Vous êtes sur la page 1sur 11

Lings, M. L. (2001). Geotechnique 51, No.

6, 555565

Drained and undrained anisotropic elastic stiffness parameters


M. L. LINGS
Nous presentons des equations qui permettent de tracer les
parametres de rigidite elastique drainee et non drainee dans
un materiau a anisotropie croisee. Nous utilisons des graphiques representant divers taux de rigidite anisotrope pour
illustrer les tendances importantes. Un nouveau graphique
en 2 dimensions montrant la limite elastique permet de
tracer les parametres elastiques mesures qui se trouvent
contenus a l'interieur. Le graphique en 2 dimensions permet
de montrer des materiaux non couples et incompressibles
ainsi que des materiaux dont le parcours de contrainte
effective initiale a une pente particuliere dans un essai
triaxial non draine. Nous etudions aussi des limites plus
restrictives que les limites elastiques et nous presentons les
amplitudes probables de differents taux de rigidite anisotrope. En general, l'anisotropie non drainee est bien inferieure a l'anisotropie drainee, c'est-a-dire qu'elle se
rapproche de l'isotropie. Dans des conditions de deformation
plane non drainee, les modules verticaux et horizontaux se
revelent identiques. Dans des conditions de deformation
plane drainee, les amplitudes des modules sont presque les
memes que celles des modules de Young.

Equations that map drained to undrained elastic stiffness


parameters in a cross-anisotropic material are presented.
Graphs of various anisotropic stiffness ratios are used to
illustrate the important trends. A new 2-D graph showing
the elastic bound enables measured elastic parameters to be
plotted, which all lie well within it. The 2-D graph enables
uncoupled and incompressible materials to be shown, plus
materials with particular initial effective stress path slopes
in an undrained triaxial test. Bounds more restrictive than
the elastic one are also discussed, and the likely ranges of
various anisotropic stiffness ratios are presented. In general,
undrained anisotropy is much less than drained: that is,
closer to isotropic. In undrained plane strain conditions,
vertical and horizontal moduli are found to be the same. In
drained plane strain conditions, the ratios of the moduli are
almost the same as the Young's moduli.

KEYWORDS: anisotropy; elasticity; stiffness.

directions respectively. This is considered preferable to either x,


y and z, or 1, 2 and 3, where further denitions are needed to
make the intended spatial directions clear.
Poisson's ratios will be represented by rather than , as it
leads to a clearer visual differentiation from the v subscripts.
The order of the subscripts follows Pickering (1970). Thus vh
indicates a horizontal strain caused by an imposed vertical
strain.
The relationship between increments of stress and strain for a
cross-anisotropic material is described by
2
3
1
hh vh
:
:
:
6 Eh
7
Eh
Ev
6
7
6
7
6
7
1
vh
hh
7
2
3 6
:
:
:
xx
6
7
Eh
Ev
6 Eh
7
6
7
7 6
6 yy 7 6
7
6
7
7 6 vh vh
1
6
7
7 6
:
:
:
6 7 6 E
7
E
E
v
v
6 zz 7 6 v
7
6
7
76
6
7
7 6
6 yz 7 6
7
1
:
:
:
:
6
7
7 6 :
6
7
7 6
G
vh
6 zx 7 6
7
4
7
5 6
6
7
1
6
7
:
:
:
:
:
xy
6
7
6
7
Gvh
6
7
6
7
4
5
2(1

)
hh
:
:
:
:
:
Eh
2
3
xx
6
7
6 yy 7
6
7
6
7
6 7
zz
6
7
7
.6
(1)
6
7
6 yz 7
6
7
6
7
6 zx 7
4
5

INTRODUCTION

The predominantly one-dimensional strain history of many soils


means that soil stiffness is likely to be anisotropic. The magnitude of the strains for which soil behaviour is truly linear elastic
is hard to estimate, but for uncemented soil it is probably very
small (Jardine, 1992). However, within the linear elastic region,
soil behaviour is also likely to be anisotropic.
Bender elements are often used to measure elastic shear
moduli, and recent developments permit both Ghh and Gvh to be
measured on the same sample (Pennington et al., 1997, 2001).
With the advent of extremely sensitive laboratory local strain
measuring devices, there is now the prospect that elastic
Young's moduli can also be measured (Cuccovillo & Coop,
1997). Beyond the linear elastic region, soil behaviour is often
markedly non-linear, but moduli at very small strain are important anchor points when characterising the degree of nonlinearity involved (Atkinson, 2000).
About 30 years ago, when it was thought that a substantial
linear elastic region existed, much work was published on the
anisotropic elastic stiffness of soils. Now, although there is a
new awareness of the very small region of applicability of linear
elastic theory, much of this earlier work appears to have been
forgotten. It is thus timely to revisit the subject.
Notation adopted
The most appropriate form of anisotropy to describe soils
with a one-dimensional strain history is cross anisotropy. Five
independent elastic parameters are required, usually chosen to
be two Young's moduli, two Poisson's ratios and a shear
modulus. Within the soil mechanics literature, various different
and sometimes confusing notations have been adopted in the
past. Variations involve the parameter symbols, and the choice
and order of the subscripts. Here, v and h subscripts will be
used for stiffness parameters to indicate vertical and horizontal
Manuscript received 16 November 2000; revised manuscript accepted 2
April 2001.
Discussion on this paper closes 1 February 2002, for further details see
inside back cover.
 University of Bristol.

xy
where the stress and strain increments are referred to rectangular Cartesian axes, with the z-axis vertical. In this equation, the
555

556

LINGS

shear modulus in the horizontal plane of isotropy has been


replaced using the relationship
Eh
Ghh
(2)
2(1 hh )

Eh/Ev
hh = 21

and the symmetry of the elastic compliance matrix (Love, 1927)


has been invoked, whereby
hv vh

(3)
Eh
Ev
From equation (3), it is evident that the choice of independent
elastic parameters is somewhat arbitrary, as any three from the
four in equation (3) could be chosen. Here, vh will generally be
used in preference to hv , because it is the parameter obtained
from a conventional triaxial compression test. Several other
authors have chosen to work with hv , but interchange between
the two is readily accomplished using equation (3).
In all the above, no mention has yet been made of either total
or effective stress. When equations are equally valid for both, as is
the case so far, no additional superscripts will be added. However,
in most of what follows, it is crucial to distinguish drained effective stress parameters from undrained total stress parameters. The
former will be indicated by the use of an effective stress prime
(e.g. E9v ), the latter by an undrained u superscript (e.g. Euv ).
Undrained superscripts are preferred to more familiar subscripts,
both for consistency of style and to avoid too many subscripts.
Consistent notation is important, because its absence in the past
has often led to confusion: equations and graphs have often relied
on accompanying text to clarify whether drained or undrained
conditions are being described.
All parameters used in the paper are elastic. However, zero
subscripts, which are commonly used to indicate very small
(zero) strains, have been omitted for clarity.
BOUNDS ON PARAMETERS

Although all ve cross-anisotropic parameters are independent, there are bounds on the values that they can take, because
of the thermodynamic requirement that the strain energy function be non-negative in an elastic material (Love, 1927). This
leads to requirements for Ev, Eh and Gvh to all be > 0, and for
1 < hh < 1. Values of Ev , Eh , vh and hh even then are not
entirely arbitrary, but must satisfy an inequality that has been
expressed in a number of ways. Pickering (1970) expressed it
using hv as
Eh
(1 hh ) 2hv 2 > 0
(4)
Ev
Raymond (1970) expressed it using vh as
Eh
1 > hh 2 vh 2
Ev

(5)

These are both equivalent, but are more usefully expressed


(Lings et al., 2000) as
Ev
(1 hh ) 2vh 2 > 0
(6)
Eh
The similarity with, yet difference from, Pickering's (1970)
expression should be noted.
These bounds on four of the ve parameters were rst shown
graphically by Pickering (1970), who presented them as a threedimensional (3-D) graph, shown in Fig. 1. All four parameters
are included by combining the two Young's moduli, giving axes
of Eh =Ev , hv and hh . Setting the inequality in equation (4) to
zero gives the bounding surface shown, which is parabolic in
both vertical and horizontal sections. There is a vertical cut-off
wall at hh 1.
An alternative 3-D graph to Pickering's (1970) `ship's bow'
shape was presented by Hooper (1975). Replacing the hv axis
with a vh axis results in an alternative `tailplane' shape shown
in Fig. 2. Again there is a vertical cut-off wall at hh 1. An
identical graph was presented by Uriel (1973), but this is not
referenced in Hooper (1975).

05
05
hh = 1

1
Eh/Ev = 05
10

hh

0
0

hv

10
10

20

Fig. 1. Bounds on elastic parameters (from Pickering, 1970)

Special cases
One major advantage of these graphical representations is
that a number of special cases can be identied and visualised.
This approach was rst taken by Pickering (1970), from which
Fig. 3 is taken. There are two planes of particular interest, to
which further reference will be made later. The rst is a plane
representing all uncoupled materials: that is, materials that
undergo no distortional strain with isotropic loading (i.e. constant deviator stress), nor any volumetric strain with deviatoric
loading (i.e. constant mean normal stress). This plane is shown
in Fig. 3 as the triangle ABC, each of whose sides lies on the
bounding surface. The equation of this plane is given by
Pickering (1970) as
Eh
1 hv hh
(7)
Ev
Within this plane there is a line CD, which represents isotropic
materials. It is important to note that, whereas all isotropic
materials are uncoupled, only a small subset of uncoupled
materials is isotropic.
The second is a plane representing all incompressible materials, the term used by Pickering (1970) specically to denote
materials that undergo no volumetric strain with isotropic loading. The equation of this plane is given by Pickering (1970) as
Eh
2 4hv 2hh 0
(8)
Ev
It is tangent to the bounding surface on line AB, but otherwise
lies outside the region of permissible elastic parameters. Thus
incompressible elastic materials lie on line AB.
UNDRAINED CONDITIONS

Ideal undrained elastic materials (i.e. where both soil grains


and water are considered to be incompressible) deform at
constant volume. The signicance of the line AB in Fig. 3 is
that all such undrained materials must lie on it, with axes in
terms of undrained total stress parameters. Therefore, as the line

ELASTIC STIFFNESS PARAMETERS


Eh/Ev

557

4
Incompressible material
3

Isotropic material

1
03

205

vh

hh

Fig. 2. Bounds on elastic parameters (from Hooper, 1975)

Eh/Ev
A

points within the permitted space can be mapped onto the


undrained line, but points on the undrained line can be reached
from an innite number of drained points within the space. The
mapping is thus one-way, from drained to undrained.
Relationships between undrained parameters
Before considering this mapping, it is helpful to clarify the
relationships between the undrained elastic parameters. This was
rst explored in detail in Gibson's (1974) Rankine lecture,
where two relations connecting the undrained elastic parameters
were given
1 Euh
2 Euv

(9)

uhh 1

(10)

uhv
C

Eh/Ev = 1

D
10

0
B
0

hh

hv

10
10

20

Fig. 3. Planes and lines representing special types of material (from


Pickering, 1970)

represents the intersection of the two planes described above, it


is clear that all ideal undrained elastic materials are both
incompressible (in the sense used here) and uncoupled. They
will therefore not distort under isotropic loading.
Figure 3 can have drained or undrained parameter axes. All
combinations of drained elastic parameters must plot somewhere
within the `ship's bow' shape; all combinations of undrained
elastic parameters must plot on the line AB. It will be shown
that a mapping exists whereby any set of drained parameters
can be converted to a set of undrained parameters. All drained

1 Euh
2 Euv

shown here with undrained superscripts. Undrained conditions


therefore reduce the number of independent parameters from
ve to three (usually Euv, Euh and Guvh ). Further considerations of
strain energy led Gibson (1974) to bounds on the Euh =Euv ratio
0<

Euh
<4
Euv

(11)

Combining these bounds with equation (10) leads to the result


1 < uhh < 1

(12)

which is also evident from line AB in Fig. 3. Equations that are


essentially the same as equations (9) and (10) have also been
given independently by Pickering (1970), Chowdhury & King
(1971) and Atkinson (1975). Chowdhury & King (1971) appear
to have been the rst to express equation (9) in the form
uvh 0:5
(13)
which can be obtained from a combination of equation (9) with
equation (3) in its undrained form. This equation was also given
by Hooper (1975), although Fig. 2 shows, incorrectly, a shaded
vertical triangular plane as representing incompressible materials (here meaning no volume change under any type of loading). They are actually only represented by the straight line on
the bounding surface. (All such materials have vh 0:5, but

558

LINGS

not all materials with vh 0:5 are incompressible: vh 0:5


is necessary but not sufcient.) Rearranging equation (10) in
the form
Euh
2(1 uhh )
Euv

(14)

enables the relationship to be plotted as shown in Fig. 4. It is


clear that the bounds given by equations (11) and (12) are
synonymous.
The ratio of the horizontal to vertical Young's moduli will be
given the symbol R, with standard superscripts to distinguish
between drained and undrained. Thus
R9

E9h
Eu
; Ru hu
E9v
Ev

(15)

This ratio is the one most commonly adopted, and is preferred


to its reciprocal because it leads to more manageable bounds in
equation (11) and Fig. 4.
EQUATIONS MAPPING DRAINED TO UNDRAINED PARAMETERS

By using the rst three rows of equation (1) in their effective


stress form, and considering the strains in various types of ideal
undrained triaxial test, it is possible after considerable manipulation to derive the following equations for undrained Young's
moduli in terms of the four drained parameters E9v , E9h , 9vh and
9hh
Euv
Euh

E9v [2(1 9hh )E9v (1 49vh )E9h ]


2(1 9hh )E9v 49vh 2 E9h

(16)

E9h [2(1 9hh )E9v 2 (1 49vh )E9v E9h ]


(1 9hh 2 )E9v 2 (1 29vh 29vh 9hh )E9v E9h 9vh 2 E9h 2
(17)

These two equations (plus a value for Guvh ) are sufcient to


describe fully the ideal undrained elastic behaviour. However,
other relationships can now easily be derived. Substitution of
equations (16) and (17) into equation (10) leads to
uhh

(1 9hh 2 )E9v 2 (9hh 29vh 29vh 9hh )E9v E9h 9vh 2 E9h 2
(1 9hh 2 )E9v 2 (1 29vh 29vh 9hh )E9v E9h 9vh 2 E9h 2
(18)

Substitution of equations (13), (16) and (17) into equation (3)


gives

Elastic
bound

35

Similarly, the ratio of undrained Young's moduli can be expressed as


Ru

(20)
Substitution using equations (17) and (18) conrms, as expected, that
Guhh

E uh/E uv

Euh
E9h

G9hh
2(1 uhh ) 2(1 9hh )

(21)

and for completeness, it should be noted that


Guvh G9vh

(22)

Equations mapping drained to undrained parameters have


been published before, although unknown to the author until
after the derivations above had been carried out. Uriel & Canizo
(1971) were the rst to publish such equations, and Bishop &
Hight (1977) derived more complete equations that involve
Skempton's (1954) B parameter, thereby enabling conditions
other than ideal undrained to be considered. The forms of these
two sets of equations are rather different from those presented
here, but careful checking shows that all three sets are equivalent. The forms adopted here, with undrained parameters expressed solely in terms of the four drained parameters E9v , E9h ,
9vh and 9hh, are considered to be more user-friendly.
Graphical representations
With the exception of the shear moduli, all the undrained
parameters above are functions of four independent variables,
but a simplication can be made by substituting R9 from equation (15). Then, by introducing U , representing the ratio of
undrained to drained Young's moduli, whereby
Euv
;
E9v

Uh

Euh
E9h

(23)

In the same way, equation (20) can be simplied to give


Ru

15

Isotropic
point

Euh
2(1 9hh )R9 49vh 2 R92
u
Ev (1 9hh 2 ) (1 29vh 29vh 9hh )R9 9vh 2 R92
(26)

It is worth noting in passing that the R and U ratios are


interrelated, since

05
Elastic
bound

0
025

(1

2(1 9hh )E9v E9h 49vh 2 E9h 2


(1 29vh 29vh 9hh )E9v E9h 9vh 2 E9h 2

9hh 2 )E9v 2

Euv 2(1 9hh ) (1 49vh )R9

(24)
2(1 9hh ) 49vh 2 R9
E9v
Eu
2(1 9hh ) (1 49vh )R9
Uh h
E9h (1 9hh 2 ) (1 29vh 29vh 9hh )R9 9vh 2 R92
(25)

050

Euh
Euv

Uv

25

075

(1 9hh )E9v E9h 29vh 2 E9h 2


(1 29vh 29vh 9hh )E9v E9h 9vh 2 E9h 2

it is possible to express equations (16) and (17) for the


undrained Young's moduli as

(1

9hh 2 )E9v 2

(19)

Uv
4

uvh Euh
Euv

uhv

0
uhh

025

05

075

Fig. 4. Permissible undrained values of uhh and Euh =Euv

U h Euh E9v Ru

U v E9h Euv R9

(27)

Equations (24) to (26) have been plotted in Figs 5 to 7 as


families of two-dimensional (2-D) graphs. The ordinate has

ELASTIC STIFFNESS PARAMETERS

559

Uv = E uv /E v

2
hh = 035

hh = 035
1

hh = 005

hh = 005

0
(a)

(b)

4
hh = 035
Uv = E uv /E v

3
hh = 035

hh = 005
hh = 005

0
0

01

02
vh
(c)

03

04

01

Euv =E9v . R9

Fig. 5. Ratio of undrained/drained vertical Young's moduli, Uv

02
vh
(d)

E9h =E9v

03

04

(a) 05, (b) 1, (c) 2, (d) 4

Uh = E uh/E h

hh = 005

hh = 005

hh = 035

hh = 035

0
(a)

(b)

Uh = E uh/E h

Elastic
bound

2
hh = 005
1

hh = 035

hh = 005

hh = 005

hh = 035

hh = 035

0
0

01

02
vh
(c)

03

04

Fig. 6. Ratio of undrained/drained horizontal Young's moduli, Uh

01

Euh =E9h . R9

02
vh
(d)

E9h =E9v

03

04

(a) 05, (b) 1, (c) 2, (d) 4

560

LINGS
4

R u = E uh /E uv

2
hh = 005

hh = 005

hh = 035

hh = 035
0
(a)

(b)

R u = E uh /E uv

hh = 005

hh = 005

hh = 035

hh = 035

Elastic
bound
0
0

01

02
vh
(c)

03

04

Fig. 7. Ratio of horizontal/vertical undrained Young's moduli, Ru

been chosen as 9vh , incremental values of 9hh dene each curve


in steps of 005 between the limits given, and each quadrant of
the graph represents a different value of R9. The ranges adopted
for each parameter are based on the spread observed within
measured data, as presented later.
The three sets of graphs show various trends in the parameters. The ratio of undrained to drained vertical Young's
moduli (Fig. 5) has a minimum value of 1, and, in theory, an
unbounded maximum value. At low values of R9, Euv is more
sensitive to the value of 9vh than of 9hh . At high values of R9,
the inuence of the two Poisson's ratios is more nearly equal.
But, for the range presented, the value of Euv is most sensitive
to the values of E9v and E9h .
The ratio of undrained to drained horizontal Young's moduli
(Fig. 6) is very different. The minimum value is again 1, but
the theoretical maximum is now affected by the combination of
drained parameters meeting the elastic bound (equation (6)).
The elastic limit is shown in Fig. 6(d), so values are most
unlikely to exceed two. In fact, with R9 as high as four, Uh
may be only just above unity. Generally, as R9 increases, Uh
decreases. At low values of R9, Euv is very sensitive to the value
of 9hh and very insensitive to the value of 9vh . At high values
of R9, it is insensitive to both.
The ratio of undrained Young's moduli (Fig. 7) has an
absolute minimum value of zero when the elastic bound is
reached, as shown in Fig. 7(d), and a likely maximum of 2
unless 9hh becomes negative. In fact, the absolute maximum of
4 (equation (11)) can be achieved only if 9hh has its limiting
value of 1. In theory at least, Fig. 7(d) shows it might be
possible to have R9  1, yet Ru , 1.
TWO-DIMENSIONAL PARAMETER PLOTS

The 3-D plots pioneered by Pickering (1970) are useful in


permitting important equations and concepts to be portrayed
graphically. They are inevitably complex to work with, and 2-D
plots would be very much easier.

01

Euh =Euv . R9

02
vh
(d)

E9h =E9v

03

04

(a) 05, (b) 1, (c) 2, (d) 4

Conventional triaxial testing


A great deal of soil testing has been performed in the triaxial
apparatus, and because of its axisymmetric nature, it replicates
the conditions necessary for cross anisotropy. Once stress conditions depart from axisymmetry, it is unlikely that perfect cross
anisotropy will remain. In conventional triaxial testingthat is,
tests on vertical samplesit is possible to measure only three
independent parameters (Graham & Houlsby, 1983). These three
parameters can be chosen in a number of ways, but Lings et al.
(2000) have suggested that there are advantages in using Ev ,
vh and Fh, where Fh is given by
Eh
Fh
(28)
(1 hh )
Combining equations (28) and (6) leads to the bound on elastic
parameters being expressed as
Ev
> 2vh 2
(29)
Fh
By using the ratio Ev =Fh as one of the axes, permitted
combinations of elastic parameters can be represented as the
area above the parabola in Fig. 8. This is in effect a 2-D
version of the Pickering (1970) plot, since Fig. 1 would remain
completely unchanged if the axes of Eh =Ev and hv were
replaced by Ev =Eh and vh (compare equations (4) and (6)).

Special cases
Once the axes in Fig. 8 have been adopted, various special
cases can be shown. Uncoupled materials described by equation
(7) are represented by
Ev
1 vh
(30)
Fh
and incompressible materials described by equation (8) are
represented by

ELASTIC STIFFNESS PARAMETERS

with the inverse expression, as proposed by Doran et al. (2000).


The slope is therefore

2
Uncoupled
175

p9
J

q 3G

15

Ev /Fh

075
Undrained
point

p9
2E9v =F9h 2(1 9vh )

q 6E9v =F9h 3(1 49vh )

05

025
Incompressible
0

075

05

025

0
vh

025

05

075

Fig. 8. 2-D plot (vh against Ev =Fh ) showing elastic bound and
special cases

Ev
2vh 0:5
Fh

(31)

In both cases the planes from Fig. 3 become straight lines in


Fig. 8. Because isotropic now looks the same as uncoupled, it is
clear that conventional triaxial tests are unable to distinguish
between isotropic and uncoupled materials, and can only deduce
the latter.
Ideal undrained materials are represented by the point shown
in Fig. 8, which is the intersection of the uncoupled and
incompressible lines. This point is equivalent to the line AB in
Fig. 3. Thus all drained materials, which must lie within the
elastic bound, map to a unique undrained point on this plot in
terms of total stress. The condition for ideal undrained behaviour given in equation (14) becomes
F uh 2Euv

or

Euv
0:5
F uh

(35)

For any particular value of ESP slope, equation (35) leads to a


simple linear equation for E9v =F9h in terms of 9vh . Fig. 9 shows
various lines of constant ESP slope, which are all found to pass
through the point with coordinates (05, 0.5)the undrained
point in total stress terms. The line representing zero ESP slope
is the same as the uncoupled line, as expected, and the line
representing innite ESP slope is the same as the incompressible line.
By showing lines of constant ESP slope in this way, it
becomes clear that the limits placed on the slope value by
Graham & Houlsby (1983) of 2=3 and 1=3 (3=2 and 3
in terms of q=p9) are not in fact absolute limits.
Measured elastic parameters
The amount of published data giving all ve drained crossanisotropic elastic parameters is still rather small. However,
such data can be plotted on the 2-D graph, although only four
of the ve parameters are involved: the independent shear
modulus plays no part. The readily available data are plotted in
Fig. 10.
The most comprehensive data are taken from Bellotti et al.
(1996), who tested dry Ticino sand in a large calibration
chamber. Anisotropic parameters were obtained from dynamic
measurements of compression and shear wave velocities. From
the large amount of data provided, only tests at a vertical

(32)

in addition to uvh 0:5 (equation (13)). The existence of a


unique undrained point on these axes shows that undrained
stiffness anisotropy cannot be discerned in conventional triaxial
tests from total stress measurements. There is only one independent undrained parameter that can be measured, which will
probably be Euv .

1/
3

4/
3

175
15

125
5/
3

2/
3

075

05

1/
6

P
025
Bound on elastic
parameters

(33)

where G and J are a shear and a coupling modulus respectively. Because J becomes zero when the material is isotropic
(strictly, when it is uncoupled ), it is more convenient to deal

2
0

E v /F h

Slope of undrained effective stress path


Several authors who have investigated anisotropic elastic
parameters have also considered the effect of these parameters
on the slope of the effective stress path (ESP) in undrained
triaxial tests (e.g. Henkel, 1971; Atkinson, 1975; Graham &
Houlsby, 1983). As parameters are thought to be elastic only at
very small strains, the relevant part of the undrained ESP is the
initial portion, after either an extended holding period or a
sharp change in the direction of the stress path (Atkinson et al.,
1990). The simplest expression for the slope of the ESP in a
p9 : q plot is given by Graham & Houlsby (1983)
q 3G

J
p9

(34)

measured with respect to the vertical. If the ESP moves upwards


and to the right, the value is positive; if it moves upwards and
to the left, the value is negative. If it moves vertically, the slope
expression is zero.
Although this is the simplest expression, involving only two
parameters, the slope can also be expressed in terms of the
three parameters E9v , 9vh and F9h, using the equations for G
and J given by Lings et al. (2000). This leads to

125

Bound on elastic
parameters

561

0
1

075

05

025

O
0
vh

025

05

075

Fig. 9. 2-D plot (9h against E9vh =F9h ) showing lines of constant ESP
slope ( p9=q)

562

LINGS
Uv =
Uv = 3
Uv = 15
Uv = 105
Uv = 10

Ticino Sand (K = 05)


Ticino Sand (K = 1)
Ticino Sand (K = 2)
Ticino Sand (K = 15)
Ham River Sand (K = 045)
Natural Gault Clay (K = 2)

2
Uncoupled

175
175
15
15

E v /F h

125

E v /F h

125

1
075
075
05

05
025
Bound on elastic
parameters

025
Bound on elastic
parameters
1

075

05

0
025

0
vh

025

05

075

Fig. 10. 2-D plot (9vh against E9v =F9h ) showing measured elastic
parameters

effective stress of 50 kPa have been presented in Fig. 10. The


main factor affecting where the data plot is the stress ratio
K ( h9 = v9 ). Because of the normalisation inherent in the
graph, other data at the same stress ratio plot in fairly similar
positions.
The data for Ham River sand have been taken from Kuwano
& Jardine (1998), who used a combination of static (using local
strain devices) and dynamic (using bender elements) measurements in a triaxial sample to obtain the ve elastic parameters.
The reported values were for a test at a vertical effective stress
of 316 kPa.
The data for natural Gault clay have been taken from Lings
et al. (2000), who, in a parallel study, also used the same
combination of static and dynamic measurements in a triaxial
sample. The in-situ vertical effective stress was around 120 kPa.
All the data plot in the right-hand half of the gure, and well
away from the elastic bound. As K and h9 increase, so F9h
increases, leading to a decrease in E9v =F9h . At the same time
there is a general decrease in 9vh .
LIKELY PRACTICAL LIMITS ON PARAMETERS

Thus far, discussion has been restricted to the limits that


exist on elastic parameters from thermodynamic considerations.
The possibility that the practical range of values is more
restricted will now be explored.
The new 2-D plot can also be used to show one of the ratios
presented earlier in Fig. 5. The ratio of undrained to drained
vertical Young's moduli can be rewritten, by rearranging equation (16), as
Uv

Euv 2E9v =F9h (1 49vh )

2E9v =F9h 49vh 2


E9v

(36)

For any particular value of Uv , equation (36) can be rearranged


to give E9v =F9h in terms of a quadratic expression for 9vh. This
leads to a family of parabolas, as shown in Fig. 11.
When Uv has its minimum value of 1, the parabola collapses
to a single vertical line at 9vh 0:5. When it has its maximum
value of innity, the parabola becomes identical to the elastic

075

05

0
025

O
0
vh

025

05

075

Fig. 11. 2-D plot (9vh against E9v =F9h ) showing lines of constant
Uv Euv =E9v

bound. Intermediate parabolas lie between these two extremes.


In all cases, the parabolas pass through the point (05, 05). For
a particular value of E9v , the parabolas show lines of constant
Euv . Note that the axes of this gure are in terms of effective
stress: each parabola indicates an innite number of combinations of the drained parameters that all lead to the same value
of the undrained vertical Young's modulus.
Comparison of Fig. 11 with Fig. 10 shows that the implied
range of values for the ratio Uv , based on drained measurements, is between 1 and 3. This upper value of 3 is simply the
highest implied so far, and should not be regarded as a limit.
Revisiting Fig. 6 for a likely range of values for the ratio Uh ,
note that the measured data show a general reduction in
Poisson's ratios as R9 increases (i.e. as the soil becomes stiffer
horizontally). Consequently, maximum values of Uh are likely
to occur at low values of R9, and are unlikely to exceed 15.
The absolute lower limit is 1, so the likely range is 1015.
As far as the ratio of undrained parameters (Ru ) is concerned,
Fig. 7 suggests a lower limit of around 05, because high values
of R9 are associated with low values of Poisson's ratio. Thus
limits on Ru might be 0520. It is interesting to compare this
with values of the undrained ratio Ru for a wide range of clays
presented by Lee & Rowe (1989). With one exception, the
results all lie within the 0520 range suggested here. Although
the reported modulus values would have been obtained without
the benet of high-resolution local strain devices, this is still
regarded as encouraging conrmation.
The likely ranges discussed above are brought together in
Table 1, along with values of the various ratios derived from
measured drained elastic parameters. When viewed in tabular
form, various trends emerge. First, where there is signicant
drained anisotropy, the degree of undrained anisotropy (Ru ) is
always signicantly less than the drained (R9): for R9 greater
than 1, Ru , R9; for R9 less than 1, Ru . R9. Second, the ratios
of undrained to drained parameters are markedly different for
the vertical and horizontal directions:
(a) Uv has a much wider range of values than Uh .
(b) Uv ratios increase as soil becomes stiffer horizontally (R9
increasing).

ELASTIC STIFFNESS PARAMETERS

563

Table 1. Range and example values of anisotropic elastic parameter ratios


Ratio of parameters
Uv Euv =E9v
Uh Euh =E9h
Ru Euh =Euv
R9 E9h =E9v
a

Likely range

HRS
(K 0:45)

Ticino sand
(K 0:5)

Ticino sand
(K 1:0)

Ticino sand
(K 1:5)

Ticino sand
(K 2:0)

Gault clay
(K 2:0)

1030a
1015
0520
0540a

103
148
075
052

127
132
086
083

131
123
114
121

157
121
120
156

165
115
133
190

291
122
166
397

Based solely on available data

(c) Uh ratios decrease as soil becomes stiffer horizontally (R9


increasing).
Other limits on parameters
Some previous authors have argued for the existence of more
restrictive bounds on drained elastic parameters than those
imposed purely from thermodynamic considerations. Three
separate physical descriptions have been used. First, Chowdhury
& King (1971, 1972) suggested that an elastic soil should not
dilate under any type of compressive loading, provided all
principal stresses are of the same sign. The two equations they
presented to dene this limiting condition are
9vh , 0:5
(37)
E9h
9hh , 1
9vh
E9v

(38)

Equation (38) can be rearranged, using equation (28), to give


E9v
. 9vh
F9h

(39)

The limits implied by equations (39) and (37) can be shown


on the 2-D plot, and are represented by the lines OP and PQ
respectively in Figs 9 and 11. Non-dilatant soils (using Chowdhury & King's denition) would lie above line OP, and to the
left of line PQ. Line OP represents materials that undergo zero
volume change under radial compression with constant axial
effective stress; line PQ represents a similar case under axial
compression with constant radial effective stress. The area to
the right of PQ is equivalent to Pickering's (1970) shaded
volume in Fig. 3, representing dilatant soils under axial compression.
Second, Graham & Houlsby (1983) concluded that the limiting slopes of the ESP in undrained triaxial compression tests
were 2=3 and 1=3 (expressed as p9=q). As shown in
Fig. 9, these are also lines OP and PQ. The limits given by
Graham & Houlsby (1983) represent ESPs with constant vertical effective stress and constant horizontal effective stress
respectively (Muir Wood, 1990): the effective stress increments
never have the same sign.
Third, Muir Wood (1990) showed that the Graham & Houlsby (1983) limits also mean that, under increasing isotropic
effective stress, all principal strains will be positive. The lines
OP and PQ correspond to volumetric compression with constant
horizontal strain and constant vertical strain respectively. The
same observation was also made earlier by Uriel & Canizo
(1971).
From the equations and graphs presented here, a fourth
description emerges, based on the minimum values of the
undrained/drained ratios of the Young's moduli. If equations
(24) and (25) are partially differentiated w.r.t. 9vh , it is found
that both have minima along lines OP and PQ. Fig. 11 shows
that Uv has minima (at constant E9v =F9h ) along line OP, and an
absolute minimum of Uv 1 along line PQ. In a similar way,
Uh has minima along line PQ, and an absolute minimum of
Uh 1 along line OP. These trends can also be seen in Figs 5
and 6. The maxima for Ru seen in Fig. 7(d) also occur along
line OP.
The minimum values of Uv 1 and Uh 1 represent Euv
E9v and Euh E9h respectively. The fact that these occur along

lines PQ and OP means that soils with such parameters undergo


no initial volume change in conventional drained tests (as
described above). This is why there is no difference between
drained and undrained stiffness.
Although the graphs presented here provide some new insights into behaviour, it is still not clear whether the limit lines
OP and PQ are anything more than an appeal to what is
considered reasonable for soil behaviour. However, it is interesting that all four approaches lead to the same limits. The data in
Fig. 10 all plot within these more restrictive bounds, and time
will tell whether other soils will be found that extend the range
of observed behaviour beyond lines OP and PQ.
PLANE STRAIN

Many forms of construction in geotechnical engineering


approximate much more closely to conditions of plane strain
than to axisymmetry. The cross-anisotropic parameters adopted
here are two Young's moduli, two Poisson's ratios and a shear
modulus. Others have worked with two constrained moduli, two
shear moduli, and a fth independent term from the stiffness
matrix (e.g. Bellotti et al., 1996). The constrained moduli,
usually denoted M v and M h, represent deformation under
uniaxial stress with zero strain in the two orthogonal directions.
In a similar way, a plane strain modulus can be dened,
denoted here by N, which represents deformation under uniaxial
stress with zero strain in just one orthogonal direction. By
considering the rst three rows of equation (1), and imposing
zero strain in one of the horizontal directions, the two plane
strain moduli can be derived
Ev
Nv
(40)
1 vh 2 (Eh =Ev )
Eh
Nh
(41)
1 hh 2
which are fairly similar to the plane strain modulus for an
isotropic soil
E
N
(42)
1 2
These equations are general, applying to both effective and total
stress parameters. The vertical and horizontal plane strain
moduli depend on just three and two of the cross-anisotropic
parameters respectively.
An interesting situation arises when equations (40) and (41)
are written in their undrained form. Substitution for the Poisson's ratios using equations (10) and (13) leads to
N uv N uh

Euv
1 0:25(Euh =Euv )

(43)

The undrained plane strain moduli are therefore the same in


both vertical and horizontal directions. This result has been
noted before by Henkel (1971) and Atkinson (1975), although
not in the terms presented here. The consequence is that, in
undrained plane strain stituations, there are only two independent parameters (Guvh and N u ) that control behaviour, compared
with three in axisymmetric situations. N u is affected by both Euv
and Euh, but changes in their value affect N uv and N uh equally.
Based on equation (43), N u is inuenced slightly more by Euv

564

LINGS

than by Euh. If the ratio Euh =Euv were to reach its absolute upper
limit of four from equation (11), the soil clearly would become
rigid (Gibson, 1974).
From equation (43), the likely consequences of introducing
anisotropy into an undrained plane strain problem can be
assessed. Four of the drained cross-anisotropic parameters (E9v ,
E9h , 9vh and 9hh ) will simply inuence the magnitude of N u ,
and hence the overall magnitude of the deformations. The
remaining parameter, Gvh , will be the one that affects distortions, and relative vertical and horizontal deformations. In
undrained analyses of tunnels, Lee & Rowe (1989) found
settlement proles to be strongly inuenced by Gvh, but hardly
inuenced at all by the ratio Euh =Euv . In more recent analyses of
tunnels in undrained London clay, Simpson et al. (1996) also
found Gvh to have a signicant effect on the width of the
settlement trough.
The ratio of plane strain parameters in their drained form can
be obtained from equations (40) and (41)
N h9 R9(1 9vh 2 R9)

1 9hh 2
N v9

(44)

When the Poisson's ratios are small, this reduces to


N h9
E9h
 R9
N v9
E9v

(45)

Even when Poisson's ratios are larger, this approximation is still


quite reasonable. Thus, in drained conditions, the ratio of the
horizontal/vertical plane strain moduli is almost the same as the
ratio of Young's moduli. This is in marked contrast to the
undrained case.
DISCUSSION

The validity of the derived relationships, which enable conversion from drained to undrained parameters, still needs to be
conrmed from actual measurements of both sets of parameters
at very small strains. Equivalent equations by Bishop & Hight
(1977), which include Skempton's (1954) B, suggest that,
provided B . 0:9, there will be little difference between ideal
undrained and actual undrained behaviour.
In conventional triaxial testing, Eh and hh cannot be measured. Some previous authors (e.g. Henkel, 1971) have therefore
resorted to making assumptions about the value of hh , usually
in terms of the other Poisson's ratios. From the relationships
and graphs presented here, it is clearly necessary to measure
9hh as an independent parameter, because it can be quite
inuential in certain cases, for example affecting Euv at high
values of R9, and Euh at low values of R9.
Direct measurement of Eh and hh requires either horizontal
subsampling and subsequent triaxial testing, or the use of a true
triaxial apparatus/cubical cell. The latter are more versatile, as
measurements can be made from various anisotropic initial
stress conditions. If initial stress states are anything other than
isotropic, triaxial tests on horizontal samples will involve departure from conditions of perfect cross-anisotropy.
SUMMARY AND CONCLUSIONS

Equations that convert drained effective stress elastic stiffness


parameters to undrained total stress elastic stiffness parameters
have been developed. Apart from the shear moduli (and
uvh 0:5), all undrained parameters are functions of the four
drained parameters E9v , E9h , 9vh and 9hh . Undrained and drained
elastic stiffness parameters, and ratios between them, are different and must not be confused with each other. It is recommended that any notation adopted always clearly distinguishes
between them.
The mapping from drained to undrained parameters is oneway, since undrained elastic behaviour is fully described by only
three parameters. The relationships can be represented graphically, which makes it clear that ratios of undrained/drained
vertical Young's moduli are generally different from the hori-

zontal ratios. The vertical ratios (Uv Euv =E9v ) cover a wider
range than the horizontal ratios (Uh Euh =E9h ), and trend in
opposite directions with increasing stress ratio. Vertical ratios
particularly can be very different from values normally encountered with isotropic materials.
Ratios of undrained Young's moduli (Ru Euh =Euv ) are usually rather different from ratios of drained Young's moduli
(R9 E9h =E9v ). In general, using this measure, undrained anisotropy is much less than drainedthat is, more nearly isotropic.
The undrained ratio has an absolute upper bound of 4, whereas
the drained ratio is unbounded.
A new 2-D graph has been introduced, on which the elastic
bound on parameters, special cases such as uncoupled and
incompressible behaviour, and a unique undrained point, can be
shown. It enables measured elastic parameters to be plotted, and
can be used to relate particular combinations of parameters to
particular initial effective stress path slopes in undrained triaxial
tests.
From a parametric study of the inuence of different drained
parameters on the undrained ones, and using available values of
measured drained elastic parameters, likely limits for the magnitude of anisotropic parameter ratios have been presented. They
are signicantly more restrictive than the elastic bounds derived
from thermodynamic considerations.
Various physical descriptions of behaviour, representing
apparently reasonable limits, have been compared and found to
be equivalent. Again, all are signicantly more restrictive than
the elastic bounds. One example is the limits placed on the
initial effective stress path slope in an undrained triaxial test by
Graham & Houlsby (1983). They turn out to be likely practical
limits, rather than elastic limits. It is not known whether these
more restrictive limits are valid, but the measured data all lie
within them.
Plane strain conditions have also been investigated, and it has
been shown that, in undrained conditions, vertical and horizontal plane strain moduli are the same. The only other independent parameter is the shear modulus Gvh , which will have a
controlling inuence on deformations in such conditions. In
drained conditions, the ratio of horizontal/vertical plane strain
moduli is almost identical to the ratio of Young's moduli.
ACKNOWLEDGEMENTS

The author is grateful to Dr Brian Simpson for several


helpful discussions regarding undrained plane strain anisotropy.

NOTATION
Eh
Ev
Fh
G
Ghh
Gvh
J
K
Mh
Mv
Nh
Nv
p9
q
R
Uh
Uv
x y
yz , zx
xx , yy
zz
hh
hv
vh
9h
9v
xx , yy

horizontal Young's modulus


vertical Young's modulus
horizontal modulus ( Eh =(1 hh ))
shear modulus parameter (Graham & Houslby, 1983)
horizontal shear modulus
vertical shear modulus
coupling modulus parameter (Graham & Houlsby, 1983)
stress ratio ( 9h = 9v )
horizontal constrained modulus
vertical constrained modulus
horizontal plane strain modulus
vertical plane strain modulus
mean effective stress
deviator stress
ratio of horizontal to vertical Young's moduli
ratio of undrained/drained horizontal Young's moduli
ratio of undrained/drained vertical Young's moduli
shear strain in horizontal plane
shear strain in vertical plane
horizontal direct strain
vertical direct strain
Poisson's ratio (horizontal to horizontal)
Poisson's ratio (horizontal to vertical)
Poisson's ratio (vertical to horizontal)
effective horizontal stress
effective vertical stress
horizontal stress

ELASTIC STIFFNESS PARAMETERS


zz
xy
yz , zx
9
Superscript u

vertical stress
shear stess in horizontal plane
shear stress in vertical plane
indicates drained effective stress parameter
indicates undrained total stress parameter

REFERENCES
Atkinson, J. H. (1975). Anisotropic elastic deformations in laboratory
tests on undisturbed London Clay. Geotechnique 25, No. 2, 357
374.
Atkinson, J. H. (2000). Non-linear soil stiffness in routine design.
Geotechnique 50, No. 5, 487507.
Atkinson, J. H., Richardson, D. & Stallebrass, S. E. (1990). Effect of
recent stress history on the stiffness of overconsolidated soil. Geotechnique 40, No. 4, 531540.
Bellotti, R., Jamiolkowski, M., Lo Presti, D. C. F. & O'Neill, D. A.
(1996). Anisotropy of small strain stiffness in Ticino sand. Geotechnique 46, No. 1, 115131.
Bishop, A. W. & Hight, D. W. (1977). The value of Poisson's ratio in
saturated soils and rocks stressed under undrained conditions. Geotechnique 27, No. 3, 369384.
Chowdhury, R. N. & King, G. J. W. (1971). Discussion: Anisotropic
elastic parameters for soil. Geotechnique 21, No. 2, 181183.
Chowdhury, R. N. & King, G. J. W. (1972). Discussion: Anisotropic
elastic parameters for soil. Geotechnique 22, No. 1, 183185.
Cuccovillo, T & Coop, M. R. (1997). The measurement of local axial
strains in triaxial tests using LVDTs. Geotechnique 47, No. 1, 167
171.
Doran, I. G., Sivakumar, V., Graham, J. & Johnson, A. (2000). Estimation of in situ stresses using anisotropic elasticity and suction measurements. Geotechnique 50, No. 2, 189196.
Gibson, R. E. (1974). The analytical method in soil mechanics. Geotechnique 24, No. 2, 115140.
Graham, J. & Houlsby, G. T. (1983). Anisotropic elasticity of a natural
clay. Geotechnique 33, No. 2, 165180.
Henkel, D. J. (1971). The relevance of laboratory-measured parameters
in eld studies. In Stressstrain behaviour of soils (ed. R. H. G.
Parry), Proc. Roscoe Memorial Symposium, Cambridge, pp. 669
675. Henley-on-Thames: G. T. Foulis.
Hooper, J. A. (1975). Elastic settlement of a circular raft in adhesive

565

contact with a transversely isotropic medium. Geotechnique 25,


No. 4, 691711.
Jardine, R. J. (1992). Some observations on the kinematic nature of soil
stiffness. Soils Found. 32, No. 2, 111124.
Kuwano, R. & Jardine, R. J. (1998). Stiffness measurements in a stresspath cell. In Pre-failure deformation behaviour of geomaterials (eds
R. J. Jardine, M. C. R. Davies, D. W. Hight, A. K. C. Smith & S. E
Stallebrass), pp. 391394. London: Thomas Telford.
Lee, K. M. & Rowe, R. K. (1989). Deformations caused by surface
loading and tunnelling: the role of elastic anisotropy. Geotechnique
39, No. 1, 125140.
Lings, M. L., Pennington, D. S. & Nash, D. F. T. (2000). Anisotropic
stiffness parameters and their measurement in a stiff natural clay.
Geotechnique 50, No. 2, 109125.
Love, A. E. H. (1927). A treatise on the mathematical theory of
elasticity, 4th edn (1st edn 1892). Cambridge: Cambridge University
Press.
Muir Wood, D. (1990). Soil behaviour and critical state soil mechanics.
Cambridge: Cambridge University Press.
Pennington, D. S., Nash, D. F. T. & Lings, M. L. (1997). Anisotropy of
G0 shear stiffness in Gault Clay. Geotechnique 47, No. 3, 391398.
Pennington, D. S., Nash, D. F. T. & Lings, M. L. (2001). Horizontally
mounted bender elements for measuring anisotropic shear moduli in
triaxial clay specimens. Geotechnical Testing Journal, 24, No. 2,
133144.
Pickering, D. J. (1970). Anisotropic elastic parameters for soil. Geotechnique 20, No. 3, 271276.
Raymond, G. P. (1970). Discussion: Stresses and displacements in a
cross-anisotropic soil. Geotechnique 20, No. 4, 456458.
Simpson, B., Atkinson, J. H. & Jovicic, V. (1996). The inuence of
anisotropy on calculations of ground settlements above tunnels. In
Geotechnical aspects of underground construction in soft ground
(eds R. J. Mair & R. N. Taylor), pp. 591594. Rotterdam: A. A.
Balkema.
Skempton, A. W. (1954). The pore-pressure coefcients A and B.
Geotechnique 4, No. 4, 143147.
Uriel, A. O. (1973). Discussion: Prediction of undrained deformations
and pore pressures in weak clay under two embankments. Geotechnique 23, No. 2, 294301.
Uriel, A. O. & Canizo, L. (1971). On the elastic anisotropy of soil.
Geotechnique 21, No. 3, 262267.

Vous aimerez peut-être aussi