K2, T2 and T3).
In Fig. 9a, a small GR deection results in the start of arch for- essential in this model because, without GR, there will be a more or
mation at the edge of the pile cap. At this location (the edge of the less even settlement of the area between the piles and the
pile cap), the differential settlement between GR and pile cap is at a concentric arches cannot develop, as shown with 2D experiments
maximum and the load starts to be attracted to the stiffer pile cap, by for example Hong et al. (2007) and Jenck et al. (2009).
resulting in an increasing pile load A. Subsequently, increasing GR The development of arching in a basal reinforced piled
deection closes the arch (b). embankment has never been observed through, for example, a
Now, the piece of GR close to the pile behaves in a relatively stiff glass wall. However, the formation of subsequent new concentric
way because it is attached to the pile and can move less freely than arches as a result of settlement underground has been observed in
the GR in the middle. Another arch therefore starts to develop in- experiments at the University of Cambridge (Casarin, 2011). In
side the rst one (c). After this, more arches develop, each one these experiments, sand was poured onto a rubber tunnel. The
smaller than the preceding one (d). Each smaller arch exerts a largest differential settlements started, in this case, in the centre of
smaller force on its subsurface than the preceding larger arch. The the tunnel. In that case, a small arch in the ll occurred rst, fol-
arches give the directions of the main principal stresses: the major lowed by a succession of larger arches.
principal stress in the tangential direction and the minor principal The theory that base deection results in concentric arches has
stress in the radial direction. also been stated by several authors presenting numerical analyses.
The creation of new arches is accompanied by increasing load For example, Han et al. (2012) carried out 2D DEM piled embank-
transfer in the direction of the piles and a reduction of the load on ment analysis and showed (in their Fig. 11) force chains that
the GR area between the piles. This results in a more or less inverse resemble concentric arches, with smaller forces in the smaller
triangular load distribution on the GR strip. arches. Vermeer et al. (2001) found main stress directions following
The process of arch development terminates in a set of concentric arches when they studied the horizontal stress distri-
concentric hemispheres which Fig. 10 shows in 3D. The GR is bution in the soil behind a soldier pile wall. A soldier pile wall of this
a b
c d
Fig. 9. Increasing GR defection results in an increasing lateral transport of load via concentric arch-shaped stress paths and an inverse triangular load distribution on the GR.
Fig. 10. New proposed analytical model: the concentric arches model. The load is transferred along the concentric 3D hemispheres towards the GR strips and then via the concentric
2D arches towards the pile caps.
kind consists of relatively weak timber laggings (comparable with The new proposed model assumes that 3D concentric arches
GR) between stiff anchored steel piles (comparable with piles). (hemispheres) are formed above the square between each four piles
Another example is Nadukuru and Michalowski (2012), who (Figs.10 and 11b). These hemispheres transfer the load outward in all
carried out discrete element simulations. After a wedge-shaped directions along the hemispheres towards the GR strips. The process
heap of particles was formed, a basal deection was prescribed. continues with the further transfer of the load along the 2D arches
In this way, maximum subsidence at the centre equal to 0.67% of above the GR strips towards the pile caps (Figs. 10 and 11c). Both the
the heaps height was observed. Nadukuru and Michalowski 3D hemispheres and the 2D arches exert a load on the GR subsurface
showed their calculated force chains in the particles. The force which increases towards the outside. The part of the load not resting
chains follow the shapes of concentric arches. They also on the GR is arching A, which is the load part transferred directly to
demonstrated that the load on the central part of the base was the pile caps, as explained in Fig. 11a. Fig. 11 depicts the three com-
reduced in the process of deection at the expense of the parts ponents of the model e (a) the load part (arching A) that is applied
farther away from the centre. Each larger arch therefore exerts a directly to the pile caps (Fig. 11a), (b) the load part that is applied on
larger stress on the base. This stress distribution resembles the the GR square between the pile caps diagonally between the pile
inverse stress distribution presented in Van Eekelen et al. caps (Fig. 11b) and (c) the load part that is applied between adjacent
(2012a,b). Sloan (2011) also concluded from his large scale tests pile caps on the GR strips (Fig. 11c) e plus the interaction that must
that it is possible that secondary arches form below the primary exist between the last two components. With this model, it is
arch. His idea is similar to the concentric arches model presented possible to approximate the observed load distribution on the piles
in the present paper. and the entire GR area between the piles.
Fig. 11. Basic idea underlying the proposed concentric arches model: distribution of the load on the GR area between the piles and the determination of arching part
A gH p$Sx $Sy FGRsquare FGRstrips going to the pile directly.
The following sections present the equations for the 2D concen-

tric arches and the 3D concentric hemispheres and these 2D and 3D
equations will then be combined to form the new arching model.

4.3. 2D concentric arches

Figs. 10, 11c and 12 show 2D concentric arches. It should be
noted that this paper states the radius for 3D hemispheres as an
upper-case R, and the radius of 2D arches as a lower-case r. The
radial stress sr in the 2D arch is found by considering the radial
equilibrium of the crown element of the 2D arch and assuming that

the stress state in the arch is uniform around the semi-circle and
that the limit state occurs in the entire arch, which gives the
tangential stress sq Kp$sr. This leads, after some derivation given
in Appendix A.1, to the following tangential stress (in kPa) for a 2D
arch in the x-direction:
sq Px2D $rKp 1 Q2D $r

" #
Kp 1
wherePx2D Kp $H 1Kp $ gHp gHxg2D (1) Fig. 13. 3D Concentric hemispheres, the tangential stress in the arches result in a
xg2D Kp 2 vertical stress exerted on the subsurface.
g
andQ2D Kp $ 4.4. 3D concentric hemispheres
Kp 2
Figs. 10, 11b and 13 show 3D concentric hemispheres. The
and, where Hxg2D (in m) is the height of the largest 2D arch (see
tangential stress (kPa) in the 3D arches is found in a similar way as
Figs. 10 and 12): for the 2D arches, as explained in Appendix A.2:
sx sx
Hxg2D for H full arching sqp0 P3D $R2Kp 1 Q3D $R
2 2
sx " #
Hxg2D H for H < partial arching (2) 22K 2Kp 2
2 whereP3D g$Kp $Hg3D p $ HHg3D $ (3)
2Kp 3
1 sin 4
and Kp g
1 sin 4 andQ3D Kp $
2Kp 3
The explanation of these equations is similar to Hewlett and
Randolphs 2D explanation (1988), except that soil weight and and where Kp is given by Eq. (2) and Hg3D (m) by:
top load are taken into account here. Hewlett and Randolph limit
the thickness of their arch and therefore the crown element to half sd s
Hg3D for H d full arching
the width of a pile cap and calculate the vertical (radial) stress 2 2
(4)
immediately below the crown element. For the concentric arches s
Hg3D H for H < d partial arching
model, however, the arch is extended downwards towards the 2
subsoil, resulting in a set of concentric arches. These arches exert a
force on their subsurface, which comprises the GR strips, in the case With surcharge load p (kPa) the tangential stress (kPa) becomes
of the 2D arches, as shown in Fig. 11c. The larger the arch, the larger (analogous to Hewlett and Randolph, 1988; Zaeske, 2001):
the exerted force. This imprint of load on the GR strip is shown in
gH z p
Fig. 12 and resembles the inverse triangle described in section 3. sq P3D $R2Kp 1 Q3D $R (5)
gH z
where z is the vertical distance between the considered point and

the GR. So far, the explanation of the 3D equations is the same as
Hewlett and Randolphs (1988). Hewlett and Randolph now limit
the thickness of the arch and therefore the crown element to half
the diagonal of a pile cap and calculate the vertical (radial) stress
immediately below the crown element which gives Hewlett and
Randolphs equation (10) (1988).
In the concentric hemispheres model, however, the arch is
extended downwards towards the subsoil, resulting in a set of
concentric hemispheres. These hemispheres exert a force on their
subsurface. The larger the radius, the larger the force exerted on the
subsurface.
4.5. Concentric arches model: combination of 2D arches and 3D

hemispheres
Fig. 12. 2D Concentric arches, the tangential stress in the arches result in a vertical The new calculation model is derived in Appendix A and sum-
stress exerted on the subsurface that resembles the simplied inverse triangle. marised in this section. Note that the equations in the appendix are
for a situation in which the piles are placed in a grid with the same of the square upon which the hemispheres exert a load, as indicated
centre-to-centre distance in both directions: sx sy, while this in Fig. 22.
section extends the equations for the situation sx s sy. The model When the area between the four piles (sx a)$(sy a) > L2x3D, the
should be applied as follows: area outside Lx3D but inside the GR square is assumed to be loaded
by gH p. This gives an extra term, FGRsq3p0, where
1. Determine the total vertical load FGRsquare (in kN/pile) exerted
FGRsq3p0 gH$ sx a$ sy a L2x3D for L2x3D <sx a$ sy a
by the 3D hemispheres on their square subsurface (Fig. 11b).
This load FGRsquare is derived by integrating the tangential stress FGRsq3p0 0 for L2x3D sx a$ sy a
of the 3D hemispheres over the area of this square (see (9)
Appendix A.3 in Eq. (53)e(85)), resulting in:
The load that does not rest on the GR square is supposed to be
transferred to the ring of GR strips and pile caps. This load is
FGRsquare FGRsquare1 FGRsquare2 FGRsquare3 therefore applied as an equally distributed surcharge load on the

gH p 2D arches. This surcharge load on the 2D arches is in kN/pile:
$ FGRsq1p0 FGRsq2p0 FGRsq3p0 (6)
gH
Ftransferred gH$sx a$ sy a FGRsq1p0 FGRsq2p0
where
FGRsq3p0 (10)
3
pP3D Lx3D 2Kp 2 L
FGRsq1p0 $ pQ3D $ x3D Distributed equally on the 2D arches, this results in a surcharge load
Kp 2 3 2
in kPa (2 full GR strips and a pile cap per pile):
FGRsq2p0 1 FGRsq2 2 FGRsq2 3 FGRsq2 4 FGRsq2 F

ptransferred transferred (11)
a$ Lx2D Ly2D a2
where
where Lx2D is the length of the part of the GR strip upon which the

2pP3D Kp Lx3D 2Kp 2D arches exert their force, as indicated in Fig. 23:
1 F GRsq2 2 1
2Kp 2
2pQ3D p3

L 3 1
x3D Lx2D sx a for H sx a
2 F GRsq2 2 1 2
3 2
!! 1
2K
Kp 1 Lx2D 2$Hxg2D for H < sx a
P3D $222Kp $Lx3Dp p P
N 1 2
3 F GRsq2 $ (12)
Kp 22Kp n0 2n 1 n 1
Ly2D sy a for H sy a
2K 2
P3D $2 22K p $Lx3D p
p
$ 2K 1 13 Kp 1 1
Kp 2 p Ly2D 2$Hyg2D for H < sy a

1 K 1 K 2
2
10 p p

1 K 1 K 2 K 3
and
42 p p p

1 K 1 K 2 K 3 K 4
sx sx
216 p p p p Hxg2D for H full arching
2 2

1 :::: K 5 :::

1320 p sx
p Hxg2D H for H < partial arching
p 2
4 F GRsq2 16 Q3D L3x3D $ 21 p ln 1 2 (13)
sy sy
where Hyg2D for H full arching
2 2

sy
22K 2Kp 2 Hyg2D H for H < partial arching
P3D g$Kp $Hg3D p $ H Hg3D $ and 2
2Kp 3
g
Q3D Kp $ 7
2Kp 3 2. Determine the total load FGRstrips (kN/pile) on the GR strips.
FGRstrips is derived by integrating the tangential load of the 2D
Hg3D (m) is the height of the largest hemisphere given in Eq. (4) and arches over the area of the GR strips (see Appendix A.3 Eq.
Fig. 10 and Lx3D is given by: (86)e(91)), resulting in:
q
2 q
2
Lx3D p1
sx a2 sy a for H 12 sx a2 sy a
2 K p
p q
2
gH p P 1 1
FGRstrip;p>0 , 2a x2D L aQ2D ,Lx2D 2
Lx3D 2$Hg3D for H < 12 sx a2 sy a gH Kp 2 x2D 4
Kp
(8) Py2D 1
FxGRstr2p0 2a L
Kp 2 y2D
where a (m) is the width of a square pile cap or the equivalent
1 2
width of a circular pile cap and FGRsquare given in kN/pile. FGRsquare1 aQ2D , Ly2D FyGRstr2p0
and FGRsquare2 (kN/pile) are indicated in Fig. 22. Lx3D (m) is the width 4
4.6. Line-shaped foundations

where

1Kp Kp 1 The concentric 2D arches also apply in the 2D situation where

Px2D Kp ,Hxg2D , gH ptransferred gHxg2D ,
Kp 2 the embankment is supported by line-shaped supporting elements,

1K p Kp 1 such as diaphragm walls or beams. The equations for this case are
Py2D Kp ,Hyg2D , gH ptransferred gHyg2D , given in Appendix A.4.
Kp 2
g When a 2D situation with line-shaped foundation is compared to
Q2D Kp , its corresponding 3D situation with square pile caps (same centre-to-
Kp 2
centre distance, same width of square pile caps and line foundation),
1 the resulting average pressure on the GR is lower for the 2D situation
FxGRstr2p0 gHasx a Lx2D sx a for H <
2 than for the 3D situation. This is different from the model of Zaeske,
1 that nds a slightly higher average pressure on the GR in the 2D case.
FxGRstr2p0 0 for H sx a
2
1
FyGRstr2p0 gHa sy a Ly2D for H < sy a
2 5. Comparison with laboratory experiments, eld tests and
1 numerical calculations
FyGRstr2p0 0 for H sy a (14)
2
5.1. Introduction
and where Lx2D and Ly2D are given in Eq. (12). A practical limi-
tation is a minimum embankment height of H 0.5sx and Step 1 of the calculation model calculates the arching expressed
H 0.5sy so that the 2D arches always have enough height to in arching A, and also calculates the load distribution on the
develop fully. Furthermore, the largest 2D arches are wide GR subsoil. Both results of step 1 should concur with measurements,
enough to rest on the pile caps, as the width of the largest arch is including the observed dependency on ll height and friction angle.
equal to sx or sy. Fig. 14 shows that this requirement is not The calculations for this section have been carried out with the
additional to EBGEO (2010) and CUR226 (2010) and only in- newly presented concentric-arches model described in section 4,
creases the minimum height in a limited, less realistic, number of EBGEO (2010) and Hewlett and Randolph (1988), all without safety
cases for the British Standard (BS8006, 2010). However, the factors.
equations for the case H < 0.5sx,y are stated here for calculations Most of the presented experimental results are measurements
in the construction phase. during a minimum of subsoil support, in other words just before a
top load increase in the experiments described in Section 2.
3. Determine the load distribution. The part transferred to the
piles directly (arching A in kN/pile) is:
5.2. Arching A
A Fpile gH p$sx $sy FGRsquare FGRstrips (15) Fig. 15 compares measured and calculated arching A of the ex-
periments described in Section 2. Calculations that agree exactly
with a measurement are located on the dotted diagonal line. The
as indicated in Fig. 11. The total load resting on GR subsoil is gure shows that the measurements agree much better with the
therefore: concentric arches model than with EBGEO.
The concentric arches model has a better match than EBGEO for
B C FGRsquare FGRstrip (16) the embankments with a ll height of 0.34e0.42 m (K2, K5, K6, K7,
T2 and T3, thus for H/(sd d) 0.51e0.62). EBGEO under-predicts
Calculation step 2 derives the GR strain from this load part B C these measurements much more than the concentric arches model.
(Van Eekelen et al., 2012b). Appendix B gives a calculation example These heights are important in design when making calculations
using the concentric arches model for step 1 and the inverse trian- for the construction phase. For the thinnest ll, K8, however, where
gular load distribution for step 2 following Van Eekelen et al. (2012b). H/(sd d) 0.33, neither model works properly, but remains on the
3.0 2.5
pile cap width a = 0.5 m centre-to-centre distance
piles sx = 2.25 m
minimum embankment height H (m)
minimum embankment height H (m)
2.5
2.0
2.0
1.5
1.5
1.0
1.0
BS8006 H0.7(sx-a)
0.5 EBGEO H0.8(sd-d)
0.5
CUR226 H0.66(sd-d)
concentric arches H0.5sx
0.0 0.0
1.0 1.5 2.0 2.5 3.0 0.25 0.50 0.75 1.00
centre-to-centre distance piles sx (m) width of pile cap a (m)
Fig. 14. Comparison of required minimum embankment heights for the new concentric arches model, BS8006 (2010), EBGEO (2010) and CUR226 (2010).
16 K2 conc. arches
K2; EBGEO 70%
14 K4; conc. arches
K4
calculated arching A (kN/pile)
arching A (percentage oftotal load, %)

K4; EBGEO
12 K5; conc. arches 60% K2 K6
K5; EBGEO K5 K9
10 K6; conc. arches
50% K8 K7
K6; EBGEO measured
8 concentric arches
K7; conc. arches
6
K7; EBGEO 40% EBGEO 2010
K8; conc. arches Hewlett & Randolph 1988
4 K8; EBGEO CUR 2010 minimum H
K9; conc. arches 30% EBGEO 2010 minimum H
2 K9; EBGEO
T2; conc. arches
0 T2; EBGEO 20%
0 5 10 15 T3; conc. arches
measured arching A (kN/pile) T3; EBGEO
10%
Fig. 15. Comparison of measured and calculated arching A in model experiments
specied in Table 1. 0%
0.0 0.5 1.0 1.5 2.0 2.5
safe side for the GR design: the models predict more load on the
H/(sd-d) (-)
GR than measured.
It should be noted that EBGEO predicted the measurements for Fig. 17. Variation of embankment height H. Comparison of measured and calculated
the thickest lls (K4 and K9, H/(sd d) 0.97) very well. The arching A% in model experiments specied in Table 1.
concentric arches model, however, gives a slight over-prediction in
cases with this relatively thick, but realistic, ll. However, in the geometry, produce a satisfying match with the results of the
following paragraphs, it is shown that measurements in two full- concentric arches model.
scale eld tests, and numerical calculations with a full-scale Fig. 16 compares full-scale eld measurements with calcula-
tions. The gure shows that EBGEO and the concentric arches
model agree better or equally well with the measurements. This is
=43o, d=0.4m, sx=1.45m, sy=1.90m H=2.60 m, =18.3 kN/m3, design load an important result, as it is dangerous to base the conclusions only
p=26.83 3
200 kPa, subsoil:1m sand, 3m soft clay, 20m sand: k= 310 kN/m on scaled model tests. It should be noted that it is difcult to
determine the friction angle 4 in the ll in the eld, while even
Houten crane (= design load)
160 though 4 has a major inuence on the calculation results in each of
the calculation models, as mentioned before in section 3.2.
arching A (kN/pile)
120
5.3. The inuence of ll height
80
measured pile 292 Figs. 17 and 18 show the inuence of embankment height H.
measured pile 285
40 EBGEO/CUR no trains Fig. 17 compares measurements of the model tests with pre-
EBGEO/CUR design load dictions. The gure shows that the concentric arches model agrees
concentr arches no trains better with the measurements than the other models. The gure
0 concentr arches design load
shows that the measurements indicate that A% increases with
embankment height and seems to stabilise for the higher em-
bankments. This nding concurs with Le Hello and Villards
(a)
=43o, average values geometry:
d=0.85m, sx=sy=2.25 m, H=1.86 m, =18.3 80%
160
kN/m3, subsoil: 17 m soft soil: k= 0 kN/m3
arching A (percentage of total load, %)
70%
Woerden
120
arching A (kN/pile)
60%
80 50%
measured pile 692 a=0.6m, sx=sy=1.5m,
measured pile 693
40% =19 kN/m3, p=0kPa, =29o
40 EBGEO/CUR (2010)
BS8006 (2010) 30% num. calc. Le Hello et al. 2009
Hewlett & Randolph (1988)
concentric arches concentric arches
0 20% Hewlett and Randolph 1988
EBGEO 2010
10% EBGEO 2010 minimum H
CUR 2010 minimum H
(b) 0%
Fig. 16. Comparison of measured and calculated arching A in two eld cases: (a)
0.0 0.5 1.0 1.5 2.0 2.5
railway Houten, Netherlands, described in Van Duijnen et al. (2010) and (b) highway H/(sd-d) (-)
exit Woerden, Netherlands, described in Van Eekelen and Bezuijen (2012c). Prediction
with surcharge load p 0 kPa, see also Appendix B for a calculation example of this Fig. 18. Variation of embankment height H, comparison of analytical models with
case. numerical calculations of Le Hello and Villard (2009).
1800
1400-1600 Test K6
1200-1400 1600
1600
1000-1200 measured
800-1000 1400 1400
EBGEO 2010
pressure on GR strip (kPa)

600-800 1200
400-600 1200 concentric arches
1000
200-400
0-200 800 1000
600
400 800
200
0.275
0 600
0.224
0.178 400
0.132
0.086 200
0.275
0.231
0.191
0.040
0.152
0.112 0
0.073
0.033
0 0.05 0.1 0.15 0.2 0.25
distance from centre between 2 piles (m)
Fig. 19. Load distribution for test K6 for p-Rb 66.1 kPa, which equals 20.0 kN/pile and minimum subsoil support: C 0.5 kN/pile, (see Table 1). Left: load distribution calculated
with concentric model in kPa. Right: cross-section through GR strip and pile. EBGEO gives an equally distributed load, which is translated in calculation step 2 into a triangular load
distribution, which is given here.
numerical calculations (2009). They developed a numerical model load on the GR strips. And the load on the GR strips is concentrated
that combined the 3D discrete element method and the nite near the pile cap in a way resembling the inverse triangular load
element method. They also found increasing arching with ll distribution. The concentric arches obviously explain the observed
height, stabilising for higher embankments. Fig. 18 shows that the concentration of load near the pile caps. The concentric model
concentric arches model is a reasonable match with the numerical agrees better with the observed load distribution than any of the
calculations of Le Hello and Villard. other available analytical models.
5.4. Load distribution on GR 5.5. Ground pressure versus depth
Fig. 19 shows the measured and calculated load distribution on Fig. 20 compares the measured ground pressure (Zaeske, 2001)
the GR strip. The gure presents the actual results from the new with the results of the concentric arches model, EBGEO, and
model. It is suggested that the step 2 calculations suggested in Van Hewlett and Randolph (1988). The gure shows that the concentric
Eekelen et al. (2012b) should be followed for design purposes using arches model over-predicts the fall in ground pressure with depth,
the simplied inverse triangular load distribution. but that it is the only model that more or less follows the measured
The result of step 1 of EBGEO is pressure on a single point of the tendency of falling pressures with depth.
GR subsurface. It is assumed that this pressure is the same
everywhere between the pile caps, not only on the GR strip. This 5.6. Parameter study
load is relatively low. For comparison purposes, the EBGEO pressure
on GR subsurface in Fig. 19 has been concentrated on the GR strip Fig. 21a shows that an increasing friction angle 4 gives
and expressed as the triangular load distribution as used in calcu- increasing arching. The gure shows the measured results for tests
lation step 2 of EBGEO. T2 and T3. These tests are the same, except for the embankment ll,
The gure shows that the measured A agrees well with the A as indicated in the gure and in Table 1. The gure shows that the
calculated with the concentric arches model. It can therefore be concentric arches model is a better t for the measurements than
concluded that the total measured B C per pile also agrees well the others. The inuence of 4 is limited in the Hewlett and Ran-
with the calculated B C, as B C total load A. The gure also dolph model for 4 > 30 . Within this model of Hewlett and Ran-
shows clearly that the concentric arches model concentrates the dolph, the situation of one of two elements is normative: the crown
EBGEO measured, with GR

0.40 Hewlett & Randolph measured; no GR
concentric arches
0.35
vert. position TPC in centre line fill (m)
TPC6
0.30
load transfer plate
0.25 TPC5
0.35 m
TPC6 TPC5
TPC4
0.20
TPC1 TPC2 TPC3 = 38o. a = 0.6
0.15 TPC4
m, sx=sy=0.50 m,
peat H=0.35 m,
or gap 0.10
=18.1kN/m3,
0.05 TPC1 p=104 kPa
22 16
25 12.5 12.5 0.00
0 25 50 75 100 125
average vertical pressure on TPC's in centre line fill (kPa)
Fig. 20. Comparison of analytical calculations with measurements in Zaeske (2001, p. 63). For comparison reasons, measurements for the situation with and without GR are given
(2001, p. 55 and 63), see also Fig. 6.
tests T2: sand fill =40.9o test K5: H=343 m

test T3: granular fill =49.0o 60%

70%

test T3
60% 50%
50%
40%
test T2
40%
30%
30%
20%
20%
10%
10%
0% 0%
15 35 55 -5 15 35 55 75 95
friction angle (o) surcharge load p (kPa)
(a) (b)
test K9: H=0.65 m measured
90% concentric arches
arching A(percentage of total load, %)

120%
EBGEO 2010
80% Hewlett & Randolph 1988
100%
70%
60% 80%
50%
60%
40%
30% 40%
20%
test K9:
20%
10% H=0.65 m
0% 0%
0.00 0.25 0.50 0.75 1.00 1.25 1.50 0.00 0.05 0.10 0.15 0.20 0.25 0.30
centre to centre distance sx pile diameter d (m)
(c) (d)
Fig. 21. Comparison of calculations and measurements in tests T2, T3, K5 and K9. See Table 1 for test specications. Parameter study: variation of (a) friction angle 4, (b) surcharge
load p. The large dots show the situation with a minimum of subsoil support measured just before a surcharge load increase, (c) centre-to-centre distance sx sy and (d) pile
diameter d.
element or the foot element as indicated in Fig. 2. For 4 < 30 , the 6. Conclusions
foot element is normative, for 4 > 30 , the crown element is
normative. For the crown element, the pressure on the subsurface In model tests, numerical studies and eld measurements of
consists of two terms: the radial stress immediately below the arch, geosynthetic reinforced piled embankments, the following features
si, and the soil weight below the arch. For 4>30 , si is so small that were observed:
the soil weight below the arch dominates. Soil weight is indepen- There is a major difference between piled embankments with or
dent of 4 and therefore constant. without GR. GR makes arching much more efcient: the load is
A% in all three models considered is independent of the sur- transferred to the piles much more efciently. With GR, the load on
charge load. This is because the models rst calculate the load the GR is concentrated on the GR strips and can be described
distribution for the situation without surcharge load (p 0 kPa) approximately as an inverse triangular load distribution on the GR
and then multiply the result by the factor (gH p)/(gH). This is strips. The difference between piled embankments with or without
shown in Fig. 21b. This gure compares the measurements and GR requires a distinction between models describing one or the
calculations for test K5. The large dots are the measurements for other situation.
the situations with a minimum of subsoil support. The gure shows The observed load distribution in the piled embankments with
that the measured A%, for the situation with a minimum of subsoil GR is neither described nor explained by any of the available
support, is indeed more or less constant: the large dots, especially analytical models.
with the higher surcharges, are located more or less on a horizontal EBGEO tends to under-predict arching, although prediction ac-
line. curacy is acceptable in one of the eld tests considered. Arching A%
Fig. 21c and d show variation in the geometric properties; the increases with embankment height and seems to stabilise for the
centre-to-centre distance sx of the piles and the pile diameter d. It is higher embankments. Consolidation of the subsoil results in an
not possible to compare this with the measurements because these increase in arching (increasing arching A). This is different from
features were not varied in the tests. The tendency in the gures results obtained using most of the available calculation models. A
conrms expectations: larger pile spacing gives less arching; larger higher friction angle of the ll gives more arching, especially during
pile cap diameter gives more arching. subsoil consolidation.
A new equilibrium model was presented in this paper: the area below : dAi r$dq
concentric arches model. It is a variation on the Hewlett and top area : dAo r dr$dq r$dq dr$dq
Randolph (1988) and EBGEO (Zaeske, 2001) equilibrium models. (17)
side area : dAs dr
The model consists of a set of concentric hemispheres and arches. volume crown element : dVzr$dq$dr
Larger hemispheres or arches exert more pressure on their
subsurface. For this 2D plane strain situation, these properties (areas
A set of concentric 3D hemispheres transfer their load to a and volume) apply for a unit length in the third dimension.
set of 2D arches between adjacent piles. These 2D arches For example dAi r$dq$1, where the 1 is 1 m. The areas dAi,
transfer their load further to the piles. The model results in dAo and dAs are therefore expressed in m2, and dV in m3. From the
a load distribution on the GR that resembles the load vertical (radial) equilibrium of the crown element, it follows that:
distribution observed in experiments, eld measurements and
numerical analysis: the load is mainly concentrated on the dq
sr $dAi sr dsr $dA0 2$sq $dAs $sin g$dV 0 (18)
GR strips with an approximately inverse triangular load 2
distribution.
The stresses sr and sq are expressed in kPa and g in kN/m3. From
The model therefore provides a satisfactory physical explana-
substituting equations (17), and assuming that sin(dq)zdq, it fol-
tion for this observed load distribution. The concentric stress arches
lows that:
were also found by several authors presenting numerical studies on
arching such as Vermeer et al. (2001), Nadukuru and Michalowski
dq
(2012) and Han et al. (2012). sr $r$dq sr dsr $r$dq dr$dq 2$sq $dr$ g$r$dq$dr
The concentric arches model explains increasing arching with 2
subsoil consolidation (GR deection). The explanation is that 0
new arches are formed in succession as GR deection progresses. (19)
However, in the limit-state version of the model presented in
this paper, the model behaves in a rigid-plastic way and no
longer describes the inuence of subsoil consolidation or dsr dsr sr sq
g 0 (20)
deformation. dr r r
The new model describes both full and partial arching, the latter
with a relatively thin embankment. where sr is expressed in kPa. From dsr =r/0, it follows that:
Agreement between measured arching A and calculations made
with the concentric arches model is good, and generally better than dsr sr sq
the EBGEO/CUR results, especially for relatively thin embankments. g (21)
dr r
This nding is important for design calculations for the construc-
tion phase.
We assume that the stress state in the arch is uniform around the
The concentric arches model is dependent on the embankment
semi-circle and that the limit state occurs in the entire arch.
height and the ll friction angle 4 in a way similar to that found in
the experiments and in the numerical calculations of Le Hello and
1 sin 4
Villard (2009). sq Kp $sr $sr (22)
Ground pressure in the embankment decreases with depth and 1 sin 4
the tendency for decreasing pressure is similar in the Zaeske
measurements (2001) and the concentric arches model. Further- This results in the 2D differential equation for the radial stress in 2D
more, this model matches Zaeskes observations better than any of arches:
the other models considered.

Parameter variation indicates that the response of the concen- dsr 1 Kp
tric arches model to variations of surcharge load and geometry is $sr g (23)
dr r
reasonable.
To solve this differential equation, it can be rewritten as
Acknowledgements dsr
pr$sr qr (24)
dr
The nancial support from Deltares and the manufacturers
Naue, TenCate and Huesker for the research was greatly appreci- where
ated. The test series was carried out as a result of close cooperation

between Deltares and the suppliers TenCate, Huesker, Naue and 1 Kp
Tensar. Their nancial support and fruitful debate have been pr and qr g (25)
r
extremely valuable. The authors are grateful for the useful com-
ments of the reviewers.
R
The left-hand and right-hand sides of Eq. (23) are multiplied by an
integration factor: e prdr . This standard procedure to solve this
kind of differential equation results in:
Appendix A. Derivation of the equations of the concentric
arches model 0 Z 1 Z
dB prdr C prdr
A.1. The 2D arch: radial equilibrium @e $sR A e q r (26)
dr
Consider Fig. 12. The areas and volume of the crown element are
(neglecting terms with a product of more than one increment): We nd:
Z Z
1 Kp 1 1
prdr dr area below : dAi p$R$dq2 p$R2 $dq2
e e r e1Kp ln r r 1Kp (27) 4 4
1
2 2 1 2 2
Thus Eq. (26) becomes: top area : dAo p$R dR $dq z p$ R2 $dq 2R$dR$dq
4 4
side area : dAs zp$R$dq$dR
d 1Kp
r $sr qr$r 1Kp g$r 1Kp (28) 1
dr 2
volume crown element : dVz p$R2 $dq $dR
4
Z (37)
g
/r 1Kp $sr g r 1Kp dr $r 2Kp C (29)
2 Kp
From the vertical (radial) equilibrium of the crown element, it
where C is a constant. Thus follows that:
g r2Kp C
sr $ (30) dq
2 Kp r 1Kp r 1Kp sR $dAi sR dsR $dA0 sq $sin $dAs g$dV 0 (38)
2
thus
From substituting equations (37), and assuming sin(dq) z dq, it
Kp 1 g follows that:
sr C$r $r (31)
Kp 2
1 2 1 2 2

sR $ p$R2 $dq sR dsR $ p$ R2 $dq 2R$dR$dq
For the weightless case we nd equation (3) of Hewlett and 4 4

Randolphs (1988): dq 1 2
sq $ $p$R$dq$dR g$ p$R2 $dq $dR 0 (39)
2 4
Kp 1
sr C$r (32)
From neglecting terms with a product of more than one increment,
The boundary condition on the outside of the 2D arch is: it follows that:
dsR 2sR sq
r Hg2D /sr g H Hg2D p (33) g 0 (40)
dR R
where Hg2D is the height of the largest 2D arch and given by Eq. (2). In the weightless case, the stress state in the arch is uniform around
Substitution of this condition into Eq. (31) gives: the semi-circle. It is assumed that the limit state occurs in the entire
arch.
K 1 g
C$ Hg2D p $Hg2D g H Hg2D p 1 sin 4
Kp 2 sq Kp $sR $s (41)
"
# 1 sin 4 R
(34)
1Kp Kp 1
/C Hg2D $ gH p gHg2D This results in:
Kp 2

dsR 2 1 Kp
thus $sR g (42)
dR R
Kp 1
2 sx g Kp 1 g which is the differential equation for the radial stress in the 3D

/sr $ gH p $r Kp 1 $r hemisphere. So far, the explanation is the same as Hewlett and
sx 2 Kp 2 Kp 2
Randolphs (1988). To solve this differential equation, Eq. (23) can
(35) be rewritten as:
As the tangential stress sq Kp$sr in kPa, we nd: 0 Z 1 Z
d B pRdR C pRdR
sq P2D $rKp 1 Q2D $r (36) @e $sR A e qR (43)
dR
where we have dened P2D and Q2D as:

with:
Kp 1

2 sx g Kp 1 2 1 Kp a
P2D Kp $ $ gH p and p R and q R g
sx 2 Kp 2 R R (44)
g
Q2D Kp $ and a 2 1 Kp
Kp 2
We nd:
Z Z
A.2. The 3D arch: radial equilibrium a
pRdR dR
e e R ealn R Ra (45)
Consider Fig. 13. The areas and volume of the crown element are
(neglecting terms with a product of more than one increment): Thus Eq. (43) becomes:

d a 22K 2Kp 2
R $sR qR$Ra g$Ra (46) P3D g$Kp $Hg3D p $ H Hg3D $ and
dR 2Kp 3
g
Q3D Kp $
2Kp 3
g
/Ra $sR $Ra1 C (47)
a1
So far, Hewlett and Randolph used the same explanation. However,
a difference is that they limit the height of their arch to half the
where C is a constant. Thus
width of the pile cap, while the equation of the concentric arches
model is extended downwards to the subsurface where R 0.
g C Hewlett and Randolph (1988) therefore use Eq. (52) for Ri < R < Ro,
sR $R a
a1 R and we use the equation for 0 < R < Ro, where for full arching:
(48)
g Ro Hg3D 0.5sd and Ri 0.5(sd d).
/sR C$R2Kp 1 $R
2Kp 3
The outer radius of the hemisphere is Ro sx/O2. The surcharge load

will rst be neglected and taken into account afterwards, analogous A.3. Derivation of load exerted on GR
to Zaeske (2001) and Hewlett and Randolph (1988). The boundary
condition on the outside of the arch is: Arching is assumed to transfer the load in two steps. The
rst step is that the load is transferred in the direction of the
R Ro Hg3D /sR g H Hg3D (49) ring of GR strips and pile caps. This is done along the 3D
hemispheres. The second step is that the load is transferred
where the arch height Hg3D is given by Eq. (4). Substitution of this further in the direction of the pile caps along the 2D arches
condition into Eq. (48) gives: between each two adjacent pile caps. The 2D and 3D arches
2Kp 1 g exert a total force on the subsurface. The total force exerted on
C$Hg3D $Hg3D g H Hg3D the GR (B C in kN/pile) may be obtained by integrating the
2Kp 3
tangential stress sq across the area of the GR. The general
" # (50) equation is:
22Kp 2Kp 2
/C g$Hg3D $ H Hg3D $ Z
2Kp 3
B C FGR sq dAGR (53)
thus GRarea

The total load B C on the subsurface, FGR, consists of two parts
22K 2Kp 2 g
/sR g$Hg3D p $ H Hg3D $ $R2Kp 2 $R along the two arching steps described above: FGRsquare and FGRstrips:
2Kp 3 2Kp 3
(51) 1. 3D hemispheres transfer the load in the direction of the ring of
GR strips and pile caps. The 3D hemispheres exert a vertical
As sq Kp sr, we nd: load on their GRsquare (indicated in Fig. 11). This is the rst
load part, FGRsquare, of the load on the subsurface.
2. The 2D arches above the GR strips transfer the load further in
sq P3D $R2Kp 2 Q3D $R (52) the direction of the pile caps. The 2D arches exert a vertical load
on the GR strips. This is the second load part, FGRstrips, of the
where we have dened P3D and Q3D as: load on the subsurface.
4 parts of FGRsquare2
Ru1= (Lx3D)/2
Ru1
Ru2
Ru2= (Lx3D R R
dR

FGRsquare1 FGRsquare1
dR area dA1 = RdR FGRsquare3
area dA2 = RdR a or Bers
Fig. 22. Integrating the tangential stress sq across the square area where the 3D hemispheres exert their load.
Total force on the GR square; determination ofpFGRsquare

Consider Fig. 22. As long as H sx a= 2, FGRsquare consists ZRu1 ZRu1
1
of the load exerted by the hemispheres only. In that case (on the FGRsq1p0 4 sq dAGR 4 sq pR dR
2
left of Fig. 22), the total force exerted on the GRsquare may be 0 0
obtained by integrating the tangential stress sq in p the 3D hemi- ZRu1
spheres across the GRsquare. When H < sx a= 2, the total 2p P3D $R2Kp 1 Q3D $R2 dR
force exerted on the GRsquare may be obtained by integrating the
0
tangential stress sq across the smaller (Lx3D$Lx3D-)square, with
width Lx3D, as indicated on the right of Fig. 22. The load FGRsq3 on
Ru1 Lx3D
P3D 2Kp 1 3 2
the area outside this Lx3D$Lx3D-square and inside the GRsquare is FGRsq1p0 2p $R Q3D $R
2Kp 3
assumed to be gH p, which will be derived for later on. Lx3D is 0
3
dened as follows: p pP3D Lx3D 2Kp 2 L
FGRsquare1 1 $ $ pQ3D $ x3D
gH Kp 2 3 2
sx a (57)
Lx3D sx a for H p
2
p (54)
sx a where sq, P3D and Q3D are given in Eq. (52). The second load
Lx3D 2$Hg3D for H < p
2 part, FGRsquare2, which rests on the area within the square but
outside the circle, is dependent on angle a. This angle a is
a function of R, and can be read from Fig. 22, as the comple-
where Hg3D in dened in Eq. (4).
For the situation in which sx s sy the integration is carried R L
mentary angle u arccos u1 arccos x3D . We therefore
out for an imaginary square with width Lx3D. This width is R 2R
determined as: nd, for a:

p L
a 2arccos x3D (58)
q 2 2R
2 q
2
L3D p1 sx a2 sy a for H 12 sx a2 sy a
2
p q
2 The force on the grey areas in Fig. 22 (outside the circle, inside the
L3D 2$Hg3D for H < 12 sx a2 sy a GR square) should therefore be determined by:
(55)
q
Considering the situation that H 1=2 sx a2 sy a2 , FGRsquare2 1 gpH $FGRsq2p0
the square with width L3D has the same diagonal as the area be-
tween the four piles. Integrating the tangential stress across this where
imaginary square gives the same or a higher force on the GR square ZRu2 ZRu2
than numerical integration of the tangential stress over the rect- FGRsq2p0 4 sq dAGRsq2 4 sq aR dR
angular between the four piles. In the remainder of this appendix,
Ru1 Ru1
the situation with sx sy applies.
The total force FGRsquare on the GR square is derived by dividing ZRu2
the square into three sections: 4 P3D $R2Kp 2 Q3D $R aR dR
Ru1
- part FGRsquare1, inside the largest circle in the Lx3D$Lx3D-square
ZRu2
(white in Fig. 22).
/FGRsq2p0 4 P3D $R2Kp 2 Q3D $R
- part FGRsquare2, outside the circle but inside the Lx3D$Lx3D-square
(dark grey Fig. 22). Ru1

- part FGRsquare3, outside the Lx3D$Lx3D-square but inside the GR p L
2 arccos x3D R dR (59)
square (light grey in the right-hand gure in Fig. 22). 2 2R
The three terms are calculated with surcharge load p 0, and

will afterwards be multiplied by the term gH p=gH to nd the
total load on the GR square:
This integral can be separated into four terms, which will be solved
FGRsquare FGRsquare1 FGRsquare2 FGRsquare3 separately:

gH p
$ FGRsq1p0 FGRsq2p0 FGRsq3p0
gH
(56) FGRsq2p0 1 F GRsq2 2 F GRsq2 3 F GRsq2 4 F GRsq2
The rst load part, FGRsquare1, which rests on the circular area in the
largest circle, is determined as follows: where
p
ZRu2 1
Z 2
2Kp 1 3$4 F GRsq2
2

1 F GRsq2 2pP3D R dR arccos rd r3 (65)
Ru1 Q3D Lx3D 3
1
ZRu2
Partial integration gives:
2 F GRsq2 2pQ3D R2 dR
Ru1
(60) 1 p

1 p2
2
ZRu2 3$4 F GRsq2 Z
2
L
3 F GRsq2 8P3D R2Kp 1 arccos x3D dR r3 arccos r2 r3 darccos r (66)
2R Q3D Lx3D 3 1
Ru1 1
ZRu2 1
As d=drarccos r p, we nd:
2 Lx3D
4 F GRsq2 8Q3D R arccos dR 1 r2
2R
Ru1 p
1
Z
2
2
3$4 F GRsq2 p r3
The rst two terms are solved as follows (Ru2 and Ru1 are indicated p pdr (67)
in Fig. 22): Q3D Lx3D 3 2 1 r2
1
We substitute:
2pP3D 2Kp Ru2 pP3D Kp L 2Kp
x3D
1 F GRsq2 R 2 1
2Kp Ru1 Kp 2 p
(61) 1 u2 1
p
2pQ3D 3 Ru2 2pQ3D p3 L 3 1 r2 u
x3D
2 F GRsq2 R 2 1
3 Ru1 3 2
/r2 u2 1
The other two integral terms are re-written as:
/ r3 dr u du
ZRu2 v (68)
3 F GRsq2 Lx3D u 1
R2Kp 1 arccos dR u u
t p2 1 1
8P3D 2R
Ru1 r 1 p
1 2
Z 2 2 2
Z
(62)
/ dr du
ZRu2 s
4 F GRsq2 L r 1
R2 arccos x3D dR 1
8Q3D 2R u 1 0
Ru1 12
We continue with solving the fourth term, 4 F GRsq2 , and substitute: This gives:
Lx3D L L Z p 1
r thus R x3D and dR x3D dr 3$4 F GRsq2 p
2R 2r 2r2 p u2 1 du (69)
Q3D Lx3D 3 2
Lx3D L 1 p 0
Lx3D x3D 2
Ru2 p 2Ru2 2$
Lp
x3D 2
Z 2 Z 2 (63) This gives, with any integral table:
/ dR dr
Ru1
Lx3D Lx3D L
x3D 1
3$4 F GRsq2 p 1 p 1 p
p 2 ln 1 2 (70)
2 2Ru1 2$Lx3D
2 Q3D Lx3D 3 2 2 2

Lx3D
where r is the cosine of the complementary angle u arccos 2R Thus
in Fig. 22. Thus
1 p p
4 F GRsq2 Q3D Lx3D 3 $ 21 p ln 1 2 (71)
1 p
6
Z 2
2 2
4 F GRsq2 Lx3D Lx3D
arccos r dr The derivation of the third term, 3 F GRsq2 , follows the same
8Q3D 2r 2r2
1 procedure as for 4 F GRsq2. With Eqs. (63) and (62) we nd:
(64)
1 p
p
Z 2
1
2 Z 2
2Kp 3
3 F GRsq2 $2
2
4 F GRsq2
/ r4 arccos rdr r2Kp 1 arccos r dr (72)
Q3D Lx3D 3 P3D $Lx3D 2Kp
1 1
We bring the power to the increment We bring the power to the increment:
N
1 X Kp 1 Kp 1 Kp 2 4
r2Kp 1 dr2Kp u2n 1 Kp 1 u2 u
2Kp n 2!
n0
p
1
Z 2 (73) Kp 1 Kp 2 Kp 3 6
2Kp $3 F GRsq2 $22Kp 3
2
u /
/ arccos r dr2Kp 3!
P3D $Lx3D 2Kp (79)
1
From this, it follows that:

Partial integration:
1
Z1 Kp 1 XN
p 1 Kp 1
p
1
Z 2 u2 1 du u2n1
Kp $3 F GRsq2 $22Kp 2 1
2
2
2n 1 n 0 n
2Kp
r arccos r j 2
r2Kp darccos r 0 0
P3D $Lx3D 2Kp
1
X
N
1 1 Kp 1
(80)
(74) n0
2n 1 n
thus
Z1 Kp 1
Kp $3 FGRsq2 $22Kp 2 p
/ 2Kp
u2 1 du
p P3D $Lx3D 22Kp
1
2 0
Kp $3 F GRsq2 $22Kp 2 Z
2

p r2Kp kp $3 FGRsq2 $22Kp 2 p X
N
1 Kp 1
pdr (75)
P3D $Lx3D 2Kp 22Kp 1 r2 P3D $Lx3D 2Kp 22Kp 2n 1 n
1 n0
P3D $222Kp $Lx3D 2Kp

/3 F GRsq2 $
Finally, the substitution of: Kp
!
p XN
1 Kp 1
p
1 u2 1 22Kp n 0 2n 1 n
p
1 r2 u P $222Kp $Lx3D 2Kp p 1
3D $ 2K 1 Kp 1
/r2 u2 1/ r3 dr u du Kp 2 p 3
K p 1 1
Kp 1 Kp 2 Kp 1 Kp 2 Kp 3
/r2Kp u2 1 10 42
3 1
Kp 1 Kp 2 Kp 3 Kp 4
/r3 u2 1
2
216
(76) 1
3 Kp 1 Kp 2 Kp 3 Kp 4 Kp 5
3 2 2 1320
/dr ur du u u 1 du
1
r Kp 1 Kp 2 Kp 3 Kp 4 Kp 5 Kp 6
9360
p u 1
p 2 1 1
r Z 12 2 12 2 //
Z
/ dr du
q (81)
r1
u 1
1 0
12
The resulting total load on the GR square for the case sxssy is given
by Eq. (6) of this paper.
gives: An alternative is to solve this numerically. This means that the
GR square is divided into a large number of small areas. The average
Z1 p
vertical stress on each increment is determined and multiplied by
Kp $3 FGRsq2 $22Kp 2 p u2 1 2 Kp 3 its area. The sum gives the total force on the GR square,
u 1 u u2 1
2
2K du
P3D $Lx3D 2Kp 2 p u with FGRsquare in kN/pile. FGRsquare1 and FGRsquare2 (kN/pile) are
0 indicated in Fig. 22, where Hg3D (m) is the height of the largest
Z1 Kp 1 hemisphere and Lx3D is the width of the square upon which the
Kp $3 FGRsq2 $22Kp 2 p hemispheres exert a load, as indicated in Fig. 22. When
/ u2 1 du
P3D $Lx3D 2Kp 22Kp Lx3D < (sx a), the area outside Lx3D but inside the GR square is
0
assumed to be loaded by gH p. This gives an extra term, FGRsq3p0,
(77) where
This last term can be rewritten with a binomial series:
FGRsq3p0 gH$ sx a2 L2x3D for Lx3D < sx a
(82)
FGRsq3p0 0 for Lx3D sx a
Z1 Kp 1 N
Z1 X
Kp 1
u2 1 du u2n du (78) Note that it is assumed that sx sy in this appendix. In section
n
0 0
n0 4.5 the equations are given for sx s sy. It is supposed that the load
not resting on the GR square is transferred to the ring of GR strips
where and pile caps. This load is therefore applied as an equally
distributed surcharge load on the 2D arches. This surcharge load on FGRstrip FGRstrip 1 FGRstrip 2
the 2D arches is, in kN/pile:
gH p
$ FGRstr 1 p0 FGRstr 2 p0 (87)
gH
Ftransferred gH$sx a2 FGRsq1p0 FGRsq2p0 FGRsq3p0
(83) where
Zr0 Zr0
where a is the width of a square pile cap or the equivalent width of
a circular pile cap. Distributed equally on the 2D arches, this results FGRstr1p0 4 sq dAGRstrip 4 sq $a dr
in a surcharge load in kPa (2 full GR strips and a pile cap per pile): 0 0
1
Z L
Ftransferred
2 x2D

ptransferred (84) 4a P2D $r Kp 1 Q2D $r dr (88)
2a$Lx2D a2
0
thus
where
1 Lx2D
P 1 2
1 s FGRstr1p0 4a 2D $r Kp Q2D $r 2
Lx2D sx a for H 2 x a Kp 2 0
(85) Kp (89)
Lx2D 2$Hg2D for H < 1 s a
2 x P2D 1 1
/FGRstr1p0 a$ 4 Lx2D Q2D $L2x2D
Kp 2 2
where Hg2D in dened in Eq. (13).
and
Total force on the GR strip; determination of FGRstrip
In this appendix, for reasons of clarity, it is assumed that sx sy.
FGRstr2p0 2gHasx a Lx2D for H < 12 sx a
The tangential stress sq (kPa) in the 2D arches above the GR strips is (90)
now given by: FGRstr2p0 0 for H 12 sx a
sq P2D $rKp 1 Q2D $r with surcharge load p > 0:
(2D arch), where

Kp
gH p P 1 1
FGRstrip;p>0 $ 4a 2D L aQ2D $L2x2D

gH Kp 2 x2D 2
1Kp Kp 1
P2D Kp $Hg2D $ gH ptransferred gHg2D $
Kp 2 FGRstr2p0
g
Q2D Kp $ (91)
Kp 2 For the situation sxssy this equations changes into Eq. (14).
(86)
where ptransferred (kPa) is the load transferred from the 3D hemi- A.4. 2D variant: line-shaped support
spheres to the 2D arches and given by Eq. (84) and Hg2D (m) is
the height of the 2D arch and given by Eq. (13). When sxssy, a The 2D equations can be worked out easily for the situation in
distinction should be made between Px2D versus Py2D and Hxg2D which the embankment is supported by line-shaped supporting
versus Hyg2D. The rest of the equation is equal to Eq. (1) in this elements (such as diaphragm walls). In this case, Eq. (86) is appli-
paper. The total force on the GR strip (without surcharge load p) cable, where ptransferred 0 kPa, as this is the load transferred from
may be obtained by integrating the tangential stress sq across the the 3D hemispheres that do not exist in the 2D case.
area of the GR strip as shown in Fig. 23. The total force on two GR The total load in kN/m0 on a 1-m-wide line foundation is (in
strips (in kN/pile) is therefore: accordance with Eq. (87)):
FGRstrip2
FGRstrip1 FGRstrip1
area dA = adr area dA = adr
Fig. 23. Integrating the tangential stress sq across the area of the GR strip where the 2D arches exert their load.
FGR line FGR line1 FGR line2 Determination of the force exerted by the 3D hemispheres on

gH p the GR square (no surcharge load yet: p 0)
$ FGR line1p0 FGR line2p0 (92)
gH
Force inside circle in GR FGRsq1;p0 11.21 kN/pile (6)
where
square (Fig. 22)
Part 1 of force on area 1FGRsq2 0.11 kN/pile (6)
Zr0 Zr0 inside Lx3D$Lx3D square,
FGR line1p0 2 sq dAGR line 2 sq $dr but outside circle (Fig. 22)
Part 2 of force on area 2FGRsq2 20.50 kN/pile (6)
0 0 inside Lx3D$Lx3D square,
1
ZL but outside circle (Fig. 22)
2 x2D
0.10
P2D $r Kp 1 Q2D $r dr
Part 3 of force on area inside 3FGRsq2 kN/pile (6)
2 (93) Lx3D$Lx3D square,
0 but outside circle (Fig. 22)
Part 4 of force on area inside 4FGRsq2 15.33 kN/pile (6)
Lx3D$Lx3D square, but outside
thus
circle (Fig. 22)

1 Lx2D Total force on area inside FGRsq2;p0 5.19 kN/pile 1FGRsq2
P 1 2 Lx3D$Lx3D square, but outside 2FGRsq2
FGR line1p0 2 2D $r Kp Q2D $r 2 circle (Fig. 22) 3FGRsq2
Kp 2 0
(94) 4FGRsq2
Kp (6)
P 1 1
/FGR line1p0 2 2D L Q2D $L2x2D Force on area outside Lx3D$Lx3D FGRsq3;p0 0.00 kN/pile (9)
Kp 2 x2D 4 square, but inside GR square)
Total force on GR square (Fig. 22) FGRsquare;p0 16.40 kN/pile (6)
and
Determination of the force transferred along the 3D hemi-
FGR line2p0 gHsx a Lx2D for H < 12 sx a spheres to the 2D arches; to be applied as surcharge load on the 2D
(95) arches.
FGR line2p0 0 for H 12 sx a
Force transferred Ftransferred 59.85 kN/pile (10)
with surcharge load p > 0: Resulting surcharge load on 2D arches ptransferred 21.20 kPa (11)
Kp
gH p P 1 1
FGR line;p>0 $ 2 2D L Q2D $L2x2D Determination of the force exerted by the 2D arches on the GR
gH Kp 2 x2D 4 strips (no surcharge load yet: p 0) and no load outside the arches

on the GR strip (Eq. (14))
FGR line2p0 (96)
Total force on GR strips FGRstrips;p0 35.97 kN/pile (14)
Determination of load distribution (no surcharge load yet: p 0).

Appendix B. Calculation example: the Woerden eld test of
Fig. 16b total force on GR for p 0 FGR B C 52.37 kN/pile (16)
Total force on pile cap with p 0 A 119.94a kN/pile (15)
a
Input parameters This is the value given in Fig. 16b.
Diameter circular pile cap d 0.85 m

Height embankment H 1.86 m Determination of load distribution (with surcharge load:
Lateral ctc distance piles sx 2.25 m
p 5 kPa).
Longitudinal ctc distance piles sy 2.25 m
Unit weight g 18.3 kN/m3
Surcharge load p 5 kPa Total force on GR for p 5 kPa FGR B C 61.60 kN/pile (16)
Cohesion c 0 kPa Total force on pile cap for p 5 kPa A 141.09 kN/pile (15)
Internal friction angle 4 43 deg Total pressure on pile cap for pA 248.63 kPa pA A/Ap
p 5 kPa
Percentage of total force transferred A% 69.6 % A% A/
to the pile cap directly for ((gH p)*s2x )
p 5 kPa
Calculated parameters Equation

q
Equivalent width pile cap a Bers 0.75 m a 1 pd 2 Step 2 with the inverse triangular load distribution according to
4
Passive earth pressure coefcient Kp 5.29 e (2) Van Eekelen et al. (2012b).
Height 3D hemisphere Hg3D 1.59 m (4)
Width square loaded by Lx3D 1.50 m (8) Additional parameters:
3D arches (Fig. 22)
Length of GR strip loaded Lx2D 1.50 m (12)
by 2D arches (Fig. 23)
Calculation parameter P2D 90.62 kPa/mKp1 (14) Subgrade reaction k 0.00 kN/m3
Calculation parameter Q2D 29.43 kN/m3 (14) GR stiffnessb J 5000 kN/m
Calculation parameter P3D 0.11 kPa/m2Kp2 (7) b
Stress and time dependent. The GR stiffness should be determined from
Calculation parameter Q3D 12.77 kN/m3 (7)
isochronic curves.
The average geometric and constitutive strains should be Filz, G.M., Sloan, J.A., McGuire, M.P., Collin, J.G., Smith, M.E., 2012. Column-sup-
ported embankments: settlement and load transfer. In: Proceedings of Geo-
equalised numerically:
Congress, Oakland, California, March, p. 24.
Filz, G., Sloan, J., 2013. Load distribution on geosynthetic reinforcement in column-
supported embankments. In: Proceedings of Geo-congress, March 2013,
s
2
California.
Px 12 sx a dz Guido, V.A., Kneuppel, Sweeney, M.A., 1987. Plate loading test on geogrid reinforced
dx x0 1 12 sx a x X
1
s a
2 x
earth slabs. In: Proceedings Geosynthetics87 Conference, New Orleans,
dx Tx pp. 216e225.
1 s
dx Han, J., Gabr, M.A., January 2002. Numerical analysis of geosynthetic-reinforced and
2 x a x0
J
pile-supported earth platforms over soft soil. Journal of Geotechnical and
Geoenvironmental Engineering 44, 44e51.
We nd: (97)
Han, J., Bhandari, A., Wang, F., 2012. DEM analysis of stresses and deformations of
geogrid-reinforced embankments over piles. International Journal of Geo-
mechanics 12, 340e350.
Handy, R.L., March, 1985. The arch in soil arching. Journal of Geotechnical Engi-
Max GR deection z(x 0) 0.022 m Van Eekelen et al., neering III (3). ASCE, ISSN: 0733-9410/85/0003-0302/$01.00. Paper No. 19547.
2012b, Eq. (15) Hewlet, W.J., Randolph, M.F., April 1988. Analysis of piled embankments. Ground
GR inclination at dz/dx at 0.332 m/m Van Eekelen et al., Engineering 22 (3), 12e18.
edge pile cap x (sx a) 2012b, Eq. (15) Hong, W.P., Lee, J.H., Lee, K.W., 2007. Load transfer by soil arching in pile-dupported
Average GR strain 3 average 0.582 % Van Eekelen et al., embankments. Soils and Foundations 47 (5), 833e843.
2012b, Eq. (18) Jenck, O., Dias, D., Kastner, R., 2009. Discrete element modelling of a granular
with correction platform supported by piles in soft soil e validation on a small scale model test
given in Eq. (97) and comparison to a numerical analysis in a continuum. Computers and Geo-
of this paper technics 36, 917e927.
Max GR strain at at
3 max 0.609 % Van Eekelen et al., Jones, C.J.F.P., Lawson, C.R., Ayres, D.J., 1990. Geotextile reinforced piled embank-
the edge of x (sx a) 2012b, Eq. (15) ments. In: Hoedt, Den (Ed.), Geotextiles, Geomembranes and Related Products.
Balkema, Rotterdam, ISBN 90 6191 119 2, pp. 155e160.
the pile cap
Kempfert, H.-G., Gbel, C., Alexiew, D., Heitz, C., 2004. German recommendations
Max tensile force Tmax at 30.5 kN/m Van Eekelen et al.,
for reinforced embankments on pile-similar elements. In: Proceedings of
at the edge of x (sx a) 2012b, Eq. (15) EuroGeo 3, Munich, pp. 279e284.
the pile cap Lally, D., Naughton, P.J., 2012. An investigation of the arching mechanism in a
Appendix of Analytical model for arching in piled embankments geotechnical centrifuge. In: Proceedings 5th European Geosynthetics Congress.
Valencia, vol. 5, pp. 363e367.
Le Hello, B., Villard, P., 2009. Embankments reinforced by piles and geosynthetics e
References numerical and experimental studies with the transfer of load on the soil
embankment. Engineering Geology 106, 78e91.
Abusharar, S.W., Zeng, J.J., Chen, B.G., Yin, J.H., 2009. A simplied method for Liikennevirasto, 2012. Geolujitetut maarakenteet, Tiegeotekniikan ksikirja, Lii-
analysis of a piled embankment reinforced with geosynthetics. Geotextiles and kenneviraston oppaita 2/2012, ISBN 978-952-255-104-7. Finnish design
Geomembranes 27, 39e52. guideline for geosynthetic reinforcement (in Finnish).
ASIRI, 2012. Recommandations pour la conception, le dimensionnement, lex- Marston, A., Anderson, A.O., 1913. The Theory of Loads on Pipes in Ditches and Tests
cution et le contrle de lamlioration des sols de fondation par inclusions of Cement and Clay Drain Tile and Sewer Pipe. Bulletin No. 31. Engineering
rigides, ISBN 978-2-85978-462-1. Experiment Station.
Blanc, M., Rault, G., Thorel, L., Almeida, M., 2013. Centrifuge investigation of load McGuire, M., Sloan, J., Collin, J., Filz, G., 2012. Critical height of column-supported
transfer mechanisms in a granular mattress above a rigid inclusions network. embankments from Bench-Scale and eld-scale tests. In: TC 211 International
Geotextiles and Geomembranes 36, 92e105. Symposium on Ground Improvement IS-GI Brussels. ISSMGE.
BS8006-1, 2010. Code of Practice for Strengthened/reinforced Soils and Other Fills. McKelvey, J.A., 1994. The anatomy of soil arching. Geotextiles and Geomembranes
British Standards Institution, ISBN 978-0-580-53842-1. 13, 317e329.
Britton, E., Naughton, P., 2008. An experimental investigation of arching in piled Naughton, P., 2007. The signicance of critical height in the design of piled em-
embankments. In: Proceedings of the 4th European Geosynthetics Congress bankments. Soil Improvement, 1e10. http://dx.doi.org/10.1061/40916(235)3.
EuroGeo 4. Edinburgh. Paper number 106. Nadukuru, S.S., Michalowski, R.L., May 2012. Arching in distribution of active load
Cain, W., 1916. Earth Pressure, Retaining Walls and Bins. John Wiley and Sons. Inc., on retaining walls. Journal of Geotechnical and Geoenvironmental Engineering,
New York. 575e584.
Carlsson, B., 1987. Reinforced Soil, Principles for Calculation. Terratema AB, Link- Public Work Research Center, 2000. Manual on Design and Execution of Reinforced
ping (in Swedish). Soil Method with Use of Geotextiles, second ed. Public Work Research Center,
Casarin, C., 2011. Private Communication, So Paulo, Brazil. pp. 248e256. (in Japanese).
Chen, R.P., Chen, Y.M., Han, J., Xu, Z.Z., 2008. A theoretical solution for pile- Rogbeck, Y., Gustavsson, S., Sdergren, I., Lindquist, D., 1998. Reinforced piled em-
supported embankments on soft soils, under one-dimensional compression. bankments in Sweden e design aspects. In: Proceedings of the Sixth Interna-
Canadian Geotechnical Journal 45, 611e623. tional Conference on Geosynthetics, pp. 755e762.
Chen, Y.M., Cao, W.P., Chen, R.P., 2008. An experimental investigation of soil arching Russell, D., Pierpoint, N., November 1997. An assessment of design methods for
within basal reinforced and unreinforced piled embankments. Geotextiles and piled embankments. Ground Engineering, 39e44.
Geomembranes 26, 164e174. SINTEF, 2002. A computer program for designing reinforced embankments. In: Proc.
Collin, J.G., 2004. Column supported embankment design considerations. In: 52nd 7th International Conference on Geotextiles, Nice 2002, France, vol. 1,
Annual Geotechnical Engineering Conference. University of Minnesota. pp. 201e204.
CUR 226, 2010. Ontwerprichtlijn paalmatrassystemen (Design Guideline Piled Sloan, J.A., 2011. Column-supported embankments: full-scale tests and design
Embankments), ISBN 978-90-376-0518-1 (in Dutch). recommendations. PhD thesis, Faculty of the Virginia Polytechnic Institute of
Deb, K., Mohapatra, S.R., 2012. Analysis of stone column-supported geosynthetic- State University, Blacksburg, VA, USA.
reinforced embankments. Applied Mathematical Modelling. Svan, G., Ilstad, T., Eiksund, G., Want, A., 2000. Alternative calculation principle for
Deb, K., 2010. A mathematical model to study the soil arching effect in stone design of piled embankments with base reinforcement. In: Proceedings of the
column-supported embankment resting on soft foundation soil. Applied 4th GIGS in Helsinki.
Mathematical Modelling 34, 3871e3883. Terzaghi, K., 1943. Theoretical Soil Mechanics. John Wiley and Sons, New York.
Den Boogert, Th., Van Duijnen, P.G., Van Eekelen, S.J.M., 2012. Numerical analysis of Van Duijnen, P.G., Van Eekelen, S.J.M., Van der Stoel, A.E.C., 2010. Monitoring of a
geosynthetic reinforced piled embankent scale model tests. Plaxis Bulletin 31, railway piled embankment. In: Proceedings of 9 ICG, Brazil, pp. 1461e1464.
12e17. Van Eekelen, S.J.M., Bezuijen, A., Oung, O., 2003. Arching in piled embankments;
EBGEO, 2010. Empfehlungen fr den Entwurf und die Berechnung von Erdkrpern experiments and design calculations. In: Proceedings of Foundations: In-
mit Bewehrungen aus Geokunststoffen e EBGEO, vol. 2. German Geotechnical novations, Observations, Design and Practice, pp. 885e894.
Society, Auage, ISBN 978-3-433-02950-3 (in German). Also available in En- Van Eekelen, S.J.M., Jansen, H.L., Van Duijnen, P.G., De Kant, M., Van Dalen, J.H.,
glish: Recommendations for Design and Analysis of Earth Structures using Brugman, M.H.A., Van der Stoel, A.E.C., Peters, M.G.J.M., 2010. The Dutch
Geosynthetic Reinforcements e EBGEO, 2011. ISBN: 978-3-433-02983-1 and design guideline for piled embankments. In: Proceedings of 9 ICG, Brazil,
digital in English ISBN: 978-3-433-60093-1. pp. 1911e1916.
Ellis, E., Aslam, R., June 2009. Arching in piled embankments. Comparison of centri- Van Eekelen, S.J.M., Bezuijen, A., Van Tol, A.F., 2011. Analysis and modication of the
fuge tests and predictive methods, part 1 of 2. Ground Engineering, 34e38. British Standard BS8006 for the design of piled embankments. Geotextiles and
Eskisar, T., Otani, J., Hironaka, J., 2012. Visualization of soil arching on reinforced Geomembranes 29, 345e359.
embankment with rigid pile foundation using X-ray CT. Geotextiles and Geo- Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012a. Model experi-
membranes 32, 44e54. ments on piled embankments Part I. Geotextiles and Geomembranes 32, 69e81.
Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012b. Model experiments on Vermeer, P.A., Punlor, A., Ruse, N., 2001. Arching effects behind a soldier pile wall.
piled embankments. Part II. Geotextiles and Geomembranes 32, 82e94 (including its Computers and Geotechnics 28, 379e396.
corrigendum Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012b2. Vllmy, A., 1937. Eingebettete Rohre, Mitt. Inst. Baustatik, Eidgen. Tech. Hochschule,
Corrigendum to Model experiments on piled embankments. Part II [Geotextiles and Zrich, Mitt. No. 9.
Geomembranes 32 (2012) pp. 82e94]. Geotextiles and Geomembranes 35: 119). Zaeske, D., February 2001. Zur Wirkungsweise von unbewehrten und bewehrten
Van Eekelen, S.J.M., Bezuijen, A., 2012. Does a piled embankment feel the passage mineralischen Tragschichten ber pfahlartigen Grndungselementen. Schrif-
of a heavy truck? High frequency eld measurements. In: Proceedings of the tenreihe Geotechnik, Uni Kassel. Heft 10 (in German).
5th European Geosynthetics Congress EuroGeo 5. Valencia. Digital version, vol. Zhang, L., Zhao, M., Hu, Y., Zhao, H., Chen, B., June 2012. Semi-analytical solutions
5, pp. 162e166. for geosynthetic-reinforced and pile-supported embankment. Computers and
Van Eekelen, S.J.M., Bezuijen, A., 2013. Dutch research on piled embankments. In: Geotechnics. ISSN: 0266-352X 44, 167e175. http://dx.doi.org/10.1016/
Proceedings of Geo-Congress, California, March 2013. j.compgeo.2012.04.001.

Geotextiles and Geomembranes: S.J.M. Van Eekelen, A. Bezuijen, A.F. Van Tol

Uploaded by

Copyright:

Available Formats

Geotextiles and Geomembranes: S.J.M. Van Eekelen, A. Bezuijen, A.F. Van Tol

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geotextiles and Geomembranes: S.J.M. Van Eekelen, A. Bezuijen, A.F. Van Tol

Uploaded by

Copyright:

Available Formats

Geotextiles and Geomembranes 39 (2013) 78e102

Contents lists available at SciVerse ScienceDirect

Geotextiles and Geomembranes

An analytical model for arching in piled embankments

Rigid arch models, such as several Scandinavian models

P part C TPC load

x aggregated, resulting in a total GR stiffness J2% 2269 kN/m in

2 TPCs behaviour suggested by this J2%. However, GR stiffness is not

m kN/m3 kPa and kN/pile

Selection of tests of series 1 (Van Eekelen et al., 2012a)

where E A% A/(A B C), with A, B and C given in kN/pile, A% in

measured ground pressure (kPa)

top load step

top load step

top load step

top load step

top load st.

measured arching A (kN/pile)

K2, arching A (kN/pile)

top load st.

The following sections present the equations for the 2D concen-

sq Px2D $rKp 1 Q2D $r

where z is the vertical distance between the considered point and

4.5. Concentric arches model: combination of 2D arches and 3D

FGRsq2p0 1 FGRsq2 2 FGRsq2 3 FGRsq2 4 FGRsq2 F

4.6. Line-shaped foundations

1Kp Kp 1 The concentric 2D arches also apply in the 2D situation where

arching A (percentage oftotal load, %)

pressure on GR strip (kPa)

5.4. Load distribution on GR 5.5. Ground pressure versus depth

EBGEO measured, with GR

tests T2: sand fill =40.9o test K5: H=343 m

arching A (percentage of total load, %)

arching A (percentage of total load, %)

arching A(percentage of total load, %)

2 sx g Kp 1 g which is the differential equation for the radial stress in the 3D

where we have dened P2D and Q2D as:

The outer radius of the hemisphere is Ro sx/O2. The surcharge load

dR area dA1 = RdR FGRsquare3

area dA2 = RdR a or Bers

Total force on the GR square; determination ofpFGRsquare

The three terms are calculated with surcharge load p 0, and

From this, it follows that:

P3D $222Kp $Lx3D 2Kp

sq P2D $rKp 1 Q2D $r with surcharge load p > 0:

(2D arch), where

Determination of load distribution (no surcharge load yet: p 0).

Diameter circular pile cap d 0.85 m

Calculated parameters Equation

You might also like