Geotextiles and Geomembranes 37 (2013) 35e44

Contents lists available at SciVerse ScienceDirect

Geotextiles and Geomembranes

journal homepage: www.elsevier.com/locate/geotexmem

Confinement effect of geocells on sand samples under triaxial compression

Rong-Her Chen*, Yu-Wen Huang, Feng-Chi Huang
Department of Civil Engineering, National Taiwan University, No. 1, Sec. 4, Roosevelt Rd., Taipei 10617, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: The confinement effect of geocells improves vastly the shear strength of granular soil. To understand the
Received 17 September 2012 mechanism better, several tests have been performed on geocell-reinforced-sand samples of two dif-
Received in revised form ferent sizes. The geocells were made of high-density polyethylene sheets, and the influencing factors
8 January 2013
examined include the shape (circular, rectangular, and hexagonal cross-sections), size and number of
Accepted 24 January 2013
cells. The effects of these variables on the compression strength of samples as well as the stressestrain
Available online 1 March 2013
behavior were investigated. It has been found that the apparent cohesion of reinforced samples vary with
the shape, size and number of cells, of which the cell size is the most significant factor. Among the cells of
Keywords:
Geocells
all shapes, the circular cells induce the highest apparent cohesion. In addition, the effectiveness of the
Compression test reinforcement is more significant at low confining pressure. This can be explained by theoretical analysis,
Compression strength which shows that the reinforced samples under low confining pressures tend to expand more and induce
Confining effect higher circumferential strain. When under high confining pressure, the samples undergo lesser dilation
Apparent cohesion leading to lower hoop forces in geocells.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction and unreinforced soil samples showed that the peak friction angle
of the soil infill and that of the composite were the same. Rajagopal
Geocells are widely applied in geotechnical engineering, such as et al. (1999) performed triaxial compression tests on granular soils
controlling erosion of slopes and river banks, enhancing bearing encased in single and multiple geocell samples with 100-mm
capacities of pavements and footings, reinforcing soft grounds and diameter. The geocells were fabricated by hand using different
slopes, and protecting shores and channel beds. The cells are often woven and nonwoven geotextiles and soft mesh to investigate the
made from high-density polyethylene or polyethylene strips ul- effect of stiffness of geocell on overall performance of the com-
trasonically welded together to give an open-cell structure for posite. It was observed that the soil developed cohesive strength
containing soil. Such confinement provided by geocells improves resulting from the confinement by the geocell, and the magnitude
vastly shear strength of granular soil; and, in turn, the increase in of the cohesive strength varied with the properties of the geocell.
soil strength provides improved bearing capacity and/or prevents They also presented a simple methodology for estimating the
soil erosion. magnitude of the apparent cohesive strength. Shen (2005) per-
The confinement of geocells has been studied mainly in the formed triaxial compression tests to investigate the effect of rela-
laboratory using triaxial compression test and uniaxial com- tive density of soils on compression strength of two different
pression test. With regard to triaxial compression tests, Bathurst aggregates reinforced with geocells. It was found both the peak
and Karpurapu (1993) reported the results of a series of large- friction angle and apparent cohesion of the reinforced soils
scale triaxial compression tests carried out on 200-mm-high iso- increased with relative density. Latha and Murthy (2007) studied
lated geocell soil composite samples and unreinforced soil samples. and compared the relative efficiency of three forms of re-
The soils tested were silica sand and limestone, with free drainage inforcements, namely, horizontal geosynthetic layers, randomly
of water in the soil samples allowed. The results illustrated stiff- oriented discrete fibers, and geocells. Uniaxial compression tests
ening effect and increase in strength imparted to the soil by the have also been conducted on large samples. Bathurst and Crowe
enhanced confinement effect. Moreover, comparison of reinforced (1994) described two large-scale uniaxial compression tests (1 m
wide
1 m long
1.44 m high) to examine the stability of multiple
layers of geocells under vertical surcharge. The geocells were
200 mm high and had a cell diameter of 200 mm. The soils tested
* Corresponding author. Tel.: þ886 2 3366 4242; fax: þ886 2 23629851.
E-mail addresses: rongherchen@ntu.edu.tw (R.-H. Chen), irene0202@msn.com were sand and limestone. They reported that the sand column and
(Y.-W. Huang), d97521007@ntu.edu.tw (F.-C. Huang). the limestone aggregate column were respectively stable up to

0266-1144/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.geotexmem.2013.01.004
36 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44

10 m and 13.5 m of equivalent heights of the structure. Wesseloo Table 1

et al. (2009) discussed the results of uniaxial compression tests Strength and elastic modulus of model and prototype materials.

performed on single cells and geocell packs of different sizes, and Model material Prototype materiala
the geocells were mainly square in shape. They found that the Tensile strength T (kN/m) 15.0 17.0
strength of the geocell composite structure was indirectly propor- Junction shear strength Sj (kN/m) 8.2 13.5
tional to the size of individual cells and that the strength reduced Junction peel strength Pj (kN/m) 6.4 9.5
Junction split strength Uj (kN/m) 8.6 12.8
with increasing number of cells in the structure. To evaluate the
Secant modulus Js (kN/m) 101 494
interaction between hoop stresses and passive earth resistance,
a
Shen (2005).
Emersleben and Meyer (2009) carried out radial load tests on single
and multiple geocell structures.
Considerable interest has also been shown in studying geocell-
strength as well as the stressestrain behavior of reinforced-sand
reinforced structures, such as shallow foundations (Dash et al.,
samples under triaxial compression. In the literature, geocells
2003; Sireesh et al., 2009; Pokharel et al., 2010; Moghaddas
adopted for testing were most commonly of circular cross-section,
Tafreshi and Dawson, 2010, 2012), slopes (Chen and Chiu, 2008;
probably because it is easier to fit cylindrical geocells into the
Leshchinsky et al., 2009), pavements (Thakur, et al., 2012;
chamber of the triaxial apparatus (Bathurst and Karpurapu, 1993;
Leshchinsky and Ling, 2013), and embankments (Zhang et al.,
Rajagopal et al., 1999; Shen, 2005). Geocells of square cross-section
2010). Nonetheless, there were few studies reporting the funda-
have also been adopted but not common (Wesseloo et al., 2009). In
mental understanding of these systems. To make up for such
practice, geocells used in the field tend to be of rhombic cross-
deficiency, this study performed tests on geocell-reinforced-sand
section. Thus, it is of interest to compare the confinement effects
samples under triaxial compression to examine the effect of vari-
of cells of different cross-sections. In this study, the geocells for
ables on compression strength and deformation characteristics of
testing were of three types, namely, cylindrical, hexagonal and
the samples. It is hoped that the results from this study will
rectangular columns. Moreover, in order to study multiple-cell ef-
improve the understanding of the confinement effect of geocells
fect, the samples were prepared to contain one cell, three cells, and
when applied to geotechnical engineering. Furthermore, the test
seven cells.
results will be good for calibrating numerical analysis, which allows
for extrapolation to realistic setup of geocells in the field.
3. Testing program and procedure
2. Materials used for testing
The variables considered in the testing program are as follows.
2.1. Soil
 Sand with and without reinforcement
The sample soil used in the present investigation was uniform,  Sample size e small sample (2.8 inches in diameter and 6-
sub-angular sand containing 99.8% silica. It is classified as poorly- inches high) and large sample (6 inches in diameter and 12-
graded sand (SP) as per the Unified Soil Classification system, with inches high)
a coefficient of uniformity Cu ¼ 2.07 and a coefficient of curvature  Cell shape e the cross-sections of cells were circular, rectangle,
Cc ¼ 1.04. The physical properties of the sand are as follows: specific and hAexagonal
gravity Gs ¼ 2.66, median grain size d50 ¼ 0.25 mm, effective grain  Cell size e cells of the same shape but different sizes
size d10 ¼ 0.15 mm, maximum dry unit weight gd,max ¼ 16.1 kN/m3,  Number of cells e different numbers of cells in same area
and minimum dry unit weight gd,min ¼ 13.3 kN/m3.
The content of the testing program is shown in Table 2. For the
2.2. Geocell material 10 tests listed, the first letter of the samples, S, represents sand; the
second letter denotes the shape of the cells; and the third letter
The material used for fabricating geocells was high-density denotes the number of cells in the sample. Samples marked with an
polyethylene sheet, for it is easy to obtain and process. The physi- asterisk are small samples of 2.8-inch diameter, while those with-
cal properties of the material are as follows: specific gravity out an asterisk are large samples of 6-inch diameter. The configu-
Gs ¼ 0.967 (ASTM D 792, 2008), thickness t ¼ 0.38 mm (ASTM D 5199, rations of all samples are depicted in Fig. 1. A photo of hexagonal
2012), and mass per unit area mA ¼ 118 g/m2 (ASTM D 5261, 2010). geocell columns is shown in Fig. 2.
The strength tests performed on this geocell material include The samples for triaxial compression were prepared according
tensile strength test on the material and the tests for three failure to the procedures of ASTM D 4767 (2011). Specifically, all samples
modes of the junction of a geocell: junction shear failure, junction
peel failure, and junction split failure (Cancelli et al., 1993). The
Table 2
tensile strength is 15 kN/m obtained from a wide-width sample, The testing program.
according to ASTM D 4595 (2011). The sample for junction strength
Soil Sample Diameter of sample Shape of cell Number of cell
tests had a width equal to the seam spacing and a length equal to
the height of cell. Forces were applied at both ends of the sample Sand S* 2.8 in. (7.1 cm) e e
Sand S 6 in. (15.2 cm) e e
until the junction failed. The results presented in Table 1 are Reinforced sand SC1* 2.8 in. (7.1 cm) Circle 1
junction shear strength Sj ¼ 8.2 kN/m, junction peel strength Reinforced sand SH1* 2.8 in. (7.1 cm) Hexagon 1
Pj ¼ 6.4 kN/m, and junction split strength Uj ¼ 8.6 kN/m. In Table 1, Reinforced sand SH3* 2.8 in. (7.1 cm) Hexagon 3
the results of another geocell material (Shen, 2005), which is also Reinforced sand SC1 6 in. (15.2 cm) Circle 1
Reinforced sand SC3 6 in. (15.2 cm) Circle 3
made of HDPE and commonly used in the field, are presented for
Reinforced sand SB3 6 in. (15.2 cm) Rectangle 3
comparison. It is noteworthy that the secant modulus (Js at 2% Reinforced sand SH3 6 in. (15.2 cm) Hexagon 3
strain) of the model material is only about 20% of the prototype Reinforced sand SH7 6 in. (15.2 cm) Hexagon 7
material, though their tensile strengths are not much different. Note: 1. S ¼ sand, C ¼ circle, H ¼ hexagon, B ¼ rectangle (or block).
As mentioned previously, this study aimed to investigate the 2. The small sample (2.8 inches in diameter) is indicated with *; otherwise the
effects of shape, size, and number of geocells on the compression sample is a large sample (6 inches in diameter).
 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44 37

were prepared under dry condition. Five layers of pre-weighted

quantity of sand were placed in the sample mold with light
tamping. Each layer of sand was 30 mm and 60 mm thick for small
and large samples, respectively. This procedure produced uniform
samples with a relative density of 55%. All samples were then
saturated and consolidated to reach equilibrium in a drained state
at the effective consolidation stress for which a strength determi-
nation is required.
All tests were carried out at a compressive strain rate of 0.1% per
minute. During the test, water was allowed to drain freely from the
sample. This is because the soil infill with geocells in the field is
often permeable and the test soil is also a pervious material. In
addition, the effect of rubber membrane increases as the sample is
compressed; therefore, the effect of the membrane should be cor-
rected so that the compression strength of the sample can be
obtained accurately. The correction of the effect of membrane,
which is 1.58 mm thick, was made according to ASTM D 4767 (2011).
Generally speaking, the correction will be of significance only when
the effective stresses are small (Henkel and Gilbert, 1952).
It also needs to note that a reinforced-sand sample is a com-
posite material, in which the stress and strain may not be as uni-
form as in an unreinforced-sand sample. Before the effect is fully
understood, the test results in the following were obtained on the
assumption that the stress and strain in reinforced-sand samples
are uniformly distributed.

4. Test results

4.1. Unreinforced sand

To examine the effect of sample size on compression strength of

sand without reinforcement, a small sample (S*) and a large sample
(S), shown in Fig. 1, were tested for comparison. The peak stresses of
the two samples occurred at axial strains of about 6% (Fig. 3a). After
that, the large sample displayed more obvious strain-softening
behavior, especially under high confining pressure (200 kPa). As
for volume change, the samples were compressed in the beginning
and subsequently dilated as axial strain increased (Fig. 3b). The
small sample tended to dilate more than the large sample; how-
ever, the dilation of both samples decreased when the samples
were subjected to higher confining pressures.
From MohreCoulomb failure envelopes, the friction angles of
Fig. 1. Different configurations of the samples used in compression tests. the small and large samples were 37 and 38 , respectively. The
cohesion intercepts were both small and thus neglected. From this
result, it can be concluded that sample size has insignificant effect
on compression strength of unreinforced sand.

4.2. Reinforced sand (2.8-inch sample)

For the small sample of reinforced sand, three tests were con-
ducted: SC1*, SH1*, and SH3* (Fig. 1). The results of the first two
samples, SC1* and SH1*, are employed to study the effect of cell
shape; these two samples had equal area confined by each geocell.
On the other hand, the effect of multiple cells are compared by
examining the results of SH1* and SH3*.

4.2.1. Effect of cell shape

In Fig. 4, the results of reinforced samples, SC1* and SH1*, and
unreinforced sample, S*, are compared. It can be seen from Fig. 4a
that the curves of reinforced samples do not show the strain-
softening phenomenon as occurred in the unreinforced sample.
The geocells changed the sand from brittle to ductile, and therefore
the reinforced samples displayed better strength. Furthermore, the
sample with circular cell, SC1*, had higher compression strength
Fig. 2. Photo of hexagonal geocell columns. than that with hexagonal cell, SH1*. Correspondingly, the volumetric
38 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44

1000
Confining pressure
… S*
— S
50 kPa
800 50 kpa
100 kPa
Deviatoric stress , Δσ (kPa)

100 kPa
200 kPa
200 kPa
600

400

200

0
0 4 8 12 16 20
(a)
Axial strain , εa (%)
(a)

5
Volumetric stress , εv (%)

Confining pressure
… S*

1 — S
50 kPa
50 kpa
100 kPa
100 kPa
200 kPa
200 kPa
-1
0 4 8 12 16 20
(b)
Axial strain , εa (%) Fig. 4. Results of compression tests on 2.8-inch samples comprising cells of different
(b) shapes. (a) Stress versus strain. (b) Volumetric strain versus axial strain.

Fig. 3. Results of compression tests on 2.8-inch and 6-inch samples of sand. (a) Stress
versus strain. (b) Volumetric strain versus axial strain. Findings on small samples are summarized as follows. First,
reinforced samples presented higher stiffness, or steeper stresse
strain of SC1* is greater than that of SH1* (Fig. 4b), because the cir- strain curve, than unreinforced samples; nevertheless, this differ-
cular shape expands more easily than the irregular hexagonal shape. ence became less significant with increasing confining pressure.
The strength parameters of the three tests are tabulated in Second, samples consisted of more small cells showed stiffer
Table 3. As can be seen, S* has c ¼ 0 and f ¼ 37 ; SC1* has behavior than those consisted of large but fewer cells. Third, among
cr ¼ 79 kPa and f ¼ 38 ; SH1* has cr ¼ 35 kPa and f ¼ 38 . Obvi- the three samples (SC1*, SH1*, and SH3*), the cylindrical sample
ously, reinforcement improved the cohesion but had little con- was the stiffest.
tribution to the friction angle of the reinforced samples. This
finding is in line with previously reported results (e.g., Bathurst and 4.3. Reinforced sand (6-inch sample)
Karpurapu, 1993). It can also be concluded that the improvement in
cohesion is attributed to the induced tensile strain in the geocell, For large samples of reinforced sand, five tests were conducted
which is a function of cell shape as well. (see Fig. 1). The three samples used for comparing the effect of

4.2.2. Effect of cell size Table 3

With regard to the effect of the size of cells, Fig. 5 shows photos Results of compression test.
of two samples after failure. The single-cell sample, SH1*, were Sample Diameter of Equivalent Apparent Friction Ratio crde
significantly deformed, while its corners were less affected. On the sample D0 (cm) diameter cohesion angle 4 ( ) (kN/m)
other hand, the sample with three cells had shorter sides, and de (cm) cr (kPa)
consequently the deformation was not as significant as that of SH1*. S* 7.1 7.1 0 37 e
Comparing the stressestrain curves under different confining SC1* 7.1 6.0 79.0 38 4.74
SH1* 7.1 6.0 35.0 38 2.10
pressures shows that the curves of SH3* are higher than those of
SH3* 7.1 3.15 44.2 39 1.39
SH1* (Fig. 6a). Correspondingly, the curves of volumetric strain for S 15.2 15.2 0 38 e
SH3* should also be higher, irrespective of the confining pressure SC1 15.2 15 33.7 37 5.06
(Fig. 6b). Likewise, the strength parameters of SH3*, cr ¼ 44.2 kPa SC3 15.2 7.0 59.4 36 4.16
and f ¼ 39 , are also higher than those of SH1* (Table 3). It is SB3 15.2 7.0 31.0 36 2.17
SH3 15.2 7.0 23.7 38 1.66
concluded that apparent cohesion decreases with the size of cells; SH7 15.2 4.62 34.5 38 1.59
however, the friction angles were about the same.
 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44 39

(a)

Fig. 5. Photos of 2.8-inch samples after test. (a) SH1*. (b) SH3*.

(b)
shape are SC3, SH3, and SB3. In order to have equal basis for com-
parison, the cell of each shape had the same area, as marked by the Fig. 6. Results of compression tests on 2.8-inch samples comprising different number
of hexagonal cells. (a) Stress versus strain. (b) Volumetric strain versus axial strain.
shaded area in Fig. 7a. With regard to the effect of number of cells,
SC1 versus SC3 and SH3 versus SH7 are compared, respectively.
connected only at three points, while the hexagons and blocks were
4.3.1. Effect of cell shape connected along the junctions between cells. As a result, the circles
Figs. 8a and b presents the stressestrain curves of three rein- expanded more freely than the hexagons and blocks and mostly
forced samples and one unreinforced sample under low and high along their outer boundaries. As seen in the photos shown in Fig. 7b,
confining pressures, respectively. The curves of the three reinforced the circles became somewhat irregular, but the hexagons still
samples under low confining pressure (Fig. 8a) were significantly maintained their original shapes. Notably, the rectangles shows quite
higher than that of unreinforced sand. Among the reinforced irregular deformation with their edges deformed much more than
samples, SC3 showed the highest strength, while the strength of their inner junctions. This demonstrates that adjacent cells
SB3 was the lowest. Notably, a rapid increase in deviatoric stress restrained one another from expansion, consequently resulting in
was observed for SC3 when axial strain exceeded 10%. In contrast, reduced development of tensile strain as well as tensile strength in
the stressestrain curves of all samples under high confining pres- the cells.
sure (Fig. 8b) were much closer, compared with those in Fig. 8a,
suggesting that the effect of reinforcement was more significant 4.3.2. Effect of cell size
under low confining pressure. The effect of cell size is examined by comparing SH3 and SH7 (see
The variations in volumetric strain for large samples were Fig. 1). As shown in Fig. 9, the stressestrain curves of SH7 are higher
similar to those for small samples (e.g., Figs. 4b and 6b). Of the three than those of SH3 under all confining pressures, though the differ-
reinforced samples, SC3 showed the highest trend of increase in ence is not significant at confining pressure of 50 kPa. Correspond-
volume, which is comparable to its highest compression strength. ingly, the curves of volumetric strain for SH7 are also higher than
However, the variations in volumetric strain for rectangular-shaped those of SH3 (Fig. 9b). In summary, using greater number of smaller
cell, SB3, were found to be not as regular as those of SC3 and SH3, cells to confine soil will induce higher compression strength as well
probably due to unsymmetrical geometry of blocks. The photos of as greater volumetric strain in the reinforced soil samples.
the three samples after tests are displayed in Fig. 7b for clarifying The other comparison is obtained from circular cells, i.e., SC1
this explanation. versus SC3. Similar observation as that for hexagonal geocells (SH3
The strength parameters of SC3, SB3, and SH3 are tabulated in versus SH7) was observed, and hence no further discussion will be
Table 3. Among them, SC3 has the highest apparent cohesion made.
cr ¼ 59.4 kPa, while SH3 has the slightly highest friction angle, The strength parameters of SC1, SC3, SH3, and SH7 are sum-
f ¼ 38 . These results are attributed to the cell shape as well as the marized in Table 3. As can be seen, the friction angles for cells of the
junction between cells. As shown in Fig. 7a, the three circles were same shape show little variations, but the apparent cohesion is
40 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44

Fig. 7. Comparison of three 6-inch samples reinforced with cells of different shapes. (a) Different configurations. (b) Photos of samples after tests (s3 ¼ 200 kPa).

much higher for samples with small cells, e.g., 59.4 kPa of SC3 5.2. Effect of cell size
greater than 33.7 kPa of SC1.
The relationship between apparent cohesion and the equivalent
diameter of a single cell, de, is shown in Fig. 11. The equivalent
5. Discussion diameter originates from the concept of the circumference per unit
area under confinement. For example, if a circular cell has a diam-
The effects of shape, size, and number of cells are discussed in eter d, then its circumference is pd and the confined area is pd2/4.
this section. In addition, improvement in the elastic modulus and Hence, the circumference per unit area is 4/d. For cells other than
compression strength of reinforced sand due to confinement of cell the circular ones, their areas are assumed equivalent to the areas of
is presented. circles, so their equivalent diameters can be obtained in this way.
Note that this relationship is considered only for a single cell.
To compare the difference in mobilization of apparent cohesion
5.1. Effect of cell shape among various shapes, the relationships between apparent cohe-
sion and the inverse of equivalent diameter are plotted in Fig. 11:
As summarized in Table 3, the friction angles of unreinforced
samples, S and S*, are 37e38 , while those of reinforced samples are a. Circular cells (for SC1 and SC1*)
36e39 , revealing insignificant effect of geocell on friction angle of
test sand. It is worthy to note that hexagonal cells induced slightly

higher friction angle than cells of the other two shapes. On the cr ¼ 4:97=de R2 ¼ 0:997 (1)
other hand, the increase in apparent cohesion due to reinforcement
of cells of all shapes is obvious. This can be explained by examining
the photos of samples after failure. For the unreinforced sample b. Hexagonal cells (for SH3 and SH3*)
(Fig. 10a), a clear failure plane can be determined as several inclined


stretches were seen on the rubber membrane. For the reinforced cr ¼ 2:92=ðde Þ0:78 R2 ¼ 0:999 (2)
sample (Fig. 10b), it did not fail when the test terminated at an axial
strain of 15%. As a result, the rubber membrane of this sample
showed only horizontal wrinkles. Evidently, the confining effect where cr ¼ apparent cohesion (kPa), de ¼ equivalent diameter of
provided by the cells mobilized the confining strength to prevent a single cell (m).
the soil from failure. As for the contribution to the compression
strength of the reinforced sand, cells of circular shapes demon- The relationships show linear relation for circular cells and
strated the best effect among those of all shapes. somewhat nonlinear relation for hexagonal cells; thus, the
 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44 41

(a) (a)

(b)
(b)
Fig. 9. Results of compression tests on 6-inch samples with different number of
Fig. 8. Results of compression tests on 6-inch samples reinforced with three cells of hexagonal cells. (a) Stress versus strain. (b) Volumetric strain versus axial strain.
different shapes. (a) s3 ¼ 50 kPa. (b) s3 ¼ 200 kPa.

versa. Therefore, for cells of circular shape, the multiple-cell effect

apparent cohesion is approximately proportional to the inverse of between one cell and three cells is estimated to be 0.82 (¼4.16/
the cell diameter, and small cells induce higher apparent cohesion. 5.06). As for samples having more than three cells, comparison is
Note that the curves are assumed to pass through the origin made between SH3 and SH7, and the result is 0.96 (¼1.59/1.66).
because a very large cell will induce very small apparent cohesion. From this observation, according to the results of the compression
In addition to that, the slope of Eq. (1) is more than that of Eq. (2), tests conducted in drained condition, multiple-cell effect seems to
also demonstrating greater effectiveness in using smooth cells of be significant only for samples comprising less than three cells.
circular shape. Although this result is consistent with the observations made by
Rajagopal et al. (1999), further research is still needed.
5.3. Multiple-cell effect
5.4. Improvement in compression strength
It is of interest also to know the difference in apparent cohesion
mobilized by a single cell and multiple cells. As seen in Fig. 11, the The effectiveness of reinforcement can be represented by
deviation of SC3 from the line through SC1 and SC1* is the effect of a deviatoric stress ratio, SR, which is defined as
multiple cells; the same effect is seen between SH1* and the curve
through SH3* and SH3. This topic is important because multiple SR ¼ ðDsr Þmax =ðDsÞmax (3)
cells are more constrained than a single cell, and the effect may be
so significant as to affect the strength parameters of reinforced soil. where (Dsr)max and (Ds)max are the maximum deviatoric stresses
Nevertheless, due to limitation of the apparatus, it is almost of reinforced and unreinforced samples, respectively. As shown in
impossible to conduct tests on samples with many cells. Despite the Fig. 12, the values of SR at confining pressure of 50 kPa are in the
difficulty, the following prediction is made according to the test range of 1.5e2.5. However, the ratio decreases as confining pres-
results in this study. sure increases, e.g., the ratios at confining pressure of 200 kPa are in
In order to eliminate the effect of cell size for direct comparison the range of 1.0e1.5. This result indicates good improvement in the
of the data, each data in Fig. 11 is expressed as a ratio of the ordinate compression strength of unreinforced soil under low confining
to the x-coordinate, i.e., cr/(1/de) or crde, as displayed in the last pressure.
column in Table 3. By doing so, theoretically, the ratio of 5.06 The other findings from Fig. 12 concern the size of sample and
obtained from the single-cell sample (SC1) can be compared with the number of cells. With regard to the size of sample, smaller
the ratio of 4.16 obtained from the three-cell sample (SC3), and vice samples have higher deviatoric stress ratios, e.g., SC1* versus SC1
42 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44

3
SC1*
SH1*
SH3*
2.5
SC1
SC3
SH3
2 SH7
SB3

SR
1.5

0.5
0 50 100 150 200 250

Confining pressure , σ (kPa)

Fig. 10. Photos of reinforced and unreinforced samples after tests (s3 ¼ 200 kPa). (a)
Unreinforced sample, S. (b) Reinforced sample, SH7. Fig. 12. Relationship between deviatoric stress ratio and confining pressure.

(also for SH3* versus SH3). Regarding the number of cells, greater
2Js εc
amount of smaller cells develops higher stress ratio, such as SH3* Ds3 ¼ (5)
D ð1  εa Þ
versus SH1* (also for SC3 versus SC1 and SH7 versus SH3).

6. Theoretical analysis where Ds3 ¼ the rubber correction in this case (kPa), Js ¼ secant
modulus of the membrane at εa (kN/m), D ¼ diameter of the
In studying the effect of rubber membranes on the measured sample at εa (m), εa ¼ axial strain (%), and εc ¼ circumferential
compression strength of saturated clay sample in undrained con- strain (%). Again, Ds3 is in functions of the reciprocal of the
dition, Henkel and Gilbert (1952) introduced two theories, com- diameter of sample, the modulus of membrane, and the
pression shell theory and hoop tension theory, to correct the deformed shape of the sample, εc/(1  εa).
compression strength of the sample. In the compression shell
theory, only the axial strain of the sample is taken into consid- In studying the behavior of reinforced granular soils, Bathurst
eration; while in the hoop tension theory, it considers both the and Karpurapu (1993) and Rajagopal et al. (1999) used the hoop
axial strain and radial strain of the sample. The equations of the two tension theory to predict the apparent cohesion for samples under
theories are as follows: drained and undrained conditions, respectively.

a. Compression shell theory

6.1. Derivation of circumferential strain
pD0 Jc εa
sr ¼ (4) In order to employ Eq. (5) for the drained tests conducted in this
Ac
study, the circumferential strain of the sample need to be derived.
Assuming a sample has an initial volume of V0, and it subse-
where sr ¼ the rubber correction (kPa), Jc ¼ compression mod-
quently decreases to V, as a result of the consolidation of soil and
ulus of the rubber membrane (kN/m), D0 ¼ initial diameter of
water drainage. Therefore,
the sample (m), εa ¼ axial strain (%), and Ac ¼ corrected area of
the sample at strain εa (m2). Note that the ratio, pD0/Ac, in Eq. (4) V
is similar to the circumference per unit confined area mentioned ¼ 1  εv (6)
V0
previously.
b. Hoop tension theory Herein, the volumetric strain, εv, is assumed positive for com-
pression and negative for extension. By substituting the diameter
100 and the length of the sample into above equation, it can be written
as
80 p
Apparent cohesion , cr (kPa)

D2 L
4 ¼ 1  εv (7)
p
60 D20 L0
4
SC1*
SC1 Rearranging this equation leads to the following:
40
SC3
SH1* rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1  εv ÞL0
20
SH3* D ¼ D0 (8)
SH3 L
SH7
Now, using axial strain εa, the equation is changed as below
0
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 20 40 60 80 100 u ð1  ε Þ ð1  εv Þ
u v
1/Equivalent cell diameter , 1/de (m-1 ) D ¼ D0 u ¼ D0 (9)
tðL0  DLÞ ð1  εa Þ
Fig. 11. Relationship between apparent cohesion and equivalent diameter of cell. L0
 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44 43

The circumferential strain, εc, which is the change in length per 800

unit circumferential length, can be expressed as test 50 kPa

test 200 kPa

pD  pD0
hoop 50 kPa
D  D0 600
εc ¼ ¼ (10) hoop 200 kPa
pD0 D0

Shear stress, τ (kPa)

By substituting the diameter, D, from Eq. (9) into the above 400
equation, it then becomes
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1  εv Þ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
200
D0  D0
ð1  εa Þ ð1  εv Þ
εc ¼ ¼ 1 (11)
D0 ð1  εa Þ
0

Accordingly, the circumferential strain can be substituted into 0 400 800 1200
Eq. (5) to calculate the induced confining stress in the reinforce- Normal stress, σ (kPa)
ment, with the secant modulus of membrane, Js, in Eq. (12) replaced (a)
by that of the geocell material.
800
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
ð1  εv Þ test 50 kPa
1 test 200 kPa
2Js ð1  εa Þ shell 50 kPa
Ds3 ¼ (12) 600 shell 200 kPa
D ð1  εa Þ

Shear stress , τ (kPa)

400
6.2. Interpretation

In the following, the compression strengths of reinforced sam- 200

ples are obtained according to Eqs. (4) and (5), respectively. Note
that it was not possible to measure the compression modulus of
rubber membrane; and hence, the compression modulus was 0
assumed to be the same as the secant modulus of the membrane
0 400 800 12 0 0
(Henkel and Gilbert, 1952).
Normal stress , σ (kPa)

a. Compression shell theory (b)

Fig. 13. Comparison between Mohr’s circles from test results and two theories (sample
s1;shell ¼ s1 þ sr (13) SH3*). (a) Hoop tension theory. (b) Compression shell theory.

b. Hoop tension theory prediction by the shell compression theory shows better agree-
ment, meaning that the sample under high confining pressure
s1;hoop ¼ Kp ðs3 þ Ds3 Þ (14) behaves like a compressed column with small volume expansions.
This is because at high confining pressure the dilation of soils is
lesser. As shown in Fig. 10b, several horizontal wrinkles appear on
where s1 is the compression strength of unreinforced soil the surface of the rubber membrane, which were mainly resulted
sample, and Kp is Rankine’s coefficient of passive earth pressure. from axial compression of the sample.

The comparison between Mohr circles from test results and

theoretical prediction are plotted in Fig. 13. The correction for 600
membrane effect, according to ASTM D 4767 (2011), has already
been subtracted from the compression strength before the appro-
priate Mohr circle is drawn. In the figure, the test results are
Mean normal stress, p' (kPa)

denoted by hollow symbols, while solid symbols represent the 400

prediction. As can be seen from Fig. 13a, the prediction by the hoop
tension theory matches well the test results under low confining
pressure of 50 kPa. On the other hand, Fig. 13b shows the prediction
by the compression shell theory is in good agreement with the test 200
results under high confining pressure of 200 kPa. In an alternative
way, the maximum mean normal stresses at various confining SH3* (Test)
SH3* (Shell)
pressures are compared in Fig. 14. This figure shows the transition SH3* (Hoop)
of the failure states with increasing confining pressure. It is clearly 0
seen the test result agrees well with the prediction by the hoop 0 50 100 150 200 250
tension theory when confining pressure is lower than 50 kPa. This
Confining pressure, σ3 (kPa)
behavior is attributed to the high tendency for the reinforcement
under low confining pressure to expand laterally under axial Fig. 14. Comparison between mean normal stresses obtained from test results and two
compression. When the confining pressure exceeds 100 kPa, the theories.
44 R.-H. Chen et al. / Geotextiles and Geomembranes 37 (2013) 35e44

7. Conclusions Dash, S.K., Sireesh, S., Sitharam, T.G., 2003. Model studies on circular footing sup-
ported on geocell reinforced sand underlain by soft clay. Geotextiles and Geo-
membranes 21, 197e219.
This study has employed tests on geocell-reinforced sand sub- Emersleben, A., Meyer, N., 2009. Interaction between hoop stresses and passive
jected to triaxial compression to examine the confinement effect of earth resistance in single and multiple geocell structures. In: GIGSA GeoAfrica
geocells. It was found that the confinement effect provided by 2009 Conference, Cape Town, September 2e5.
Henkel, D.J., Gilbert, G.C., 1952. The effect of rubber membranes on the measured
geocells related mainly to the mobilization of the tensile strength in triaxial compression strength of clay samples. Géotechnique 3, 20e29.
the geocells, which varies with the volumetric strain induced. The Leshchinsky, B., Ling, H.I., 2013. Numerical modeling of behavior of railway bal-
induced volumetric strain is affected by the size and shape of the lasted structure with geocell confinement. Geotextiles and Geomembranes 36,
33e43.
cell, confining pressure, as well as multiple-cell effect. The findings Leshchinsky, D., Ling, H.I., Wang, J.-P., Rosen, A., Mohri, Y., 2009. Equivalent
are summarized as follows. seismic coefficient in geocell retention systems. Geotextiles and Geo-
membranes 27, 9e18.
Latha, G., Murthy, V.S., 2007. Effects of reinforcement form on the behavior of
 The apparent cohesion of reinforced samples varies mainly with geosynthetic reinforced sand. Geotextiles and Geomembranes 25, 23e32.
the size, shape, and number of cells, of which cell size is the most Moghaddas Tafreshi, S.N., Dawson, A.R., 2010. Comparison of bearing capacity of
significant factor. Approximately linear relationships were a strip footing on sand with geocell and with planar forms of geotextile rein-
forcement. Geotextiles and Geomembranes 28, 72e84.
found between the apparent cohesion and the inverse of the Moghaddas Tafreshi, S.N., Dawson, A.R., 2012. A comparison of static and cyclic
equivalent diameter of the cell. Moreover, circular cells induced loading responses of foundations on geocell-reinforced sand. Geotextiles and
the highest apparent cohesion among cells of all shapes, while Geomembranes 32, 55e68.
Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L., Halahmi, I., 2010. Investigation
the apparent cohesion induced by hexagonal cell was the lowest.
of factors influencing behavior of single geocell-reinforced bases under static
 The difference in friction angle between the reinforced and loading. Geotextiles and Geomembranes 28, 570e578.
unreinforced-sand samples was insignificant. For cells of vari- Rajagopal, K., Krishnaswamy, N.R., Latha, G., 1999. Behavior of sand confined with
ous shapes, hexagonal ones showed slightly higher friction single and multiple geocells. Geotextiles and Geomembranes 17, 171e184.
Shen, C.W., 2005. The mechanical characteristics of geocell-reinforced earth. Master
angle, owing to many corners of hexagons. Thesis. Department of Civil Engineering, National Taiwan University (in
 The constraint by adjacent cells prevents the cells from Chinese).
expanding laterally, and consequently the mobilization of the Sireesh, S., Sitharam, T.G., Dash, S.K., 2009. Bearing capacity of circular footing on
geocell-sand mattress overlying clay bed with void. Geotextiles and Geo-
tensile strength of geocells is restrained in multi-cell configu- membranes 27, 89e98.
rations. This effect may be significant and should be considered Thakur, J.K., Han, J., Pokharel, S.K., Parsons, R.L., 2012. Performance of geocell-
when using the data from tests on a single cell. reinforced recycled asphalt pavement (RAP) bases over weak subgrade under
cyclic plate loading. Geotextiles and Geomembranes 35, 14e24.
 With increase in confining pressure on the reinforced soil, the Wesseloo, J., Visser, A.T., Rust, E., 2009. The stressestrain behavior of multiple cell
reinforcing effect becomes less significant, meaning that the geocell packs. Geotextiles and Geomembranes 27, 31e38.
reinforcement is more effective under low confining pressure. Zhang, L., Zhao, M., Shi, C., Zhao, H., 2010. Bearing capacity of geocell reinforcement
in embankment engineering. Geotextiles and Geomembranes 28, 475e482.
 Comparison between theoretical and test results show that the
behavior of reinforced soil under low confining pressure can be
Notations
predicted by the hoop tension theory. The behavior of rein-
forced soil samples under high confining pressure is more
similar to a stiff column under axial compression. Basic SI units are given in parentheses
Ac: corrected cross-sectional area (m2)
Cc: coefficient of curvature (dimensionless)
Acknowledgments Cu: coefficient of uniformity (dimensionless)
c: cohesion intercept (N/m2)
This work was supported by the National Science Council of Tai- cr: apparent cohesion (N/m2)
D: diameter of sample at εa (m)
wan (99-2221-E-002-121). We would also like to thank the three Dr: relative density (%)
anonymous reviewers for their valuable and constructive D0: initial diameter of sample (m)
suggestions. d: diameter of circular cell (m)
de: equivalent diameter of cell (m)
d10: effective grain size (m)
References d50: median grain size (m)
Gs: specific gravity (dimensionless)
ASTM D 792, 2008. Standard test methods for density and specific gravity (relative Jc: compression modulus (N/m)
density) of plastics by displacement. American Society for Testing and Mate- Js: secant modulus (N/m)
rials, West Conshohocken, Pennsylvania, USA. Kp: Rankine’s coefficient of passive earth pressure (dimensionless)
ASTM D 4595, 2011. Standard test method for tensile properties of geotextiles by L: length of sample at εa (m)
the wide-width strip method. American Society for Testing and Materials, West L0: initial length of sample (m)
Conshohocken, Pennsylvania, USA. Pj: junction peel strength (kN/m)
ASTM D 4767, 2011. Standard test method for consolidated undrained triaxial p0 : mean normal stress (kN/m)
compression test for cohesive soils. American Society for Testing and Materials, Sj: junction shear strength (kN/m)
West Conshohocken, Pennsylvania, USA. T: tensile strength of geocell material (kPa)
ASTM D 5199, 2012. Standard test method for measuring the nominal thickness of t: thickness of geotextile (m)
geosynthetics. American Society for Testing and Materials, West Conshohocken, Uj: junction split strength (kN/m)
Pennsylvania, USA. εa: axial strain (%)
ASTM D 5261, 2010. Standard test method for measuring mass per unit area of εc: circumferential strain (%)
geotextiles. American Society for Testing and Materials, West Conshohocken, εv: volumetric strain (%)
Pennsylvania, USA. f: friction angle of soil ( )
Bathurst, R.J., Crowe, R.E., 1994. Recent case histories of flexible geocell retaining gd,max: maximum dry unit weight (N/m3)
walls in North America. In: Tatsuoka, F., Leshchinsky, D. (Eds.), Recent Case gd,min: minimum dry unit weight (N/m3)
Histories of Permanent Geosynthetic-reinforced Soil Retaining Walls. Balkema, mA: areal density of geotextile (kg/m2)
Rotterdam, pp. 3e19. r: density of soil (kg/m3)
Bathurst, R.J., Karpurapu, R., 1993. Large scale triaxial compression testing of geocells- s: normal stress (N/m2)
reinforced granular soils. Geotechnical Testing Journal 16 (3), 296e303. Ds: deviatoric stress (N/m2)
Cancelli, A., Rimoldi, P., Montanelli, F., 1993. Index and performance tests for geo- sr: rubber correction or increment in axial stress (N/m2)
cells in different applications. In: Cheng, S.C.J. (Ed.), Geosynthetic Soil Testing s1: axial stress (N/m2)
Procedures, ASTM STP 1190, pp. 64e75. s3: confining pressure (kPa)
Chen, R.H., Chiu, Y.M., 2008. Model tests of geocell retaining structures. Geotextiles Ds3: rubber correction or increment in confining pressure (kPa)
and Geomembranes 26, 56e70. s: shear stress (N/m2)