Sunteți pe pagina 1din 25

JAMES, R. G. & BRANSBY. P. L. (1971). Gtotechniqus 21, No. 1, 61-83.

A VELOCITY FIELD FOR SOME PASSIVE EARTH


PRESSURE PROBLEMS
R. G. JAMES* and P. L. BRANSBY*

SYNOPSIS
A simple velocity field is developed for the passive Un simple champ de vitesse est tree pour le prob-
earth pressure problem of the displacement of a l&me de la pression passive du sol dans le deplace-
rough plane wall into a mass of dry sand. Three ment dune paroi brute, dans une masse de sable sec.
modes of wall displacement are considered : wall rota- Trois modes de deplacement de paroi sont consi-
d&es: la rotation de la paroi sur la partie superieure,
tion about the top, wall rotation about the toe and La
la rotation du mur sur le pied et la translation.
wall translation. The validity of the velocity field validite du champ de vitesse dans les trois cas est
for the three cases is discussed. The pattern of the discutee. Le modele off& par le champ de vitesse
velocity field and the velocity boundary conditions et par les conditions de la zone limite de vitesse per-
enables predictions to be made of the strain fields met de faire des predictions des champs de defor-
associated with the three modes of wall movement mation dans les trois modes de mouvement de paroi
and these predictions are compared with experimen- et ces predictions sont comparees 8. des observations
tal observations. experimentales. Le champ de vitesse predit &gale-
The velocity field also predicts
ment la position et la serie des surfaces de rupture
the location and sequence of the rupture surfaces
qui se produisent dans un sable dense et qui aboutis-
that occur in dense sand and leads to an understanding sent & une comprehension des differences dans les
of the differences in the experimentally observed champs deformation et dans les distributions de con-
strain fields and wall stress distributions. trainte dans la paroi.
This Paper follows on from an earlier paper (James Cet Expose suit un article anterieur (James et
and Bransby, 1970) in which the boundary stresses Bransby, 1970) dans lequel les contraintes de zone
and the predicted principal stress directions in the limite et les directions predites des contraintes
principales dans la masse Btaient comparees avec des
mass were compared with measured boundary
contraintes mesurees de zones limites et avec des
stresses and with measured strain increment direc- mesures des directions daccroissement de deforma-
tions. The present Paper makes direct comparison tion. Le present Expose fait une comparaison
between predicted and measured strains for dif- directe entre les deformations predites et mesurees
ferent modes of wall movement. pour differents modes de mouvement de paroi.

INTRODUCTION

In order to determine the displacements of retaining walls under passive load at the working
load condition, it is necessary to consider what stresses, strains and displacements are caused
throughout the soil mass. At the moment there is little theoretical work in which the strains
and deformations caused by wall movement are calculated, while experimental data of the
strain and displacement fields are of recent origin (Arthur, 1962; James, 1965; Bransby, 1968;
Lord, 1969). The data do not encourage theoretical approaches to retaining wall problems as
the deformation behaviour is extremely complex.
It was therefore very difficult to explain and predict the gross differences in behaviour that
were observed when rough plane walls were rotated about different points in their lengths into
sand masses.
Conventional stress analyses of retaining wall problems (such as the Coulomb method and
the log spiral method) and the method of Sokolovski (1960, 1965) do not distinguish between
different wall movements so long as the angle of internal friction + of the sand and the angle of

* Department of Engineering, University of Cambridge.


61
62 R. G. JAMES AND P. I.. BRANSBY

NOTATION

voids ratio 1. .
v =%IYril incremental strains in r, # polar
wall height co-ordinate system
coefficient of earth pressure at i,, i, major and minor principal incre-
rest mental strains
(al+%)/2 8 inclination of wall to vertical,
(~I-42 rotation about toe
thickness of transition layer
~splacements in the Yand # direc- 4 angular velocity of wall, rotation
tions respectively in polar co- about toe
ordinate system B inclination of wall to vertical,
cumulative volumetric strain rotation about top
incremental volumetric strain, $ angular veIocity of wall, rotation
li,+tt, about top
total velocity of element of soil
components of velocity along the angle of dilation, sin-I( --d/p)
01and /3zero extension lines (TV,oy, ~~~ stresses in Cartesian co-ordinate
horizontal displacement of wall system
horizontal velocity of wall @IFCT3 major and minor principal
cumulative maximum shear stresses
strain angleof internalfriction,sin-l (t/s)
incremental maximum shear inclination of @ zero extension
strain, I+ - $1 $
direction to x axis
angle of wall friction
incremental strains in Cartesian Note: 8, 4, B, e are taken positive for
co-ordinate system rotation into the sand

wall friction 6 are the same in all cases. The predictions of the Sokolovski method relating
to the passive earth pressure problem considered in this Paper have been given by James and
Bransby (1970). However, the analysis in that paper was concerned only with the stress dis-
tribution within the sand mass, whereas this Paper is primarily concerned with the prediction
of the distribution of strains within the mass.
Shield (1953), Drnovsek (X965) and Poorooshasb et al. (1966, 1967) have considered dis-
placement fields associated with statically admissible stress fields. However, Shield and
Dmovsek adopted the unsatisfactory material idealization proposed by Drucker and Prager
(1952). Paorooshasb et al. developed solutions using mathematical models based on the
observed behaviour of sand in the triaxial apparatus, but did not discuss the differences in
strain fields, rupture surface mechanisms and wall stress distributions caused by different wall
displacements.
The Authors therefore decided to simplify the analysis and did so by making an unusual
material idealization and by considering only the deformation aspects of the problem. Such
an approach can explain the gross features of the data and allows qualitative prediction of the
soil behaviour in different wall problems. No attempt is made to obtain the exact solution,
although it is anticipated that the development of such exact solutions will be helped by the
insight gained by this approach.
The work of this Paper forms a small part of the research discussed by Roscoe (1970) who
outlined a simple method of predicting the strain field from known boundary conditions. This
approximate method is described in detail in this Paper as it represents one stage of the
development of the theory which gives results that may be directly compared with experi-
mental observations.
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 63

METHOD OF APPROACH

The upper and lower bound theorems of the theory of plasticity are extremely powerful
tools which enable limits to be placed on the collapse load of plastic structures, and it is these
theorems that gave the stimulus for the method of approach. Methods of analysis of soil
mechanics problems which find either upper or lower bound solutions have been suggested by
several authors (e.g. Drucker and Prager, 1952; Sokolovski, 1960, 1965) but the relevance of
the solution depends on whether soil may reasonably be idealized as a perfectly plastic
material. Even if it were to be accepted that the stress behaviour of sand may be approxi-
mated by a simple rigid plastic idealization based on the Mohr-Coulomb failure criterion, it is
well known that the deformation behaviour of sand does not meet the requirements of the
idealization. Experimental data (e.g. Barden and Khayatt, 1966; Cole, 1967; Poorooshasb et
aE., 1966, 1967) indicate that there is substantially less expansion during shear than that
required by the perfectly plastic idealization and thus the theorems of plasticity are no longer
strictly relevant (Davis, 1968).
However, the limit theorems suggest methods of analysis for soil problems that can be
investigated by comparison of theory with experiment. James and Bransby (1970) use the
Sokolovski method of analysis and discuss the validity of the idealization of stress behaviour
and the accuracy of a lower bound method for one of the passive earth pressure problems
discussed. A method suggested by the upper bound approach is now considered.
To obtain an upper bound solution for a perfectly plastic material, a compatible deforma-
tion mechanism is selected for the problem and then the collapse load associated with the
mechanism is calculated using the yield condition for the material.
This Paper concentrates on the first part of an upper bound calculation procedure. De-
formation fields are established for three passive earth pressure problems and the consequences
of adopting these deformation fields are compared with experimental observations. It is
found that such a procedure gives insight into soil behaviour and explains the gross differences
in behaviour that are observed in the three problems. However, it must be remembered that
the method is approximate as no account is taken of the stress equilibrium equations. Fur-
ther, an anomaly emerges in the analysis of one of the problems.
Of vital importance for the analysis is the choice of the material idealization. For con-
ventional methods of analysis it is only necessary that an idealization be made of behaviour
under stress, and it is usually considered that for sand it is sufficient to determine a value of
the angle of internal friction +. In the present analysis only deformations are to be con-
sidered, and in the same way, a single parameter determining deformation behaviour is
required. During the discussion of possible material idealizations, both the stress and
deformation parameters are considered so that the generality of the idealization adopted is
clear.
Perhaps the simplest material idealization is that the soil is a perfectly plastic material
which yields when the Mohr-Coulomb failure criterion is obeyed. For dry sand the Coulomb
criterion states that yielding will occur when
t=ssin+ . . . . . . . . . (1)
where 4 is the angle of internal friction of the soil and

compressive stresses and strains being taken as positive. The normality condition of the
64 R. G. JAMES AND P. L. BRANSBY

theory of pksticity requires that at yieId the plastic strain increment vector be norma to
the yield surface and thus
ir= -3sind . . . . . . . . . . . . (2)
where ir = i,+i, = <,+I?,
p = &-<a = [(ix - dJ2 + (jXY)2]l2
Defining the angle of dilation (Bent Hansen, 1958) as v=sin-l( -G&f, i.e. v is positive for
expansion of the material, equation (2) may be restated as
v=$ . . . , . . . . . - (3)
Unfortunately the observed behaviour of soil does not correspond with this elegant and con-
venient idealization. Soils are found to dilate at a rate considerably less than that required by
equation (3) and thus the choice of this idealization is unreasonable for an analysis in which
soil deformations are of paramount importance.
A second convenient idealization, adopted by Brinch Hansen (19533 for example, is that
of an idealized material which yields at the Mohr-Coulomb failure condition (equation (1)) but
which deforms such that
v=o . . . . . . 0 . . * (4)
It must be noted that the normality condition is not satisfied and thus that the upper and lower
bound theorems of the theory of plasticity may not be used. The constant volume condition
implied by equation (4) is valid where the soil is effectively undrained, or where the deforming
mass of soil is at the critical void ratio state (Roscoe et al., 1958). However, the idealization
is not adequate for analyses of retaining walls in a mass of sand which can and does suffer
significant volumetric strain during deformation.
It is therefore necessary to consider a material idealization which allows deformation such
that neither equation (3) nor equation (4) is obeyed. Bent Hansen (1958) proposed a material
which yields with angle of internal friction 4 and deforms with angle of dilation v where d;# v.
This idealized material is not perfectly plastic as normality does not hold, and thus theorems
of plasticity are not relevant. The idealisation is more flexible than the two discussed pre-
viously and thus can be arranged to give better agreement with the observed soil behaviour.
The applicability of this idealization for dense and loose sand subjected to passive earth
pressure is examined subsequently.
It may be noted that if 4 and v were not considered as soil constants, but ahowed to vary
throughout deformation, any observed stress ratio-shear strain-volumetric strain relation-

Expmion
t

Shear rtrrin
(4
* Fig. 1. (a) Stress ratio, (b) volumetric
(7) strain-shear strain curves from a
typical drained test on dense sand
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 65
ship (Fig. 1) could be approximated by assuming that $ and v were constant over a series of
small increments of shear strain y. Thus is could be taken that C$=c,& and v = v, (see Fig. 1)
over the increment of shear strain Ay, while 4 = &,, and v = v, for an increment of strain of Ay,.
Thus the study of an idealized material which has 4 #v seems to have great potential.

EXPERIMENTAL EVIDENCE FOR MATERIAL IDEALIZATION

Ex$erimental techniques
The experimental data discussed were obtained from tests in which a plane rough wall was
rotated about its top or its toe into a mass of dry sand with a level unloaded surface. The
sand was contained in a large narrow glass-sided tank and was constrained to deform in plane
strain. The wall extended the whole width of the tank (7i in.) and was 12-13 in. high.
Strains in the sand mass were determined by an X-ray technique (Roscoe et al., 1963).
Lead shot markers were embedded in the sand so as to form a regular grid and radiographs
were taken of the sand sample containing the markers at successive increments of wall rota-
tion. The positions of the images of the markers on the radiographs were measured and the
magnitudes of the strain increments in the element of sand defined by each triangular mesh of
lead shot were calculated for each increment of wall rotation. Images of rupture surfaces also
appear on the radiographs of dense sand samples. Dense sand dilates during shear and, as a
consequence, zones of high deformation show up as dark regions on the radiographs.
The maximum width of the tank is limited by the maximum penetrative power of X-rays
produced by the machine available, while the height of the wall must be sufficient to allow
adequate coverage of the strain field by triangular meshes of shot of the minimum size that
can be measured accurately. Thus determination of the strain field requires that the wall
should be as high as possible, but the sand sample should be narrow. These requirements
imply that the wall height to width ratio will be high. It is therefore possible that friction on
the gIass side plates will have some effect on the measured wall stresses. However, the simi-
larity between the data of Schofield (1959) and those of Arthur (1962) and James (1965) is
encouraging. Schofield did experiments in which a rough wall 6 in. high was rotated about its
top into a mass of sand with a level surface. The wall was made in three independent sections
(total width 14 in.), but measurements were taken only from the central section. Arthurs
tests were on a wall 6 in. x 6 in. which was contained in a glass-sided tank 6 in. wide while
Jamess data were obtained from a wall 13 in. high by 74 in. wide inside a glass-sided tank

0.6

+0.02 L
Fig. 2. Stress ratio-shear strain relationships
obtained from drained simple shear appara- Fig. 3. Volumetric strain-shear strain relation-
tus tests at constant normal load on ships for simple shear apparatus tests shown
Leighton Buzzard sand (Cole, 1967) in Fig. 2
3
66 R. G. JAMES AND P. L. BRANSBY

7$ in. wide. All three workers observed that the point of action of the maximum force on the
wall was at approximately 5 of the wall height from the toe of the wall, while the rupture sur-
face mechanisms observed by James and Arthur were similar to that which could be inferred
from Schofields data.
The large height to width ratio of the wall is a necessary sacrifice that must be made so that
strain data may be obtained. However, it seems from the limited data available that the
main features of behaviour are the same whether tests are performed with wide or narrow
walls.
Wall stresses are measured by a number of earth pressure cells of the type described by
Arthur and Roscoe (1961). These cells measure the magnitudes of the normal and shear
forces on the cell as well as the eccentricity of the normal force.
The sand used in the experiments is a rounded coarse Leighton Buzzard sand (nominally
lying between no. 14 and no. 25 B.S. sieves) which has a maximum voids ratio of O-79 and a
minimum voids ratio of 0.49. Typical stress ratio-shear strain-volumetric strain data
(Cole, 1967) from drained (constant normal load) tests on loose and dense samples of sand in the
simple shear apparatus are shown in Figs 2 and 3. The specific gravity of the grains is 266.
Stress and strain data obtained from the earth pressure experiments are given after dis-
cussion of the theoretical velocity fields, but full analyses are given by James (1965) and
Bransby (1968).

The angle of dilation v (i.e. sin-r (-$1~)) for elements of sand may be determined from the
slope of the v against y relationship for the sand during deformation. Data obtained from a
test in which a rough wall was rotated about its toe into dense sand are given as the experi-
mental points in Fig. 4. The line AA is drawn to represent the trend of the data. Lines BB
and CC were obtained from data of tests in which the wall was rotated about its top into dense
sand and, despite the different modes of wall rotation in the different tests, lines AA, BB and
CC are nearly coincident. The data from all tests indicate that, after an initial small strain,

Test PH
ep = 0.77
Ho .IZ.I"

0.06i

Fig. 4. Relationship between volumetric Fig. 5. Mean volumetric strain-shear


strain and shear strain observed in strain curva for elements of sand near
sand elements when a rough wall is a rough wall rotated about its toe into
rotated about its top or its toe into loose sand
dense sand. Test data from Bransby
(1968), James (1965) and Arthur (1962)
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 67

mental)

strain

(a)

Fig. 7. Velocity projections


on the slip lines
cz direction
-
.
CY

.
l, direction

\ B direction

(b)
Fig. 6. (a) Mobrs circle of strain increment,
(b) directions in the physical plane for a Fig. 6. Portion of the slip
material which dilates at Y line field

the slope of the v against y relationship for the dense sand is approximately constant with
v=20.
Data obtained from a test in which a rough wall was rotated into loose sand are given in
Fig. 5. The slope of the v against y curve varies for shear strains up to y = 0.05, but thereafter
the curve is approximately horizontal, i.e. v =O.
Further analysis and discussion is given by James and Bransby (1970) but it seems that
the v constant approximation is a reasonable basis for the derivation of theoretical velocity
fields.

VELOCITY FIELDS FOR v CONSTANT MATERIAL

Equations fog the velocity Jield

The equations governing velocity fields for v constant material are considered and the
pattern of the chosen velocity field is described.
The Mohr circle of strain increment for the material dilating at v is shown in Fig. 6(a),
while the corresponding directions in the physical plane are shown in Fig. 6(b). The two
directions of zero linear strain are indicated by the CYand /3zero extension directions, while the
angle of inclination of the p zero extension direction to the x axis is denoted by #. Consider
the total velocity1 V at any point P in the velocity field (Fig. 7). It is convenient to denote

1 Behaviour is assumed to be independent of time but, following the conventions of plasticity theory,
displacements are interpreted as velocities.
68 R. G. JAMES AND P. L. BRANSBY
the orthogonal projections of the velocity vector at P on to the a: and ,6 zero extension lines
by Va and V, respectively. The signs of the velocity projections are chosen so that an anti-
clockwise rotation of the positive direction along the CIline through (n/Z)+V transforms it into
the positive direction along the /3line.
It may be shown (Shield, 1953; Roscoe, 1970) that the relationships between V, and VD
along either zero extension direction which ensure that the directions are directions of zero
linear strain are
dVa - (V, tan v+ V, set v) d# = 0 along an CIline . . . . (5a)
dV, + (V, set Y + V, tan v) d# = 0 along a /3line . . . . (5b)
These equations are especially convenient for the discussion of velocity fields in which at
least one family of zero extension lines is straight and for simplicity discussion in this Paper is
confined to such fields. Shield shows that a line of discontinuity in the velocity field must
coincide with either the CIor the /3zero extension directions and thus for brevity the CIand /3
directions are subsequently described as slip lines. Shield also shows that the change in
velocity across a slip line is inclined at an angle v to the slip line. Both statements may be
confirmed quickly for the simple cases treated here. Consider a portion of the velocity field
where adjacent /3slip lines (& /3,and p3) are cut by a series of straight CC lines (al, cc2and +,
Fig. 8). Equation 5(a) is used to see whether the /3lines can be lines of discontinuity, i.e. slip
lines, in the velocity field. As the CLlines are straight, i.e. d$ = 0 along each c(line, equation 5(a)
reduces to dV, = 0. Thus so long as V, is maintained constant along say the c(~line, the total
velocities at points 1, 2 and 3 on lines ,&, /I2and /3amay be different. The Vu constant con-
dition implies that any increment of total velocity between points 1,2 and 3 on line CL~ must be
perpendicular to the CL~ line, i.e. inclined at v to the fiZslip line. The fact that the discontinuity
in velocity between points on either side of the slip line is inclined to the slip line means that
the transition layer between the regions on either side of the slip line must be appreciably
thick. This follows because the material in the transition layer must supply the volume
change associated with the component of the velocity discontinuity normal to the slip line.

Form of velocity jield


The choice of the pattern of the slip line field is a matter of judgement and experience.
However, it is arbitrary so long as the specified boundary conditions can be accommodated,
but the reliability of any predictions made naturally depends on the relevance of the chosen
slip line field to the particular boundary value problem considered. A slip line field (Fig.
9(a)) of a form similar to that of the stress fields of Prandtl and Sokolovski for cohesionless
weightIess soil was selected for the retaining wall problem. This slip line field has the advant-
age that the slip lines are curved and of the general shape that would be expected in the passive

Fig. 9 (a). General slip line field Fig. 9 (b). Slip line field after wall boundary
condition is taken into account
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 69
failure of a rough retaining wall, but the field is reasonably simple mathematically. Of course
it would be possible to select more complex velocity fields, but it is felt that additional com-
plexity is not justified until the possibilities of the simple velocity field have been exhausted.
It is assumed that the wall is perfectly rough and that sand immediately adjacent to the
wall moves so that there is no relative velocity between the wall and the sand. This assump-
tion seems a reasonable approximation as the face of the wall is covered with sandpaper, and
it has the consequence that the wall is a line of zero extension. Zone 1 of the field therefore
vanishes and the velocity field reduces to that shown in Fig. 9(b).
It is not necessary that this particular assumption about the roughness of the wall should
be made, although it is convenient and the data of Bransby (1968) show that it is reasonable
for one mode of wall rotation (i.e. about toe). Alternatively it could be assumed that the wall
was perfectly smooth or that the inclination of the slip lines at the wall was determined by a
function of 6 and 4 (Sokolovski, 1965).
Both families of slip lines are straight in region 3 and are of the pattern of the stress
characteristics in a Rankine zone. In region 2, a zone of radial shear, the p lines are straight
and pass through 0, the top of the wall, while the CIlines are logarithmic spirals with pole 0.
It is assumed that the sand remains rigid outside the lowest C(line ABC. As the /3 lines are
straight in both regions d#=O along all /? lines and so equation 5(b) requires that dVO =0
along each /3 line. However, V, = 0 at points on ABC and so V, = 0 everywhere, i.e. the velo-
city is everywhere perpendicular to the /3 lines. This fact may be used to simplify equation
5(a) into a form which is used later. Since the total velocity V is everywhere perpendicular
to the p lines, Va may be written as
I, = Vcos v . . . . . . . . (6)
Taking note that VIg= 0 everywhere, substitution of equation (6) in equation S(a) gives
dV--(Vtanv) d# = 0 . . . . . . . . (7)
Let V on the CCslip line at the general point D on the wall be VW. Consider the variation of
the magnitude of V along the CIline DEF in the zone of radial shear. Equation (7) may be
integrated as
V = K exp (# tan V) . . . . . . . . (8)
where K is a constant determined by the condition that V= V, at D. Thus on the a line DEF
in the zone of radial shear
I/= V,exp[($+@tanV] . . . . . . . (9)
where 0 is the inclination of the wall to the vertical as shown in Fig. 9(b). For point E, on the
edge of the zone of radial shear, equation (9) gives
Vn = I/, exp [(&++e) tan V] . . . . . . (10)
The knowledge of VIEgiven by equation (10) allows the value of V along the CIline EF in zone 3
to be established. The C(lines are straight in zone 3, i.e. d#= 0. Therefore along EF equation
(7) reduces to dV =O, and thus the total velocity of points along EF is constant at a value given
by V=V,.
The meaning of equation (9) can be interpreted if the radius from 0 of any point P on the
CYline is denoted by Y (see Fig. 10) and the radius of point D on the wall by rW. Then, because
DE (Fig. 9(b)) is a log spiral, for all points on DE
y =r,exp[(#+8)tanV] . . . . . . (11)
Substitution of equation (11) in equation (9) gives
v
-= v
_w . . . . . . . . . (12)
Y rw
70 R. G. JAMES AND P. L. BRANSBY

This implies that, as the velocity of any point on the a: line is proportional to its radius from 0,
the 01line rotates as a rigid body about 0.

Velocity fields for three modes of wall dis@acement


The velocities of points on one 01line have been established in terms of the velocity of the
point at the intersection of the Q!line and the wall, and the variation of the velocity through-
out the mass may now be determined. Three separate cases are considered: rotation of the
wall about its toe, rotation of the wall about its top and translation of the wall.

Rotation of the wall about its toe. The velocity of particles of sand immediately adjacent
to the wall is perpendicular to the wall and of magnitude
VW = (H-r&j * . . . . . . . (13)

where H is the wall height (length OA in Fig. 9(b)), 8 is the angular velocity of the wall and Y,
is the radius from 0 of any point on the wall.
Consider a point P at a radius Y in the zone of radial shear lying on a p line inclined at 4.
Suppose that P lies on a typical 01line, which cuts the wall at point D at a radius rWfrom 0,
while the velocity at D is VW. Three relationships hold. First equation (9) gives the velocity
at P in terms of VW. Second V, may be evaluated from equation (13) so long as Y, is known.
Third equation (11) gives rWin terms of Y. Thus the velocity at P may be determined as
V = V, exp [($+B) tan v] = (H-r,)4 exp [(#+@) tan v] . . (14)
V = He exp [(I/+(?) tan v]-rB . . . . . . . . . . (15)
Equation (15) establishes the variation of the total velocity V in terms of the co-ordinates 7, ~5
in the zone of radial shear. The velocity of points on the edge of zone 3 (line OB, which is
inclined at angle 4,) is given by
V, = H@ exp [(#B-!-@) tan v]-& . . . . . (16)
The velocities of all points on one c( line in region 3 are constant so equation (16), which defines
the boundary velocities, also gives the velocities of all points in region 3.

Rotatiolz of the wall abogt its top. The assumed velocities of sand particIes immediately
adjacent to the wall are given by
v, = 73 . . . . . . . . . (17)
where 4 is the angular velocity of the wall. For convenience the inclination of the wall to the
vertical when it rotates about its top is defined by the angle d, where 0 is taken as positive for
rotation into the sand. Using equations (9) and (11) as before, the velocities everywhere in
region 2 are obtained as
V = -r,@ exp [($-et) tan V] = rB . . . . . (18)
Thus the entire region rotates about 0 as a rigid body and there is a discontinuity in velocity
across line AB. In consequence a transition layer forms between the moving and stationary
blocks of sand separated by AB. The 0: velocities in region 3 are constant along each a line
and are defined by the a velocities at line OB
VB = r& * . . . . * . . . (19)

T7a~slat~on of the wall. The velocities of sand particles immediately adjacent to the wall
are given by
VW= jJ . . . . . * . . . (20)
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 71

where 9 is the horizontal wall velocity. The velocities in the zone of radial shear may be
obtained as
V = jl exp [(#) tan V] . . _ . . . . (21)
The magnitude of the velocity is constant everywhere on OB (as V is independent of Y) and so
the velocity in region 3 is perpendicular to the ,6 lines and everywhere of magnitude
V, = 3 exp [(&) tan V] . . . . . . . (22)
Equation (22) implies that there is a discontinuity in velocity across ABC varying from 9 at
the wall to j exp (& tan V) in region 3, and thus that there is a transition layer along ABC
between moving and still blocks of sand.

STRAIN FIELDS ASSOCIATED WITH DERIVED VELOCITY FIELDS

The strain fields associated with the derived velocity fields are of little importance when
discussing the deformation of a perfectly plastic material, for stresses are independent of strain.
However, in soil the magnitudes of the strains are of great interest both because the stress ratio
varies with strain and because rupture surfaces will form only along regions of high strain. A
derivation of the strain fields associated with the three velocity fields will allow predictions to
be made of the different patterns of the rupture surfaces in the three cases.
Strain fields will be derived for three zones-zones 2 and 3 and the transition layer-for the
three wall displacement modes. Rotation of the wall about its toe is considered first. Polar
co-ordinates are used for the analysis of the zone of radial shear (Fig. 10) and the velocities in
the r and 3 directions are denoted bv u and v respectively. When the wall rotates about its
toe it is known from equation (15) that
u=o
. . . . (23)
v = IId exp [(#+ 0) tan V]-re .

..
>
and so the strain rates may be calculated as

; T =-&LO

1
ar
av
i&X ----=
u -$titanvexp[($+@ tanv] (24)
r ra*
Pr$ = _zt_av+! = -He exp [(#+ 0) tan ~1
Y al/ ar Y Y
The volumetric and shear strain rates are then

ir = -ytanuexp[(#+U) tanv]
. . (25)
v =gsecvexp[(#+8)tanv]
Y

y Fig. 10 (left). Co-ordinate


system for analysis of
zone of radial shear

Fig. 11 (right). Co-ordi-


nate system for analy-
sis of zone 3
72 R. G. JAMES AND P. L. BRANSBY

Thus, in the zone of radial shear, the strain rates are inversely proportional to radius along
any one radial line.
The strain rates in zone 3 may be calculated most conveniently if new axes Bx and By are
drawn so that the x axis coincides with the CIlines (Fig. 11).
Velocities are constant along an u line in the zone at values given by equation (16). If R is
defined as the distance of any point on OB from B, equation (16) may be rewritten as
I, = Rd . . . . . . . . . (26)
since the length OB is H exp [(#e+ 0) tan ~1. Thus in region 3 the velocities u and v in the
direction of the new x and y axes are
u = R(Bcosv = ye
v = Rti sin v = ye tan v
. . . . . (27)
>
The strain rates are then
EX = 0
cy = -4tanV . . . . . . . (28)
Y.W = - e 1
and are constant everywhere in the region. The volumetric and shear strain rates are
d = -4tanV
. . . . . . . .
3 = 4 set Y >
There is no velocity discontinuity and hence no transition layer between sand above and
below line ABC when the wall rotates about its toe and thus the strain rates have been estab-
lished in all zones.
The strain field associated with rotation of the wall about its top is considered now.
Equation (18) shows that zone 2 rotates about 0 as a rigid body and thus strain rates are zero
everywhere within the zone. However, there is a velocity discontinuity along the Q line ABC
(Fig. 9(b)) and there must be a transition layer between moving and still blocks of sand which
suffers high deformation.
The approximate magnitude of the strains in any such layer can be estimated if certain
simplifying assumptions are made. Let the thickness of the transition layer be T and the
change in velocity across the layer, which is inclined at v to the layer, be U. Select co-
ordinates such that the x axis is tangential to the layer and they axis perpendicular to it. If
the strain rates are assumed uniform across the layer, they may be determined as
i, = 0
I
ty = -2 sinv
T . . . . . . .
I
--
u cos v
YXY =
T J
The volumetric and shear strain rates are
d = u smv
T
. . . . . . . (31)
p=;
1
Equation (31) may be used to determine the approximate strain rates in the transition
layer extending from the toe of the wall. It is assumed that the thickness of the transition
layer is known and is of constant magnitude. The discontinuity in velocity across the layer
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 73

varies from He at A (Fig. 9(b)) to He exp [(I,&- 0) tan V] along BC and so the shear strain
rate varies from
HI?
y=-?;-atA . . . . . . . . . . . . (32)

to
+H@
T exp [(&- 0) tan V] along BC . . . .

The strain rates are constant throughout region 3 at values given by


ti = &tan v
li = psecv . * . . . . . . (34)
>
Thus for wall rotation about the top, for which the width of the transition layer is con-
siderably less than the wall height, equations (32) and (34) show that the strain rates in zone 3
are much less than those in the transition layer. This behaviour may be contrasted with that
associated with rotation of the wall about its toe, where there is no transition layer and the
strain rates are large in the zone of radial shear.
The case of wall translation is now considered. Strain rates have to be determined for
three regions but, as for the other two cases, it is found that the strain rates are zero in one
region. The strain rates in the zone of radial shear may be determined in terms of the hori-
zontal velocity of the wall as

5=-j- tan v
r exp [* tan V]

. . . . . (35)
p = jyexp [#tanv]
1

It is interesting that the strain rates derived for wall translation are identical to those pro-
duced by wall rotation about the toe at a velocity 6 = j/H (see equation (25)). This fact is
evident when it is remembered that wall translation can be obtained by the superposition of
wall rotation about the toe and rotation about the top, and that the latter condition produces
no strain in the interior of the radial zone.
The shear strain rates in the transition layer vary from

v=satA ............ (36)

to j=$exp[&tanv] alongBC ...... . (37)

The velocity of every particle in region 3 is constant (equation (22)) and thus the whole region
moves as a rigid body.
The strain rate field for wall translation is thus a combination of those fields derived for
wall rotation about the top and about the toe and it shows the features of both. This is
important in discussing the observed rupture surfaces and the predicted strain fields and wall
stress distribution.
Equation (34) for wall rotation about the top states that the material within zone 3 con-
tracts during deformation and thus the assumption concerning the sand behaviour is violated.
Hence, although the velocity field is found to be satisfactory for wall translation and for rota-
tion about the toe, it is unsatisfactory for rotation about the top. In reality the velocity field
observed in the experiments when the wall is rotated about its top is more complex than that
assumed here and so far no satisfactory simple theoretical idealization has been found.
However, the incompatibility in the assumed field is in a region of low strain and so it may not
represent a serious deviation from the real behaviour. It is therefore worth while continuing
74 R. G. JAMES SND P. L. BRANSBY

the development for wall rotation about the top alongside the other two cases, with the proviso
that the results be treated with caution.
Before the strain rate fields for the three cases are compared with the experimental data
it is necessary to examine the regions of applicability of the strain rate fields derived. The
fields have been developed with the inherent assumption that the variation of the strain rate
in the sand mass is continuous, This assumption is normally valid for loose sand, but it is
well known that a continuous mass of dense sand will degenerate into quasi-rigid blocks
separated by rupture surfaces soon after the shear strain in the mass exceeds that required to
cause the sand to reach its peak stress ratio condition. The derived velocity field will not
necessarily apply for this second regime. Thus in dense sand the predictions of the velocity
field can strictly be tested only at small wall displacements. However, the velocity fields
can be used to give qualitative indications of behaviour at larger strains and to predict the
positions and patterns of the rupture surfaces.

OBSERVED RUPTURE SURFACES COMPARED WITH PREDICTIONS FROM VELOCITY FIELD


The formation of rupture surfaces in dense sand may be observed by the X-ray technique.
Narrow bands of sand suffer gross shear deformation and as a result dilate sufficiently for the
change of density to become evident on the radiographs. Radiographs are available for the
three modes of wall movement discussed and the predictions are compared for each case. The
interpretation of equations (23)-(37) is discussed and the predicted patterns of rupture surfaces
are established.
A typical stress ratio-shear strain relationship for dense sand is shown in Fig. 2. It is ex-
pected that a mass of dense sand will strain reasonably continuously so long as no region in the
mass sustains a shear strain greater than that required to reach the peak stress ratio condition
(point B). However, as soon as strains increase beyond the peak stress ratio condition, sand
within the rupture surfaces will move towards condition C and at large strains will approach
the critical state condition. Sand immediately adjacent to the rupture surface will have
unloaded from the peak condition to some state D. There is now a discontinuity of strain and
deformation across the rupture surface.
Rupture surfaces will therefore form in regions of high strain and will be orientated along
those directions in the velocity field across which discontinuities of velocity can occur, i.e. zero
extension lines. This argument from considerations of the velocity field alone is confirmed by

Fig. 12 (left). Predicted rupture surface mechanisms in dense


sand for wall rotation about its toe

Fig. 13 (below). Rupture planes observed on radiographs


exposed at three stages in a test in which a rough wall was
rotated about its toe into dense sand
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 75
the finding of James (1965) that rupture surfaces form along the zero extension directions and
not along the stress characteristics of conventional analysis. (This finding of James is also
supported by the data of Bransby, 1968).
Equations (23)-(37) give the strain rate field consequent on any wall velocity. For sim-
plicity, it is assumed that this field is maintained throughout deformation. Thus equations
derived in terms of the incremental wall velocity (4, 9) and incremental strains (ti, p) may be
rewritten in terms of the total wall displacement (0,~) and the total strains (v, y). The
experimental data confirm that this simplifying assumption is reasonable until rupture planes
form.
The case of wall rotation about its toe is considered first. Equation (25) states that the
shear strain rate (and hence the shear strain) is inversely proportional to radius from the top
of the wall. In consequence rupture surfaces will be formed close to the top of the wall after
small wall rotations. Then as wall rotation continues the high strain region will extend down-
wards and rupture surfaces will be observed lower in the mass, i.e. the order of formation of the
rupture surfaces will be as shown in Fig. 12. A rupture surface will form only along the CC line
(ABC, Fig. 9(b)) extending from the toe after large wall rotations.
The observed positions of the rupture surfaces at three stages of a test on dense sand
(e, = 0.51) in which the wall rotates about the toe are shown in Fig. 13. The data confirm that
the overall pattern of behaviour predicted by the velocity field occurs. It is interesting that
rupture surfaces are observed along both families of zero extension lines. No rupture surfaces
are observed near the toe of the wall, even after 7 of wall rotation.
The velocity field suggests that region 2 rotates as a rigid body when the wall rotates about
its top. However, significant strain is expected in the transition layer between moving and
still blocks of sand. The field also predicts deformation in zone 3 but it is now shown that the
strain in this zone is considerably less than that in the transition layer. It is assumed that the
width of the transition layer T is H/3. Equation (32) indicates that in the transition layer
near the toe of the wall the shear strain y at any angle of wall rotation 8 is

y = H; = 38 . . . . . . . . (38)

However, the predicted strain in zone 3 is I


y= 0secv . . . . . . . . * (39)
so that y = 1.068 for v = 20. Rupture surfaces should therefore be observed along the transi-
tion layer (as shown by the continuous line in Fig. 14) but no rupture surfaces are expec-
ted in the interior or zones 2 or 3. A typical rupture surface mechanism observed by James
(1965) is shown as the broken line in Fig. 14 and its location is in reasonable agreement with

Fig. 14. Predicted rupture surface mechanism Fig. 15. Predictedrupture surface mechanism
and an observedrupture mechanism(James, in dense sand for wall translation
1965) in dense sand for wall rotation about
the top
76 R. G. JAMES AND P. L. BRANSBY

that predicted by the velocity field. The observed angle of breakout of 35 to the horizontal
in Fig. 14 is in agreement with the theoretical value of [45 - (v/2)] with Y= 20. The classical
Rankine theory would require a breakout angle of [45 - (+/2)] = 20.5 for Leighton Buzzard
sand with a peak $ of 49 (see Fig. 14).
The velocity field for wall translation predicts that both rupture surface mechanisms will
occur. Equation (35) indicates that strain in the zone of radial shear is inversely proportional
to radius from the top of the wall, while equation (36) indicates that considerable shear strain
occurs in the transition layer. Thus rupture surfaces should form at an early stage of the test
near the top of the wall and also near the toe of the wall along the transition layer. As wall
displacement increases the rupture surface mechanism in the zone of radial shear is expected
to extend downwards from the top of the wall. The expected pattern and order of formation
of the rupture surfaces is shown in Fig. 15. A typical radiograph obtained by Lucia (1966)
when he translated a rough wall (H =4 in.) into dense sand is given in Fig. 16. It is clear that
both mechanisms occur. Previous radiographs show that the rupture surface extending
from the toe of the wall formed first, while rupture surfaces in the zone of radial shear appeared
later. Unfortunately neither Lucia (1966) nor May (1967) measured strains in the sand mass,
and so a detailed description of this unusual deformation mechanism cannot be given.
The predictions of the location and sequence of the rupture surfaces in the failure mechanism
could be made only after the strain fields associated with different boundary displacements
had been developed, for the basic slip line pattern (and hence the shape of the rupture surfaces)
is assumed to be the same for all three cases. Thus for soil mechanics problems it is not
sufficient to establish only the slip line patterns; the strain and the strain rate fields have an
equal or a greater significance.

COMPARISON OF OBSERVED AND PREDICTED SHEAR STRAINS

Experimental data are available to check the detailed predictions of the magnitudes of the
strains for two of the three cases considered. Bransby (1968) gives data from an investigation
of the rotation of a wall about its toe into loose and dense sand, while James (1965) gives similar
data for the rotation of the wall about its top into dense sand.
The strain data are normally presented in diagrams showing contours drawn through all
points which suffer a certain value of strain. Contour maps are drawn for shear and volu-
metric strains and strain increments and it is impractical to present all the data here. Three
contour maps have been chosen to check the velocity field predictions, but many more data
are given by James (1965) and Bransby (1963).
The cumulative shear strain contours observed at 8=7 when the wall rotates about the
toe into loose sand are shown in Fig. 17(a). (E ar1ier incremental data indicate that there is
little change in pattern of the strain rate field during wall rotation.) The sand sustains
intense shear deformations, especially near the top of the wall, with no tendency for rupture
surfaces to form.
Equation (25) predicts that strain inside the zone of radial shear will be inversely propor-
tional to radius from the top of the wall but, if it is assumed that the sand is deforming at
constant volume (v=O), the strain will be constant at any one radius. However, for zone 3
equation (29) indicates that the strain will be constant and of small magnitude.
The predicted contours for a v=O material at 0=7 are shown in Fig. 17(b). A glance
shows the differences rather than the similarities between the predictions and the observed
data, but it is claimed that the predicted field gives a reasonable indication of the positions of
the contour lines. Most of the irregularities of the observed contours are due to small experi-
mental errors in the determination of the strain, and these errors become exaggerated when
unsmoothed contours are drawn through the data.
The major differences between the predictions and the data are in the interior of zone 3 and
Fig. 18 (a). Shear strain contours at 9 = lo in
a test in which a rough wall rotates about
its toe into dense sand

Fig. 18 (b). Predicted shear strain contours at


9 = 1 when a rough wall rotates about
its toe into dense sand

Test LE(Jmer. 1965)

c9'=,4~

eo 0.55

& = 13"

Fig. 19 (a). Contours of shear strain from a


test on dense sand in which the wall was
rotated about the top

Fig. 19 (b). Predicted strain field at 9 = l-G0


when a rough wall rotates about its top into
dense sand
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 77
along the edge of the zone of radial shear. These differences indicate some of the inadequacies
of the simple velocity field that has been selected and could be used as a basis for the develop-
ment of a more complex velocity field.
A similar comparison can be made between the predictions for and the observations of the
behaviour of dense sand. It is taken that dense sand deforms with v=20. Comparisons
are made using data obtained at an early stage of the test (8= 1) when the deformation is
reasonably continuous. The shear strain contours obtained when the wall was rotated about
its toe into dense sand are shown in Fig. 18(a) and the predicted contours are shown in Fig.
18(b). The same general comments as made for the comparison for loose sand apply: there
are discrepancies between the observations and the predictions but the main features of
behaviour in the zone of radial shear are predicted reasonably well.
The thickness of the transition layer must be assumed before any strain predictions may be
made for rotation of the wall about its top. It is assumed that T = H/3. The predicted strain
field is shown in Fig. 19(b) and the experimental data at 0= 1.4 for a test on dense sand are
given in Fig. 19(a). The agreement between the prediction and the data depends critically on
the choice of the width of the transition layer and on the assumption that the strain is uniform
across the layer. The observed variation of strain in the sand mass is more complex than the
simple pattern given by the assumed velocity field and part of this additional complexity is
caused by the strict compatibility requirements that must be met by the observed strain field.
However, there is a marked resemblance between Figs 19(a) and 19(b) and they both differ
from those for the rotation of the wall about its toe (Figs 18(a) and 18(b)).
The most striking prediction of the velocity field is that, when the wall rotates about the
toe, strains in the zone of radial shear are inversely proportional to radius. The accuracy of
this prediction can be assessed if the magnitude of strains at points along a radius at 45 to the
horizontal are plotted against radius. The data from tests on loose and dense sand (Figs
20(a) and 21(a)) indicate that strains decrease rapidly with increasing radius, as predicted.
A more rigorous check can be made when I/y is plotted against r and the predicted straight line
relationship with a cut off at the edge of the zone of radial shear is plotted. Fig. 20(b) indi-
cates that the loose sand mass deforms so that the magnitude of the strain is nearly inversely
proportional to radius, and that the magnitudes of the strains are close to the predictions.
The agreement is not so good for dense sand (Fig. 21(b)), but even so there is a straight line
relationship between l/y and r for r < 7 in.
The corresponding data obtained from rotation of the wall about its top into dense sand
are shown in Fig. 21 (a) for 0= 1.4. The agreement between measured and predicted strains
is not so good for this mode of rotation, presumably partly as a result of the incompatibility
of the assumed velocity field, but the data indicate the wide differences of strain behaviour
between the two tests.

QUALITATIVE PREDICTIONS OF WALL STRESS DISTRIBUTION

It is considered whether the predicted strain fields give any indication of the probable wall
stress distributions when the wall is displaced in the various modes. The discussion is
restricted to a consideration of the behaviour of dense sand (although the argument could
easily be extended to loose sand) since the changes of stress ratio during deformation (Fig. 2)
are dramatic and have considerable influence on the shape of the measured stress distributions.
Elements of sand at different points on the stress ratio-shear strain curve for dense sand
are classified into three groups for convenience of discussion. Elements of sand are considered
to be pre-peak when the strains are small and the stress ratio in the sand is low, i.e. for
position OA of the curve in Fig. 2. Elements of sand are taken as being at peak stress ratio
conditions for positions AB of the curve, and after B the sand is post-peak and the stress ratio
60 c

Fig. 20 (a). Shear strain of points on a radius at 45 to the Pig. 20 (b). Reciprocal of shear strain on radius at 45 to
horizontal from the top of the wall at 8 = 7 in test PH the horizontal from top of wall at 8 = 7 in test PH
(e, = 0.77) when the wall rotates about the toe (e, = 0.77) when the wall rotates about the toe

0.06'

1
0.04
t

Pig. 21 (a). Shear strain at points on a radius at 45to the Pig. 21 (b). Reciprocal of shear strain at points on radius
horizontal from the top of the wall in test PE (e, = 0.51) at 45 to the horizontal from top of wall in test PE (e, =
at 8 = I for rotation about the toe and at 8 = 1.4 for O-51) at 0 = 1 for wall rotation about the toe
rotation about the top (test LE, e, = O-55)
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 79
TOP -> Normal
stress

b
Fig. 22. Expected wall stress distribution at two
stages of a test in which a rough wall rotates
about its toe into dense sand

(b) Later in thetest

has dropped considerably from its peak value. These distinctions are not precise, but serve as
a basis for discussion of the sand state at different points in the mass.
Two observations enable the probable wall stress distribution to be determined. First it
is known that a small increase in 4 of the soil behind the wall causes the magnitude of the pas-
sive force on the wall to increase considerably. Second it is assumed that the maximum stress
that can occur at any depth on the wall is proportional to depth. This does not imply that
the maximum stress is observed simultaneously at all points on the wall. The deformation
conditions might be such that elements adjacent to some points on the wall are unstrained and
that the stress ratios, and hence stresses, are low in these regions. When all sand is deforming
at its peak stress ratio and when wall friction is mobilized at 6 everywhere on the wall, the
maximum stress is observed at every point and the total force on the wall is a maximum.
The situation corresponds to that considered by classical analysis, and the assumption of
linearity is valid. Bransby (1968) investigated the magnitudes of the peak wall stresses
measured at different depths on a wall when it rotates about its toe and concludes, for a situa-
tion in which a considerable zone of sand suffers intense deformation, that the peak stresses are
approximately linear with depth, although the peak stresses at different depths are measured
at different wall rotations.
Rotation of the wall about its toe is considered first. The predicted strain field (e.g. Fig.
17(b)) suggests that strains will be very high near the top of the wall at an early stage of the
test, but will be much smaller at greater depths in the zone of radial shear. (The strains are
small in zone 3.) It is expected therefore that, for small angles of wall rotation, the sand near
the top of the wall will be at its peak condition while sand lower in the mass will be pre-peak.
As a consequence high stresses will be measured on the top of the wall and proportionally
lower stresses will be measured near the toe of the wall as shown in Fig. 22. At the later stage
of the test, strains in the sand near the top of the wall will be so great that the sand will be in
its post-peak condition and + will have fallen below its peak value. Thus the wall stresses on
the top of the wall will also have decreased below their early peak values (Fig. 22). However,
sand lower in the mass will have achieved the peak stress ratio condition and so the peak stress
will be measured at that depth. At greater depths the sand will be pre-peak and 4 will be low
and so the wall stresses will be substantially less than their peak values. Stress distributions
measured at two stages of the test (Fig. 23) are in fair agreement with the conclusions of this
rather qualitative discussion, and the data plotted in Fig. 24 indicate that there are wide dif-
ferences in angle of wall rotation when the peak stresses are measured for different points on
the wall.
The velocity field indicates that the failure of the sand in the zone of radial shear is a pro-
gressive phenomenon which continues as wall rotation increases. Two likely consequences
are suggested. First the maximum force on the wall is high because the whole mass contri-
butes to this force. Second a Iarge wall rotation is required to reach the peak loads for the
strain in the body of the mass to be sufficient for the sand to be at its peak stress ratio.
80 R. G. JAMES AND P. L . BRAN S #BY

- Sand
surface

a-

Stress : Ib/sq. in.


I I
15 25 35 45 55
0c7p"

Fig. 23. Measured normal stress distribution on Fig. 24. Angle of wall rotation (0,,) re-
centre line of wall at two stages in test PE quired to mobilise the peak normal
(e. = 0.51, Ho = 12.1 in.) when a rough wall stress on pressure cells at different
rotates about its toe depths on the wall in test PE

Peak moment is measured on a wall rotating about its toe into dense sand at 0=5i
(Bransby, 1968) while, when a wall of approximately the same height rotates about the top,
peak moment is measured at 0 = 1.6. The peak forces on the wall are also substantially
greater when the wall rotates about the toe.
The predicted variation of strain in the sand mass when the wall rotates about the toe
(equation (25)) indicates that the strain at geometrically similar points near walls of different
heights is dependent only on the angle of wall rotation; it does not depend on wall height.
The velocity field therefore predicts that 0 is the significant displacement parameter governing
the behaviour of walls rotated about the toe, and thus that the stress and strain distributions
caused by rotation of walls of different heights are similar at the same angle of wall rotation.
Data from a test on a wall half the height of the walls used in the main test programme tend
to confirm this hypothesis. It was found that the wall rotation required to achieve the peak
moment was the same for the large or small walls, while the peak stresses on geometrically
similar points of the large or small walls were measured at the same wall rotation (Fig. 24).
The rupture surface mechanisms observed in both cases were similar in shape and the mechan-
isms were observed to form at the same angle of wall rotation. Thus when the wall rotates
about the toe, the qualitative predictions of wall stress distributions that may be made from
the velocity field are confirmed by the experimental data.
The strain field associated with wall rotation about the top indicates that sand in the tran-
sition layer suffers high deformations, but that sand in the zone of radial shear is unstrained.
It is expected therefore that the stresses in the zone of radial shear will change little from their
initial K, values. Consequently stresses on the top portion of the wall will be small. How-
ever, the deforming mass of sand in the transition layer will transmit high stresses to the wall
near the toe. The expected normal stress distribution will therefore be of the form shown in
Fig. 25(a). The shape of the stress distribution should change little during the test as the
velocity field suggests that the sand in the zone of radial shear should remain unstrained.
However, the magnitude of the stresses will change as the sand in the transition layer passes
through its pre-peak, peak and post-peak states. A typical observed stress distribution just
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 81
before the peak moment is measured on the wall is shown in Fig. 25(b), and it is of a form con-
sistent with that suggested by the velocity field.
The apparent thickness of the transition layer can be estimated from the observed angle of
wall rotation eP at the instant the peak force is measured on the wall. It is assumed that the
peak force on the wall is measured when the sand in the transition layer is at its peak stress
ratio, and that this peak stress ratio occurs at a shear strain of yP = 0.15 for the Leighton Buz-
zard sand used in the experiments (Cole, 1967). Then, rewriting of equation (36) gives

T 2% . . . . . . . . . (40)
YP
James (1965) observed that 0; = 1.6 for H = 13 in. and so for his tests T,, = 2.4 in. or 0*19H.
The data of Schofield (1959) and Arthur (1962) obtained from tests on walls 6 in. high suggest
that 0; may be taken as 0bE2.4 for H =6 in. The apparent width of the transition layer is
then T, = 1.7 in. or 0.28H. It is interesting that the width is neither a constant proportion of
the wall height nor a constant value independent of wall height. A constant width of transi-
tion layer would imply that the wall rotation to peak load were inversely proportional to wall
height; the absolute displacement of the toe would then be the significant displacement para-
meter. In contrast, if the width of the layer were a constant proportion of the wall height, the
wall rotation to peak load on the wall would be independent of wall height and the angle of
wall rotation would then be the significant displacement parameter as found for rotation of
the wall about its toe. However, as neither of these simple hypotheses enables wall rotation
to peak for walls of different heights to be correlated, great care must be taken when extrapo-
lating model studies to the full size.
The measured wall stress distribution when the wall translates should be a combination of
the distributions already considered. It is expected therefore that the wall stresses will be
higher than those measured for either of the wall rotations considered, so long as the wall
height and the maximum wall displacement are the same in the three cases. The data presented
by Roscoe (1970) suggest that this is so only up to the attainment of peak load on the wall at a
displacement corresponding to that required to cause failure by rotation about the top.
Thereafter the simple velocity field will not be relevant.
The stress distribution on the wall before failure should be of the form shown in Fig. 26 but
the shape of the distribution varies as the test proceeds. Unfortunately, few experimental
data are available, but James (1967) suggests that the distribution is not linear and is of the
shape indicated in Fig. 26. The difference in angle of wall rotation required to mobilize peak

Normal
Normal
IW~SS
: Iblrq.in.

.__
(4 (b) Toe -

Fig. 25. (a) Expected distribution, (b) measured distri- Fig. 26. Expected wall stress
bution of normal stress when a rough wall is rota- distribution in a test in which
ted about its top into dense sand (e. = 0.55, a rough wall translates into
Ho = 13 in.) dense sand
82 R. G. JAMES AND P. L. BRANSBY

moment on the wall for the two wall rotations (rotation about top t$=1*6 for H=13 in.,
rotation about toe f?,= 5.5) suggests that the stress and strain fields associated with the forma-
tion of the rupture surface extending from the toe control the wall behaviour for all but the
smallest walls, especially as for this mechanism 0; decreases as H increases. The velocity field
suggests that the failure mechanism and the stress distribution are complicated functions of
wall rotation and wall height, and indicates that care must be taken in predicting the behaviour
of a full size structure from the behaviour of the model.

CONCLUSIONS
The data from tests in which a wall was rotated about its top or its toe indicate that the
simplifying assumption that the angle of dilation is constant is a reasonable approximation for
the deformation of both dense and loose sand.
Derivation of the slip line patterns and the velocity fields is not sufficient to predict the
behaviour of a sand mass; a study of the associated strain fields provides much extra informa-
tion as well as providing a check that the velocity field is satisfactory.
Although the slip line patterns are assumed to be the same, the strain fields associated with
the velocity fields for rotation of a wall about its top or its toe or for wall translation indicate
that considerable differences in behaviour are to be expected for the three modes of wall dis-
placement, while indicating that the simple slip line field is not strictly relevant for rotation of
the wall about the toe.
The velocity field enables predictions to be made of the location and sequence of the rup-
ture surfaces in the strikingly different failure mechanisms that are produced in dense sand
by the three modes of wall displacement.
The predicted strain fields are in moderate agreement with the observed strain fields in
regions of high deformation.
The measured wall stress distributions are consistent with those expected from the pre-
dicted strain fields.
The velocity field indicates that wall height may have a significant bearing on wall be-
haviour when the wall translates or rotates about the top. However, the velocity field indi-
cates that behaviour should be independent of height when the wall rotates about the toe.
The derived velocity and strain fields give great insight into the behaviour of walls which
are displaced in various ways.
It must be emphasized that the stress equilibrium equations have not been considered in the
analysis and so the displacement solutions are not exact, particularly as the choice of the velo-
city field was somewhat arbitrary and governed by the desire to keep the analysis straight-
forward. More complicated velocity fields could be developed and some account could be
taken of the stress equilibrium equations, but it is felt that this Paper gives a new approach
to retaining wall problems and good agreement with the published data.

ACKNOWLEDGEMENT
The work outlined was performed as part of the programme of research of the Cambridge
Soil Mechanics Group under the direction of the late Professor K. H. Roscoe. The Authors
wish to acknowledge the debt they owe to Professor Roscoe for his guidance and support.

REFERENCES
ARTHUR, J. R. F. (1962). Strains and lateral force in sand. Ph.D. thesis, University of Cambridge.
ARTHUR, J. R. F. & ROSCOE, K. H. (1961). An earth pressure cell for the measurement of normal and shear
stresses. Civ. Engng Publ. Wks Rev. 56, 765-770.
BARDEN, L. & KHAYATT, A. J. (1966). Incremental strain rate ratios and the strength of sand in the triaxial
test. Gdotechnique 16, No. 4, 338-357.
BRANSBY, P. L. (1968). Stress and strain in sand caused by rotation of a model wall. Ph.D. thesis, Uni-
versity of Cambridge.
VELOCITY FIELD FOR SOME PASSIVE EARTH PRESSURE PROBLEMS 83
COLE, E. R. L. (1967). The behaviour of soils in the simple shear apparatus. Ph.D. thesis, University of
Cambridge.
DAVIS, E. H. (1968). Theories of plasticity and the failure of soil masses. In Soil mechanics : selected toflics,
I. K. Lee, ed. London: Butterworth.
DRP~OVSEK,J. (1965). Geschwindigkeitsfelder hinter Stiitzmauern, die durch aktiven Erddruck bean-
sprucht sind. Bautechnik 3, 96-100.
DRUCKER, D. C. & PRAGER, W. (1952). Soil mechanics and plastic analysis or limit design. Q. appl.
Maths 10, 157-165.
HANSEN, BENT (1958). Line ruptures regarded as narrow rupture zones; basic equations based on kine-
matic considerations. Proc. Conf. Earth Pressure Probl., Brussels 1, 39-49.
HANSEN, J. BRIP~CH(1953). Earth pvesswe calculation. Copenhagen: Danish Technical Press.
TAMES, I. P. (1967). Stress-displacement relationship for sand subjected to passive pressure. Ph.D. thesis,
Uni&rsiiy of illanchester. _
JAMES, R. G. (1965). Stress and strain fields in sand. Ph.D. thesis, University of Cambridge.
1.4~~s. R. G. & BRAHSBY. P. L. (1970).
_1
Experimental and theoretical investigation of a passive earth
pressure problem. Giotechnique 20: ;?;o. 1; 17-37.
LORD, J. A. (1969). Stresses and strains in an earth pressure problem. Ph.D. thesis, University of Cam-
bridge.
LUCIA, J. B. H. (1966). Passive earth pressure and failure in sand. Research project report, Mechanical
Sciences Tripos, University of Cambridge.
MAY, J. (1967). The cutting of soils. Research project report, Post-graduate course in soil mechanics,
University of Cambridge.
POOROOSHASB,H. B., HOLUBEC, I. & SHERBOURNE,A. N. (1966). The yielding and flow of sand in triaxial
compression, part 1, Can. Geotechnical J. 3, No. 4, 179-190.
POOROOSHASB,H. B., HOLUBEC, I. & SHERBOURNE,A. N. (1967). The yielding and flow of sand in triaxial
compression, parts 2 and 3. Calz. GeotechnicaZ J. 4, Ko. 4, 376-397.
ROSCOE, K. H. (1970). Tenth Rankine Lecture: The influence of strains in soil mechanics. Gdotechnique
20, No. 2, 129-170.
ROSCOE, K. H., ARTHUR, J, R. F. & JAMES, R. G. (1963). The determination of strains in soils by an X-ray
method. Civ. Engng Publ. Wks Rev. 58, 873-876 and 1009-1012.
ROSCOE, K. H., SCHOFIELD,A. N. & WROTH, C. P. (1958). On the yielding of soils. G&technique 8, No. 1,
22-52.
SCHOFIELD,A. N. (1959). The development of lateral force during the displacement of sand by the vertical
face of a rotating model foundation. Ph.D. thesis, University of Cambridge.
SHIELD, R. T. (1953). Mixed boundary value problems in soil mechanics. Q. appl. Maths. 11, 61-75.
SOKOLOVSKI,V. V. (1960). Statics of soil media. London: Butterworth.
SOKOLOVSKI,V. V. (1965). Staticsofgranuhr media. Oxford: Pergamon.

S-ar putea să vă placă și