Académique Documents
Professionnel Documents
Culture Documents
Bearing Capacity of Spatially Random Soil The Undrained Clay Prandtl GRIFFITHS 2001
Bearing Capacity of Spatially Random Soil The Undrained Clay Prandtl GRIFFITHS 2001
4, 351±359
By merging elasto-plastic ®nite element analysis with ran- En associant une analyse d'eÂleÂments ®nis eÂlasto-plastiques et
dom ®eld theory, an investigation has been performed into une theÂorie du champ aleÂatoire, nous avons enqueÃte sur la
the bearing capacity of undrained clays with spatially vary- capacite porteuse des argiles non draineÂes ayant une reÂsis-
ing shear strength. The object of the investigation is to tance au cisaillement variant dans l'espace. Cette investiga-
determine the extent to which variance and spatial correla- tion a pour but de deÂterminer les effets sur les statistiques
tion of the soil's undrained shear strength impact on the de capacite porteuse de la variance et de la correÂlation
statistics of the bearing capacity. Throughout this study, spatiale de la reÂsistance au cisaillement non draine du sol.
bearing capacity results are expressed in terms of the bear- Tout au long de cette eÂtude, nous exprimons les valeurs de
ing capacity factor, Nc , in relation to the mean undrained capacite porteuse en termes de facteur de capacite porteuse,
strength. For low coef®cients of variation of shear strength, Nc , par rapport au moyen de reÂsistance non draineÂe. Pour
the expected value of the bearing capacity factor tends to les coef®cients bas de variation de reÂsistance au cisaillement,
the Prandtl solution of Nc 5:14. For higher values of the la valeur attendue du facteur de capacite porteuse tend aÁ la
coef®cient of variation, however, the expected value of the solution de Prandt1 de Nc 5:14. Cependant, pour des
bearing capacity factor falls quite steeply. The spatial corre- valeurs plus eÂleveÂes du coef®cient de variation, la valeur
lation length is also shown to be an important parameter attendue du facteur de capacite porteuse baisse de manieÁre
that cannot be ignored. The results of Monte Carlo simula- assez marqueÂe. Nous montrons eÂgalement que la longueur
tions on this non-linear problem are presented in the form de la correÂlation spatiale est un parameÁtre important qui ne
of histograms, which enable the interpretation to be ex- peut eÃtre neÂgligeÂ. Nous preÂsentons les reÂsultats des simula-
pressed in a probabilistic context. Results obtained in this tions de Monte-Carlo sur ce probleÁme non lineÂaire sous
study help to explain the well-known requirement that bear- forme d'histogrammes, ce qui permet d'exprimer l'interpreÂ-
ing capacity calculations require relatively high factors of tation dans un contexte probabiliste. Les reÂsultats obtenus
safety compared with other branches of geotechnical design. dans cette eÂtude aident aÁ expliquer une neÂcessite bien
connue : les calculs de capacite porteuse demandent des
KEYWORDS: bearing capacity; limit state design/analysis; numer- facteurs de seÂcurite relativement eÂleveÂs par rapport aux
ical modelling; plasticity; shear strength; statistical analysis autres branches de conception geÂophysique.
ì cu
8
ìln cu ln ì cu ÿ 12 ó 2ln cu (6)
B
Q
q = Q/B
Rollers
Rollers
2B
Fixed
5B
δv /B : × 10–3
material property ®elds.
10
PARAMETRIC STUDIES
Analyses were performed using the mesh of Fig. 2 with the
input parameters in the following ranges:
0:125 < È cu , 1
15
(14)
0:125 < COVcu < 4
dividing it by the mean undrained shear strength, ì cu . The Fig. 3. Typical load/deformation curves corresponding to different
reader should bear in mind the Prandtl solution of 5´14 when realisations in the bearing capacity analysis of an undrained clay
viewing this ®gure. It is clear that a majority of the curves with È cu 1 and COVcu 1
¯atten out at bearing capacity values below the Prandtl solution.
This trend will be con®rmed in all the results shown in this
paper. for low values of COVcu , m Nc tends to the deterministic Prandtl
Figure 4 shows a typical deformed mesh at failure with a value of 5´14. For higher values of COVcu , however, the mean
superimposed greyscale corresponding to È cu 1, in which bearing capacity factor falls steeply, especially for lower values
lighter regions indicated stronger soil and darker regions in- of È cu . For example, in a highly variable case where È cu 0:5
dicated weaker soil. In this case the dark zones and the light and COVcu 4, the predicted m Nc value is less than unityÐ
zones are roughly the width of the footing itself, and it appears over ®ve times smaller than the Prandtl value! For the recom-
that the weak (dark) region near the ground surface to the right mended upper limit of COVcu 0:5 suggested by Lee et al.
of the footing has triggered a quite non-symmetric failure (1983) and others, the m Nc value is closer to 4, corresponding
mechanism. The shape of the non-symmetric mechanism is to a more modest reduction of 20%. What this implies from a
emphasised further by the plot of displacement vectors for the design standpoint is that the bearing capacity of a heterogeneous
same realisation, shown in Fig. 5. soil will on average be less than the Prandtl solution that would
For each combination of È cu and COVcu , nsim 1000 rea- be predicted assuming the soil is homogeneous with its strength
lisations of the Monte Carlo process were performed, and the given by the mean value. The in¯uence of È cu is also
estimated mean (m Nc ) and standard deviation (s Nc ) of the pronounced with the greatest reduction from the Prandtl solu-
resulting 1000 bearing capacity factors from equation (4) were tion being observed with values around È cu 0:5. As the value
computed. of È cu is reduced further towards zero, there is evidence of a
Figure 6(a) shows how the estimated mean bearing capacity gradual increase in the value of m Nc , as shown in Fig. 6(b).
factor, m Nc , varies with È cu and COVcu . The plot con®rms that, From a theoretical point of view, it could be speculated that, as
Fig. 4. Typical deformed mesh and greyscale at failure with È cu 1. The darker regions indicate weaker soil
BEARING CAPACITY OF SPATIALLY RANDOM SOIL 355
Fig. 5. Displacement vectors at failure for the same case shown in Fig. 4. The non-symmetric shape of the failure mechanism is clearly visible
5·5
Prandtl, 5·14
Prandtl, 5·14
5
5·0
4·5
4
4·0
3·5
3
c
c
mN
mN
3·0
2·5
2
Θc = 0·5
u
2·0 Θc = 1·0
u
Θc = 2·0 COVc = 0·125
u u
1·5 Θc = 4·0 COVc = 0·25
u 1 u
Θc = 8·0 COVc = 0·5
u u
Θc = ∞ COVc = 1
1·0 u u
COVc = 2
u
COVc = 4
0 u
0·5
10–1
2 3 4 5 6 7 8 9
1
2 3 4 5 6 7 8 9
10 0 0·5 1 1·5 2 2·5 3 3·5 4
Θc
COVc u
u
(a) (b)
Fig. 6. (a) Estimated mean bearing capacity factor, m Nc , as a function of undrained shear strength statistics, È cu and COVNc . (b) More
clearly shows the increase in mN c as È cu ! 0
È cu becomes vanishingly small, the mean bearing capacity next. In this case, the distribution of qf will be statistically
factor will continue to increase towards the deterministic Prandtl similar to the underlying distribution of cu but magni®ed by
solution of 5´14. The explanation lies in the fact that as the 5´14. The mean bearing capacity will therefore be given by
spatial correlation length decreases, the weakest path becomes ì qf 5:14ì cu (15)
increasingly tortuous and its length correspondingly longer. As
a result, the weakest path starts to look for shorter routes hence m Nc 5:14 for all COVcu .
cutting through higher-strength material. In the limit, as Figure 7 shows the in¯uence of È cu and COVcu on the
È cu ! 0, it is expected that the optimum failure path will be estimated coef®cient of variation of the bearing capacity factor,
the same as in a uniform material with strength equal to the COVNc s Nc =m Nc . The plots indicate that COVNc is positively
mean value, hence returning to the deterministic Prandtl solu- correlated with both COVcu and È cu . This ®gure also indicates
tion. that the correlation length, È cu , has a signi®cant in¯uence on
Also included in Fig. 6(a) is a horizontal line corresponding COVNc . For small correlation lengths COVNc is small and rather
to the analytical solution that would be obtained for È cu 1. insensitive to COVcu ; however, for higher correlation lengths
This hypothetical case implies that each realisation of the COVNc increases quite consistently until it reaches the limiting
Monte Carlo process involves an essentially homogeneous soil, maximum value corresponding to È cu 1, de®ned by the
albeit with strength varying only from one realisation to the straight line where COVNc COVcu .
356 GRIFFITHS AND FENTON
5 For the particular case shown in Fig. 8, the ®tted log-normal
distribution has the properties m Nc 3:31 and s Nc 2:08:
Θc = 0·125
u hence from equations (5) and (6) the underlying normal dis-
Θc = 0·25
u
Θc = 0·5
tribution is de®ned by mln Nc 1:03 and sln Nc 0:58. Equation
4
u
Θc = 1·0 (17) therefore gives p(Nc , 5:14) 0:85, indicating an 85%
u
Θc = 2·0 probability that the actual bearing capacity will be less than the
Prandtl value.
u
Θc = 4·0
Figure 9 gives a summary of p(Nc , 5:14) for a range of
u
Θc = 8·0
u
3
Θc = ∞
u
values of È cu and COVcu . The ®gure indicates a wide spread of
probability values with respect to È cu , with the highest prob-
abilities corresponding to the lowest values of È cu . For exam-
c
COVN
0·4 1·0
0·9
0·8
Prandtl, 5·14
0·3
0·7
p (Nc < 5·14)
0·6
f (Nc)
0·2 0·5
0·4 Θc = 0·5
u
Θc = 1·0
0·3 u
Θc = 2·0
u
0·1 Θc = 4·0
0·2 u
Θc = ∞
u
0·1
0 0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9
0 2 4 6 8 10 12 14 10–1 1 10
Nc COVc
u
Fig. 8. Histogram and log-normal ®t for the computed bearing Fig. 9. Graph showing the probability p(Nc , 5:14) that the bearing
capacity factors when È cu 2 and COVcu 1. The log-normal capacity factor will be lower than the Prandtl solution based on the
function has the properties m Nc 3:31 and s Nc 2:08 mean strength
BEARING CAPACITY OF SPATIALLY RANDOM SOIL 357
strength variance would suggest a 50% probability that the 1·0
bearing capacity would lie below the Prandtl value. 0·9
In order to remove this anomaly, the results have been
reinterpreted in Figs 10 and 11 to compare the computed 0·8
bearing capacity factor with the Prandtl solution after it has 0·7
been reduced by a factor F. The factor F is equivalent to a
u
0·6 Θc = ∞
(Fig. 13(b)), and for a soil with COVcu 0:5 (a value at the
u
0·5
upper end of the recommended range), the required factor of
0·4 safety increases further to F 3, as shown in Fig. 13(c). In the
0·3 case of COVcu 1 (Fig. 13(d))Ða value that might be consid-
ered exceptionally high for most soilsÐthe need for even higher
0·2 factors of safety is indicated.
0·1 An important observation from Fig. 13 is that the correlation
length, È cu , becomes increasingly relevant to the probabilistic
0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 interpretation of the bearing capacity problem as COVcu gets
10–1 1 10
larger. This is clear from the way the curves are bunched
together when COVcu 0:125 (Fig. 13(a)) yet are quite diver-
COVc
u
Fig. 10. Graph showing the probability p(Nc , 5:14=2) that the
gent when COVcu 1 (Fig. 13(d)).
bearing capacity factor will be lower than the Prandtl solution In all of Fig. 13, the `crossing-over' of the lines correspond-
based on the mean strength incorporating a factor of safety F 2 ing to different È cu values implies that high values of È cu are
bene®cial to design at low values of F by giving lower
probabilities of design failure, but may be a liability at higher
values of F. The explanation lies in the fact that smaller
1·0 correlation lengths lead to smaller values of COVNc , as shown
0·9
in Fig. 7. Increasing F will therefore result in a steeper fall in
Θc = 0·5 the probability of design failure as the factored bearing capacity
u
0·8 Θc = 1·0
u
factor rapidly passes through the `bunched up' distribution.
Θc = 2·0 In addition to the expected trend, which shows p(Nc
0·7 u
Θc = 4·0 , 5:14=F) decreasing as F increases for all È cu , the curves
p (Nc < 5·14/3 )
u
0·6 Θc = ∞ also con®rm that a factor of safety of 3 is able to reduce the
u
0·5 probability of design failure to negligible levels for all soils in
the recommended range of 0:1 , COVcu , 0:5. These results
0·4
may help explain, in a probabilistic context, why factors of
0·3 safety used in bearing capacity calculations are typically much
0·2 higher than those used in other limit state calculations in
geotechnical engineering, such as slope stability and earth
0·1 pressures.
0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9
10–1 1 10
COVc
u
CONCLUDING REMARKS
The paper has shown that soil strength heterogeneity in the
Fig. 11. Graph showing the probability p(Nc , 5:14=3) that the form of a spatially varying log-normal distribution can signi®-
bearing capacity factor will be lower than the Prandtl solution based cantly reduce the mean bearing capacity of a strip footing on
on the mean strength incorporating a factor of safety F 3 undrained clay.
358 GRIFFITHS AND FENTON
1·0 1·0
0·9 0·9
0·8 0·8
Θc = 0·5
0·7 Θc = 0·5 0·7 u
Θc = 1·0
u
Θc = 1·0
p (Nc < 5·14/F )
0·2 0·2
0·1 0·1
0 0
(a) (c)
1·0 1·0
0·9 0·9
Θc = 0·5
u
0·8
Θc = 0·5
0·8 Θc = 1·0
u
u Θc = 2·0
0·7 Θc = 1·0 0·7 u
u
Θc = 2·0 Θc = 4·0
p (Nc < 5·14/F )
0·4 0·4
0·3 0·3
0·2 0·2
0·1 0·1
0 0
1 1·5 2 2·5 3 1 1·5 2 2·5 3
F F
(b) (d)
Fig. 13. Graphs showing the relationship between p(Nc , 5:14=F) and F for a soil with COVcu (a) 0´125, (b) 0´25, (c) 0´5 and (d) 1