You are on page 1of 17

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

net/publication/229859454

Yield design of dry‐stone masonry retaining structures—Comparisons with


analytical, numerical, and experimental data

Article  in  International Journal for Numerical and Analytical Methods in Geomechanics · October 2008
DOI: 10.1002/nag.697

CITATIONS READS

27 3,190

3 authors:

Anne-Sophie Colas Jean-Claude Morel


Université Gustave Eiffel Ecole Nationale des Travaux Publics de l'Etat
19 PUBLICATIONS   151 CITATIONS    140 PUBLICATIONS   4,060 CITATIONS   

SEE PROFILE SEE PROFILE

Denis Garnier
École des Ponts ParisTech
54 PUBLICATIONS   585 CITATIONS   

SEE PROFILE

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

2019 Researcher Links Workshop on Investigation of the Impact of Occupant Behavior on Building Performance in the UK and China View project

The use of waste to manufacture construction materials to built sustainable cities and enhance circular economy View project

All content following this page was uploaded by Jean-Claude Morel on 07 November 2017.

The user has requested enhancement of the downloaded file.


INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS
Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
Published online 19 March 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nag.697

Yield design of dry-stone masonry retaining


structures—Comparisons with analytical,
numerical, and experimental data

A. S. Colas1 , J. C. Morel1, ∗, † and D. Garnier2


1 Département Génie Civil et Bâtiment (CNRS, URA 1652), Université de Lyon, École Nationale des Travaux
Publics de l’État, rue M. Audin, 69518 Vaulx-en-Velin Cedex, France
2 Laboratoire des Matériaux et des Structures du Génie Civil (CNRS, UMR 113), Université de Paris-Est,

École Nationale des Ponts et Chaussées, 6-8 Avenue B. Pascal, 77455 Marne-la-Vallée Cedex 2, France

SUMMARY
Over the past few years, there has been a steady increase in the interest in dry-stone walling not only to
preserve existing constructions but also to build new ones. Yet, dry masonry’s expansion is slowed down by
the lack of scientific knowledge to assess its reliability. This study aims at contributing to the construction
of this scientific frame using a simplified model based on yield design and homogenization, which can
be directly exploited for engineering purposes. A new analytical expression of the ultimate load is thus
established. Then, the validity of the method is assessed by comparisons with limit equilibrium analysis,
distinct element method, and field trials. Finally, possible improvements of the model are discussed.
Copyright q 2008 John Wiley & Sons, Ltd.

Received 12 February 2007; Revised 12 November 2007; Accepted 26 December 2007

KEY WORDS: dry-stone masonry; retaining wall; masonry dam; yield design; homogenization

1. INTRODUCTION

Dry-stone walling is an ancient, vernacular form of construction that can be found in many areas
around the world. The structures are created by fitting interlocking stones together without mortar
to build a wall or a vaulting [1].

∗ Correspondence to: J. C. Morel, Département Génie Civil et Bâtiment (CNRS, URA 1652), Université de Lyon,
École Nationale des Travaux Publics de l’État, rue M. Audin, 69518 Vaulx-en-Velin Cedex, France.

E-mail: morel@entpe.fr

Copyright q 2008 John Wiley & Sons, Ltd.


1818 A. S. COLAS, J. C. MOREL AND D. GARNIER

Throughout the industrial revolution, a very large number of dry-stone edifices were raised as
retaining walls or rural constructions, but in the early 1900s dry masonry was largely abandoned
on behalf of industrial techniques. However, over the past few years, there has been a resurgence of
interest in dry-stone construction not only for the maintenance and assessment of existing structures
but also to provide the information required to encourage new projects. This phenomenon can
largely be accounted for by the low environmental impact of local material construction [2]. Yet,
the greatest problem has proved to be the lack of scientific knowledge about dry-stone construction
behaviour. Thus, in addition to dry-stone associations and masons’ promotional activities, scientific
studies have been launched to simulate dry-stone walls.
Dry-stone buildings are extremely difficult to model given their random nature, and their stability
is likely to depend on a lot of different factors such as material properties, block arrangement, or
the loading process. Although there have been only a few studies on dry construction design, two
different approaches can be distinguished.
Dry-stone walls are often assumed to behave as homogeneous rigid structures that yield to solid
mechanics. Wall safety is actually checked by simple limit equilibrium analysis and qualitative
visual examination. Limit equilibrium analysis corresponds to a simple but pragmatic approach
that has globally proved satisfactory, but it fails to take into account the characteristics of dry-
stone internal behaviour. Thus, some scientists have incorporated characteristics of dry-stone walls
into the basic limit equilibrium analysis, in an attempt to construct a more realistic model [3–5].
However, these simulations still rely on observations and empiricism and thus are not implemented
in a rigorous framework.
A second attempt to model dry-stone walling assumes that it is made of distinct pieces of stone
linked by contact laws and thus resorts to micromechanics. Dickens and Walker [6] pioneered
this method using the distinct element code UDEC to model walls at Great Zimbabwe National
Monument. Harkness et al. [7] have applied this technique to Burgoyne’s field trials [8], and their
study was later expanded on [9–11]. UDEC is based on the distinct element method (DEM). This
program is a powerful device to simulate masonry behaviour under loading until failure and can
supply plentiful information on the stress distribution of the structure or even the velocity of the
blocks. However, it proves to be quite complex and is not yet ready as a practical solution for the
calculation of ultimate bearing capacities.
Investigations on masonry have led to the development of an alternative to micro and macrome-
chanics by the way of homogenization. It relies on a multi-scale approach and results in deriving
physical characteristics of the homogenized masonry from those of its constituent materials. Pande
et al. [12] first developed this method in order to find elastic properties of periodic masonry. In
1995, Antoine [13] has implemented the homogenization of periodic masonry in a rigorous way. A
homogenization of non-periodic masonry has recently been performed by Cluni and Gusella [14]
to deal with old masonry. These studies rely on elastic characteristics and provide the behaviour
law of the masonry. In the same way, de Buhan and de Felice [15] have undertaken a homogeni-
zation of periodic brick structures in the framework of yield design theory and determined a yield
strength criterion for this kind of masonry.
This later approach by de Buhan and de Felice has been adopted in the present paper because
it complies with both theoretical and practical expectations for dry-stone modelling. The work
presented here initiates the use of a multi-scale approach to assess dry-stone retaining structure
stability under an external loading. First, the structure is homogenized, which boils down to finding
a macroscopic yield criterion based on the microscopic structure of the wall. Then, wall stability is
evaluated, thanks to yield design theory. This method proves to be relevant for simulating dry-stone

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1819

walls since the heterogeneity of the structure is taken into account while maintaining a pragmatic
process.
First, investigations concentrate on the yield design of dry-stone retaining structures. Then, the
results obtained are compared with those found by a simple limit equilibrium analysis and finally
validated on experimental and numerical data.

2. YIELD DESIGN OF DRY-STONE STRUCTURES

The model presented in this work utilizes yield design theory [16, 17]. Yield design provides the
ultimate bearing capacity of a structure under a multi-parameter loading, which is based on the sole
knowledge of the resistance criteria of the constituent material. It results in constructing a virtual
velocity field with respect to kinematic conditions and then writing the balance between external
forces and maximum resisting works. This method was first devoted to soil mechanics [18, 19], but
it has been expanded on rock mass [20, 21] and periodic masonry [15, 22]. The following section
describes the procedure to model a dry-stone retaining structure using this method.

2.1. General description of the model


Yield design theory requires only three kinds of parameters, which are geometry of the system,
loading mode, and resistance of the constituent material.
The system considered in this study is a retaining structure of height h, thickness at the top l,
batter 1 , and counter-slope 2 (Figure 2). It is simulated as a homogenized continuous medium.
The wall is founded onto rigid bedrock, and the bed joints are inclined with an angle  from
horizontal.
This system is subjected to its constant weight  and to an h l high external load. The external
pressure p l on the structure is assumed to have a triangular distribution inclined at an angle
of  from the back face and is proportional to its unit weight l and pressure coefficient K l .
This distribution corresponds to water pressure (l = w , K l = 1,  = 0) and can thus be used to
model masonry dams. On the other hand, this loading can also be linked to Caquot’s earth
pressure distribution (l = s , K l = K a ,  = s ); therefore, the applications of the model could
be extended to dry-stone earth-retaining walls, in a first approximation. Comparisons with
limit equilibrium and distinct element simulations will tend to corroborate the validity of this
hypothesis.
In this study, dry-stone construction is simulated by a regular and periodic structure. This
hypothesis leads to an idealization of the structure that can be justified as regard to the apparent
regularity (linear bed joints, use of pins to prevent blocks from rotating) of the well-built dry-stone
masonry (Figure 1). Further validations of this approximation will be discussed later.
Therefore, the resistance criterion found by de Buhan and de Felice [15] in their analysis of
periodic block masonry can be used here. This criterion results from a homogenization process
of the masonry implemented within yield design theory. It is based on the strength capacities
of the constitutive material, the contact laws between the blocks—considered as Mohr–Coulomb
strength criterion depending on the cohesion C and the friction angle  of the joints—and the
virtual kinematics (here, rigid translation) imposed to each block. In the case of dry masonry, the
cohesion C can be set to 0 as joints prove to be purely frictional, and the criterion by de Buhan

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1820 A. S. COLAS, J. C. MOREL AND D. GARNIER

(a) (b) (c)

Figure 1. Simulation of dry masonry: (a) dry-stone masonry; (b) dry-stone representative element; and
(c) idealized representative element.

Figure 2. Failure mechanism of the homogenized dry-stone wall.

and de Felice becomes




⎪ −D11 0


⎨ 2a
hom ( D ) = 0 if tan |D11 | b D22 (1)



⎪ 1
⎩ |D12 + D21 | tan D11 + D22
tan 

where hom is called support function and represents the maximum work density of the homoge-
nized masonry macroscopic strain rate D.
Considering these three parameters, it is possible to determine the failure strength domain using
yield design. This analysis focuses on providing the ultimate backfill height h l+ that the system
can stand.

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1821

2.2. Failure mechanism of the structure


A rigid block failure mechanism (Figure 2) will be considered in this study because it is a simple
model that is kinematically admissible. This choice proves to be accurate as regard to empiricism
[8, 23] and photogrammetric measurements achieved on experimental walls by Villemus et al. [5].
The wall breaks along OO with normal n, where O is the toe of the wall and O  is a point
of the back face, where coordinates −lr and h r depend on  (the angle formed by OO from
X 1 ) so that tan  ∈ [0, (h +l(1+1 ) tan )/(−h tan +l(1+1 ))]. The lower part of the wall OO C
remains fixed, whereas the trapezium OABO is given a virtual velocity
v(X ) = − e3 ∧ X (2)
where  is the velocity and (>0) is the angular velocity of point O.

2.3. Work of external forces


The work of external forces W e is defined by
 
W = .v dV + (
.n).v dS
e
(3)
V S

Considering the virtual failure mechanism presented above, W e can be expressed as a cubic
polynomial in h l :
W e = p3 (, , )h l 3 + p2 (, , )h l 2 + p1 (, , )h l + p0 (, , ) (4)

2.4. Maximum resisting work


On the other hand, the maximum resisting work W mr can be expressed as
 
W =  ( d ) dV + hom (n, 'v() dS
mr hom
(5)
V S

where hom ( d ) is the maximum work density of the wall strain rate d. As a rigid block failure
mechanism is considered here (2), it can be inferred that no deformation occurs in the wall (d = 0)
and, consequently,
hom ( d ) = 0 (6)

where hom (n, 'v() is the maximum work density of the velocity jump 'v( across the failure line
OO with normal n. The homogenized criterion (1) established earlier can be used here

hom (n, 'v() = 0 (7)


provided n and 'v( respect the following inequations:
−n 1 v1  0
2a
tan |n 1 v1 |  n 2 v2 (8)
b
1
|n 1 v2 +n 2 v1 |  tan n 1 v1 + n 2 v2
tan 

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1822 A. S. COLAS, J. C. MOREL AND D. GARNIER

Considering (6) and (7), the maximum resisting work (5) becomes
W mr = 0 (9)
under conditions (8) where 'v(X )( = − e3 ∧ X :
2
1  0 and  tan 
1
  (10)
h +l(1+1 ) tan  1
0  tan  min ,
−h tan +l(1+1 ) tan 

2.5. Ultimate external load


The kinematic approach of yield design theory [16, 17] asserts that the work of the corresponding
external forces in a kinematically admissible failure mechanism has to remain lower than the
maximum resisting work of this mechanism:
W e W mr (11)
Now it has been established that W mr = 0; therefore, the structure will prove to be potentially
stable for any load height h l ensuring that W e is negative. As W e is a cubic polynomial in h l ,
Cardano’s method can be used to find the positive root h l0 verifying
∀h l ∈ [0, h l0 ], W e (h l )0 (12)
The upper-bound value of h l is given by the minimum of h l0 with respect to the kinematic
parameters , , and under conditions (10):

h l+ = min {h l0 (, , )} (13)


,,

The study of h l0 provides opt , opt , and opt so that

h l+ = h l0 (opt , opt , opt ) (14)

Numerical studies on h l0 prove that h l0 is minimal for → 0 (sliding failure mode) or  → 0


(overturning failure mode). Therefore, only these two particular cases have to be treated to find
the ultimate backfill height.

2.5.1. Sliding failure mode. For a sliding failure ( = 0), the work of external forces W e is
a quadratic polynomial in h l . It has a unique positive root h l0 sli depending on  and , and

verifying (12)

 1 2 /1
sli
h l0 (, ) = h r + l((2+1 +2 )h −(1+1 +2 )h r ) (15)
K l l cos(+2 ) 1−tan(+2 )2 /1
sli with respect to  and
The ultimate backfill height by sliding h lsli is found when minimizing h l0
sli (, ) increases in  and  / , it is minimal when  and  / achieve
 under (10). As h l0 2 1 2 1
their minimal values, which means that opt = 0 and (2 /1 )opt = tan .

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1823

It follows that

 1 tan(+)
h lsli = h r 0 + l((2+1 +2 )h −(1+1 +2 )h r 0 ) (16)
K l l cos(+2 ) 1−tan(+2 ) tan(+)

where
tan 
h r 0 = h r (0) = − l(1+1 +2 )
1−tan 2 tan 

2.5.2. Overturning failure mode. On the other hand, for a rigid body rotation ( = 0), W e remains
a cubic polynomial. It follows that

W e = h l 3 +q1 ()h l +q0 () (17)


ov verifying (12) that can also be given by Cardano’s method.
It has a unique positive root h l0
This root depends only on the failure angle . The optimum value opt is completed by finding
ov () equals zero.
where the derivative of h l0
Thus, the ultimate backfill load to overturn the wall can be expressed as

h lov = h l0
ov
(opt ) (18)

As previously said, sliding and overturning represent the two limits for the ultimate height; thus,

h l+ = min{h lsli , h lov } (19)

It can be noted that the analytic solution (19) does not depend on the block size parameters a
and b, which amounts to saying that the ultimate load is independent of the slenderness ratio of
the blocks in the case of dry-stone masonry retaining structures under linear pressure distribution.
This result could prove to be changed for other wall geometries or different types of loading (e.g.
seismic loading).
Yield design modelling of dry-stone structures enables one to find an original analytical expres-
sion of the ultimate bearing loads of the wall, depending on the sole resistance of the raw material.
In addition, this technique gives an indication of the failure mode (sliding or overturning, failure
angle) of the structure.

3. COMPARISON WITH LIMIT EQUILIBRIUM

Yield design theory has provided an analytical expression of the ultimate load to prevent a dry-
stone structure from collapsing. This model can be evaluated by comparing it with a simple limit
equilibrium analysis, which is the current method used by engineers to design gravity retaining
structures.
In this model, it is necessary to consider the wall as a monolithic, rigid solid in order to perform
a limit equilibrium analysis. The structure is subjected to its weight P, the external loading pressure
P l on its back, and the reaction R of the foundation (Figure 3).

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1824 A. S. COLAS, J. C. MOREL AND D. GARNIER

Figure 3. Limit equilibrium analysis of dry-stone wall.

The sliding limit of the wall is reached when the sum of the forces acting on the structure equals
zero, that is,
 K l l K l l cos(+2 ) 2
l((2+1 +2 )h +(1+1 +2 )h r 0 )+ sin(+2 )h l 2 − hl = 0 (20)
2 2 2 tan(+)

Therefore, the ultimate backfill height h lsli happens to be the same as that achieved with yield
design theory (16) because yield design results are performed with the wall sliding along the
foundation (opt = 0). Nevertheless, it has to be noted that limit equilibrium assumes that the
sliding occurs along bed joints, whereas yield design gives this condition as a result.
Similarly, the overturning equilibrium is reached when the sum of the moments acting on the
structure equals zero. Thus, this sum can be expressed as a polynomial that is linked up with that
found by yield design (17) where setting  = 0. Since the ultimate backfill height found by yield
design is computed with the optimal value opt , it is obviously lower than that determined by
limit equilibrium analysis.
Therefore, the yield design method provides better prediction of dry-stone wall ultimate capac-
ities by taking into account the possible internal failure of the masonry and thus represents a step
to make the link between theory and practice. In addition, the rigorous framework in which this
model is implemented makes it possible to expect further improvements based on this present
work.

4. APPLICATION TO FULL-SCALE EXPERIMENTAL DRY-STONE STRUCTURES

The analytical results previously established are used to assess the stability of experimental full-
scale retaining structures and thus evaluate the relevancy of the yield design model.
As part of his PhD thesis, Boris Villemus has carried out experimental studies on full-scale dry-
stone retaining structures under water pressure with the intent to understand the masonry failure
process [5]. Although these experimental tests do not represent real structures, they can be related
to the case of masonry dams, which are made with a dry rockfill masonry. Five walls were built
with different geometries and raw materials and loaded by water pressure until they collapsed. The
pathologies have been measured and recorded throughout the experimental study. The geometrical
and physical characteristics of these walls are summarized in Table I.

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1825

Table I. Geometrical and physical characteristics of Villemus’s walls [5].


Wall’s characteristics Wall 1 Wall 2 Wall 3 Wall 4 Wall 5
Height h (m) 1.85 1.85 3.85 1.90 4.10
Thickness l (m) 0.60 0.90 1.20 0.65 1.16
Batter 1 (deg.) 8.5 0 8.5 6.8 8.5
Counter slope 2 (deg.) 0 0 0 0 0
Wall unit weight  (kN/m3 ) 15.4 15.0 15.7 16.0 18.0
Joint friction angle  (deg.) 36 36 36 36 28.5
Joint inclination  (deg.) 0 0 0 −4 −8.5

Table II. Comparison of ultimate backfill height for different methods.


Wall 1 Wall 2 Wall 3 Wall 4 Wall 5
Model S O S O S O S O S O

YD h l+ (m) 1.75 1.82 1.90 1.86 3.62 3.75 1.97 1.87 4.04 4.00
 (deg.) (0) (16.0) (0) (15.6) (0) (16.0) (0) (16.1) (0) (17.1)
LE h l+ (m) 1.75 1.85 1.90 1.89 3.62 3.81 1.97 1.92 4.04 4.14
Ex. h l+ (m) 1.59 — 1.75 3.22 — 1.84 3.48 —
 (deg.) (u) (u) (11.5) (7.5) (8.5)

Ex. failure

S = Sliding, O = Overturning.
YD = yield design; LE = limit equilibrium; Ex. = experiments.
Note: • (u) means that the failure angle of the wall was not measured in case of walls 1 and 2.
• entries in bold indicate the ultimate loading height.

It should be noted that all the data presented here were accurately measured on site and can
be found in [5]. Therefore, this work can be directly exploited to test the validity of yield design
theory on dry-stone retaining structures. Numerical computations have been performed on both
yield design and limit equilibrium analyses. The critical backfill height and the failure angle of
the wall found for each failure mode are collected and compared with the actual failure height and
mode that have occurred (Table II).
Theoretical critical heights are of similar order with experimental ones, and yield design results
happen to be in closer agreement with experiments than limit equilibrium analysis. It is interesting
to point out that yield design also succeeds in predicting the type of failure, except for wall 5.
The five walls collapsed under water pressure, mainly by sliding, even though the two failure
modes give numerical results that are similar in order. The failure angle of the wall is about 10◦
to horizontal.
Sliding ultimate heights reach the same values with both models, as previously expected. In
reality, yield design provides a zero optimal failure angle and slightly overestimates the sliding
ultimate height (error rate from 7 to 16%). These differences with experimental data could be

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1826 A. S. COLAS, J. C. MOREL AND D. GARNIER

due to the fact that the blocks are given a rigid translation motion in the homogenization of the
masonry. Thus, the rotation of the bed joints, which actually occurs when dry-stone walls fail
by sliding, is not assimilated in this model. Improvements could be achieved by changing the
masonry’s yield criterion.
For overturning failures, yield design provides lower values than limit equilibrium, meaning that
yield design is better for safety. In addition, it accurately estimates the critical height (error less
than 6%) and the potential angle opt of overturning. This improvement is particularly interesting
for practical concerns as most dry-stone retaining walls fall by overturning.
Numerical tests have confirmed that yield design could provide closer predictions of the ultimate
bearing capacity and failure mode of the system than simple limit equilibrium analyses, although
both methods require the same supply of data and computation. Similarly, the consistence with field
trial results enables the practical validation of this method as an effective tool to assess dry-stone
retaining structure stability when loaded by water pressure. In practice, these results mean that
yield design can be used in an attempt to simulate masonry dams, with the differences previously
pointed out.

5. COMPARISON WITH DEM

A second means of ensuring the validity of the yield design model consists in verifying its
robustness towards other simulation processes. A parametric analysis based on experimental data
will be conducted and compared with the DEM. This analysis relates to Burgoyne’s field trials [8].
In 1834, Lieutenant-General Sir John Burgoyne undertook full-scale tests on four dry-stone
walls. Each wall was made of granite and of the same height (6.1 m) and volume (19.2 m3 ), but
with different cross sections (Figure 4). The bed joints of each wall were perpendicular to the
front face. These four walls were backfilled with uncompacted earth and their deformations were
observed. These experiments were used to evaluate dry-stone wall behaviour and more especially
the influence of geometry on their stability.
Since they remain the most important full-scale trials in scientific history, many studies have
been devoted to these experiments. In 2000, Harkness et al. [7] first modelled Burgoyne’s walls
using the distinct element code UDEC. Their results were then widely exploited. Expanding on this
work, Zhang et al. [10] reported on a comparison between distinct and finite element modelling of
Burgoyne’s walls. Powrie et al. [9] and Claxton et al. [11] have reproduced Harkness et al. results
in order to assess the influence of parametric variations on wall behaviour, using, respectively,
deformable and rigid block distinct element models.

Figure 4. Section profiles of Burgoyne’s experimental walls (front face on the right).

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1827

The lack of information concerning the physical characteristics of the backfill or even the
wall has led us to concentrate our comparisons on the sensitivity of the different models towards
parametric variations out of any experimental consideration. In this study, the same parametric tests
as those of Powrie et al. [9] and Claxton et al. [11] were carried out on the yield design model.
These tests aimed at investigating the properties that could control stability, and also showing how
the model could behave, and comparing its evolution with that from the DEM. This will enable
one to assess the choice of Caquot’s distribution as a representation of soil pressure.

5.1. Comparison with Powrie et al. study


Powrie et al. [9] have concentrated their study on modelling Burgoyne’s walls A and B with a
deformable block model. According to Burgoyne’s report [8], these two walls remained stable at the
end of the filling. Powrie et al. have decided to explore what could happen if diminishing the stone
and soil friction angles and the joint inclination. The original values of geometrical parameters are
taken from Burgoyne [8], and those of physical characteristics (not given by Burgoyne) are given
by Harkness et al. [7] and recorded in Table III.
In this paper, the same investigations were undertaken with the yield design and limit equilibrium
models and compared with distinct element results. Limit equilibrium analysis remains to be the
standard measure that maximizes the critical height.
Powrie et al. have evaluated the influence of joint friction angle on walls A and B with horizontal
( = 0◦ ) and inclined joints ( = −11.3◦ ). The friction angle has been reduced from the initial value
of 45◦ to 20◦ in steps of 1◦ or 2◦ , while the other parameters remained constant, and the wall
behaviour has been simulated for each variation (Figure 5).
First, it can be noted that the results provided by yield design are linear, whereas the DEM
results are given as an interval. This is due to the choice of Powrie et al. to build the backfill by
0.31 m high layers. It is also necessary to point out that the authors have stopped their computation
as soon as the backfill height overcame the wall height and then considered that the wall was
stable. The hatched zone in Figures 5 and 6 represents the potential stability of the walls not
calculated by Powrie et al.
When joints are horizontal, walls A and B fail by sliding as the friction angle of the stone is
reduced below 22◦ . These results are similar for both methods.
Yet, with inclined joints, both walls A and B remain stable in the yield design model whatever
the friction angle of the stone, which differs from distinct element results. Two phenomena can
account for this difference. For low values of , this is due to the overestimation of sliding height
in yield design, as previously noted in Section 4. Further calculations on yield design show that
wall A would slide for 11◦ and wall B for 12◦ , which are not represented here because they
are too low to stand for real block friction angles. As  increases, the difference can be ascribed
to the little dependence of  for overturning ultimate height, which results from the choice of a
rigid block failure mechanism.

Table III. Original values of uncertain physical parameters of


Burgoyne’s walls by Harkness et al. [7].

Parameter Unit weight (kN/m3 ) Friction angle (deg.)


Wall 22.7 45
Backfill 15.5 28

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1828 A. S. COLAS, J. C. MOREL AND D. GARNIER

(a)

(b)

Figure 5. Effect of wall friction angle on walls A (a) and B (b).

Powrie et al. have also studied the effect of backfill properties on the wall. They have modelled
walls A and B for a soil friction angle varying from 28◦ to 20◦ (Figure 6).
It is interesting to see that the results provided by the different models are quite close for this
parametric variation. The two models act in a similar manner: diminishing the friction angle of the
backfill leads to a decrease in the stability of the wall, which tends to overturn. Wall A collapses
for l below 22◦ (24 according to Powrie et al.) and wall B fails when l 25◦ (27 for Powrie
et al.). There is an average difference of 0.3 m (4%) between the ultimate heights.
On the other hand, the inclination of the bed joints has no effect on the ultimate height in
this case.

5.2. Comparison with Claxton et al. study


Claxton et al. have also investigated the effect of different parameters on wall stability with the
distinct element code UDEC. They have chosen to assess the stability of wall D (Figure 4) with a
rigid block model, depending on wall thickness, unit weight, friction angle, or backfill properties.
The backfill layers of this study were 0.61 m high.
The same conclusions as for the effect of the stone (Figure 7(a)) and backfill (Figure 7(b))
friction angles can be drawn, although these parametric variations were conducted on another wall
(wall D) with a different geometry and using a different model (rigid block).

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1829

(a)

(b)

Figure 6. Effect of soil friction angle on walls A (a) and B (b).

Besides, the effects of the unit weights of the wall (Figure 7(c)) or the backfill (Figure 7(d))
appear only in yield design calculations. Wall stability decreases with diminishing density and
while increasing the unit weight of the soil. The yield design model proves to be more sensitive
to variations because it provides continuous analytical solutions, whereas the DEM gives discrete
numerical results. More precise computations could be achieved by diminishing fill layers in the
discrete element simulation, at the expense of a longer calculation time.
As for the influence of wall thickness (Figure 7(e)), the two models happen to be in total
agreement, increasing wall thickness reinforces stability.
This parametrical analysis shows that the yield design model acts mainly in a similar manner
with the DEM at failure. Yield design has encountered some difficulties in taking into account the
influence of the joint friction angle. However, the analytical form makes it possible to account for
the sensitivity of the model and provides immediate and precise results. These conclusions tend to
prove that Caquot’s model meets the requirements of soil pressure on dry-stone retaining walls.
Thus, these findings have highlighted the influence of different parameters on wall stability. As
for geometry, a large battered wall with inclined joints is more likely to resist external pressures.
Also, a high density of the wall contributes to its stability. On the other hand, the higher the backfill
earth pressure coefficient and unit weight, the less resistant the wall.

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1830 A. S. COLAS, J. C. MOREL AND D. GARNIER

Figure 7. Stability of wall D depending on joint friction angle (a), backfill friction angle (b), wall unit
weight (c), backfill unit weight (d), and wall thickness (e).

6. CONCLUSIONS

Dry-stone masonry is a widely expanded form of construction, and an important heritage can be
found all around the world. However, no structural design method has been validated until recently.
Limit equilibrium analysis proves to be limited when considering internal failure within the wall,
and DEM is too complex for practical determination of ultimate loads.
This study has presented a method to assess dry-stone retaining structure stability under external
pressure, thanks to homogenization and yield design theory. This technique takes into account the
heterogeneity of the structure while keeping a macroscopic approach. Yield design provides an
analytical expression of the ultimate backfill height only depending on geometrical and physical
characteristics of the wall and the backfill and requiring no calculation time to compute numerical
results. Therefore, this method is directly applicable for practical design.
The reliability of this model has been tested on field trials and compared with simulations
found in literature, with two main conclusions emerging. First, numerical tests have proved the
robustness of yield design over other models. When compared with limit equilibrium analysis, it
proves to be more secure, as expected, and provides complementary information on the failure
mode. In addition, there is close agreement with DEM results at failure, although yield design
requires less parameters and computation. Based on a homogenization process, this study makes

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
YIELD DESIGN OF DRY-STONE STRUCTURES 1831

the link between macro and micromechanical approaches. Furthermore, experimental assessment
has shown the validity of yield design modelling of dry-stone retaining structures.
These results tend to validate the yield design method as an efficient means of stability assessment
for dry-stone structures. Yield design modelling is a relevant alternative, outlining a strict analytical
but practical approach, close to reality and easy to carry out since requiring only few parameters
and no programming. This method accurately simulates retaining structure failures by overturning,
but it could be interesting to undertake further investigations on sliding to improve masonry dam
modelling. Ongoing research also concentrates on dry-stone earth retaining walls by setting up a
yield design model on both backfill soil and retaining wall to assess the overall stability of the
system.

ACKNOWLEDGEMENTS
The authors would like to thank the Centre in Innovative Construction Materials at the University of Bath
and more especially Paul MacCombie and Chris Mundell for their precious advice and revisions. They
would also like to acknowledge Claude-Henri Lamarque (ENTPE) for his support and valuable comments.

REFERENCES
1. Lassure C, Repérant D. Cabanes en pierres sèches de France. Édisud, Aix-en-Provence, 2004.
2. Morel JC, Mesbah A, Oggero M, Walker P. Building houses with local materials: means to drastically reduce
the environmental impact of construction. Building and Environment 2001; 36(10):1119–1126.
3. Cooper MR. Deflections and failure modes in dry-stone retaining walls. Ground Engineering 1986; 19(8):40–45.
4. Villemus B. Étude des murs de soutènement en maçonnerie de pierres sèches. Ph.D. Thesis, ENTPE-INSA, Lyon,
2004.
5. Villemus B, Morel JC, Boutin C. Experimental assessment of dry-stone retaining wall stability on a rigid
foundation. Engineering Structures 2007; 29(9):2124–2132.
6. Dickens JG, Walker PJ. Use of distinct element model to simulate behaviour of dry-stone walls. Structural
Engineering Review 1996; 8(2/3):187–199.
7. Harkness RM, Powrie W, Zhang X, Brady KC, O’Reilly MP. Numerical modelling of full-scale tests on drystone
masonry retaining walls. Géotechnique 2000; 50(2):165–179.
8. Burgoyne J. Revetments of retaining walls. Corps of Royal Engineering Papers 1853; 3:154–159.
9. Powrie W, Harkness RM, Zhang X, Bush DI. Deformation and failure modes of drystone retaining walls.
Géotechnique 2002; 52(6):435–446.
10. Zhang X, Koutsabeloulis NC, Hope S, Pearce A. A finite element analysis for the stability of drystone masonry
retaining walls. Géotechnique 2004; 54(1):57–60.
11. Claxton M, Hart RA, McCombie PF, Walker PJ. Rigid block distinct-element modeling of dry-stone retaining
walls in plane strain. Journal of Geotechnical and Geoenvironmental Engineering 2005; 131(3):381–389.
12. Pande GN, Liang JX, Middleton J. Equivalent elastic moduli for brick masonry. Computers and Geotechnics
1989; 8(3):243–365.
13. Antoine A. Derivation of the in-plane elastic characteristics of masonry through homogenization. International
Journal of Solids and Structures 1995; 32(2):137–163.
14. Cluni F, Gusella V. Homogenization of non-periodic masonry structures. International Journal of Solids and
Structures 2004; 41(7):1911–1923.
15. de Buhan P, de Felice G. A homogenization approach to the ultimate strength of brick masonry. Journal of the
Mechanics and Physics of Solids 1997; 45(7):1085–1104.
16. Salençon J. Calcul à la rupture et analyse limite. Presses de l’École Nationale des Ponts et Chaussées, Paris,
1983.
17. Salençon J. An introduction to the yield design theory and its application to soils mechanics. European Journal
of Mechanics – A/Solids 1990; 9(5):477–500.
18. de Buhan P. Approche fondamentale du calcul à la rupture des ouvrages en sols renforcés. Thèse d’État, Université
Paris VI, 1986.

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag
1832 A. S. COLAS, J. C. MOREL AND D. GARNIER

19. Jellali B, Bouassida M, de Buhan P. A homogenization method for estimating the bearing capacity of soils
reinforced by columns. International Journal for Numerical and Analytical Methods in Geomechanics 2005;
29(10):989–1004.
20. Bekaert A, Maghous S. Three-dimensional yield strength properties of jointed rock mass as a homogenized
medium. Mechanics of Cohesive-Frictional Materials 1996; 1(1):1–24.
21. Bekaert A, de Buhan P, Maghous S. Failure design of jointed rock structures by means of a homogenization
approach. Mechanics of Cohesive-Frictional Materials 1998; 3(3):207–228.
22. Sab K. Yield design of thin periodic plates by a homogenization technique and an application to masonry walls.
Comptes Rendus Mécanique 2003; 331(9):641–646.
23. Constable C. Retaining walls—an attempt to reconcile theory with practice. American Society of Civil Engineers
1875; 3:67–75.

Copyright q 2008 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1817–1832
DOI: 10.1002/nag

View publication stats

You might also like