Geotextiles and Geomembranes 6 (1987) 53--80

Stabilization of Very Soft Soils Using High Strength

Geosynthetics: the Role of Finite Element Analyses

R. K. Rowe
Geotechnical Research Centre, Faculty of Engineering Science, The University of Western
Ontario, London, Ontario, Canada N6A 5B9
and
K. L. Soderman
Golder Associates, 500 Nottinghill Road, London, Ontario, Canada N6K 3PI

ABSTRACT

The application of finite element (FE) techniques for the analysis of re-
inforced embankment behaviour is reviewed. Details such as the choice of
finite element and constitutive models as well as the validation of finite
element results against benchmark solutions are discussed.
Results from the authors' finite element analyses are examined, and it is
shown that FE analysis can be particularly useful for identifying the
mechanisms of failure and also for indicating why a geosynthetic reinforcing
material may substantially improve stability for a certain foundation
strength profile whereas for different foundation strength profiles the same
reinforcement may give rise to negligible improvement in embankment
stability.
The use of plasticity solutions for estimating the maximum effect of
reinforcement is illustrated. The results of finite element analyses are then
used to demonstrate that although the collapse load calculated from
plasticity theory can be attained for very highly reinforced embankments, in
many situations failure will occur at embankment heights well below the
collapse height. It is then demonstrated that the failure height for a reinforced
embankment is related to the modulus of the reinforcement but is not very
sensitive to the modulus of the soil.
The development of strain in the geotextile is examined and it is demon-
strated both from field evidence and theoretical analysis that the reinforce-
ment plays a relatively small role at low load levels since the soil is essentially
elastic. Significant strain in the geotextile begins to develop with increasing
plasticity and in fact most of the strain is developed after a contiguous plastic
53
Geotextiles and Geomembranes 0266-1144187l$03-50 © 1987, Elsevier Applied Science
Publishers Ltd, England. Printed in Great Britain
54 R. K. Rowe, K. L. Soderman

region is developed in the soil, since beyond this point the reinforcement is all
that prevents collapse from occurring. As a consequence, the strains
developed in reinforcement for a given embankment height will largely
depend on the height of the embankment ielative to the height at which
contiguous plasticity occurs, and hence will be sensitive to the magnitude and
distribution of the actual shear strength in a soil deposit.

1 INTRODUCTION

It has now been well established that the provision of geosynthetic re-
inforcement within an embankment or dike constructed on very soft soils
can substantially improve stability and allow construction to heights con-
siderably in excess of that which would be practicable without reinforcement
(e.g. see Refs 1-4). Numerous simple design methods based on considera-
tion of limit equilibrium have been proposed (e.g. see Refs 5-7). An
indication of the maximum improvement in performance which can be
achieved by reinforcement can also be obtained from classical plasticity
bearing capacity solutions.
Neither the limit equilibrium analyses nor the plasticity solutions provide
information concerning the embankment deformations and reinforcement
strains which may be associated with a given reinforcing scheme. In reality,
of course, reinforced embankments are a composite system involving at
least three components: namely, the foundation, the embankment fill and
the reinforcement. The performance of the reinforced embankment will
d e p e n d on the interaction between these components and to a large extent,
this interaction will arise from strain compatibility requirements at the
interface between the various components. This then raises the question as
to how important is consideration of interaction and to what extent will the
performance of the reinforced system depend on deformations and strains in
the reinforcement. A second, but related, question concerns the magnitude
of the strains and deformations that are to be expected under working
conditions.
These questions could all be answered by the construction and monitoring
of a large number of full scale field test embankments. Unfortunately, the
cost of performing and adequately monitoring a sufficiently large number of
full scale embankments is so large that this is not practical. Finite element
techniques provide a cost effective alternative to construction of a large
n u m b e r of test embankments by allowing us to perform numerical simula-
tions of e m b a n k m e n t construction for a wide range of different situations
with the objective of providing a 'data base' which can be used to validate
 Stabilizing soft soils using high strength geosynthetics 55

approximate methods of analysis and to indicate the conditions under which

these analyses are applicable.
Finite element techniques have the potential to allow us to:
improve our understanding of observed behaviour in field trials;
model the complete response of a reinforced embankment up to collapse;
examine the effects of changes in the elements of the system (i.e. the
properties of the reinforcement and the soil); and
investigate changes in construction procedures and the nature of the
system.
The finite element technique is well recognized as being a very powerful
tool and numerous examples of its application to modelling of reinforced
embankments can be found in the literature (e.g. see Refs 2, 4, 8-18).
However, the use of finite element analyses for performing studies which
could answer the questions raised above is subject to some important
constraints.
If one is to model the response of the embankment up to collapse, then it
is essential to adopt a finite element formulation and constitutive model
which (a) models the stress-dependent properties of the fill material; (b)
correctly models plastic failure and plastic flow in the fill and the foundation;
and (c) allows for potential slip at the soil-reinforcement interface.
It should also be recognized that there are many different finite element
models with different characteristics and capabilities. The only way to
determine if a particular formulation and program is likely to give reason-
able results is to 'validate' the program against (a) limiting analytical
benchmark solutions; (b) available data relating to the construction and
performance of full scale embankments; and/or (c) centrifuge test data.
Finally, it is noted that although a finite element study may be a cost-
effective alternative to a full scale field study, it is not without cost. A good
elasto-plastic FE analysis requires an experienced 'driver' and considerable
data preparation time.
The primary roles of finite element analysis are (a) for research into the
behaviour of reinforced embankments including validation of simplified
methods of analysis; and (b) for complementing conventional analyses on
large/important projects or where the anticipated conditions are such that
the validity of simplified approaches may be questioned.
The objective of this present paper is to review some of the important
factors to be considered in the use of the finite element technique to analyze
reinforced embankments and to discuss some of the authors' findings which
have resulted from the use of the finite element technique in the analysis of
reinforced embankments.
56 R. K. Rowe, K. L. Soderman

2 SELECTION OF FINITE E L E M E N T AND

CONSTITUTIVE M O D E L

Modelling of the discrete components of a reinforced soil system (i.e. the

soil, the reinforcement and the soil-reinforcement interface) involves con-
sideration of both the type of finite element and the constitutive relationship
that will be adopted. This is particularly so if the analysis is intended to be
carried to collapse.

2.1 The soil

There are many different types of finite elements which could potentially be
used to represent the soil. Restricting attention to plane strain conditions, it
has been shown 19'2°that constant strain, linear strain, quadratic strain and
cubic strain triangles can all give accurate predictions of undrained collapse
for soft foundations, although if constant strain triangles are used, they must
be used in a crossed triangle formation (i.e. rectangles are divided into four
isosceles triangles--see Ref. 19).
A n u m b e r of quadrilateral elements could also be used but care is needed
in the choice of element and integration rule. The most commonly used
quadrilateral elements have either four nodes or eight nodes (see Ref, 21).
The four noded element is not suitable for predicting collapse loads. 2° The
eight noded element is most commonly used in conjunction with a reduced
(2 x 2) integration rule suggested by Zienkiewicz. 2~This approach has the
beneficial effect of avoiding the phenomenon of 'locking' and potentially
improving the calculation of collapse loads while decreasing the cost of the
computation (as compared with an analysis using full 3 x 3 integration).
However, as indicated by Nagtegaal and De J o n g 22 and Sloan, 23 the use of
reduced (2 x 2) integration with eight noded quadrilateral elements can
give rise to unrealistic deformation patterns and care is required to ensure
satisfactory results are obtained.
The choice of which 'suitable' element to use will depend on consideration
of efficiency and convenience. If the concern is simply to determine the
bearing capacity factors for rigid footings on clay, then it has been demon-
strated 2°'24that the high order (e.g. 15 noded cubic strain) triangles are the
most efficient, in that they allow the determination of accurate collapse
loads with a minimum number of nodes and elements. However, in the
analysis of reinforced embankments, the size of elements (particularly in the
fill) will be dictated by constraints related to simulation of the construction
process and geometry, and in these cases the potential gain in efficiency may
be lost because of the large number of degrees of freedom in the fill that are
required to model construction in small fill lifts. Thus the choice of element
 Stabilizing soft soils using high strength geosynthetics 57

ceases to be a clear-cut decision, and although the present authors' work has
indicated that there are computational advantages to using cubic strain
triangles for the analysis of bearing capacity problems where shear strength
increases with depth, for analyses of reinforced embankments on these same
foundations it has been found to be convenient and efficient to use low order
(constant strain) triangles for modelling the soil in the foundation and fill.
Constitutive models for the soil may be subdivided into two categories
namely, non-linear elastic and (non-linear) elasto-plastic. Non-linear elastic
(e.g. hyperbolic) models can be expected to provide acceptable results at
low stress levels (e.g. when there is a large 'factor of safety'); however, since
they are based on elastic theory, they cannot correctly model plastic failure
and plastic strains within the soil mass (it is noted that the use of a cohesion
intercept c and friction angle 4~in a hyperbolic model does not imply that the
model is a plasticity model---e.g, see Ref. 25). Consequently, these models
are not suitable for calculating collapse heights.
Numerous plasticity formulations have been proposed in the literature.
The simplest of these involves a Mohr--Coulomb failure surface and a
non-associated flow rule. 26This model has been successfully applied in the
analysis of geotextile reinforced embankments. 2.1°,12,27This form of analysis
can be readily modified to include the consideration of a non-linear failure
envelope commonly encountered with granular materials. 28 These models
can be expected to model the soil behaviour up to and including failure. By
examining the results of studies performed using this class of model it is
possible to assess the magnitude of the strains to be expected prior to
collapse of the structure and hence to make some initial assessment of
potential significance of strain softening. However, this class of model is not
suitable for modelling strain-softening behaviour, and indeed the modelling
of localization and strain softening even for unreinforced materials requires
considerable additional research.
In modelling the behaviour of the embankment, it is essential to consider
the variation in soil stiffness with increasing stress level during construction,
since this can have a significant influence on the stresses and displacements
developed within the reinforced embankment. The simplest way of
modelling this is to adopt a non-linearity based on Janbu's equation, viz.

( E / P . ) = K(o'/Pa) m (1)

where E is the Young's modulus of the soil, tr is th6 minor principal stress or
the mean stress depending on the details of the formulation, Pa is atmos-
pheric pressure and K and m are the material parameters. This non-linearity
is included in the 'hyperbolic' model and can also be readily included in
elastic-plastic models. It should be recognized that the modelling of 'yield'
58 R. K. Rowe, K. L. Soderman

implicit in eqn (1) is only approximate and is not appropriate for situations
where there may be cyclic loading. 29 However, there is considerable
evidence to suggest that this approach can provide reasonable results for
problems involving monotonic loading (as is generally the case in modelling
e m b a n k m e n t construction).

2.2 The reinforcement

The reinforcement can be modelled using a one-dimensional bar element.

Non-linearity of the stress strain behaviour and yield can also be readily
modelled by making the element stiffness a function of stress (or strain)
level. Breakage (snap) of the reinforcement can also be modelled; however,
this involves the redistribution of stresses developed in the reinforcement
prior to breaking, and erroneous stress distributions can be obtained unless
particular care is taken with the numerical algorithm used to redistribute
these stresses.

2.3 The soil-reinforcement interface

The interaction between the soil mass and the reinforcement can be
modelled by introducing soil-reinforcement interface elements. This can be
achieved in a number of ways, including the use of joint elements, nodal-
compatibility slip elements or by substructuring. C o m m o n approaches to
modelling the soil-reinforcement interface involve three nodes at each point
along the reinforcement; one attached to the soil above the reinforcement,
one on the reinforcement, and one attached to the soil below the reinforce-
ment. The nodal-compatibility slip element (which may be formulated
initially in terms of normal and tangential springs with very high stiffnesses)
(a) ensures compatible displacement between a pair of dual nodes (one
attached to the soil and one attached to the reinforcement) until a Mohr-
Coulomb failure criterion is reached, and (b) replaces the compatibility
conditions by a failure condition and dilatancy equation once the interface
strength is exceeded. Joint elements allow relative deformation of the soil
and reinforcement, prior to failure of the interface, based on some assumed
constitutive relationship of what is in effect an interface layer between the
reinforcement itself and the general soil continuum (e.g. Andrawes e t al. 3°'31
used a hyperbolic model to represent the interface behaviour).
In its simplest form, the joint element may be comprised of a pair of
normal and tangential springs. Clearly, as the stiffness of a joint element
increases, it tends to a nodal-compatibility slip element, and the distinction
between the two is related to the question of whether a distinct interface
layer exists or whether the deformations at the interface (prior to failure) are
 Stabilizing soft soils using high strength geosynthetics 59

simply due to the interaction between the reinforcement and the soil on
either side of the interface. If there are good experimental data indicating that
a distinct interface layer exists with experimentally defined stress-strain
characteristics, then this can be readily modelled as a joint element or as a
thin layer of continuum element (with slip still being modelled using a
nodal-compatible slip element). In the absence of this data, a nodal-
compatibility slip element would seem appropriate.
A n y modelling of interface behaviour must consider three possible
mechanisms of failure as noted below.
(a) If there is insufficient anchorage capacity, failure will occur at the soil
reinforcement interface above and below the reinforcement as the
reinforcement is pulled out of the soil. This 'pullout' mode involves
displacement of the reinforcement relative to the soil on both sides of
the reinforcement. This is not often a problem for typical sheets of
reinforcing material. However, it may occur if the reinforcement is in
the form of strips or grids, or if insufficient room is available to anchor
very high strength fabrics.
(b) If the shear strength of the soil reinforcement is less than the shear
strength of the soil alone, then failure may occur by sliding of the soil
along the upper surface of the reinforcement, as the upper soil mass
moves relative to both the reinforcement and the underlying soil.
This rarely occurs.
(c) The soil below the reinforcement (usually the soft foundation) may
be squeezed out from beneath the lowest reinforcement layer (and
the entire reinforced embankment). In this case, the soft foundation
soil may move relative to the reinforcement and the overlying soil.
This commonly occurs in reinforced embankment analysis.
If the reinforcement is in the form of a sheet, completely separating the
soil above and below the reinforcement, then the interface resistance can be
readily determined by direct shear tests. 32 In this case, provision for slip at
the interface is the same irrespective of the mechanism of failure (that is,
direct shear or pullout). However, if the reinforcement takes the form of a
geogrid, with openings which are large compared to the grain size of the soil,
or if the reinforcement consists of separate reinforcing strips, then special
care is required to correctly model the failure mechanism. For these
materials, the interface shear resistance in direct shear (e.g. if there is sliding
of the soil along the upper surface of the reinforcement) may be sub-
stantially higher than the interface resistance in pullout. 32In modelling these
materials, it is necessary for the formulation of the interface element to be
such that it can detect whether it is in a direct shear or pullout mode and to
then select the appropriate interface parameters to model this mode of
60 R. K. Rowe, K. L. Soderman

shearing. Thus the behaviour of the interface element on one side of the
reinforcement is related to the behaviour of the interface element on the
other side (since the mode of shearing can only be assessed by consideration
of the direction of shear on either side of the reinforcement).
For planar reinforcement, independent movement of the soil may occur
above and below the reinforcement following either a direct shear or pullout
failure. For strip reinforcement, independent movement of the soil above
and below the plane of reinforcement can only occur during a direct shear
m o d e of failure. Pullout of strips is really a three-dimensional phenomenon,
in which the strips move relative to the soil around them but the soil between
strips remains continuous. As noted by Naylor and Richards, 33the common
approach of using a conventional joint element (or nodal compatibility
element) implicitly treats the strips as an equivalent two-dimensional sheet
and will cause serious error since it interrupts the transfer of shear stress
through the soil.
Since pullout of strips does represent a truly three-dimensional situation,
it can only be approximately modelled in a two-dimensional analysis. A
n u m b e r of different approaches can be adopted. For example, Naylor and
Richards 33 proposed a composite formulation which ensured continuity of
shear stress in the soil after pullout by introducing a 'conceptual shear zone'.
A n alternative approach implemented by the present authors in their
formulation involves an interface element which has a node above the
reinforcement, a node on the reinforcement and a node below the reinforce-
ment. Prior to slip, normal and tangential compatibility between the soil and
reinforcement is enforced by means of very stiff springs. The normal and
shear stresses 'above' and 'below' the reinforcement are automatically
monitored. If a pullout mode o f failure occurs (as inferred by the direction of
shear above and below the reinforcement together with a Mohr-Coulomb
failure criterion), then the computer program automatically enforces
compatibility between the soil nodes 'above' and 'below' the reinforcement
(thereby maintaining continuous transfer of shear stress in the soil) while
allowing slip between the reinforcement node and the two soil nodes. The
normal force between these nodes is used to assess the normal forces acting
on the strip; the corresponding shear resistance (based on a Mohr--Coulomb
failure criterion) between the strip and soil is applied to both the upper and
lower soil node, and as an equilibrating force to the node on the soil strip.
Since the strip covers only a small area of the soil, the Mohr-Coulomb
parameters must be adjusted to take account of the actual surface area, per
unit width of the embankment, which is in contact with the soil.
For the remainder of this paper, attention will be restricted to sheets of
geosynthetic reinforcement. Rowe and Mylleville 34 discuss strip
reinforcement.
 Stabilizing soft soils using high strength geosynthetics 61

3 NUMERICAL PROCEDURE

The previous section discussed factors associated with the selection of the
finite elements and constitutive model to be used. The validity of finite
element results will, of course, be dependent on the use of a suitable finite
element model. However, the choices of load step size and finite element
mesh together with the numerical procedure adopted for ensuring that the
failure criteria, the flow rule and total equilibrium are all satisfied at the end
of each load increment are equally important.
The construction sequence adopted in the finite element simulation
should follow, as nearly as practicable, that which would be anticipated in
practice. Typically (but not necessarily---e.g, in the case of dredging) this
will involve construction of the embankment in a number of horizontal
layers or lifts. Each of these layers will be simulated by a row of finite
elements where construction of the layer involves 'turning on gravity' (i.e.
increasing the unit weight of the layer from zero up to the 'design value') in a
number of increments. For non-linear problems involving both plasticity
and stress-dependent fill stiffness, the choice of load step size should be such
that any further reduction in load step size will not significantly affect the
results of the analysis. This can be established by repeating an analysis for a
number of different load step sizes. For the analyses to be discussed in the
following section, the construction simulation involved up to 16 lifts (layers)
of fill and a total of up to 250 load steps (increments) being placed.
The finite element mesh should be selected such that it is sufficiently
refined (i.e. has enough degrees of freedom) in the critical zone where
collapse occurs to ensure that the collapse height and mechanism are deter-
mined to sufficient accuracy. A preliminary indication of how this critical
zone may change due to reinforcement can be obtained by looking at critical
circles from limit equilibrium analyses for the unreinforced case, and by
examining the plasticity solutions for a rigid footing (to be discussed in a
following section).
The validity of the finite element mesh and procedure can be assumed by
comparing the finite element collapse height for an unreinforced embank-
ment with that from conventional limit equilibrium analyses and by
comparing the collapse height for a very heavily reinforced embankment
with that estimated from plasticity solutions for rigid footings.

4 EFFECT OF R E I N F O R C E M E N T ON FAILURE MECHANISM

As intimated in the preceding section, the inclusion of high modulus re-

inforcement in an embankment can significantly change the failure
62 R. K. Rowe, K. L. Soderman

mechanism. For example, Figs 1 and 2 show the velocity field at collapse
obtained by Rowe and Soderman ~5 from a finite element analysis of an
unreinforced (J = 0) and a heavily reinforced (reinforcement modulus
J = 4000 kN/m) embankment* having a crest width of 30 m (2:1 side slopes)
and resting on a soil with an undrained shear strength which increases
linearly from a surface value c* of 7-69 kPa at a rate p* = 1.54 kPa/m. In
these 'velocity fields' the arrows indicate the direction and relative
magnitude of the soil movements at the onset of collapse.
In Fig. 1, the velocity field indicates that at a collapse height of 3 m, the
failure mechanism for the unreinforced embankment begins a few metres
from the shoulder and extends to a depth of about 2 m. For comparison
purposes, the critical circle from a simplified Bishop limit equilibrium
analysis is also shown and it can be seen that the two methods of analysis give
a very similar failure mechanism and collapse height.
The collapse height obtained for the heavily reinforced e m b a n k m e n t
(J = 4 0 0 0 k N / m ) is approximately twice that for the unreinforced
e m b a n k m e n t . The reason for this substantial increase in collapse height is
evident from a comparison of Figs 1 and 2, which shows that the reinforce-
m e n t forces the collapse mechanism down into the stronger soil at depth. In
fact, increasing the modulus from 0 to 4000 kN/m moves the edge of the
mechanism from near the shoulder to near the centreline of the embank-
m e n t and forces it from a depth of about 2 m to a depth of between 8.5 and
9m.
The preceding example is one in which the finite element results
d e m o n s t r a t e that high modulus reinforcement can substantially improve the
stability of an e m b a n k m e n t and also, by inspection of the change in failure
mechanism, provides an intuitive feel for why this improved performance
was realized. However, the finite element solution can be equally useful for
indicating situations where even high modulus reinforcement may give rise
to very little improvement in stability. To illustrate this, Figs 3 and 4 show
the velocity fields at collapse obtained by Rowe and Soderman ~3 for an
unreinforced e m b a n k m e n t constructed on a deposit of clay having a
uniform strength with depth. The primary difference between the two sets of
results is the thickness of the deposit, which corresponds to a depth to crest
width ratio ( D / B ) of 0.33 and 0.55 for Figs 3 and 4 respectively.
A n examination of the displacement components at the interface between
the fill and foundation in Figs 3 and 4 reveals that for D / B = 0-33, the
c o m p o n e n t of horizontal displacement is substantially greater than for

*For the analyses discussed in this paper, the fill properties were assumed to be given by
friction angle ~ = 32°, dilatancy ~b= 0%unit weight,/ = 20 kN/m 3, Poisson's ratio u = 0.35
and dimensionless stiffnessparameters K = 100 and m ~ 0.5.
 Stabilizing soft soils using high strength geosynthetics 63

0 I 2 3m
I i i |

i!i!i!i!i!i!i
Fig. 1. Velocity field at collapse for the case C*o = 7.69 kPa, pc* = 1.54 kPa/m and J = 0
(unreinforced) (after Ref. 15).

0246m
I I I |

i!!tilt111l'm.
:.• :.:.:.:.:: . :::::~, i ~ . - % ~ , . 7 _ 4 ~ . , ' Y , , ~ . ~ : : " : ::. :.:. :.:.:.:.: .:.: ::. :. : :. :.:
. : . . . ,,,, : ~" , ''',. ~, ', ,, ~a. -~~. , ~, f. .,.,, '- /~,-<- " ,~.' ".'.' '., ~ . . . .
• . "......., "... "........... • .., • . . .

• ",' " , " ' " '." ".°,' "," °,'.'.',' ".'.' ",'," ' , ' , , ' , " ' ' , ' • , " ,

" ",' " " ' ",'," "o'0° " " ' ' ",',' " "," ' ' • ' • ' • " ' •, ', ' , • ,
• ,,,,,,, • • • ,.,,, • • • ,., ,,,.,,-,,., . • • • ..,,, • • .., , , ' , .
. . . . . . . . . . , , , . . . . . . . . , . . . . . . . . . . . .
", . . . . . . . . , . . . . . , , . , . . . . . , . . . . . . . . . .

Fig. 2. Velocity field at collapse for the case of C*uo= 7.69 kPa, p~* = 1.54 kPa/m and
J = 4000 kN/m (after Ref. 15).

Io 6.Srr, =I
/--PLANE OF GEOTEXTILE

T
I .~.':~: . ~ .,,-,.,,-~-,~-~_.,~-~-:~,-.-...-..........
~" :.~'Z~.-."~7-_'.-.'. : " : : : 4.5m
.t.

Fig. 3. Velocity field after collapse, D/B = 0.33 (after Ref. 13).
64 R. K. Rowe, K. L. Soderman

D/B = 0.55. If a strong reinforcing geotextile was to be inserted at the

interface between the fill and foundation, then for D/B = 0.33 one would
expect that the reinforcement would resist these lateral movements thereby
mobilizing forces in the reinforcement, changing the failure mechanism to
give a more rigid downward movement of the embankment and,
consequently, increasing the collapse height. However, for D/B --- 0.55,
there is very little lateral movement at the interface and the inclusion of a

P.A E OE EXT,.E
.-p
-[
ZSm

,,

1
Fig. 4. Velocity field after collapse, D/B = 0.55 (after Ref. 13).

reinforcing geotextile would not be expected to significantly change the

failure mechanism or the collapse height for this case.
The results presented in this section show that the improvement in
stability which can be achieved by using a given high strength geosynthetic
reinforcement will depend on the undrained strength profile and relative
depth of the deposit, a n d as a consequence, it may be very unwise to
extrapolate potential benefit of reinforcement obtained in field trials on one
site to another site.

5 ESTIMATING THE MAXIMUM IMPROVEMENT

IN C O L L A P S E H E I G H T T H A T CAN BE O B T A I N E D
USING R E I N F O R C E M E N T

A number of investigators 35'36have developed bearing capacity factors for

rigid footings. These solutions have considered the effect of increasing
undrained strength with depth as well as the effect of the relative thickness
of the soil deposit. The bearing capacity factors determined by them have
been synthesized and plotted in Fig. 5 in terms of the dimensionless quantity
pcb/cuo where pc is the rate of change in cu, b is the effective width of the
footing, and cu the undrained strength directly beneath the footing (see Fig.
5). Since a reinforced embankment can never be reinforced beyond the
 Stabilizing soft soils using high strength geosynthetics 65

35

50

25 -r . . . . . . . -)f. . . . . . . . . . . . . . . . . . . . . . . . . . /
=- Rough
20
Nc
15 -ff o.
onO ,o-<,o
I0
No= 11.3 + 0 . 3 8 4 Pcb/Cuo
5
5.14-- 5 -
" ~x/-ff

0 I I I I I I I I I
0 5 I0 15 20 25 30 35 40 45 50

CUO

Fig. 5. Bearing capacity factor for non-homogeneous soil (synthesized from results by Davis
and Booker, 35and Matar and Salencon.36

point of being rigid, these solutions place a limit on the improvement in

stability which can be achieved using high s t r e n g t h reinforcement. In
conjunction with a finite element analysis, these solutions may serve two
purposes.
Firstly, prior to the analysis, they can be used to estimate the maximum
height of e m b a n k m e n t which can be achieved and this information will
influence the design of the finite element mesh. Secondly, after an analysis,
these solutions can be used as a check on the collapse height obtained from
the finite element analysis for a highly reinforced e m b a n k m e n t .
Since the plasticity solutions are for a rigid footing of width b and since the
e m b a n k m e n t will generally have a trapezoidal shape, an approximation
must be m a d e to determine the equivalent width of the e m b a n k m e n t . From
plasticity considerations, the pressure at the edge of a rigid footing is
(2 + 7r)Cuo. It is assumed here that the effective width of the footing b will
extend between the points on either side of the e m b a n k m e n t where the
applied pressure 7h is equal to (2 + zr)Cuo. Thus, the thickness h, where the
applied pressure is (2 + 7r)Cuo, is
h = (2 + lr)Cuo/Y (2)
and hence (from Fig. 6)
b = B+2n(H-h) (3)
66 R. K. Rowe, K. L. Soderman

lq b "I

I "r"--.,+ Ouo u

/'/'/ff////////////f/////////////////////l/////////////////////////
Fig. 6. Definition of variables used to estimate collapse height for a perfectly reinforced
embankment.

I- b ~ l
I
, I . CUO _ C u

11/iiii../iiiii/i/ii111111111z1111/11

d
b

0.5

l I I II *''I , i , *i,,.I i , t J,,.l I I I iJaiJ

0
0.I I I0 I00 I000
Pc~
ClIO

Fig. 7. Effect of non-homogeneity on depth of the failure zone beneath a rough rigid footing
(modified from Ref. 36).

where B is the crest width, H is the embankment height and n is the

cotangent of the slope angle.
The bearing capacity q, of the rigid footing of width b is given by

q~ = Nccuo+q~ (4)
where q, is a uniform surcharge pressure applied to the soil surface outside of
the footing width and Nc is the bearing capacity factor obtained from Fig. 5.
Inspection of Fig. 6 shows that the triangular edge of the embankment is
providing a surcharge that would increase stability. What is required is an
estimate of q, in terms of the pressure applied by this triangular distribution.
Figure 7 shows the depth d to which the failure mechanism is expected to
extend based on Matar and Salencon. 36 From an inspection of typical
characteristic fields, it is found that the lateral extent of the plastic region
 Stabilizing soft soils using high strength geosynthetics 67

involved in the collapse of a rigid footing extends a distance x from the

footing where x is approximately equal to the minimum of d as determined
from Fig. 7, and the actual thickness of the deposit D, i.e.
x = min(d,D) (5)

Thus distributing the applied pressure due to the triangular distribution over
a distance x gives

q~ = nyh2/2x for x > nh (6a)

and

q~ = ( 2 n h - x ) y h / 2 n h for x <- nh (6b)

This value may then be compared with the average applied pressure qa due
to the embankment over the width b, viz.

q~ = y [ B H + n(H z - h2)]]b

At collapse, the ratio qu/qa should be equal to unity. Thus, to determine

the collapse height, H¢, for a given crest width, side slope and strength
profile, it is necessary to adopt a procedure in which H is assumed, qu and qa
are calculated, and the ratio of q J q , determined. If qo/q~ is greater than
unity, then H should be increased (if less than unity, then H is decreased)
until the critical height H = Hc is achieved wherein q~/q~ = 1. This is the
procedure to be adopted initially to determine the maximum height of
embankment that may need to be modelled.
For the purposes of estimating the maximum possible factor of safety
(defined here as FS = q J q , ) for a given embankment geometry and soil
profile, or for checking the reasonableness of a finite element collapse load,
qu and qa can be determined directly from eqns (2)-(7), and hence the ratio
qJqa can be determined. In the case where the finite element analysis
indicates collapse, the ratio of qJq~ calculated for the embankment
geometry at collapse should be approximately equal to unity for a perfectly
reinforced embankment (i.e. provided that failure is not controlled by
failure or yield of the reinforcement).
To illustrate the application of this checking technique, consider the
highly reinforced embankment whose velocity field at collapse is shown in
Fig. 2. Figure 8 shows the vertical displacement at a point a beneath the
shoulder of the embankment, as the height of the embankment is increased
(i.e., as y H increases, where H is the total thickness of the fill above point a
at any time). Results are given corresponding to fabric modulus values of
1000, 2000 and 4000 kN/m.
68 R. K. Rowe, K. L. Soderman

180 I I I I

160
140
Applied 120
Pressure ,m
"YH I00
(kPo) 8O
6O I I* I N_ft" ,.54kPo/,
,Sm zI ~1~
40 • Contiguous Plosticity -~-~r~///z////~////'//
• Maxknum Net Height
20 .... Collopse

0 I I 1 I
0 I 2 3 4
Vertical Deflection of Point a (m)

Fig. 8. Applied pressure versus deflection for various reinforcement moduli (modified from
Ref. 15).

In each case, the deformations of the embankment became indeterminate

and collapse occurred at an applied pressure of approximately 120 kPa. The
collapse pressure was independent of the geotextile modulus and was in fact
controlled by shear failure at the interface between the geotextile and the
underlying clay foundation. Thus, for the case where B = 30 m, Hc = 6 m,
3/ = 20 k N / m 3, C,o = 7.69 kPa, pc = 1.54 kPa/m and n = 2, from eqns (2)
and (3) it is found that h = 1.98 m and b = 46 m. This gives pc b/c,o equal to
9.2 and hence from Fig. 5 Nc = 12.3. For pcb/c,o = 9.2, Fig. 7 gives x = 0.2
b = 9.2 m, and hence from eqn (6a), qs = 8.5 kPa, and so from eqn (4) the
ultimate bearing capacity is 103 kPa. This may be compared with the
average applied pressure in the finite element analysis, which from eqn (7) is
106 kPa. Thus the calculated applied pressure exceeds the ultimate bearing
capacity from plasticity theory by about 3%. This level of 'error' associated
with the numerical solution is considered to be acceptable for typical finite
e l e m e n t analyses.

6 U S E O F F I N I T E E L E M E N T ANALYSIS T O D E T E R M I N E
FAILURE AND COLLAPSE HEIGHTS

As d e m o n s t r a t e d in the previous section, classical plasticity solutions can be

used to determine the maximum improvement in stability which can be
achieved by the use of reinforcement. Furthermore, the results shown in
Fig. 8 suggest that for a perfectly reinforced embankment, the collapse load
 Stabilizing soft soils using high strength geosynthetics 69

is independent of the modulus of the geosynthetic. However, this collapse

load does not take account of possible failure of the reinforcement or
pullout of the reinforcement, nor does it consider the magnitude of the
deformations that may develop prior to 'collapse'. This then raises the
question as to what is the collapse height for a given situation and whether
there is a height less than the collapse height at which the embankment may
be deemed to have failed, even though, in the strict sense of plasticity
theory, it has not collapsed. To answer this question, it is useful to define the
term 'contiguous plasticity' and to look at the sequence of events as a
reinforced embankment is constructed up to collapse as determined from a
finite element analysis.
The term 'contiguous plasticity' relates to the situation where there is
general plastic failure within the soil in the region of a potential collapse
mechanism (i.e. the shear strength of the soil is mobilized along the
potential collapse mechanism). For an unreinforced embankment, collapse
and failure coincide with the development of contiguous plasticity and a
corresponding collapse mechanism (see Fig. 1). However, for a reinforced
embankment, the development of contiguous plasticity is only the first step
towards failure and collapse. The height at which contiguous plasticity is
developed is dependent upon the geotextile modulus as indicated in Fig. 8,
but the effect is not large. The effect of the geotextile modulus is really only
appreciable after the development of contiguous plasticity. Thus although
the collapse height is the same for the three cases examined in Fig. 8, the
deformations prior to collapse differ substantially.
The importance of considering deformations in any assessment of failure
can be demonstrated by replotting the results given in Fig. 8 in the form of
net fill height above original ground level (i.e. the thickness H minus the
vertical deflection for that fill thickness) against the fill thickness H as shown
in Fig. 9. For the unreinforced embankment, the maximum net fill height is
about 3 m and occurs at the onset of contiguous plasticity. Notice that this
also corresponds to the collapse height and in fact the net fill height at the
onset of collapse is only slightly less than the fill thickness. For the reinforced
embankments, it can be seen that the maximum net fill height does depend
on the geotextile modulus. Any attempt to place additional fill after
attaining this maximum height will result in the loss of net height. This
corresponds to a controlled failure. In order to support the stresses imposed
by the additional fill, large deformations of the geotextile and the soil must
develop. This can continue in a controlled manner until collapse eventually
occurs. However, from a practical standpoint, the fill thickness of the
embankment at failure should be considered to correspond to the fill
thickness at the time the maximum net fill height is obtained and not the fill
thickness required to achieve 'true' (i.e. uncontrolled) collapse.
70 R. K. Rowe, K. L. Soderman

I I I I I I .~
I
I
Collapse
II
J • 4000kN/rn I
Net
Emb.42 - 2ooo~
Height
Hnet 3
(m)

I / • Contiguous Plosficity
/ ~ Maximum Net Fill Heigh, il
l i I I I I
0 I 2 3 4 5 6
Embankment Thickness, H (m)
Fig. 9. Net fill height versus fill thickness for various reinforcement moduli (modified from
Ref. 15).

From the foregoing it is evident that the failure height of an embankment

on a given soil profile is directly related to the modulus of the geotextile. As
might be expected, the maximum force mobilized in the geotextile at failure
increases with increasing geotextile modulus. However, because the
deformation pattern also changes with increasing geotextile modulus, these
forces do not correspond to a unique strain for soil profiles where the
strength increases with depth.
Since the failure height of an embankment, as defined above, is related to
deformations, one would expect that the modulus of the soil would have
some influence on the calculated failure height for a given level of
reinforcement.
The ratio of undrained Young's modulus to undrained shear strength
(Eu/cu) for many soft clays lies in the range 125 -< EJc~ <- 500. 37The results
presented in Figs 1, 2, 8 and 9 were obtained for E,/cu = 125. Figure 10
shows the effect of varying EJc. from 125 to 500. This fourfold increase in
soil modulus reduced the embankment thickness required for contiguous
plasticity by about 6% and reduced the maximum geotextile strain at con-
tiguous plasticity from 2-5% to 1.0%. The embankment thickness required
to cause failure and the maximum geotextile strain at failure ( - 7 . 5 % ) were
to all practical purposes the same, although the maximum net height was
slightly smaller for the lower value of Eu/c~. It may be concluded that the
failure pressures and maximum geotextile strains at failure obtained for
 Stabilizing soft soils using high strength geosynthetics 71

7 I I I I I I
%0 " lOkl:~ Nominol Strength
Pc • 2 kPa ,I
6
¢ ~ . 7.69kPa=
1.54 kPo/m I Foct°~ed Strength
5
Net Euo" ,5000 kPo 1
PE" IOOO kPo/m ~ I
Emb. 4
Height
Eu,c.. 1,,"
I I guo"12~o kPo
Hnet 3
(rn)

• Contiguous Plosticity
/ t Moximum Net Fill Hehjht (F'oilure)
I I I I I I
00 I 2 3 4 5 6 7
Embankment Thickness, H (m)

Fig. 10. Net embankment height versus embankment thickness for two values of E,/cu,
J = 4000 kN/m.

EJc~ = 125 are also reasonable for values of EJco in the range 125 < Eu/c~
< 500. The explanation for this lies in the fact that the strains which control
failure are predominantly plastic strains and, since changing the modulus
really only changes the elastic strains (and these are small), there is no
significant effect.

7 D E V E L O P M E N T OF STRAIN IN T H E R E I N F O R C E M E N T
AND FACTORS OF SAFETY

Rowe and Soderman 2 have compared the calculated and observed per-
formance of both a reinforced and an unreinforced test embankment con-
structed on a 3-3 m thick layer of very soft organic clay (cu = 8 kPa) at
Almere, The Netherlands.~ The construction procedure for the reinforced
e m b a n k m e n t involved (a) placing a high modulus geotextile (J = 2000 kN/
m) directly on the clay foundation, (b) excavation of a ditch beside the
proposed e m b a n k m e n t and placing the soil on top of the geotextile near the
edge to form a retaining bank (see insert to Fig. 11), and (c) placing
hydraulic fill until failure occurred. The strains in the geotextile were
monitored during construction. Figure 11 shows the observed and cal-
culated increase in strain at point A (see insert) after the commencement of
hydraulic filling. As can be seen, very little strain is developed during initial
72 R. K. Rowe, K. L. Soderman

I I I I
~

Embankment [ RetoininQBank
Height
(m)

Observed •
(SCW, 1981 )

Calculated
(Rowe ~ Soderman, 1984)

~A I l I I
0
0 I 2 3 4
Increase in Strain ( % )

Fig. 11. Observed and calculated increase in strain with increasing embankment height--
Almere embankment (strains relative to commencement of hydraulic filling).

stages of loading (until the fill height reaches about 1 m). There is then a
gradual increase in strain with increasing fill height between 1 and 2 m, and a
rapid increase in strain for fill heights in excess of 2 m. Collapse occurred
w h e n the fill height was 2.75 m. The fill height depends upon the position
where it was measured (see insert to Fig. 11). The m i n i m u m and m a x i m u m
heights at collapse were 2.75 m and 3 m respectively. T h e lower height is
used herein as the reference height.
T h e analysis indicates that for fill heights less than I m, the clay is largely
elastic and the lateral deformations are small. As the fill is increased from
1 m to 2 m, there is extensive growth of the plastic region within the clay,
giving rise to an increase in lateral and vertical deformations which is, in
turn, reflected by increased strain in the fabric. At a given height, the
geotextile reduces plasticity within the soil. For example, in the un-
 Stabilizing soft soils using high strength geosynthetics 73

reinforced case, the analysis predicted failure at a fill height of 1.8 m. At the
same height in the reinforced embankment, the displacements are smaller
and the plastic region was not contiguous. A contiguous plastic region in the
soil was calculated to occur at a fill height of 2.05 m. This is approximately
15% higher than the corresponding height for the unreinforced embank-
ment. The development of a contiguous plastic region represents the first
stage of collapse for the embankment; the embankment is now completely
d e p e n d e n t upon the geotextile for the support of additional fill. Further fill
was added until, at a height of approximately 2.7 m, the analysis indicated
that the shear strength of the soil-geotextile interface was reached and
pullout occurred beneath the clay retaining bank. This was followed by
collapse of the embankment. Since the geotextile force at point A was
controlled by the clay-geotextile interface strength and pullout to the right
of A (see insert to Fig. 11), there could not be any increase in strain at this
point.
Three observations can be drawn from this case. Firstly, the use of a high
modulus geotextile substantially increased the collapse height (from 1-75 m
unreinforced to 2.75 m reinforced). Secondly, the geotextile did not play a
significant role, and hence did not experience significant increases in strain,
until a large area of plastic failure had developed in the underlying founda-
tion. Finally, the agreement between the calculated and observed behaviour
was very encouraging.
The observed rapid development of strain once contiguous plasticity is
reached is to be expected. However, since the height at which contiguous
plasticity occurs is only slightly increased by the inclusion of a high modulus
fabric (in the Almere case from 1-8 m unreinforced to 2-05 m reinforced),
this raises the question as to what magnitude of strains may be developed at
failure and what magnitude of strains would be developed under working
conditions.
In the context of this paper, the 'factor of safety' is defined as the divisor
necessary to reduce the soil strength to such a point that failure would just
occur. Thus, for example, if the expected undrained shear strength para-
meters for a foundation soil were Cuo = 10 kPa and pc = 2 kPa/m, then the
allowable height for an embankment which has a 'factor of safety' of 1.3
would be the height at which failure occurs in an analysis conducted for the
factored parameters C*o = 10/1-3 = 7.69 kPa and p* = 2/1.3 = 1.54 kPa/
m. To illustrate the implications of this, Figs 12 and 13 show the develop-
ment of the maximum strain in the geotextile with increasing embankment
height H, as calculated for the factored parameters (Ca*, pc*) and the
expected parameters (Cuo,pc) respectively.
In b o t h cases, the trends indicated by the analysis are similar to that
observed at the Almere test embankment. Thus the geotextile strain in-
74 R. K. Rowe, K. L. Soderman

140 I I I I

~
120

ILURE
I00

Applied 80 -.1~ CONTIGUOUS PLASTICITY

Pressure
qCH
(kPa) 60

40

20

0 1 I 1 1
0 2 4 6 8 I0
Maximum Geotextile Strain ( % )

Fig. 12. A p p l i e d p r e s s u r e v e r s u s m a x i m u m geotextile strain for the case o f ~'*uo = 7.69 kPa,
p* = 1-54 k P a / m (J = 4000 k N / m , B = 30 m, n = 2, D = 15 m, ~/ = 20 kN/m3).

220 I I I I I I

9_00 Results For Actual Volues

Cuo = I0 kPo
180 Pc = 2 kPa/m
160 3- 4000kN/m ~ t

140 Failure
APPLIED 120
PRESSURE -,e-. Contiguous Ploslicity
"YH IO0
(kPo) 80
60
40
2O
0 I I I I I I
0 2 4 6 8 I0 12 14
MAXIMUM GEOTEXTILE STRAIN (%)

Fig. 13. A p p l i e d p r e s s u r e v e r s u s m a x i m u m geotextile strain for the case o f Cuo = 10 kPa,

pc = 2 k P a / m (J = 4000 k N / m , B = 30 m, n = 2, D = 15 m, ~/ = 20 kN/m3).
 Stabilizing soft soils using high strength geosynthetics 75

creases slowly until the height corresponding to contiguous plasticity is

reached. The embankment is then completely dependent on the geotextile
for stability and the strain increases rapidly until failure occurs. The primary
difference between these analyses and the Almere analysis was that here the
fill was all assumed to be granular and hence failure was controlled by the
foundation and geotextile modulus and not by pullout.
Using factored parameters (C,*o= 7-69 kPa, p* = 1.54 kPa/m), failure
occurred at a pressure y H o f 118 kPa (i.e. H -~ 6 m) and the corresponding
geotextile strain at failure was 7.5%. Thus if one adopts a factor of safety of
1.3, then this height of about 6 m would be the allowable height for an
e m b a n k m e n t with design parameters Cuo = 10 kPa, pc = 2 kPa/m. It should
also be noted that if the permissible strain in the fabric was 10%, then this
would also correspond to a factor of safety of approximately 1.3 against
reaching the permissible strain (i.e. FS = 10/7.5 = 1-33).
Figure 13 shows the results of an analysis performed using the expected
parameters (Cuo, pc). It can be seen that under expected working conditions
(i.e. at an allowable pressure of 118 kPa based on the analyses for the
factored parameters), the maximum geotextile strain is only about 3%. This
is considerably less than the maximum geotextile strain of 7.5% which
occurred at the pressure of 118 kPa under failure conditions for the factored
strength parameters. This situation arises because most of the strain is
developed after contiguous plasticity is reached. Using the factored para-
meters (see Fig. 12), contiguous plasticity occurred at an embankment
height of 4 m, while failure did not occur until a height of 6 m (i.e. 50%
higher than the contiguous plasticity height) was reached, with the geo-
textile strains increasing from 1.5% to 7.0% as the embankment height
increased from 4 m to 6 m. With the expected parameters, the height at
which contiguous plasticity would occur is 5.5 m, which is only about 10%
below the design height of 6 m. Under these working conditions, the 'factor
of safety' with respect to an assumed permissible stress of 10% is 3.3 as
opposed to 1.3 based on factored conditions.
This example illustrates that the strains in the geotextile are very sensitive
to the height of the embankment relative to the height at which contiguous
plasticity would occur. Thus, for a given design height, the geotextile strains
will be very sensitive to the magnitude of the actual shear strength of the soil.
The failure pressures obtained for these two cases can be used to contrast
two possible definitions of the factor of safety. With the definition of the
factor of safety adopted above (i.e. based on factored shear strength
parameters), the allowable pressure is 118 kPa for the case of nominal
strengths Cuo = 10 kPa and pc = 2 kPa/m, a geotextile of modulus 4000 kN/
m and an overall factor of safety of 1.3. Altematively, the factor of safety
could be defined as the ratio of the failure pressure obtained using nominal
76 R. K. Rowe, K. L. Soderman

strengths to the allowable pressure. This definition yields a factor of safety of

1.42 at the allowable pressure of 118 kPa suggested previously. This calcula-
tion indicates the latter definition of the safety factor is less conservative
than the definition based on factored strength. Since the major uncertainty
in most e m b a n k m e n t designs is the shear strength of the foundation, the
authors would recommend the use of factored strength to determine
allowable design heights.
It is also noted that the magnitude of strain in the geosynthetic reinforce-
ment at failure will depend on the geometry of the embankment, the shear
strength characteristics of the foundation and the modulus of the re-
inforcement. As a consequence, there is no simple 'limiting geotextile strain'
that can be safely or economically applied to all situations (e.g. see Ref. 15).

CONCLUSION

The finite element technique is a very powerful toot which can be used in the
analysis of reinforced embankment behaviour. However, the applicability
of the results for any given situation will depend on the details of the specific
finite elements and constitutive model which are used, as well as on the care
with which the numerical analysis is conducted. Thus finite element schemes
should be validated against available benchmark solution (as illustrated in
this paper) and, wherever possible, against observed field performance.
Results from a number of finite element analyses performed by the
authors have been discussed. The finite element results were shown to be
particularly useful for identifying the mechanisms of failure and also for
indicating why a geosynthetic reinforcing material may substantially
improve stability for a certain foundation strength profile, whereas for
different foundation strength profiles the same reinforcement may give rise
to negligible improvement in embankment stability.
The use of plasticity solutions developed for a rigid footing, for estimating
the maximum effect of reinforcement was illustrated in the paper. The
results of finite element analyses were used to demonstrate that although the
collapse load calculated from plasticity theory can be attained for very
highly reinforced embankments, in many situations failure will occur at
e m b a n k m e n t heights well below the collapse height. It is then demonstrated
that the failure height for a reinforced embankment is related to the
modulus of the reinforcement. As might be expected, the maximum force
mobilized in the geotextile at failure increases with increasing geotextile
modulus. However, because the deformation pattern in the soil also changes
with increasing modulus (due to different levels of plasticity in the soil),
 Stabilizing soft soils using high strength geosynthetics 77

these forces do not correspond to a unique strain that could be generalized

as a standard limit on strain to be used in limit equilibrium analyses.
It is shown that although failure has been defined in terms of a dis-
placement criterion (specifically, failure is deemed to have occurred when
the addition of fill would not give rise to any additional increase in
e m b a n k m e n t height above original ground level), it is shown that the failure
height is not significantly influenced by the modulus of the soil over a typical
range of modulus values expected in most field situations. It is concluded
that this situation arises because failure is controlled by plastic strains in the
soil (not elastic strains) and the modulus of the reinforcement.
The development of strain in the geotextile was examined, and it was
demonstrated both from field evidence and theoretical analysis that the
reinforcement plays a relatively small role at low load levels since the soil is
essentially elastic. Significant strain in the geotextile begins to develop with
increasing plasticity and in fact most of the strain is developed after a
contiguous plastic region is developed in the soil, since beyond this point the
reinforcement is all that prevents collapse from occurring. As a con-
sequence, the strains developed in reinforcement for a given embankment
height will largely depend on the height of the embankment relative to the
height at which contiguous plasticity occurs, and hence will be sensitive to
the magnitude and distribution of the actual shear strength in a soil deposit.

ACKNOWLEDGEMENTS

The work described in this paper forms part of a general programme of

research into reinforced soil funded by the Natural Science and Engineering
Research Council of Canada under Grant No. A1007. The authors also
gratefully acknowledge the valuable assistance of Dr J. R. Booker and Mr S.
K. Ho.

REFERENCES

1. Study Centre for Road Construction (SCW), Stability of slopes constructed

with polyester reinforcing fabric, Arnhem, The Netherlands (1981).
2. Rowe, R. K. and Soderman, K. L., Comparison of predicted and observed
behaviour of two test embankments, International Journal of Geotextiles and
Geomembranes, 1 (1984), pp. 157-74.
3. Fowler, J. and Koerner, R. M., Stabilization of very soft soils using geo-
synthetics, Proceedings of Geosynthetics '87, New Orleans, 1 (1987), pp.
289--300.
78 R. K. Rowe, K. L. Soderman

4. Humphrey, D. N. and Holtz, R. D., Use of reinforcement for embankment

widening, Proceedings of Geosynthetics '87, New Orleans, 1 (1987), pp. 278-
88.
5. Jewell, R. A., A limit equilibrium design method for reinforced embankments
on soft foundations, Proceedings of the Second International Conference on
Geotextiles, Las Vegas, 2 (1982), pp. 671-6.
6. Fowler, J., Theoretical design considerations for fabric reinforced
embankments, Proceedings of the Second International Conference on
Geotextiles, Las Vegas, 2 (1982), pp. 665-70.
7. Ingold, T. S., An analytical study of geotextile reinforced embankments,
Proceedings of the Second International Conference on Geotextiles, Las Vegas,
2 (1982), pp. 683-8.
8. Bell, J. R., Greenway, D. R. and Vischer, W., Construction and analysis of a
fabric reinforced low embankment on muskeg, Proceedings of the First Inter-
national Conference on the Use of Fabrics and Geotechnics, Paris, France
(1977), pp. 71-6.
9. Ohta, H., Mochinaga, R. and Kurihara, N., Investigation of soft foundations
with surface reinforcement. Proceedings of the Third Australia-New Zealand
Conference on Geomechanics, Wellington, New Zealand, l (1980), pp. 123-8.
10. Rowe, R. K., The analysis of an embankment constructed on a geotextile,
Proceedings of the Second International Conference on Geotextiles, Las Vegas,
2 (1982), pp. 677-82.
II. Boutrup, E. and Holtz, R. D., Analysis of embankments on soft ground
reinforced with geotextiles, Proceedings of the Eighth European Conference on
Soil Mechanics and Foundation Engineering, ECSMFE, Helsinki (23-26 May,
1983).
12. Rowe, R. K., MacLean, M. D. and Soderman, K. L., Analysis of a geotextile
reinforced embankment constructed on peat, Canadian Geotechnical Journal,
21 (1984), pp. 563-76.
13. Rowe, R. K. and Soderman, K. L., An approximate method for estimating the
stability of geotextile reinforced embankments, Canadian Geotechnical
Journal, 22(3) (1985), pp. 392-8.
14. Rowe, R. K. and Soderman, K. L., Reinforced embankments on very poor
foundations, International Journal of Geotextiles and Geomembranes, 4(1)
(1986), pp. 65-81.
15. Rowe, R. K. and Soderman, K. L., Reinforcement of embankments on soils
whose strength increases with depth, Proceedings of Geosynthetics '87, New
Orleans (1987), pp. 266-77.
16. Schaefer, V. R. and Duncan, J. M., Evaluation of the behaviour of the
Mohicanville Dike No. 2. Virginia Polytechnic Institute and State University
Report to the US Army Corps of Engineers (1986), 161 pp.
17. Monnet, J., Gourc, J. P. and Mommesin, M., Study of soil geotextile inter-
action--a reinforced embankment, Proceedings of the International
Symposium on Numerical Models in Geomechanics, Ghent (1986), pp. 539-42.
18. Chou, N. N. S., Tzong, W. H. and Siel, B. D., The effectiveness of tensile
reinforcement in strengthening an embankment over soft foundation,
Proceedings of Geosynthetics '87, New Orleans, 1 (1987), pp. 332--40.
19. Nagtegaal, J. C., Parks, D. M. and Rice, J. R., On numerically accurate finite
element solutions in the fully plastic range, Comp. Meth. Appl. Mech. Eng., 4
(1974), pp. 153-77.
 Stabilizing soft soils using high strength geosynthetics 79

20. Sloan, S. W. and Randolph, M. F., Numerical prediction of collapse loads

using finite element method, International Journal of" Numerical Analytical
Methods in Geomechanics, 6 (1982), pp. 47-76.
21. Zienkiewicz, O. C., The Finite Element Method in Engineering Science, 3rd
Edition, McGraw-Hill, London (1977).
22. Nagtegaal, J. C. and De Jong, J. E., Some computational aspects of elastic-
plastic large strain analysis, International Journal of Numerical Methods in
Engineering, 17 (1981), pp. 15--41.
23. Sloan, S. W., Plastic collapse calculations using high-order elements,
Proceedings of the International Conference Accuracy Estimates and Adaptive
Refinements in Finite Element Computations (ARFEC), Lisbon (19-22 June
1984), pp. 301-13.
24. Sloan, S. W., The numerical analysis of incompressible and plastic solids using
finite elements, Ph.D. Thesis, University of Cambridge ( 1981).
25. Duncan, J. M., Hyperbolic stress-strain relationships, Proceeding of ASCE,
Workshops on Limit Equilibrium, Plasticity and Generalized Stress-Strain in
Geotechnical Engineering (1980), pp. 443--60.
26. Davis, E. H. Soil mechanics--selected topics, In: Theories of plasticity attd
failure of soil masses, I. K. Lee (Ed.), Butterworths (1968), Chapter 6.
27. Rowe, R. K., Reinforced embankments: analysis and design, Journal Geo-
technical Engineering Division, ASCE, 110 (1984), pp. 231-46.
28. Rowe, R. K., Lo, K. Y. and Tham, L., The analysis of tunnels and shafts in
dense (oil) sands, Proceedings of the Fourth International Conference on
Numerical Methods in Geomechanics, Edmonton (1982), pp. 587-96.
29. Zylynski, M., Randolph, M. F., Nova, R., Wroth, C. P., On modelling the
unloading-reloading behaviour of soils, International Journal on Numerical
and Analytical Methods in Geomechanics, 2 (1978), pp. 87-93.
30. Andrawes, K. Z., McGown, A., Mashhour, M. M. and Wilson-Fahmy, R. F.,
Tension resistant inclusion in soils, Journal Geotechnical Engineering, ASCE,
106, No. GT12 (1980), pp. 1313-26.
31. Andrawes, K. Z., McGown, A., Wilson-Fahmy, R. F., Mashhour, M. M., The
finite element method of analysis applied to soil-geotextile systems, Pro-
ceedings of the Second International Conference on Geotextiles, Las Vegas, 2
(1982), pp. 695-700.
32. Rowe, R. K., Ho, S. K. and Fisher, D. G., Determination of soil-geotextile
interface strength properties, Proceedings of the Second Canadian Symposium
on Geotextiles and Geomembranes, Edmonton (September 1985).
33. Naylor, D. J. and Richards, H., Slipping strip analysis of reinforced earth,
International Journal for Numerical and Analytical Methods in Geomechanics,
2 (1978), pp. 343-66.
34. Rowe, R. K. and Mylleville, B., The analysis of steel reinforced embankments
on soft clay foundations, Proceedings of the Sixth International Conference in
Geomechanics, Innsbruck, Austria, April 1988, pp. 1273--8.
35. Davis, E. H. and Booker, J. R., The effect of increasing strength with depth on
the bearing capacity of clays, Geotechnique, 23 (4) (1973), pp. 551--63.
36. Matar, M. and Salencon, J., Capacit6 portante a une semelle filante sur sol
purement coh6rent d'6paisseur limit6e et de coh6sion variable avec la pro-
fondeur, Annales de l'Institute Technique du Batiment et des Travaux Publics,
Suppi6ment No. 352 (Juillet-Aout 1977), Serie: Sols et Fondations, No. 143,
pp. 95-107.
80 R. K. Rowe, K. L. Soderman

37. Duncan, J. M. and Buchigani, A. L., An engineering manual for settlement

studies, Geotechnical Engineering Report, Department of Civil Engineering,
University of California at Berkeley (1976), 94 pp.
38. Janbu, N., Soil compressibility as determined by oedometer and triaxial tests,
European Conference on Soil Mechanics and Foundation Engineering,
Wiesbaden, Germany, 1 (1963), pp. 19-25.