Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Zhang, X., Koutsabeloulis, N. C., Hope, S. & Pearce, A. (2004). Géotechnique 54, No.

1, 57–60

TECHNICAL NOTE

A finite element analysis for the stability of drystone masonry retaining


walls
X . Z H A N G  † , N. C . KO U T S A B E L O U L I S  , S . H O P E  a n d A . P E A R C E 

KEYWORDS: numerical modelling and analysis element modelling approach (Harkness et al., 2000); reason-
ably close agreements have been obtained in terms of the
critical stable height of backfill and the failure character-
INTRODUCTION istics. In this technical note, the behaviour of Burgoyne’s
Masonry-faced walls are common in many countries, parti- wall D, a plain vertical wall of uniform thickness of 1 m
cularly in hilly areas. Many of these structures are used to (Fig. 1) has been simulated by using the finite element code
support highways or buildings. The stability of a drystone VISAGE TM (1999).
retaining wall is likely to depend on factors not being taken To model the behaviour of wall joints, a finite element
into account in conventional methods of analysis, for exam- analysis uses techniques that are different from those used
ple in a limit equilibrium analysis, because the deformation with a discrete element analysis. The discrete element
behaviour of a drystone wall involves: approach UDEC (Itasca, 1996) is a type of numerical analy-
sis based on discontinuum modelling, which explicitly simu-
(a) large plastic strain
lates the discontinuities of the wall, including their geometry
(b) interaction between wall, backfill and base
and mechanical properties. The finite element analysis ap-
(c) the placement process of backfill
plied in this study is based on continuum modelling, with
(d ) the behaviour of joints within a wall.
which a discontinuous medium, for example the wall, is
For such types of structure, a numerical modelling analysis treated as an equivalent continuum or is simulated using
may provide a useful way to understand the failure mechan- joint elements. In this study, two series of analyses have
isms involved and to quantify the degree of wall stability if been carried out to simulate the wall. In one, the wall was
the numerical modelling analysis can reproduce the behav- treated as an equivalent continuum, and lower stiffness and
iour of wall deformation observed in fields and/or field weaker strength were used to account for the presence of
trials. wall joints; in the other, joint elements were used to simulate
Four full-scale drystone masonry retaining walls were the wall. In this case, the parameters of wall joints, such
built and tested at Kingstown (Burgoyne, 1853), and the as their spacing, direction, stiffness and strength, were
behaviour of these walls was reproduced by using a discrete simulated.

Fig. 1. Cross-section and failure mechanisms of the wall tested by General J. Burgoyne,
1834 (from Burgoyne, 1853)

Manuscript received 25 July 2003; revised manuscript accepted 23


September 2003.
Discussion on this paper closes 1 July 2004, for further details see
p. ii.
 VIPS (Vector International Processing Systems Limited), Brack-
nell, UK.
† Department of Earth Science and Engineering, Imperial College,
London, UK.
58 ZHANG, KOUTSABELOULIS, HOPE AND PEARCE
This numerical analysis followed the same process of FEMGV 6.2-01: VIPS 8-AUG-2002 16:05 f2_wall
backfill placement as used in the field trial; the geometry of MODEL: WALL1b
the wall was similar to that used in the field trial and the 1m 8m
discrete element analysis; and the parameters for the backfill

4 ⫻ 0·5 m
were the same as those used in the discrete model. This
technical note presents the procedure for modelling such a

Base and back wall


drystone retaining wall, and the results obtained are com-
pared with those from the field trials and the discrete

1m
element analysis (see Fig. 2). Wall

6m
Backfill

1m
1m
DESCRIPTION OF NUMERICAL ANALYSIS
All of the analyses were carried out in plane strain. A

1m
typical analysis had 1456 four-noded quadrilateral elements
and 1537 nodes. A finite element mesh is given in Fig. 3, Y
and the boundary conditions and dimensions of the wall Z
X
simulated Burgoyne’s experiments and were the same in the
discrete element analysis. Placement of backfill was mod- Fig. 3. Typical finite element grid of the wall and the boundary
elled by adding the soil in layers one layer by one, as in the conditions used in this study.
actual construction process. The wall was constructed in
advance of the backfill placement, which was simulated by
adding soil layers in lifts of 1 m at a time for the first 4 m, deformation during backfill placement. In all of the analyses
and then 0.5 m until the wall collapsed. For each of these it was assumed that the pore water pressures throughout the
added soil layers 50 load-steps were used to monitor the backfill were zero at every stage.
The wall, the soil and the natural back wall and base were
Wall
all modelled as elastic/Mohr–Coulomb plastic materials. The
values of the material properties used are given in Table 1.
The wall and the bedrock in Burgoyne’s experiments were
both of granite. Intact granite has a mass density of about
2650 kg/m3 (Goodman, 1980), and the wall had a rather
lower mass density of about 2270 kg/m3 , as quoted by
Burgoyne (1853).
Backfill
Natural wall
RESULTS
For the equivalent continuum model the deflection of the
wall increased with increasing height of the backfill. The
deformation of the wall resulted in the plastic failure of the
Base
backfill soil. A distinctively, highly deformed region formed
within the backfill soil with considerable subsidence against
the wall, particularly within the upper part. During the
Fig. 2. Geometry and boundary conditions used in the discrete placement of backfill up to a height of 5 m the wall was
element analyses by Harkness et al. (from Harkness et al., 2000) stable, but the wall failed with a further layer of 0.5 m.

Table 1. Material properties used in the finite element analyses


Property: intact rock/soil Soil Natural Wall
bedrock
(granite) Equivalent Joint
continuum element
Intact rock/soil
Mass density: kg/m3 1550 2650 2270 2270
Young’s modulus: MPa 8 40 000 50 40 000
Poisson’s ratio 0.32 0.3 0.3 0.3
Angle of internal friction: degrees 28 45 30 45
Cohesion: MPa 0 100 0.2 100
Tensile strength: MPa 0 Infinite 0 0
Joints
Normal stiffness: MPa/m – – – 500
Shear stiffness: MPa/m – – – 100
Friction angle: degrees – – – 40
Dilation angle: degrees – – – 0
Cohesion: kPa – – – 50–70
Tensile strength: kPa – – – 50–70
Spacing: m – – – 0.3 for
vertical joints
0.25 for
horizontal
joints
STABILITY OF DRYSTONE MASONRY RETAINING WALLS 59
These analyses show that the backfill had a critical stable 0
height of 5 m for an increment of backfill height of 0.5 m
with the properties used. This agreed closely with the results ⫺0·1
obtained in Burgoyne’s field trials and the discrete element
analysis, where the critical stable height was 4.88 m for an
increment of 0.305 m in backfill height (Harkness et al., ⫺0·2
2000). Also, this analysis indicated that the wall failed in an
overturning fashion, which was observed in Burgoyne’s field

Displacement x: m
⫺0·3
trials (see Fig. 1) and in the discrete element model.
In the analyses with joint elements, the same critical Using joint elements
stable height of backfill of 5 m was achieved with the ⫺0·4 for the wall
parameters used in Table 1. Fig. 4 shows a comparison of
the horizontal displacements at the wall crest, for both the ⫺0·5
finite element analyses and the discrete element analyses,
during the placement of backfill. Apparently, the deformation
behaviour simulated in the finite element analysis with joint ⫺0·6 Using an equivalent
continuum model
elements was very close to that simulated in the discrete for the wall
element analysis, although the same critical backfill height ⫺0·7
was obtained in both the finite element analyses. This
suggested that the deformation of wall joints played an
⫺0·8
important role in the behaviour of wall deformation. Fig. 5 0 10 20 30 40 50
shows the histories of the horizontal deflection at the wall
Load step
crest, for the equivalent continuum model and the joint
element model in the finite element analyses, during the
Fig. 5. Comparison of horizontal deflection at wall crest during
placement of backfill from 5 to 5.5 m with 50 load-steps, at placement of backfill from 5 to 5.5 m with 50 load steps, at
which the wall failed. Using an equivalent continuum model which the wall failed
for the wall, the failure of the wall was dominated by ductile
deformation, and the horizontal deflection developed
smoothly from 0.1 m to about 0.7 m. Using joint elements
for the wall, the failure of the wall was characterised by stability of the wall, further numerical analyses were carried
brittle deformation behaviour, which was observed in fields out using the finite element code VISAGE TM . In these
and field trials (Burgoyne, 1853; Wong & Ho, 1997). The analyses, joint elements were used and the other parameters
wall collapsed suddenly when the horizontal deflection at the were kept unchanged. Slight variation in the mortar strength
wall crest approached a value of 0.12 m. has led to significant changes in the stability of the wall.
These results indicated that the deformation behaviour of Where the mortar strength (both the cohesion and tensile
a fractured medium could be simulated with joint elements strength) decreased from 60 to 50 kPa, the critical stable
in a finite element analysis. It is interesting to note that very height of backfill reduced from 5 to 4 m. On the other hand,
close agreement of simulating a complex structure, involving the critical stable height of backfill changed from 5 to 5.5 m
large plasticity, multi-loading process and joint deformation, where the mortar strength increased from 60 to 70 kPa.
could be achieved using two different types of numerical Field observation (GEO Report, 1982), experimental
modelling method. However, different numerical analytical results (Burgoyne, 1853) and numerical analyses (Harkness
approaches have their own merits and disadvantages. et al., 2000; Powrie et al., 2002) reveal that there were
To examine the effects of mortar strength of joints on the triangular fragments of the masonry remaining at the lower

0·8 Finite element analysis


(VISAGE) with an equivalent
continuum model for the wall
0·7
Horizontal deflection at the crest: m

0·6

0·5

0·4

Discrete element
0·3
analysis (UDEC)
Critical stable
0·2 backfill height
Finite element analysis
(VISAGE) with a joint element
0·1
model for the wall

0
0 1 2 3 4 5 6
Backfil height: m

Fig. 4. Comparison of deformation behaviour simulated using finite element analyses (VISAGE TM ) and discrete element
analyses (UDEC)
60 ZHANG, KOUTSABELOULIS, HOPE AND PEARCE
6
6

5
5

4
4
Wall height: m

Wall height: m
3
3

2
2

1
1

0
0 0·02 0·04 0·06 0·08 0
⫺50 ⫺25 0 25 50

Horizontal deflection: m Vertical stress, y: kPa


(a) (b)

Fig. 6. (a) Horizontal deflection along the outer wall surface; (b) vertical stress distribution within
inner wall surface.

inner corner of the walls. This is because the joints of a This highlights the fact that more realistic behaviour of a
drystone wall have a low strength, enabling failure mechan- fractured medium can be simulated if joint elements are
isms involving rupture surfaces passing through the wall to used.
develop. Fig. 6(a) shows the stable horizontal deflection Before failure, tensile stress develops within the inner wall
along the outer wall surface of the wall with a critical surface, indicating that a potential rupture is likely to form
backfill height of 5 m. The profile of horizontal deflection through the wall toe at an angle of about 458 with the base.
indicates that the deformation of the wall increased sharply This is in agreement with field observation.
above a height of 1 m. Fig. 6(b) shows the distribution of
the vertical stress along the inner wall surface where the
wall was stable with a critical backfill height of 5 m. The
inner wall surface was in tension where the wall height was ACKNOWLEDGEMENTS
below 1.75 m and the maximum value of the tensile stress The authors would like to thanks V.I.P.S. Limited for
in the vertical direction was located at about 1 m of the wall permission to publish this paper, and two anonymous re-
height. These features indicated that a potential rupture was viewers for their valuable comments.
likely to develop within the wall from the maximum tensile
stress point on the back face to the toe of the wall—a dip of
about 458 with the base. This was observed both in the field
trials (Fig. 1) and in the discrete element model (Harkness REFERENCES
et al., 2000; Powrie et al., 2002). Burgoyne, J. (1853). Revetments or retaining walls. Corps R. Engrs
Paper 3, 154–159.
GEO Report (1982). Report on the stability of old masonry retain-
CONCLUSIONS ing walls in Hong Kong, Vol. 1. Hong Kong: Geotechnical
Control Office, Public Works Department.
The analyses discussed in this note, which were carried Goodman, R. E. (1980). Introduction to rock mechanics. New York:
out to model the deformation behaviour of the wall using John Wiley & Sons.
the finite element code VISAGE TM , have shown that there Harkness, R. M., Powrie, W., Zhang, X., Brady, K. C. & O’Reilly,
are close agreements of the main features obtained using the M. P. (2000). Numerical modelling of full-scale tests on dry-
finite element analysis and those observed in field trials. stone masonry retaining walls. Géotechnique 50, No. 2,
Also, these results are very similar to those obtained in a 165–179.
discrete element analysis. Itasca (1996). Universal Distinct Element Code, Version 3.0. Min-
In the finite element analyses, the same critical stable nesota, USA: Itasca Consulting Group Inc.
height is obtained by using both an equivalent continuum Powrie, W., Harkness, R. M., Zhang, X. & Bush, D. I. (2002).
Deformation and failure modes of drystone retaining walls.
model for the wall and a joint element model. However, the
Géotechnique 52, No. 6, 435–446.
deformation behaviour obtained using a joint element model VISAGE TM (1999). The VISAGE TM System Version 7.0: User’s
for the wall is closer to that observed in the field trials and Guide. Bracknell: V.I.P.S. (Vector International Processing Sys-
the discrete element analysis than that obtained using an tems) Limited.
equivalent continuum model. In the joint element model the Wong, H. N. & Ho, K. K. S. (1997). The 23 July 1994 landslide at
wall collapsed suddenly, with the feature of brittle failure. Kwun Lung Lau, Hong Kong. Can. Geotech. J., 34, 825–840.

You might also like