You are on page 1of 12

Stress characteristics for shallow footings in cohesionless slopes

J. GRAHAM
Department of Civil Engineering, University of Manitoba, Winnipeg, Man., Catlada R3T 2N2
M. ANDREWS
Geotech Associates, 29 Auzonville Road, Turzaputza, Trinidad, West Indies
AND

D. H. SHIELDS
Department of Civil Engineering, University of Manitoba, Winnipeg, Man., Canada R3T 2N2
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

Received May 4 , 1987


Accepted November 2, 1987

Footings placed at shallow depth in bridge approach embankments can be used to reduce the cost and to improve the
performance of bridge supports. The embankments frequently terminate in slopes dropping to underpass level, and the footings
therefore have lower capacity than that for footings on level ground. The needed design procedures have not been well
validated.
The paper describes a new solution using the method of stress characteristics for footing capacity in cohesionless slopes. The
soil is assumed to have c = 0, (b = constant, y > 0. Particular attention is paid to modelling the asymmetric nonfailing zone
immediately beneath the footing. Solutions for various slope angles and friction angles have been obtained for footings placed
at the crest of the slope with D/B = H/B = 0, and for H/B = 0.5, 1.0, and 2.0; D/B = 0.5 and 1.0. The theoretical bearing
capacities are compared with experimental values from two series of laboratory tests.
Key words: bearing capacity, footings, failure, bridge abutments, slopes, sand, cohesionless, stress characteristics.

Des semelles mises en place ? faible


i profondeur dans les remblais d'approche de ponts peuvent &treutilisCes pour rCduire le
coQtet amCliorer la performance des appuis de ponts. Les remblais se terminent frkquemment en pente vers le niveau inferieur,
For personal use only.

et les semelles ont ainsi une capacitC portante inferieure 2 celle dCveloppte sur une surface au niveau. Les procCdures de calcul
requises n'ont pas Ctt bien CprouvCes.
Cet article dCcrit une nouvelle solution utilisant la mCthode des caractCristiques de contraintes pour le calcul de la capacitC
portante de semelles dans des pentes de sols pulvCrulents. L'on suppose que le sol a c = 0, (b = constante, y > 0. L'on porte
une attention particuliCre au modelage de la zone situCe immkdiatement sous la semelle et non assujettie 2 la rupture. Des
solutions pour diffkrents angles de pente et de frottement ont CtC obtenues pour des semelles placCes en crEte avec D/B =
H/B = 0, et pour H/B = 0,5, 1,O et 2,O; D/B = 0,5 et 1,O. Les capacitCs portantes thCoriques sont comparkes aux valeurs
expCrimentales donnCes par deux series d'essais en laboratoire.
Mots clis : capacitC portante, semelles, rupture, culCes de pont, pentes, sable, pulvCrulent, caractCristiques des contraintes.
[Traduit par la revue]

Can. Geotech. J . 25, 238-249 (1988)

Introduction on large-scale model footings. "Remarkable agreement" with


the Shields et al. (1977) results has recently been reported by
Engineered structures sometimes require shallow footings on Kimura et al. (1985), who carried out centrifuge tests at 30 g
sloping ground. These include structures placed on benches cut on a model footing 40 mm wide (effective width, 1200 mm).
into the slope, retaining walls, transmission towers, and bridge Graham (1974) showed that an alternative form of analysis
piers. Highway overpass bridges frequently connect with using stress characteristics allows more careful modelling of
approach embankments about 10 m high that terminate in a boundary and field conditions for failures in sand. The method
sloping face dropping down to the underpass level. In these has recently been used to model closely spaced interfering
cases, supporting the end span of the bridge in the approach fill footings (Graham et al. 1984) and the effects of footing size on
can produce savings in construction costs and reductions in capacity using a critical state strength model (Graham and
pavement maintenance costs compared with the use of deep Hovan 1986). Stress characteristics have been used only rarely
foundations (Shields et al. 1980). in slopes (for example, Sokolovskii 1965; Reddy and Jagan-
Bauer et al. (1981) noted that the available theoretical solu- nath 1985), and in each case, to overcome important computa-
tions for bearing capacities in slopes gave widely differing tional difficulties, the soil was assumed to retain some
answers and were thought to be unduly conservative. Two of cohesion. The method does not seem to have been used previ-
them (Meyerhof 1957; Giroud and Tran 1971) permitted the ously in cohesionless sand when the footings are placed right at
footing to be placed elsewhere than at the crest of the slope. the sand surface, and this forms the objective of the work
Both these solutions use limit equilibrium analysis of assumed described here.
failure surfaces. However, this method can only accommodate The paper describes a new analysis for bearing capacity in
rather unconvincing modelling of the boundary conditions and cohesionless slopes that takes particular account of the stress
failure domain properties (Graham 1974). In response to these conditions immediately beneath the footing. The results are
perceived weaknesses in the analyses, Lebegue (1973), compared with experimental data published previously by
Shields et al. (1977), and Bauer et al. (1981) reported test data Lebegue (1973) and by Shields et al. (1977).
Printed in Canada 1 Imprim6 au Canada
A ET AL. 239

The method of stress characteristics 812


where qu = 2 J o a,dx/B, and the bearing capacity coefficient
Details of computational procedures have been given else- N,($) depends on the angle of shearing resistance 4 . When the
where (Graham 1968) and will only be briefly reviewed here. footing is not at the ground surface but at depth D , the bearing
The solution combines the differential equations for stress capacity can be written:
transmission in plane strain at a point:

where Nyq = N,,(+,D/B). In this case the numerical proce-


dures are simpler. However, the boundary conditions still need
with the Coulomb - Mohr relationship to be carefully considered if good results are to be obtained
from the analysis. It is this question of careful attention to the
[2] ax,,= a(l F sin 4 cos 2$); T ~ ,= T,, =a sin 4 sin 2$ boundary conditions of the solution that marks the particular
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

for the T, a relationships when the soil is failing.' The param- feature of the present work on bearing capacity in slopes. The
eter a is the mean stress (al + a3)/2, and $ is the inclination of symbol N,, will be used throughout the paper for the bearing
a l with the vertical. The sand is assumed to operate always capacity coefficient. When D/B = 0, N,, = N,.
under drained conditions. All stresses are therefore effective
stresses. The common (')-convention for effective stress will Asymmetry of the nonfailing "trapped wedge"
be discarded for convenience, even though effective stresses Andrews (1986) showed that most existing analyses for foot-
are implied throughout the paper. ings in slopes assume a nonfailing zone of soil immediately
Substitution of [2] into [I] leads to hyperbolic differential beneath the footing, with the zone being symmetric on either
equations with product terms of a, $ that suggest logarithmic side of the centre line. As is common, the zone will be called
transforms an "elastic wedge." In contrast with the symmetry that is seen
beneath footings on level ground, photographs from physical
models on slopes (Peynircioglu 1948; Mizuno et al. 1960;
to separate the variables. This leads to four simultaneous equa- Giroud and Tran 1971; and Kimura et al. 1985) show that the
tions, which are solved numerically in related pairs along spe- failure zones are asymmetric (Fig. 1). The downslope zone
cific stress characteristic directions for (x,z,o,$) at previously NPQR in Fig. 1 is larger than the zone LRST rising to the hori-
For personal use only.

unknown points in the failure domain: zontal ground surface behind the crest. This means that the
elastic wedge LMNR must also be asymmetric. Vesic (1973)
[4] dtldz = b, along- the direction dxldz = tan ($ + ,LL)
92+@2" ,I and Kimura et al. (1985) show that only a single wedge is
formed beneath the footing, not two smaller wedges as in some
[5] dqldz = a , along the direction@= tan ($ - p)
analyses. The detailed geometry of the wedge is not immedi-
where a , b = -t sin ($ $ p)/(2a os ($ 7 p) ) . Note that ately obvious, and a number of alternatives will be studied in
p = (45" - 4/2), so $ $ p are the directions of "slip lines" following paragraphs.
along which shear failure takes place. The shape and stress dis- Determinate rectilinear failure zones NPQ and LST in Fig. 1
tribution of the failure zones are developed as the analysis pro- extend below the slope surface and the horizontal ground sur-
ceeds. There is no need for the successive searching for face respectively. From Mohr circle considerations the inclina-
most-critical surfaces that is typical of limit equilibrium tions of the major principal stress can be obtained as
methods. To provide definite integrals, known boundary con-
$ = 0.5{a - a - arc sin (sin alsin 4 ) )
ditions or previous calculated values of (x,z,a,$) must be
available on each of the slip lines. Nonlinearity of the stress and
functions can be taken into account by iterative procedures for
$ (Sokolovskii 1965) and for a (Graham 1968).
The log-transform in [3] leads to difficulties with the numer- along NQ and LS respectively. The values of a along these sur-
ical procedures when there is a zero-surcharge boundary. faces can also be obtained from Mohr circles. These form the
Graham and Stuart (1971) gave a convenient solution for starting boundary conditions for detailed stress characteristic
calculating the N,-coefficient for footings on the surface of calculations in the transition zones NQR and LSR, which end
cohesionless sand with self-weight (c = 0, 6 = constant > 0, with the stresses along NR and LR respectively. Under the
y > 0). It involves (1) starting from a finite surcharge; (2) influence of self-weight in the failure zones the "radial" slip
extending the solution out into the domain; (3) shrinking the lines extending from N and from L are curved, and the
calculated field (and hence the surcharge) back towards the "spiral" lines QR and SR are not logarithmic spirals. They
origin; and (4) again extending the solution into the domain, have to be determined using numerical forward integration of
under the influence of gravity forces. In this way, the effect of [4] and [5].
the assumed surcharge is successively reduced until finally the As mentioned earlier, the geometry of the wedge LMNR is
computed stresses in the failure zone are proportional to the indeterminate, and has to be chosen using one of the assump-
distance from the comer of the footing. At this stage the solu- tions outlined in the following sections. This involves, for
tion is independent of surcharge and the calculated ultimate example, defining wl, or in Fig. 1 and it can be expected that
bearing capacity q, of a footing of width B on the surface of a they will vary with a and with 4 . Once this choice for wl and or
cohesionless sand can be expressed by +
has been made, then XI tan wl = Xr tan o r ; XI Xr = B; and
+
e = (X, - Xl)/2 (X, XI). The distribution of vertical stress
FGH across the base LMN of the footing is obtained from the
vertical stresses calculated along LR and NR. Note that the
'Notation is given in the List of symbols at the end of the paper. vertical stress G at M should be the same whether it is calcu-
240 CAN. GEOTECH. I. VOL. 25, 1988
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

FIG. 1 . Schematic of failure zones for footing at crest of slope.


For personal use only.

FIG.2. Geometry of the asymmetric nonfailing trapped wedge LNR.

lated as avlby way of TSR coming from beneath the horizontal at the edge of the footing. In many ways this is the reason
ground surface TL, or as a,, by way of PQR from the slope for differences between solutions for "smooth" footings
surface PN. This will form an important condition for the solu- using o, = o, = 45" + $12 and "rough" footings using o, =
tions shown later. W, = 6.
The following section will examine four possible altema-
tives for determining the geometry of the trapped wedge for
Modelling of the wedge LMNR footings on slopes. In each case the angle wl corresponding to
Graham and Stuart (1971) showed that the capacity of a sur- the part of the failure zone extending out to the level ground
face footing on horizontal ground depends on the assumptions behind the crest is assumed equal to $ (Graham and Stuart
made for the geometry of the trapped wedge beneath the foot- 1971; Graham and Hovan 1986). The reasonableness of this
ing. This is because the stress distribution on the footing base assumption will be discussed later.
depends strongly on the angle at the pole of the transition zone Model 1 assumes that even though the boundaries SR and
GRAHAM ET AL.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

FIG. 3. Continuity conditions in the trapped wedge LNR.


For personal use only.

QR of the radial transition zones in Fig. 2 are not logarithmic various assumed values of w,. As w, decreases in Fig. 4a, the
spirals, the ratio X1/X,can be estimated as if they are. That is, angle at the pole N of the radial transition zone increases
it assumes that the deviations from logarithmic spiral behav- and the corresponding values of a,, increase. These can
iour caused by gravity forces in the sand have similar effects in be expressed as the variation of N,, = a,,lyX, versus the
both transition zones. If, for computational purposes, the +,
assumed values of w, (Fig. 4b). N,, varies with with a , and
length SL is fixed in Fig. 2, then so is LR and avl from the with a,.
Terzaghi (1943) solution for rough surface footings on level +;
Once again, the conditions are invoked that wl = Zl (= XI
ground. Using a logarithmic spiral asumption for RQ, it is then tan +) = Z, (= X, tan w,); and avl = a,, (Fig. 3). This allows
possible to calculate or(and X,) using the conditions MR = ro, +
calculation of a unique value of w, for each a and (Fig. 5a).
sin w, = rol sin wl;avl = a,,. This permits an estimate of X,/X,, The variation of w, with each of the independent parameters is
and w,, remembering that wl = +. Once this is done, the formal quite significant in Fig. 5a. However, if w, is converted into
stress characteristic solution is carried out in the usual way the eccentricity e, or especially the ratio Xl/X,, as in Fig. 5b,
from LS towards LR and from NQ towards NR. Integrating the +
then the variation with friction angle is greatly diminished.
vertical components of the stresses along LR and NR provides In this way the asymmetry of the trapped wedge can be
the required prediction of the failure load on LMN, and there- uniquely related to the slope angle a for all +-angles. Figure 5b
fore of N,,. also shows a similar result obtained from model 3 calculations.
Model 2 uses direct calculations for the logarithmic spiral Figure 6 presents typical trapped zones obtained from the
surfaces ql = rolexp (01tan 4); rir= ro,exp (8, tan +) in Fig. 2; +
four different models. In Fig. 6a, = 45" and a = 30°, while
and aj = 5 exp (2Aq tan +) for the rate at which the mean in Fig. 6b, += 30°, a = 20". The eccentricity suggested by
stress builds up in the radial transition zones. Once again, all four models is significant. It must therefore be included in
+
w1 = 4; XI tan = X, tan w,; and avl = a,,. an analysis of bearing capacity on slopes if the results are to be
Model 3 is closely similar to models 1 and 2 except that the usable.
magnitude of avi at R is assumed to be established by the
N,-values proposed by Graham and Stuart (1971) using stress
characteristics. Validity of the models for the trapped wedge
Model 4 is considered to be more rigorous than models 1-3 All four models assume that the footing fails by moving ver-
because it avoids the previous assumption that the ratio Xl/X, tically downwards without rotation and develops failure zones
for soil with self-weight will be the same as that when the soil extending outwards to the slope surface NP and the horizontal
is weightless. In this case the size of the failure zone on the left ground surface LT in Fig. 1 . (It is understood that if in reality
below the horizontal ground surface in Fig. 3 is again estab- the sand is compressible, the failure zones will not extend to
lished using the stress characteristic solution for N, presented the soil surface, and "punching" failure will occur (Vesic
by Graham and Stuart (1971). The difference from model 3 is 1973).) The models assume that the displacements preceding
in the treatment of the failure zone extending to the slope sur- failure are small enough to be considered negligible. That is,
face. The procedure is shown in Fig. 4a. It uses stress charac- the sand behaviour is rigid-plastic. This will also be discussed
teristics to calculate a series of a,, values (1, 2 , 3 , etc.) for in a later section.
CAN. GEOTECH. J . VOL. 25, 1988

TABLE1. Comparison of models 1-4 and photographic evidence for trapped wedge asymmetry

Observed Model 1 Model 2 Model 3 Model 4


A u t h o r * ~ c u o , w , e w , w , e w , o , e w , w , e w , w , e

*(I) Peynircioglu (1948); (2) Giroud and Tran (1971); (3) Kimura et al. (1985).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

s c a n e d out for 10" d a d 45". When a < loo, w ,


' ~ n a l ~ s ewere w, and e = 0.0.

3 i.. STRESS 4 . Nevertheless, it can be concluded from Table 1 that o, is


: ....,,, DISTRIBUTIONS much closer to 4 than it is to 45" +
412 (except for Kimura
c~,,
et al.). Hence the choice of ol = 4 for the present work is rea-
---- sonable. Also, o, is less than wl (except, once again for Kimura
...... .... et al.). Therefore, the photographic evidence supports the
M N models proposed earlier in which the trapped wedges are
asymmetric with shapes generally like those shown in Fig. 6.
It will be remembered from the previous section that models
1- 3 were based on the geometry of the failure zones assuming
the sand to be weightless. Model 4 took full account of self-
weight in the failure zones and will therefore be considered the
preferred model. This choice is made even though the agree-
ment between the observed asymmetry in Table 1 and the
For personal use only.

model 4 values is less good than with some of the other


models.
Calculations have been performed using model 4 for 4 =
30, 35, 40, and 45", and for 10" Ia! 5 ( 4 - 5)" in incre-
ments of 5". Figure 7a shows values obtained for footings with
their front edge right at the crest of the slope (D/B = 0;
H/B = 0). As expected, the bearing capacity decreases
towards zero as a approaches 4 . Figure 7a also shows values
obtained by Graham and Stuart (1971) for level ground with
a! = 0. The latter results clearly form a limit for the behaviour
when a! > 0 and are compatible with the present calculations.
Figure 7b compares the present results with solutions by pre-
vious authors. The values shown in the figure are for 4 = 40".
The authors' results lie towards the upper limit of previously
published theoretical solutions. It was pointed out earlier that
the stimulus for this study of footings in slopes came from an
understanding that existing theories were mostly conservative
(Bauer et al. 1981). Thus it is in fact helpful that the new
analysis presented in Fig. 7a gives values higher than most
existing solutions.
Modelling for embedment (D/B > 0) and setback (H/B > 0)
FIG.4. Ny,, wr relationships for the model 4 analysis.
While footings at the crest of slopes provide a relatively
amenable problem for analysis, in practice most footings are
Table 1 compares calculated values of the trapped wedge placed at some depth D into the slope (Fig. 8a) and at some
parameters o,, w,, and e with the limited number of values horizontal distance H back from the crest (Fig. 8b). For D/B <
available from published photographic evidence. The evidence 1.0 the embedment can be modelled as a simple normal
is limited, but the trend towards increasing asymmetry of the surcharge along TL, and a surcharge with normal and shear
trapped wedge (denoted by e) with increasing slope angle is components along NP. Values for shallow embedment are
clear. This conclusion supports the theoretical findings actually generated during the numerical scaling procedures
reported in Table 2. It should be remembered that measured described earlier that lead to the results for zero embedment in
displacement fields do not, in general, correspond well with Fig. 7a, so they are easily obtained. During this step the asym-
the shape of calculated stress fields (Serrano 1972). Photo- metry of the wedge LNR is assumed to have the model 4
graphed failure zones appear to be significantly affected by values shown in Fig. 5. That is, it is not affected by shallow
friction between the sand and the glass container, and by dif- embedment. Calculated values of N,, for H/B = 0; with
ferences between the angle of dilation v and angle of friction D/B = 0.0, 0.5, and 1.O, are shown in Fig. 9.
GRAHAM ET AL

I I I I
(a)
+=450 a
40' A
- 35' 0 -

-
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

MODEL 4 RESULT

I I I I

SLOPE ANGLE a! (DEG) SLOPE ANGLE a! (DEG)


FIG.5. Asymmetry of the trapped wedge using model 4, for various +-values: (a) w, versus a;(b) X,/X, versus a.

TABLE2. Averaged variation of X,/X, 3. Critical setback distance H, below which failure zones exit
TABLE
and e/B with slope angle cr for various from slope face (after Graham et al. 1984)
+-angles (model 4 results)
+
Angle of shearing resistance (deg) 30 35 40 45
Slope angle a Eccentricity Critical setback distance Hc/B 1.2 2.0 2.7 3.5
For personal use only.

(deg) X,/Xr e (xB)

10 for H/B = 0.0, 0.5, 1.0, and 2.0 and D/B = 0.0, 0.5,
and 1.0.
As would be expected from the earlier discussion, the value
+
of N,, for a given reduces with a , increases with D/B, and
increases with H/B. Figures 9 and 10 also show the "level
ground" ( a = 0) values for N,, given by Graham and Stuart
(1971) for D/B = 0.0, 0.5. The trend of the new results is
clearly correct. As H/B increases or a decreases the results
The modelling for the horizontal setback distance H requires form the new calculations on slopes converge towards the level
more careful consideration. Two limiting conditions can be ground values in the earlier study.
readily identified as shown in Fig. 86. If the footing is at the
crest with H = 0, then the geometry of the trapped wedge is
given by Fig. 5 and the failure zone emerges from the slope Comparison with large-scale test results
along the length NP. Second, when the footing is placed at Only a limited number of carefully controlled test experi-
L"N" the failure is just contained beneath the level ground ments of footings on slopes have been reported in the litera-
surface N"P" and the presence of the slope has no influence ture. Lebegue (1973) tested a 200 mm wide by 1000 mm long
on the computed bearing capacity. That is, for H 2 N"P" = footing. The tests were camed out in a 3000 mm X 3000 mm
Hc the footing behaves like the footings on level ground (in plan) box. The footing was located at the crest of the
described by Graham and Stuart (1971). Values of Hc are 3000 mm long slope, hence H/B = 0. Readings were taken
given in Table 3. The distance N"P" is the distance at which from only the centre segment of the three-section footing, to
shallow (surface) parallel footings begin to "interfere" and simulate plane strain conditions. He used two different sand
gain additional capacity (Graham et al. 1984). In this case, the densities and three slope angles. After some smoothing of the
wedge L"NUR" is symmetric, with the angles at L" and N" data to take account of variability in preparing the sand, the
equal to 4. experimental results can be presented as shown in Fig. 11. The
To the authors' knowledge there is at present no rational way theoretical results in the figure are for D/B = 0 and H/B = 0,
of predicting how the asymmetry of L'N'R' or the position of and are identical with those in Fig. 7a. The trend of the theoret-
P' varies with a , +, and H/B. In the present work it has been ical results is in broad agreement with the experimental values,
assumed that as N' moves from N to N" (0 I H I Hc), then but they underestimate the decrease in N,, with increasing a.
(1) P' varies linearly between P and P"; and (2) X,/X, varies On detailed examination, the experimental values are higher
linearly between the values for model 4 in Fig. 5b for crest than analytical values for the same +-value. It should be
loading and the value 1.0 for the symmetric, level ground case. remembered that the model studies fail the sand under plane
The resulting positions of P' and R' in Fig. 8b define the extent strain conditions while the +-values quoted by Lebegue (1973)
of the failure zone P'Q'R'. Conditions along N'P' are treated are from triaxial tests. (The pressure range in the triaxial tests
as a triangular surcharge N'NP' having normal and shear com- is not reported. Values of 6 can be expected to decrease with
ponents. Results of these computations are given in Figs. 9 and increasing pressure.)
CAN. GEOTECH. J. VOL. 25, 1988
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

MODEL
I ... ..... ... ; ' -\ .
.'..>:::.
~ ~ 3 0 0
For personal use only.

FIG.6. Typical predictions for the asymmetry of the trapped wedge from models 1-4: (a) 4 = 45", a = 30"; (b) 4 = 30°, a = 20".

It is widely understood (see, for example, Wroth and ties that would be associated with a more sophisticated soil
Houlsby 1985) that $ in plane strain is larger than that in tri- model in which $ varied with mean pressure levels.
axial tests (+,,). Quite good agreement is obtained if the exper- More extensive tests have been reported by Shields et al.
imental results in Fig. 11 are compared with the analytical (1977, 1980). The tests were performed in a large laboratory
results using $ = $,, x 1.1. Similar suggestions were made by sandbox 15 m long, 2 m wide, and 2.2 m high using 500 kN of
Graham and Stuart (1971) for checking surface footings on crushed silica sand. The slope was formed by depositing air-
level ground with a = 0. dry sand from a rotating drum spreader. Two sand densities
Lebegue also tested some smaller footings with B = 60 mm were achieved, 14.85 and 15.75 kN/m3, respectively 70 and
and length of 200 mm in plane strain. Despite N,, being 90% density index. Careful attention to the pressure range
a dimensionless parameter, it does depend on the size of the during testing suggested $,,-values of 37 and 41 " respectively
footing, decreasing with increasing breadth B for the same for the two different sand densities.
density of sand. The shear strength of the sand is only approxi- A 300 mm broad footing stretched across the full 2 m width
mated by the common assumption of a straight strength enve- of the sandbox to create two-dimensional loading. To elimi-
+
lope, that is, that is independent of stress level. At failure, nate end effects at the sides of the box, the footing was made in
larger footings transfer larger stresses to the soil, mobilize three equal segments that were pressed into the sand at the
lower average $-values in the failure zones, and have smaller same rate, with load and displacement readings being taken
N,,-values. This question of "scale effect" in sands has been only from the central segment. Only 2:l slopes (a! = 27")
addressed recently by Graham and Hovan (1986), who incor- were examined, but the positions of the footings were altered
porated a critical state strength model into their analysis of sur- in the range 0 r D/B r 3; 0 r H/B r 5.
face footing capacity. However, in this paper it has been Comparison between the theoretical and experimental results
thought preferable to focus attention on a new understanding of is shown in Fig. 12 using the form of presentation originally
the asymmetric wedge LMNR (for example, in Figs. 2, 3, and proposed by Shields et al. (1977). Tests and analyses were per-
6) and to leave for a future occasion the additional uncertain- formed for discrete ranges of D/B and H/B, and then lines of
GRAHAM ET AL.

SLOPE 6 : l 4:l 2:l 1.5: 1 SLOPE 6 : 1 4 : l


-
1000 I I I I I I 1000 I I I I I I
- (a)
-
- (b) -
D/B=O.O
- - A CHEN AND DAVIDSOld (1973) -
H/B ~ 0 . 0 - 0 MIZUNO E T A L . (1960)
400 - * MEYERHOF (1957) -
GIROUD AND T R A N (1971)
B O W L E S (1975)
- A KOVALEV (1964)
-
r HANSEN (1970)

: %!
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

A L L VALUES FOR $=40°, D / B = H / B = O

I I I

SLOPE ANGLE Cr. (DEG) SLOPE ANGLE a! (DEG)


FIG.7. Bearing capacity coefficient N,, versus slope angle a: (a) predictions for D/B = H/B = 0 for various +-values; (b) comparison with
+
previously published values, = 40".
For personal use only.

FIG. 8. Schematic of failure zones for embedment and setback: (a) D/B > 0; (b) H/B > 0.

constant Nyq interpolated between the known values. The the values of Hc predicted by Graham and Stuart (1971) with
values in the figure are for the front edge of the footing in the experimental findings. The contours of N,, (experimental)
the position shown. Once again, the trend of the results is become parallel with the ground surface (signifying Hc has
encouraging, with the theoretical results perhaps not taking been reached) at roughly the predicted distances.
quite enough account of the influence of the reduction in Table 4 examines whether the experimental and analytical
capacity caused by the slope. It is also interesting to compare results should be compared on the basis of +,,
or 1.1 X +,,.
246 CAN. GEOTECH. J. VOL. 25, 1988

H/B=O.O ----- H/B=0.5 * a = O (GRAHAM AND STUART 1971)


1000 I I
I (b)
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

10 I I I
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
a (DEG) a (DEG) a (DEG)
FIG.9. Predicted N,,-values: (a) D/B= 0; ( b ) D/B = 0.5; (c)DIB = 1.0. (HIB = 0.0, 0.5.)

H/B=l.O ----- H / B = 2 . 0 * a = O (GRAHAM AND STUART 1971)


1000
For personal use only.

FIG. 10. Predicted N,,-values: (a) D/B = 0; (b) D/B = 0.5; (c) D/B = 1.0. (H/B = 1.0, 2.0.)

The theoretical values have been interpolated from the results the theoretical values of N,, should be increased to account
given earlier in Figs. 9 and 10. In the compact sand the for this added depth of embedment.
average ratio between theoretical and experimental values for
N,, is 1.33 using +,,
and 2.82 using 1.1 x +,,.
A similar con-
Discussion and conclusions
clusion can be drawn for the dense sand, where the ratios are
1.04 and 1.88 respectively. In contrast with Lebegue's results Prediction of the bearing capacity of footings requires (1)
discussed previously, the agreement between experimental and identification of an appropriate failure mechanism, particularly
theoretical results is much better in this case when +,,
is used. as it relates to the indeterminate zone immediately beneath the
Shields et al. (1977) took particular care to relate the +-values footing; (2) availability of a suitable analytical model; (3) an
to the stress levels in their tests (see also Graham and Stuart appropriate strength model for the soil; and (4) validation of
1971). the results by comparison with large-scale laboratory tests,
It should be remembered that neither Lebegue nor Shields full-scale field tests, or centrifuge tests.
et al. considered the increase in embedment 6D that the test The paper has examined how the modelling of boundary
footings underwent to reach failure. In the case of the tests of conditions affects the calculation of bearing capacity on
Shields et ul., the footing generally settled 20 or 30 mm, which slopes. In particular it has identified that the trapped elastic
represents 0.07B-0.10B. It can be argued (Graham and Stuart zone is asymmetric to the footing base. A solution has been
197I), particularly in the case'of surface footings (D = 0), that developed on the assumptions (Fig. 1) that (1) the edge LR of
GRAHAM ET AL.

TABLE4. Comparison of experimental and theoretical values of N,,

Values of N,,
Geometry Experimental* TheoreticalT Theorylexperiment
H/B DIB Col. (1) Col. (2) Col. (3) (2) + (1) (3) + (1)

(a) Compact sand


29 64
58 140
84 176
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

48 100
74 148
100 196
(b) Dense sand
64 140
140 200
176 320
100 200
148 300
196 380
148 260
208 350
260 450
*~rom Shields et a[. (1973).
;~uthors,from Andrews (1986). Compact sand: col. (2) I#J = +,,
= 37", col. (3) +=
1.14[, = 41"; dense sand: col. (2) 4 = &,, = 41°, col. (3) 4 = 1.1& = 45".
For personal use only.

"punching" shear failure in the sand (Vesic 1973). At present


LEBEGUE (1973) there is inadequate experimental evidence to justify these
=40°
c # I ~ ~ e assumptions fully. They do, however, form a rational and
coherent framework for allowing the calculations to proceed.
The analysis used the method of stress characteristics for
+
sand with c = 0, = constant, y > 0. The numerical proce-
dures are known to give essentially the same results as the
more usual logarithmic spiral methods provided the boundary
and domain conditions are specified identically (Graham
1974). However, the method does offer the particular advan-
tage in the present work that the boundary conditions can be
modelled more convincingly (see also Graham and Stuart
1971) and this has formed the principal thrust of the work pre-
sented here using only simple constant-+ strength relationships.
Validation of the theoretical results is incomplete. It has
been known for some time that existing theories tend to under-
estimate the capacity of footings on slopes. Thus it is encour-
aging in Fig. 7b that the present results lie above the majority
of previous theoretical solutions. The new analytical results
have been compared with the experimental results presented by
Lebegue (1973) (Fig. 11) and by Shields et al. (1977)
(Fig. 12). It was shown that the trend and magnitude of the
SLOPE ANGLE CY (DEG) theoretical results is largely correct in both cases. In Fig. 11
+
the results were compared on the basis of = 1.1 x +,,,an
FIG. 11. Comparison of predicted values of N,, with experimental empirical correction that takes some account of the plane strain
values from Lebegue (1973) for B = 200 mm. conditions under which the footings were tested. However, in
Fig. 12 and Table 4 the agreement is much better if +,, is used,
although it has been noted that no account is taken of how
+
the trapped zone slopes at w, = to the horizontal, as in a settlements preceding failure affect the capacity of the
normal footing on level ground; (2) the failure zones going out footings. When experimental or field results are reported in the
to the level ground and the slope surface (LT and NP respec- literature for use in calibrating analytical procedures, it is
tively) start from a common point R; and (3) the vertical stress essential to give the grain size distribution of the sand, its
at M is the same whether it is calculated from LRST or from mineralogy, its density at emplacement, the type and stress
NRQP, that is, a,, = q,,. It also assumes that the footing range of the shear testing, and some estimate of the curvature
moves downwards with "general" shear failure and not of the strength envelope.
CAN. GEOTECH. J. VOL. 25, 1988
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

( a 1 COMPACT SAND C$ =37O

( b ) DENSE SAND C$ =41°


For personal use only.

VALUES OF N y q : THEORETICAL AUTHORS


EXPERIMENTAL----- SHIELDS E T A L . (19771

I
FIG. 12. Comparison of predicted values of N,, with experimental values from Shields et al. (1977) for B = 300 mm.

Further testing is needed, This could be valuably done in a tions of Ontario. Ingrid Trestrial and Prabir Mitra provided
centrifuge, which couples the advantages of convenient model technical support.
sizes with elevated stress levels that correspond with field-
scale applications. This work is currently in progress with the ANDREWS, M. 1986. Computation of bearing capacity coefficients for
centrifuge at the Laboratoire central des ponts et chaussCes in shallow footings on cohesionless slopes using stress characteristics.
Nantes, France. M.Sc. thesis, University of Manitoba, Winnipeg, Man.
However well tests are performed, they are only usefi~lif the ATKINSON, J. H., and BRANSBY, P. L. 1978. The mechanics of soils:
sand behaviour is adequately characterized. Thus, future an introduction to critical state soil mechanics. McGraw Hill (UK)
Ltd., Maidenhead, England.
model test programs also need to pay close attention to the
BAUER,G. E., SHIELDS, D. H., SCOTT,J. D., and GRUSPIER, J. E.
properties of the sand. It is suggested that this needs to take 1981. Bearing capacity of footings in granular slopes. Proceedings,
account of the influence of stress level as well as initial 10th International Conference on Soil Mechanics and Foundation
density on the shearing resistance. Newer approaches to sand Engineering, Stockholm, Vol. 2, pp. 33-36.
characterization, such as those proposed by Atkinson and BEEN,K., and JEFFERIES, M. G. 1985. A state parameter for sands.
Bransby (1978), Bolton (1986), and Been and Jefferies (1985), GCotechnique, 35: 99- 112.
need to be adopted by geotechnical modellers whether their BOLTON, M. D. 1986. The strength and dilatancy of sands. GCotech-
work is with physical models or with computer models. The nique, 36: 65-78.
additional strength testing program is not large but needs to be BOWLES, J. E. 1975. Spread footings. In Foundation engineering
carefullv done. handbook. Edited by H. F. Winterkorn and H.-Y. Fong. Van
In summary, new theoretical results for footings near the Nostrand Reinhold, New York, NY.
CHEN,W. F., and DAVIDSON, H. L. 1973. Bearing capacity determi-
crests of slopes can be compared favourably with two sets of nation by limit analysis. ASCE Journal of the Soil Mechanics and
published large-scale model tests. However, the basis for this Foundations Division, 99(SM6): 433-449.
comparison is inconsistent-in one case $, can be used while GIROUD, J. P., and TRAN,V.-N. 1971. Force portante d'une fonda-
in the other it has to be empirically increased to correspond tion sur une pente. Annales de 1'Institut Technique du Bdtiment et
with plane strain conditions. des Travaux Publics, Supplement 283 -284.
GRAHAM,J. 1968. Plane plastic failure in cohesionless soils.
Acknowledgements GCotechnique, 18: 301 -3 i6.
1974. Plasticity solutions to stability problems in sand. Cana-
The work has been supported by research grants and fellow- dian Geotechnical 11: 238-247,
ships from the Natural Sciences and Engineering Research GRAHAM, J., and HOVAN, J. -M. 1986. Stress characteristics for bear-
Council of Canada, the University of Manitoba Transport ing capacity in sand using a critical state model. Canadian Geotech-
Institute, and the Ministry of Transportation and Communica- nical Journal, 23: 195 -202.
ET AL. 249

GRAHAM, J., and STUART, J. G. 1971. Scale and boundary effects in SOKOLOVSKII, V. V. 1965. Statics of granular media. Pergamon
foundation analysis. ASCE Journal of the Soil Mechanics and Press, New York, NY.
Foundations Division, 97(SMll): 1533- 1548. TERZAGHI, K. 1943. Theoretical soil mechanics. Wiley, New York,
GRAHAM, J., RAYMOND, G. P., and SUPPIAH,A. 1984. Bearing NY.
capacity of three closely-spaced footings on sand. GCotechnique, VESIC,A. S. 1973. Analysis of ultimate loads of shallow foundations.
34: 173- 182. ASCE Journal of the Soil Mechanics and Foundations Division,
HANSEN, J. 1970. A revised and extended formula for bearing capac- 99(SM1): 43 -75.
ity. Danish Geotechnical Institute, Copenhagen, Denmark, Bul- WROTH,C. P., and HOULSBY, G. T. 1985. Soil mechanics-property
letin 28. characterization and analysis procedures. State of the art report.
KIMURA,T., KASAKABE, O., and SAITOH,K. 1985. Geotechnical Proceedings, 1lth International Conference on Soil Mechanics and
model tests of bearing capacity problems in a centrifuge. GCotech- Foundation Engineering, San Francisco, CA, Vol. 3, pp. 1-54.
nique, 35: 33 -45.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Université de Sherbrooke on 06/30/14

KOVALEV, I. V. 1964. De la rksistance ultime des fondations IimitCes


par un talus. Traduction du russe, extrait du recueil des travaux de
LIIZLT, fascicule 225, "Nekotovye vo prosy tonnele-i-mosto-
List of symbols
stroeniy," Leningrad, URSS, 1964. French translation by Labora- sin ($ + u 4 cos ($ F ~r.))
~ ) l { 2 tan
toire central des ponts et chaussges, Paris, France, 1973. footing breadth, depth
LEBEGUE, Y. 1973. Essais de fondations superficielles sur talus. Pro- cohesion
ceedings, 8th International Conference on Soil Mechanics and eccentricity of base of trapped nonfailing zone from
Foundation Engineering, Moscow, Vol. 4.3, p. 3 13. footing centre line ( X B )
MEYERHOF, G. G. 1957. The ultimate bearing capacity of founda-
setback distance from crest, critical setback inside
tions on slopes. Proceedings, 4th Intemational Conference on Soil
which the slope affects the footing capacity
Mechanics and Foundation Engineering, London, Vol. 1, pp.
384-386. bearing capacity coefficient for D/B = 0,
MIZUNO,T., TOKUMITSU, Y., A N D KAWAKAMI, H. 1960. On the D/B $ 1.0
bearing capacity of a slope of cohesionless soils. Soils and Founda- average footing contact pressure at failure
tions, 1: 30-37. radius, initial radius, pole angle for logarithmic
PEYNIRCIOGLU, H. 1948. Tests on bearing capacity of shallow foun- spiral
dations horizontal top surfaces of sand fills and the behaviour of horizontal, vertical coordinates
For personal use only.

soils under such foundations. Proceedings, 2nd International Con- base dimensions o n left, right side of trapped zone
ference on Soil Mechanics and Foundation Engineering, Rotter- (Fig. 3)
dam, Vol. 3, pp. 144-205.
depth of trapped zone coming from left, right side
REDDY, A. S., and JAGANNATH, S. V. 1985. Bearing capacity of strip
footings in depressions. Proceedings, Indian Geotechnical Confer- (Fig. 3)
ence, Roorkee, Vol. 1, pp. 77-82. slope angle, deg
SERRANO, A. A. 1972. The method of associated fields of stress and unit weight
velocity, and its application to earth pressure problems. Proceed- angle of shearing resistance (friction), dilation
ings, 5th European Conference on Soil Mechanics and Foundations normal, shear stress
Engineering, Madrid, Vol. 1, p. 77. inclination of major principal stress to vertical
SHIELDS,D. H., SCOTT,J. D., BAUER,G. E., DESCHENES, J. H., 45" - 412
and BARSVARY, A. K. 1977. Bearing capacity of foundations near
slopes. Proceedings, 9th Intemational Conference on Soil
(In a)/2 tan 4 + $
base angles at left, right of trapped wedge (Fig. 3)
Mechanics and Foundations Engineering, Tokyo, Vol. 2, pp.
715-720. Subscripts:
SHIELDS, D. H., DESCHENES, J. H., SCOTT,J. D., and BAUER,G. E. 1, 3, v major, minor, vertical
1980. Advantages of founding bridge abutments on approach fills. 1 r i j left, right, initial, general j t h
Roads and Transport Association of Canada Conference, Sep-
tember 1978. Published RTAC Forum 3, No. 1, Spring 1980. Effective stresses are implied throughout the paper

You might also like