Вы находитесь на странице: 1из 13

Indian Geotech J (OctoberDecember 2013) 43(4):331343

DOI 10.1007/s40098-013-0053-7

ORIGINAL PAPER

A Multiple Regression Equation for Prediction of Bearing


Capacity of Geosynthetic Reinforced Sand Beds
G. Madhavi Latha Amit Somwanshi
K. Hariprasad Reddy

Received: 18 January 2013 / Accepted: 18 March 2013 / Published online: 28 March 2013
 Indian Geotechnical Society 2013

Abstract Several researchers have studied the effects of


various important parameters like number of layers of
reinforcement (N), depth of the upper most layer of reinforcement from the base of the footing (u), spacing of the
reinforcement layers (h), width of reinforcement layer (b),
tensile strength of reinforcement (T) and density of soil (c)
on the ultimate bearing capacity of footings on reinforced
soils. However, very few of them have attempted to
develop equations to estimate the bearing capacity of the
footings on reinforced soils by considering all the above
parameters. Review of the existing state of the art in this
area is presented in this paper and the ability of the existing
analytical models to predict the ultimate bearing capacity
of reinforced sand beds is verified using results from laboratory model tests. A multiple regression equation to
estimate the ultimate bearing capacity of square footing
resting on geosynthetic reinforced sand beds based on the
laboratory model studies is presented. This equation is
verified against the experimental results of model studies
and also with the experimental measurements of other
researchers. Statistical analysis of the model test data
revealed that the aperture size of the reinforcement is very

G. M. Latha (&)
Department of Civil Engineering, Indian Institute of Science,
Bangalore 560012, India
e-mail: madhavi@civil.iisc.ernet.in
A. Somwanshi
Garware Wall Ropes Ltd., Pune, India
e-mail: amitbs84@gmail.com
K. H. Reddy
Wipro Limited, Bangalore, India
e-mail: hariprasad.reddy@wipro.com

influential in altering the bearing capacity of reinforced


sand beds. The regression equation developed in this study
is reasonably good in predicting the bearing capacity of
square footing on dense sand reinforced with planar
geosynthetics.
Keywords Multiple regression analysis  ANOVA 
Bearing capacity  Reinforced sand  Square footing 
Geosynthetics

Introduction
Design of shallow foundations requires calculation of
bearing capacity and settlements. Bearing capacity depends
on the strength of the soil and the settlements depend on the
compressibility of the soil. Reinforcing the soils with geosynthetics is beneficial because it has the advantages of both
decreasing the settlements and increasing the bearing
capacity. It is the pioneering work of Binquet and Lee [4]
that marked the beginning of systematic research in the field
of reinforced earth beds. Subsequently many researchers
have reported the beneficial effects of using soil reinforcement on the performance improvement of shallow foundations [1, 7].
Optimum quantity of reinforcement and the position,
form and configuration of the reinforcement play vital role
in finding cost-effective solutions for most of the geotechnical engineering problems involving reinforced soils.
A number of researchers have presented the theoretical
analyses for determining the bearing capacity of shallow
foundations on reinforced soil. Some of the important
studies in this area are: Binquet and Lee [5], Huang and
Tatsuoka [9], Huang and Menq [8], Wayne et al. [20],
Kumar and Saran [10], Michalowski [14], Kumar and

123

332

Indian Geotech J (OctoberDecember 2013) 43(4):331343

Typical test set-up and layout of the multi-layered


geosynthetic reinforced sand bed adopted in the model tests
is shown in Fig. 1. N layers of reinforcement, square in
shape and having side width b, are placed at specific depths
while preparing the sand bed for each model test. The
depth of placement of the top most reinforcement layer
from the bottom face of the footing is measured as u and
the vertical spacing between consecutive layers of reinforcement is measured as h. The total depth of reinforcement zone, dr, below the bottom face of the footing
becomes u ? (N - 1) h. All these parameters are expressed in non-dimensional form in terms of the footing width
as u/B, h/B, b/B and dr/B (reinforcement depth ratio). In all
the tests reported in this paper, u/B and h/B were kept
equal. The footing settlements and surface deformations
(heave/settlement) were measured at different locations.
One model test is conducted on square footing resting on
unreinforced sand bed. After that, series of model tests
were carried out on footings resting on geosynthetic reinforced sand beds by varying different parameters such as
type of geosynthetic material, depth of reinforcement (dr),
depth to the top most reinforcement layer from the base of
the footing (u), vertical spacing between consecutive layers
of reinforcement (h) and width of reinforcement layers (b).
Complete details of the model tests and the results are
presented in Latha and Somwanshi [12].

Walia [11] and Chen [6]. Very few researchers attempted


to develop regression models for reinforced sand beds [3].
Though many researchers have studied the load bearing
capacity of reinforced sand foundations and its substantial
increase with the geosynthetic reinforcement, the need for
more detailed studies concerning the design of soil reinforcement is increasingly apparent. Also, an empirical
equation for quick estimate of bearing capacity of reinforced earth beds considering the quantity, type, position
and spacing of reinforcement would be extremely helpful
for preliminary designs.
In the present study an attempt is made to develop a
regression equation based on laboratory model load test
data for estimating the bearing capacity of square footing
on geosynthetic reinforced sand beds. Verification of the
equation with experimental data of other researchers is also
presented.

Model Studies
Laboratory model plate load tests were conducted on sand
beds prepared in a steel test tank with inside dimensions
900 9 900 9 600 mm. The model footing was 25 mm
thick rigid steel plate and measured 150 9 150 mm, square
in shape. Poorly graded sand is used in the tests and the
average unit weight and relative density of the sand were
kept uniform as 15.6 kN/m3 and 70 % respectively in all
model tests. Four types of geogrids were used to reinforce
the sand beds: Weak biaxial geogrid (WG), Strong biaxial
geogrid (SG), Uniaxial geogrid (UG), and geonet (GN).
The tensile properties of these geosynthetics were determined from standard multi-rib tension test [2] and are listed
in Table 1. Since these tests are model tests, strength and
stiffness of the geogrid have to be scaled down to represent
the field conditions. The reason for using GN is to scale
down the tensile strength of the reinforcing layer for the
model tests as weaker geogrids are not fabricated. Weaker
biaxial geogrid has secant modulus almost one third of the
strong biaxial geogrid, representing low stiffness and GN
has ultimate tensile strength five times less than that of
stronger geogrid, representing low strength.

Prediction of Bearing Capacity Using Existing


Analytical Solutions
Chens Method
Chen [6] developed an analytical solution for determining
the ultimate bearing capacity of reinforced soil foundations. This solution is based on the stability analysis of
reinforced soil foundation assuming the failure in the
reinforced zone. If the strength of the reinforced zone is
slightly larger than that of the underlying unreinforced
zone or if the reinforcement depth ratio (d/B, where d is the
depth of reinforced zone and B is the width of the

Table 1 Properties of geosynthetics


Property

Type of geosynthetic material


WG

SG

UG

GN

Ultimate tensile strength (kN/m)

20

40

40

7.6

Failure strain (%)

18

17

13

2.4

Secant modulus at 5 % strain (kN/m)

160

450

434

317 (failed at 2.4 % strain)

Mass per unit area (g/m2)

220

530

380

139

Aperture size (mm)

35 9 35

30 9 30

220 9 17.2

1.5 9 1.5

123

Indian Geotech J (OctoberDecember 2013) 43(4):331343

333

Fig. 1 Test set-up and layout


and configuration of
reinforcement layers in the
model tests

foundation) is relatively large, failure occurs in the reinforced zone as shown in Fig. 2.
Chen [6] discussed two different reinforcing mechanisms: horizontal confinement effect and tension membrane effect. Only horizontal confinement effect is
discussed here as this type of reinforcing mechanism most
likely controls the performance of the reinforced sandy soil
used in the present study. To include the contribution of
reinforcement, the method of superposition can be used and
an additional item DqT is then added to classical bearing
capacity formula for strip footing on unreinforced soil. The
bearing capacity formula now takes the following form:
quR cNc qNq 0:5cBNc DqT

where qu(R) is the ultimate bearing capacity of reinforced


soil foundation; DqT is the increased bearing capacity due
to the tensile force of the reinforcement; c is the cohesion
of soil; q is the surcharge load; c is the weight of soil; B is
the width of footing; and Nc, Nq and Nc are bearing
capacity factors, which are dependent on the friction angle
of soil /.

Based on the stability analysis of reinforced soil foundations, an expression is derived to determine the increased
bearing capacity (DqT) due to the insertion of two or more
layers of reinforcement. For strip footings, the increased
bearing capacity DqT can be calculated as:
DqT

N
X
4Ti u i  1h
i1

B2

where Ti is the tensile force in the ith reinforcement layer;


N is the number of reinforcement layers; u is the top layer
spacing of reinforcement; h is the vertical spacing between
reinforcement layers. It should be noted that all reinforcement layers must be placed above the failure zone to
contribute to improving the performance of the soil
foundation.
For square footings, the increased bearing capacity DqT
can be calculated as:
n
X
12Ti u i  1hrT
;
3
DqT
B2
i1
where

Fig. 2 Failure in reinforced


zone [6]

123

334

Indian Geotech J (OctoberDecember 2013) 43(4):331343




u i  1h
p /
tan

rT 1  2
B
4 2


B
p /

 for u i  1h\ tan


2
4 2
1 u i  1h
rT 
2
2Hf



B
p /

 for u i  1h  tan
;
2
4 2

where Hf is the depth of failure surface and can be


evaluated as:
B
Hf
ecosp=4/=2 tan / cos /
2 cosp=4 /=2

The tensile force developed along the reinforcement can


be evaluated based on the measured strain along the
reinforcement.
In the absence of a rigorous solution for the vertical
settlement distribution at a certain depth, it may be
assumed that the shape of reinforcement at that certain
depth is of the form as shown in Fig. 3 for sand. The
reinforcement beneath the footing is assumed to move
downward uniformly (lines bc). The reinforcement located
outside the boundary (lines aa0 and dd0 ) is considered to
have negligible strain.
Based on the measured strain distribution along the reinforcement, Chen [6] proposed that the slope of the boundary
lines aa0 and dd0 can be taken as about 2:1 (vertical: horizontal), which is the same as the simplified 2:1 stress distribution lines. Since the distribution of vertical settlement is
now known, next step is to determine the amount of settlement
at a certain depth beneath the footing (Se).
Using this simplified strain influence factor distribution
diagram given by Schmertmann et al. [16], the elastic
settlement Se in sand can be calculated as:
Ie Dz
Se C1C2C3q  cDf R
Es
cDf
q  cDf
 t 
C2 1 0:2 log
0:1
C1 1  0:5

C3 1:03  0:03L=B  0:73


Fig. 3 Simplified distribution
of vertical settlement in sand [6]

123

7
8
9
10

where C1 is a correction factor for the depth of embedment;


C2 is a correction factor for secondary creep in sand; C3 is a
correction factor for the footing shape; Ie is the strain
influence factor; Dz is the incremental depth; Es is the
elastic modulus of sand; t is the time since the application
of load (year) (t C 0.1 year); L is the length of footing; B is
the width of the footing.
Based on the above assumptions and analysis, the
average strain in reinforcement at a certain footing settlement can be calculated as:
Lab Lbc Lcd  Lad
Lad
q
Lab Lcd S2e z=22

12

Lbc B

13

Lad B z

14

eavg

11

where Se is the settlement at a depth of z beneath the center of


footing; z is the depth of reinforcement = u ? (i - 1)h. The
average tensile force, Tavg, developed in reinforcement can
then be evaluated by the following equation:
Tavg Jeavg ;

15

where J is the tensile modulus of the reinforcement.


The measured strain by Chen [6] showed that the strain
distribution along the reinforcement is not uniform. The
tensile strain is the largest at the point beneath the center of
the footing and decreases with the distance away from the
center of footing. A triangular distribution is assumed to
approximately describe the real strain distribution along the
reinforcement. The maximum strain in this triangular distribution can be calculated as:
emax 2eavg

16

For the model footings tested in present study, the


ultimate bearing capacities are calculated using this
analytical model and the results are presented in Table 2.
Comparison of the measured and estimated bearing
capacities for the laboratory model tests and also the
percentage error for each case are presented in Table 2.
The average error in the prediction of Chens analytical
model for the model tests carried out in present work is
18 %. Example calculations for the model test where sand

Indian Geotech J (OctoberDecember 2013) 43(4):331343

335

bed was reinforced with five layers of GN represented by


S.No. 5 in Table 2 are presented below.
B = 0.15 m, Df = 0.0 m, c = 15.6 kN/m3, / = 44,
N = 5 layers of GN, u/B = 0.4, h/B = 0.4, qu = 270.3 kPa
(unreinforced, at s/B = 17.6 %), J = 317 kN/m (secant
modulus at 2.4 % of tensile strain)
Step 1 Calculating the settlement at all the layers of
reinforcement:
C1 1;

C2 1;

Calculating the tensile forces in all the layers of


reinforcement:

First layer:
Lab

Lbc B 150 mm
Lad B z 150 120 270 mm
Lab Lbc Lcd  Lad
Lad
61:43 150 61:43  270
1:06 %

270

eavg

C3 1

Calculation of settlement for all the layers of reinforcement


is shown in Table 3.
Step 2

q q
Lab Lcd S2e z=22 13:22 120=22
61:43 mm

emax 2eavg 2  1:06 2:12 %


e

u
tanp=4/=2
Bu
2

u2

emax 1:73 %

T2 Je 317  0:0173 5:47 kN/m


Third layer:

q q
Lab Lcd S2e z=22 5:92 180=22
90:19 mm

q q
Lcd S2e z=22 23:462 60=22
38:08 mm

Lbc B 150 mm

Lbc B 150 mm

Lad B z 150 60 210 mm

Lad B z 150 180 330 mm

Average strain:

eavg

Lab Lbc Lcd  Lad 38:08150 38:08  210

210
Lad
7:69 %

Lab Lbc Lcd  Lad


Lad
90:19 150 90:19  330
0:117 %

330

eavg

emax 2eavg 2  7:69 15:38 %

emax 2eavg 2  0:117 0:234 %


e

Strain at the triangle


soil wedge faces ac and bc as shown
u
u2
in Fig. 2. e tanp=4/=2
emax 8:05 % but maximum tensile
Bu
2
strain recorded in reinforcement material during tension test
is 2.4 %.

u
tanp=4/=2
Bu
2

u2

emax 0:234 %

T1 Je 317  0:024 7:6 kN=m

T3 = Je = 317 9 0.00234 = 0.74 kN/m


Strain developed in fourth layers is very low and no
strain is developed in fifth layer, so contribution of these
layers in negligible.
Step 3 Calculating the increased bearing capacity DqT:

Second layer:

For first two layers:

Table 2 Comparison of
ultimate bearing capacities of
reinforced sand beds observed
in model tests with the values
predicted using Chen [6]

S. no.

Reinforcement
type

B (m)

u/B = h/B

Ultimate bearing
capacity qu(R) (kPa)
(measured)

Ultimate bearing
capacity qu(R) (kPa)
(calculated)

Error (%)

WG

0.15

0.67

540

378

30.0

WG

0.15

0.67

441

378

14.3

UG

0.15

0.67

405

526

29.9

SG

0.15

0.67

585

509

13.0

GN

0.15

0.4

675

554

17.9

GN

0.15

0.5

630

479

24.0

7
8

GN
GN

0.15
0.15

0.67
1.0

3
2

495
405

451
300

8.88
25.9

GN

0.15

0.67

459

451

1.7

123

336

Indian Geotech J (OctoberDecember 2013) 43(4):331343

Table 3 Calculation of settlement for all the layers of reinforcement as per Chen [6]

First layer (at a depth of z1 = u)

Dz (mm)

Es (kPa)

z (mm)

Ie

IeDz/Es

60

3,000

90

1.9

0.038

60

3,000

150

1.35

0.027

60

3,000

210

0.82

0.0164

60

3,000

270

0.27

0.0054

60

3,000

330

0.0868

Se1 C1 C2 C3 q  cDf R IeEDz


111270:3  00:0868 23:46mm
s
Second layer (at a depth of z2 = u ? h)

60

3,000

150

1.35

0.027

60

3,000

210

0.82

0.0164

60

3,000

270

0.27

0.0054

60
R

3,000

330

0
0.0488

Se2 C1 C2 C3 q  cDf R IeEDz


111270:3  00:0488 13:2mm
s
Third layer (at a depth of z3 = u ? 2 h)

60

3,000

210

0.82

60

3,000

270

0.27

0.0054

60

3,000

330

0.0164

0.0218

Se3 C1 C2 C3 q  cDf R IeEDz


111270:3  00:0218 5:9mm
s
Fourth layer (at a depth of z4 = u ? 3 h)

60

3,000

270

0.27

60

3,000

330

0.0054
0
0.0218

Se4 C1 C2 C3 q  cDf R IeEDz


111270:3  00:0054 1:46mm
s
Fifth layer (at a depth of z5 = u ? 4 h)

DqT

Settlement at this layer is zero as no strain is developed at this depth

h

i
ui1h
n 12Tiu i  1h 1  2
tan p4  /2
X
B
B2

i1

280 kPa
For third layer:
Hf

B
ecosp=4/=2 tan / cos / 0:202 m
2 cosp=4 /=2

DqT

n 12Tiu i  1h
X
i1

Step 4

B2

1
2

i
 ui1h
2Hf

4 kPa

Calculating the ultimate bearing capacity of


reinforced sand:

quR qu DqT 270:3 280 4 554:3 kPa


Wayne et al.s Method
Wayne et al. [20] developed an analytical solution for
determining the ultimate bearing capacity of reinforced soil
foundation for four different failure modes; failure above
reinforcement, failure between reinforcements, failure

123

along reinforcement and failure through reinforcement.


Model studies in present study fall in the category of two
layer soil system of strong soil layer over weak soil layer.
If strength of the reinforced zone is much higher than that
of the underlying unreinforced zone and the reinforcement
depth ratio (d/B) is relatively small, a punching shear
failure occurs in the reinforced zone followed by a general
shear failure in the unreinforced zone as shown in Fig. 4.
The ultimate bearing capacity for square footings on a
reinforced soil foundation was given as:


4ca sa d
2Df Ks ss tan /t
2
quR qb
2ct d 1
B
d
B
n
P
4 Ti sT tan d
 ct d
17
i1
B
where qu(R) is the ultimate bearing capacity of reinforced
soil foundation; qb is the ultimate bearing capacity of the
underlying unreinforced soil; ca is the unit adhesion of
upper layer; d is the thickness of reinforced zone; D is the
inclination angle of the passive force with the horizontal;
B is the width of footing; ct is the unit weight of soil in
reinforced zone; /t is the friction angle of reinforced zone;
Ti is the tensile force in the ith layer of reinforcement; n is

Indian Geotech J (OctoberDecember 2013) 43(4):331343

337

Fig. 4 Assumed mode of


failure for footing on a two layer
soil system with horizontal
reinforcement [20]

the number of reinforcement layers; Df is the embedment


depth of the footing; Ks is the punching shear coefficient,
which depends on the ultimate bearing capacity of soil in
both the reinforced zone and the underlying unreinforced
zone. sa, ss and sT are shape factors for punching shear
resistance. A better estimate of the punching shear coefficient, Ks can be obtained from the design charts of
Mayerhof and Hanna [13]. Valsangkar and Meyerhof [19]
showed that for circular footing a shape factor may be
taken as unity for the punching failure of footings in both
sand and clay layers. For square footings, this approximation can still be applied for conservative design [6].
For the model footings tested in present work, the ultimate bearing capacities are calculated using the analytical
model of Wayne et al. [20] and the results are presented in
Table 4. Comparison of the measured and estimated
bearing capacities for the laboratory model tests and also
the percentage error for each case are presented in the
table. The average error in the prediction of Wayne et al.s
analytical model for the model tests carried out in present
work is 85 %. Example calculations for the model test
where sand bed was reinforced with 5 layers of GN represented by S.No. 5 in Table 4 are presented below:
B = 0.15 m, Df = 0.0 m, c = 15.6 kN/m3, / = 44,
N = 5 layers of GN, u/B = 0.4, h/B = 0.4, qu = 270.3 kPa
(unreinforced, at s/B = 17.6 %), J = 317 kN/m (secant
modulus at 2.4 % of tensile strain)
Step 1 The tension developed in the reinforcement at
different levels (from calculation steps described
for Chens method):
T1 7:6 kN/m;
Step 2

T2 5:47 kN/m;

T3 0:74 kN/m

Calculating the ultimate bearing capacity of the


underlying unreinforced sand:

Nc 258:08 for friction angle / 44


qb 0:4cBNc 241:6 kPa
Step 3

Calculating the ultimate bearing capacity of the


reinforced sandy soil:

d 44


4ca sa d
2Df Ks ss tan /t
2ct d2 1
quR qb
B
d
B
n
P
4 Ti sT tan d
 ct d 955 kPa
i1
B

Ks 20;

ca 0;

Huang and Menqs Method


Huang and Menq [8] developed an analytical solution for
determining the ultimate bearing capacity of reinforced soil
foundation using the deep-footing and wide-slab
failure mechanisms. The basic concept of this failure
mechanism is that the bearing capacity of footing (width:
B) on reinforced soil foundation is equivalent to that of a
wider footing (width: B ? DB) at the depth of d (total
depth of reinforcement) on unreinforced soil foundation as
shown in Fig. 5.
Huang and Menq [8] provided a relationship to determine the ultimate bearing capacity of a surface footing on
reinforced soil foundation as follows:
quR g  c  B DB  Nc c  d  Nq

18

where qu(R) is the ultimate bearing capacity of reinforced


soil foundation; g is an coefficient depending on footing
shape; c is the dry unit weight of soil; B is the width of
footing; Nc, Nq are bearing capacity factors; d is the total
depth of reinforcement; DB is the increase of footing width
due to the inclusion of reinforcement, which is equal to
(2 9 d) tan a; a is the stress distribution angle due to the
wide-slab effect as shown in Fig. 5 and can be calculated
using the following equation:
tana 0:680  2:071h=b 0:743CR 0:030l=B
0:076N;

19

where h is the vertical spacing between reinforcement


layers; CR is the covering ratio of reinforcement (area of
reinforcement divided by the area of soil covered by
reinforcement); l is the length of reinforcement; N is the
total number of reinforcement layers.

123

338
Table 4 Comparison of
ultimate bearing capacities
of reinforced sand beds
observed in model tests with the
values predicted using
Wayne et al. [20]

Indian Geotech J (OctoberDecember 2013) 43(4):331343

Ultimate bearing
capacity qu(R) (kPa)
(calculated)

Error (%)

S.no.

Reinforcement
Type

B (m)

u/B = h/B

Ultimate bearing
capacity qu(R) (kPa)
(measured)

WG

0.15

0.67

540

998

WG

0.15

0.67

441

718

62.8

UG

0.15

0.67

405

1,163

187.2

SG

0.15

0.67

585

1,144

95.6

GN

0.15

0.4

675

955

41.5

GN

0.15

0.5

630

855

35.7

7
8

GN
GN

0.15
0.15

0.67
1.0

3
2

495
405

802
655

62.0
61.7

GN

0.15

0.67

459

1,081

135.5

84.8

Fig. 5 Failure mechanism of


reinforced soil foundation (after
Huang and Menq [8])

For the model footings tested in present work, the


ultimate bearing capacities are calculated using the analytical model of Huang and Menq [8] and the results are
presented in Table 5. Comparison of the measured and
estimated bearing capacities for the laboratory model tests
and also the percentage error for each case are presented
in the table. The average error in the prediction of Huang
and Menqs analytical model for the model tests carried
out in present work is 60 %. Example calculations for the
model test where sand bed was reinforced with 5 layers of
GN represented by S.No. 5 in Table 5 are presented
below:
B = 0.15 m, Df = 0.0 m, c = 15.6 kN/m3, / = 44,
N = 5 layers of GN, u/B = 0.4, h/B = 0.4, l/B = 5.93, d =
Table 5 Comparison of
ultimate bearing capacities of
reinforced sand beds observed
in model tests with the values
predicted using Huang and
Menq [8]

123

0.3 m, w = 0.000275 m, W = 0.0015 m, CR = 0.1833,


qu = 270.3 kPa (unreinforced, at s/B = 17.6 %)
CR is the cover ratio = w/W; w the width of longitudinal ribs; W the center-to-center spacing of the longitudinal ribs.
Step 1 Calculating the predicted value of tan a:
tana 0:680  2:071h=b 0:743CR 0:030l=B 0:166
Step 2

Calculating the increase of footing width due to


the inclusion of reinforcement:

DB 2  d tan a 0:1 m
Step 3

Calculating the ultimate bearing capacity of the


reinforced sand:

S.No. Geogrid Type B (m) u/B = h/B N b/B

Ultimate bearing
Error (%)
Ultimate bearing
capacity qu(R) (kPa) capacity qu(R) (kPa)
(measured)
(calculated)

WG

0.15

0.67

4 5.93 540

744

37.8

2
3

WG
UG

0.15
0.15

0.67
0.67

3 5.93 441
4 5.93 405

502
959

13.8
136.8

SG

0.15

0.67

4 5.93 585

759

29.7

GN

0.15

0.4

5 5.93 675

1127

67.0

GN

0.15

0.5

4 5.93 630

926

47.0

GN

0.15

0.67

3 5.93 495

586

18.4

GN

0.15

1.0

2 5.93 405

-74

118.2

GN

0.15

0.67

4 5.93 459

828

80.4

10

GN

0.15

0.5

4 5.0

586

900

53.6

11

GN

0.15

0.5

4 4.0

540

871

61.3

Indian Geotech J (OctoberDecember 2013) 43(4):331343

339

g 0:4; Nc 258:1; Nq 154:9


quR g  c  B DB  Nc c  d  Nq 1,127 kPa
Multiple Regression Analysis
Multiple linear regression analysis of the results from
model tests is carried out considering all the influential
parameters as observed in experiments and an equation is
proposed to estimate the bearing capacity of the footings on
reinforced sand. The statistical analysis is conducted using
the software Minitab 15 [15]. Details of the statistical
techniques used in the analysis for the multiple regression
are explained in following sections.

the critical values of the theoretical distribution (dependent


on which F is used) to determine the p value. Anderson
Darling test gives us the p value which is the actual
probability of rejecting null hypothesis when it is true. By
comparing this p value with significance level (a), we take
decisions whether to confirm data follows normal
distribution or not. Typically a value is considered as
0.05, which means that up to 5 % error is acceptable in
taking such decisions. Any p value greater than a value
infers that there is not enough evidence to reject the null
hypothesis and hence we go with the null hypothesis
assumption.
Signification Test for the Overall Model

Regression Model
Regression analysis is a statistical technique used to
examine the relationship between a dependent variable and
a set of independent variables. In multiple linear regression
analysis, it is hypothesized that this relationship is linear
and has the following form:
yi b0 b1 xi1 b2 xi2    bk xik ei ;

20

where i = 1, 2,,n and n is the number of observations; yi


is dependent variable; xi1, xi2,, xik are independent variables; b0, b1,,bk are unknown regression coefficients;
ei is random error. It should be noted that this model is
called linear because of its linearity in regression coefficients, not in the independent variables.
Normality Check
Before carrying out multiple linear regression analysis, the
data should be checked for normality as regression expects
data should follow normal distribution. Normality Check
can be done using AndersonDarling test [18]. In this
check, data is plotted in the normal probability plot and if
all the data points form a straight line then it infers that the
data follows normal distribution.
The AndersonDarling test determines whether data
comes from a specific distribution. The formula for the test
statistic A to assess if data {Y1 \ \ YN} (this data
must be ordered) comes from a distribution with cumulative distribution function F is
A2 N  S,
where
S

N
X

2k  1=N ln FYk ln1  FYN1k 

21

k1

The critical values for the AndersonDarling goodnessof-fit test are dependent on the specific distribution that is
being tested. The test statistic can then be compared against

Significance test for the overall model is a test to determine


the effectiveness of the entire model, i.e. whether the linear
relationship exists between the dependent variable and
independent variables. This is generally done by testing the
null hypothesis: H0: b1 = b2 = = bk = 0 against the
alternative hypothesis H1: at least one of the bj is non-zero.
The null hypothesis implies that none of the independent
variables are linearly related to the dependent variable in
the assumed multiple regression equation. The alternative
hypothesis suggests at least one of the independent variables is linearly related to the dependent variable. This
hypothesis can be tested by comparing Mean Square
Regression (MSR) and Mean Square Error (MSE). This test
is an F statistic. The best way for this test is to use Analysis
of Variance (ANOVA). ANOVA table is generally used for
the ANOVA calculations and it has the general form as
shown in Table 6.
The terms displayed in Table 6 are defined and computed as follows:
n
X
SST
yi  y2 total sum of squares
i1

SSE

n
X

yi  ^yi 2 sum of squares due to error

i1

SSR

n
X

^yi  y2 sum of squares due to regression

i1

SSE
mean square due to error
nk1
SSR
mean square due to regression
MSR
k

MSE

where ^yi are the predicted values, y is the mean of


dependent variables.
The three sums of squares are related by the formula:
SST SSR SSE

22

Rejecting null hypothesis (H0) if p value \ a-value;


failing to reject null hypothesis, if p value C a-value.

123

340

Indian Geotech J (OctoberDecember 2013) 43(4):331343

Table 6 General form of ANOVA table for multiple linear regression


Degree of
freedom
Regression k

Sum of
squares

Mean
square

SSR

MSR

MSR/
MSE

MSE

Error

n-k-1

SSE

Total

n-1

SST

where SEb^j is the standard error of the regression coeffi^ , and cjj is diagonal element of (X0 X)-1 correcientb
j
^.
sponding tob
j

Check the p values against the individual factors.


Rejecting null hypothesis (H0) if p value \ a-value; failing
to reject null hypothesis, if p value C a-value. a is the
significance level.
Variables Included in the Analysis and Selection
Technique

Goodness of Fit of the Model


The quality of the fit can be measured by the sum of the
squares of the residuals (ei). The residual is defined as:
ei yi  ^yi

23

A good fit should have small residuals. However, this


quantity is dependent on the units of yi. Thus the coefficient
of multiple determination, R2, is generally used to measure
the goodness of fit.
Pn
SSR
SSE
yi  ^yi 2
1
1  Pi1
R2
24
n
SST
SST
y2
i1 yi  
R2 ranges from 0 to 1. The closer it is to 1, the better the fit.
R2 equal to 1 means perfect linear relationship exists
between the dependent variable and independent variables,
while R2 equal to 0 indicates independent variables have no
impact on the dependent variable.
MSE is the estimate of standard error, i.e. s2 = MSE.
The smaller the values of s2 are, the better the fit. This
measurement provides an excellent indication of the
quality of the fit when the prediction is important for the
model. In most cases, both R2 and s2 needs to be considered
to assess the goodness of fit.
Significance Tests for Regression Coefficients
If null hypothesis (H0) in significance test for the entire
model is rejected, it only indicates at least one of the bj is
non-zero. Then the additional tests are needed to determine
which of these bjs are not equal to zero. Significance tests
for individual regression coefficients would be useful for
this determination. This is generally done by testing the null
hypothesis: H0: bj = 0 against the alternative hypothesis
H1: bj = 0. If null hypothesis (H0) is not rejected, it indicates the independent variable xj can be removed from the
regression model. This test is a t statistic test and can be
written as
t

^
^
b
b
j
j
p ;
SEb^j
cjj MSE

123

25

Variables used in the regression analysis of the data from


model tests are selected as top layer spacing ratio or
normalized vertical spacing between layers (u/B = h/B),
normalized length of reinforcement layer (l/B), Number of
reinforcing layers (N), tensile modulus of reinforcement
(J), ratio of aperture area to footing area (Aa/Af) and settlement ratio (s/B). To make all the variables dimensionless, the tensile modulus (J) is normalized to 100 kN/m and
J
then used in the model, i.e. 100 kN/m
is used as a variable in
the model instead of J.
Regression Analysis of Model Tests
Based on the experimental results a multiple regression statistical analysis was conducted to develop a bearing capacity
ratio (BCR) model that can facilitate the estimation of bearing
capacity of a square footing resting on geosynthetic reinforced
sand bed from the bearing capacity of the same footing on
unreinforced sand bed. In developing the BCR model, all the
geosynthetic layers were assumed to have enough length to
fully mobilize its tensile contribution. Also, the effect of relative density of sand bed is not studied in the model and hence
the density of sand bed is not a variable in the model. Only
linear curve fitting is performed in this study. The significance
level a is set to 0.05 in this study.
A linear regression model used to investigate the effects
of all variables is given in the form of following equation:
 
 
s
b
h
BCR b1 b2
b3
b4
b5 N
B
B
B


 
J
Aa
b6
b7
26
100 kN/m
Af
Normal probability plot from AndersonDarling test is
shown in Fig. 6. In this case, p value is 0.454 which is
greater than the significance level (a) 0.05 and this assures
that the data is following normal distribution.
The multiple regression analysis is then conducted on
this model and the results yielded the following equation:

Indian Geotech J (OctoberDecember 2013) 43(4):331343

341

 
 
b
h
0:135
 0:628
B
B
B


 
J
Aa
0:0794N 0:0855
 3:91
100kN/m
Af

BCR 0:691 4:62

s

27
To determine the effectiveness of the entire model,
significance test for the overall model is performed.
Table 7 presents the results of ANOVA calculations.
SSR
70:9 %
SST
p
s MSE 0:237063
R2 value explains how good the regression model is.
71 % of the linear variation in BCR is being explained by
the factors that are considered in the analysis. The data
used for regression analysis is obtained from laboratory
model tests, in which 100 % repeatability cannot be
guaranteed. The materials involved, methods of sample
preparation and manual testing would impose certain limitations on the results and hence R2 value[70 is reasonable
for this case.
The null hypothesis is rejected because p value is\0.05.
This suggests that at least one of the independent variables
is linearly related to the dependent variable. To determine
the independent variables with linear relation, significance
tests for individual regression coefficients are then conducted. The results of these significance tests are summarized in Table 8. The null hypothesis is rejected because
p value is \0.05 for all the factors. This means that all
R2

independent variables are linearly related to the dependent


variable. Significance tests give the variance inflation factorVIF. If the VIF value is greater than 5 then there is
multi-colinearity between the variables. In this case VIF is
\5 for all the regression coefficients and hence the Multicolinearity is absent and we can proceed with the regression model (Minitab Inc. [15]).
As observed from Table 8, the most influential factor in
the estimation of bearing capacity after the settlement level
is the normalized aperture area (Aa/Af). To the authors
knowledge, none of the models reported so far consider the
effect of aperture size of geosynthetic material in the calculations. Present study clearly indicates the significance of
this parameter in the design.
Verification of Regression Model
To verify the regression model, the model is checked
against the experimental results of Chen [6]. A series of
laboratory model tests for square footings on reinforced
sand were reported by Chen [6]. The bearing capacity
values for these tests are predicted by the regression model
developed in the present study and the results are compared
against the measured values as shown in Table 9. The
predicted values at lower settlement levels for all the tests
are relatively higher compared to measured data. However
there is very good agreement between the predicted values
and measured data at higher settlement levels for tests in
which u/B or h/B is \0.5. For higher values of u/B or h/B,
the regression equation predicted higher BCR values.

Fig. 6 Normal probability plot


of BCR

123

342

Indian Geotech J (OctoberDecember 2013) 43(4):331343

Table 7 ANOVA table for nonlinear model for sand bed


Degree of
freedom
Regression

Sum of
squares

Mean
square

7.8200

1.3033

23.19 0.000

0.0562

Error

57

3.2033

Total

63

11.0233

Table 8 Significance test for individual regression coefficients


Variable

Regression coefficient

SE

p-value

VIF

u/B

-0.6275

0.1834

0.001

2.068

0.07939

0.03472

0.026

2.172

b/B

0.13498

0.03369

0.000

3.609

J/100 kPa

0.08854

0.03697

0.020

3.034

1.060

0.000

2.786

0.5656

0.000

1.023

Aa/Af
s/B

-3.914
4.6157

Comparison of Regression Model with Existing


Analytical Models
Ultimate bearing capacity of reinforced footing predicted
using Huang and Menq [8], Wayne et al. [20] and Chen [6]
are compared with laboratory model test results along with
the predictions using the regression equation developed in
this study in Fig. 7 for the case of square footings resting on
GN reinforced sand beds. Bearing capacity improvement
factor (If) in the plot is defined as the ratio of bearing
pressure of the reinforced soil at a given settlement and the
bearing pressure of unreinforced soil at the same settlement
value. As observed from the plot, predictions using Chen [6]
and using the regression model developed in the present
study are in good agreement with laboratory measurements.

All other methods overestimated the performance of reinforced sand beds.


According to Chen [6] and Sharma et al. [17], if the
strength of the reinforced zone is slightly larger than that of
the underlying unreinforced zone or if the reinforcement
depth ratio (d/B) is relatively large, the failure occurs in the
reinforced zone. In the present laboratory model testing
program, all the tests are conducted at d/B C 2.0, which is
relatively high. Also, due to the flowability of sand, failure
of reinforced sand most likely occurs in the reinforced zone
with the orientation of reinforcement close to the horizontal
direction and hence the assumptions are justified.
The deviations observed in case of Huang and Menq [8]
and Wayne et al. [20] can be explained with deep footing effect. The deep footing effect suggests that the
performance of reinforced soil is very similar to that of
unreinforced soil with a footing having an additional
embedment depth equal to the depth of the reinforced zone.
The additional bearing capacity contributed by the forces
acting on the lateral surfaces of equivalent deep footing
(deep footing effect) can be evaluated by the wideslab effect, which means the equivalent deep footing
having a base wider than the original footing. Deep
footing effect is explicitly or implicitly implied in the
design methods of Huang and Menq [8] and Wayne et al.
[20]. This effect results in an almost linear increase of the
BCR with increasing the number of reinforcement layers or
total depth of the reinforcement because of relatively high
friction angle of sand and hence no convergence can be
obtained (Sharma et al., [17]). However, the proposed
regression model and the analytical model of Chen [6]
overcome this shortcoming and stabilize as the number of
reinforcement layers increases.

Table 9 Verification of regression model for experimental data from Chen [6]
Test no.

SGG11-1

Reinforcement configuration

N = 1, BasXgrid11

u/B = h/B

b/B

J/100 kN/m

Aa/Af

@ s/B = 3 %

@ s/B = 12 %

BCR (exp)

BCR (exp)

BCR (reg)

BCR (reg)

0.167

10

2.68

0.0357

1.30

2.37

2.67

2.79

SGG11-2

0.33

10

2.68

0.0357

1.33

2.27

2.7

2.68

SGG11-3

0.5

10

2.68

0.0357

1.16

2.16

1.06

2.58

SGG11-4

0.667

10

2.68

0.0357

1.1

2.05

0.94

2.47

SGG11-5

1.0

10

2.68

0.0357

1.04

1.84

0.86

2.26

SGG31-1

N = 1,

0.2

10

0.8

0.0238

1.36

2.11

2.51

2.53

SGG31-2

2 9 Miragrid 8XT

0.3

10

0.8

0.0238

1.39

2.05

2.49

2.46

SGG31-3

0.4

10

0.8

0.0238

1.32

1.99

SGG31-4

0.5

10

0.8

0.0238

1.20

1.92

1.89

2.34

SGG12

BasXgrid11

SGG13
SGG14
exp experimental, reg regression

123

0.33

10

2.68

0.0357

1.5

2.35

2.76

2.76

3
4

0.33
0.33

10
10

2.68
2.68

0.0357
0.0357

1.57
1.63

2.43
2.51

2.79
2.95

2.84
2.92

Indian Geotech J (OctoberDecember 2013) 43(4):331343

343

References

Fig. 7 Comparison of predictions of regression model with analytical


models and experimental results

The regression model developed in the present study


could be used for the preliminary design calculations of the
footings on dense sand beds reinforced with geosynthetics.
However, several simplifying assumptions are made in the
development of model. Hence this model should be verified
using field tests on full-scale reinforced soil foundations
before applying it to actual design.

Summary and Conclusions


A simple equation based on multiple regression of the data
from laboratory model load tests is proposed to estimate
the bearing capacity of square footings on dense sand
reinforced with geosynthetics. Based on the statistical
analysis on experimental test results of square footing on
reinforced soil bed, it is observed that the aperture size of
the reinforcement is influential in controlling the BCR and
cannot be ignored in the designs. The regression equation is
verified against the experimental measurements of earlier
researchers and is found to be predicting the ultimate
bearing capacity of square footings on reinforced beds
reasonable well. The regression equation developed in the
present study could be used for the preliminary design
calculations of the footings on dense sand beds reinforced
with geosynthetics.

1. Akinmusuru JO, Akinbolade JA (1981) Stability of loaded footings on reinforced soil. J Geotech Eng Div ASCE
107(6):819827
2. ASTM Standard D 6637 (2001) Standard test method for determining tensile properties of geogrids by the single or multi-rib
tensile method. American Society for Testing and Materials,
Pennsylvania
3. Ber AK, Ghosh A, Ghosh A (2005) Regression model for bearing
capacity of a square footing on reinforced pond ash. Geotext
Geomembr 23(3):261285
4. Binquet J, Lee KL (1975) Bearing capacity tests on reinforced
earth slabs. J Geotech Eng Div ASCE 101(12):12411255
5. Binquet J, Lee KL (1975) Bearing capacity analysis on reinforced
earth slabs. J Geotech Eng Div ASCE 101(12):12571276
6. Chen Q (2008) An experimental study on characteristics and
behaviour of reinforced soil foundation. PhD thesis, Louisiana
State University, Louisiana
7. Fragaszy RJ, Lawton E (1984) Bearing capacity of reinforced
sand subgrades. J Geotech Eng Div ASCE 110(10):15001507
8. Huang CC, Menq FY (1997) Deep-footing and wide-slab effects
in reinforced sandy ground. J Geotech Geoenviron Eng ASCE
123(1):3036
9. Huang CC, Tatsuoka F (1990) Bearing capacity of reinforced
horizontal sandy ground. Geotext Geomembr 9(1):5182
10. Kumar A, Saran S (2003) Closely spaced footings on geogrid
reinforced sand. J Geotech Geoenviron Eng ASCE
129(7):660664
11. Kumar A, Walia BS (2006) Bearing capacity of square footings
on reinforced layered soil. J Geotech Geol Eng 24:10011008
12. Latha GM, Somwanshi A (2009) Bearing capacity of square
footings on geosynthetic reinforced sand. Geotext Geomembr
27:281294
13. Meyerhof GG, Hanna AM (1978) Ultimate bearing capacity of
foundations on layered soils under inclined load. Can Geotech J
15(4):565572
14. Michalowski RL (2004) Limit loads on reinforced foundation
soils. J Geotech Geoenviron Eng ASCE 130:381390
15. Minitab Inc (2007) Meet Minitab 15. Reference manual for
Minitab 15. State College PA
16. Schmertmann JH, Hartman JP, Brown PR (1978) Improved strain
influence factor diagrams. J Geotech Eng Div ASCE
104(8):11311135
17. Sharma R, Chen Q, Abu-Farsakh M, Yoon S (2009) Analytical
modeling of geogrid reinforced soil foundation. Geotext Geomembr 27:6372
18. Stephens MA (1974) EDF statistics for goodness of fit and some
comparisons. J Am Stat Assoc 69:730737
19. Valsangkar AJ, Meyerhof GG (1979) Experimental study of
punching coefficients and shape factor for two-layered soils. Can
Geotech J 16(4):802805
20. Wayne MH, Han J, Akins K (1998) The design of geosynthetic
reinforced foundations. In: Proceedings of ASCEs 1998 annual
convention & exposition ASCE GSP 76, pp 118

123

Вам также может понравиться