Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/343502045

Experimental and Statistical Study on Single and Groups of Stone Columns

Article  in  Key Engineering Materials · August 2020


DOI: 10.4028/www.scientific.net/KEM.857.399

CITATION READS

1 189

3 authors:

Maki Jafar Al-Waily Mohammed Yousif Fattah


Al-Furat Al-Awsat Technical University University of Technology, Iraq
26 PUBLICATIONS   123 CITATIONS    432 PUBLICATIONS   2,656 CITATIONS   

SEE PROFILE SEE PROFILE

Maysa Salem
Al-Furat Al-Awsat Technical University
6 PUBLICATIONS   3 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Response of Circular Footing on Dry Dense Sand to Impact Load with Different Embedment Depths View project

My Ph. D. Work View project

All content following this page was uploaded by Maki Jafar Al-Waily on 13 October 2021.

The user has requested enhancement of the downloaded file.


Key Engineering Materials Submitted: 2020-02-03
ISSN: 1662-9795, Vol. 857, pp 399-408 Revised: 2020-02-15
© 2020 Trans Tech Publications Ltd, Switzerland Accepted: 2020-03-01
Online: 2020-08-07

Experimental and Statistical Study on Single and Groups of Stone


Columns
Maki J. Mohammed Al-Waily1,a, Mohammed Y. Fattah2,b
and Maysa Salem Al-Qaisi1,c
1
Al-Musaib Technical College, Al-Furat Al-Awsat Technical University, Babylon, Iraq
2
Civil Engineering Department, University of Technology, Baghdad, Iraq
maki_jafar@yahoo.com, bmyf_1968@yahoo.com, cmayssa11salim@yahoo.com
a

Keywords: Laboratory, Statistical models, stone column, single, group.

Abstract. In the present study, 24 laboratory models on soft clay treated with stone columns were
carried out. The results for each case are analysed for the purpose of constructing a statistical model
linking the variables studied. The experiments showed that the stress concentration and bearing
capacity of soil treated with stone column increase with increasing the undrained shear strength (cu),
number of columns and L/d ratio. The models represent a single stone column and a group of stone
columns. The studied variables are three dependent variables, the stress concentration ratio (n),
bearing capacity of soil treated with stone column (q) and the settlement improvement ratio (Sr) due
to the existence of stone columns. The independent variables are six: the undrained shear strength of
clay soil, with three values (6, 9 and 12 kPa), the number of stone columns (1, 2, 3 and 4 columns)
and the length (L) to the diameter (D) of the stone column or (L/D) ratio in two values (6 and 8).
Besides, the bearing capacity of the soil treated (q) with stone columns and the settlement
improvement ratio were used in some statistical models as independents. After regression analysis, a
set of equations that correlate the previous variables have been suggested. The incepted values for
dependent variables are close to the laboratory results.

Introduction
Soft soils require improvement of their properties if they have to be utilized as foundation. For this
purpose, different soil improvement techniques such as preloading, sand drains, dynamic compaction
and stone column have been used as economical alternative to deep foundations. The most attractive
method appears to be the stabilization of soft soil by installation of stone column that have been
widely used in many countries during the last three decades. This is related to feasibility and the
suitability of stone column technique. Stone columns were known in France in the 1830s, they have
recently been rediscovered and the mechanism of their behavior under load is not well understood.
This method consists of forming vertical holes in the ground that are filled with crushed rock to form
columns or “piles” confined by the soil. They are ideally suited for improving soft clays and silts and
also for loss sand [1-4].
Over the last six decades, many researchers have proved through theoretical studies, laboratory
model tests, and full scale (field) tests that stone columns are highly efficient in improving many
ranges of soil with increasing bearing capacity, minimizing expected long-term decline, increasing
soil stability and differing from piles, because it acting as a drain for surrounding soil. Another
important factor in the use of stone columns is that it is less economical than the piles and does not
require installation in the soil to complex equipment, but the installation equipment is very easy and
varied. The raw materials of the stone column of crushed stone, stone or gravel are available in almost
a place. Many researchers have tried to use additional materials (additives) with the crush stone such
as lime, cement, bitumen geotextile or steel in very few percentages and others. All additives have
proved to be more efficient in increasing the efficiency of the stone column [5-7].
The experimental tests carried out by [8] showed that the stone columns with L/D=8 provided a
stress concentration ratio n of 1.4, 2.4, 2.7, and 3.1 for the soil having a shear strength cu=6 kPa,
treated with single, two, three, and four columns, respectively. The values of n were decreased to 1.2,

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications Ltd, www.scientific.net. (#540854688-12/07/20,22:36:36)
1
400 Geotechnical Engineering Development

2.2, 2.5, and 2.8 when the L/D = 6. The values of n increase when the shear strength of the treated
soil was increased to 9 and 12 kPa. Fattah et al. [9] found experimentally that whether a column was
floating or end bearing (resting on a rigid stratum), encasement of the column by a geogrid was most
effective in improving the bearing ratio of reinforced soil by approximately between 1.29 and 1.83
times that of untreated soil, reducing the settlement by approximately 0.71 - 0.45 times that of
untreated soil for spacing, s = 2.5d. The bearing improvement ratio (bearing capacity of treated-to-
untreated soil) increased with decreasing spacing of stone columns. A higher improvement ratio was
achieved for the models reinforced with stone columns at s = 2.5d.
The following equation was obtained by carrying out statistical analysis using the SPSS (Statistical
Package for the Social Sciences) program form the present experimental work and previous studies
data. The following equation was obtained with a determination coefficient (R2) = 0.871 to estimate
the bearing capacity of floating stone column group installed in clays of different undrained shear
strengths between (4–25) kPa and with different diameters and L/D ratios constructed by cased bored
method [10]:

qu = 15.34 Cu 0.401 As0.266 Ns0.084 (L/D)0.526 (1)


where:
cu: the undrained shear strength, kPa,
As: the area replacement ratio,
Ns: No. of stone columns, and
L/D: length to diameter ratio.
In the current work, 24 laboratory model tests were carried out and the results of the tests were
analysed to show the behaviour of each model and in order to construct a statistical model linking the
variables studied.

Experimental Work
Figure 1 shows full details about the experimental set up that is used to conduct all model tests by
the aid of soil box with dimensions 1100 x 1000 x 800 mm3. The soil used in the experimental work
consisted of 48% clay, 42 silt% and 10% sand. The clay soil bed is prepared in three values of
undrained shear strength, (6, 9 and 12 kPa). The stone column is built using crushed stone of particle
size about (1/6 to 1/7) of the diameter of column. Tow values of the length to diameter ratio of stone
column, one of them (L/D = 6) which represent the case of floating stone column, and the other (L/D
= 8) to represent the full penetration case are used.
The load is applied in increments, each one equal to (1/10) the failure load. At each load increment,
the total stress on the system of soil plus stone column and the stress on the stone column alone are
recorded. The stress on soil surrounding the stone column is computed based on these previous
stresses. The loading tests are carried out in five cases of model tests, untreated soil and soil treated
with single or two, three and four stone columns for three shear strengths and two values of (L/D)
ratio i.e. there are one model test for untreated soil and 24 model tests for soil treated with stone
columns.
Key Engineering Materials Vol. 857 401

2
1- Loading frame.
2- Hydraulic jack.
3-Handle.
4
10 4- Proving ring to read total stress
5- Proving ring to read the stress
on stone column
6- Dial gauge to read the settlement
5 7- Plate loading test on soil.
3 8- Plate loading test on stone column.
9 6 9- References to fix the dial gauge.
10- Timer.
11- Stone column with dia. = 5o mm
8 7 12- Clay soil bed.
13- Soil box (1100 x 1000 x 800) mm3
14- Rigid base.

12
13 11

14

Figure 1. Schematic diagram of the experimental set-up.

Results of Tests
The analysis of the results of the model tests included computing the following parameters:
1) Stress concentration ratio, which is known as the ratio between the applied stresses on the stone
column to the applied stress on the soil surrounding the column, or (n =σc/σs).
2) Bearing capacity of the soil treated with stone column (qtr), which is calculated at the same value
of settlement under applied stress for models, untreated soil and soil treated with stone columns.
3) Settlement improvement ratio (Sr) which is defined as the ratio of settlement of treated soil (Str)
to the settlement of untreated soil (S) or (Sr= Str/S).

The undrained shear strength of the soil prepared in the soil box tests ranged from 6 to 12 kPa.
This means that the soil improvement included very soft clays. The settlement improvement achieved
by the stone column was calculated at the same stress increment applied to the treated soil and to the
untreated soil. The values of the settlement at the end of the tests were used. Table 1 summarizes
results of all 24 model tests. The two following points are observed from this table:
• The stress concentration and bearing capacity of soil treated with stone column increase with
increasing the undrained shear strength(cu), number of columns and L/d ratio.
• The settlement improvement ratio (Sr) decreases when the cu, number of columns and L/d ratio
increase.

Statistical Models
Statistical Models for Stress Concentration Ratio
First model (standard multiple regression).By the aid of the program STATISTICA the results
of the regression analysis to establish the relationship between the stress concentration ratio n as a
dependent variable and the [shear strength (cu), number of columns (No. of col.), (L/D) ratio, the
402 Geotechnical Engineering Development

bearing capacity of the treated soil (qtr) and the settlement reduction ratio (Sr)] as a most significant
independent variables. To choose the appropriate model, the predicted model was:

n = - 0.496 - 0.007 cu + 0.057 Ns + 0.085 (L/D) + 0.014 q - 0.470 Sr (2)

Table 2 displays more details about (R, R2, adjusted R2 and the standard error of estimate).
Second model (standard multiple regression-zero intercept). In the second statistical model
consider the stress concentration ratio n as a dependent variable and the shear strength (cu), number
of columns (No. of col.), (L/D) ratio, the bearing capacity of the treated soil (qtr) and the settlement
reduction ratio (Sr) are the most significant independent variables. Also, Table 2 displays more details
about (R, R2, adj. R2 and the standard error of estimate). The model predicted the following equation:

n = 0.014 cu + 0.041 Ns + 0.069 (L/D) + 0.036 q - 0.937 Sr (3)


Table 1. Summary of model tests results.
cu No. of Stress concentration qtr
Case L/D Sr
(kPa) columns ratio (n) (kPa)
1 6 1 6 1.2 37.2 0.73
2 9 1 6 1.5 46.0 0.53
3 12 1 6 1.6 56.5 0.68
4 6 2 6 2.2 49.0 0.4
5 9 2 6 2.3 54.2 0.4
6 12 2 6 2.8 58.4 0.67
7 6 3 6 2.5 55.8 0.31
8 9 3 6 2.6 67.5 0.27
9 12 3 6 3.4 74.6 0.43
10 6 4 6 2.7 67.6 0.17
11 9 4 6 2.7 73.3 0.23
12 12 4 6 3.0 77.0 0.36
13 6 1 8 1.4 38.8 0.65
14 9 1 8 1.5 46.4 0.57
15 12 1 8 1.8 53.7 0.76
16 6 2 8 2.4 52.7 0.33
17 9 2 8 2.6 58.9 0.39
18 12 2 8 3.1 70.3 0.49
19 6 3 8 2.9 60.1 0.24
20 9 3 8 3.1 70.2 0.24
21 12 3 8 3.8 80.8 0.36
22 6 4 8 3.2 70.8 0.19
23 9 4 8 3.6 78.0 0.18
24 12 4 8 3.6 90.3 0.27

Third model (stepwise multiple regression). This analysis was based on the statistical approach
using stepwise multiple regression analysis method to establish a relationship between n and the most
significant independent variables (L/D and q). The stepwise method reduces the number of variables
initially present to those that are actually significant. The model predicted the relationship as:
n = - 0.302 + 0.0304 q + 0.0736 (L/D) (4)
Fourth model (stepwise multiple regression-zero intercept). The fourth statistical model the
depending on results of the (stepwise multiple regression-zero intercept). The (L/D) ratio, bearing
Key Engineering Materials Vol. 857 403

capacity q and Sr ratio are the most significant independent variables. The model predicted the
relationship as:
n = 0.060 L/D + 0.041q - 0.900 Sr (5)

Fifth model (stepwise multiple regression-zero intercept). The fifth model is based on the
results of the (stepwise multiple regression-zero intercept). The bearing capacity qtr is the most
significant independent variable. The model predicted the relationship as:
n = 0.042 q (6)
Sixth Model (Stepwise Multiple Regression-Zero Intercept). The last statistical model is
depended on the results of the (stepwise multiple regression-zero intercept). The settlement reduction
ratio (Sr) is the most significant independent variable. The model predicted the relationship as:
n = 4.775 Sr (7)

Summary of Statistical Models for Stress Concentration Ratio n


Tables 2 and 3 show a summary of the statistical models for stress concentration ratio n.
Table 2. Summary of statistical models of stress concentration ratio n.
Model R R2 Adjusted R2 Std. Error of the Estimate
1 0.930 0.866 0.828 0.311
2 0.995 0.990 0.987 0.305
3 0.930 0.864 0.843 0.297
4 0.995 0.990 0.988 0.291
5 0.993 0.985 0.985 0.328
6 0.803 0.645 0.630 1.6234
Table 3. Stress concentration ratio models.
Model Stress concentration ratio n
1 n = - 0.496 - 0.007 cu + 0.057 Ns + 0.085 (L/D) + 0.014 q - 0.470 Sr
2 n = 0.014 cu + 0.041 Ns + 0.069 (L/D) + 0.036 q - 0.937 Sr
3 n = - 0.302 + 0.0304 q + 0.0736 (L/D)
4 n = 0.060 L/D + 0.041q - 0.900 Sr
5 n = 0.042 q
6 n = 4.775 Sr

Statistical Models for Bearing Capacity of Treated Soil with Stone Column
First model (standard multiple regression). To establish a relationship between the bearing
capacity q as dependent variable and [the stress concentration ratio n, shear strength (cu), number of
columns (No. of col.), (L/D) ratio and settlement ratio (Sr)] as significant independent variables, the
appropriate model has been chosen. Table 4 displays more details about (R, R2, adjusted R2 and the
standard error of estimate. The model was predicted as:
q = 10.176 + 2.424 n + 2.781cu + 6.6 Ns + 1.464 (L/D) - 15.01Sr (8)
Second Model (Standard Multiple Regression-Zero Intercept). The results of (Standard
multiple regression-zero intercept) connecting the stress concentration ratio n, the shear strength (cu),
number of column (No. of Col.), (L/D) ratio and settlement reduction ratio (Sr) are the most significant
independent variables, showed that the model predicted the following equation:
404 Geotechnical Engineering Development

q = 2.365 n + 2.7 cu + 7.925 Ns + 2.07 (L/D) - 6.582 Sr (9)


Third Model (Stepwise Multiple Regression-Zero Intercept). Tables 4 and 5 present the results
of (Stepwise multiple regression-zero intercept). The stress concentration ratio n and the settlement
reduction ratio (Sr) are the most significant independent variables. The model predicted the following
equation:
q = 21.3 n + 17.26 Sr (10)
Summary of statistical model for bearing capacity. Tables 4 and 5 show the summary of
Statistical models for bearing capacity q.
Table 4. Summary of statistical models of the bearing capacity q.
Model R R2 Adjusted R2 Std. Error of the Estimate
1 0.988 0.976 0.970 2.376
2 0.999 0.999 0.998 2.471
3 0.995 0.991 0.990 6.365
Table 5. Bearing capacity models.
Model Bearing capacity models
1 q = 10.176 + 2.424 n + 2.781cu + 6.6 Ns + 1.464 (L/D) - 15.01Sr
2 q = 2.365 n + 2.7 cu + 7.925 Ns + 2.07 (L/D) - 6.582 Sr
3 q = 21.3 n + 17.26 Sr

Statistical Models for Settlement Reduction Ratio Sr


First Model (Standard Multiple Regression)
By the aid of the program STATISTICA (Version 5.5A), the results of the regression analysis can
be summarized in Tables 6 and 7 to establish the relationship between the settlement reduction ratio
Sr as a dependent variable and the (stress concentration ratio n, shear strength (cu), number of
columns, (L/D) ratio and the bearing capacity of the treated soil) as significant independent variables,
to choose the appropriate model. Table 6 displays more details about (R, R2, adjusted R2 and the
standard error) of estimate. The model predicted the following equation:
Sr = 0.66 - 0.024 n + 0.058 cu + 0.004 Ns + 0.012 (L/D) - 0.013 q (11)
Second Model (Standard Multiple Regression-Zero Intercept).
Tables 6 and 7 show the results of (Standard multiple regression-zero intercept). The stress
concentration ratio n, shear strength (cu), number of columns, (L/D) ratio and the bearing capacity of
the treated soil (q) are the most significant independent variables. The model predicted the following
equation:
Sr = - 0.117 n + 0.078 cu + 0.065 Ns + 0.087 (L/D) - 0.013 q (12)
Third Model (Stepwise Multiple Regression).
Tables 6 and 7 show the results of (Standard multiple regression). The shear strength (cu), number
of columns, and the bearing capacity of the treated soil (q) as significant independent variables. The
model predicted the following equation:
Sr = 0.726 + 0.054 cu - 0.018 Ns - 0.012 q (13)
Summary of Statistical Models for Sr
Tables 6 and 7 show the summary of statistical models for bearing capacity q.
Key Engineering Materials Vol. 857 405

Table 6. Settlement reduction ratio Sr models.


Model Sr
1 Sr = 0.66 - 0.024 n + 0.058 cu + 0.004 Ns + 0.012 (L/D) - 0.013 q
2 Sr = - 0.117 n + 0.078 cu + 0.065 Ns + 0.087 (L/D) - 0.013 q
3 Sr = 0.726 + 0.054 cu - 0.018 Ns - 0.012 q
Table 7. Summary of statistical models of settlement reduction ratio Sr.
Model R R2 Adjusted R2 Std. Error of the Estimate
1 0.942 0.887 0.855 0.07
2 0.977 0.954 0.942 0.108
3 0.940 0.883 0.866 0.067

Appropriate Models
Stress Concentration Ratio n
When the data fitted are very well, then the explanatory power of the model is measured by the
coefficient of multiple determination R2=0.990 (adj. R2 = 0.988), and the standard error of estimate
is 0.291 (Table 2). The Average Absolute Relative Error (AARE) method is also used to choose the
appropriate model, the following formula is used in this method:
nobserved (t ) − n predicted (t )
ARE (t ) = (14)
nonserved (t )
where: ARE(t) is the absolute relative error for model (t)
1 t = 24
AARE = ∑ ARE ( t ) (15)
24 t =1
The Average Absolute Relative Error (AARE) values of the six statistical models of n are:
• 2.26 for the first model.
• 2.26 for the second model.
• 2.37 for the third model.
• 2.22 for the fourth model.
• 3.15 for the fifth model.
• 14.92 for the sixth model.
According to the previous values of AARE, the most suitable model with most significant
independent variables is the fourth model or:
n = 0.060 L/D + 0.041q - 0.900 Sr
Figure 2 shows the relation between the predicted and observed values of n, according to the
pervious equation.
406 Geotechnical Engineering Development

Figure 2. Predicted versus observed n values.

Bearing Capacity (q)


The Average Absolute Relative Error (AARE) method is used to choose the appropriate model
The AARE values of the three statistical models of q are:
• 0.66 for the first model.
• 0.74 for the second model.
• 1.95 for the third model.
The final suitable model with most significant independent variables is:
q = 10.176 + 2.424 n + 2.781cu + 6.6 Ns + 1.464 (L/D) - 15.01Sr
The AARE of 0.66, which indicates that the predicted errors, is, on average, 6.6% of the predicted
values, is considered to be a good result. Figure 3 shows the relation between the predicted and
observed values of the bearing capacity according to the pervious equation.

Figure 3. Predicted versus observed q values.


Key Engineering Materials Vol. 857 407

Settlement Reduction Ratio Sr


The Average Absolute Relative Error (AARE) method is used to choose the appropriate model
The AARE values of the three statistical models of Sr are:
• 3.08 for the first model
• 4.76 for the second model
• 3.04 for the third model
According to the previous values of AARE, the most appropriate model with most significant
independent variables is the third model or
Sr = 0.726 + 0.054 cu - 0.018 Ns - 0.012 q
The explanatory power of the model is measured by the coefficient of multiple determination R2
= 0.883 (adj. R2 = 0.886), and the standard error of estimate is 0.067. Figure 4 shows the model
efficiency. It can be noticed that the model predicts the settlement reduction ratio Sr efficiently as
shown by slope of the best fit line (slope = 0.95).

Figure 4. Predicted versus observed Sr values.

Conclusions
The experimental work and statistical analysis of data revealed the following conclusions:
• Stress concentration ratio, which is known as the ratio between the applied stresses on the stone
column to the applied stress on the soil surrounding the column, or (n =σc/σs).
• Bearing capacity of the soil treated with stone column (qtr), which is calculated at the same
value of settlement under applied stress for models, untreated soil and soil treated with stone
columns.
• Settlement improvement ratio (Sr) which is defined as the ratio of settlement of treated soil (Str)
to the settlement of untreated soil (S) or (Sr= Str/S).
• A number of models has been derived to estimate the stress concentration ratio, stone column’s
bearing capacity and settlement reduction ratio with good correlation coefficients.
408 Geotechnical Engineering Development

References
[1] Hughes, J. M. O., & Withers, N. J., Reinforcing of soft cohesive soils with stone columns, Ground
engineering. Vol. 7, No. 3, (1974) 42 – 49.
[2] Madhav, M. R., and Vitkar, P. P, Strip footing on weak clay stabilized with a granular trench or
pil, Can. Geotech. J.,15(4), (1978) 605–609.
[3] Aboshi, H., Ichimoto, E., Enoki, M. & Harada, K., The compozer: a method to improve
characteristics of soft clays by inclusion of large diameter sand columns, proceedings International
Conference on soil reinforcement; Reinf. Earth and other techniques. Vol. 1, Paris, (1979) 211 – 216.
[4] Indian Standards (IS), Indian standard code of practice for de-sign and construction for ground
improvement-guidelines. Part 1: Stone columns, IS 15284 (Part 1), New Delhi, India (2003).
[5] Greenwood, D. A., Mechanical improvement of soils below ground surface”, Proceedings Ground
Engineering Conference, Institution of Civil Engineers, (1970).
[6] Jellali, B., Bouassida, M., and De Buhan, P., A homogenization method for estimating the
bearing capacity of soils reinforced by stone columns, Int. J. Numer. Analyt. Meth. Geomech. ,29(10),
(2005) 989–1004.
[7] Kempfert, H. G., “Ground improvement methods with special emphasis on column-type
techniques”, Int. Workshop on Geotechnics of Soft Soils (2003).
[8] Fattah, M. Y., Shlash, Kais T., Al-Waily, Maki J., Stress Concentration Ratio of Model Stone
Columns in Soft Clays, Geotechnical Testing Journal, ASTM, Vol. 34, No. 1, (2003) 61-71, Paper
ID GTJ103060Available online at: www.astm.org.
[9] Fattah, M. Y., Zabar, B. S., Hassan, H. A., Experimental Analysis of Embankment on Ordinary
and Encased Stone Columns, International Journal of Geomechanics, ASCE, Vol. 16, No. 4 (2016),
DOI 10.1061/(ASCE)GM.1943-5622.0000579 , 04015102.
[10] Fattah, M. Y., Al-Neami, M. A., Al-Suhaily, A. S., Estimation of Bearing Capacity of Floating
Group of Stone Columns, Engineering Science and Technology, an International Journal, Vol. 20,
(2017) 1166–1172, http://dx.doi.org/10.1016/j.jestch.2017.03.005, Elsevier.

View publication stats

You might also like