You are on page 1of 12

Geolextiles and Geomembranes 13 (1994) 43 54

Finite Element Modelling of Pull-Out Tests with Load and


Strain Measurements

I. Yogarajah* & K. C. Y e o $
Department of Civil Engineering, University of Strathclyde, Glasgow UK G4 0NG

(Received 1 March 1993 accepted 19 May 1993)

ABSTRACT

This paper examines the load and strain distributions along a geogrid rein-
forcement during a pullout operation. Loads and strains along various
sections o f the geogrid reinJbrcement were measured at different pull-out
displacements, with respect to dfferent anchorage lengths. Numerical
modelling was then carried out to simulate the operation. Comparisons
between the measured and the simulated load and strain distributions are
presented. The validity o f the simulation and its application are discussed.

INTRODUCTION

The stress transfer mechanism between soil and grid reinforcements


involves a frictional resistance at the soil reinforcement interface and
bearing (passive) resistance at the grid apertures. In reinforced soil struc-
tures, such as reinforced soil walls and embankments, the primary
mechanism of stress transfer is frictional, with bearing resistance playing
an important role when grid reinforcements are employed.
To assess the soil-reinforcement interface frictional angle, in-soil
performance tests such as pull-out tests and/or shear box tests are
commonly employed. Although the choice of test used to obtain the rele-
vant interface frictional angle is a matter of debate, pull-out tests have
*Present address: L & M Geotechnic, 63 Hillview Avenue, 05-01 Lam Soon Industrial
Building, Singapore 2366.
:~Present address: Geotechnical Engineering Office, Hong Kong.
43
Geotextiles and Geomembranes 0266-1144/94/$06.00 ~ 1994 Elsevier Science Publishers
Ltd, England. Printed in Ireland.
44 1. Yogarajah, K.C. Yeo

been acknowledged by researchers to provide a better simulation of the


behaviour of reinforcements in reinforced soil structures (Garbulewski,
1990; Venkatappa & Kate, 1990). Frictional forces assessed are based on
an apparent friction angle 6, which is dependent on the soil and properties
reinforcement.
In many experimental pull-out tests, either loads or strains along the
reinforcements are measured. The availability of a single load or strain
measurement prevents the elastic moduli along the various sections of the
reinforcement from being determined. Therefore, many finite element
simulations (Kwok, 1987; Bauer & Mowafy, 1988; Handel et al., 1990)
apply a single elastic modulus, usually determined from index load-strain
curves of the material, over the whole length of reinforcement to predict
the measured behaviour (load or strain), while ignoring its compatriot
(strain or load). This method has proved to be valid for metallic reinfor-
cements where the load-strain relationship is not time-dependent, but
when applied to relatively extensible polymeric reinforcements this
assumption could be doubtful.
This paper describes the experimental procedures for pull-out tests,
where both loads and strains were measured along polymeric grid rein-
forcements. The measured loads and strains were used to determine the
elastic moduli along various sections of the reinforcement. Numerical
modelling of the test was then carried out by applying these experimen-
tally determined elastic moduli. Comparison of results are then made by
employing a single elastic modulus along the reinforcement length and the
numerical analysis repeated. The requirement to measure both loads and
strains to carry out a proper finite element simulation of a pull-out test is
then discussed.

PULL-OUT MECHANISM

Theoretical analysis and experimental observations (Mashhour, 1979;


Jewel et al., 1984) have shown that in pull-out tests, maximum deforma-
tions along the reinforcement occur at the loaded edge, with deformations
decreasing with distance from the loaded edge. The consequence of this is
that the pull-out force developed along the reinforcement causes shear
stresses within the soil, compelling the magnitudes and directions of the
principal stresses in the soil to be continuously altered. When polymeric
reinforcements are used in the tests, the interpretation of the results are
further complicated by the stress-strain properties of the material. This
complex behaviour has made the interpretation of pull-out test results
very difficult. To overcome this, it is useful to carry out numerical
modelling to analyse the behaviour of the pull-out operation.
Finite element modelling of pull-out tests 45

Two parameters are required for carrying out such an analysis the
elastic modulus or moduli of the reinforcement and the shear modulus at
the soil/reinforcement interface. In previous experimental pull-out tests
carried out by researchers (Mashhour, 1979; Palmeira & Miligan, 1990),
only loads or strains were measured, which meant that the actual elastic
moduli along the various sections of the reinforcement cannot be
obtained. In addition, the tests provided no information on the shear
stresses developed along the reinforcement. It is therefore not possible to
provide the elastic moduli and the shear modulus from the experiment for
the finite element simulation. The behaviour obtained from the simula-
tion, therefore, cannot be quantified and may not be valid for design
purposes.
The pull-out test mechanism is shown schematically in Fig. l(a). When
a load is applied to the front end of the reinforcement, shear stresses at the
soil-reinforcement interface are developed as the reinforcement is
strained. Load is then transferred progressively along the entire length of
the reinforcement. The rate of transfer and the magnitude of the load
passed are dependent on the reinforcement length due to the variation of
shear stresses along the reinforcement length. A typical load-strain varia-
tion along a reinforcement length is shown in Fig. l(b).
With long reinforcements, the far end of the reinforcement could lay
dormant for small pull-out displacements and may only be activated at

Gn

~4

pull out force ~ "--- ~

T| > T2 > T3 • T4

(a)

I
W
m

Distance along geogrld

(b)
Fig. 1. Pull-out test mechanism.
46 I. Yogarajah, K.C. Yeo

large displacements. When a constant rate of deformation is applied on


the reinforcement, high frictional resistances develop over the entire length
to prevent displacement. These high frictional resistances cause a very
high strain rate to develop at the loaded end and sharply reduced strain
rates towards the rear. The high strain rate causes a high load with low
strain at the loaded end of the reinforcement, due to the visco-elastic
nature of the material. When the load applied is greater than the rupture
load of the reinforcement, the reinforcement ruptures. On the other hand,
if a shorter length of reinforcement is used, the frictional resistance along
the reinforcement is lower, thus allowing displacement to occur over the
entire length. A lower strain rate is developed at the loaded end and a
gradually reduced strain rate occurs towards the rear of the reinforcement.
Slippage may occur if the load generated is higher than the frictional
resistance developed along the reinforcement length.
Both these factors (the length of reinforcement and the rate of defor-
mation) are interdependent and dictate the stresses and strains along the
reinforcement, which in turn affects the elastic moduli of the reinforce-
ment and the shear modulus at the soil reinforcement interface. To enable
accurate modelling of the pull-out behaviour, it is important to relate the
shear modulus to the experimental data. If the elastic moduli of the rein-
forcement can be determined, numerical modelling can be carried out to
back-analyse both the measured loads and strains along the reinforcement
and thus provide a shear modulus through calibration. By modelling
various reinforcement lengths, shear moduli for different anchorage
lengths can be determined. With sufficient experiments, various sets of
elastic and shear moduli for a range of reinforcement lengths can be
determined and used for future design applications.

TEST APPARATUS, SOIL AND R E I N F O R C E M E N T PROPERTIES

Test apparatus

The layout of the test apparatus is shown in Fig. 2. A rectangular steel


tank of 2m (length) x 0.5 m (width) x l m (height) with two reinforced
glass panels of thickness 12mm was mounted on the 2m sides. The back
end of the tank was sealed, while the front end was mounted with a
supported plywood facing with a slot at the mid-height to allow the
geogrid to pass through. Brushes were mounted on the slot to prevent the
loss of soil during the pull-out operation. The pull-out force was provided
by a jack mounted outside the tank.
Finite element modelling Of pull-out tests 47

Glass sides

2000 ~I

Fig. 2. Pull-out test apparatus.

Soil

Uniformly graded Leighton Buzzard sand was used for the tests. The sand
was deposited in the tank by means of a sand spreader to obtain a density
of 17-5 kN/m 3. The density was measured by placing sand pots at various
depths during the sand placement and subsequently measuring the density
of the sand in the pots. The peak angle of friction measured by triaxial
tests at the same density was 47 ° .

Reinforcement

The reinforcement used was Tensar SR 80 geogrid. As a result of the


visco-elastic nature of the polymer, the load-strain behaviour of the
geogrid is time-dependent. The appropriate load-strain properties for
design can be obtained by isochronous curves by using Sherby-Dorn plots
(Yeo, 1985).
In each test, three strips of reinforcements were used. The central strip
was three ribs wide, and is referred to as the instrumented strip. Two five-
rib wide strips referred to as dummy strips were placed on either side of
the instrumented strip, to prevent edge effects from the sides of the tank.

INSTRUMENTATION

To reflect the correct in-soil conditions, instrumentation was employed to


measure both loads and strains along the reinforcement. Load cells were
made of high-yield steel plates, 20 mm long x 10 mm wide, attached with
high-elongation strain gauges, forming a Wheatstone bridge. The geogrid
was cut at various sections, and clamps of width equivalent to the instru-
mented sample were used to connect the load cells to the reinforcement.
48 1. Yogarajah, K.C. Yeo

Strains were measured with Showa high-elongation strain gauges directly


attached to the polymeric geogrid. The gauges were connected to form a
Wheatstone bridge, with the two active gauges attached to the geogrid,
and the two d u m m y gauges attached to an isolated piece of similar geogrid
placed in the same environment as the test grid. The detailed locations of
the load and strain instrumentation for the two lengths of reinforcement
used in the test are shown in Fig. 3.
Two 20-mm diameter wire conduits were laid on either side of the
instrumented sample to accommodate and protect the wires. The conduits
were greased and membraned to prevent any frictional interaction with
the soil. Pull-out tests were carried out with and without the conduits, and
the measurements checked to ensure full conformance. The instrumenta-
tion was connected to an electronic datalogger programmed to scan the
results at l-rain intervals.
Calibration of the strain gauges and load cells was carried out prior to
the tests.

TEST P R O C E D U R E

The tank was initially filled to half-height by the sand spreader at the
appropriate intensity, and the test grid with the greased conduits carrying
the wires were laid over the sand. The test grid and the conduits were
attached to a clamp, which in turn was connected to the pull-out jack. The
tank was then filled to the required height. Two tests with different rein-
forcement lengths of 1.8m (Test A) and 1.1 m (Test B) were carried out
with a uniform overburden pressure of 6.5 k N / m 2. Both the tests were
carried out at a displacement rate of 1 mm/min.

TEST R E S U L T S

The load and strain results obtained in the tests are shown in Figs 4 7.
The strain and load distributions for all the tests follow approximately the

l.lm sample ~Load cells


Strain~auges /

1.8m sample
Fig.3.Positionsof straingaugesand loadcellsalongthereinforcement.
Finite element modelling of pull-out tests 49
25
x 10mrn Dlsld. (Expt.)
+ 10ram DtspI.(FEM)
20 • 20ram Dlapl. (F..xpt)
2omm Dlepl(FEM)

o 10 ,+
\\

0.5 1 1.5
Distance along geogrid (m)
Fig. 4. Load distribution along the geogrid (1.8 m).

same trend, with maximum values recorded at the loaded end and values
reducing to zero at the free end of the reinforcement. In Test A, the loaded
end of the reinforcement recorded a load of 2 4 k N / m and a strain of
1.55% (caused by the high strain rate imposed). Thereafter, the load
decreased sharply, while the strain gradually reduced (caused by the low
strain rate), Figs 4 and 5. In the case of Test B, the loaded end of the

2.5

X lOmm DislN. (Expt.)


+ 10mm DIspl. (FEM)
2 • 20ram DIspl. (Expt)
20mm Displ (FEM)

A 1.5

x\

0.5
\x

0 "~- X
0.5 1 1.5
Distance along geogrid (m)
Fig. 5. Strain distribution along the geogrid (1.8 m).
50 1. Yogarajah, K.C. Yeo

16

14

12

10
E
!\r\ \
~6 \

4
\x

0 i 'z. ~ ,
0 0.5 1 1.5
Distance along geogrid (m)
Fig. 6. Load distribution along the geogrid (1-1 m).
reinforcement recorded a load of 15kN/m and a strain of 2.3%. There-
after, both the loads and strains decreased gradually (Figs 6 and 7). The
experimental results agree with the mechanism previously explained.

FINITE ELEMENT ANALYSIS

The CRISP finite element program developed at the University of


Cambridge (Britto & Gunn, 1990) was used to model the pull-out test.
2.5

,~ x lOmrn Diapl. (Expt.)


+ lOrnm DispI.(FEM)
2 • 2OmmDispL (Expt)
~ + ~ ~ 20mm Displ (FEM)

A 1.5

ta0
1

0,5 \ •
\ x

0
0 0.5 1 1.5 2
Distance along geogdd (m)
Fig. 7. Strain distribution along the geogrid (1.1 m).
Finite element modelling o f pull-out tests 51

The finite element mesh used is shown in Fig. 8. It consists of 232 nodes
and 210 elements. The soil mass was modelled using quadrilateral
elements with a M o h r - C o u l o m b failure criteria, while a bar element with
linear axial stiffness was used to model the reinforcement. Joint elements,
as proposed by G o o d m a n et al. (1968), were used to model the interface
between the soil and reinforcement.
The input parameters for the simulation are summarised in Table 1. The
backfill properties were obtained from triaxial tests. The apparent fric-
tional angle at the soil-reinforcement interface (6) was taken as 0.9 0'p.
Three different elastic moduli were applied along three sections over the
length of reinforcement. The elastic modulus for each section was
obtained from the ratio of measured loads and strains of the particular
sections, as shown in Table 1. Shear moduli (k~) of 1 x 106 k N / m 2 (Test
A) and 1 × 104kN/m 2 (Test B) with a residual shear value (ksre~) of
1 × 103kN/m 2 and the normal modulus (kn) of l × 107kN/rn 2 were
applied. The residual shear modulus k~re~ was selected based on the
assumption that partial slippage would occur once the shear strength was
reached.

Table 1
Parameters Employed for the Finite Element Simulation

Test Length (m) Modulus ( k N / m ) k,1 k~ k ....


( k N / m 2) (kN/m2) ( k N / m e)

A 0-0 0.4 8.0 × 105 10 v 10(~ 103 43


0.4-0.8 4.5 x 105
0.8 1-8 3.0 × 105
B 0.0 0.4 6.5 × 10 5 10 7 10 4 10 3 43
0.4-0.8 6-2 × 105
0.8 1.1 6.0 × 105

Soil: c = 0 k N / m 2 0'p 47 E = 80 × 103 k N / m 2

II IIIII
Re£nforcement Interface

// ../, /,i/,/.,////,/,/ /~ ;//

i s°iI I
I Ii
Fig. 8. Finite element mesh employed for analysis.
52 I. Yogarajah, K.C. Yeo

The results of the finite element simulation are in good agreement with
the experimental results for the load and strain distributions (Figs 4-7).
The shear modulus obtained in Test A was higher than that obtained in
Test B. This is because in Test A the reinforcement length is longer, thus
providing a higher shear (frictional) resistance.
Comparison was made by carrying out an additional finite element
analysis for Test B using a single elastic modulus for the reinforcement.
All the other parameters applied for the model were similar to those
applied in the earlier simulation. An elastic modulus of 4 × 10 2 kN/m 2 was
used throughout the length of the reinforcement. The modulus was
obtained from the secant modulus of isochronous curves at a temperature
of 10°C and a strain of 10% (Kwok, 1987). The results are shown in Figs 9
and 10. The load distribution obtained agrees with the measured data
while the strain distribution obtained was consistently higher than the
measured data (Fig. 10).

CONCLUSION

The measurement of loads and strains has led to a better understanding of


the reinforcement behaviour during the pull-out test. Because of the visco-
elastic nature of the polymeric material, it is vital to have both the
measurements to enable adequate finite element modelling of the reinfor-

16
X lOrnrn DispL (Expt,)
14 ! 10ram DispI.(FEM)
2Omrn Displ. (Expt)
12 20mm Disp~ (FEM)

A10
z
~8

6 \ e

'

0 0.5 1 1.5 2
Distance along geogrid (m)
Fig. 9. Load distribution along the geogrid using a single elastic modulus throughout the
length (1.1 m).
Finite element modelling of pull-out tests 53

2.5-~

g'ls 2' ~
', . \
" \X

ol
0
+'t
0.5 1 1.5 2
Distance along geogrid (m)
Fig. 10. Strain distribution along the geogrid using a single elastic modulus throughout the
length (1.1 m).

cement behaviour. With the evaluation of the elastic moduli of the rein-
forcement (obtained from the measurements), the shear modulus of the
interface can then be calibrated. The technique can subsequently be used
in the modelling of reinforced soil structures. In addition, the technique of
applying varying elastic moduli can be used to simulate the visco-elastic
behaviour of the polymeric reinforcement.

ACKNOWLEDGEMENTS

The authors would like to express their gratitude to Professors M c G o w n


and Andrawes for the use of the laboratories and for their invaluable
assistance.

REFERENCES

Bauer, G.E. & Mowafy, Y.M. (1988). The interaction mechanism of granular
soils with geogrids. In Proceedings of the International Conference Numerical
Methods in Geomechanics, Innsbruck, pp. 1263 72.
Britto, A.M. & Gunn, M.J. (1990). Critical State Soil Mechanics via Finite
Elements. Ellis Horwood, Chichester, UK.
Garbulewski, K. (1990). Direct shear and pullout resistance at the geotextile
interface. In Proceedings of the Fourth International Conference on Geotextiles,
Geomembranes and Related Products, The Hague, Netherlands, pp. 737~12.
54 I. Yogarajah, K.C. Yeo

Goodman, R.E., Taylor, R.L. & Brekke, T.L. (1968). A model for the mechanics
of jointed rocks. Journal of Soil Mechanics Engineering and Foundations Divi-
sion, ASCE, 94, 637 59.
Handel, E., Schweigger, H.F. & Yeo, K.C. (1990). A simple thin-layer element to
model soil geotextile interaction. In Proceedings o["the International ReinJorced
Soil Conference, Glasgow, UK.
Jewel, R., Milligan, G.W.E., Sarsby, R.W. & Dubois, D. (1984). Interaction
between soil and geogrids. In Proceedings of the Symposium on Polymer Grid
Reinforcement in Civil Engineering, London, pp. 18 29.
Kwok, C.M. (1987). Finite element study of embankments on soft clay. PhD
thesis, Sheffield University, UK.
Mashhour, M. (1979). The behaviour of model granular embankments with and
without fabric inclusions. PhD thesis, University of Strathclyde, UK.
Palmeira, E. M. & Milligan, G. W. E. (1990). Scale and other factors affecting the
results of pull-out tests of grids buried in sand. Geotechnique, 39 (3), 511-24.
Venkatappa, Rao G. & Kate, J.M. (1990). Interface friction evaluation of some
Indian geotextiles. In Proceedings of the Fourth International Conference on
Geotextiles,. Geomembranes and Related Products, The Hague, Netherlands, p.
793.
Yeo, K.C. (1985). The use of polymeric geogrids in soil reinforcement. PhD
thesis, Strathclyde University, UK.

You might also like