You are on page 1of 19

PERFORMANCE OF A STONE COLUMN FOUNDATION

By James K. Mitchell, 1 F. ASCE a n d T i m o t h y R. H u b e r , 2 M . ASCE


Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

ABSTRACT: Vibro-replacement stone columns were used to support a large


wastewater treatment plant founded on up to 48 ft (15 m) of soft estuarine
deposits. Support was required for distributed foundation stresses up to 3,000
psf (145 kPa). The basic design requirement was a loading of 30 tons (265 kN)/
stone column with a settlement of less than 0.25 in. (6 mm) under that load in
a test on a single column within a group. Column spacings ranged from a
4 ft x 5 ft (1.2 m by 1.5 m) pattern under the most heavily loaded areas, to a
7 ft x 7 ft (2.1 m by 2.1 m) pattern under lightly loaded areas. Twenty-eight
single column load tests were done during the installation of the 6,500 stone
columns to evaluate load-settlement behavior. Laboratory tests were done to
provide soil property data needed for finite element predictions of both the load
test behavior and settlements of the completed structures. Predicted load test
settlements were somewhat greater than those recorded during the load tests,
but agreement was generally good. The installation of stone columns led to a
reduction in settlements to about 30-40% of the values to be expected on un-
improved ground. The settlement of a large uniformly loaded area of improved
ground was predicted to be about ten times that measured in a load test on a
single column within the area. Measured settlements varied from 1.0-2.4 in.
(25-60 mm) for a soft sediment thickness of 32-35 ft (10-11 m). A settlement
of about 2.5 in. (64 mm) was predicted by the finite element analysis. Settle-
ment predictions using other, simpler methods gave values which agreed rea-
sonably with both the measured values and the finite element predictions.

INTRODUCTION

The vibro-replacement method for constructing compacted gravel stone


columns for increasing the load carrying capacity of compressible fine-
grained soils was used in Santa Barbara, Calif, to support a new 11,000,000
gal/day (42,000,000 L/day) wastewater treatment plant situated within
a historic tidal estuary. Completed in 1976, this was the first major use
of stone columns on the west coast of the United States. Over 6,500
stone columns were installed on a 12-acre (5-ha) site to support various
structures ranging in size up to 250 ft by 500 ft (76 m x 152 m), with
uniform foundation loads of up to 3,000 psf (145 kPa).
In order to verify the design and anticipated performance of the Santa
Barbara Treatment Plant foundation, 28 field load tests were performed
on individual stone columns constructed at the site. The load testing pro-
gram is described in detail by Engelhardt and Golding (4) and Staal and
Engelhardt (13).
The study described herein was undertaken to ascertain how well the
load settlement behavior of the field load tests could be predicted and
to permit estimates of the settlements of the completed structure which
covered large loaded areas. It involved the use of nonlinear finite ele-
'Prof.
2
and Chmn. of Civ. Engrg., Univ. of California, Berkeley, Calif.
Assoc, Terratech, Inc., Monterey, Calif., formerly with Geotechnical Con-
sultants, Inc., San Francisco, Calif.
Note.Discussion open until July 1,1985. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manu-
script for this paper was submitted for review and possible publication on Jan-
uary 26, 1984. This paper is part of the Journal of Geotechnical Engineering, Vol.
Ill, No. 2, February, 1985. ASCE, ISSN 0733-9410/85/0002-0205/$01.00. Paper
No. 19491.
205

J. Geotech. Engrg. 1985.111:205-223.


ment techniques to perform parametric analytical studies for comparison
with measured field settlement data and for evaluation of stone column-
soft ground interactions. A comprehensive description of the project is
presented by Mitchell and Huber (10), so only the details necessary for
analysis of the load-settlement behavior are presented in this paper.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

SANTA BARBARA WASTEWATER TREATMENT PLANT

A plan view of the Santa Barbara Wastewater Treatment Plant layout


is presented in Fig. 1. The largest structure within the plant area is the
biological units building which houses the sedimentation and aeration
tanks. This structure measures approximately 500 ft (152 m) long by 250
ft (76 m) wide, with base elevations of structural foundations varying
from - 1 to +5 ft (-0.3 to 1.5 m).
The structures are founded on reinforced concrete slabs and footings
supported over a gravel stress distribution and drainage blanket over-
lying the stone columns. Foundation loads vary considerably, even within
the same structures, with design bearing pressures ranging from 1,000-
3,000 psf (50 to 145 kPa). Settlements of up to 4 in. (100 mm) under a
pressure of 1,500 psf (72 kPa), and 14 in. (360 mm) under a pressure of
3,000 psf (145 kPa), were estimated if the soft estuarian deposits were
loaded in their natural state.
Light industrial and warehouse type structures were located adjacent
to the site that would be damaged by settlement if the ground-water
table were lowered. The close proximity of the site to downtown Santa
Barbara precluded excessive noise or vibrations from the construction.
As the City was under a mandated time constraint to upgrade its waste-

T _ v r -T~V-\
-40 -35 -30 -Z5 -20 ^"

\ SECONDARY \ PRIMARY
\ D SEDIMENTATION SEDIMENTATIONl

\ "
\

100 t 2QO

AI PA' OEOTECHNICAL SECTION


v__ SUBSURFACE CONTOURS (ELEV. IN FEET) OF
"~^. SUITABLE FOUNDATION MATERIAL
1-
I" *! APPROXIMATE LOCATION OF FUTURE STRUCTURES

FIG. 1.Site PlanSanta Barbara Wastewater Treatment Plant

206

J. Geotech. Engrg. 1985.111:205-223.


water treatment capacity and quality, it was imperative that site prep-
aration and foundation construction be completed within six months.
For these and other reasons (10) including cost, stone columns were se-
lected for support of the structures in lieu of the more conventional pile
foundation or remove and replace alternatives.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

SITE CONDITIONS

The plant site is located immediately adjacent to the Pacific Ocean at


an approximate average elevation of 8 ft (2.5 m) above mean sea level.
Ground water typically occurs within 5 ft (1.5 m) of the ground surface.
The site is underlain by estuarine deposits of recent Holocene geologic
age (less than 10,000 yr old). These recent estuarine deposits are un-
derlain by older marine deposits of Late Pleistocene geologic age. The
sediment profile under the major biological units structure is presented
in Fig. 2.
The estuarine deposits typically consist of lenticularly interlayered soft
to firm silty clay, and sandy clay and loose to medium dense clayey sand
and silty sand, with occasional layers of sandy silt. Layers and lenses
of fine to coarse grained sand with minor amounts of gravel are locally
present, usually in the lower portions of the deposits. The total thick-
ness of the soft estuarine deposits increases progressively from the
northeast to the southwest across the site, with thicknesses in structural
areas varying from approximately 20-48 ft (6-15 m) beneath the struc-
tural foundation grade.
The older marine deposits which underlie the estuarine deposits
throughout the site consist primarily of medium dense to dense clayey
sand, and silty sand with lesser amounts of firm to stiff sandy clay and
sandy silt. Although the total depth of the older marine sediments is
unknown, geologic evidence suggests they may extend as deep as 2,000

FIG. 2.Soil Conditions Beneath Biological Units Structure


207

J. Geotech. Engrg. 1985.111:205-223.


ft (600 m) beneath the ground surface. A typical soil profile for the upper
50 ft (15 m) of deposits is shown in Fig. 3.
The site is in a very seismically active area, with nine major active or
potentially active faults located within a distance of 9 mile (15 km). The
San Andreas fault is located approximately 40 mile (60 km) to the north-
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

east. Earthquakes to a Richter magnitude of 6.5-7.0, with a resultant


maximum acceleration at the site of 0.25 g, were considered probable.
Liquefaction of sandy zones within the soft estuarine deposits was con-
sidered likely during an earthquake.

STONE COLUMN DESIGN AND CONSTRUCTION

The stone column method for ground reinforcement to increase bear-


ing capacity and reduce settlements of foundations on compressible soils
is now well established. Recent comprehensive references on the tech-
nology of stone column design and construction include those by Mitch-
ell (9), Barksdale and Bachus (2), and Greenwood and Kirsch (6).
Santa Barbara stone columns were constructed by the vibro-replace-
ment method in which 1.5-2.5 ft (0.50-0.75 m) diameter holes were jet-
ted into the in-situ soil using a vibroflot. The holes were then filled in
stages with 0.5-4 in. (12-100 mm) gravel which was compacted by vi-
bration.
The stone columns were arranged in square and rectangular patterns
in plan at each of the building locations. Center to center spacings be-

MOISTURE DRY
GE0L0SIC DEPTH ENGINEERING CONTENT 0ENSITY SPT
CLASSIFICATION UI.Km.)SYBBOL CLASSIFICATION
(%) (Ib./tuJl) V

Er^. CLAYEY SAND ( S O


M-
WM SANDY SILT I M L I

ESTUARINE
DEPOSITS
m& SILTY SAND (SM)

SILTYCLAY ( C L l
AND
21

46
95

TO
9

3
CLAYEY S I L T l M L l
-5
'C'u^'if-':-
SANDY SILT IML) 33 85 6

fr^#jfff CLAYEY SAND ( S C ] 52 49 6

IMS SANDY S I L T l M L l AND


SILTYCLAY ( C D

SILTY SAND ISMl 28 84 23


INTERLAYERED WITH
10
CLAYEY SAND (SC)
AND SANDY CLAY (CLl
21 97 19

;^i^$i. SANDY CLAY ( C L l 23 106


OLDER MARINE
y^y-:.
: : !
DEPOSITS SILTY SAND (SM
INTERLAYERED WITH
CLAYEY SAND (SCI
AND CLAYEY S I L T l M L l
15 ::?.':'V.
CLAYEY SAND (SCI 25 105

FIG. 3.Typical Soil Profile


208

J. Geotech. Engrg. 1985.111:205-223.


tween columns were established using a design loading of 30 tons (265
kN) per stone column. The densest pattern consisted of stone columns
installed in a 4 ft by 5 ft (1.2 m by 1.5 m) pattern which provided for
an approximate uniform foundation load of 3,000 psf (145 kPa). The most
widespread pattern contained stone columns installed in a 7 ft by 7 ft
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

(2.1 m by 2.1 m) pattern which provided an allowable bearing load of


approximately 1,225 psf (60 kPa).
The 7 ft by 7 ft spacing was the widest spacing allowed for this project.
This criterion was based on soil liquefaction considerations under pos-
sible earthquake loading. As with the vibrocompaction process used for
densifying sand deposits, the vibro-replacement method will also den-
sity layers of clean sand present in the native soils. However, beyond a
radius of approximately 4 ft (1.2 m) from the vibroflot, the compaction
process was not considered to impart enough vibration to effectively
density the in-situ clean sand pockets at this site.
All stone columns extended completely through the soft estuarine de-
posits and penetrated a minimum of 1 ft (0.3 m) into the underlying
older marine deposits. The lengths of the stone columns varied from 30-
49 ft (9-15 m). The diameters of stone columns exposed for load testing
or inspection following construction varied from 32-48 in. (0.81-1.22 m)
or both, with an average of 42 in. (0.07 m).
Structural loads were transferred to the stone columns through a sandy
gravel distribution blanket at the ground surface. One purpose of the
distribution blanket was to prohibit stress concentrations on the bottom
of the structural foundations. Thus the blanket's thickness was varied
in the design from 1-3 ft (0.3-1 m), dependent on the stone column
spacing. The use of compacted sandy gravel also allowed for adequate
drainage of the stone columns.

FIELD LOAD TESTS

To evaluate load-settlement behavior, 28 vertical load tests were per-


formed on individual stone columns within groups of columns. The per-
formance criterion for each load test was that the test column be able to
support the design load of 30 tons (265 kN) without exceeding 0.25 in.
(6.4 mm) settlement. In foundation areas in which this settlement cri-
terion was exceeded due to poor soil conditions, additional tests were
done, and if the criterion was not satisfied on the retest, additional stone
columns were installed. Details of the test procedures and test results
are presented by Engelhardt and Golding (4), Staal and Engelhardt (13),
and Mitchell and Huber (10).
Data pertaining to each of the 28 field load tests are presented in Table
1. Load settlement curves for each test are presented by Staal and En-
gelhardt (13). The relatively large degree of variation observed between
some of the settlements listed in Table 1 can be attributed mainly to the
nonuniformity of the soil properties found in the estuarine deposits and
the variable depth to the firm older marine deposits.
A summary of the load test results to be used for comparison with
the finite element analyses is given in Fig. 4. Ranges are shown for the
applied pressure versus settlement curves obtained from tests on col-
umns within three different stone column spacing patterns. The column
209

J. Geotech. Engrg. 1985.111:205-223.


TABLE 1.Summary of Load Test Results
Design
Number load, in Settle- Test Stone Eleva- Elevation
of days pounds ment at footing column tion top bottom
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

Test from Instal- per design area, in Stone column diam- of stone of stone
num- lation to square load, in square spacing pattern, eter, in column, column,
ber load test feet inches feet In feet inches in feet in feet
(D (2) (3) (4) (5) (6) (7) (8) (9)
1A 5 2,000 0.14 29.0 5X6 _ +6 -38
2A 8 1,540 0.33 38.7 6 x 6.5 +2.5 -37
3A 5 2,400 0.07 26.5 5x5 +7.5 -27
4A 9 1,580 0.17 32.6 5.5 x 6 +2.5 -34
5A 10 2,830 0.20 18.7 4 X 5.3 -0.5 -29
6A 12 1,220 0.15 47.8 7x7 42 +4.5 -33
7A 8 1,560 0.23 38.5 5.5 x 7 48 +4.0 -40
8A 5 1,670 0.48" 36.4 6x6 +1.5 -40
9A 11 1,200 0.24 38.2 6 x 6.5 40 -0.5 -42
10A 15 1,980 0.13 30.0 5.5 x 5.5 36 +4.5 -38
11A 11 2,400 0.25 25.5 5x5 40 -3.0 -36
12A 9 -1,250 0.24 47.8 radial pattern 38 -4.0 -40
13A iq -1,600 0.12 33.4 radial pattern 32 +4.5 -40
IB 13 2,000 0.11 29.7 5X6 +6.0 -38
2B 33 3,000 0.19 18.7 4X5 -0.5 -41
3B 33 1,540 0.13 38.2 6 x 6.5 47 -0.5 -41
4B 55 1,540 0.08 33.4 5.5 x 6 37 +2.5 -36
5B 30 1,540 0.11 33.4 5.5 x 6 40 +2.5 -34
6B 37 1,600 0.19 38.2 6 X 6.3 38 -1.0 -40
7B 17 1,200 0.07 38.2 6 X 6.5 38 -1.0 -42
8B 42 1,410 0.14 41.7 6.5 x 6.5 45 +4.5 -33
9B 42 2,600 0.12 22.5 4 x 5.75 41 -0.5 -26
10B 16 2,670 0.21 22.5 4.5 x 5 +6.0 -30
11B 26 1,220 0.10 47.8 7X7 +4.5 -38
12B 20 -3,800 0.22 15.6 radial pattern -4.0 -40
2C 5 1,540 0.44 38.7 6 X 6.5 +2.5 -37
8C 9 1,620 0.25 36.4 6X6 42 -1.0 -40
2E 5 1,540 0.19 32.6 5.5 X 6 +2.5 -36
a
Post test excavation disclosed organic, uncompacted trash fill under loading platform.

APPLIED PRESSURE (POUNDS/FOOT2)


OOP 2000 3000 5000

20 ft. per stone column

o.A

I
I? 0.4

FIG. 4.-Ranges of Stone Column Load Test Results

210

J. Geotech. Engrg. 1985.111:205-223.


spacings correspond to stone column patterns having tributary areas of
20, 33, and 49 sq ft (1.9, 3.1, and 4.6 m 2 ), with a maximum variance of
0.75 sq ft (0.1 m 2 ).
It should be noted that the load versus settlement curves for the in-
dividual load tests and, therefore, the summary curves presented in Fig.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

4, do not account for all the settlement which could have developed for
each load step. Although most of the settlement under a load increment
had taken place by the time the criterion of less than 0.010 in./hr (0.25
mm/hr) settlement before addition of the next load increment was reached,
there was still additional settlement which could occur if each load in-
crement were of longer duration. For example, in Load Test 11B the final
40-ton loading was sustained for just over 113 hr. An additional 25%
settlement was measured after the settlement rate had dropped below
0.010 in./hr (13). The load changing criterion at a settlement rate of 0.010
in./hr or less was arbitrary and was selected to avoid unduly long test-
ing times.

FINITE ELEMENT ANALYSIS

Predictions of the load test results by a finite element analysis were


compared with the in-situ field test results to evaluate the ability of the
method to model the actual behavior of the stone column foundation
effectively. The model was then used to study the effects of varying
loading conditions and soil properties, and to predict the settlement of
the completed treatment plant.
The finite element analysis of the stone column foundation system
was accomplished using an axisymmetric computer program developed
by J. M. Duncan at the University of California at Berkeley. Nonlinear,
stress-dependent material properties can be used in the analysis.
Unfortunately, an axisymmetric computer program cannot precisely
model a system of several stone columns surrounding a central column,
as is required for analysis of a load test on a single column in a group.
Therefore, to account for the presence of other individual columns, cy-
lindrical rings of elements containing stone column material properties
were included in the finite element model.
The basic axisymmetric finite element mesh used to represent the in-
dividual load tests is shown in Fig. 5. The mesh shown in this figure
represents a stone column spacing pattern of 5.75 ft by 5.75 ft (1.75 m
by 1.75 m), or approximately 33 sq ft (3.1 m2) of ground surface area,
or tributary area, per stone column. The surface dimensions of the con-
centric rings of stone column material surrounding the central stone col-
umn were calculated so that the relationship between stone column sur-
face area to total surface area remained constant. Assuming the average
stone column to have a radius of 1.75 ft (0.53 m), the 5.75 ft by 5.75 ft
(1.75 m by 1.75 m) stone column pattern has approximately 9.6 sq ft (0,9
m2) of exposed stone column material for every 33 sq ft (3.1 m 2 ) of total
surface area. In other words, the stone column replacement ratio is 29%
for this pattern. This calculation is shown in Fig. 6 for the first concentric
ring of stone column material surrounding the central stone column.
Calculated in this manner, the thickness of the first stone column ring
outside the central stone column is approximately 1.5 ft (0.46 m). Suc-

211

J. Geotech. Engrg. 1985.111:205-223.


LOADING PLATFORM STONE COLUMNS
DEPTH
III) 7S l i ' i n ' L5' 4.5' . IJ5' 3.75' 175' 4 . 2 5 '

FILL

COHESIONLESS
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

. ESTUARINE
DEPOSITS

COHESIONLESS OLDER
MARINE
DEPOSITS
COHESIVE
4

mr A W/T
NOTE: 1 ft.= 0.305 m

t
(j! of Column

FIG. 5.Basic Finite Element Mesh

cessive rings, each removed 5.75 ft (1.75 m) away from the immediately
interior ring, have thicknesses calculated in a similar manner.
The layering of soil types within the sediments was modeled by al-
ternating horizontal layers of cohesive and cohesionless soils. Although
the thickness of actual soil layers in these deposits varied from a few
inches to several feet across the site, the analysis was done assuming
layers with thicknesses of 3-6 ft (1-2 m) in the estuarine deposits,, and
3-8 ft (1-2.4 m) in the older marine deposits. The relative proportions
of cohesive and cohesionless soil types in the estuarine deposits were
determined during subsurface exploration to be approximately equal;
whereas, the older marine deposits contained approximately 60% more
cohesionless soil than cohesive.
To represent uniform loading across a large area reinforced with stone
columns, the material properties of the concrete loading platform and
212

J. Geotech. Engrg. 1985.111:205-223.


ACTUAL PLAN VIEW OF STONE COLUMNS

TOTAL SURFACE AREA III


PATTERN=(5.75t1)'=33sq.tt.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

EXPOSED STONE COLUMN


AREA = (1.7511]'>9.6lq.f1.

.-. PERCENTAGE OF STONE COLUMN


ARE* TO TOTAL ABEA' 2 3 %

TOTAL STUDY AREA ' n 18.511.1*


= 227.0 sq.ft.

REQUIRED SURFACE AREA OF STONE


\ COLUMN MATERIAL =6.29 (227.0)
' =65.8 sq.ft.

, REQUIRED RING AREA OF STONE


J COLUMN MATERIAL =65.8 iq.fl.

/ REOUREO RING THICKNESS

/ ,_**! ,l6
2J-I5.75I1.)

FIG. 6.-Calculation of Stone Column Material Surface Area

the adjacent fill depicted in Fig. 5 were changed to correspond to those


of a compacted gravel distribution blanket. The distribution blanket was
assumed to have the same material properties as the gravel in the stone
columns. Vertical pressure was applied across the top of the entire model
to simulate the uniform foundation loadings. Alternatively, the analysis
could be done as a one-dimensional compression of a single column and
its surrounding tributary area, both subjected to uniform stress incre-
ments at the ground surface.

MATERIAL PROPERTIES

Upon completion of the stone column installation, exploratory borings


were drilled between the individual stone columns, and undisturbed soil
samples were obtained by pushing thin-walled Shelby tubes into the
underlying sediments. Although a large degree of variation of soil types
was found within both the estuarine and older marine deposits, the soil
samples were classified as either cohesive or cohesionless, depending
on whether the predominant soil type was clay or sand, for simplicity.
The cohesionless soils were assumed to be free-draining within the
stone column environment, and were tested only in drained triaxial
compression. The cohesive soil was assumed not to be completely free-
draining during the short duration of a load test, so both consolidated
drained and undrained triaxial tests were performed. The undrained test
213

J. Geotech. Engrg. 1985.111:205-223.


U. S. S T A N DA R D SIEVE SIZE

o
o s ? S 1 1
5 d c(
100 s E Z z
C9
g i
N : ' '

"^ :
:
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

i f 80
1:
t r Pi

3=70
>- :
m
60 ; :
j:
EC
Uj50
; i
;
H 40 |:| ; t? ^
F V * '.'?>
' - ^

UJ
H r &.
,"'vl ^ ' 1* !
..'A
:?; ^ P
llj
;gg
,
3 !^ * PI
V'<:
-' ?"- '<^;;:^22J
:
"pss^.i
3C 0 1001 : 10 : 1.0 ' 0.1 i 0.01 0.0133
| GRAIN-SIZE IN MILLIMETERS:

i GRAVEL i SAND
CO SILT
| COARSE j FINE ;C0ARSE: MEDIUM : FINE j

PARTICLE SIZE DISTRIBUTION

RANGE OF PARTICLE SIZES FOR STONE COLUMN GRAVEL


AS DELIVERED TO THE SITE

RANGE OF PARTICLE SIZES DETERMINED FOR CONSTRUCTED


STONE COLUMNS

FIG. 7.Stone Column Gravel Grain-Size Distribution

results were used to define soil strength and deformation parameters for
the short-term load test; whereas, the drained test results were used to
estimate long-term settlements under the structural loads.
Consolidated, drained triaxial compression tests were performed on
12-in. (305-m) diameter specimens reconstituted from the gravel used for
the stone columns. The angular to rounded gravel, which was obtained
from the alluvial valleys of the Santa Ynez and Ventura Rivers, had the
range of grain size distributions shown in Fig. 7. Shown also in Fig. 7
are the results of gradation tests done on samples taken from stone col-
umns after construction. The fines were concentrated in annulus at the
perimeter of the column. The strength and stress-deformation charac-
teristics of both gradations were comparable.
Space limitations preclude inclusion of the detailed test results herein;
however, the nonlinear behavior of the in-situ soils and stone column
gravels was approximated well by hyperbolic stress-strain and volume
change parameters determined using the methods of Duncan and Chang
(3). A summary of the properties and deformation parameters for each
of the five soil types is presented in Table 2. The soil parameters listed
in this table were obtained by averaging the results of several triaxial
compression tests for each soil. They are in agreement with parameters
for similar soils published by others (8,14).
214

J. Geotech. Engrg. 1985.111:205-223.


TABLE 2.Summary of Soil Properties and Deformation Parameters
Soi Properti 3S

Dry den-
sity, in Cohesion,
Soil pounds in pounds Friction Plas-
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

classifi- per cubic per square angle, in Liquid ticity


cation foot foot degrees limit index K n Rf G F d
(1) (2) (3) (4) (5) (6) (7) (8) 0) (10) (11) (12)
Drained Tests
Estuarine
cohesive
Estuarine
cohesion-
less 0.69 0.24 13.4
Older ma-
rine cohe-
sive 150
Older ma-
rine cohe-
sionless 190 0.90

Undrained Tests
Estuarine
cohesive
Older ma-
rine cohe-
sive 19 31 10 315 0.90
Drained Tests, Stone Column Material
Sandy
gravel 116 390 0.59

Note: Initial tangent modulus, ;: E,* - Kpa

(<n - 0-3)/
Failure ratio, Rf. Rf -
( i - *3)uit

Initial Poisson's ratio, v,: v,- = G F log

Tangent Poisson's ratio, v*: v; ;


d(ai - CT3)

Rfjai - g2)(l ~ sin <f>>'


E, 1
1c cos tj) + 2CT3 sin 4>
c = cohesion; ^3 = confining pressure;
p = atmospheric pressure; (ui - uty = compressive strength;
4> = friction angle; (oi - a3)ult = asymptotic value of stress difference.

In addition to the soil parameters listed in Table 2, a value for the at-
rest earth pressure coefficient, K0, was needed for the finite element
analysis. For the estuarine deposits between stone columns, and for the
stone columns themselves, K0 was assumed to equal 1.0. This relatively
high value was used in consideration of the vibro-compaction stone col-
umn construction process which increases lateral pressures above the
natural in-situ values which existed prior to construction. The at-rest
coefficient was taken as K0 = 0.5 for the older marine deposits, since
stone column construction did not affect these deposits.

215

J. Geotech. Engrg. 1985.111:205-223.


INDIVIDUAL LOAD TEST ANALYSES

Analytical versus Field Load Test Results.Calculated load versus


settlement curves for the three different stone column spacings are pre-
sented in Fig. 8. For comparison, the ranges of field load versus settle-
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

ment curves from Fig. 4 for these same three spacings are also shown.
The settlements calculated by the finite element analysis generally
overestimate the actual load test settlements somewhat; however, the
agreement is reasonably good. One reason could be that some of the
soil labeled as cohesionless for the finite element analysis, and therefore,
tested under drained conditions in the laboratory testing program, could
have contained enough silt or clay or both to prohibit it from draining
freely during the short-lived field load tests. Another reason could be
the load test procedure which provided for increasing loads once the
settlement rate under a given load had decreased to 0.010 in./hr (0.25
mm/hr). Had longer load increment durations been used, greater set-
tlements would have been recorded.
Settlements Without Stone Columns.To estimate the settlement of
an unimproved site, a finite element analysis was performed on a load
test model similar to the one depicted in Fig. 5, but without stone col-
umns. This was accomplished by changing the material properties of the
stone column elements to correspond to the unreinforced cohesive and
cohesionless native soil layers.
To simulate the native soil conditions existing prior to stone column
placement, a few changes were required in the soil parameters used in
the finite element analysis. Since the density of the cohesionless soils
present in the estuarine deposits increased with the installation of stone
columns, the cohesionless soils had less strength in the native condition
than they had after the stone columns had been placed. This lower
strength was modeled by lowering the modulus number, K, for the
cohesionless estuarine soil from 260 to 220. As there were no undis-
turbed soil samples obtained prior to stone column placement available
for laboratory testing, this reduction in the modulus number was esti-
mated based on a correlation with values for similar soil types presented

APPLIED PRESSURE (POUNDS/FOOTZ)


O IOOO 2O00 3000. __ IOOO_ _ 5000
1 1 1 1 1

^^
N
fc^ N. SS. ^ f H \ \ \ \ V > t > i z - ^

kl 0.2
- ^fii
v*^*^W\r N
Vsr--Nrfi \ \ \ \ \
,20 ft. per stone cotunn

^^5r\\\\\ //ZV// -

s
y^=^3^\\\

43 ft. per stone <^^JxE*EL


column \^zz
Hit^ III vAVA\
y/////&>L

% vvv/s^' '%wS
: , -

Y\ ^ww
W/^
0.4 1
33 fir per stone column
Predicted Curve
Predicted Settlement Under Design Load
0.6 i i . i 1 I i 1 I I

FIG. 8.Comparison of Measured and Predicted Load-Settlement Behavior of a


Single Loaded Stone Column Within a Group

216

J. Geotech. Engrg. 1985.111:205-223.


APPLIED FOOTING PRESSURE (POUNDS/FOOT2)
0 IOOO 2000 3000
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

FIG. 9.Predicted Load-Settlement Behavior for Load Test With and Without Stone
Columns

by others (8,14). Additionally, the at-rest lateral earth pressure coeffi-


cient, K0, was reduced from 1.0 to 0.5.
The load versus settlement curves calculated to simulate load tests per-
formed on both unimproved and stone column foundations are pre-
sented in Fig. 9. The stone column spacing conforms to the 5.75 ft by
5.75 ft stone column pattern. Settlements calculated for a load test on a
single column within a group of stone columns were approximately 30%
of the settlements calculated for the same loadings if applied to an un-
improved foundation.
Settlements with Increased Stone Column Strength.The hyperbolic

APPLIED FOOTING PRESSURE (POUNDS/FOOT 2 ) APPLIED FOUNDATION PRESSURE (POUNDS/FOOT2)


1000 2000 -0 DOO 2000 3000

MODULUS NUMBER, K, FOR


'STONE COLUMNS INCREASED '
y F R 0 M 260 TO 6 0 0

SETTLEMENTS CALCULATED FOR


UNIFORM LOADING ACROSS A
SETTLEMENTS CALCULATED FOR STONE COLUMN FOUNDATION WITH
A STONE COLUMN SPACING OF -
5.75 FEETX5.75 FEET SPACING PATTERN
5-75 FEETx 5.75 FEET

FIG. 10.Effect of Decreased Stone FIG. 11.Predicted Load-Settlement


Column Stiffness on Load-Settlement Curves for Uniform Loading of a Large
Behavior of a Single Loaded Stone Col- Area Stabilized With Stone Columns
umn in a Group

217

J. Geotech. Engrg. 1985.111:205-223.


strength parameters used to represent the properties of the constructed
stone columns determined directly from laboratory tests gave a modulus
number K, of 390 for the gravel. This value is somewhat lower than
typically obtained for other compacted gravels (8,14). To evaluate the
effect of a stiffer stone column gravel on settlements in a single column
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

load test, a finite element analysis was done using a K value for the
gravel of 600. The other parameters were kept the same.
The load versus settlement curve calculated with the higher stone col- I
umn modulus is shown on Fig. 10. As can be seen by examining Fig.
10, the effect of increasing the stiffness of the stone column gravel by a
little more than 50% (K increased from 390 to 600), resulted in a reduc-
tion in calculated settlements of approximately 10%. This suggests that
the results are relatively insensitive to the value of K used for the com-
pacted stone, which means that, in practice, it need not be measured to
great precision.

UNIFORM LOADING ANALYSES

Settlements.An estimation of ultimate settlements caused by uni-


form loading over a large area was obtained by performing two different
finite element analyses using the model for the 5.75 ft by 5.75 ft spacing
pattern and the standard soil parameters presented in Table 2. The load
versus settlement curves are presented in Fig. 11.
One analysis was made using undrained cohesive soil properties, and
the other assumed the cohesive soil to be drained. A comparison of the
two load versus settlement curves reveals that the use of drained co- i
hesive soil properties gives a 55%-60% greater settlement at any given j
load than the same analysis using undrained property values. Such be-
havior would be expected, and the difference between the two curves
would depend on the relative proportions of cohesive and cohesionless
soils in the profile. j
A comparison between the settlements calculated for a single column I
load test and the settlements for the uniform foundation loading con-
dition enabled the prediction of future building settlements from the re- \
suits of individual field load tests. Reference to Figs. 8 and 11 shows I
that for the 5.75 ft by 5.75 ft stone column pattern modeled with un-
drained cohesive soil properties, the settlements obtained from the uni- J
form loading analysis should be approximately 6-1/2 times greater than j
the settlements calculated for a load test on a single stone column. How-
ever, ultimate structural settlements are more accurately modeled with j
drained cohesive soil properties. Thus a more correct estimation of ul- '
timate structural settlement from the results of field load tests can be j
obtained by comparing the calculated settlement for the single column ;
load test modeled with undrained cohesive soil properties to the results |
of the uniform foundation loading analysis using drained cohesive soil i
properties. On this basis, the ratio of ultimate foundation settlement due ,
to loading of a large area compared to the settlement of a single load |
test is increased to approximately 10, a value consistent with both field ,'
experience (K. Engelhardt, personal communication) and predictions by [
other methods, as noted subsequently. i

218

J. Geotech. Engrg. 1985.111:205-223.


Settlement Reduction.The potential settlement in the absence of stone
columns was calculated by the finite element method for comparison
with settlements calculated for uniform loading. As with the corre-
sponding analysis for a single column load test, a modulus number of
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

220 was used for cohesionless soil in the estuarine deposits, and the at-
rest pressure coefficient was reduced from 1.0 to 0.5. The load versus
settlement curves calculated for uniform loading on the unimproved site
and for the improved stone column foundations are compared in Fig.
12. Both the curves shown were obtained using drained cohesive soil
properties. The settlements with stone columns equal approximately 30%
of the settlements calculated for uniform loading on the unimproved
site. The predicted settlements of uniformly loaded unimproved ground
using the finite element analysis agreed well with the settlement esti-
mates made using conventional settlement analysis methods.
Vertical Stress Distribution.The distribution of vertical stresses among
the stone columns and native soil on horizontal planes beneath the sur-
face was calculated for uniform surface loading on a stone column foun-
dation and drained soil properties. Within the stone column-native soil
system, the vertical stresses from uniform foundation loads are redis-
tributed approximately according to the ratio between the elastic moduli
of the stone column and of the native soil. The resultant ratio between
the vertical stress in the stone column and the vertical stress in the na-
tive soil is called the stress concentration ratio, n.
In the finite element model used in this analysis, the first layer of
elements located beneath the distribution blanket assumed cohesionless
soil. The stress concentration ratio in this layer is approximately 2. The
layer immediately beneath this was modeled as cohesive, and the stress
concentration here is approximately 3. The stress concentration ratios
did not vary appreciably with depth for a given soil type. The difference

APPLIED PRESSURE (POUNDS/FOOT2)


1000 2000 3000

15.0

FIG. 12.Predicted Foundation Settlements With and Without Stone Columns

219

J. Geotech. Engrg. 1985.111:205-223.


in stress distributions between the cohesionless and cohesive native soil
layers reflects the comparatively higher stiffness of the cohesionless soil.

MEASURED SETTLEMENTS
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

Settlement data were recorded during the period since completion of


the biological units structure at the plant in late 1976 through the end
of 1982. Values recorded over this period are shown in Fig. 13.
Measured settlements across the biological units structure vary from
a low of 0.4 in. (10 mm) at the northeast corner, where the soft sedi-
ments are thinnest, 15 ft (5 m), and foundation pressures are low, to
approximately 2.5-3.0 in. (65-75 mm) at the southwest corner where
the soft sediments are thickest, 45 ft (14 m). In the area of the biological
units where the soft sediment thickness is approximately 35 ft (11 m),
corresponding to the thickness of the estuarine deposits used in the fi-
nite element analysis, the observed settlements varied from approxi-
mately 1.0-2.4 in. (25-60 mm). The corresponding settlement for a large
loaded area calculated by the finite element analysis was approximately
2.5 in. (64 mm).
Overall, the measured settlements are somewhat less than the pre-
dicted values. However, the agreement is reasonable, since the actual
loadings are probably somewhat less than the design loads used for set-
tlement prediction. Additionally, the individual settlement points which
were used to record the periodic settlements were not established until
after completion of the structural skeleton.

STRUCTURAL SETTLEMENTS PREDICTED BY OTHER METHODS

The prediction of settlements for a stone column foundation support-


ing a large loaded area can be made by a variety of procedures. A num-
ber of these methods are summarized by Mitchell (9), Greenwood and
Kirsch (6), and Barksdale and Bachus (2). The application of some of
these methods to the biological units structure is described as follows.
Experience-Based Methods.Experience derived charts presented by
Greenwood (6), and reproduced by Mitchell (9) and Barksdale and Ba-
chus (2), present settlement of treated ground as a percentage of un-
treated ground settlement. A settlement reduction to about 30% of the

1.6 -'illI.I i.o w- 0.4


SECONDARY PRIMARY
SEDIMENTATION AERATION SEDIMENTATION
TANKS TANKS TANKS 1.0

2.6 1.9
l.0
40.8

"~| 2.4
1.8 < > RECORDED SETTLEMENTS
1976 TO DEC. 1982 (INCHES)

FIG. 13.Settlements Recorded at Santa Barbara Wastewater Treatment Plant


through 1982

220

J. Geotech. Engrg. 1985.111:205-223.


untreated ground value is predicted for the conditions at this site. Un-
treated ground settlements of 4 in. (100 mm)-14 in. (360 mm) were es-
timated, depending on the foundation loadings and thickness of soft
material. Thus, the predicted settlement of the improved foundation would
be in the range of 1 in. (30 mm)-4 in. (100 mm), depending on these
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

same factors.
As noted earlier, K. Engelhardt (personal communication) predicted
from observations on previous projects that the settlement of large loaded
area should be about 10 times greater than that measured under the
design load in a single column load test. As the single column load test
settlements were limited to 0.25 in. (6 mm), a settlement of 2.5 in. (64
mm) would be predicted. ,
Elastic Theory.A pile group/analysis was made based on the elastic
theory approach developed by Poulos (11). This method proceeds from
the condition of an incompressible pile in a homogeneous elastic me-
dium through a series of corrections to account for pile compressibility,
a firmer bearing stratum under the pile, consolidation of the soft soil,
and group effects. This approach led to the prediction that the settle-
ment of a large loaded area should be about 5 to 10 times greater than
that of a loaded single column in a group. When applied to single col-
umn load test results, this gives values of 1.2-2.5 in. (32 to 64 mm).
Reduced Stress Methods.Methods proposed by Priebe (12) and
Aboshi, et al. (1) [summarized by Mitchell, (9)], are based on the concept
that vertical stress concentrations on the stone columns give a reduced
average stress on the soft soil. Important parameters in estimates by
these methods are the stress concentration factor and the stone column
replacement ratio. From field measurement on several projects it has
been determined that the ratio of vertical stress in the stone to that in
the soft ground, is typically in the range of 2-6, with values of 3-4 usual.
Priebe and Aboshi, et al. also present methods for determination of this
factor based on column properties and spacing.
When these methods are applied to the Santa Barbara conditions, it
is found that the settlement of the treated ground should be about one-
third that of the untreated ground, by Priebe's method; whereas, the
Aboshi et al., method should predict 40-50% of the untreated ground
value, depending on the column spacing.
Summary.From the settlement predictions made by the various
methods, including the finite element analysis, it may be seen that the
variation among values predicted is quite small. On the average, it would
appear that a settlement of about 2.4 in. (60 mm) would be anticipated
for areas loaded to the full design value where the soft sediments are
of the order 32-35 ft (10-11 m) thick. This compares with actual values
of 1.0-2.4 in. (25-60 mm) at the site.

CONCLUSIONS

Axisymmetric finite element analyses of the vibro-replacement stone


column foundation installed at the site of the new Santa Barbara Waste-
water Treatment Plant have led to the following conclusions:
1. Load test settlements calculated by the finite element method for
initial settlement conditions, using undrained clay properties and drained
221

J. Geotech. Engrg. 1985.111:205-223.


properties for sand and stone columns, are somewhat higher than the
average settlements observed during actual field load tests conducted
on similar stone column spacing patterns. However, the overall results
obtained from the finite element analysis indicate reasonable agreement
between calculated and observed settlements for the individual load tests.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

Time limitation considerations indicate that complete settlement may not


have developed for all the load steps.
2. Settlements calculated by the finite element method for the stone
column foundation were approximately 30% of the settlements calcu-
lated for the same loading on an unimproved site for a 5.75 ft by 5.75
ft (1.75 by 1.75 m) stone column pattern.
3. The effect of increasing the stone column stiffness by about 50%,
to give modulus values in closer agreement with published data for coarse
gravel, resulted in a reduction of calculated settlements of approximately
10%. The fact that the modulus used to model the stone column gravel
has such a limited effect on the calculated settlement values indicates
that the accurate determination of its value is not necessary for esti-
mating structural settlements.
4. Ultimate long-term settlements calculated for uniform loading across
the foundation were approximately 10 times greater than the settlements
obtained from single simulated load test analyses.
5. The finite element analysis yielded computed stress concentration
ratios, i.e., the ratio of vertical stress carried by the stone column to that
of the soft ground between columns, at any depth of 2-3.
6. Measured settlements in the area of the biological units where the
soft sediment thickness corresponds to that used in the finite element
analysis, 32-35 ft (10-11 m), varied from approximately 1.0-2.4 in. (25-
60 mm). Although these values are somewhat less than the finite ele-
ment predicted value of approximately 2.5 in. (64 mm), the results are
reasonable considering that the actual loads are probably somewhat less
than the design loads; and the reference datums used to record the
structural settlements were established following construction of the bi-
ological units' structural skeleton.
7. Settlement predictions using several other, more simplified meth-
ods gave values that agreed reasonably well with both the finite element
predictions and the measured values. This lends support to the use of
the simpler methods in practice.

ACKNOWLEDGMENTS

The assistance and cooperation of many individuals and agencies as-


sociated with this study are gratefully acknowledged. The writers are
particularly indebted to Messrs. Joseph Gonzalez and Ivar Staal of Geo-
technical Consultants, Inc., who made available company files, records,
reports, and data from which this study developed. Special thanks are
also extended to Engineering Science, Inc., design engineers, and the
City of Santa Barbara for whom the geotechnical investigation and load
test program were performed and who gave approval for obtaining spe-
cial soil samples necessary for the completion of the laboratory testing
program.
222

J. Geotech. Engrg. 1985.111:205-223.


Finally, the writers are grateful for the assistance a n d suggestions of
the late Prof. Kenneth L. Lee u n d e r w h o s e direction this research pro-
gram was initiated.
Downloaded from ascelibrary.org by North Carolina State University on 12/17/12. Copyright ASCE. For personal use only; all rights reserved.

APPENDIX.REFERENCES

1. Aboshi, H., Ichimoto, E., Harada, K., and Emoki, M., "The Compozer: a
Method to Improve Characteristics of Soft Clays by Inclusion of Large Di-
ameter Sand Column," Collogue Inter, sur le Reinforcement des Sols,
E.N.P.C, Paris, France, 1974, pp. 211-216.
2. Barksdale, R. D., and Bachus, R. C , "Design and Construction of Stone Col-
umns, Volume I," U.S. Department of Transportation, Federal Highway
Administration Report No. FHWA/RD-83/026, 1983, p. 210.
3. Duncan, J. M., and Chang, C-Y., "Nonlinear Analysis of Stress and Strain
in Soils," Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 96,
No. SM5, Sept., 1970.
4. Engelhardt, K., and Golding, H. C , "Field Testing to Evaluate Stone Col-
umn Performance in a Seismic Area," Geotechnique, Vol. 25, No. 1, Mar.,
1975.
5. Greenwood, D. A., "Mechanical Improvement of Soils Below Ground Sur-
face," Proceedings of Conference on Ground Engineering, Institute of Civil En-
gineers, London, 1970, pp. 11-12.
6. Greenwood, D. A., and Kirsch, K., "Specialist Ground Treatment by Vibra-
tory and Dynamic MethodsState of the Art," Advances in Piling and Ground
Treatment for Foundations, Institution of Civil Engineers, London, England,
1983.
7. Huber, T. R., "Finite Element Analysis of the Load-Deformation Behavior of
A Vibro-Replacement Stone Column Foundation," thesis presented to the
University of California, at Los Angeles, Calif., in 1978, in partial fulfillment
of the requirements for the degree of Master of Science in Engineering.
8. Kulhawy, F. H., Duncan, J. M., and Seed, H. B., "Finite Element Analysis
of Stresses and Movements in Embankments During Construction,' Report
No. TE 69-4, Office of Research Services, University of California, Berkeley,
Calif., 1969.
9. Mitchell, J. K., "Soil Improvement," State of the Art Report, Proceedings of
the Tenth International Conference on Soil Mechanics and Foundation Engineering,
Stockholm, Sweden, Vol. 4, 1981, pp. 509-565.
10. Mitchell, J. K., and Huber, T. R., "Stone Column Foundations for a Waste-
water Treatment PlantA Case History," Geotechnical Engineering, Vol. 14,
1983, pp. 165-185.
11. Poulos, H. G., "Load Settlement Predictions for Piles and Piers," Journal of
the Soil Mechanics and Foundation Division, ASCE, Vol. 98, No. SMD, 1972,
pp. 879-897.
12. Priebe, H., "Abschatung des Stezungoverhaltens eares durch Stopfver-
dichrung vergessenten Baugrundis," Die Bautechnik, 53, H.5.S., 1976, pp. 160-
162.
13. Staal, I., and Engelhardt, K., "Discussion, Ground Treatment by Deep Com-
paction," Institution of Civil Engineers, London, England, 1976, pp. 128-
133.
14. Wong, K. S., and Duncan J. M., "Hyperbolic Stress-Strain Parameters for
Nonlinear Finite Element Analyses of Stress and Movements in Soil Masses,"
Department of Civil Engineering, Institute of Transportation and Traffic En-
gineering, University of California at Berkeley, July, 1974.

223

J. Geotech. Engrg. 1985.111:205-223.

You might also like