ARTICLE IN PRESS

Geotextiles and Geomembranes 23 (2005) 261–285

www.elsevier.com/locate/geotexmem

Technical Note

Regression model for bearing capacity of a square

footing on reinforced pond ash
Ashis Kumar Bera, Ambarish Ghosh, Amalendu Ghosh
Bengal Engineering College (a Deemed University), Howrah 711 103, India
Received 11 May 2004; received in revised form 20 August 2004; accepted 18 September 2004
Available online 23 November 2004

Abstract

This paper presents the regression analysis of the bearing capacity of a square footing on
reinforced pond ash. A power model has been developed to estimate bearing capacity of a
square footing ðqrs Þ on reinforced pond ash at any settlement using all possible regression
techniques based on 2088 model test data to select the significant subset of the predictors.
From the regression analysis, the predictors viz., the bearing capacity of a square footing ðqs Þ
at the same settlement on unreinforced pond ash, settlement to width of the footing ratio ðs=BÞ
in percentage, number of layers (N) of reinforcement, friction ratio (f), i.e. the ratio of the
pond ash—geotextile interface friction angle (c) to the direct shear friction angle of pond ash
ðfd Þ; depth of the upper most layer of reinforcement from the base of the footing to width of
the footing ratio ðu=BÞ; length of reinforcement sheet to width of footing ratio ðLs =BÞ and
vertical spacing of the reinforcement to width of the footing ratio ðSv =BÞ are found to be
significant to the model. The adjusted coefficient of determination ðR2adj Þ for the proposed
model is found to be 0.9448 and for the regression model 72% data out of 2088 observed data
have prediction with less than 15% error. The additional set of 99 data was used for validation
of the model.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Bearing capacity; Square footing; Regression; Power model; Pond ash; Geotextiles

Corresponding author.
E-mail addresses: ashis@civil.becs.ac.in (A.K. Bera), ghoshambarish@hotmail.com (A. Ghosh),
amalendughosh@hotmail.com (A. Ghosh).

0266-1144/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.geotexmem.2004.09.002
 ARTICLE IN PRESS
262 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

Nomenclature

B width of footing
Dlr depth of last reinforcement layer below base of the footing at which qrs
attains maximum value
Ep percentage of error
f friction ratio ðc=fd Þ
Ls sheet length of the reinforcement
N number of layers of reinforcement
p number of independent variables
qs bearing capacity of square footing on unreinforced pond ash at any
settlement
qrs bearing capacity of square footing on reinforced pond ash at any
settlement
q^ rs predicted value of the dependant variable qrs
qru ultimate bearing capacity of square footing on reinforced pond ash
s settlement of the footing
Sh horizontal spacing between two consecutive reinforcement strips
Sv vertical spacing between two consecutive layers of reinforcement
u depth of first layer of reinforcement below footing
x1 ; x2 ; . . . ; xp are independent variables
y is dependent variable
x1 ; x2 ; . . . ; xp are unknown parameters
x is a matrix ðn  pÞ made up of observations on each of p independent
variables
y is n  1 vector of observations
x is p  1 vector of unknown parameters
x0 is transpose of x
x1 is inverse of x
fd direct shear friction angle of pond ash
c pond ash geotextile interface friction angle

1. Introduction

The solution of bearing capacity problems are getting significantly more important
with the advent of research to improve the foundation soil conditions by introducing
new materials into the soil in various forms. Although the concept of reinforced soil
exists since biblical times and the first commercial use of reinforced earth was
developed by Vidal (1966), more recently this concept is being widely used to
improve the bearing capacity of footing on soil. Significant published valuable
literature documents model testing of the bearing capacity of shallow foundation on
reinforced soil.
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 263

Bearing capacity of small scale model footings on reinforced sand has been
studied by Binquet and Lee (1975a), Akinmusuru and Akinbolade (1981), Fragaszy
and Lawton (1984), Guido et al. (1985), Ramanatha Ayyar et al. (1988), Mahmoud
and Abdrabbo (1989), Omar et al. (1994), Abedin et al. (1997), Yamamoto
and Kusuda (2001), Dash et al. (2001), Shin et al. (2002), and Boushehrian and
Hataf (2003).
Adams and Collin (1997) investigated the potential benefits of geosynthetic
reinforced foundations using large scale model footing load tests on geosynthetic
reinforced sand.
Small scale model footing tests on reinforced clay were studied by Samtani and
Sonpal (1989), and Mandal and Sah (1992).
Raza et al. (1999) studied the shape effect of surface footings in terms of load
carrying capacity and settlement on treated and reinforced fly ash through model
tests.
Fly ash is a solid waste commonly produced from the combustion of coal at
thermal power plants. Generally there are two types of ash: fly ash and bottom ash.
When these two ash materials are mixed together and transported into the ash pond
they are commonly called pond ash. Every year huge amounts of pond ash are
produced by thermal power plants throughout the world, which creates problems of
disposal, as it is an environmental hazardous. To partially solve the problem,
reinforced pond ash can be used as a foundation medium. However, there still is a
paucity of data regarding the bearing capacity of shallow foundations on reinforced
pond ash.
A number of researchers have presented the theoretical analyses for determining
the bearing capacity of shallow foundations on reinforced soil. Binquet and Lee
(1975b) discussed the analysis of the bearing capacity of a strip footing on a granular
soil containing horizontal layers of reinforcements. Dixit and Mandal (1993) studied
the bearing capacity of geosynthetic reinforced soil using the variational method,
while Otani et al. (1998) carried out bearing capacity analysis of strip footing on
geosynthetic reinforced cohesive soil by using a rigid plastic finite element method.
Yamamoto and Kusuda (2001) employed approximate solutions for bearing
capacity of reinforced soil using the upper-bound theorem of limit analysis.
Boushehrian and Hataf (2003) investigated the bearing capacity of circular and ring
footings on reinforced sand by finite element method. Michalowski (2004) suggested
a method for estimating the bearing capacity of strip footings over foundation soils
reinforced with horizontal layers of geosynthietics.
A review of the literature indicates that the laboratory model tests and also
theoretical analyses for bearing capacity of strip, rectangular, circular and square
footings on different types of soils viz., sand, clay and fly ash reinforced with
different types of reinforcing materials—metallic, plastic, geosynthetic, jute-cloth
and other natural fibres were carried out by the above researchers. The effects of
various vital parameters viz., number of layers (N) of reinforcement, depth of the
upper most layer of reinforcement from the base of the footing (u), horizontal
spacing of the reinforcement (Sh), vertical spacing of the reinforcement (Sv), friction
ratio (f), i.e. the ratio of the pond ash–geotextile interface friction angle ðcÞ to the
 ARTICLE IN PRESS
264 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

direct shear friction angle of pond ash ðfd Þ; length of reinforcement sheet (Ls),
bending stiffness, tensile strength of reinforcement and density of soil on the ultimate
bearing capacity ðqru Þ of footing on reinforced soil as well as the bearing capacity
ðqrs Þ of footing on reinforced soil with respect to settlement (s) have been studied by
the above authors. To the present development of empirical formula to estimate the
qrs by considering all the above parameters has not received much attention. In the
present study an attempt has been made to develop a regression model on the basis
of the experimental data, for estimating the bearing capacity ðqrs Þ of square footing
on reinforced pond ash corresponding to any s=B ratio in terms of qs (the bearing
capacity of square footing on unreinforced pond ash corresponding to the same s=B
ratio), s=B; N, u=B; Ls =B; f and S v =B for the purpose of using this model in design
analyses.

2. Method of regression analysis

Empirical estimation methods are common in all specialties within Civil

Engineering discipline. Generally, a function is assumed to estimate a relation-
ship between the dependent variable and independent variables. The coefficients
of the function are estimated by least squares—regression analysis using the
observed data.
A general linear model is assumed in the following form:
y ¼ x1 x1 þ x2 x2 þ    þ xp xp : (1)
Where, y is dependent variable, x1 ; x2 ; . . . ; xp are independent variables, x1 ; x2 ; . . . ; xp
are unknown parameters and p is the number of independent variables. This model is
linear in the parameters, xj ; j ¼ 1 to p
In practice, n observations would be available on y with the corresponding n
observations on each of the p independent variables. Thus n numbers of equations,
in the form of Eq. (1), can be written, one for each observation. Essentially, n
equations will be solved for the p unknown parameters. Thus n must be equal to or
greater than p ðnXpÞ: The n equations are:
y1 ¼ x1 x11 þ x2 x12 þ    þ xp x1p ;
y2 ¼ x1 x21 þ x2 x22 þ    þ xp x2p ;
..
.
yn ¼ x1 xn1 þ x2 xn2 þ    þ xp xnp ; ð2Þ
where, yi is the ith observation on y and xij is the ith observation on the jth
independent variable. Eq. (2) can be written as
X
p
yi ¼ xj xij ; for i ¼ 1 to n: (3)
j¼1
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 265

The equations are presented in the matrix form as below:

8 9 8 98 9
>
> y1 >
> >
> x11 x12 x13    x1p > > >
> x1 > >
>
> >
> >
> >
> >
> >
>
>
> y >
> >
> x21 x 22 x 23    x 2p >
> >
> x >
>
> 2>
< = < > >
=< >
> 2
=
y3 ¼ x31 x32 x33    x3p x3 ;
>
> > > .. > > .. >
> .. >
> > >
> > ..
> .. .. >
>
>
>
>
>
>
>
>
> . >
> >
> . . .    . >
> >
> . >>
>
: >
> ; > : >
;>: > ;
yn xn1 xn2 xn3    xnp xp

i:e: y ¼ x x : (4)
np p1
n1

Where, y is n  1 vector of observations, x is a matrix ðn  pÞ made up of

observations on each of p independent variables, and x is p  1 vector of unknown
parameters.
For Eq. (4) to have an intercept term, it is necessary that xi1 ¼ 1 for i ¼ 1 to n;
while x1 becomes the intercept. In the following development, it is assumed that
xi1 ¼ 1 for i ¼ 1 to n: Now the assumed model can be written in the form of:

yi ¼ x1 þ x2 xi2 þ    þ xp xip ; for i ¼ 1 to n: (5)

The above Eq. (5) is a special case of Eq. (4).

The values of the unknown Pparameters x; can be estimated by minimizing the sum
of the squares of the errors, e2i ; Where, ei ¼ ðyi  y^ i Þ; y^ i is the predicted value of yi :
X
0

S¼ e2i ¼ y  x x^ y  x x^ ;

the symbol ‘‘^’’ indicating the predictor.

Differentiating this expression with respect to x^ and equating the partial derivative
to zero, the following expressions are obtained.


0 ¼ 2 x0 y  x x^

or

x0 y ¼ x0 x x^ (6)

Which represents the normal equations.

Now,

ðx0 x Þ1 x0 y ¼ ðx0 x Þ1 ðx0 xÞ x^ ¼ x^

or; x^ ¼ ðx0 x Þ1 x0 y : ð7Þ

The x0 x matrix is made up of the sum of the squares and cross products of the
independent variables. For the p  p matrix, x0 x is to be inverted, the rank being p.
 ARTICLE IN PRESS
266 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

3. Indices for model assessment

Assessment of regression model can be done through estimation of multiple

coefficient of determination (R2 ) as well as adjusted multiple coefficient of
determination ðR2adj Þ and also through estimation of standard error ðE s Þ:

3.1. Estimation of multiple coefficient of determination (R2) and adjusted multiple

coefficient of determination (R2adj)

A measurement of how well the multiple regression line fits the data or what
proportion of the total variation has been explained by regression line can be
assessed through a statistics called multiple coefficient of determination (R2) and it is
used universally.
Now multiple coefficient of determination, R2 ; can be expressed as
sum of squares due to regression
R2 ¼ :
sum of squares about the mean
For determination of multiple coefficient of determination ðR2 Þ; Draper and Smith
(1966) demonstrated an equation that can be written in matrix notation as:

Total sum of the squares;

X

0

y2i or y0 y ¼ nȳ2 þ x^ 0 x0 y nȳ2 þ y0 y  x^ x0 y ð8Þ
P
So that the three components of the total sum of the squares, y0 y or y2i are:

1. nȳ2 ; the0 sum of the square due to the mean,

^ ¼ e0 e ¼ P e2 ¼ Pðy  y^ Þ2 the sum of squares
2. y0 y  x^ x0 y ¼ ðy  x x^ Þj ðy  x xÞ i i i
of0 deviations from
P  regression
or the residual sum of squares, and
2
3. x^ x0 y nȳ2 ¼ y^ i  ȳ ; the sum of squares due to regression.

The sum of the squares about the mean or the sum of the squares corrected for the
mean is y0 y nȳ2
Therefore,

0 
x^ x0 y nȳ2
R2 ¼ : (9)
ðy0 y nȳ2 Þ

The range of R2 is 0 to 1.
Sometimes with addition of explanatory variables to the model, value of R2 will
never decrease, even if the additional variables are explaining an insignificant
proportion of the variation in y. The addition of this explanatory variable is not
desirable. Then an alternative measure of the goodness of fit that is useful in multiple
regressions is R2 adjusted ðR2adj Þ:
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 267

Now, R2adj can be expressed as

0

y0 y  x^ x0 y =ðn  pÞ
R2adj ¼ 1 

y0 y nȳ2 =ðn  1Þ

0

y0 y  x^ x0 y ðn  1Þ
¼1

y0 y nȳ2 ðn  pÞ

or
 
ðn  1Þ
R2adj ¼ 1  ð1  R Þ 2
: (10)
ðn  pÞ
Where,
0


0

x^ x0 y nȳ2 y0 y þ x^ x0 y  y0 y nȳ2
R2 ¼
 ¼

y0 y nȳ2 y0 y nȳ2

0

y0 y  x^ x0 y
¼1
 :
y0 y nȳ2

3.2. Estimation of standard error (Es)

The efficiency of the regression line can be assessed through the estimation of
standard error. An unbiased estimate for the variance, E 2s is given by:
P 2
ei
E 2s ¼
ðn  pÞ
or
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2
ei
Es ¼ : (11)
ðn  pÞ
Estimation of E s indicates that the smaller that the value is, the better the
prediction is.

4. Inferences of multiple regression coefficients

Whether the multiple regression coefficients are significant or insignificant

can be assessed through F statistics and t statistics. By applying the F test
the significance of multiple regression coefficient as a whole can be assessed
and significance of partial regression coefficient can be assessed through t
statistics.
 ARTICLE IN PRESS
268 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

4.1. Significance of multiple regression coefficients as a whole

Sometimes, the assumed regression equation may prove to be statistically not

significant. Whether this is so, is determined by a comparison of the variance
contributed by the regression and the error variance, using the F test. In other words,
this test indicates the suitability of the assumed model for predicting data. The null
hypothesis is that all the partial regression coefficients, x1 ; x2 ; . . . ; xp are equal to
zero, i.e. Ho: x1 ¼ x2 ¼    ¼ xp ¼ 0 versus the alternative hypothesis Ha: at least
one of the values of x is non-zero. From the definition of R2 ; F can be expressed in
the following form:
½R2 =ðp  1Þ
F cal ¼ : (12)
½ð1  R2 Þ=ðn  pÞ
The decision rule for the ‘F’ test is:

Reject Ho if F cal 4F ð1  a; p  1; n  pÞ:

Accept Ho if F cal pF ð1  a; p  1; n  pÞ:
where F ð1  a; p  1; n  pÞ is a value chosen from the F table for the appropriate
level of significance, a: Acceptance of the null hypothesis implies that the explanatory
variables in the regression equation are of little or no use in explaining the variation
in the dependant variable. Rejection of the null hypothesis implies that at least one of
the explanatory variables helps explain the variation in dependant variable.

4.2. Significance of a partial multiple regression coefficients

If the null hypothesis in the case of the F test is rejected then t statistics should be
performed to evaluate the contribution of individual variables to explain the
variation of dependant variable. From the t statistics, if any regression coefficient
found not to be statistically significant then a new regression equation is proposed
eliminating the corresponding variable of statistically insignificant regression
coefficient.
The decision rule for the ‘t’ test is:

Reject Ho if tcal 4tð1a=2; npÞ or tcal o  tð1a=2; npÞ :

Accept Ho if tð1a=2; npÞ ptcal ptð1a=2; npÞ :
Where, tð1a=2; npÞ is a value chosen from the t table for the appropriate level of
significance, a:
The t statistics is given in the following expression:

x^ i
tcal ¼ : (13)
sx^ i
2
Where, sx^ i is the positive square root of C 1
ii E s :
And C ii is the ith diagonal element of ðx xÞ1 :
1 0
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 269

5. Selection of best ‘‘fit’’ regression model

A common practice in selecting a multiple regression model is to perform several

regressions on a given set of data using different combinations of the independent
variables. The model may be linear or non-linear and all the independent variables
may or may not be included in the best model. The regression, that fits the data best,
is then selected. The selection of appropriate model mainly depends on the judgment
of the researcher, including incorporation of relevant theory and knowledge of the
subject under study, and careful examination of scatter plots, residual plots (The
differences between the true and predicted values for the point in the sample are
called the residuals.), and also regression diagnostics (Dielman, 2001).
Partial information about the relationship between the variables can be obtained
by forming a scatter plots matrix (Devore and Farnum, 1999). This is just a
collection of two-dimensional scatter plots, arranged in a square array, in which each
variable is plotted against every other variable. The matrix gives the clear indication
of whether any single predictor might be related to dependent variable (y), whether
the relationship might be linear or non-linear. The scatter plots may also reveal any
strong relation between any particular pair of predictors.
The variable selection techniques in the regression model are tools to aid the
researcher in sifting through a number of explanatory variables to determine
which one he or she feels should be included in the regression equation. The
reliability of the technique depends on the judgment of selecting the variables giving
emphasis on the theory of the subject. Some of the variable selection techniques are
as follows:
(i) all possible regressions;
(ii) backward elimination;
(iii) forward selection and
(iv) stepwise regression.

Of the above listed methods, the all-possible regressions technique is considered to

be the best method because it examines every possible model from the given list of
variables (Dielman, 2001). In this method all possible equations are formed taking
every possible combination of independent variables. If all of the equations required
to have an intercept term, 2p1 regression equations would have to be formed where
p is the number of independent variables one of which is always equal to one to
produce the intercept term. Generally, the commonly used criteria to help in
choosing between the alternative regression models are:
(i) The best equation yielding the largest value of R2 or R2adj among the regression
models.
(ii) The model with the smallest standard error ðE s Þ are considered to be the best.
(iii) Finally, the real test, of how good the resulting regression model is, depends on
the ability of the model to predict the dependent variable for observation on the
independent variables that were not used in estimating the regression coefficients
(Haan, 1994).
 ARTICLE IN PRESS
270 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

6. Transforming non-linear models

Non-linear models can be transformed to linear models. For example;

The non  linear model y ¼ zxx (14)
can be linearized by using a logarithmic transformation, which is given as
follows:
log y ¼ log z þ x log x (15)
or
yt ¼ zt þ xt xt ; (16)
where
yt ¼ log y; zt ¼ log z; xt ¼ x and xt ¼ log x:
Standard regression techniques can now be used to estimate zt and xt for
Eq. (16). Then the values of z and x can be computed for the above general
expressions.

7. Regression modelling for bearing capacity of a square footing on reinforced pond

ash

Data used in the regression analysis are obtained from experimental model
tests of 80 mm square footing on pond ash reinforced with geotextiles performed
by the authors at the Civil Engineering Department, Bengal Engineering College,
India. The properties of the materials used in the present investigation are given
below:
Pond ash: In the present investigation pond ash, used as a foundation medium,
has been procured from the thermal power plants at Budge Budge in West Bengal,
India. The chemical composition of Budge Budge pond ash in percentage by
weight is as follows: SiO2:50.50, Al2O3:25.01, Fe2O3:0.71, CaO:9.73, MgO:4.18
and Loss on ignition (LOI):9.80. A grain size analysis of pond ash has been
conducted in the laboratory as per ASTM Standard D 422 (1990). From the grain
size analysis it has been found that 4%, 72% and 24% of particles of the pond ash
are gravel size, sand size and silt size, respectively. Uniformity coefficient (C u )
and coefficient of curvature (C c ) are 3.0 and 1.46, respectively. The pond ash used
in the present work is non-plastic and its specific gravity is 2.16. The maximum
dry density (MDD) and optimum moisture content (OMC) of pond ash have been
found to be 10.40 kN/m3 and 37%, respectively, from the compaction test conducted
as per ASTM standard D 698 (1991). The value of angle of internal friction at OMC
and MDD is 361.
Geotextile: Four types of geotextile sheets viz., jute geotextile (made of jute—a
natural fibre) and synthetic geotextiles: GT—320, GT—240 and GT—135 have been
used in the present investigation. The engineering properties of geotextiles as
determined in the laboratory are: Thickness of jute geotextiles, GT—320, GT–240
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 271

and GT–135 are 1.25 mm, 0.75 mm, 0.55 mm and 0.30 mm, respectively. The
breaking strength in warp direction of jute geotextiles, GT—320, GT—240
and GT—135 are 18, 40, 30, and 16 kN/m; and corresponding elongation at
break are 14%, 32%, 40%, and 36%, respectively. Whereas breaking strength in
weft direction of jute geotextiles, GT—320, GT—240 and GT—135 are 16, 36, 28,
and 12 kN/m while corresponding elongation at break are 12%, 30%, 34% and
27%, respectively.
Friction ratio: Friction ratio (f) is a vital design parameter for reinforced earth
structures. The basic mechanism of reinforced earth involves the generation
of frictional forces between the soil and reinforcement. The angle of internal
friction angle ðfd Þ of pond ash and the interface friction angle ðcÞ between pond
ash and geotextile sheet have been determined through a series of direct shear
tests. These direct shear test are conducted at OMC (37%). In case of jute—
grotextile the values of friction ratio, ‘f’ at different dry densities of pond ash
viz., 9.30, 9.80, 10.50 and 10.70 kN/m3 are 0.9230, 0.9125, 0.8902 and 0.8809,
respectively; whereas in case of GT—320, GT—240 and GT—135 the values
of friction ratio, ‘f’ at dry density of pond ash 9.30 kN/m3 are 0.7949, 0.8974
and 0.7179, respectively.
Based on existing experimental data, the regression models have been developed
for estimating the bearing capacity (qrs ; expressed in kN/m2) of a square footing at
different s=B ratios on reinforced pond ash. The parameters used in the models are
bearing capacity of square footing (qs ; expressed in kN/m2) at different s=B ratios on
unreinforced foundation medium, i.e. pond ash, ðs=BÞ in percentage, N, u=B; f, Ls =B
and S v =B: From the scatter plots matrix, it is revealed that the relationship is non-
linear. In the present study a non-linear power model has been chosen, which is given
below in the multiple variable form:
x
y ¼ x0 xx11 xx22    xpp : (17)

Where, y is the dependent variable; x1 ; x2 ; x3 ; . . . ; xp are independent variables,

x0 ; x1 ; x2 ; x3 ; . . . ; xp are the partial regression coefficients; p being the number of
predicator variables.
Eq. (17) is typically fit to experimental data by taking the logarithms of both
sides and expressing them in a relationship with the linear structure of equation as
follows:
log y ¼ log x0 þ x1 log x1 þ x2 log x2 þ    þ xp log xp : (18)

Eq. (18) can be best fit using the least squares technique. Goodness of fit statistics,
such as R2 or R2adj and E s of the Eq. (18) are determined, which reflect the efficiency
of the model. Further for assessing the accuracy of the model (Eq. (17)) the
percentage of error (E p ) is calculated.
Where,
 
y  y^
Ep ¼  100: (19)
y
 ARTICLE IN PRESS
272 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

7.1. Numerical model for qrs in terms of qs and s/B

Based on the experimental data, a power model has been developed to estimate qrs
relating qs and s=B ratio in percentage as given below:

s b2
qrs ¼ b0 ðqs Þb1 : (20)
B
The values of multiple regression coefficients (b0 ; b1 and b2 ), coefficient of
determination ðR2 Þ; standard error ðE s Þ and percent predictions having less than 3%
error (PE3) corresponding to S v =B; u=B; Ls =B; f and N are presented in Table 1. For
all the cases, the coefficient of determination ðR2 Þ has been found to be 0.99. Fig. 1
illustrates a typical graph of observed data versus predicted data of bearing capacity
of square footing on reinforced pond ash for Sv =B ¼ 0:3125; N ¼ 3; u=B ¼ 0:3125;
Ls =B ¼ 7 and f ¼ 0:7179 with maximum percentage of error of less than 3%. From
this analysis it is revealed that there exits a strong relation between qrs and the
predictors qs and s=B:
From the above model (Eq. (20)) qrs can be estimated by using the values of qs and
s=B for specific values of other different bearing capacity parameters like N, u=B; f,
S v =B and Ls =B: The authors have greater comfort with this model because one has
to conduct field testing to obtain qs to be used in the model to assess qrs ; which will
take into consideration the types of pond ash, compaction procedure, etc. But other
than the specific values of N, u=B; f, Sv =B and Ls =B (Table 1), the model is not
useful. To avoid this limitation a generalized model to predict qrs is developed in the
following section.

7.2. Numerical model for qrs in terms of qs, s/B, N, u/B, Sv/B, f and Ls/B

A regression model has been developed to estimate qrs considering all of the vital
bearing capacity parameters viz., qs ; s=B; N, u=B; f, Ls =B and S v =B: Analyzing the
scatter plot matrix obtained from the experimental data, a non-linear power model
has been chosen as follows:

a12  a15
a11 s a13 a14 Ls ðu=BÞ ðS =BÞ
qrs ¼ a10 qs N f a16 a17v : (21)
B B

All possible regressions procedure is used to select the best subset of predictors.
In the present investigation the number of data points, n ¼ 2088 and number of
predictors (independent variables)=8, one of which is always equal to 1 to produce
the intercept term. Then numbers of possible equations are 2ðp1Þ ¼ 2ð81Þ ¼ 128:
But from the physical significance point of view in the present analysis every
possible equation should contain a10 qas 11 ðs=BÞa12 ; because qrs is the improved form
of qs due to inclusion of reinforcement into the foundation medium and qrs is
mobilized only when settlement(s) takes place. Then the number of equations
reduces to 25 ¼ 32:
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 273

Table 1
Constants in multiple regression, coefficient of determination (R2 ), standard error (E s ) and PE3 from
Eq. (20)

Sv =B Ls =B f u=B N No. of obs. b0 b1 b2 R2 Es PE3a

0.3125 7 0.9230 0 1 20 1.094 1.0821 0.1399 0.99 0.017 60

2 22 2.839 0.7201 0.3319 0.99 0.008 100
3 22 26.56 0.0145 0.9258 0.99 0.011 96
4 17 23.96 0.0788 1.1553 0.99 0.004 100
5 16 54.13 0.3160 1.3442 0.99 0.004 100
6 20 8.388 0.1553 1.1277 0.99 0.007 100
7 21 22.09 0.0042 1.1113 0.99 0.008 100
8 19 6.996 0.3088 0.9187 0.99 0.010 95
0.3125 7 0.9230 0.1250 1 24 1.193 1.0729 0.0742 0.99 0.016 71
2 21 14.43 0.3555 0.5172 0.99 0.014 86
3 21 7.838 0.4878 0.5478 0.99 0.014 81
4 19 35.06 0.0048 0.9846 0.99 0.013 94
5 17 46.97 0.2588 1.3478 0.99 0.005 100
6 18 9.068 0.2038 1.0232 0.99 0.012 94
7 21 52.07 0.2327 1.2755 0.99 0.019 62
8 20 47.96 0.1867 1.2263 0.99 0.028 60
0.3125 7 0.9230 0.6325 1 23 3.403 0.6688 0.2934 0.99 0.008 100
2 21 6.177 0.4957 0.4724 0.99 0.011 86
3 22 28.03 0.04613 0.9114 0.99 0.012 77
4 22 13.43 0.2507 0.7706 0.99 0.006 100
5 22 12.36 0.2361 0.8791 0.99 0.025 59
6 18 49.41 0.1223 0.9919 0.99 0.016 72
0.3125 7 0.9230 0.9375 1 22 1.883 0.8509 0.1444 0.99 0.015 68
2 22 3.270 0.6769 0.2969 0.99 0.014 77
3 21 4.953 0.5257 0.4967 0.99 0.011 100
4 21 5.851 0.4839 0.4669 0.99 0.019 57
0.3125 7 0.9230 1.25 1 22 2.093 0.8945 0.0246 0.99 0.014 82
2 20 0.899 1.1314 0.1609 0.99 0.014 70
0.3125 7 0.9230 1.5625 1 21 1.094 1.0853 0.1579 0.99 0.013 90
2 19 1.226 1.0657 0.1468 0.99 0.013 84
3 20 1.543 1.0193 0.1528 0.99 0.008 95
0.3125 7 0.9230 0.3125 1 24 5.8465 0.5409 0.397 0.99 0.006 100
2 23 0.8123 1.0369 0.254 0.99 0.013 91
3 20 12.499 0.1055 1.052 0.99 0.012 85
4 18 1.1371 1.0442 0.348 0.99 0.025 44
5 21 79.630 0.4496 1.533 0.99 0.005 100
6 18 8.8512 0.4457 0.785 0.99 0.013 94
7 13 171.75 0.3194 1.115 0.99 0.009 100
0.3125 7 0.9125 0.3125 1 21 1.8956 0.8358 0.2823 0.99 0.009 100
2 21 2.6580 0.7326 0.4162 0.99 0.007 100
3 21 33.422 0.1185 0.8443 0.99 0.005 100
4 21 41.425 0.0413 0.9626 0.99 0.008 100
 ARTICLE IN PRESS
274 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

Table 1 (continued )

Sv =B Ls =B f u=B N No. of obs. b0 b1 b2 R2 Es PE3a

0.3125 7 0.8902 0.3125 1 21 0.9601 0.0901 0.2873 0.99 0.023 86

2 21 6.9399 0.4913 0.6471 0.99 0.023 57
3 21 1.9832 0.7653 0.5379 0.99 0.014 81
0.3125 7 0.8809 0.3125 2 20 0.8731 0.9164 0.3369 0.99 0.025 60

0.1250 7 0.9230 0.3125 2 30 2.1189 0.8364 0.2191 0.99 0.017 87

0.1250 7 0.9230 0.3125 3 26 13.947 0.2513 0.6886 0.99 0.020 85
0.6250 7 0.9230 0.3125 2 28 37.878 0.1824 1.0279 0.99 0.017 86
0.6250 7 0.9230 0.3125 3 30 51.62 0.2736 1.1657 0.99 0.013 90
0.9375 7 0.9230 0.3125 2 29 9.5789 0.2601 0.6315 0.99 0.015 93
0.9375 7 0.9230 0.3125 3 29 3.84 0.5066 0.5603 0.99 0.014 79
1.2500 7 0.9230 0.3125 2 28 7.8870 0.2471 0.9601 0.99 0.012 86
0.3125 6 0.9230 0.3125 1 20 1.786 0.8636 0.1849 0.99 0.003 100
2 21 4.533 0.5255 0.5473 0.99 0.013 81
0.3125 5 0.9230 0.3125 1 20 1.671 0.8813 0.1722 0.99 0.005 100
2 22 4.855 0.5499 0.4907 0.99 0.008 100
3 23 9.232 0.3390 0.7327 0.99 0.007 100
4 22 64.95 0.2845 1.2422 0.99 0.004 100
5 20 189.1 0.6147 1.5029 0.99 0.004 100
6 20 40.64 0.1458 1.1739 0.99 0.023 60
7 22 20.36 0.04058 1.0566 0.99 0.031 65

0.3125 3 0.9230 0.3125 1 21 0.991 1.04599 0.0332 0.99 0.010 100

2 22 5.161 0.5271 0.4898 0.99 0.003 100
3 20 9.543 0.3000 0.7195 0.99 0.009 95
4 20 194.9 0.6334 1.5075 0.99 0.005 100
5 20 195.2 0.6315 1.5249 0.99 0.008 100
6 20 39.30 0.1437 1.1666 0.99 0.017 70
7 20 37.57 0.1217 1.1512 0.99 0.020 65
0.3125 2 0.9230 0.3125 1 21 0.753 1.0917 0.0288 0.99 0.013 86
2 20 0.528 1.2069 0.0181 0.99 0.027 50
3 21 5.000 0.5180 0.5068 0.99 0.004 100
4 21 18.70 0.1195 0.8314 0.99 0.002 100
5 21 38.11 0.1084 1.0597 0.99 0.007 100
0.3125 7 0.7179 0.3215 1 22 3.754 0.6049 0.3592 0.99 0.008 91
2 22 6.356 0.4583 0.5062 0.99 0.006 96
3 21 8.972 0.3398 0.6860 0.99 0.004 100
4 22 45.248 0.1386 1.0638 0.99 0.007 96
5 21 79.049 0.2952 1.1731 0.99 0.005 100
6 22 25.539 0.0432 0.9810 0.99 0.009 91
7 22 18.923 0.1177 0.9458 0.99 0.014 91
0.3215 7 0.8974 0.3215 1 22 0.7446 1.0854 0.0706 0.99 0.020 81
2 22 3.2134 0.6561 0.4068 0.99 0.013 100
3 21 4.4714 0.5564 0.5691 0.99 0.013 95
4 22 39.405 0.1154 1.0938 0.99 0.015 91
5 22 22.829 0.0712 0.9549 0.99 0.062 90
6 22 24.000 0.0535 1.0243 0.99 0.076 90
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 275

Table 1 (continued )

Sv =B Ls =B f u=B N No. of obs. b0 b1 b2 R2 Es PE3a

0.3215 7 0.7949 0.3215 1 22 1.0428 1.0241 0.0415 0.99 0.010 96

2 22 4.2648 0.5928 0.416 0.99 0.009 96
3 22 12.517 0.2259 0.845 0.99 0.015 76
4 22 23.523 0.02185 1.024 0.99 0.009 96
5 21 31.145 0.0452 1.084 0.99 0.009 91
6 22 16.218 0.1714 0.929 0.99 0.019 82
a
PE3 percent predictions having less than 3% error.

Now the possible models are as follows:

s a12  a15
Ls ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 N a13 f a14 a16 a17v ; (M.1)
B B

s a12  a15
a13 a14 Ls ðu=BÞ
qrs ¼ a10 qas 11 N f a16 ; (M.2)
B B

s a12  a15
a13 a14 Ls ðS =BÞ
qrs ¼ a10 qas 11 N f a17v ; (M.3)
B B

s a12
ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 N a13 f a14 a16 a17v ; (M.4)
B

s a12  a15
Ls ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 N a13 a16 a17v ; (M.5)
B B

s a12  a15
a14 Ls ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 f a16 a17v ; (M.6)
B B

s a12  a15
Ls
qrs ¼ a10 qas 11 N a13 f a14 ; (M.7)
B B

s a12
ðu=BÞ
qrs ¼ a10 qas 11 N a13 f a14 a16 ; (M.8)
B

s a12
ðS =BÞ
qrs ¼ a10 qas 11 N a13 f a14 a17v ; (M.9)
B

s a12  a15
Ls ðu=BÞ
qrs ¼ a10 qas 11 N a13
a16 ; (M.10)
B B

s a12  a15
Ls ðS =BÞ
qrs ¼ a10 qas 11 N a13 a17v ; (M.11)
B B
 ARTICLE IN PRESS
276 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

10000

f = 0.7179
u / B = 0.3125
Ls / B = 7
1000 Sv / B = 0.3125
N=3
Observed qrs

100

10

0 % variation line

1
1 10 100 1000 10000
Predicted qrs

Fig. 1. Typical observed qrs versus predicted qrs from Eq. (20).

s a12
ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 N a13 a16 a17v ; (M.12)
B
 a15

s a12
Ls ðu=BÞ
qrs ¼ a10 qas 11 f a14a16 ; (M.13)
B B

a12  a15
a11 s a14 Ls ðS =BÞ
qrs ¼ a10 qs f a17v ; (M.14)
B B

s a12
ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 f a14 a16 a17v ; (M.15)
B

s a12 L a15
s ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 a16 a17v ; (M.16)
B B

s a12
qrs ¼ a10 qas 11 N a13 f a14 ; (M.17)
B

s a12  a15
Ls
qrs ¼ a10 qas 11 N a13 ; (M.18)
B B

s a12
ðu=BÞ
qrs ¼ a10 qas 11 N a13 a16 ; (M.19)
B

s a12
ðS =BÞ
qrs ¼ a10 qas 11 N a13 a17v ; (M.20)
B
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 277

s a12  a15
a14 Ls
qrs ¼ a10 qas 11 f ; (M.21)
B B

s a12
ðu=BÞ
qrs ¼ a10 qas 11 f a14 a16 ; (M.22)
B

s a12
ðS =BÞ
qrs ¼ a10 qas 11 f a14 a17v ; (M.23)
B

s a12 L a15
s ðu=BÞ
qrs ¼ a10 qas 11 a16 ; (M.24)
B B

s a12 L a15
s ðS =BÞ
qrs ¼ a10 qas 11 a17v ; (M.25)
B B

s a12
ðu=BÞ ðS =BÞ
qrs ¼ a10 qas 11 a16 a17v ; (M.26)
B

s a12
qrs ¼ a10 qas 11 N a13 ; (M.27)
B

s a12
qrs ¼ a10 qas 11 f a14 ; (M.28)
B

s a12 L a15
s
qrs ¼ a10 qas 11 ; (M.29)
B B

s a12
ðu=BÞ
qrs ¼ a10 qas 11 a16 ; (M.30)
B

s a12
ðS =BÞ
qrs ¼ a10 qas 11 a17v ; (M.31)
B

s a12
qrs ¼ a10 qas 11 : (M.32)
B
Based on 2088 numbers of experimental data using logarithmic transformation,
the multiple regression coefficients of the above models are determined. The values
of R2 ; R2adj ; and E s of the logarithmic transformed models are calculated. After
examining the values of R2 and R2adj for all of the 32 numbers of models, it has been
found that, the values of R2 ð¼ 0:9450Þ and R2adj ð¼ 0:9448Þ are the highest in case of
the relevant parameters. The value of standard error, E s ð¼ 0:079Þ for model (M.1)
has been found to be the minimum considering the values of E s for all the models.
The percent predictions having less than 5%-, 10%-, 15%-, 20%- and 25% error
(viz., PE5, PE10, PE15, PE20 and PE25) are calculated. From these calculated values, it
is found that model (M.1) gives the highest values of PE5, PE10, PE15, PE20 and PE25,
the corresponding values being 30, 52, 72, 86 and 97, respectively. From the above
observation, it is opined that model (M.1) is the best one. Thus the final equation to
 ARTICLE IN PRESS
278 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

10000

1000
Observed qrs

100

10
0 % variation line

1
1 10 100 1000 10000
Predicted qrs

Fig. 2. Observed qrs versus predicted qrs from Eq. (22).

estimate qrs is obtained as given below:

s 0:3647  0:1111
0:1901 0:1465 Ls
qrs ¼ 1:9165q0:7777
s N f 0:8105ðu=BÞ 0:5171ðSv =BÞ : (22)
B B
The plot of predicted data from Eq. (22) versus observed data is shown in Fig. 2.
For the analysis purposes, after performing logarithmic transformation the above
model (M.1) can be written as follows:

s
logðqrs Þ ¼ logða10 Þ þ a11 logðqs Þ þ a12 log þ a13 logðNÞ
 
B  
Ls u Sv
þ a14 logðf Þ þ a15 log þ logða16 Þ þ logða17 Þ: ð23Þ
B B B

7.2.1. Significance of the multiple regression coefficients as a whole (F test)

Significance of the multiple regression coefficients as a whole of the Eq. (23) was
tested using the F test. The values of ‘F’ has been calculated by putting the values of
R2 ; p and n in Eq. (12). The values of ‘F’ (calculated) has been found to be 5145.
From the table of F distribution with a ¼ 0:05; F (0.95, 7, 2080)=2.01. Therefore,
F cal ¼ 5145 and this is greater than the tabulated F ð0:95; 7; 2080Þ ¼ F critical ¼ 2:01;
which rejects the null hypothesis, i.e. at least one of the explanatory variables of
Eq. (23) helps explain the variation of qrs :

7.2.2. Significance of the partial multiple regression coefficients

The significance of the partial multiple regression coefficients of Eq. (23) were
performed through the t statistics. Table 2 presents the summary of the t statistics of
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 279

Table 2
Values of t statistics for different parameters of model (M.1)

Parameters Coefficients Standard error t statistics tcritical ¼ tð0:975; 2080Þ

Intercept Log ða10 Þ ¼ 0:2825 0.030310 9.32

qs a11 ¼ 0:7777 0.017880 43.48
S=B a12 ¼ 0:3646 0.012990 28.05
N a13 ¼ 0:1901 0.006590 28.81 1.96
f a14 ¼ 0:1465 0.054909 2.67
Ls =B a15 ¼ 0:1111 0.012325 9.01
u=B log ða16 Þ ¼ 0:091 0.005648 16.15
Sv =B log ða17 Þ ¼ 0:286 0.011197 25.57

the coefficients; the decision rule for rejection of the null hypothesis is satisfied, i.e.
all the explanatory variables help explain the variation of qrs :

7.2.3. Validation of the model

For validation of the proposed model, for estimating qrs ; the model has been tested
with 99 additional experimental data that were not used in developing the model.
The result is presented in Table 4. From the above result it is found that a good
agreement exits between predicted qrs and observed qrs :

7.2.4. Applicability of the proposed model

The predictors are chosen in such a way that the estimated value of qrs from the
proposed model will be close to that of the field value. The authors have greater
comfort with this model because one has to conduct field testing to obtain qs to be
used in the model to assess qrs ; which will take into consideration the types of pond
ash, compaction procedure, footing size in addition to friction ratio ‘f’, which is to be
determined from laboratory direct shear test, and this will also take into
consideration the reinforcement mechanism between the pond ash and types of
reinforcement.

7.2.5. Limitation of the values of the predictors

The above model can be used for estimation of qrs in the field, and the predictor
variables should be chosen by considering physical conditions of the field otherwise
it would become uneconomical. Table 3 presents comparative study for the optimum
values of u=B; Ls =B and Dlr =B (Dlr being the depth, measured from the underside of
the footing, of the last layer of the reinforcement for which the maximum value of qrs
is obtained) obtained from the previous researchers and the present investigation.
From this table it is found that the optimum value for u=B lies between 0.175 and
0.5. The maximum value of number of layers (N) depends on value of u=B and Sv =B:
So values of N should be calculated from the equation given below:
Dlr ¼ u þ ðN  1Þ  S v ;
Dlr  u
or; N ¼ 1 þ : ð24Þ
Sv
 ARTICLE IN PRESS
280 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

Table 3
Optimum values of u=B; Dlr =B and Ls =B from several authors

References u=B Dlr =B Ls =B Reinforcing Foundation

material medium

Binquet and Lee (1975a) 0.3 2.0 — Aluminium foil Sand

Akinmusuru and 0.5 1.75 — Fibre strip Sand
Akinbolade (1981)
Fragaszy and Lawton — — 6–7 Aluminium foil Sand
(1984)
Guido et al. (1985) — — 2.5 Geotextile Sand
Ramanatha Ayyar et al. — — 5 Coir rope Sand
(1988)
Mandal and Sah (1992) 0.175 — — Geogrid Clay
Yetimoglu et al. (1994) 0.25 — — Geogrid Sand
Adams and Collin (1997) 0.48 — — Geogrid Sand
Shin and Das (1999) 0.25–0.5 — — Geogrid Sand
Hataf and Baziar (2000) 0.315 — — Waste tire material —
Shin et al. (2002) 2 — Geogrid Sand
Present investigation 0.25–0.3125 1.75 5–7 Geotextiles (natural Pond ash
and synthetic)

From Table 3, it is also found that Dlr lies in between 1.75 B and 2 B. The
maximum value of Sv is so chosen that the last layer of reinforcement should be
within the value of Dlr ; otherwise uneconomical conditions would be created. From
Table 3 it is further observed that the optimum value of sheet length lies between 5B
and 7B and beyond this length it becomes uneconomical. So, the length of the
reinforcement should be used within the range of 5B and 7B. Value of friction ratio
(f) is to be found out from laboratory testing. In the proposed model the value of
friction ratio (f) used lies within the range of 0.7179 and 0.9230 (Table 4).
It is recommended that for a particular field condition, the results from the
regression model should be checked for at least one series of model tests to ensure
that the regression model is acceptable for the particular field condition.

7.2.6. Illustrative example to estimate qrs using the proposed model

To illustrate the example, calculations are presented for the following problem.
Estimation of qrs of square footing (2 m  2 m) resting on reinforced pond ash.
The following data are given: (1) bearing capacity ðqs Þ of a square footing
(2 m  2 m) at s=B of 8%=120 kN/m2; and (2) friction ratio, f ð¼ ðc=fd Þ ¼ 0:9:
The following step-by-step procedure illustrates the calculation procedure.
Step 1. Assume Dlr =B ¼ 1:80ð1:75pDlr =Bp2:00Þ;
Sv =B ¼ 0:3 ðS v =BpDlr =BÞ;
u=B ¼ 0:3 ð0:175pu=Bp0:5Þ;
Ls =B ¼ 5 ð5pLs =Bp7Þ:
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 281

Table 4
Comparison of q^ rs (predicted, using additional data not used in developing the model) and corresponding
observed qrs

qs (kN/m2) S/B (%) N f Ls/B u/B Sv/B q^ rs (kN/m2) qrs (kN/m2) Ep

74 3.20 2 0.9230 7 0.3125 0.1250 100 101 +1.2

84 3.74 2 0.9230 7 0.3125 0.1250 118 115 2.8
95 4.30 2 0.9230 7 0.3125 0.1250 136 134 1.8
107 4.98 2 0.9230 7 0.3125 0.1250 158 153 2.7
116 5.46 2 0.9230 7 0.3125 0.1250 173 170 1.5
130 6.36 2 0.9230 7 0.3125 0.1250 201 192 4.7
152 7.80 2 0.9230 7 0.3125 0.1250 243 225 8.1
169 9.13 2 0.9230 7 0.3125 0.1250 280 252 11.0
262 24.38 2 0.9230 7 0.3125 0.6250 405 355 14.1
266 25.75 2 0.9230 7 0.3125 0.6250 418 374 11.7
269 26.98 2 0.9230 7 0.3125 0.6250 429 395 8.6
273 28.13 2 0.9230 7 0.3125 0.6250 440 417 5.5
277 29.58 2 0.9230 7 0.3125 0.6250 454 429 5.9
282 30.73 2 0.9230 7 0.3125 0.6250 466 442 5.5
226 15.48 3 0.9230 7 0.3125 0.6250 330 273 20.8
233 16.73 3 0.9230 7 0.3125 0.6250 348 297 17.3
233 16.83 3 0.9230 7 0.3125 0.6250 349 321 8.8
239 18.02 3 0.9230 7 0.3125 0.6250 365 345 5.8
245 19.21 3 0.9230 7 0.3125 0.6250 380 361 5.2
250 20.56 3 0.9230 7 0.3125 0.6250 396 385 2.9
254 21.76 3 0.9230 7 0.3125 0.6250 409 413 +0.8
258 23.09 3 0.9230 7 0.3125 0.6250 424 435 +2.5
261 24.32 3 0.9230 7 0.3125 0.6250 436 458 +4.7
265 25.49 3 0.9230 7 0.3125 0.6250 448 483 +7.2
268 26.71 3 0.9230 7 0.3125 0.6250 461 509 +9.5
59 2.50 3 0.7179 7 0.3125 0.3125 71 68 4.9
85 3.75 3 0.7179 7 0.3125 0.3125 109 100 8.8
108 5.00 3 0.7179 7 0.3125 0.3125 146 130 12.1
129 6.25 3 0.7179 7 0.3125 0.3125 181 165 10.3
147 7.50 3 0.7179 7 0.3125 0.3125 215 195 10.7
164 8.75 3 0.7179 7 0.3125 0.3125 248 223 11.4
179 10.00 3 0.7179 7 0.3125 0.3125 278 251 10.8
192 11.25 3 0.7179 7 0.3125 0.3125 307 281 9.1
204 12.50 3 0.7179 7 0.3125 0.3125 334 312 6.9
214 13.75 3 0.7179 7 0.3125 0.3125 359 337 6.5
222 15.00 3 0.7179 7 0.3125 0.3125 382 364 5.2
230 16.25 3 0.7179 7 0.3125 0.3125 404 392 3.0
237 17.50 3 0.7179 7 0.3125 0.3125 424 414 2.5
243 18.75 3 0.7179 7 0.3125 0.3125 443 436 1.7
248 20.00 3 0.7179 7 0.3125 0.3125 461 457 1.0
252 21.25 3 0.7179 7 0.3125 0.3125 478 477 0.4
256 22.50 3 0.7179 7 0.3125 0.3125 494 499 +1.0
260 23.75 3 0.7179 7 0.3125 0.3125 510 519 +1.8
263 25.00 3 0.7179 7 0.3125 0.3125 525 538 +2.5
267 26.25 3 0.7179 7 0.3125 0.3125 540 554 +2.6
31 1.25 4 0.7179 7 0.3125 0.3125 35 37 +4.4
59 2.50 4 0.7179 7 0.3125 0.3125 75 66 13.2
85 3.75 4 0.7179 7 0.3125 0.3125 115 95 20.3
108 5.00 4 0.7179 7 0.3125 0.3125 154 134 15.2
 ARTICLE IN PRESS
282 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

Table 4 (continued )

qs (kN/m2) S/B (%) N f Ls/B u/B Sv/B q^ rs (kN/m2) qrs (kN/m2) Ep

129 6.25 4 0.7179 7 0.3125 0.3125 192 163 17.5

59 2.50 2 0.7948 7 0.3125 0.3125 67 71 +5.5
85 3.75 2 0.7948 7 0.3125 0.3125 102 106 +3.4
108 5.00 2 0.7948 7 0.3125 0.3125 137 139 +1.3
129 6.25 2 0.7948 7 0.3125 0.3125 170 165 3.2
147 7.50 2 0.7948 7 0.3125 0.3125 202 194 4.4
164 8.75 2 0.7948 7 0.3125 0.3125 233 217 7.1
179 10.00 2 0.7948 7 0.3125 0.3125 262 239 9.2
192 11.25 2 0.7948 7 0.3125 0.3125 288 261 10.3
204 12.50 2 0.7948 7 0.3125 0.3125 314 279 12.4
214 13.75 2 0.7948 7 0.3125 0.3125 337 300 12.3
222 15.00 2 0.7948 7 0.3125 0.3125 359 318 13.0
230 16.25 2 0.7948 7 0.3125 0.3125 380 333 13.9
237 17.50 2 0.7948 7 0.3125 0.3125 399 353 13.1
243 18.75 2 0.7948 7 0.3125 0.3125 417 370 12.6
248 20.00 2 0.7948 7 0.3125 0.3125 434 383 13.1
252 21.25 2 0.7948 7 0.3125 0.3125 449 404 11.3
256 22.50 2 0.7948 7 0.3125 0.3125 465 419 11.0
204 12.50 1 0.8974 7 0.3125 0.3125 280 286 +2.3
214 13.75 1 0.897 7 0.3125 0.3125 301 300 0.4
222 15.00 1 0.8974 7 0.3125 0.3125 320 314 2.2
230 16.25 1 0.8974 7 0.3125 0.3125 339 326 3.9
237 17.50 1 0.8974 7 0.3125 0.3125 356 336 5.8
243 18.75 1 0.8974 7 0.3125 0.3125 372 347 7.3
248 20.00 1 0.8974 7 0.3125 0.3125 387 358 7.9
252 21.25 1 0.8974 7 0.3125 0.3125 401 372 7.9
256 22.50 1 0.8974 7 0.3125 0.3125 415 382 8.6
260 23.75 1 0.8974 7 0.3125 0.3125 428 391 9.2
263 25.00 1 0.8974 7 0.3125 0.3125 440 400 10.0
267 26.25 1 0.8974 7 0.3125 0.3125 453 411 10.3
31 1.25 2 0.8974 7 0.3125 0.3125 32 35 +7.9
59 2.50 2 0.8974 7 0.3125 0.3125 68 61 11.2
85 3.75 2 0.8974 7 0.3125 0.3125 104 104 0.4
108 5.00 2 0.8974 7 0.3125 0.3125 139 138 1.0
129 6.25 2 0.8974 7 0.3125 0.3125 174 168 3.2
147 7.50 2 0.8974 7 0.3125 0.3125 206 198 4.3
164 8.75 2 0.8974 7 0.3125 0.3125 237 225 5.5
179 10.00 2 0.8974 7 0.3125 0.3125 266 248 7.2
179 10.00 1 0.7179 7 0.3125 0.3125 226 200 13.1
192 11.25 1 0.7179 7 0.3125 0.3125 249 217 15.0
204 12.50 1 0.7179 7 0.3125 0.3125 271 233 16.3
214 13.75 1 0.7179 7 0.3125 0.3125 291 248 17.3
222 15.00 1 0.7179 7 0.3125 0.3125 310 261 19.0
230 16.25 1 0.7179 7 0.3125 0.3125 328 272 20.7
237 17.50 1 0.7179 7 0.3125 0.3125 344 286 20.3
243 18.75 1 0.7179 7 0.3125 0.3125 360 296 21.6
248 20.00 1 0.7179 7 0.3125 0.3125 374 306 22.3
252 21.25 1 0.7179 7 0.3125 0.3125 388 319 21.8
256 22.50 1 0.7179 7 0.3125 0.3125 401 328 22.2
260 23.75 1 0.7179 7 0.3125 0.3125 414 339 22.2
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 283

Step 2. Calculate N (Ref. Eq. (24))

1:8  0:3
N ¼1þ ¼ 6:
0:3
Step 3. Calculate qrs (Ref. Eq. (22))
qrs ¼ 1:9165ð120Þ0:7777 ð8Þ0:3647 ð6Þ0:1901 ð0:9Þ0:1465
ð5Þ0:1111 0:8105ð0:3Þ 0:5171ð0:3Þ
¼ 216 kN=m2
Therefore, the bearing capacity of a square footing (2 m  2 m) on the given pond
ash reinforced with geotextile for the given parameters is 216 kN/m2.

8. Conclusions

Pond ash reinforced with geotextiles may find potential application as safe
foundation medium. Numerous research has been carried out on the bearing
capacity of footing on sand or clay. However, there is paucity of data regarding
prediction of bearing capacity on reinforced or unreinforced pond ash. This study
presents a regression model based on 2088 data points, to predict the bearing
capacity of square footing on pond ash reinforced with geotextiles. The significance
of different predictor variables, governing the bearing capacity is also examined
through statistical analysis.
On the basis of analysis of the results obtained from the present investigation, the
following conclusions may be drawn:
1. Power model obtained through regression analysis of the model test results of this
study may be used for predicting qrs :
2. A non-linear power model has been developed to estimate the qrs in terms
of qs and s=B for different parameters viz., N, f, u=B; Ls =B and S v =B: For all
the cases values of R2 have been found to be 0.99 and for most of the cases PE3
is 100%.
3. All possible regression techniques have been used to select the best subset of
predictors and from this analysis all the parameters governing bearing capacity
viz., qs ; s=B; N, f, u=B; Ls =B and S v =B are included in the best model. It is also
revealed from this statistical analysis that all these predictors have significant
contribution to qrs :
4. The best model, obtained through the present analysis considering all the
predictor variables, has shown that 86% data have less than 20% error.
5. The authors have greater comfort with this model because one has to conduct
field testing to obtain qs to be used in the model to assess qrs ; which will take
into consideration the types of pond ash, compaction procedure, etc. in addi-
tion to other parameters like friction ratio ‘f ‘which is to be determined
from laboratory direct shear testing, and this will also take into considera-
tion the reinforcement mechanism between the pond ash and types of
reinforcement.
 ARTICLE IN PRESS
284 A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285

6. This model is applicable for square footings on geotextile reinforced pond

ash having the values of friction ratio (f), Dlr =B; u=B and Ls =B within the range
of 0:7179pf p0:9230; 1:75pDlr =Bp2:00; 0:175pu=Bp0:5; and 5pLs =Bp7;
respectively, whereas value of S v =B should be chosen with in the value of Dlr =B:
Beyond these ranges of the parameters and for other field conditions, the
model should be checked through for at least one set of laboratory model
test results.

References

Abedin, Z., Hassan, M., Dewan, A.S., 1997. Bearing capacity of a jute cloth reinforced composite sand
bed. Proceedings of the 14th International Conference on Soil Mechanics and Foundation
Engineering, Vol. 3. Humbarg, pp. 1553–1556.
Adams, M.T., Collin, J.G., 1997. Large model spread footing load tests on geosynthetic reinforced soil
foundations. Journal of Geotechnical and Geoenvironment Engineering, ASCE 123 (1), 66–72.
Akinmusuru, J.O., Akinbolade, J.A., 1981. Stability of loaded footings on reinforced soil. Journal of
Geotechnical Engineering, ASCE 107 (6), 819–827.
ASTM, D 422 (1990). Test Method for Particle-Size Analysis of Soils. Philadelphia, PA, pp. 6–12.
ASTM, D 698 (1991). Test Method for Laboratory Compaction Characteristics of Soil Using Standard
Effort. Philadelphia, PA, pp. 28–35.
Binquet, J., Lee, K.L., 1975a. Bearing capacity test on reinforced earth slabs. Journal of Geotechnical
Engineering, ASCE 101 (12), 1241–1255.
Binquet, J., Lee, K.L., 1975b. Bearing capacity analysis on reinforced earth slabs. Journal of Geotechnical
Engineering, ASCE 101 (12), 1257–1276.
Boushehrian, H.J., Hataf, N., 2003. Experimental and numerical investigation of the bearing capacity of
model circular and ring footings on reinforced sand. Geotextiles and Geomembranes 21 (4), 241–256.
Dash, S.K., Krishnaswamy, N.R., Rajagopal, K., 2001. Bearing capacity of strip footings supported on
geocell-reinforced sand. Geotextiles and Geomembranes 19 (4), 235–256.
Devore, J.L., Farnum, N.R., 1999. Applied Statistics for Engineers and Scientists. International Thomson
Publishing, Inc., USA.
Dielman, T.E., 2001. Applied Regression Analysis for Business and Economics. Duxbury. Thomson
Learning, Inc., USA.
Dixit, R.K., Mandal, J.N., 1993. Bearing capacity of geosynthetic reinforced soil using variational
method. Geotextiles and Geomembranes 12 (6), 543–566.
Draper, N.R., Smith, H., 1966. Applied Regression Analysis. Wiley, New york.
Fragaszy, J.R., Lawton, E., 1984. Bearing capacity of reinforced sand subgrades. Journal of Geotechnical
Engineering, ASCE 110 (10), 1500–1507.
Guido, V.A., Biesiadecki, G.L., Sullivian, M.J., 1985. Bearing capacity of geotextile reinforced
foundation. Proceedings of the 11th International Conference on soil Mechanics and Foundation
Engineering, San Francisco, CA, Vol. 3, pp. 1777–1780.
Haan, T.C., 1994. Multiple linear regression. Statistical Methods in Hydrology. Affiliated, East–West
Press Pvt. Ltd., New Delhi, pp. 197–219.
Hataf, N., Baziar, A., 2000. Use of tire shreds for bearing capacity improvement of shallow footings on
sand. Proceedings of the Third International Conference on Ground Improvement Techniques,
Singapore, pp. 189–194.
Mahmoud, M.A., Abdrabbo, F.M., 1989. Bearing capacity tests on strip footing resting on reinforced
sand subgrades. Canadian Geotechnical Journal 26, 154–159.
Mandal, J.N., Sah, H.S., 1992. Bearing capacity tests on geogrid—reinforced clay. Geotextiles and
Geomembranes 11 (3), 327–333.
Michalowski, L.R., 2004. Limit loads on reinforced foundation soils. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE 130 (4), 381–390.
 ARTICLE IN PRESS
A.K. Bera et al. / Geotextiles and Geomembranes 23 (2005) 261–285 285

Omar, M.T., Das, B.M., Puri, V.K., Yen, S.C., Cook, E.E., 1994. Bearing capacity of foundation on
geogrid reinforced sand. Proceedings of the 13th International Conference on Soil Mechanics and
Foundation Engineering, New Delhi, India, Vol. 2, pp. 1279–1282.
Otani, J., Ochiai, H., Yamamoto, K., 1998. Bearing capacity analysis of reinforced foundations on
cohesive soil. Geotextiles and Geomembranes 16 (4), 195–206.
Ramanatha Ayyar, T.S., Joseph, J., Beena, K.S., 1988. Bearing capacity of sand reinforced with coir rope.
Proceedings of the First Indian Geotextiles Conference on Reinforced Soil and Geotextiles, IIT
Bombay, India, pp. A.11–A.16.
Raza, S.A., Khan, M.A., Ahmad, M.S., Sharma, A., 1999. Behaviour of footing resting on fly ash bed
reinforced with geofibres and treated with Lime, cationic surfactant. In: Dayal, U., Shina, R., Kumar,
V. (Eds.), Fly Ash Disposal and Deposition; Beyond 2000 A.D. Narosa Publishing House, New Delhi,
India, pp. 204–210.
Samtani, N.C., Sonpal, R.C., 1989. Laboratory tests of strip footing on reinforced cohesive soil. Journal of
Geotechnical Engineering, ASCE 115 (9), 1326–1330.
Shin, E.C., Das, B.M., 1999. Bearing capacity of strip foundation on geogrid—reinforced sand,
Proceedings of the 11th International Conference on soil Mechanics and Geotechnical Engineering,
Seoul, Korea, pp. 189–192.
Shin, E.C., Das, B.M., Lee, E.S., Atalar, C., 2002. Bearing capacity of strip foundation on geogrid-
reinforced sand. Geotechnical and Geological Engineering 20 (2), 169–180.
Vidal, H., 1966. La terre Armee, Anales de I’ Institute Techanique du Batiment et des Travaux Publiques,
France, July–August, pp. 888–938.
Yamamoto, K., Kusuda, K., 2001. Failure mechanisms and bearing capacities of reinforced foundations.
Geotextiles and Geomembranes 19 (3), 127–162.
Yetimoglu, T., Wu, J.T.H., Saglamar, A., 1994. Bearing capacity of rectangular footings on geogrid-
reinforced sand. Journal of Geotechnical Engineering, ASCE 120 (12), 2083–2099.