Professional Documents
Culture Documents
net/publication/239410939
CITATIONS READS
44 2,676
4 authors, including:
William Powrie
University of Southampton
408 PUBLICATIONS 7,092 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Infrastructure Slopes: Sustainable Management And Resilience Assessment: iSMART View project
Achilles: Assessment, Costing and Enhancement of Long - Life, Long-Linear Assets View project
All content following this page was uploaded by William Powrie on 21 July 2015.
In this paper, the factors controlling the deformation of Dans cet exposé, nous cherchons les facteurs contrôlant
drystone retaining walls are investigated by means of la déformation des murs de soutènement en pierres
discrete element analyses. It is shown that toppling fail- sèches au moyen d’analyses d’éléments discrets. Nous
ure of unweathered drystone retaining walls is likely to montrons que les murs de soutènement en pierres qui
occur in a brittle manner, with wall crest deflections not n’ont pas été exposées au vieillissement climatique seront
exceeding 1% of the backfill height until the factor of enclins à basculer en se fragilisant, les fléchissements de
safety (based on soil strength) falls below 1·05. A com- la crête du mur ne dépassant pas 1% de la hauteur de
pressible sub-base and weathering of the blocks will both remblayage, jusqu’à ce que le facteur de sécurité (basé
tend to reduce the backfill height at failure to below that sur la résistance du sol) tombe en dessous de 1·05. Une
indicated by a limit equilibrium analysis. Bulging failure base de fondement compressible et le vieillissement clima-
is more likely to be associated with a deterioration in tique des blocs auront tendance à réduire la hauteur de
block joint stiffness due to weathering than a compres- remblayage au point de rupture à un niveau inférieur à
sible sub-base, although the latter will decrease the celui indiqué par une analyse d’équilibre limite. Le
reduction in joint stiffness needed to cause bulging fail- bombement du mur sera probablement associé à une
ure. Bulging is much less brittle than toppling, and the détérioration de la rigidité des joints entre blocs à cause
proximity to failure of bulging walls could in some du vieillissement plutôt qu’à une base de fondement
circumstances be assessed on the basis of the size of the compressible, bien que cette dernière diminue la perte de
bulge. rigidité des joints, perte nécessaire pour causer le bombe-
ment. Le bombement est bien moins fragile que l’écroule-
ment et le proximité des murs bombés de la rupture
KEYWORDS: numerical modelling and analysis; retaining pourrait, en certains cas, être évaluée en fonction de la
walls; soil/structure interaction dimension du bombement.
435
Delivered by ICEVirtualLibrary.com to:
IP: 152.78.179.112
On: Wed, 22 Sep 2010 16:56:19
436 POWRIE, HARKNESS, ZHANG AND BUSH
rock face) was prevented from moving horizontally, but was
free to move in the vertical direction. In all of the analyses
it was assumed that the pore water pressure throughout the
backfill was zero at every stage.
(Cooper, 1986). In the analyses described by Harkness et al. 10 Masonry density = 2270 kg/m3
(2000) the masonry blocks from which the wall was con- Friction angle of masonry joints = 45˚
Backfill friction angle = 20˚
Deflection of masonry at the top: cm
(a) the reduction of resisting moment due to the weight (b) the effect on the mechanism of failure within the
of the wall when the fulcrum is moved in from the toe wall—allowing blocks to roll, for example, where they
of the wall by the amount of the corner rounding would otherwise have enough effective width or height
(Cooper, 1986) to resist this tendency.
JOINT INCLINATION
Burgoyne’s walls A and B both had a 1 in 5 batter to the
front face. As the blocks used to construct each wall were
rectangular, construction of the walls with a planar (rather
than stepped) front face meant that the block joints were Fig. 3. Effects of lateral extent of backfill on stability of wall A:
similarly inclined at an angle of 11:38 (i.e. 1 in 5), except at (a) deflection with different ratios, R, of width to depth; (b) x-
the very bottom of the wall. Analyses have been carried out displacement contours during failure with R 0:52, and (c)
with R 1:64
for each wall with both inclined and horizontal masonry
joints, to investigate the influence of joint inclination on
overall wall behaviour. The results for the baseline properties
(458 joint friction angle, 288 soil friction angle and the backfill height of 6·1 m (20 ft) was achieved. The results for
stiffness moduli as given in Table 1) are shown in Fig. 4, the two walls are given in Fig. 5, in which the deflection of
and indicate very little effect of joint inclination for these the top of the wall is plotted against backfill height for all
joint and soil strengths. These results are part of a wider stable heights and all values of ö9.
study in which both soil and joint friction were varied for On the reasonable assumption that the active pressure of
the two joint inclinations, and are discussed later. dry backfill against a wall is approximately proportional to
depth, the total active force against the wall would be
expected to be proportional to the square of the depth,
EFFECTIVE FRICTION ANGLE OF THE BACKFILL whereas the moment of the force about its toe would be
Both wall A and wall B were stable at a backfill height of proportional to the depth cubed. Thus for a 5% reduction in
6·1 m (20 ft) using the baseline parameters (Table 1) with a backfill height from, say, 6·1 m (20 ft) to 5·79 m (19 ft), the
backfill friction angle of 288. A series of analyses was active force against the wall would fall by about 10% and
carried out for each wall in which the friction angle of the the moment of that force by about 14%. To increase the
backfill was reduced in steps of 18 or 28 to 208 to explore force or moment to their original values, the active pressure
the effect on wall movement and on the maximum backfill at the new backfill height would need to be increased by
height that the wall could support. For each backfill friction reducing the backfill friction angle. From an initial effective
angle a new analysis was carried out with the backfill built friction angle of 288, the required reductions in ö9 would be
up in stages of 0·3 m (1 ft) until the wall failed or the full approximately 2:68 and 4:08 (to 25:48 and 248) for force and
6 25˚
20˚
28˚
10 m (32 ft 10 in)
(a) 0
8 (a)
4 22˚
28˚
27˚
1 m (3 ft 3 in) 24˚
2
1m
0
10 m (32 ft 10 in) 4·2 4·4 4·6 4·8 5·0 5·2 5·4 5·6 5·8 6·0 6·2
Stable backfill height: m
(b)
(b)
Deflection of masonry at the top: cm
Wall B
4 strength) of approximately 1·04. Thus a wall with a factor of
safety of just 1·04 may have deflected by only 1% of its
2 Wall A height: this is discussed later.
The results of a second series of analyses, similar to that
described above but with horizontal and vertical joints, are
0
4·2 4·4 4·6 4·8 5·0 5·2 5·4 5·6 5·8 6·0 6·2
shown in Fig. 6. The stability of the walls shows very little
Stable backfill height: m change from the results of the first series (that is, not
(c) discernible at the 0·305 m (1 ft) resolution of the backfill
height changes), but for both walls the deflection at the top
Fig. 4. Cross-sections of walls with horizontal joints: (a) wall A; is rather less with horizontal joints than with inclined joints.
(b) wall B. (c) Comparison of stability. Compare with Fig. 1 for This is probably because the mode of deformation of the
inclined joints former (at least, before failure) includes sliding on the
joints: the mobilisation of active pressure by parallel out-
ward movement of the wall requires less movement of
the top of the wall than if the wall is constrained (by the
moment respectively. Either of these reductions could be resistance of the joints) to rotate about its base. The
consistent with the results for changes in maximum stable difference between the deformation patterns at failure is
backfill height given in Fig. 5, and the results do not there- shown in Fig. 7 for wall B. Fig. 7(a) shows rotation (by a
fore clearly distinguish between force and moment in caus- mechanism passing through the toe) for the wall with
ing failure. However, the mode of failure (by overturning) inclined joints, and Fig. 7(b) shows sliding on a joint near
suggests that moment instability is the more likely cause. the base for the wall with horizontal joints. It is worth
The brittleness of both walls is illustrated by the fact that, noting that relative sliding near the base of walls with
in all analyses leading to failure, any wall tilting more than horizontal joints would in reality be more noticeable than
0:28—that is, a top deflection greater than about 2 cm or the top-of-wall deflection of either giving, perhaps, better
0·3% of its height of 6·1 m (20 ft)—failed when subjected to warning of approaching instability.
a further 0·305 m (1 ft) of backfill. The converse appears to
hold (except for the analyses in which a reduced extent of
backfill was modelled) in that a wall with a deflection of JOINT FRICTION ANGLE
less than 0·3% of its height was able to withstand a further Analyses were carried out for both walls with both
0·305 m increment of backfill. Furthermore, walls with the inclined and horizontal block joints, for a fixed backfill
maximum computed stable deflection of 6 cm (1% of wall friction angle of 288 and block joint friction angles ranging
height) failed when the backfill friction angle was reduced from 208 to 458. The interface friction angle between the
by 18. A 18 reduction in backfill friction angle from 288 wall and the backfill remained at 288.
corresponds to a reduction in factor of safety (based on soil For the walls with inclined joints, the results of these
0
(a)
Deflection of masonry at the top: cm
8
(a)
20–27˚ Will fail by overturning 20˚
6 Masonry density = 2270 kg/m3
28˚ Stable at 6·1 m
Friction angle of masonry joints = 22˚
Backfill friction angle = 28˚
4 28˚ Corner radius of masonry blocks = 1 cm –16
22˚ 26˚
2 24˚ 27˚
–14
–12
0
–10
4·2 4·4 4·6 4·8 5·0 5·2 5·4 5·6 5·8 6·0 6·2 –8
Stable backfill height: m
–6
(b)
–4
–2
Fig. 6. Effect of backfill friction angle on stability of (a) wall A,
(b) wall B: horizontal joints
analyses are shown in Fig. 8. Fig. 8(a) shows that, in the (b)
case of wall A, reducing the block interface friction angle
from 458 to 268 had no significant effect on the deflection of Fig. 7. Comparison of internal failure deformation of wall B
the wall at a given backfill height. When the block joint with different inclinations of joints within local areas around
friction angle was reduced (somewhat unrealistically) below toe (see Figs 1(b) and 4(b): x-displacement contours (cm) and
rotation of blocks. (a) inclined masonry joints; (b) horizontal
268, however, wall A failed by overturning and the maxi-
masonry joints
mum stable backfill height was reduced to 5·8 m (19 ft). The
maximum stable backfill height then reduced by approxi-
mately 0·305 m (1 ft) for each 28 reduction in block joint
friction angle below 268. than its counterpart with inclined joints, in terms of the
For wall B with inclined joints (Fig. 8(b)), the apparent maximum stable backfill height. This is probably a result of
sensitivity to a reduction in block friction angle from 458 to the different patterns of soil strain (and hence lateral stress
448 suggests that at a backfill height of 6·1 m (20 ft) this distributions) implied by Fig. 7, which are in turn due to the
wall is only marginally stable. As the block joint friction increased opportunity for internal sliding deformation in the
angle was reduced towards 258, the wall displacement at a walls with horizontal joints. (Small changes in the backfill
backfill height of 5·8 m (19 ft) increased very slightly, but height due to settlements associated with different patterns
overturning remained the dominant mode of deformation of wall movement may also have had a significant effect on
and failure. As the block joint friction angle was reduced walls of marginal stability.) In extreme cases of low block
below 258, the mode of deformation and failure changed to joint friction (,248), the failure mode for wall A was
sliding. As with wall A with inclined joints, each 28 reduc- changed from overturning to sliding by making the joints
tion in block joint friction angle below 258 reduced the last horizontal.
stable height by approximately 0·305 m (1 ft).
With horizontal joints, both walls were stable at a backfill
height of 6·1 m (20 ft) for joint friction angles between 248 WALL JOINT STIFFNESS
and 458 (Fig. 9). Sliding occurred when the joint friction Bulging is often encountered in ageing dry-stone retaining
angle fell below 248 in both cases. For wall A, the maximum walls. It is an essentially flexible mode of deformation, in
deflection at the top with a backfill height of 6·1 m (20 ft) which the deformability of the wall is much greater than that
was about 1·7 cm for all block joint friction angles between of the individual masonry blocks. This suggests that the
248 and 458. For wall B, however, the maximum deflection interfaces between the blocks may play an important role in
at a backfill height of 20 ft increased from 3·7 cm to 6·7 cm bulging. A low friction angle between the blocks is likely to
as the block joint friction angle was decreased from 458 to lead to sliding rather than bulging failure (see the earlier
248. With a joint friction angle of less than 248, wall A slid discussion and Cooper, 1986), indicating that bulging is
along the base, whereas in the case of wall B sliding likely to be associated with a reduction in block interface
occurred along the base and also at the lower horizontal stiffness, rather than strength.
joints. Masonry blocks usually have roughly trimmed surfaces,
In general, as the joint friction angle is reduced, there is a resulting in a contact area between adjoining blocks of less
tendency for a wall with horizontal joints to perform better than 10% of the total, even under high normal stresses
Fig. 9. Effect of friction angle of masonry joints on stability of COMPRESSIBILITY AND STRENGTH OF THE
(a) wall A, (b) wall B: horizontal joints SUB-BASE
The potential importance of a compressible or weak sub-
base in the development of a bulging failure in a drystone
(Goodman, 1976). Experimental measurements (Goodman, retaining wall is highlighted by Cooper (1986). To investi-
1976; Bandis et al., 1983) show that rock surface stiffness is gate the effects of sub-base compressibility and strength,
related to surface topography in addition to the materials analyses were carried out for walls A and B using the base-
properties of the rock blocks. Generally, a rough surface has line parameters, but with the different sub-base stiffnesses
0 5 cm
(a)
5·79 m 6·1 m
0
suffer a bearing capacity failure (Table 3). This confirms the
need to distinguish between wall collapse arising from tilting
–1 due to the compressibility or localised failure of the part of
–2
the sub-base under the toe of the wall, and wall collapse due
to a general bearing capacity failure, as recognised by
–3 Cooper (1986).
2 m above base Wall collapse resulting from bearing capacity failure or
–4
excessive sub-base compressibility was always by toppling
–5 or sliding. It was not possible to promote bulging failure, or
4 m above base even significant bulging deformation, simply by reducing the
–6 strength and stiffness of the sub-base. This suggests that
At the top
–7 bearing capacity failure or compressibility of the sub-base is
0 2 4 6 8 10 12 14 unlikely to be the sole cause of wall bulging. However, it is
Notional time possible that the effects of foundation compressibility or
(c)
bearing failure combined with a reduction in the block
interface stiffness within the wall could promote bulging
Fig. 10. Displacements of wall A with soft joints at (a) stable
backfill height of 5·79 m; (b) unstable backfill height of 6·1 m;
failure: this is investigated below.
(c) deflection history With a softer and weaker sub-base (i.e. sand), large move-
ments of the foundation started to occur at a lower backfill
height for wall A than for wall B. This is not surprising,
given the smaller base width of wall A. However, wall A
and strengths shown in Table 2. The sub-base in the vicinity stabilised following a degree of deformation, resulting in a
of the wall and the lower blocks of the wall are meshed to greater stable backfill height for wall A than for wall B, but
0·1 m (as detailed in Harkness et al., 2000) to provide with a larger deflection (Table 3). This may be consistent
adequate resolution for yielding and deformation. The var- with a greater restoring moment due to friction on the back
ious sub-base parameters are representative of materials of the wall in the case of wall A than wall B, as discussed
ranging from hard rock to medium sand, any of which might in a later section. Bearing failure could in reality be
be used as a foundation in practice (Bowles, 1977; prevented or postponed by the provision of a wider founda-
Goodman, 1980).§ tion block where the sub-base material is of reduced strength
or stiffness.
§
It is recognised that the thickness of the sub-base (between the
wall and the rigid bottom boundary) is limited. The compressibility
of the sub-base will depend on its thickness as well as on its EFFECT OF A REDUCED WALL JOINT STIFFNESS AND
stiffness. The use of a thicker sub-base layer in the mesh would A COMPRESSIBLE SUB-BASE
therefore probably have given quantitatively, if not qualitatively, A series of analyses was carried out using the geometry
different results. of wall A to investigate the combined effects of a reduction
Table 3. Deformation and stability of walls A and B with different sub-base properties
Wall Stable height: m; ft Deflection: cm Wall rotation Failure mode
A B A B A B A B
Hard rock 6·1; 20 6·1; 20 1·12 1·94 F F S S
Medium rock 6·1; 20 6·1; 20 1·25 2·0 F F S S
Very soft rock 5·8; 19 5·8; 19 2·16 2·63 F F Fw Fw
Dense sand 5·49; 18 5·19; 17 6·83 3·69 F F Fb Fb
Medium sand 4·88; 16 4·58; 15 18·1 3·35 B F Fb Fb
F, forward rotation of wall; B, backward rotation of wall; S, stable; Fw , failure in wall; Fb ,
failure in base.
in block joint stiffness and construction on a softer/weaker (in terms of displacements) at factors of safety considerably
sub-base. The combinations of joint stiffness and sub-base below those required by a modern design code. This is
stiffness and strength investigated, and the results of the consistent with the statement made by Cooper (1986) that ‘it
analyses, are summarised in Table 4. is unusual to find walls with large pre-failure deformations
These results indicate that the interaction of joint weath- due to toppling’. It should also come as no surprise in view
ering and sub-base compressibility/strength is quite complex. of Terzaghi’s observation, based on tests on large retaining
They suggest that on a firm sub-base (very soft rock), a walls, that a deflection of 1% of the wall height is normally
reduction in joint stiffness may not always have a detrimen- enough to bring a dry sand backfill to the active state
tal effect on wall stability. While the initial reduction in (Terzaghi, 1934; 1936).
joint stiffness led to a reduction in the maximum stable Second, the results of the numerical analyses (in terms of
height from 6·1 m (20 ft) to 5·8 m (19 ft), further reductions both displacements and the maximum stable backfill height)
in joint stiffness actually resulted in a restoration of the were very sensitive to the frictional strength (ö9) of the
maximum stable height to 6·1 m (20 ft). This confirms that backfill. Underestimating the backfill friction angle by as
bulging can act to help stabilise a wall, for the reasons little as 18 could lead to the underestimation of the factor of
discussed below. safety on soil strength by 4%: the uncertainty involved in
In contrast, on the dense sand sub-base (which consis- selecting parameters for the back-analysis of an existing wall
tently exhibited bearing capacity failure), progressive soft- is likely to be rather greater than this.
ening of the joints led to a monotonic decrease in the Third, many drystone retaining walls, particularly in hilly
maximum stable height. As noted previously, the wall with and mountainous areas, act primarily as facings, with a
hard joints failed by toppling. On softening the joints, the relatively narrow backfilled zone occupying the space be-
mode of wall failure changed from toppling to moving out tween the wall and a steep, stable cut face (O’Reilly et al.,
at the base (as a result of bulging) combined with bearing 1999). In these circumstances, the reduced backfill width
failure of the sub-base. will reduce displacements and may increase the maximum
stable height, probably as a result of friction at the vertical
rock face helping to support the backfill. One potential
DISCUSSION AND IMPLICATIONS FOR PRACTICE problem with a narrow backfill between a wall and a low-
The results reported in this paper go some way towards permeability natural rock face is that the relatively small
explaining the perception that the application of a conven- voidage could allow pore water pressures to build up very
tional design code to a drystone retaining wall leads to an rapidly in the event of flooding from the retained surface,
over-conservative assessment of its stability. First, although especially if the backfill is not well drained (Wong & Ho,
the failure by toppling of a wall with sound foundations and 1997).
unweathered blocks may be well predicted by a conventional In contrast to the brittle nature of toppling failure, quite
limit equilibrium analysis (Harkness et al., 2000), toppling large bulging deformations can be sustained with the wall
deformations remain small (less than 1% of wall height) apparently remaining stable (Cooper, 1986). In the numerical
until the factor of safety on soil strength falls below about analyses described in this paper, bulging was always asso-
1·05. In other words, these walls can perform satisfactorily ciated with a reduction in block joint stiffness. A weak or
of the wall than similar walls with inclined joints (i.e. (g) A reduction in sub-base strength and stiffness may
normal to the front face of the wall), though ultimate cause failure of the wall at a lower backfill height than
failure for both walls occurred at the same soil friction a simple limit equilibrium analysis would suggest. With
angles. Walls with horizontal joints had the possible a weaker and more compressible sub-base, but reason-
advantage that movement at the base of the wall prior ably stiff wall joints, failure would still be by
to failure could give a warning of incipient instability toppling—precipitated either by bearing capacity failure
whereas the very small movement of the top of the wall or by the compression of the sub-base beneath the toe.
for both walls could be regarded as insignificant. Wall bulging is unlikely to occur solely as a result of
However, at inter-block friction angles likely to be the deformation of the sub-base, but the reduction in
encountered in practice (~ 458) there was little if any joint stiffness needed to cause bulging failure decreases
significant impact. with the stiffness and strength of the sub-base.
(d ) A real drystone wall is likely to have a non-rigid sub-
base and weathered blocks. Both of these will tend to
reduce the backfill height at failure to below that
ACKNOWLEDGEMENTS
indicated by a limit equilibrium analysis—by between
The work described in this paper was carried out in the
20% and 25% for the two walls investigated in this
Department of Civil and Environmental Engineering at the
study on a sub-base representative of a medium sand.
University of Southampton, under contract to the Highways
The only circumstance in which a simple limit
Agency. The views expressed are those of the authors, not
equilibrium analysis is likely to underestimate the
the Highways Agency. The authors are grateful to Drs K. C.
backfill height at failure (by 5% in this study) is when
Brady, M. R. Cooper and M. P. O’Reilly for helpful discus-
the wall acts essentially as a facing, with only a limited
sions and advice.
backfill zone between it and a stable rock face.
(e) Weathered blocks were simulated by increasing the
degree of block corner rounding and reducing either
the strength or the stiffness of the joints. Increasing the NOTATION
degree of block corner rounding may have an adverse R ratio of backfill width to overall wall height
impact on a wall already of marginal stability: in the x horizontal distance
case of wall A, an increase in the degree of corner y, z depth
rounding reduced both the maximum stable displace- ö9 effective backfill friction angle
ment and the backfill height at failure. However, it did
not alter the mechanism of wall failure, which
remained by toppling. REFERENCES
( f ) A large reduction in joint strength was needed to have Bandis, S. C., Lumsden, A. C. & Barton, N. R. (1983). Funda-
a significant impact on wall stability: the mechanism of mentals of rock joint deformation. Int. J. Rock Mech. Min. Sci.
wall failure then changed from toppling to sliding. Wall Geomech. Abstr. 20, 249–268.
bulging could be achieved only by reducing the joint Bowles, J. E. (1977). Foundation analysis and design, 2nd edn.
New York: McGraw-Hill.
stiffness. A substantial reduction in joint stiffness was Burgoyne, J. F. (1853). Revetments or retaining walls. Corp of
needed to cause bulging failure, which could be made Royal Engineers Papers 3, 154–159.
to occur at a lower backfill height than that at which a Cook, N. G. W. (1992). Natural joints in rock: mechanical, hydrau-
limit equilibrium analysis would suggest failure by lic and seismic behaviour and properties under normal stress.
toppling. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 29, 198–223.