See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/235652081

Design methods for geosynthetic anchor trenches on the basis of true scale
experiments and discrete modelling

Article  in  Canadian Geotechnical Journal · December 2004

DOI: 10.1139/t04-063

CITATIONS READS

28 445

2 authors:

Pascal Villard Bruno Chareyre

University Joseph Fourier - Grenoble 1 Université Grenoble Alpes
88 PUBLICATIONS   949 CITATIONS    105 PUBLICATIONS   979 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

dessiccation and unsaturated soils View project

ECM small grant scheme View project

All content following this page was uploaded by Bruno Chareyre on 11 October 2016.

The user has requested enhancement of the downloaded file.

 1193

Design methods for geosynthetic anchor trenches

on the basis of true scale experiments and
discrete element modelling
Pascal Villard and Bruno Chareyre

Abstract: Analytical models are proposed for analysing the mechanisms whereby geosynthetic anchors fail. The cases
of run-out and L- or V-shaped anchor trenches are considered. An analytical expression is obtained for the tension that
may be mobilized in the anchor as a function of the geometry and mechanical properties of the materials. Two hypoth-
eses relating to the nature of the soil are taken into account, assuming failure either through relative slippage between
the soil and the inclusion (cohesive soil) or by deformation of the anchoring soil mass (frictional soil). The proposed
analytical formulations are compared with the numerical results of discrete-element simulations and with the experi-
mental results of full-scale pullout tests. The results obtained show the validity of the proposed analytical formulations
for designing anchors in trenches.
Key words: geosynthetics, anchorage, pullout, design, failure, discrete elements.
Résumé : Des modèles analytiques sont proposés pour analyser les mécanismes de rupture des ancrages géosynthéti-
ques. On traite le cas des ancrages simples par recouvrement, et des ancrages en tranchée en L ou en V. Une expres-
sion analytique de la tension mobilisable dans l’ancrage est obtenue, en fonction de la géométrie et des propriétés
mécaniques des matériaux. Deux hypothèses relatives à la nature du sol sont prises en considération en supposant soit
une rupture par glissement relatif sol–inclusion (sol cohésif), soit par déformation du massif d’ancrage (sol pulvéru-
lent). Les formulations analytiques proposées sont comparées aux résultats numériques de simulations par éléments
discrets, ainsi qu’à des résultats expérimentaux d’essais d’extraction en vraie grandeur. Les résultats obtenus démon-
trent la validité des formulations analytiques proposées pour le dimensionnement des ancrages en tranchée.

Mots clés : géosynthétiques, ancrage, extraction, dimensionnement, rupture, elements discrets.

Villard and Chareyre

1205

Introduction the system, it is necessary to estimate the tension that can be

The stability and permanence of geosynthetic lining sys-
tems on slopes depend partly on the efficiency of the an-
chors holding the geosynthetic sheets at the top of the slope.
The role of the anchor is to withstand the tension generated
by friction along the slope (Fig. 1a). This can be done using
a simple run-out anchor (Fig. 1b). The geosynthetic sheets
are often installed in trenches, however, as in Figs. 1c and
1d, to optimize the dimensions of the anchor zone and en-
sure effective anchorage. This being the case, the geometry
of the anchorage system consists of a number of segments
with different orientations, forming an L or V shape. To size
Received 25 August 2003. Accepted 1 June 2004. Published
on the NRC Research Press Web site at http://cgj.nrc.ca on
17 December 2004.
P. Villard and B. Chareyre.1,2 Laboratoire Interdisciplinaire
de Recherche Impliquant la Géologie et la Mécanique
(LIRIGM), Joseph Fourier’s University of Sciences, 38041
Grenoble CEDEX 9, France.
1 analytical models for anchor failure. The anchoring capaci-
Corresponding author (e-mail: bruno.chareyre@ec-lyon.fr).
2
Present address: Ecole Centrale de Lyon, Unité mixte de
recherche, Centre National de la Recherche Scientifique
5513, 36, avenue Guy de Collongue, 69134 Ecully CEDEX,
France.
Can. Geotech. J. 41: 1193–1205 (2004) doi: 10.1139/T04-063
mobilized in the anchor (the anchoring capacity) according
to its geometry and the properties of the constituent materi-
als.
Experimental and numerical studies were undertaken at
the Research Institute for Agricultural and Environmental
Engineering (CEMAGREF) Regional Centre in Bordeaux,
France, and the Laboratoire Interdisciplinaire de Recherche
Impliquant la Géologie et la Mécanique (LIRIGM),
Grenoble, France, to understand the anchoring mechanisms.
Briançon et al. (2000) showed that the existing analytical
formulas were often ineffectual for predicting the strength
provided by an anchor trench. Subsequently, Chareyre et al.
(2002) demonstrated the complex mechanisms involved
during the pullout of geosynthetic sheets. These complex
mechanisms depend not only on the characteristics of the
soil–geosynthetic interface but also on the behaviour of the
anchoring soil mass. Chareyre (2003) carried out a paramet-
ric study to investigate the effect of the nature of the soil, the
geometry of the anchor, and the properties of the interface
on the anchorage. On this basis, this paper proposes simple
ties determined based on the formulation are compared with
those from experimental results and numerical simulations.
The experimental results used for this study were obtained
using the CEMAGREF anchoring facility (Fig. 2), with geo-
© 2004 NRC Canada
1194
Fig. 1. Types of system studied: (a) typical geometry, (b) run-out
anchor, (c) L-shaped anchor, (d) V-shaped anchor.
synthetic sheets anchored in trenches of various shapes. Two
different kinds of soils were used, namely a sand and a
clayey silt. The geosynthetic tested was a needle-punched
nonwoven geotextile reinforced in one direction. The details
of the experimental work were described in Briançon
(2001).
The numerical model used to simulate the behaviour of
anchored geosynthetic sheets is based on the discrete-
element method (DEM) developed by Cundall and Strack
(1979). The DEM assumes a set of particles interacting with
one another at the points of contact and can be used to simu-
late large relative displacements at the interface and large
deformations in the soil (shear bands, crushing, or overall
rotation). The calculation method adopted to model the an-
chors is described in Chareyre et al. (2002) and Chareyre
and Villard (2002).
Analytical modelling of failure
New analytical formulations are proposed in the paper for
determining the anchoring capacities of various geometries
(run-out anchors, L-shaped anchors, and V-shaped anchors).
For each geometry, the failure mechanism is analysed in the
context of two different hypotheses. In the first, it is as-
sumed that the anchoring soil mass is rigid (i.e., undergoes
only elastic strain) with the exception of the cover soil,
which can shift during pullout. Pullout failure therefore cor-
responds to a relative displacement between the soil and
geosynthetic. In the second, it is assumed that both the strain
and failure of the anchoring soil mass are taken into account
by means of a block mechanism. This section gives an ac-
count of the existing and proposed analytical formulations.
Existing design methods
Researchers who have proposed analytical formulas for
designing run-out or trench-type anchors (Hulling and
Sansone 1997; Koerner 1998; Guide technique 2000) drew
on various hypotheses for determining the maximum force
Tmax that can be mobilized in the anchor. Certain hypotheses
are common to all the authors, and others vary from one to
another. These hypotheses are summarized in the following
subsections.
Hypotheses common to all researchers
The geometry of the anchor is represented schematically
by linear segments (numbered in an increasing order from
the outside towards the inside of the soil mass).
The anchor fails only by relative displacement at the soil–
geosynthetic interface.
Can. Geotech. J. Vol. 41, 2004
The shear stresses τ that can be mobilized at the interface
are equal to the maximum stresses τmax corresponding to the
slip limit state (on one or both sides of the geosynthetic).
Friction is governed by a Mohr–Coulomb interface law:
τmax = σ n tan δ , where δ is the friction angle, and σ n is the
normal stress acting at the interface before pullout.
The contribution of the ith segment to the total anchorage
can be assimilated to a force Fi calculated by integrating the
shear stress of intensity τmax on either side of that portion of
the geosynthetic sheet.
Hypotheses specific to individual researcher
For Hulling and Sansone (1997), the normal stress σ n at
the interface can be evaluated from the initial stress state of
the soil mass. The normal stress at any point is calculated by
taking into account the weight of the underlying soil (of
density γ) and the coefficient K0 of the earth at rest (Fig. 3a).
According to this principle, the sum of the individual fric-
tion values Fi is calculated independently for each segment
of the anchor. The tensions provided by each segment are
then added, neglecting any possible effect of change in di-
rection. If the anchor consists of three consecutive segments,
then
3
[1] Tmax = ∑ Fi
i =1
Koerner (1998) takes into account a different stress state
from the initial one on the failure of L-shaped anchors
(Fig. 3b):
(1) The inclination of the tensile force T1 produces an in-
crease in the normal stress under the first segment of the
anchor. This increase is assumed to be equivalent to the
vertical component of T1. This means increasing the ten-
sion that can be mobilized by a factor of 1/[1 − sin(β)
tan(δ)], where β is the slope angle, and δ is the friction
coefficient at the interface.
(2) On either side of the vertical segments, the normal
stress is the active or passive earth pressure, depending
on the side considered. For a three-segment anchor,
3
[2] Tmax = ∑ Fi /[1 − sin(β) tan(δ)]
i =1
In Guide technique (2000), as shown in Fig. 3c, it is con-
sidered that the tension in the sheet is multiplied at each
change in direction by a factor of exp(λ tan δ), where λ is the
difference in orientation between two segments in radians.
This exponential factor is derived from an analogy with the
friction of a wire on a cylinder. Implicitly, this means taking
into account an increase in the normal stress σ n at each
bend. Hryciw (1990) gives a demonstration of this relation.
Outside the bends, the normal stress at the interface corre-
sponds to the initial stress state. Lastly, for three segments,
with Ki being the coefficient associated with bend i, one ob-
tains
[3] Tmax = K1[ F1 + K2( F2 + K3 F3)]
© 2004 NRC Canada
Villard and Chareyre
Fig. 2. Experimental apparatus of Briançon et al. (2000).
Fig. 3. Hypotheses for normal stress state at the interface for an
L-shaped anchor: (a) Hulling and Sansone (1997), (b) Koerner
(1998), (c) Guide technique (2000).
Summary
Various case studies have shown that some of these hy-
potheses are only appropriate under certain condition. In
particular, relative displacements at the interface do not al-
ways occur, the friction at the soil–geosynthetic interface
may be only partially mobilized if the failure occurs in the
1195
soil, and the normal stresses acting on the interfaces at fail-
ure can be very different from the initial stresses.
Simple run-out anchors
The geometry of a simple run-out anchor is as shown in
Fig. 4. The tension T1 is mobilized by friction at the soil–
geosynthetic interfaces while the upper layer (block A) tends
to move at the same time as the sheet.
The static equilibrium of block A above the sheet (Fig. 5)
can be used to determine the maximum value of T1′ (eq. [4]).
P1 is the weight of the soil above the sheet; Rt is the tensile
force mobilized in the soil at the end of the anchor; and γ
and σt are the density and limit tensile stress in the soil, re-
spectively. The forces are expressed per unit width of the an-
chor. It should be noted that the strength of the upper layer
may be greater than P1 tan(δ). In this case, it is assumed that
the cover soil remains integral with the soil mass and the
force Rt that can be mobilized is equal to P1 tan(δ), in confor-
mity with eq. [5]:
[4] T1′ = P1 tan(δ) + Rt
[5] Rt = min[Hσt , P1 tan(δ)]
The tension T1 required to pull out the sheet (applied par-
allel to the slope) is obtained by considering the angle effect.
Two failure mechanisms are considered as outlined in the
next two subsections.
Assumption of a rigid soil mass
It is assumed that near the change in direction, the geo-
synthetic moves in relation to the underlying soil, which is
considered to be fixed. It is therefore assumed that the stress
state at the interface verifies τ = σ n tan(δ). In this hypothesis,
the problem is analogous to that of a wire rubbing against
the arc of a circle. It is thus possible to introduce a weight-
ing coefficient K1, the value of which is given in eq. [6],
such that K1 = T1 / T1′ . It should be noted that this result is
obtained by applying the approach recommended in the
Guide technique (2000) to a particular case:
[6] K1 = exp(β tan δ)
Assumption of soil mass failure
In this case, it is assumed that the forces acting near the
change of angle are likely to shear the soil along the slip line
shown in Fig. 6, and that the moving block is sufficiently
small for the volume (weight) and surface (surrounding
© 2004 NRC Canada
1196
Fig. 4. Geometry of a run-out anchor.
Fig. 5. Static equilibrium of block A above the geosynthetic
sheet. R1, reaction of the soil.
Fig. 6. Failure of soil near a change of angle. α, direction of the
failure line; φ, internal friction angle of the soil.
stress) forces applied to the block to be neglected in relation
to T1 and T1′. Then, considering the Mohr–Coulomb law of
friction along the slip line, a limit equilibrium of the forces
in the sheared zone yields eqs. [7] and [8]:
[7] T1 − T1′ cos(β) − R sin(α + φ) = 0
[8] − T1′ sin(β) + R cos(α + φ) = 0
where φ is the internal friction angle of the soil. By combin-
ing,
[9] T1 / T1′ = [cos(β) + sin(β) tan(α + φ)]
The most critical angle α (the one that minimizes the ratio
T1 / T1′) is zero, and the ratio T1 / T1′ corresponding to this
mechanism is therefore given by eq. [7]. T1 / T1′ must in the-
ory be 1 or more. Certain pairs (β, φ) may, however, give val-
ues of less than 1 in eq. [10]. In this case, it is assumed that
the soil is very unstable near the bend and does not provide
any additional resistance. A tension ratio of 1 is therefore as-
sumed.
[10] T1 / T1′ = [cos(β) + sin(β) tan(φ)]
The possibility of failure within the soil mass does not ex-
clude failure at the interface. The predominant mechanism is
that which leads to the lowest ratio.
To summarize, the weighting coefficient can be used to
take into account the effect of change of direction in the fol-
lowing way: (i) for a rigid mass,
Can. Geotech. J. Vol. 41, 2004
Fig. 7. Geometry of an L-shaped anchor. The bends in the an-
chor are labelled 1, 2, and 3.
[11] T1 = K1T1′
where K1 = exp(β tan δ); and (ii) for a deformable mass,
[12] T1 = K1*T1′
where K1* = min[exp(β tan δ); cos(β) + sin(β) tan(φ)].
By combining eq. [4] with eqs. [11] or [12], depending on
the assumptions made, it is possible to determine the anchor-
ing capacity. If the soil mass is rigid, T1 = K1T1′; if failure
occurs close to the bend in the soil mass, T1 = K1*T1′ if K1* >
1, otherwise T1 = T1′. In the case of anchors with multiple
changes in orientation, coefficients similar to those in
eqs. [11] and [12] will be derived associated with each bend.
The variables incorporating this possibility of failure in the
soil are indicated with asterisks.
L-shaped anchors
Consider an L-shaped anchor in a rectangular trench as
shown in Fig. 7. The sum of the friction values on the upper
side of segment 1, RL, is a function of the geometry and
characteristics of the soil. Let H and P1 be the height and
weight, respectively, of block A, then:
(1) If Hσt > P1 tan φ, the soil cover (block A) is sufficiently
strong not to be dissociated from the rest of the soil
mass (block B). During pullout, it is assumed that
RL = P1 tan φ.
(2) If Hσt < P1 tan φ, block A will be dissociated from the
rest of the soil mass after cracking at bend 2. Given the
fact that the anchor is mobilized progressively as the
sheet pulled out, the failure of the upper block of soil
will occur before that of the anchor. In this case, only
friction at the base of the soil layer will be considered in
determining the anchoring capacity, and it is assumed
that RL = 0.
Two failure mechanisms are considered as outlined in the
following subsections.
Failure mechanism 1: assumption of a rigid soil mass
In the case of a rigid soil mass, failure involves the rela-
tive displacement of the inclusion in relation to the mass,
through slip at the interface. Thus the maximum forces that
can be mobilized correspond to the limit equilibrium state
at all points of the soil–geosynthetic interface (i.e.,
τ = σ n tan δ ). The value of the anchoring capacity is there-
fore determined by considering the distribution of the nor-
mal stress σ n on each segment of the interface.
Segment 3 — At the initial state it is assumed that the
weight P2 of block B of the soil mass rests entirely on the
© 2004 NRC Canada
Villard and Chareyre 1197

Fig. 8. Notation of forces implied in static equilibrium of the Fig. 9. Subhorizontal failure in soil below the upper bend.
system.

last segment of the sheet. When the tension in the sheet in-
creases, the action of the geosynthetic at bend 3 results in an
uplift force on block B. The vertical component of this uplift
force is equal to T3 (Fig. 8) and opposes weight P2. It results
in a reduction in the normal stresses acting on segment 3 of
the anchor, and the final normal stress is equal to P2 – T3.
Hence, by using the notations in Fig. 8, T3′ is given in
eq. [13], and the relation between T3 and T3′ at the limit slip
state is defined in eq. [14]:
[13] T3′ = 2(P2 − T3) tan(δ)
[14] T3 = K3T3′
where P2 is the weight of the soil above the sheet (P2 =
γB(H + D), and K3 is the change-of-angle coefficient defined
as in eq. [11] by K3 = exp(π / 2 tan δ). Therefore,
[15] T3′ = 2P2 tan(δ)/[1 + 2K3 tan(δ)]
[16] T3 = 2P2K3 tan(δ)/[1 + 2K3 tan(δ)]
Segment 2 — The horizontal forces exerted by the soil on
the vertical portion of the sheet are denoted Fh and obtained
from eq. [17]. The friction force (proportional to Fh) that can
be mobilized on each vertical side is added to T3 to give the
tension T2′ (given by eq. [18]). The assumption of a limit
slip state at the second bend gives eq. [19]:
[17] Fh = γ K0D (H + D/ 2)
[18] T2′ = T3 + 2 Fh tan(δ)
[19] T2 = K2T2′
⎧⎪ ⎡ 2K3B (H + D) ⎤
[22] T1 = K1 tan(δ) ⎨γ K2 ⎢ K0D (2H + D) + ⎥
⎩⎪ ⎣ 1 + 2K3 tan(δ) ⎦
RL ⎫⎪
+ P1 + ⎬
tan(δ) ⎪
⎭
where K1 is the change-of-angle coefficient defined by K1 =
exp(β tan δ).
Failure mechanism 2: assumption of soil mass failure
With this anchoring geometry, no failure is envisaged
specifically near segment 3. It is estimated that the uplift
mechanism described previously remains valid, allowing the
tension T2′ to be calculated (eq. [18]). In contrast, for the up-
per bend, the possibility of failure is envisaged by taking the
subhorizontal failure line of Fig. 9 as a simplifying assump-
tion. This failure scheme, which is similar to that adopted
for the failure of a soil bend but extended to a larger area,
was derived from the results of simulations (Chareyre 2003).
The stability of block C situated between the geosynthetic
and the failure line is then considered. It is assumed that the
thickness of block C is such that the total mass of the mov-
ing soil is close to that of the cover soil (block A). By writ-
ing the limit equilibrium of the system, an expression for the
corresponding tension T1′* is obtained (eq. [23]). For a given
configuration, T1′* must be compared with the tension T1′ in
eq. [24] to determine the most critical failure mode. The
value to be adopted in calculating the anchoring capacity is
the lower of the two values T1′ and T1′*. Lastly, the tension
T1 that can be mobilized at the head of the system is ob-
tained by applying the coefficient K1*, taking into account
the change in sheet angle near the side slope (eq. [25]):

where K0 is the earth pressure coefficient at rest, K2 is the ⎧⎪ ⎡

[23] T1′ * = tan(φ) ⎨P1 + tan δ ⎢ γ K0D (2H + D)
change-of-angle coefficient defined by K2 = exp(π / 2 tan δ), ⎪⎩ ⎣
and D is the height of the trench (see Fig. 7).
Segment 1 — The proposed mechanism for segment 1 is 2γ K3B (H + D) ⎤ ⎫⎪
the same as that for the run-out anchor, with Rt = RL T1′ and + ⎥ ⎬ + RL
1 + 2K3 tan(δ) ⎦ ⎭⎪
then T1 obtained by adding the friction in the upper part
(eq. [20]) to T2 and including the effect of bend 1 (eq. [21]).
The expression for T1 based on the geometric data is given ⎧ ⎡
[24] T1′ = tan(δ) ⎨P1 + γ K2 ⎢ K0D(2H + D)
in eq. [22]: ⎩ ⎣
[20] T1′ = T2 + P1 tan(δ) + RL 2K3B (H + D) ⎤ RL ⎫⎪
+ ⎥+ ⎬
[21] T1 = K1T1′ 1 + 2K3 tan(δ) ⎦ tan(δ) ⎪⎭

© 2004 NRC Canada

1198 Can. Geotech. J. Vol. 41, 2004

[25] T1 = K1* min(T1′ ; T1′ *) Fig. 10. Geometry of a V-shaped anchor. The bends in the an-
chor are labelled 1, 2, and 3.
V-shaped anchors
In this geometry, the soil cover is split into two blocks, A
and B (Fig. 10). During pullout, block A is dissociated from
block B and moves in conjunction with the underlying
geosynthetic sheet. Depending on the inclination ψ of the
anchor trench bottom and the friction angle at the interface
δ, block B may move along (mechanism 1: ψ ≤ δ) or remain
in position (mechanism 2: ψ ≥ δ). The calculation of T2′ and
then T1 is conducted regarding these two possible mecha-
nisms. Fig. 11. Static equilibrium of block B (mechanism 1).

Assumption of a rigid mass

Expression for T2′ in the case ψ ≤ δ
If ψ ≤ δ, it is assumed that block B is integrated with the
underlying geosynthetic sheet and moved along with it dur-
ing pullout. At failure, friction forces only act on the right
side of block B, following detachment between the block
and the left side of the trench (as in Fig. 11). Block B being
subjected to the action of its own weight P2 equal to
γB[2H + D]/2, of the contact reaction of intensity R2 acting Fig. 12. (a) Static equilibrium of block B (mechanism 2).
on the lower side of the geosynthetic, and of the tension T2′, (b) Equilibrium of the V wedge.
the limit equilibrium condition is expressed in eqs. [26] and
[27]:
[26] R2 cos(δ) − P2 cos(ψ) = 0
[27] T2′ − R2 sin(δ) − P2 sin(ψ) = 0
This gives
[28] T2′ = P2[cos(ψ) tan(δ) + sin(ψ)]
This calculation is only valid if the tensile strength inside
block B is high enough to effectively uplift the whole block
(assumption of a rigid mass). It can be verified as an accept-
able approximation, however, in cases where there is only
partial uplift.
Expression for T2′ in the case ψ ≥ δ
Consider the case where the inclination of the sides of the
anchor trench is greater than the soil–geosynthetic friction
angle. When the anchor fails, block B remains in position,
and the upper and lower interfaces of the geosynthetic sheet
are subjected to maximum shear stress within the trench.
The static equilibrium of block B is shown in Fig. 12. By [35] T3 = 2K3 sin(δ) P2 sin(ψ − δ)/ A
considering the limit equilibrium of each segment,
eqs. [29]–[32] are obtained: [36] T2′ = 2 sin(δ) P2[K3 sin(ψ + δ) + sin(ψ − δ)]/ A

[29] T3′ = 2R3 sin(δ) where

[30] T2′ = T3 + 2R2 sin(δ) [37] A = sin(2ψ) + 2 sin(δ)[(K3 + 1) cos(δ)

− 2 cos(ψ) cos(ψ − δ)]
[31] R3 sin(ψ − δ) − R2 sin(ψ + δ) + T2′ cos(ψ) = 0
Expression for T1
[32] R3 cos(ψ − δ) + R2 cos(ψ + δ) + T2′ sin(ψ) − P2 = 0 The anchoring capacity T1 is obtained by considering suc-
The condition relating to the change of angle also gives cessively the effects of bend 2 (eq. [38]), the weight of block
A (eq. [39]), and bend 1 (eq. [40]); thus,
[33] T3 = K3T3′
[38] T2 = K2T2′
where K3 = exp[2ψ tan(δ)].
Solving the five equations gives [39] T1′ = T2 + P1 tan(δ)
[34] T3′ = 2 sin(δ) P2 sin(ψ − δ)/ A [40] T1 = T1′ K1

© 2004 NRC Canada

Villard and Chareyre 1199

Fig. 13. Simulated load (T1) – displacement (U0) curves (three different initial states) and mean curve.

where K2 and K1 are the change-of-angle coefficients equal Table 1. Characteristics of soils 1 and 2
to exp[ψ tan(δ)] and exp[β tan(δ)], respectively. The expres- (Briançon 2001).
sion for T1 based on the geometric characteristics is thus
φ (°) c (kPa) γ (kN/m3)
[41] T1 = K1{P1 tan(δ) + γ K2B (H + D/ 2) Soil 1 34 11.7* 16.7*
× [cos(ψ) tan(δ) + sin(ψ)]} Soil 2 41 0 16.7
Note: c, cohesion intercept.
if ψ ≤ δ and *Corrected value.

[42] T1 = K1{P1 tan(δ) + 2γ K2 sin(δ) B (H + D/ 2)

Fig. 14. Final geometry of the run-out anchor (U0 = 0.07 m) in
× [K3 sin(ψ + δ) + sin(ψ − δ)]/Α} the simulation where β = 22°.
if ψ ≥ δ (A is defined in eq. [37]).

Assumption of soil mass failure

For this geometry, it would be possible to deduce a failure
line similar to that proposed for the L-shaped anchor, where
the entire surface portion is affected. However, the worst-
case mechanism for V trenches is obtained by assuming that
only local failures occur, near bends 1, 2, and 3. To account
Fig. 15. Increase in Tmax versus angle β: comparison of experi-
for this, the coefficients K1, K2, and K3 in eq. [41] or eq. [42]
mental, simulated (DEM), and analytical results.
(including in the expression for A) should be replaced by the
coefficients K1*, K*2 , and K*3 from eqs. [43]–[45]. If one of
the coefficients in eqs. [43]–[45] is less than 1, it is replaced
by 1 in the equations:
[43] K1* = min[exp(β tan δ); cos(β) + sin(β) tan(φ)]

[44] K*2 = min[exp(ψ tan δ); cos(ψ) + sin(ψ) tan(φ)]

[45] K3* = min[exp(2ψ tan δ); cos(2ψ) + sin(2ψ) tan(φ)]

Validation of the analytical models

In this section, the analytical formulations presented pre-
viously are compared with the results of numerical models
and in certain cases with the results of full-scale experi-
ments, whenever these have been carried out (Briançon
2001). no failure in the soil mass (thus validating the assumption of
All the experimental and numerical results showed that a rigid soil mass). Conversely, in a purely frictional soil
when an anchor in a cohesive soil (clayey silt) fails, there is (sand), failure is accompanied by major deformation in the

© 2004 NRC Canada

1200 Can. Geotech. J. Vol. 41, 2004

Fig. 16. Comparison between the tests and the simulations for an L-shaped anchor and both types of soil.

soil (failure of the soil mass). The analytical formulations Fig. 17. Simulated evolution of an L-shaped anchor: (a) initial
are compared with the numerical and experimental results state, (b) with soil 1, (c) with soil 2.
with the assumption of either a nondeformable soil mass or
failure of the soil mass depending on the nature of the an-
choring soil.

Numerical and physical models

Anchor failure was simulated using a two-dimensional
discrete-element model (DEM). A commercial program
(PFC2D, HCItasca, Minneapolis, MN) was used. In the
DEM, the soil is represented by circular particles that inter-
act with one another via their contact points, and its consti-
tutive behaviour is defined via micromechanical parameters
of the contact laws. The geosynthetic was modelled by
means of a dynamic spar elements method (DSEM), pro-
posed by the authors, which has been coupled with the DEM
code (Chareyre and Villard 2004). The advantage of DSEM
elements is their ability to reproduce the behaviour of the
geosynthetic and its interface directly. In particular, there are
no compression forces or bending moments in the elements.
To enable the numerical and experimental results to be
compared, an attempt was made to reproduce in the simula-
tions the behaviour of the two soils used during the series of
tests, namely a frictional cohesive soil (soil 1) and a purely
frictional soil (soil 2). The main characteristics of these soils
are given in Table 1. The values associated with soil 1 (prop-
erties of the soil and results of pullout tests) are given as used indiscriminately to refer to the real material or the cor-
corrected data related to a soil density γ of 16.7 kN/m3. The responding numerical model. The geosynthetic used for the
experimental values involving forces were multiplied by γ/γ r, tests had a tensile stiffness modulus J of 624 kN/m (mea-
where γ r is the real density of soil 1, so that the various re- sured at 12% of strain), and the soil–geosynthetic friction
sults can be compared at constant density from one soil to angle δ was equal to 34° with both soils. J and δ were intro-
another. To simulate the behaviour of soils 1 and 2, the duced directly in the DSEM model.
numerical model was calibrated so that simulation of The main results obtained for each of the anchors mod-
homogeneous reference tests (biaxial tests) reproduced the elled are the anchor failure kinematics, the tension at various
characteristics given in Table 1. The calibration procedure is points of the geosynthetic sheet, and the pullout curve (pull-
described by Chareyre and Villard (2002). It should be out force as a function of sheet head displacement). The
pointed out that, to obtain internal friction angles in the arrangement of the model particles is based partly on a ran-
model that are as high as those in the experiments, it was dom draw, producing variability in the numerical results.
necessary to combine the soil particles in pairs and obtain This is why the simulation procedure is repeated three or
elongated “grains.” Hereafter, the terms soil 1 and soil 2 are four times for each anchor (generation of the initial state and
© 2004 NRC Canada
Villard and Chareyre 1201

Fig. 18. Experimental and simulated load–displacement curves and analytical prediction of the pullout strength (with two values of in-
terface friction angle δ): (a) soil 1, (b) soil 2.

Fig. 19. Tensile force mobilized in inclusion: simulated and analytical results with soil 1. The bends in the anchor are labelled 1, 2,
and 3.

© 2004 NRC Canada

1202 Can. Geotech. J. Vol. 41, 2004

Fig. 20. Tensile force mobilized in inclusion: simulated and analytical results with soil 2. The bends in the anchor are labelled 1, 2,
and 3.

pullout). A trend curve is obtained (Fig. 13) by calculating Fig. 21. Simulated evolution of a V-shaped anchor: (a) initial
the mean tension for a given displacement. Only the mean state, (b) with soil 1, (c) with soil 2.
curves are represented hereafter.

Simple run-out anchors: influence of ␤

A series of pullout tests was carried out on straight an-
chors, using different values of the slope angle β. These tests
were carried out only with soil 2 and are used as a reference
for studying the effect of the angle. The corresponding nu-
merical simulations were then performed. The final geome-
try simulated in the case of β = 22° is represented in Fig. 14.
For each value of β, the simulations include calculation of
the maximum value Tmax(β) of the force T1 mobilized during
pullout. The effect of a change in direction of the sheet
at the bend is then calculated by means of the ratio
Tmax(β)/Tmax(β = 0). Tmax(β = 0) is obtained, as in eq. [4],
from the value of P1 in the simulation, and with Rt = 0.
All the numerical and experimental results are summa-
rized in Fig. 15. The correlation between the models and
experiments is reasonable.
The results corresponding to the change-of-angle coeffi-
cients K1 and K1* (eqs. [11] and [12]) are shown for compari-
son purposes in Fig. 15. It can be seen that K1* (assumption
of a deformable soil mass) produced a better simulation for “soil mass failure” hypothesis in the case of purely frictional
this case (soil 2). soil.
The analytical formulations developed previously are
L-shaped anchors compared in Fig. 18 with the simulations and available ex-
The geometry of the L-shaped anchors is defined by L = perimental data (eqs. [13]–[22] for soil 1 and eqs. [23]–[25]
1.1 m, D = 0.5 m, B = 0.5 m, H = 0.25 m, and β = 22° using for soil 2). The results for two different interface friction an-
the notations as in Fig. 7. Numerical simulations were car- gles, 18.6° and 34°, are presented (it should be noted that
ried out to compare with the available experimental data for tan(34) ≈ 2 tan(18.6)). It is noted that the analytical methods
two types of soil (clayey silt and sand). The numerical re- provide a good approximation with Tmax. The numerical ap-
sults are close to the experimental results for both soils plications were carried out with values of K0 derived from
(Fig. 16). Figure 17a shows the initial arrangement of the the simulations (K0 = 0.6 in soil 1 and 0.7 in soil 2).
model particles. Figures 17b and 17c show the particle ar- In most design methods in the literature, Tmax is consid-
rangements obtained after pullout of the sheet in the case of ered to be proportional to tan(δ). The simulations, however,
the cohesive frictional soil and the purely frictional soil, re- show a nonlinear dependency: (i) in Fig. 18a, the maximum
spectively. With soil 1, there is no major deformation of the tension is more than doubled when the friction coefficient is
soil mass. With soil 2, in contrast, there is severe deforma- multiplied by two; (ii) in Fig. 18b, the maximum tension is
tion. This confirms the pertinence of the “rigid soil mass” increased by only about 50% when the friction coefficient is
hypothesis in the case of cohesive frictional soil and the multiplied by two.
© 2004 NRC Canada
Villard and Chareyre
Fig. 22. Experimental and simulated load–displacement curves and analytical prediction of the pullout strength (with two values of the
interface friction angle): (a) soil 1, (b) soil 2.
It is particularly noted that these nonlinearities are also
found expressed in the analytical models. In the first case,
the nonlinearity is attributed to the terms that are exponen-
tially dependent on the friction, and in the second case to the
term dependent on the soil friction angle.
Figures 19 and 20 compare the distribution of tension in
the anchor (in a state such that T1 ≈ Tmax) with the values
T3′, T3, …, T1 taken from the analytical models. In the simu-
lations with soil 1 (clayey silt), the change in tension at the
bends displays a comparable increase to that predicted by
the analogy with wire friction (i.e., multiplication by
exp(π / 2 tan δ) at bends 2 and 3 and by exp(β tan δ) at bend
1). In the case of soil 2 (sand), the increase in tension at
bend 1 is correctly taken into account by the coefficient K1*
(eq. [12]). It may also be noted that the effect of bend 2 is in
conformity with the failure mechanism described in Fig. 9.
This good correspondence between all results suggests that
1203
the hypotheses proposed accurately represent the mecha-
nisms highlighted by the simulations.
V-shaped anchors
As this anchor configuration has not been tested at full
scale, only numerical and analytical comparisons are carried
out. The geometry of the V-shaped anchors studied here is
defined by L = 0.8 m, D = 0.25 m, B = 0.8 m, H = 0.25 m,
and β = 22°. The initial state of the model and the final state
after pullout are presented in Fig. 21 for each soil. As be-
fore, there is very little deformation with soil 1 and severe
deformation with soil 2.
The numerical and analytical results are compared in
Fig. 22, with two different angles of friction at the interface
(δ = 34° and δ = 18.6°), covering failure mechanisms 1 and
2, respectively. It can be noted that the two analytical meth-
© 2004 NRC Canada
1204 Can. Geotech. J. Vol. 41, 2004

Fig. 23. Tensile force mobilized in inclusion: simulated and analytical results with soil 1. The bends in the anchor are labelled 1, 2,
and 3.

Fig. 24. Tensile force mobilized in inclusion: simulated and analytical results with soil 2. The bends in the anchor are labelled 1, 2,
and 3.

ods provide good approximations of the maximum tensions
obtained in the simulations.
The evolution in tension along the anchor corresponds
relatively well to the predicted values (Figs. 23, 24). In
particular, the proposed analytical model reproduces the
nontrivial trend consisting of an increase in T3′ when the in-
terface friction decreases to β = 18.6°.
Discussion
Two concurrent failure mechanisms have been taken into
account for the analyses: (1) instability at the soil–
geosynthetic interface, and (2) instability in the anchoring
soil mass. The experimental and numerical results show that
mechanism 1 governs the failure when the soil has cohesion,
so the assumption of a rigid soil mass is suitable to estimate
the anchoring capacity. In contrast, in the absence of cohe-
sion, it is necessary to consider 1 and 2 as two possible fail-
ure mechanisms, the anchoring capacity being obtained from
the most critical mechanism.
In practice, however, it is difficult to quantify cohesion of
soil, and the most appropriate failure hypothesis will depend
© 2004 NRC Canada
Villard and Chareyre
on the design criterion chosen. When anchoring a geo-
membrane for impermeability purposes, the role of the de-
sign is to prevent the anchoring capacity from exceeding the
tensile strength of the geomembrane. Considering a cohesive
soil mass (undrained behaviour) and failure mechanism 1
alone will lead to a conservative design in such a case, since
it gives the highest estimation of the anchoring capacity. In
contrast, when a geotextile is to be anchored for reinforce-
ment purposes (e.g., multilayered liners), the anchorage is
designed in accordance with the minimum strength to be
provided. In this case, a conservative design will be obtained
by neglecting the cohesion (drained behaviour) and consid-
ering mechanisms 1 and 2.
Conclusion
Analytical modelling has been proposed to estimate the
pullout strength provided by run-out anchorages and anchor-
ages in L- or V-shaped trenches. The modelling was based
on the results of true-scale pullout tests and numerical DEM
simulations, which have shown a number of important fea-
tures of deformation and failure for this type of system:
(1) The normal stresses acting on the interfaces at failure
can be very different from the initial stresses.
(2) The friction at the soil–geosynthetic interface may be
only partially mobilized if the failure occurs in the soil.
(3) Tensioning of the sheet is accompanied by an increase
in the stresses at the bend of the interface. The ratio of
forces before and after the bend is an exponential func-
tion of the bend angle (mechanism 1) or a function of
soil properties (mechanism 2).
(4) The soil–geosynthetic interaction in a particular area of
the anchor may be transferred to another area in terms
of forces at the interface (e.g., right and left segments in
the V-shaped anchor).
This study has shown that determining the anchoring ca-
pacity based on interface friction alone can result in a poor
design. Mechanical properties of the anchoring soil must be
regarded. Accordingly, the influence of the cohesion and
friction angle of the soil has been taken into account in the
proposed modelling.
This paper demonstrates the pertinence of discrete ele-
ment modelling for studying the behaviour of soil–inclusion
systems at a macroscopic level. In this particular application,
the numerical results were in good agreement with the ex-
perimental results (from both qualitative and quantitative
standpoints), and meaningful data were obtained concerning
View publication stats
1205
the basic mechanisms involved in failure of the anchorage
systems.
References
Briançon, L. 2001. Stabilité sur pentes des dispositifs géosynthéti-
ques — caractérisation du frottement aux interface et applica-
tions. Ph.D. thesis, Université Bordeaux I, Talence, France.
Briançon, L., Girard, H., Poulain, D., and Mazeau, N. 2000. De-
sign of anchoring at the top of slopes for geomembrane lining
systems. In EuroGeo 2: Proceedings of the 2nd European Geo-
synthetics Conference, Bologna, Italy, 15–18 October 2000.
Edited by A. Cancelli, D. Cazzuffi, and C. Soccodata. Italian
Geotechnical Society, Rome. Vol. 2, pp. 645–650.
Chareyre, B. 2003. Modélisation du comportement d’ouvrages
composites sol–géosynthétique par éléments discrets — applica-
tion aux ancrages en tranchées en tête de talus (Discrete element
modelling of composites soil–geosynthetics structures — appli-
cation to anchor trenches at the top of slopes). Ph.D. thesis, Uni-
versity of Grenoble I, Grenoble, France.
Chareyre, B., and Villard, P. 2002. Discrete element modelling of
curved geosynthetic anchorages with known macro-properties.
In Numerical Modelling in Micromechanics via Particle Me-
thods: Proceedings of the 1st International PFC Symposium,
Gelsenkirchen, Germany, 6–7 November 2002. Edited by H.
Konietzky. A.A. Balkema, Rotterdam, The Netherlands.
pp. 197–203.
Chareyre, B., and Villard, P. 2004. Dynamic spar elements and
DEM in 2D for the modelling of soil–inclusion problems. Jour-
nal of Engineering Mechanics, ASCE. In press.
Chareyre, B., Briançon, L., and Villard, P. 2002. Theoretical versus
experimental modelling of the anchorage capacity of geotextiles
in trenches. Geosynthetics International, 9(2): 97–123.
Cundall, P.A., and Strack, O.D.L. 1979. A discrete numerical
model for granular assemblies. Géotechnique, 29(1): 47–65.
Guide technique. 2000. Etanchéité par géomembranes des ouvra-
ges pour les eaux de ruissellement routier — guide complémen-
taire. Service d’Études techniques des Routes et autoroutes
(SETRA) and Laboratoire central des Ponts et chaussées
(LCPC), Paris, France.
Hryciw, R.D. 1990. Load transfer mechanisms in anchored geo-
synthetic systems. University of Michigan Research Report to
the Air Force Office of Scientific Research, Grant No. 88-0166,
University of Michigan, Ann Arbor, Mich.
Hulling, D.E., and Sansone, L.J. 1997. Design concerns and per-
formance of geomembrane anchor trenches. Geotextiles and
Geomembranes, 15: 403–417.
Koerner, R.M. 1998. Designing with geosynthetics. 4th ed.
Prentice-Hall, Inc., Upper Saddle River, N.J. pp. 487–494.
© 2004 NRC Canada