SOILS AND FOUNDATIONS Vol. 51, No. 4, 749–760, Aug.

2011
Japanese Geotechnical Society

DEFORMATION CHARACTERISTICS OF DRY HOSTUN SAND

WITH PRINCIPAL STRESS AXES ROTATION

M. BLANCi), H. DI BENEDETTOii) and S. TIOUAJNIiii)

ABSTRACT
Coaxiality between the principal directions of the stress tensor and the principal directions of the plastic strain incre-
ment tensor is assumed in conventional plasticity models. In order to investigate coaxiality, or non-coaxiality, between
these two principal directions, a series of drained tests on dry Hostun sand was carried out using a precision Hollow
Cylinder Apparatus (HCA). The applied stress path includes large Principal Stress Axes Rotation (PSAR). Two of the
three principal stresses are kept constant. Therefore, among the three principal stresses, only the intermediate principal
stress, which is conˆning pressure (same pressure outside the hollow cylinder for internal and external lateral surfaces),
changes during loading. During these tests, at diŠerent stress levels, elastic (or quasi-elastic) properties are also investi-
gated, using small amplitude quasi-static cycles. These small cycles are performed in two diŠerent directions by succes-
sively changing only the axial stress szz or the shear stress suz. Elastic experimental properties are well simulated using
the DI Bendetto-GeoŠroy-Sauzeat (DBGS) hypo-elastic model, which takes into account PSAR. For each test, the
elastic part of deformation is calculated using the DBGS model and removed from global strain so that it is possible to
to focus only on the irreversible part (plastic part). Then, the principal directions of stress and plastic strain increment
are compared. Experimental results show that there is no coaxiality between these directions. This observation attests
to the existence of a non-coaxial plasticity. In addition, the coupling between tge coaxial and non-coaxial part is clearly
shown. Experimental results reveal that the plastic strain part is very important for the ˆrst large amplitude cycles and
remains greater than the elastic part even after 20 cycles.

Key words: deformation, drained shear, elasticity, laboratory test, plasticity, repeated loads, sand, torsion, yield
(IGC: D6)

1990 and Gutierrez et al., 1991, among others). These

INTRODUCTION studies show a diŠerence between the principal directions
The focus on the deformation characteristics of sandy of stress and the plastic strain increment. This means that
grounds over the last four decades has been especially im- the standard elasto-plastic hypothesis of coaxiality is not
portant in that it has provided a framework for inves- relevant. There is a non-coaxial plastic behaviour.
tigating settlement and liquefaction during earthquakes. Non-coaxiality has been often studied in order to ex-
During this time, the eŠects of cyclic loading have been plain shear band formation in the strain hardening
taken into account very carefully (Tastusoka and Ishi- region. Rudnicki and Rice (1975), Tsutsumi and
hara, 1974). These conditions may take place with the Hashiguchi (2005), Yu and Yuan (2006) and Yu (2006)
continuous rotation of the principal stress axes (Ishihara developed non-coaxial models in a two dimensional stress
and Towhata, 1983). Principal Stress Axes Rotation plan. Recently, Qian et al. (2008) extended these models
(PSAR) is also a current loading path, which has an im- in a general three dimensional tensorial space. Non-coax-
portant impact on geotechnical constructions (Jardine, ial plastic models also allow for a better understanding of
1994). Elastic and plastic deformations are dependent on some sand characteristics. For example, Gutierrez and
the rotation of principal stress axes. For these reasons, Ishihara (2000) performed a better evaluation of energy
PSAR eŠects need to be studied and modelled correctly. dissipation thanks to non-coaxiality.
The reproduction of PSAR conditions in a laboratory re- In this article, the authors propose to study the in-
quires complex devices. DiŠerent campaigns used Hollow ‰uence of PSAR on both elastic and plastic deforma-
Cylinder Apparatuses (HCA) to test sand samples (Ishi- tions. An experimental campaign on drained dry Hostun
hara and Towhata, 1983, Miura et al., 1986, Pradel et al., sand using large amplitude cyclic loadings with a continu-
i)
Researcher, Physical Modelling in Geotechnics Group, GER, IFSTTAR, LUNAM University, France (formerly PhD Student, Ecole
Nationale des Travaux Publics de l'Etat, University of Lyon) (matthieu.blanc＠ifsttar.fr).
ii)
Professor, D áepartement G áenie Civil et B âatiment (CNRS 3237), Ecole Nationale des Travaux Publics de l'Etat, University of Lyon, France.
iii)
PhD Student, ditto.
The manuscript for this paper was received for review on June 18, 2010; approved on March 31, 2011.
Written discussions on this paper should be submitted before May 1, 2012 to the Japanese Geotechnical Society, 4-38-2, Sengoku, Bunkyo-ku,
Tokyo 112-0011, Japan. Upon request the closing date may be extended one month.

749

This is an Open Access article under the CC-BY-NC-ND license.

750 BLANC ET AL.

ous rotation of the principal stress axes was conducted. 1983) shows that these dimensions allow for reasonable
The HCA, developed at the Ecole Nationale des Travaux homogeneity of stress and strain tensors within the sam-
Publics de l'Etat (ENTPE) laboratory and named ``T4C ple. Equations used for the calculation of homogenous
StaDy,'' has the same outside pressure for the internal stress and strain tensor are given in Sauzeat (2003). Two
and external lateral surfaces. This choice has the advan- neoprene membranes (0.5 mm thickness) constitute the
tage of creating homogenous stress and strain ˆelds wi- lateral sides, while two rigid platens close the sample at
thin the sample. However, only the two extreme principal the top and the bottom. The top cap, connected to the
stresses can be kept constant during the rotation, while press piston, is mobile in rotation and axial translation.
the intermediate one, s2＝srr, changes. Therefore, Axial and torsional loading is ensured by a servo-con-
peculiar analyses have to be introduced. Di Benedetto et trolled hydraulic press. Axial and shear load cells are in-
al. (2001) showed the importance of PSAR on the evolu- corporated in the piston. Measurements accuracies of ax-
tion of elastic deformation. An elastic investigation was ial and shear stresses can be respectively estimated equal
also conducted using small amplitude quasi-static cyclic to 0.1 kPa and 0.05 kPa (Duttine, 2005). The lateral sides
loadings in diŠerent directions. (both internal and external) of the sample are subjected
The ``T4C StaDy'' HCA is described at ˆrst. Then, the to atmospheric pressure while vacuum pressure down to
test procedure is introduced and the test results are ex- 20 kPa is applied inside the sample. This pressure is con-
posed. A general hypo-elastic model (developed within trolled by an electronic vacuum regulator from an elec-
the laboratory and named DBGS) is presented; simula- tronic signal input. Stress state of the hollow cylindrical
tions carried out with this model are compared with ex- sample is presented in Fig. 2. The stress tensor is given in
perimental measurements, revealing a non-negligible part Eq. (1) in ( …er, …eu, …ez) sample axes:
of non-coaxial elastic deformation. This elastic simulated srr 0 0 Pc 0 0 
strain is removed from the total strain to obtain the plas-
tic part. The general decomposition in the coaxial and  
s(ruz)＝ 0 suu suz ＝ 0 Pc t  (1)
non-coaxial part of diŠerent tensors is exposed. Then the
 0 suz szz 0 t Pc＋sa
amplitudes of the total and non-coaxial plastic strain in- with Pc conˆning pressure, sa axial stress minus or-
crements are compared. Finally, this evolution is studied thoradial stress (＝szz－suu) (Fig. 3) and t shear stress in
with a number of large cycles to investigate the cyclic the ( …eu, …ez) directions.
hardening properties.

HOLLOW CYLINDRICAL APPARATUS

``T4C STADY''
The hollow cylinder sample in the HCA ``T4C StaDy''
(Fig. 1) has a height of 12 cm, an outer diameter of 20 cm
and an inner diameter of 16 cm. A numerical analysis of
stress and strain distributions in the HCA (Hight et al.,

Fig. 2. Applied loads in a hollow cylindrical sample—Idealized stress-

es and strains on an element

Fig. 3. Analysis of stress state: principal stresses—deˆnition of angle

as of major principal stress direction in ( …eu, …ez) axes (left) and
Fig. 1. Schematic view of the HCA ``T4C StaDy'' Mohr's circle representation (right)
 BEHAVIOUR WITH AXES ROTATION 751

The accurate local systems of strain measurement (Fig. Table 1. Grading characteristics of Hostun sand used in the present
1) include 14 non contact transducers. Vertical and angu- study
lar displacements are measured on two levels thanks to Passing Diameter (mm) Coe‹cients Void ratios
two light duralumin rings less than 30 g in weight, which D*10 D*30 D*60 Cu** C** e*** e***
c min max
are hung on 3 points of the outer membrane. Rings have Hostun 0.26 0.32 0.37 1.42 1.06 0.648 1.041
aluminum targets at which the non-contact transducers
* Dx deˆned by xz passing particle size
are aimed (8). Radial displacements (outer and inner) are
** Coe‹cient of uniformity: Cu＝D60/D10 and coe‹cient of curvature:
also measured by a set of non-contact transducers (6) Cc＝(D30)2/(D10D60)
pointing through the membranes towards aluminum foils *** Hostun: after Flavigny et al. (1990)
placed inside the membranes in contact with sand. To in-
sure better accuracy, all non-contact transducers (1 mm
or 2 mm range) can be moved manually by micrometric
screws during the test. The resulting estimated strain ac-
curacy is better than 0.0005z (Duttine, 2005). However,
as an obvious counterpart, the transducers need to be
repositioned during testing. The form of the strain tensor
in the sample axes ( …er, …eu, …ez) is given in Eq. (2) and the
terms of the strain tensor are represented in Fig. 2.
 er 0 0 

e(ruz)＝ 0 eu g/2  (2)
 0 g/2 ez 

Fig. 4. Representation of the loading path which remains on the same

Principal Stress Convention and PSAR Angle Mohr-Coulomb surface in deviatoric normalized p plane: I1＝tr(s)
Major s1, Intermediate s2 and Minor s3 principal ＝ 3 (a)—Evolution of intermediate principal stress parameter b
stresses can be expressed with Pc, sa and t: as a function of 2as (b)

 sa s2a＋4t2
s1＝sIII＝Pc＋ ＋
 2 2 Stress Path
s2＝sI＝Pc (3)
The aim of the experimental campaign is to study the
s3＝sII＝Pc＋sa－ s2a＋4t2 eŠects of PSAR on the elastic and plastic behaviours of
 2 2
dry sand. During the test, the angle of PSAR as increases
Principal directions ( …e1, …e2, …e3) are associated to principal continuously at a constant rate while the principal stress-
stresses (s1, s2, s3). Directions …e2 and …er always coincide es s1 and s3 remain always constant. Due to the HCA
and PSAR occurs in ( …eu, …ez) directions. To simplify the ``T4C StaDy'' design, intermediate principal stress s2
rotation tensor writing, ( …eI, …eII, …eIII) base (corresponding (which corresponds to conˆning pressure Pc) varies be-
to principal stresses (sI, sII, sIII) deˆned in Eq. (3) is in- tween the two other principal stresses (Figs. 3, 4). Then
troduced. Then, PSAR occurs around …eI. The angle of the intermediate principal stress parameter b＝(s2－s3)/
PSAR as is deˆned as the angle between the major prin- (s1－s3) and also the Lode angle change during the test.
cipal direction …eIII (＝ …e1) and the vertical …ez: The following relationship shows the link between as and
b (Fig. 4(b)):
2as＝arctan Ø »
2t
sa
(4)
b＝
1－cos (2as)
(5)
2
In order to apply this particular stress path, the radius
TEST PROCEDURES and center of the Mohr-Coulomb circle, respectively: tmax
Tested Material ＝ (sa/2)2＋t2 and Pc＋sa/2 (Fig. 3), are kept unchanged
The tested material is an air dried poor graded sand, during the test. These conditions are obtained experimen-
called Hostun sand from its original location in France. tally by monitoring the three applied stresses (Pc, sa, t)
This sand is quartz dominated and sub angular in shape with cyclic sinusoidal evolutions at diŠerent phase lags.
with grading curve characteristics as summarized in Table In this way, the mobilized Mohr-Coulomb angle qmob
1. Deposit is made by air pluviation. The vibration remains constant during PSAR (Fig. 4(a)):
method is used to obtain the desired initial void ratio. s 1－ s 3 tmax
Moulds are removed while applying a partial vacuum of sin (qmob)＝ ＝ (6)
s 1＋ s 3 sa
20 kPa (corresponding to an eŠective pressure of 20 kPa) Pc＋
within the sample. Then samples are isotropically con- 2
solidated to the desired conˆning pressure. After removing the moulds and instrumenting the sam-
ple, isotropic consolidation is applied until an eŠective
pressure of 80 kPa (path from point O to point O? in
752 BLANC ET AL.

Fig. 5. Stress path of test n92 in diŠerent axes: (sa; 2t) and ( p; sII)
—Representation of current stresses (Pc, sa, t) and principal stress-
es (sI, sII, sIII) as a function of 2as during PSAR

Table 2. Characteristics of the PSAR tests conducted in this study

test tmax Pc＋sa/2 qmob e0 number of

n9 (kPa) (kPa) (9) cycles
1 20 60 19.5 0.65 20
2 20 60 19.5 0.84 20

3 25 55 27.0 0.67 18
4 25 55 27.0 0.83 15

Figs. 4, 5) is achieved. The purpose of this consolidation

is to remove irreversible deformations due to an isotropic
plastic mechanism. Then the conˆning pressure is
decreased to an isotropic state (point O! in Figs. 4, 5)
Fig. 6. Representation of the 4 tests in (ez－eu; g) axes (origins of
whose value depends on the test itself. A deviatoric path strains are taken at the beginning of the PSAR, point A)
starts with a classical compression triaxial loading until
point A. At this state, loading with PSAR begins. Four
tests are preformed at two Mohr-Coulomb angles qmob is the same for all tests.
and two initial void ratios e0 (Table 2). For each test, a
continuous rotation at a constant angular rate a· s＝ Elastic Investigation
0.0189/s is applied. This stress path is presented in Fig. 5 In order to evaluate elastic deformation, an elastic in-
in diŠerent axes: (sa; 2t) and (p; sII) (where p is the mean vestigation is conducted at the end of the test. Sixteen in-
pressure, and sII the square of the second stress invariant vestigation points are selected on the circular stress path
sII＝ J2＝ sijsij). Evolutions of the current stresses (Pc, sa, (Fig. 5). Once the investigation point is reached, a creep
t) and principal stresses (sI, sII, sIII) are also plotted as a period of approximately two hours is observed in order to
function of as. consume most of the viscous deformations that would
For each test, between 15 and 20 cycles are performed. otherwise occur during the elastic investigation period.
Figure 6 represents the strain paths of the four consi- Then two types of small quasi-static cyclic loadings are
dered tests plotted in (ez-eu; g) axes. The origins of the applied, i) a ``pure axial loading'' where only szz changes
strains are taken at the beginning of PSAR (i.e., for as＝ and ii) a ``pure torsional loading'' where only t changes.
09or at point A). By comparing tests n9 1 and 2 (qmob＝ The strain amplitudes of these cycles are small enough
19.59 3 and 4 (qmob＝27.09
) with tests n9 ), the in‰uence of (about 10－5) to insure a negligible plastic part (Duttine,
qmob is considered. Strain amplitudes increase with qmob. 2005; Di Benedetto et al., 2001). Tthis elastic investiga-
On the other hand, by comparing tests n9 1 and 3 (medi- tion procedure is explained in more detail in Duttine et al.
um density samples) with tests n92 and 4 (dense samples), (2007).
more deformations are observed for medium density
samples (as expected). However, the general strain shape
 BEHAVIOUR WITH AXES ROTATION 753

Table 3. Constants of DBGS model (Duttine, 2005)

ELASTIC STRAIN DETERMINATION
Cg b n0 m
It is supposed that the strain increment is divided into
two parts: an elastic strain increment deeij and a plastic 1003 1.68 0.175 0.44
strain increment depij:
deij＝deeij＋depij (7)
In order to obtain the correct plastic deformation values,
the elastic part must be evaluated accurately and removed
from the total deformation. A hypo-elastic model devel-
oped at ENTPE and named the DBGS (Di Benedetto et
al., 2001) is presented and compared with experimental
elastic investigations.

Hypoelastic DBGS Model

This rheological model is given by a hypo-elastic for-
mulation. The general incremental form, for HCA geo-
metry, of the hypo-elastic DBGS model (Di Benedetto et
al., 2001) is given from Eq. (8) to Eq. (12):
deeij＝M eijkldskl (8)

 deer M err M eru M erz M erg  dsrr 

 deeu
e
de z
＝

M eur
M ezr
dg / 2 M egr
e
M euu
M ezr
M egu
M euz
M ezz
M egz
 
M eug
M ezg
M egg
dsuu
dszz
2 dt
(9)

where dee and ds are the elastic strain and stress incre-
ments. M e is the rheological compliance tensor which is
independent of stress increment direction. Moreover, ex-
perimental results exhibit a symmetrical tensor M e (M eij＝
M eji) (Duttine et al., 2007). The DBGS model gives the fol- Fig. 7. M eij values obtained for test n92 from quasi-static small ampli-
lowing anisotropic (orthotropic) and symmetrical expres- tude cyclic loadings in diŠerent directions (pure axial (a) and pure
sion of the tensor M e: torsion (b)) performed at diŠerent stress levels i) data points and ii)
simulation with DBGS hypo-elastic model (lines and dashed lines)
1 S̃v･S̃p＋tS̃v･tS̃p
M e＝ (10)
f (e) 2 isotropic stress state.
with f (e) a function of the void ratio: f (e)＝Cg(b－e)2/(1 Previous experimental campaigns conducted on dry
＋e) where sCg; btare two of the material's constants. Hostun sand at diŠerent void ratios and conˆning pres-
The transpose function is denoted by ``t'', and s S v; S p t sures (Duttine, 2005) allowed for four of the constants of
are the tensors originally deˆned by Hardin and Blan- the model for this sand to be evaluated (Table 3).
dford (1989) which take the following forms for HCA ge- The APPENDIX includes a method which can be used
ometry when expressed in the principal stress axes ( …eI, …eII, to manipulate this tensor with PSAR when the ( …er, …eu, …ez)
…eIII): base and the principal base ( …eI, …eII, …eIII) do not coincide.
 1 － n0 － n0 0 
Experiments Analysis and Simulation
S̃v＝
－ n0 1 － n0 0
－ n0 － n0 1
 0 0
0
0 1＋n0
 (11) As exposed previously, two types of small quasi static
cyclic loadings are applied. One is purely in the axial
direction (only dszz is not nil) giving four terms of the
1  elastic tensor (M erz, M euz, M ezz, M egz), while the other one is
0 0 0 purely in torsion (only dt is not nil) in order to directly

 
sIm obtain four other terms (M erg, M eug, M ezg, M egg) (Eq. (9)).
1
m
0 0 0 To remain within the elastic domain, the amplitude of
s II
S̃p＝ 1 (12) quasi-static loading is 1 kPa (resp. 0.5 kPa) for axial
0 0 m
0 (resp. torsional) loading. Strain amplitudes remain also
sIII
1 below some 10－6, considered as the elastic limit. Ex-
0 0 0 m/2
m/2
perimental results for test n9 2 are presented. In Fig. 7,
 s sIII
II  the evolution with as of the 8 terms of the elastic tensor
where sm; n0tare two constants which stand respectively obtained from experimental data is plotted for test n9 2.
for power coe‹cient and Poisson's ratio value for an Some terms show important changes during PSAR. For
754 BLANC ET AL.

reasonably acceptable even for important PSAR.

¿dee¿,ds ＝ M e2＋M e2＋M e2＋M e2 dszz

zz rz uz zz gz
 e e2 e2 e2 e2
(13)
¿de ¿,dt＝ M rg ＋M ug＋M zg＋M gg 2 dt


¿dee¿exp
,ds ¿dee¿exp
,dt
j,ds ＝ －
zz
1; j ,d t ＝ －1 (14)
¿dee¿DBGS ¿dee¿DBGS
zz
,ds zz ,dt

For test n9 2, the simulated elastic strain increment vec-

tor dee obtained for a stress increment vector ds (which is
the diŠerence between the stress tensor at as and as＋das
with das＝1.89 ) is plotted (Fig. 9) in the (deez－deeu; dge)
axes for diŠerent stress levels in the (sa; 2t) axes. The
simulated elastic strain increment direction is very close
Fig. 8. Elastic strain deviation j between DBGS model and ex- to the tangent vector at the circle described in the (sa; 2t)
perimental values obtained for an axial stress increment (only dszz axes. As explained later, this tangential direction is a non-
»0) and for a shear stress increment (only dt»0) at the 16 elastic
coaxial direction. Therefore, during the test, the quasi
investigation points
totality of elastic deformation is in the non-coaxial direc-
tion. For this reason, as underlined previously, accurately
evaluating the elastic part is a key point in the study of
plastic deformation, especially its non-coaxial part.

PLASTIC MECHANISMS
Plastic deformation is obtained after removing the
elastic deformation (simulated by DBGS model) from the
total deformation. Figure 10 represents the total (Fig.
10(a)), elastic (Fig 10(b)) and plastic (Fig. 10(c)) strain
paths during 3rd cycle of test n92. Origins of strains are
taken at the beginning of the considered cycle (i.e., for as
＝09). During this cycle, few volume changes happen and
there is nearly no volume variation between the beginning
and the end of this cycle. Meanwhile, as it is usual for a
medium loose sample (e0＝0.84), the sample contracts
slightly after one cycle. This volume variation is too small
to have an impact on the value of function f (e) (Eq. (10)).
Considering this, the hypothesis of no volume change
during cyclic loading was chosen. As a result, simulated
Fig. 9. Simulated elastic strain increment vectors dee (obtained for das elastic strains remain the same for all cycles. Figure 11
＝1.89) in (deez－deeu; dge) axes for diŠerent stress levels in (sa; 2t) represents the elastic, plastic and total strains in the (ez－
axes (test n92) eu; g) axes for cycle 3 and cycle 20 of test n9 2. In this
ˆgure, the 3rd (Fig. 11(a)) and the 20th (Fig. 11(b)) cycles
can be compared. The plastic strain amplitudes decrease
example, M ezz, which is independent of sII (the only prin- signiˆcantly between the 3rd and 20th cycles (almost by a
cipal stress which changes during the test), is nearly dou- factor of 3). Meanwhile, even after 20 cycles, plastic
bled between as＝09and as＝909 . On Fig. 7, simulations deformation remains greater than elastic deformation.
made with the DBGS model are also plotted for test n9 2 Two main plastic mechanisms (Eq. (15)) are consi-
using constants of Table 3. dered:
In order to evaluate the accuracy of DBGS model
depij＝depc pnc
ij ＋deij (15)
predictions, simulated (¿dee¿DBGS ,ds ) and experimental
zz

(¿dee¿exp
,ds )
zz
norms of the elastic deformation obtained for The ˆrst terms corresponds to one (or more) coaxial plas-
an axial stress increment (only dszz is not nil) (Eq. (13)) tic mechanism(s), noted by ``pc'', which maintains coaxi-
are compared. The deviation j,ds (Eq. (14)) between
zz ality between the principal directions of stress and the
DBGS model and experimental values is plotted in Fig. 8 plastic strain increment. It is classically accepted that this
for the 16 elastic investigation points. The same analysis mechanism takes the evolutions of mean pressure and
is conducted for the norm of the elastic deformation ten- deviatoric stress slightly. In order to minimize the eŠect
sor due to a shear stress increment (only dt is not nil) of the mean pressure, an isotropic pre-consolidation (up
(Fig. 8). The deviations are always smaller than 25z. to 80 kPa) is performed before each test. The deviatoric
Considering the experimental scatter due to device ac- eŠects can be considered relatively small. Indeed, in the
curacy, the DBGS model simulations can be considered normalized deviatoric plan (I1＝tr(s)＝ 3 ), the applied
 BEHAVIOUR WITH AXES ROTATION 755

Fig. 10. Total (a), elastic (b) and plastic (c) strains during cycle 3 of
test n92 as a function of as

stress path, remains on the same Mohr-Coulomb surface

(Fig. 4(a)) (qmob is constant, Eq. (6)). Nevertheless,
Fig. 11. Representation of total, elastic (simulated by DBGS model)
Lode's angle u or b (Eq. (5)) value changes during the
and plastic strains in (ez－eu; g) axes (origins of strains are taken at
test, which may be the result of deformation due to a clas- the beginning of the cycle) for the 3rd cycle (a) and the 20th cycle
sical coaxial mechanism. However, this doesn't change (b) of test n92 (points ABCD are deˆned in Fig. 4)
the proposed analysis of non-coaxial plastic mechanism.
The second terms of Eq. (15) corresponds to a non-
coaxial mechanism, noted by ``pnc'', which indeed plays proposed by Rudnicki and Rkce (1975) has been largely
a very important role during tests with PSAR. It is stud- accepted to simulate this non-coaxial eŠect. In this theo-
ied in the next section. It should be noted here that, when ry, a non-coaxial plastic deformation increment depnc ij is
expressed in principle stress axes, a non-coaxial X ''nc'' ten- assumed to be proportional to the non-coaxial stress in-
sor has nil terms on the diagonal. It is orthogonal to a crement dsnc
ij :
coaxial X ''c'' tensor as X ``nc''
ij .X ``c''
ij ＝0.
1
depnc pnc
ij ＝deij ＝ dsnc
ij (16)
H pnc
NON-COAXIAL FLOW RULE where H pnc is the plastic modulus which governs the non-
As noted in introduction, more and more experimental coaxial mechanism. It must be noted that this mechanism
evidence shows the existence of non-coaxial plastic be- is purely deviatoric; no volume variation is introduced:
haviour. Unlike usual plastic theory where principal depnc pnc
ij ＝deij . In the next section, the general decomposi-
directions of stress and the principal directions of plastic tion in a coaxial and a non-coaxial part is presented.
strain increment are the same, there is a non-coaxial
strain increment vector which responds to a non-diagonal Non-coaxial Decomposition
loading stress increment tensor (when expressed in the The general mathematical decomposition of any tensor
principal directions of stress). The yield vertex theory X in a coaxial part ``c'' and a non-coaxial part ``nc'' rela-
756 BLANC ET AL.

tively to a stress basis is given by Eq. (17).

ads＝as＋459 (20)
 ``c'' Xkldkl Xklskl XklSkl

X ＝ dij＋ sij＋ Sij The DBGS model formulation gives an elastic strain in-
 ij 3 sklskl SklSkl (17)

``nc'' ``c'' crement dee that is mainly non-coaxial. As such, the elas-
X ij ＝Xij－X ij tic part of strain increment impacts directly on the ampli-
where d is Kroneker delta, s is the deviatoric part of s and tude of the non-coaxial strain increment. For this reason,
S is a tensor orthogonal to I (identity) and s (Eq. (18)). a good evaluation and simulation of elastic deformation
are very important. In this paper, great care has been
dij＝ 1 if i＝j

0 if i»j{ taken on that point.

1 Experimental Results Analysis

s ＝s － tr(s)d
ij ij ij (18)
3 In Blanc (2011), similar trends were presented for the
S ＝s s － 1 J d －J s 3
three other tests. The focus was on the 3rd and 20th cycles
in order to study evolution with the number of cycles.
 ij ik kj 2 ij ij
3 J2
The ˆrst two cycles were not considered since creep
I, s and S are orthogonal to X ``nc'' (Eq. (19)). periods were applied at diŠerent levels. In order to com-
pare the strain increments, the variation of stress rotation
X nc nc nc
ij dij＝0; X ij sij＝0; X ij Sij＝0 (19)
for each increment is always das＝1.89(i.e., every 100s
X is either thestress increment ds, the strain increment because tests are performed at constant rotation rate a· s＝
de, the elastic strain increment dee or the plastic strain in- 0.0189/s).
crement dep. The plastic strain increment vector of the 3rd cycle is
In the case of the performed tests, the decomposition plotted in Fig. 13. The amplitude and direction of this
of plastic strain increments (and the stress increment) can vector are not constant during the cycle. Using the
be represented in Fig. 12 at a current stress level (sa; 2t) decomposition presented in Fig. 12, it is easy to observe
when considering the axes (depz－depu ; dgp). The rotation that plastic strain increment vector has not only a coaxial
angle of principal plastic strain increment directions ade is p component but also a non-coaxial (tangential) part. It
deˆned between the vertical direction (represented by denotes the existence of non-coaxial plastic behaviour. In
horizontal axis in Fig. 12) and plastic strain increment dep Fig. 14(a), rotation angles of the stress as, stress incre-
direction. The normal at the circular loading path ment ads and plastic strain increment ade are presented p

represents the coaxial direction, whereas the tangent during this 3rd cycle. Fig. 14(b) shows the evolution of
represents the non-coaxial direction. The plastic strain in- the plastic strain increment direction ade compared to the
p

crement vector is projected on these two directions to ob- coaxial direction as. It shows a cyclic shape evolution,
tain its coaxial and non coaxial parts (Eq. (17)). Figure 12 which has maximum values (points M1 and M2 in Fig.
shows also the stress increment ds, which is totally non- 14(b)) of about 109after pure axial compression-exten-
coaxial in these axes. The following relationship is ob- sion (i.e., after points A and C or for as＝09and 909 ).
tained between the rotation angles of the principal stress
directions as and the principal stress increment directions
ads:

Fig. 12. Decomposition of plastic strain increment vector in a coaxial

and a non-coaxial vector in (sa; 2t) axes (for stress) and (depz－depu ; Fig. 13. Plastic strain increment vectors dep (obtained for das＝1.89)
dgp) axes (for strain)—Deˆnition of principal plastic strain incre- in (depz－depu ; dgp) axes for diŠerent stress levels in (sa; 2t) axes (test
ment rotation ade p
n92)
 BEHAVIOUR WITH AXES ROTATION 757

Fig. 14. Principal axes rotations of: stress as, stress increment ads and
plastic strain increment ade (a) and evolution of directions of plas-
p

tic strain increment ade -as (b), as a function of as (test n92–3rd cy-
p

cle)

These maxima are obtained when the radial stress incre- Fig. 15. Components of non-coaxial deviatoric stress increment tensor
(a) and non-coaxial plastic strain increment tensor (3rd cycle (b)
ment dsrr is nearly nil. In contrast, dep is close to the and 20th cycle (c)) as function of as (obtained for das＝1.89)
coaxial direction (points m1 and m2 in Fig. 13(b)) when
dsrr is nearly maximum (points B and D). At points A
and C (sa＝09and 909) coaxial stress increment is nil (ds1 2 2 2
2 ¿dep¿＝ 2depr ＋2depu ＋2depz ＋dgp
2
(23)
＝ds2＝ds3＝0). Figure 14 reveals that, in points A and
C, the plastic strain increment has a non-negligible coaxi- Due to non-coaxial decomposition, the radial parts of
al part (a non coaxial part would give ade －as＝459 p). It these tensors are always nil (dsnc pnc
r ＝0 and der ＝0). In ad-
nc pnc nc pnc
indicates coupling between the coaxial and non-coaxial dition, dsz (and dez ) and dsu (and deu ) are always op-
plastic mechanisms. This coupling may be due to a diŠer- posite. dsnc nc
z and dsu are nil at points A and C. As the non-
ence between the principal strain (from isotropic states) coaxial tensor has one dimension, only the amplitude
directions ae and the principal stress directions as. ¿depnc¿ can be studied. While the amplitude of the non-
The components of the non-coaxial plastic stress incre- coaxial stress increment dtnc
max is constant during the entire
ment tensor (Fig. 15(a)) and the non-coaxial plastic strain test, the non-coaxial plastic strain increment amplitude
increment tensor (Figs. 15(b), (c)) are plotted as a func- ¿depnc¿, changes with as. The maximum amount of non-
tion of as. Norms of these increment vectors, dtnc max and coaxial plastic deformation occurs at around as＝309and
dgpnc
max , deˆned in Eqs. (21) and (22), are also plotted. 1209and the minimum is at around as＝709and 1609 , in-
dicating that the plastic non-coaxial modulus H pnc is not
Ø dsnc
»
nc 2
z －dsu
dtnc
max＝ ＋(dtnc)2 (21) constant during a cycle. It is, however, generally consi-
2
dered as a constant, as can be seen in Eq. (16).
dgpnc
max＝ 2 ¿de
pnc
¿＝ (depnc pnc 2
z －deu ) ＋(dg
pnc 2
) (22) The norm of the non-coaxial plastic strain increment
¿depnc¿ (resp. norm of non-coaxial plastic strain incre-
758 BLANC ET AL.

Fig. 17. Ratio between the norm of plastic strain increment ¿dep¿ (a)
and norm of plastic non-coaxial strain increment ¿depnc¿ (b) during
cycle ``x'' and the same values during the 3rd cycle

ment ¿dep¿) is plotted in Fig. 16(a) (resp. Fig. 16(b)) for

the 3rd and the 20th cycle. The ratio of these two norms,
¿depnc¿ and ¿dep¿, is presented in Fig. 16(c). The non-
coaxial part changes signiˆcantly during one cycle, from
20z to 90z of the total plastic amplitude. Even when the
loading is purely non-coaxial (i.e., in points A and C),
there is some coaxial deformation (around 15z). A non-
coaxial stress increment induces coaxial deformations,
suggesting a link between the coaxial and non-coaxial
plastic mechanisms. A comparison between the 3rd and
Fig. 16. Comparison of the norm of plastic non-coaxial strain incre-
ment ¿depnc¿ (a), norm of plastic strain increment ¿dep¿ (b), their the 20th cycles allows the following interesting observa-
ratio (c) and direction of plastic strain increment ade -as (d) for 3rd
p tions. The non-coaxial plastic strain increment keeps the
and 20th cycles of test n92 (obtained for das＝1.89) same form between the 3rd and the 20th cycle but its am-
plitude is reduced by a factor of about 3 (Fig. 16(a)). This
is proof of a non-coaxial cyclic hardening. The direction
 BEHAVIOUR WITH AXES ROTATION 759

lowing conclusions may be proposed:

i) A hypo-elastic model DBGS, developed at the EN-
TPE laboratory, is a good predictor of the evolution
of the hypo-elastic tensor M e and also elastic defor-
mation even for important rotations of principal
stress directions. A simulation of elastic strains ob-
tained by this model provides important non-coaxial
deformation. Hence, a good simulation of elastic
deformation is essential to correctly estimate the
non-coaxial part of plastic deformation.
ii) The tests performed, with the continuous rotation of
the principal stress directions, shows huge non-coax-
ial plastic strain increments. Traditional plastic theo-
ry, where principal stress and principal plastic strain
increment have the same direction, cannot be accept-
ed. Therefore, a non-coaxial plastic mechanism must
Fig. 18. Cyclic hardening parameter bp as a function of cycle number
(ref. 3rd cycle)
be considered.
iii) For a constant non-coaxial stress increment applied
during tests, the non-coaxial plastic strain increment
of the plastic strains increment (Fig. 16(d)) and the response changes continuously. No proportionality
proportion of non-coaxial plastic strain increment (Fig. was able to be found between the norm of the non-
16(c)) change within one cycle but the values remain un- coaxial stress increment ¿dsnc¿ and the norm of the
changed for diŠerent cycles. In addition, the total and non-coaxial plastic strain increment ¿depnc¿.
non-coaxial plastic strain increment amplitudes decrease iv) Due to device limitations, the experimental stress
with the number of cycles. path is not purely non-coaxial. Indeed, the radial
The ratio b p (resp. b pnc), deˆned in Eq. (24), between stress increment is not nil except for diŠerent stress
the norm of the total (resp. non-coaxial) plastic strain in- levels during loading: for pure compression and ex-
crement during the xth cycle and the one during the 3rd tension conditions (i.e., points A and C, Fig. 4).
cycle is presented in Fig. 17 as a function of the cycle However, even in points A and C where stress incre-
number. ments are purely non-coaxial, the plastic strain incre-
ment has a non-negligible coaxial part. It indicates a
¿depx¿ coupling between the coaxial and non-coaxial plastic
b p＝
¿dep3¿ mechanisms. This coupling may be due to the diŠer-
¿depnc
x ¿
ence between the principal stress axes and principal
b pnc＝ (24) strain (calculated from the isotropic state) axes. If
¿depnc
3 ¿
this is the case, it is valueless to model non-coaxial
For a given cycle, both the ratios b p and b pnc can be plasticity by using a proportional relationship (Eq.
considered as nearly constant and equal: b p＝b pnc. This (16)) between the non-coaxial parts of strain and
ratio, called the cyclic hardening parameter b p, decreases stress increments (i.e., between dsnc and depnc).
with the number of cycles (from 80z for the 4th cycle to v) Amplitudes of plastic coaxial and non-coaxial strain
30z for the 20th cycle). Figure 18 represents the evolu- increments decrease in the same proportion with the
tion of b p as a function of the number of cycles. As the number of large amplitude cycles. Cyclic hardening
reference is the 3rd cycle, the plot starts at x＝3 with b p＝ is nearly the same for both coaxial and non-coaxial
1. plastic mechanisms. In addition, even after 20 cy-
cles, plastic deformation remains more important
than elastic deformation.
CONCLUSIONS
This paper describes an experimental campaign con-
ducted in order to study the impact of the Principal Stress REFERENCES
Axes Rotation (PSAR) on deformation characteristics. A 1) Blanc, M. (2011): Etude experimentale et modelisation du compor-
series of drained tests on dry Hostun sand performed tement des sols avec rotation d'axes principaux de contraintes,
Ph.D thesis, ENTPE, University of Lyon. (in French) (to be pub-
with a Hollow Cylinder Apparatus (HCA) called ``T4C
lished).
StaDy'' was carried out with a continuous rotation of 2) Di Benedetto, H., GeoŠroy, H. and Sauzeat, C. (2001): Viscous
principal stress directions. The elastic behaviour was also and non viscous behaviour of sand obtained from hollow cylinder
investigated under these conditions with small amplitude tests, Advanced Laboratory Stress-Strain Testing of Geomaterials,
quasi-static cycles in diŠerent directions. The focus was (eds. by Tatsuoka et al.), Balkema, 212–221.
3) Duttine, A. (2005): Comportement des sables et des melanges
on the non-coaxial part of strain increment tensors (elas-
sable/argile sous sollicitations statiques et dynamiques avec et sans
tic and plastic) which refers to the deviation of principal 9rotation d'axes:, Ph.D thesis, ENTPE, University of Lyon (in
stress from principal strain increment directions. The fol-
760 BLANC ET AL.

French). 12) Miura, K., Miura, S. and Toki, S. (1986): Deformation behaviour
4) Duttine, A., Di Benedetto, H., Pham Van Bang, D. and Ezaoui, A. of sand under principal axes rotation, Soils and Foundations, 26(1),
(2007): Anisotropic small strain elastic properties of sands and mix- 36–52.
ture of sand-clay measured by dynamic and static methods, Soils 13) Pradel, D., Ishihara, K. and Gutierrez, M. (1990): Yielding and
and Foundations, 47(3), 457–472. ‰ow of sand under principal stress axes rotation, Soils and Founda-
5) Flavigny, E., Desrues, J. and Palayer, B. (1990): Note technique: le tions, 30(1), 87–99.
sable d'Hostun RF, Revue Francaise de Geotechnique, 53, 67–70. 14) Qian, J. G., Yang, J. and Huang, M. S. (2008): Three-dimensional
6) Gutierrez, M., Ishihara, K. and Towhata, I. (1991): Flow theory for noncoaxial plasticity modeling of shear band formation in ge-
sand during rotation of principal stress direction, Soils and Foun- omaterials, Journal of Engineering Mechanics, 134(4), 322–329.
dations, 31(4), 121–132. 15) Rudnicki, J. W. and Rice, J. R. (1975): Conditions for the localiza-
7) Gutierrez, M. and Ishihara, K. (2000): Non-coaxiality and energy tion of deformation in pressure-sensitive dilatant materials, Journal
dissipation in granular materials, Soils and Foundations, 40(2), of the Mechanics and Physics of Solids, 23, 371–394.
49–59. 16) Sauzeat, C. (2003): Comportement des sols en petites et moyennes
8) Hardin, B. O. and Blandford, G. E. (1989): Elasticity of particulate deformation, Ph.D thesis, ENTPE, University of Lyon (in French).
materials, Journal of Geotechnical Engineering, 115(6), 788–805. 17) Tatsuoka, F. and Ishihara, K. (1974): Drained deformation of sand
9) Hight, D. W., Gens, A. and Symes, M. J. (1983): Development of a under cyclic stresses reversing direction, Soils and Foundations,
new hollow cylinder apparatus for investigating the eŠects of prin- 14(3), 51–65.
cipal stress rotation in soils, G áeotechnique, 33(4), 355–383. 18) Tsutsumi, S. and Hashiguchi, K. (2005): General non-proportional
10) Ishihara, K. and Towhata, I. (1983): Sand response to cyclic rota- loading behaviour of soils, International Journal of Plasticity,
tion of principal stress directions as induced by wave loads, Soils 21(10), 1941–1969.
and Foundations, 23(4), 11–26. 19) Yu, H. S. and Yuan, X. (2006): On a class of non-coaxial plasticity
11) Jardine, R. J. (1994): One perspective on the pre-failure deforma- models for granular soils, Proceedings of the Royal Society A:
tion characteristics of some geomaterials, Keynote Lecture in Pre- Mathematical, Physical and Engineering Science, 462(2067),
Failure Deformation Characteristics of Geomaterials, (eds. by 725–748.
Mitachi and Miura), Balkema. 20) Yu, H. S. (2006): Plasticity and Geotechnics, Springer.

APPENDIX
For principal stress axes rotation, the sample axes ( …er, …eu, …ez) and principal stress axes ( …eI, …eII, …eIII) do not coincide. An
expression of the hypo-elastic tensor M e in ( …er, …eu, …ez) axes is given below. Substituting Eqs. (11) and (12) in Eq. (10)
gives the hypo-elastic tensor M e in ( …eI, …eII, …eIII) axes (Eq. (A1)).
 1 － n0 1
Ø 1
＋ m »－ Ø n0 1 1
＋ m » 0


 
smI m
2 sII sI m
2 s III sI

M e(I, II, IIII)＝

1 Ø
－n0 1
m
1
＋ m
2 sI sII » 1
sIIm
－
Ø n0 1
m
1
＋ m
2 s III sI » 0
(A1)
Ø
f (e) －n0 1
m
1
＋ m
2 sI s III » － n0 1
m Ø 1
＋ m
2 sII s III » 1
smIII
0

1 ＋ n0
0 0 0
 (sIIsIII)m/2
Incremental stress tensor is expressed in ( …eI, …eII, …eIII) by Eq. (A2).
1
ds(I, II, III)＝P(I, II, III)ª(r, u, z).ds(r, u, z).P－
(I, II, III)ª(r, u, z) (A2)
with P the transformation rotation matrix from ( …eI, …eII, …eIII) to ( …er, …eu, …ez):
1 0 0 
P(I, II, III)ª(r, u, z)＝ 0 cos (as) sin (as)  (A3)
0 －sin (as) cos (as)
For HCA analysis, only four terms of the strain and stress tensors are non nil. These tensors are represented by a four
component vector. With this notation, the relationship (Eq. (A2)) between stress increments in ( …eI, …eII, …eIII) and ( …er, …eu,
…ez) becomes:
ds(I, II, III)＝Q.ds(r, u, z) (A4)
with
1 0 0 0 

 
2 2
0 cos (a) sin (a) － 2 cos (a) sin (a)
Q＝ 2 2
(A5)
0 sin (a) cos (a) 2 cos (a) sin (a)
0 2 cos (a) sin (a) － 2 cos (a) sin (a) cos2 (a)－sin2 (a) 
Then the hypo-elastic tensor M e expression in ``( …er, …eu, …ez) base'' is:
M e(ruz)＝Qt.M e(I, II, III).Q (A6)