Vous êtes sur la page 1sur 15

Gudehus, G. & Nübel, K. (2004). Géotechnique 54, No.

3, 187–201

Evolution of shear bands in sand


G . G U D E H U S * a n d K . N Ü B E L †

Localisation of deformation in narrow shear bands is a La localisation de déformation dans des bandes de cisail-
fundamental phenomenon of granular material behav- lement étroites est un phénomène fondamental du com-
iour. When modelling granular materials, shear localisa- portement des matières granulaires. Pour la modélisation
tion has therefore to be considered. The model must des matières granulaires, la localisation du cisaillement
include an adequate constitutive law with an intrinsic doit donc être prise en compte. Le modèle doit inclure
length determining the thickness of the localised shear une loi constitutive adéquate avec une longueur intrinsè-
zones, and a representation of microscopic inhomogene- que déterminant l’épaisseur des zones de cisaillement
ities triggering shear localisation. In the framework of a localisées ; il doit aussi inclure une représentation des
Cosserat continuum the behaviour of non-cemented gran- inhomogénéités microscopiques déclenchant une localisa-
ular material can be described by a hypoplastic law tion de cisaillement. Dans le cadre de travail d’un con-
including the evolution of non-symmetric stresses and tinuum de Cosserat, le comportement d’une matière
couple stresses. Void ratio fluctuations are introduced by granulaire non cimentée peut être décrit par une loi
a physically justified probability density function. The hypoplastique incluant l’évolution des contraintes non
assumption is that fluctuations are minimal in the densest symétriques et des contraintes couplées. Des fluctuations
state, and maximal in the critical state. It is shown that du taux de pore sont introduites par une fonction de
shear bands can evolve spontaneously in the interior of a densité de probabilité justifiée sur le plan physique. Nous
granular body even under homogeneous stress or displa- supposons que les fluctuations sont minimes dans l’état le
cement boundary conditions. Experimental verifications plus dense et maximales dans l’état critique. Nous mon-
are given. Numerical results are compared with labora- trons que les bandes de cisaillement peuvent évoluer
tory tests performed with a Cornforth and a Hambly- spontanément à l’intérieur d’un corps granulaire même
type biaxial apparatus. Shear localisation in the experi- sous une contrainte homogène ou des conditions limites
ment is visualised by means of an optical method called de déplacement. Nous fournissons les vérifications expéri-
particle image velocimetry. Further numerical results are mentales. Nous comparons les résultats numériques avec
compared with experimental results taken from publica- les essais de laboratoire réalisés avec un appareil biaxial
tions. Conrford et de type Hambly. La localisation de cisaille-
ment dans l’expérience est visualisée au moyen d’une
KEYWORDS: deformation; earth pressure; friction; limit state méthode optique appelée vélocimétrie d’image de parti-
design/analysis; numerical modelling and analysis; shear cule. D’autres résultats numériques sont comparés avec
strength les résultats expérimentaux publiés.

INTRODUCTION grains’ thickness. He found that the evolution of shear bands


The notion of slip surface, shear band or narrow shear zone did not lead to slip lines with constant jm as usually
was introduced by Coulomb (1773). Rankine (1856) pre- presumed. Moreover, he initiated attempts to obtain limit
sumed a static limit condition to hold in a sand body with a states by numerical calculations of deformation fields. As
free plane surface. Sokolovsky (1942) extended this theory, such problems turned out to be mathematically ill-posed,
using slip lines as characteristics. Without using Coulomb’s various methods of regularisation have been proposed, in-
extremum principle, his resulting earth pressure Ea have cluding polar quantities.
been formally obtained. It has not been proven, however, The evolution of shear bands was first numerically calcu-
why and under what conditions slip lines can occur. Darwin lated with a Cosserat extension by Mühlhaus (1987), and
(1883) made an important contribution that has remained then by Tejchman (1989), using elastoplasticity. Parallel to
scarcely noticed until now. By means of model tests with further work along this line, Huang (2000), Tejchman &
fine dry sand and a wall rotating around its foot, he found Gudehus (2001), Huang & Bauer (2002) and Huang et al.
(taking the slope angle for j) that none of the theories of (2002) improved hypoplastic constitutive relations and car-
his time suited his observations. The procedure of filling had ried out systematic studies.
a strong influence on Ea , which he called the historical The present paper contains a selection from the thesis
effect, following the advice of C. Maxwell. by Nübel (2002), and consequences drawn by both authors.
Roscoe (1970) clarified several features of shear localisa- The constitutive relation is briefly outlined with an as-
tion in sand by means of innovative experiments. He showed sumed shearing along a rough boundary. Fluctuations of
that void ratio, shear deformation and mobilised friction void ratio are then introduced, together with their reduc-
angle, jm , increase together in zones of approx. 10–20 tion by compression or cyclic shearing. The first example
with shear banding is a slender biaxial sample. Stretching
of more compact rectangles leads to several dominating
bands. This is shown for a biaxial sample and a strip with
Manuscript received 29 July 2002; revised manuscript accepted 20 free surface. Cases of active and passive earth pressure are
October 2003.
Discussion on this paper closes 1 October 2004, for further details
presented in the next section, followed by two further
see p. ii. cases: vertical punching of a strip, and the active and
* Institute of Soil and Rock Mechanics, University of Karlsruhe, passive trapdoor problem. In all cases theoretical and
Germany. experimental results are strikingly similar. Finally, conclu-
† Ed. Züblin AG, Stuttgart, Germany. sions are outlined.

187
188 GUDEHUS AND NÜBEL
"   #
BOUNDARY LAYER AND CONSTITUTIVE RELATIONS 3p n
Consider shearing of dry sand along a rough, plane and ec ¼ ec0 exp  (2)
hs
rigid boundary (Fig. 1). The tangential displacement, u (Fig.
1(a)), may be linearly distributed outside a boundary layer
of thickness db, and may start at the boundary with u ¼ 0 wherein the granulate hardness, hs , is used for scaling p.
and @u/@z ¼ 0 for z ¼ 0. The average rotation, ø (Fig. There are upper and lower bounds of possible void ratios, ei
1(b)), is constant outside the boundary layer, and tends to and ed , which decrease with p/hs as ec . The position of e in
zero at the boundary. There is a couple stress, M, in the the e–ln p plot can be described by the pressure-dependent
boundary layer (Fig. 1(c)), and also a non-symmetric portion density index, Id ¼ (ec  e)/(ec  ed ).
of shear stress,  xz   zx (Fig. 1(d)). Balance of angular Figure 2(a) and (b) may be compared with critical state
momentum requires @M/@z ¼  xz   zx . The angle of soil mechanics (CSSM; Schofield & Wroth, 1968). CSSM
shearing, ª ¼ @u/@z, is plotted in Fig. 1(e). There is a presumes
curvature, k ¼ @ø/@z, inside the boundary layer (Fig. 1(f)).  
Volume changes are excluded. Shear and normal stress p
ec ¼ ecr  º ln (3)
(abbreviated  ¼  xz ,  ¼ zz ) are the same on wall-parallel pr
planes for equilibrium (Fig. 1(g)). The other stress compo-
nents (xx and  yy ) are not plotted. All quantities are for critical void ratios with reference values ecr and pr and a
presumed to be constant along the x-axis. slope º. The reference stress hs in equation (2) is objective,
The polar quantities may be imagined as those in a whereas pr is subjective. The asymptotes of equation (2) for
clamped beam, as in Fig. 1(g). M is like the bending p ! 0 and p ! 1 are physically reasonable, whereas
moment; the non-symmetric shear stress,  xz   zx , corre- equation (3) holds only for a certain range of p. The upper
sponds to the horizontal internal shear force. The curvature, and lower bounds, ei and ed , decrease with p by equation (2)
k, is related to the local rotation, ø, by k ¼ @ø/@z as for a with factors ei0 and ed0 instead of ec0 . The slope º ¼ de/
beam. Outside the boundary layer with polar quantities ø ¼ d ln(p/pr ) is not the same therefore as presumed in CSSM.
ª holds as usual. Explanations of polar quantities for two As in CSSM, ei can be related with an isotropic compres-
dimensions are given by Mühlhaus (1987) and Tejchman sion of a loose skeleton. e . ei would be possible only with
(2001), for example. macropores. The lower bound of plasticity is not specified in
Figure 2 shows the evolution of states for the kinematics CSSM, whereas e , ei in hypoplasticity would mean crack-
of Fig. 1(a,b,e,f) with constant strain rate. The sand is ing due to shearing, as in dry masonry.
assumed to be rate-independent, so that stress paths can be The hypoplastic constitutive relation describes the evolu-
labelled only by order and not by real time. The initial state tion of effective stress components with the evolution of
may be isotropic with pressure p and void ratio e. Consider strain components by a set of differential equations. Critical
first the zone z > db without polar effects. Typical stress states can be reached asymptotically by monotonic shearing.
paths as calculated with the hypoplastic law tend to critical The stress component ratio tends to be constant by equation
states with (1), and correspondingly for cylindrical deformations, as in
jj ¼  tan jc (1) CSSM. Thus the determination of jc is easy. For shearing
with constant e the mean pressure, p, decreases (A, Fig.
(see Fig. 2(a)). The mean pressure p ¼ (xx +  yy + zz )/3 2(b)) or increases (B) if e . ec or e , ec respectively holds
tends to that for e ¼ ec as sketched in Fig. 2(b). The critical initially. For shearing with constant  or p, the void ratio
void ratio ec decreases with p/hs via decreases (C) or increases (D) if e . ec or e , ec

z z z z

db db db db

0 u 0 ω 0 M 0 τxz ⫺ τxz
(a) (b) (c) (d)

τ
z z

db db

0 γ 0 κ x
(e) (f) (g)

Fig. 1. Variables in a sheared strip


EVOLUTION OF SHEAR BANDS IN SAND 189
c
τ M
B c

D
E
B
A C

ϕ A
0 σ 0 τ ⫺ τzx
(a) (c)

e db
A dg
C
D
ei
B ec
ed

0 ln ( p/hs) 0 Id
1
(b) (d)

Fig. 2. Evolution of: (a) stress component; (b) void ratio and pressure; (c) polar
stresses; (d) boundary layer thickness (mean grain size d g )

respectively holds initially. The initial contraction for e , ec close to ep at the beginning, the asymptote ep is approached
is dropped in Fig. 2(b) for simplicity. The peak of / for in the usual range of p. As shown in Fig. 4(a) this requires a
this case (D) is indicated in Fig. 2(a); it is related to shear loose initial packing. Such curves can be used for determin-
localisation as outlined further below. The cases A, B, C and ing hs and n in equation (2). For hard minerals hs exceeds
D are qualitatively the same as in CSSM. 1 GPa; then the turning point p ¼ hs /3 of e against ln p by
For an oedometric compression e tends to ep by equation equation (2) is far above the usual p-range. n ranges from
(2) with ep0 instead of ec0 , with ec0 , ep0 , ei0 . If e is about 0.2 for angular to 0.3 for round grains.

1 Id 0

Id ⫽ 0·5

0·2

ed ec ei e

(a) (b)

Fig. 3. The density index, Id , scatters randomly (a) with a number distribution (b)
depending on the average I d

0·16
ep
0·80 Loose 0·004
Loose Shear components

Mean

se
e

normal comp.
Dense Dense
Dense Loose

0·50 0 0
10 100 1000 0 2 4 6 0 2 4 6
p: kPa ε 1: % ε 1: %
(a) (b) (c)

Fig. 4. Evolution of (a) void ratio, e, (b) void ratio fluctuations, s e , and (c) stress fluctuations, só , with oedometric compression
190 GUDEHUS AND NÜBEL
The lower bound ed of e is approached by cyclic shearing Shearing along a plane rough boundary has been analysed
with constant pressure, as shown for example in Fig. 5(b). numerically (Tejchman & Gudehus, 2001; Huang & Bauer,
ed0 for p ¼ 0 is close to the conventional emin for hard 2002; Huang et al., 2002) with different initial and boundary
grains, and ec0 is then close to emax . These and further rules conditions. As might be expected, the thickness, db, of the
for determining the hypoplastic parameters are given by boundary layer with polar stresses is a multiple of dg : for
Herle & Gudehus (1999). An exponent  in the range 1 ,  example, db /dg  5 for rounded grains. db /dg increases
, 2 is needed for modelling compression. More important nearly in proportion to cp , which can be used to estimate cp .
is an exponent Æ in the range 0.1 , Æ , 0.3. Together with For a given cp , db /dg increases if the initial density index,
jc and the density index Id , Æ determines the peak friction Id , decreases (Fig. 2(d)). db /dg is lowest for densest states
angle, jp . As the actual jp depends on shear localisation, a (Id ¼ 1) and explodes for Id ! 0. This means that shear
precise determination of Æ requires a more detailed analysis, banding can occur only if the void ratio at the onset is lower
as the one in the section ‘Slender biaxial samples’ below. than critical.
Abrasion and fragmentation of grains may be neglected for
p/hs , approx. 103 .
Consider now the evolution of polar stresses (Fig. 2) in FLUCTUATIONS AND THEIR DECREASE
the boundary layer. With distributions like those in Fig. 1, For solving boundary value problems with hypoplasticity,
the rates ª_ and k_ of shear strain and curvature are assumed an initial state field must be given with void ratio e and
to be stationary. Starting from arbitrary small quantities, the stress components. It is helpful for the analysis of shear
polar stresses tend to critical values with a ratio determined banding to start with fluctuating e-fields. Stress fluctuations
by k_ =ª_ . The non-polar stress paths are close to those of Fig. can be generated as shown below.
2(a), and the magnitude of the polar quantities depends on Figure 3(a) shows a rectangle subdivided into rectangular
e ¼ ec as p in Fig. 2. Equation (1) is replaced by the more elements with a randomly distributed density index, Id . The
general friction condition (Huang et al., 2002) range of e is subdivided into 100 steps. The numbers, N, of
 2  2 cells for each step are replaced in Fig. 3(b) by continuous
 M= d g þm[ð   zx Þ2 =] curves. These number distributions are taken over from
þ ¼1 (4)
 tan jc c2p Shahinpoor (1981).They are exponential, with a cut-off at ed
and a decay towards ei above ec . In accordance with
where dg is a representative grain size, for example d50 from observations, the scattering increases with decreasing aver-
the grain size distribution. This is a characteristic length of age density index, I d . The spatial distribution also depends
the material. m is a constant, with order of magnitude 1, on the element size, but this lack of objectivity had no
which can be calculated from the constitutive relation. The influence on our results, as was found by comparative
polar friction coefficient, cp , increases with the angularity of calculations. The size of the smallest cell is about three
the grains. Imagine a single grain in the interior of a grain grain sizes, and there is a fractal distribution of fluctuation
skeleton rotated from outside. The resisting moment will lengths.
tend to a limit that increases with angularity and with Figure 4 shows the evolution of void ratios and fluctua-
density. tions calculated for an oedometric compression (the average
The hypoplastic constitutive relation with polar quantities axial shortening, 1 , increases from 1 ¼ 0). The initial state
is formulated in such a way that the desired asymptotes are is isotropic with p ¼ 78 kPa and loose or dense with a
obtained as attractors of solutions. The asymptote in equa- distribution of e as in Fig. 3(b). For loose sand, the spatial
tion (4) is approached by pure shearing with curvature (Fig. average, e, of e approaches ep for proportional compression
2(c)). The asymptote by equation (1) is approached by pure described by equation (1) with ep0 instead of ec0 , whereas
shearing without curvature; the polar stresses then tend to the ep line is not approached for dense sand (Fig. 4(a)). The
zero. The latter property ensures that the non-polar hypo- scattering, s e , of e decreases with compression, more for
plastic relation is included consistently and suffices outside loose than for dense sand (Fig. 4 (b)). This is due to the
shear bands. The analysis of experiments with shear localisa- common asymptote s e ! 0 for p ! 1 (which cannot be
tion has shown that a good fit is obtained with the only reached): the higher s e is (and the lower Id therefore, cf. Fig.
additional parameter, cp , in the range from 0.5 to 3 for 3(b)), the stronger is its decrease by compaction.
rounded to angular grains and m ¼ 1. Figure 4(c) shows the calculated evolution of stress fluc-

σ1

0·16
(b) (c)
ε1
0·80

ε2
0·70
se

0·08
e

(a)
(d)
0·60

ed (e)

0·50 0
⫺2 0 2 ⫺2 0 2
ε 1: % ε 1: %

Fig. 5. Evolution of void ratio, e, and its fluctuation, s e , under constant pressure and alternating lateral strain (a); (b, c)
initially loose specimen; (d, e) initially dense specimen
EVOLUTION OF SHEAR BANDS IN SAND 191
tuations s . Assuming zero at the beginning, high values are
generated with a small strain. This is typical for an assembly
of hard grains enclosed by rigid moving boundaries. The
deviatoric stress fluctuation is bigger for loose sand, and is
suppressed only by very high pressures. The normal stress A
fluctuation is smaller and less regular, and is also reduced
with increasing pressure.
This can be related to so-called stress chains in granular
assemblies (Behringer & Miller, 1997; Behringer et al.,
2001). They are more marked for loose than for dense
skeletons: the larger the fluctuations of void ratio, the larger
therefore are the fluctuations of stress. The stress deviator
fluctuates more than the mean normal stress, as the latter is
an average over all directions. The stress fluctuations tend to
zero for p ! 1. s decreases more rapidly by compression
if it is higher. The hypothetical limit e ! 0 for p ! 1 B
under compression is thus an attractor for void ratio and
stress fluctuations: both tend to zero.
Starting with the same initial states as for Fig. 4, void
ratios are far more reduced by cyclic lateral strains, 2 ,
under constant pressure. This is shown in Fig. 5 for 1 ¼
78 kPa and mean 2 up to 2%. The mean vertical strain
(Fig. 5(a)) is alternating with a cumulative increase. Initial
fluctuations of e are taken similarly as in Fig. 3(b) with Id ¼
0 and Id ¼ 0.9. The evolution of mean void ratio, e, and
scattering, s e , is plotted in Fig. 5(b) and (c) for the loose
initial state, and in Fig. 5(d) and (e) for the dense state. C
Independently of the initial state, e tends to ed depending on
p by equation (1) with ed0 instead of ec0 . The scattering
tends to zero for e ! ed in accordance with Fig. 3(b). More
cycles are needed for the approach with higher initial void
ratio and with smaller strain amplitudes. (a) (b) (c) (d)
The state of maximum cyclic densification is thus an
attractor, with implied zero density fluctuation. The lower Fig. 6. Calculated evolution of shape and state during biaxial
bound, ed , in the e–ln p plot (Fig. 2(b)) is a kind of phase compression: (a) deformed mesh; (b) density index, Id ; (c)
:
volume strain rate, v; (d) mean couple stress, M
limit. For e , ed the granular skeleton would behave like
dry masonry with uniaxial strength and cracking by shear-
ing. 7 for the overall average and for the interior of the govern-
ing shear band.
Just before the 1 – 2 peak (A) volumetric strain rate,
SLENDER BIAXIAL SAMPLES _ 1 þ _ 2 , and couple stress, M, show a pattern (Fig. 6(c) and
The numerical results with polar effects shown in this (d)), whereas deformation and density, Id , (Fig. 6(a) and (b))
paper have a common numerical background. Quadrilateral do not. Just after the peak (B), the patterns are more
finite elements including Cosserat rotations and the hypo- marked, but still do not appear for deformation and density.
plastic relation were implemented in an ABAQUS code by There is a strong dilation inside the shear band, whereas the
Huang (2000). Time integration of the constitutive relation average e increases to the initial value after a small decrease
is performed with an implicit time integration scheme. To (Fig. 7(b)). The rotation increases inside the shear band after
prevent volumetric locking a selective reduced integration is the peak (Fig. 7(c)). The fluctuation of e increases in
applied (Zienkiewicz et al., 1971). The material parameters the shear band after the peak and remains low outside
have been determined with disturbed samples (Herle & (Fig. 7(d)).
Gudehus, 1999). The polar parameters have been estimated The pattern formation, which is a kind of self-organisa-
as outlined in the previous section. The size of the elements, tion, may be explained as follows. Alongside with fluctua-
visible in Figs 6(a) and 19(b), is about 3dg ; this was found tions of normal stresses, a minute fluctuation of couple
to suffice for obtaining mesh-independent shear bands. The stress, M, arises, say of amount M0 . For an assumed shearing
initial void ratio was randomly distributed according to Fig. with constant rate _ 1 ¼ 1 and dilation ratio  ¼ _1 =_2 the
3 by a Monte Carlo method. The boundary conditions were hypoplastic evolution equation for M can be written as
changed in steps with a predefined loading factor. In general, 
f b  1 þ  2 pffiffiffiffiffiffiffiffiffiffiffiffiffi
for the non-linear process multiple iterations were needed to _
M¼M  2 a^ f d 1 þ 2 (5)
obtain equilibrium. Quadratic convergence was obtained with 3p 3p
a Newton–Raphson scheme for the set of linearised alge-
braic equations. where p ¼ (1 + 2 + 3 )/3 is the mean pressure, f b is
The calculated spatial evolution in a slender sample with proportional to hs (p/hs )1 n , fd ¼ (1  Id )Æ is calculated
constant lateral pressure 3 under biaxial compression is with Æ  0.2, and a^  0:15 depends on the friction angle
shown in Fig. 6. The average initial density index is rather jc . Equation (5) leads to exponential growth for
high, I d ¼ 0:87, and the density fluctuation is low (cf. Fig.
3(b)). These results are very similar to experimental ones by  1 3º þ 
. (6)
Desrues (1998) and Finno et al. (1997). The evolution of  2 1  3º
some quantities with average axial strain 1 is plotted in Fig.
using the empirical result 2  (1 + 3 )/3 (e.g. Finno et al.,
192 GUDEHUS AND NÜBEL
400 0·20
A B C A B C A B C
0·75
Inside shear Inside
σ1 ⫺ σ 2

band

ω
e
Complete sample
Complete
0 0·55 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
ε2: % ε2: % ε2: %
(a) (b) (c)

0·20 0·006 2 ⫻ 10⫺4


A B C A B C A B C

Inside Inside

M: kNm
Shear components

se

Complete Complete

Normal components
0 0 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
ε 2: % ε2: % ε2: %
(d) (e) (f)

Fig. 7. Calculated evolution of state during biaxial compression, averaged over the whole sample and inside the governing shear band

pffiffiffiffiffiffiffiffiffiffiffiffiffi
1997) and the abbreviation º ¼ a^ f d 1 þ 2 . As  depends and maximal scattering (Fig. 7(e)), which corresponds to
on 1 /3 via the constitutive relation (close to Rowe’s experimental observations by Oda & Kazama (1998) and
stress–dilatancy relation), equation (6) is an implicit inequal- numerical discrete element simulations (Viggiani et al.,
ity for 1 /2 . It yields a 1 /2 below peak for Id near 1, and 1997). After the peak the particle rotations grow continu-
always 1 /2 below the limit by equation (1) for Id . 0. ously (Fig. 7(c)), the stress deviator comes close to the
Thus we have a spontaneous polarisation more or less critical value (Fig. 7(e)), the normal stress fluctuations
before the peak. It starts at random locations where equation approach a maximum (Fig. 7(d)), and the couple stress
(6) is first satisfied, according to initial and induced fluctua- decreases (Fig. 7(f)).
tions of void ratio and stress. Some parts of the sample thus The shear band thickness, ds, can be estimated as follows.
reach a maximal stress ratio, max|/| ¼ tanjp , on planes Calculated mean values of shear rate, ª_ , and couple stress,
with inclination Łp  458 + jp /2, earlier than the average M, along the band are plotted in Fig. 9 against its normal.
(Fig. 8(a) and (b)). The peak friction angle, jp , is deter- Approximating by Gaussian distributions, ds may be defined
mined by Id (Herle & Gudehus, 1999) and scatters as Id by means of the second moments. ds is not the same for ª_
does. There are short narrow zones of thickness ds where as for M, as was noticed earlier (Tejchman & Gudehus,
max|/| has been reached (Fig. 8(c)). The dilation ratio, 2001). An approximation of ds is obtained from shearing
_=ª_ (Fig. 8(d)), also goes through a peak. Id decreases, along a rough wall with constant pressure (Huang et al.,
and with it jp . Neighboured zones at peak unite owing to 2002). The thickness, db, of a boundary layer (second
positive feedback. section above) is about one half of ds .
One of the narrow shear zones has the lowest average jp The position of the governing shear zone is random. It
and goes through first. It wins in the competition, the further can be at any height, or the mirror image of the one in Fig.
deformation is concentrated there. Well beyond the peak, 6, and can also be reflected at the top or bottom plate. It is
changes of shape and state are strongly localised (stage C in triggered by a combination of highest void ratio and highest
Fig. 6). As can be seen in Fig. 7, inside the governing shear stress ratio. Only the shear band inclination, Ł, can be
zone the void ratio approaches the critical value (Fig. 7(b)) estimated in advance. Repeated numerical calculations like

τ
γ·
θa
d
s
2θa
σ3
ϕa σ
σ3 σ1 σ τ ε·
θa
θa

(a) (b) (c) (d)

Fig. 8. Some parts of the sample reach a maximal stress ratio earlier than the average
EVOLUTION OF SHEAR BANDS IN SAND 193
0·3
Id ⫽ 0·97
Id ⫽ 0·47
xn Id ⫽ 0·13
γ·
0·2

N
0·1
M

0
32 36 40 44
xn ϕp: degrees
(a) (b)
Fig. 10. Frequency distribution, N, of the friction angle j p , for
Fig. 9. Definition of the thickness of a shear band different overall initial density index, I d , from 150 simulated
biaxial tests with random void ratio distribution

those for Fig. 6 yield Ł between approx. 568 and 608 for the
same conditions (Łp is slightly reduced by the mesh). considered as a soil constant, even for given initial Id .
Experiments by Yoshida et al. (1994) yield Ł  628 for the Comparative calculations have shown that the findings
same data. Bifurcation analysis with an assumed uniform shown in Figs 6, 7 and 10 are not numerical artefacts. The
shear zone leads to Ł  638. All this agrees with earlier initial distribution of void ratio, the arrangement of elements
findings (Vardoulakis et al., 1978), so that Ł ¼ 458 + jp /2 is and the convergence criteria had been varied, which had no
not more than a crude upper bound. influence upon the results.
The overall peak friction angle, jp , can only be estimated.
It is determined from the peak of 1  3 for constant 3
(as in Fig. 7(a)), and it depends strongly on the initial COMPACT BIAXIAL SAMPLES AND EXTENDED
overall density index, I d . Results of 150 simulated biaxial STRIPS
tests are plotted in Fig. 10. The frequency distribution has a Tests were made in a Hambly-type biaxial apparatus (as
low scattering; the mean, jp , depends quite markedly on I d . described by Topolnicki et al., 1990) and analysed with our
The results would be different for other loading paths up to numerical model. The sample is kept rectangular by a pair
peak. Properly speaking, jp is determined by the actual Id of fixed and two pairs of movable smooth, hard plates (Fig.
(Herle & Gudehus, 1999) in the shear band, and j decreases 11(a)). Karlsruhe sand (e.g. Herle & Gudehus, 1999) with
to jc when Id tends to zero (critical state). Owing to the d50 ¼ 0.45 mm was used. A silicone grease between the
historical element (Darwin, 1883), jp therefore must not be rubber mould around the sample and the plates kept the

Low ∆V High

x2

x3 x1

(b) (c)

A A

A–A

(a)

(d) (e)
PIV visualisation Calculated

Fig. 11. Experimental and calculated evolution of primary and secondary shear band patterns of Karlsruhe sand for plane
strain with a Hambly-type biaxial apparatus: (a) experimental device; (b, d) experimental shear band patterns; (c, e)
calculated shear band patterns
194 GUDEHUS AND NÜBEL
friction there very low, so that the measured pressures are et al., 1998). Knowing the displacement field the deforma-
principal stresses. The mean confining pressure has been tion field can be calculated (Nübel & Weitbrecht, 2002). A
kept below 100 kPa by servo control. PIV image of volume strain, ˜V, is shown in Fig. 11(b).
Digital photographs of the sample were taken through the Four shear zones forming a rhomboid are more or less
top glass plate, and evaluated with particle image velocime- visible. Closer inspection shows further less marked bands
try (PIV) (Nübel & Weitbrecht, 2002). As shown in Fig. 12 parallel to the dominant ones. It is worth mentioning that
the grains can serve as tracers. For evaluating a 2D deforma- the PIV method to observe deformation fields is limited to
tion field, digital images are taken by means of a CCD two dimensions. Experimental studies demonstrated that
camera from consecutive stages. The digital image is divided shear band patterns can be also visualised in three dimen-
into small disjointed sub-areas called interrogation cells. If sions by computed tomography (Alshibli et al., 1984;
the deformation of the specimen between two consecutive Desrues et al., 1996).
images is sufficiently small it can be supposed that the Calculated patterns are shown in Fig. 11(c) and (e). The
characteristic patterns of the interrogation cells do not coincidence of both orientation and thickness with the
substantially change their characteristics but only their loca- experiment is good. The positions are random both in theory
tion. Therefore a local displacement vector can be deter- and experiment. The similarity of thick bands in Fig. 11(b)
mined for each interrogation cell between two consecutive and (c) can be considered as a validation of our theory. The
images by means of a cross-correlation function (e.g. Raffel evolution is similar to that in Fig. 6, with the difference,
however, that dislocations at the boundaries are excluded by
the plates on all sides.
Area of interest (AOI) The shear bands are reflected at the smooth boundaries.
Polar quantities of opposite signs in two crossing bands
cancel each other in the crossing area. The further shear
bands in Fig. 11(b)–(e) are secondary: that is, they appear
after the dominating ones. The band thickness is roughly
twice that of Fig. 2(d). A strong fluctuation of deviatoric
stress has been observed and also calculated.
Figure 13 shows a model test with uniform stretching of
the base of a horizontal sand layer (initial height 15 cm)
with a free surface. The walls are hard and smooth, two side
(a) (b) walls are from glass, and a rubber membrane at the bottom
lies on a hard smooth plate. The sand was filled in densely
Fig. 12. (a) Area of interest and (b) interrogation cell for PIV (I d  0:85) with thin horizontal marker layers of black

(a)

(b) (c)

Fig. 13. Sand layer with free surface and stretched base: (a) shear bands; (b) surface waves; (c) shear bands visualised
with PIV
EVOLUTION OF SHEAR BANDS IN SAND 195
grains. After 10% stretching two families of parallel shear maximum as explained with Fig. 3(b), which is thus justi-
bands had developed (Fig. 13(a)). The surface had become fied. The scattering of pressures, s (Fig. 15(e)), starts from
wavy with maximal slopes near the outcrops of the shear assumed zero and tends to a plateau. The couple stress, M12
bands (Fig. 13(b)). PIV images (Fig. 13(c)) show the thick- (Fig. 15(f)), grows exponentially up to a maximum and then
ness and distance of the shear bands in more detail. tends to zero (cf. Fig. 7(f)). In the asymptote the whole field
The calculated evolution (Fig. 14) agrees with the essen- is paved by fully dilated shear bands, and the average couple
tial observations. Two primary families of bands appear stress is zero as those of crossing bands (cf. Fig. 11(e))
alongside, with a minor crisping of the free surface (Fig. compensate each other. The fluctuation of M (not plotted)
14(a)). Further stretching leads to more marked secondary reaches a maximum, and also that of deviatoric stress and
bands and deeper offsets at the outcrops of the primary void ratio. An evolution of shear bands as in Fig. 15(a) and
bands (Fig. 14(b)). The critical void ratio (Id ¼ 0) is nearly (b) was experimentally observed by Desrues et al. (1996)
reached in the latter. The band thickness is nearly twice that and Mokni & Desrues (1998).
of Fig. 2(d), taking the polar constant am ¼ 1.0, which This evolution is typical for self-organisation of dissipa-
replaces cp in equation (4) and can thus be determined. tive dynamic systems. There is a pattern formation with
The extension of a strip has also been calculated with characteristic lengths and fractal nesting, and deterministic
constant lateral pressure and rigid smooth top and bottom chaos in the end. For these critical phenomena, the soil
plates (Fig. 15). This corresponds to a flat biaxial sample. A mechanics critical state is a kind of strange attractor. In
primary zigzag pattern evolves, followed by secondary and simpler words, a sand body can be brought to an overall
tertiary patterns (Fig. 15(a) and (b)). Inclinations and thick- critical state by large uniform boundary stretching, and then
nesses are the same as before. The distance of the primary the disorder is maximal. On the way there, a regular shear
bands is evidently given by the size of the sample; the band pattern can be achieved.
following bands are nested in the primary pattern.
The evolution of state quantities may be explained with
some sketches. The mean void ratio (Fig. 15(c)) tends to the VICINITY OF MOVING WALLS
critical value for the given pressure, ec by equation (2), An example of a smooth rigid wall translated back from
along with 1 /2 ¼ tan2 (458 + jc /2) and 2  (1 + 3 )/3. an initially dense sand layer is shown in Fig. 16 by contours
The scattering of void ratio s e (Fig. 15(d)) tends to a of shear distortion. The calculated results for two wall

1 Id 0

(a) (b)

Fig. 14. Calculated surface and density index, Id , fields for two stretching stages of the test of Fig. 13: (a) ˜l/l
10%; (b) ˜l/l 20%

1 Id 0 e
ec sσ

0 ⫺ε3 0 ⫺ε3
(c) (e)
(a)

se M12

0 ⫺ε3 0 ⫺ε3
(b) (d) (f)

Fig. 15. Calculated biaxial compression of a flat sample with lateral constant pressure: (a, b) evolution of shear bands; (c,
d) average and scattering of void ratio; (e) scattering of pressure; (f) couple stress
196 GUDEHUS AND NÜBEL

Low εp High
u ⫽ 0·2 cm u ⫽ 0·4 cm
u u

PIV section

Finite element simulation Finite element simulation


(a) (b)

Experiment Experiment

(c) (d)

Fig. 16. Active wall of 6 cm height under translation, numerical and experimental results

displacements (Fig. 16(a), (b)) are in good agreement with Two examples of a smooth rigid wall rotated around its
the PIV results (Fig. 16(c), (d)). The deformation is almost foot back from an initially dense sand layer are shown in
totally localised to a single, practically plane, shear zone. Fig. 17. The calculated contours of volume expansion (Fig.
This reaches a thickness of about twice that of a boundary 17(a)) are close to those observed with X-rays by Milligan
layer as given in Fig. 2(d). The dilation inside tends to a (1974) (Fig. 17(b)). There are three dominant shear bands,
critical state. The inclination Ła ¼ 648 is close to that of a nearly with Coulomb inclination, and steps at their outcrops
slender biaxial sample (cf. Fig. (6)). (cf. Figs 13 and 14).
These findings again support our theory, but not the con- A closer inspection shows that there are substantial differ-
ventional earth pressure theory. The resultant force, Ea , can ences from Rankine’s theory. Fig. 17(c) shows calculated
correctly be obtained with a suitable peak friction angle, jp , density contours for a small wall rotation so that the free
and Ka ¼ tan2 (458  jp /2), but jp depends on the overall surface is still practically horizontal. Apart from three
initial density index, nearly as for slender biaxial samples dominant shear bands there are two families of less marked
(Fig. 10). The proper jp for Ka depends on the actual I d bands. The mobilised friction angle along them is lower and
in the shear band, and tends to jc for Id ! 0. The earth not uniform. With increasing wall rotation the earth pressure
pressure distribution (not plotted) is linear with depth only if distribution becomes increasingly non-linear.
there is a smooth rigid bottom behind the wall foot. Other- Stronger deviations from classical theories are obtained
wise it increases linearly in the upper half and decreases for passive cases. This is shown in Fig. 18 for a smooth
strongly with further depth. The latter finding is known from rigid wall pushed into an initially dense layer. The contours
more conventional finite element calculations and experi- of shear strains from calculation (Fig. 18(b)) agree with
ments. The grain-size effect is missed therein, however: the those observed with PIV (Fig. 18(a)). Instead of a single
wall displacement up to Ea depends on the ratio of grain wedge, a curved slip band starting horizontally at the foot is
size, dg, and wall height, h, in particular for h/dg , approx. obtained, together with three weaker bands from the wall
200 (Tejchman, 1997). The pressure distribution also de- top. As has been observed by Roscoe (1970), this kind of
pends on dg /h.

1 Id 0

ω ω

Simulation Experiment Simulation


(a) (b) (c)

Fig. 17. Calculated and experimental (Milligan, 1974) density decrease behind a smooth rotating wall
EVOLUTION OF SHEAR BANDS IN SAND 197
Low εp High
PIV section
u ⫽ 1·0 cm
u

Experiment Finite element simulation

Fig. 18. Passive wall translation, numerical and experimental strain deviators

progressive failure cannot be approximated by a classical of passive earth pressure, Kp , decreases with increasing ratio
formula; what is the representative mean j therein? of wall height, h, and grain size, dg, as the slip band
The calculated earth pressure distribution is linear with thickness increases with dg (cf. Fig. 2(d)).
depth only for a smooth bottom. Otherwise it grows far The numerical calculation of stress and deformation fields
stronger than linear along the lower half of the wall. As with for such cases has turned out to be more difficult than
the active case, the results depend also on the grain size, dg, presumed 30 years ago (Simpson & Wroth, 1972). Our
at least for h/dg , approx. 200. The shear band is far calculations with such delicate features as those in Fig.
thicker than in the previous cases. 20(b) are numerically robust, which demonstrates that the
A similar dominant shear band has been obtained by regularisation by polar quantities serves the purpose. Having
pushing a rough rigid wall downwards into an initially dense achieved such a good agreement of theory and experiment,
sand layer (Fig. 19). This can be seen from dislocations of one could even go on with numerical instead of physical
thin dark layers in an experiment (Fig. 19(a)) and likewise model tests. However, the mesh size increases with (h/dg )2
from calculated distortions of a finite element mesh (Fig. for plane strain, as any possible shear zone of thickness ds
19(b)). More revealing is the evolution of localised dilation . approx. 10dg has to be subdivided.
(Fig. 20). The patterns of density decrease, as observed with
X-rays (Fig. 20(a)), are in remarkable agreement with the
calculated ones (Fig. 20(b)). A first narrow shear zone VICINITY OF VERTICALLY TRANSLATED
propagates nearly horizontally from the wall foot (stage A). HORIZONTAL STRIPS
It comes to a stop with further wall displacement, where- Figure 21 shows density contours in originally dense
upon a second curved band appears (stage B). Along with horizontal sand layers on a rigid bottom with a yielding
the latter a weaker band appears slightly above, and an strip, called an active trapdoor. X-ray observations (Fig.
opposite one ending at the wall top. Later on, a second band 21(a)) by Graf (1984) reveal nearly the same arc-shaped
of the latter kind appears, together with weak crossing bands shear bands as in our calculations (Fig. 21(b)). In the latter
(stage C). some other, wider, arcs are also visible. It has been calcu-
More precisely than reported by Roscoe (1970), the lated and observed with PIV that the region under the main
Karlsruhe experiment shown in Fig. 20(a) seems to justify arc is dilated. For the passive trapdoor—that is, for punching
the use of two sliding wedges for the calculation of Ep in a strip from the bottom—the agreement of measured and
(Gross, 1980). The slip plane inclinations can in fact be calculated density contours is even more striking (Fig. 22).
estimated from the assumption that they make Ep minimal Curved shear bands propagate from the edges of the strip
as compared with other inclinations. However, the mobilised upwards and sideways. Conventional static and kinematic
friction angle must not be assumed equal along all the slip limit state methods fail for such cases—a known difficulty
surfaces, and as given from the initial density. Actually the for tunnels and silos. Conventional numerical methods fail
bottom slip zone is almost fully dilated, the rising one less, for lack of regularisation.
and the one dipping from the wall top least. The coefficient The last case of the present selection is punching of a

(a) (b)

Fig. 19. (a) Passive earth pressure experiment by Schwing (1991); (b) calculated deformation
198 GUDEHUS AND NÜBEL
Low ∆V High

(a) (b)

Fig. 20. (a) Experimental and (b) calculated evolution of volume strain for the experiment of Fig. 19

Dense Id Loose
Dense Id Loose

30 cm
30 cm
40 cm
40 cm

(a) (b) (a) (b)

Fig. 21. (a) Experimental (Graf, 1984) and (b) calculated Fig. 22. (a) Experimental (Graf, 1984) and (b) calculated
density fields above of a trapdoor translated downwards by density fields above of a trapdoor translated upwards by u
6 mm 6 mm

rigid strip into an originally dense horizontal sand layer. with the passive trapdoor there is a loss of symmetry. A
Shear bands, as observed (Fig. 23(a)) for example by Vesic simplified analysis with three or four slip wedges, or with a
(1973), are similarly obtained by calculation (Fig. 23(b)). A wedge under the strip and with Ep at its one flank, might
rather undeformed wedge slides downwards and sideways, appear to be justified. More than for Ep , however, this would
and induces further shear bands similar to those in Fig. work only with the actual mobilised friction angles. These
20(b). The real wedge flanks are steeper than the calculated depend not only on the initial density, but also on position
ones, which is presumably due to a higher base friction at and on grain size, dg . The conventional calculation with Ng
the strip in the experiment. Weaker and steeper shear bands is therefore not justified.
evolve from the right edge and get stuck at some depth. As The dependence of Ng on dg is known from 1g and
EVOLUTION OF SHEAR BANDS IN SAND 199
Dense Id Loose

Foundation

(a) (b)

Fig. 23. Calculated evolution in comparison with an experiment by Vesic (1973): (a) experiment; (b) simulation

centrifuge model tests. This is obtained with our theory, in more precise boundary condition for the rather smooth base
fairly good agreement with the results of model tests by of the strip.
Tatsuoka et al. (1990) (Fig. 24). The calculated curves More detailed insight can be gained by another example
depend strongly on the base friction coefficient of the strip, of the same kind (Fig. 25). This time the pattern is more
and on the additional friction factor for the polar quantities. symmetric. Two dominant shear bands enable the sand to be
This enables a good matching, but prevents precise predic- pushed aside; two other strong bands get stuck at some
tions. A better coincidence may be obtained by means of a depth. Two families of weaker bands are reminiscent of

400
400
S.L.B. Sand, 1g

B ⫽ 1·2 cm
2P/(γB 2)
2P/(γB 2)

200 B ⫽ 2·4 cm
200
B ⫽ 1·0 cm
B ⫽ 2·5 cm
B ⫽ 10 cm

B ⫽ 10·0 cm

0 0·2 0·4
u/B 0 0·2 0·4
u/B
(a) (b)

Fig. 24. (a) Experimental (Tatsuoka et al., 1990) and (b) calculated influence of the footing breadth on
2P/(ªB2 )

Low εp High

(a) (b)

Fig. 25. Punching of a strip foundation: (a) experiment (Tatsuoka et al., 1990); (b) calculation
200 GUDEHUS AND NÜBEL
Sokolovsky’s (1942) theory, but cannot justify it as the Minutes Proc. Inst. Civ. Engng, 350–378.
mobilised friction angle is not uniform and not even known. Desrues, J. (1998). Localisation patterns in ductile and brittle
The competition of shear bands and the survival of a few is geomaterials. In Material instabilities in solids (eds R. de Borst
similar to those for slender biaxial samples (Fig. 6). and E. van der Giessen), pp. 136–157. John Wiley & Sons.
Desrues, J., Chambon, R., Mokni, M. & Mazerolle, F. (1996). Void
ratio evolution inside shear bands in triaxial sand specimens stud-
ied by computer tomography. Géotechnique 46, No. 3, 529–546.
CONCLUSIONS Finno, R. J., Harris, W. W., Mooney, M. A. & Viggiani, G. (1997).
Our investigations permit a number of conclusions, as Shear bands in plane strain compression of loose sand. Géotech-
follows. nique 47, No. 1, 149–165.
Graf, B. (1984). Theoretische und experimentelle Ermittlung des
(a) Proportional compression, and densification by alternat- Vertikaldrucks auf eingebette Bauwerke, No. 96 of Veröffentli-
ing deformations with constant pressure cause a chungem des Instituts für Boden- und Felsmechanik der Uni-
reduction of fluctuations. versität Karlsruhe.
(b) The observed evolution of shear bands in various model Gross, H. (1980). Berechnung von Erdwiderstandsbeiwerten nach
tests with dry sand is well modelled by means of der Gradientenmethode. Bauingenieur 55, 469–472.
hypoplasticity with polar quantities. Herle, I. & Gudehus, G. (1999). Determination of parameters of a
(c) Higher void ratios have a stronger spatial fluctuation: hypoplastic constitutive model from grain properties. Mechanics
this is allowed for by suitable random distributions for of Cohesive-Frictional Materials 4, No. 5, 461–486.
Huang, W. (2000). Hypoplastic modelling of shear localisation in
the onset and obtained with the calculated evolution. granular materials. PhD thesis, Graz University of Technology.
(d ) The fluctuation of deviator stress increases with the Huang, W. & Bauer, E. (2002). Numerical investigation of shear
fluctuation of void ratio. localization in a micro-polar hypoplastic material. Int. J. Numer.
(e) Polar stresses evolve together with shear bands, starting Anal. Methods Geomech. 27, 325–352.
from rough boundaries, sharp edges or spontaneously Huang, W., Nübel, K. & Bauer, E. (2002). A polar extension of
inside from small fluctuations. hypoplastic model for granular material with shear localization.
( f ) Inside shear bands the sand tends to a critical state with Mech. Mater. 34, 563–576.
maximum spatial fluctuations of all quantities. Milligan, G. W. E. (1974). The behaviour of rigid and flexible
(g) A sand body that is uniformly stretched from its retaining walls in sand. PhD thesis, University of Cambridge.
Mokni & Desrues, J. (1998). Strain localization measurements in
boundary tends towards an overall critical state with a
undrained plane-strain biaxial tests on hostun RF sand. Mech-
fractal evolution of shear band patterns. anics of Cohesive-Frictional Materials 4, 419–441.
(h) Conventional mechanisms with slip surfaces are roughly Mühlhaus, H. B. (1987). Berücksichtigung von Inhomogenitäten im
reproduced: realistic forces could thus be obtained with Gebirge im Rahman einer Kontinuumstheorie. No. 106 of Ver-
actual peak friction angles, but these are not known in öffentlichungem des Instituts für Boden- und Felsmechanik der
advance. Universität Karlsruhe.
Nübel, K. (2002). Numerical and Experimental Investigation of
A lot of further work is desirable. For sand bodies larger Shear Localization of Granular Material, No. 159 of Veröffen-
than those in the models the mesh size is far too small. tlichungem des Instituts für Boden- und Felsmechanik der Uni-
Remeshing and interface elements for the localisation zones versität Karlsruhe.
will be necessary. Then limitations and extensions of simpli- Nübel, K. & Weitbrecht, V. (2002). Visualization of localization in
fied design models can be worked out; scattering and grain skeletons with particle image velocimetry. J. Test. Eval.
uncertainty will play a key role. ASTM. 30, No. 4, 322–329.
A deeper physical understanding is achieved with our Oda, M. & Kazama, H. (1998). Microstructure of shear bands and
its relation to the mechanisms of dilatancy and failure of dense
approach, and can be widened. Strong scattering is inherent
granular soils. Géotechnique 48, No. 4, 465–481.
with granular bodies; pattern formation and deterministic Raffel, M., Willert, C. & Kompenhaus, J. (1998). Particle image
chaos can appear. Scale effects are implied: faulting of the velocimetry – a practical guide. Springer, 1998.
earth crust is also a kind of shear banding. Rankine, W. (1856). On the stability of loose earth. Phil. Trans R.
Soc. London 147, No. 1, 74ff.
Roscoe, K. H. (1970). The influence of strains in soil mechanics.
ACKNOWLEDGEMENT Géotechnique 20, No. 2, 129–170.
The work has been supported financially by the DFG Schofield, A. & Wroth, C. P. (1968). Critical state soil mechanics.
(Deutsche Forschungsgemeinschaft, German Research Coun- London: McGraw-Hill.
Schwing, E. (1991). Standsicherheit historischer Stützwande, inter-
cil), under contract No. Gu 103/52-1. nal report. Institut für Boden- und Felsmechanik der Universität
Karlsruhe.
Shahinpoor, M. (1981). Statistical mechanical considerations on
REFERENCES storing bulk solids. Bulk Solids Handling 1, No. 1, 31–35.
Alshibli, K. A., Sture, S., Costes, N. C., Frank, M. L., Lankton, M. Simpson, B. & Wroth, C. P. (1972). Finite element computations
R., Batiste, S. N. & Swanson, R. A. (1984). Assessment of for a model retaining wall in sand. Proc. 5th Eur. Conf. Soil
localized deformations in sand using X-ray computed tomo- Mech. Found. Engng, Madrid, 85–94.
graphy. Geotech. Test. J. 23, No. 3, 274–299. Sokolovsky, V. V. (1942). Statics of earthy media (Russian). Mos-
Behringer, R. P. & Miller, B. (1997). Stress fluctuations for sheared cow: Izdatelstvo Akademii Nauk.
3D granular materials. In Proc. Powders and Grains 3rd Com. Tatsuoka, F., Nakamura, S., Huang, C. C. & Tani, K. (1990).
(eds R. P. Behringer and J. T. Jenkins), pp. 333–336. Rotterdam: Strength anisotropy and shear band direction in plane strain tests
Balkema. on sand. Soils Found. 30, No. 1, 35–54.
Behringer, R. P., Geng, J., Howell, D., Longhi, E., Reydellet, G., Tejchman, J. (1989). Scherzonenbildung und Verspannungseffekte in
Vanel, L., Clement, E. & Luding, S. (2001). Fluctuations in Granulaten unter Berücksichtigung von Korndrehungen, No. 117
granular materials. In Proc. Powders and Grains 4th Com. (ed. of Veröffentlichungem des Instituts für Boden- und Felsmecha-
Y. Kishino), pp. 347–354. Rotterdam: Balkema. nik der Universität Karlsruhe.
Coulomb, M. (1773). Essai sur une application des règles des Tejchman, J. (1997). Modelling of shear localization and auto-
Maximis & minimis à quelques Problèmes des Statiques, relatifs genous dynamic effects in granular bodies, No. 140 of Veröffen-
à l’architecture. Mémoires présentés à l’Académie des Sciences tlichungem des Instituts für Boden- und Felsmechanik der
VII. Report, Paris, Editions Science et Industrie, 1971. Universität Karlsruhe.
Darwin, G. H. (1883). On the horizontal thrust of a mass of sand. Tejchman, J. (2001). Shearing of an infinite narrow granular layer
EVOLUTION OF SHEAR BANDS IN SAND 201
between two boundaries. Bifurcation and Localization theory in Viggiani, G., Mooney, M. A., Bardet, J. P. & Chambon, R. (1997).
geomechanics. (ed. H. B. Mühlhaus), pp. 139–146, Lisse: Swets A numerical investigation of local volume change and void ratio
& Zeitlinger, 2001. evolution leading to failure of granular media. Proceedings of
Tejchman, J. & Gudehus, G. (2001). Shearing of a narrow granular the International Symposium on Deformation and Progressive
layer with polar quantities. Int. J. Numer. Anal. Methods Geo- Failure in Geomechanics (eds Asaoka, Adachi and Oka), pp.
mech. 25, No. 1, 1–28. 401–406. Pergamon.
Topolnicki, M., Gudehus, G. & Mazurkiewicz, B. K. (1990). Yoshida, T., Tatsuoka, F., Siddiquee, M. S. A & Kamegai, Y.
Observed stress–strain behaviour of remoulded saturated clay (1994). Shear banding in sands observed in plane strain com-
under plane strain conditions. Géotechnique 40, No. 2, 155–187. pression. Proc. 3rd Int. Workshop on Localisation and Bifurca-
Vardoulakis, I., Goldscheider, M. & Gudehus, G. (1978). Formation tion Theory for Soils and Rocks (eds R. Chambon, J. Desrues
of shear bands on sand bodies as a bifurcation problem. Int. J. and I. Vardonlakis), pp. 165–180. Rotterdam: Balkema.
Numer. Anal. Methods Geomech. No. 2, 99–128. Zienkiewicz, O. C., Tylor, R. L. & Too, J. M. (1971). Reduced
Vesic, A. S. (1973). Bearing capacity theory from experiments: integration technique in general analysis of plates and shells.
Discussion. J. Soil Mech. Found. Engng ASCE 99, 575–577. Int. J. Numer. Methods Engng 3, 275–290.

Vous aimerez peut-être aussi