Watts, K. S., Johnson, D., Wood, L. A. & Saadi, A. (2000). GeÂotechnique 50, No.

6, 699±708

An instrumented trial of vibro ground treatment supporting strip foundations

in a variable ®ll
K . S . WAT T S  , D. J O H N S O N y , L . A . WO O D { a n d A . S A A D I §

Full-scale instrumented load tests have been carried out to Des essais de charge entieÁrement instrumenteÂs ont eÂte effec-
study the installation and performance of vibro stone col- tueÂs pour eÂtudier l'installation et la performance des co-
umns supporting a strip foundation in a variable ®ll and the lonnes de pierre vibro soutenant des fondations bande dans
performance of a similar strip foundation on untreated un remblai variable et la performance de fondations bande
ground. The results from site investigation and laboratory similaires sur un sol non traiteÂ. Nous preÂsentons les reÂsultats
testing are presented, and the vibro design principlesÐin de l'enqueÃte sur le site et des essais en laboratoire et nous
particular the use of a modi®cation to a standard formula deÂcrivons les principes de conception vibro, en particulier,
that re¯ects the radial densi®cation of soil around the stone l'utilisation d'une variante de la formule standard qui re¯eÁte
columnÐare outlined. The treated foundation performance la densi®cation radiale du sol autour de la colonne de pierre.
is compared with predicted behaviour; calculated settlements La performance des fondations traiteÂes est compareÂe au
are in reasonable agreement with measured values, but comportement preÂvu; les tassements calculeÂs preÂsentent une
stress measurements suggest that current design assumptions assez bonne concordance avec les valeurs mesureÂes mais les
signi®cantly overestimate the degree of stress concentration valeurs de contrainte suggeÁrent que les suppositions concep-
on the stone columns. The degree of radial densi®cation of tuelles actuelles surestiment consideÂrablement le degre de
®ll around the stone columns varied according to installation concentration de la contrainte sur les colonnes de pierre. Le
technique and soil type. The implications of the trial ob- degre de densi®cation radiale du remblai autour des co-
servations for existing design methods are considered. lonnes de pierre varie en fonction de la technique d'installa-
tion et du type de sol. Nous consideÂrons les implications des
KEYWORDS: design; ®eld instrumentation; full-scale tests; ground observations expeÂrimentales pour les meÂthodes conceptuelles
improvement; settlement; soil structure interaction. existantes.

INTRODUCTION examine the design principles adopted, in particular the use of a

The ground treatment techniques most commonly adopted in modi®cation to a standard formula that re¯ects the radial
the UK are the various deep vibro processes collectively de- densi®cation of soil around the stone column, and to present
scribed as vibro, which involve the construction of stone observations from the ®rst phase of the trial to load 250 mm
columns in the ground using a powerful torpedo-shaped vibrat- thick foundations. The design and behaviour of the strip founda-
ing poker (St John et al., 1989). Much of this work is carried tions after stiffening and reloading have been described and
out to treat shallow variable ®lls, and the introduction of stone analysed by Wood et al. (1996). Site investigation activities,
columns is designed to stiffen the existing ®ll and create more instrumentation, ground treatment, foundation construction,
consistent foundation conditions, thus reducing total and differ- loading and monitoring were carried out during the period
ential settlement. March 1990 to June 1991. Further site investigations were
It is dif®cult to predict the behaviour of foundations on carried out between January and May 1995. The layout of the
variable ®lls in which vibrated stone columns have been in- trial foundations, site investigation activities and instrumentation
stalled. The interaction of the modi®ed ®ll, stone columns and is shown in Figs 1 and 2.
the foundation is complex, and considerable reliance is placed
on experience of similar applications. Current design procedures
adopt a relatively simpli®ed view of this complex interactive SITE INVESTIGATION
system, and the purpose of this paper is to compare a typical An initial site investigation comprising ®ve trial pits and two
design with detailed ®eld observations. The distribution of stress boreholes identi®ed ash overlying cohesive ®ll to depths be-
between stone columns and intervening ground beneath the tween 3 m and 5 m. There was stiff glacial till below the ®ll.
foundation is critical for the prediction of foundation perform- Further boreholes drilled for instrumentation included SPTs and
ance, and is calculated from the relationship between the material sampling. Dynamic probing (DP), described in DIN
stiffness of the stone column and the surrounding soil. In 4094 (1974) and BS 1377: Part 9 (1990), and geophysics were
practice this is dif®cult to assess with con®dence. A ®eld trial carried out during ground treatment and the foundation con-
to study the installation and subsequent performance of vibro struction phases of the trial. The site investigation data from the
stone columns in a variable ®ll was carried out by Bauer immediate vicinity of the treated and untreated test foundations
Foundations Ltd., the Building Research Establishment Ltd are summarized in Figs 3 and 4 respectively, and illustrate the
(BRE) and South Bank University to substantiate the ®ndings of variability of the ®ll.
a study carried out by Wood (1990a) utilizing ®nite element DP tests carried out at several positions on line X±X
methods to model strip foundation behaviour on vibrated stone indicated 3±4 m depth of ®ll close to Trial Pit 4, and a survey
columns. A paper outlining the ®eld trial was published by along Y±Y indicated similar ground conditions. DP tests along
Watts et al. (1992). The objective of the current paper is to the line Z±Z suggested variable ®ll between 5 m and 6 m in
depth. Geophysical testing using continuous surface shear Ray-
leigh waves (Abbiss, 1981) was carried out at two locations
Manuscript received 24 May 2000; revised manuscript accepted 7 along line X±X and one location on Y±Y, and enabled dynamic
September 2000. soil properties to be calculated as a function of depth. There is
Discussion on this paper closes 22 February 2001, for further details see
p. ii. broad agreement between borehole logs and DP tests regarding
 Building Research Establishment, UK. the depth of the ®ll/natural clay interface, and geophysics also
y
Roger Bullivant Ltd. helped in identifying soil strata boundaries. Subsequent vibro
{
South Bank University. poker penetration provided a further indicator of depth to the
§
Formerly South Bank University. ®rm clay. A laboratory testing programme was carried out at
699

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 700 WATTS, JOHNSON, WOOD AND SAADI

BH3
9
TP4 BH5 TP3
X X
6
8 10 1 2 3 4 5 7

BH1 BH2
BH4
Mill access road Y Y

TP1

Z Z

15
18
Scale: m 17
0 10 16
Stone columns: 1–14 dry, 15–18 wet installation
6
Borehole
Dynamic probing
Geophysics

Test foundation strip

Extent of excavation
Site boundary
Edge of roadway
Trial pit
N TP5

Fig. 1. Plan of trial site

East DPH DPH West

N100 N100
TP4 0 10 0 10 TP3
0
BH5
Loose Ash, gravel, Loose
1
ash boulders ash
fill 6 fill
2 fill
Depth below OGL: m

Soft Soft clay,

3 silt 13 boulders
fill fill
4 GWL
8
5 Firm, silty
SPT 'N'

glacial
6 fill

7
Boundaries of granular fill/soft fill/
8 Scale: m
natural till deduced from geophysics
0 2
Base of fill indicated by BH1 and BH2
Base of stone columns
Section on X–X

Fig. 3. Fill pro®le along line of treated foundation

boulders. The granular ®ll also contained pockets of soft silty

clay. The lower cohesive ®ll comprised silty clay with dispersed
Fig. 2. Instrument layout granular fragments. Particle size analyses were carried out on
most of the samples, and the results are plotted in Fig. 5
together with the ranges of soil particle size generally accepted
South Bank University to categorize the ®ll and underlying as suitable for treatment by vibro methods (NHBC, 1995). The
strata. Engineering parameters for the ®ll and natural soil were organic content of the ®ll ranged from 1´4% to 4%. Classi®ca-
obtained to con®rm design assumptions and predict foundation tion tests on the cohesive ®ll indicated inorganic clay of
behaviour. The granular ®ll material contained a high proportion medium plasticity. The soil properties obtained from the ®eld
of black ash, some small pieces of sandstone and limestone, and laboratory investigations are listed in Table 1 with addi-
slate, burnt slag, clinker, brick and concrete fragments, sand tional properties used to back-analyse treatment performance
and gravel and broken sandstone, and sandstone cobbles and shown in Fig. 6.

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 VIBRO GROUND TREATMENT SUPPORTING STRIP FOUNDATIONS 701
East DPH DPH West
N100 N100
0 0 10 0 10
BH1 BH4 BH2
Ash,
1 boulders 16 Ash, cinder, 6 Ash, gravel,
fill brick, stone 4 sandstone
2 Soft sandy fill boulders
clay, 2 Soft clay, fill
Depth below OGL: m

3 fill stones,
brick GWL 82 Sandstone

SPT 'N'
4 fill boulders
fill
Firm 64
5 clay Firm, silty
glacial
Silty,sandy
6 fill
clay
7 15
Silty Firm, silty
8 laminated clay
clay
9 Weathered
Weathered sandstone
10 sandstone

Boundaries of granular fill/soft fill/ Scale: m

0 2
natural till deduced from geophysics

Section on Y–Y

Fig. 4. Fill pro®le for the untreated foundation

TREATMENT DESIGN
A preliminary design, of the kind representative of current
practice, was based on the type of limited site investigation
information that might ordinarily be available when preparing a
tender for a small-scale project. Only data from borehole 1 and
the trial pits excavated as part of the original investigation of
the site were used. The design consisted of: Fig. 6. Soil properties used to analyse treatment performance

(a) an assessment of the ultimate capacity of an individual

stone column, and hence the factor of safety against At the trial site the water table was situated at the base of the
bulging failure of that stone column ®ll. The total pressure on the top of the column can then be
(b) an assessment of column spacing written as
(c) a prediction of the settlement of the loaded composite stone
column/soil system. Pc ˆ K pc (K 0 ãs h ‡ 4cu ) ÿ ãc :h (2)
In predominantly granular soils, such as the ash ®ll at horizon 1
Stone columns are designed on the assumption that they are on this site, the current vibro design practice assumes cu equal
acted upon by a triaxial stress system and are assumed to be in to zero, and hence
a state of shear yield at a critical depth h. Many authors have
proposed approximations for the stress system. However, the Pc ˆ K pc (K 0 ãs h) ÿ ãc :h (3)
approach given by Hughes & Withers (1974) for cohesive soils In granular soils, the assumption that horizontal stresses are
is often adopted. Their operating equation can be rewritten as governed by K 0 substantially underestimates the strength of the
ó v9c ˆ Kpc (K0 [ãs h ÿ uso ] ‡ uso ÿ us ‡ 4cu ) (1) column. If the soil is free draining then the resistance to radial
mm
1·18

3·35

37·5
µm

150

212

300
425

600

6·3
63

10

14

20
28

50
63
75
2

100

90

80
Suitable for
70 stone column
Percentage passing

method
60
Suitable for
50 in situ
densification
40
Ash
30
fill Cohesive fill
20

10

0
µm 2 6 20 60 200 600 2 6 20 60 200 mm
Fine Medium Coarse Fine Medium Coarse Fine Medium Coarse
Clay Cobbles
Silt Sand Gravel

Fig. 5. Particle size analysis of the granular and cohesive ®ll

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 702 WATTS, JOHNSON, WOOD AND SAADI
expansion will be increased by column formation, and in applied foundation load of 123 kN=m2 for column spacing of
granular soils there will be a zone of improved ground around 1´8 m. The calculated settlement pro®le for the untreated strip
each stone column. Thus the use of K 0 is inappropriate in the was based on a Boussinesq stress distribution for a strip
situation after column construction and prior to loading the foundation. In clay soils settlement was calculated using the
stone column. This is recognised by Baumann & Bauer (1974) formula ó v mv H, where ó v is the average stress in the layer, mv
with the use of a coef®cient K s that has a value between the at- is the coef®cient of volume compressibility for the soil (Table
rest and passive coef®cients. 1), and H is the layer thickness, while in the granular layer
In equation 3, K pc is calculated using the average angle of (horizon 1 in Fig. 7) the formula (ó v H)=Es was used. Using
internal friction for the stone column material and improved the stiffness and compressibility parameters shown in Fig. 7 this
zone or annulus around the stone column, and is then denoted indicates up to 40 mm settlement under these loading condi-
by K p(ave). If Dc ˆ stone column diameter, Di ˆ diameter of the tions. The same Boussinesq stress distribution was used to
improved zone, then Di ˆ n 3 Dc where n ˆ diameter ratio. A estimate settlement beneath the treated strip. In order to calcu-
typical value for n of 1´5 has been used in this design. Once late the post-treatment settlement in the granular ®ll an appro-
ultimate stone column capacity has been calculated, the proce- priate average modulus of deformation, EAve , for the stone
dure developed by Baumann & Bauer is the commonly adopted column/improved annulus/unmodi®ed soil system in the granu-
method used to determine the distribution of stress between the lar ®ll was calculated from
stone columns and the surrounding soil. In this design the EAve 3 A0 ˆ (Ei 3 Ai ) ‡ (Es 3 As ) (5)
possible bene®t of a zone of improved ground around the stone
columns in granular soils has been considered. Accordingly Settlement within the granular layer was then calculated using
modi®cations have been introduced into the original Baumann the formula (ó v H)=EAve . Within the treated clay layer (Horizon
and Bauer formula such that it is now 2) settlement without treatment was calculated as described
    above using the parameters shown in Fig. 7, and a settlement
Es a
1‡2 (K s )ln reduction factor after Priebe (1995) was applied to calculate
Pi Ei r post-treatment settlements. Settlement beneath the treated depth
ˆ     (4)
Ps Es a (Horizon 3) is the same as for the untreated strip. In Fig. 7 the
2 (K i )ln
Ei r predicted settlement for the treated strip is based on immediate
settlement for the granular ®ll layer and immediate and primary
In Fig. 7 the model and parameters for the soil and stone consolidation settlement in the underlying clay horizons.
column adopted in the preliminary design process are shown.
Based on these values, ó v9c for a column is 340 kN=m2 . The
Baumann & Bauer analysis gives Pi and Ps as 248 kN=m2 and GROUND TREATMENT
12 kN=m2 respectively and thus a ratio of 21. The factor of Initial treatment was carried out using the dry top-feed
safety against bulging failure, ó v9c =Pi , is 1´4 at the maximum `vibrodisplacement' technique using a hydraulically operated

Fig. 7. Soil model and predicted settlements

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 VIBRO GROUND TREATMENT SUPPORTING STRIP FOUNDATIONS 703
Table 1. Principal soil parameters and test procedures
Parameter Soil type Test method

Ash ®ll Cohesive ®ll Natural clay Stone column

(0±2´25 m) (2´25±3´75 m) (Till)
Bulk unit weight: 16´0 20´6 21´2 13´0 Cohesive samples extruded from
kN=m3 U100 tubes. Ash ®ll sample density
range estimated from recompacting
Water content: % Ave ˆ 18:0 22´6 21´6 0 disturbed material for shear box tests
(see below)
Dry unit weight: 13´1 16´8 17´4 13´0
kN=m3
Plastic limit: % ± 20 19 ± BS 1377: Part 2: 1990, Test 5.3
Liquid limit: % ± 37 38 ± BS 1377: Part 2 : 1990, Test 4.3
PI: % ± 17 19 ±
Organic content: % 4´1 1´4 ± ± BS 1377: Part 3: 1990, Test 3
Laboratory vane tests on U100
Undrained shear 18±78 75±80 samples
strength: kN=m2 ±
(Ave ˆ 40) Hand vane testing in-situ below
treated foundation
Effective cohesion, C9 ˆ 5 C9 ˆ 0 Multi-stage drained triaxial
C9: kN=m2 ö9 ˆ 26 ö9 ˆ 28

Shear strength, Loose: ö9 ˆ 36 Loose: ö9 ˆ 45 300 mm drained shear box tests

ö9: degrees Dense: ö9 ˆ 46 Dense: ö9 ˆ 50
4´3 at 0´0 m Secant modulus at depths stated, based Hydraulic (Rowe) cell BS 1377:
7´2 at 1´5 m on trend lines plotted through results Part 6: 1990, Test 3
from 5 tests on ash ®ll (horizon 1) and
Constrained 4´2 at 1´6 m 13 tests on clay ®ll (horizon 2) and 75 mm samples from U100 tubes
Modulus, D: MPa 5´4 at 2´8 m natural clay (horizon 3) in Fig. 7 tested in one-dimensional
consolidation apparatus BS 1377: Part
5´4 at 2´9 m 5: 1990, Test 3
6´9 at 5´0 m

vibrating poker with a maximum power output of 86 kW and a INSTRUMENTATION, FOUNDATION CONSTRUCTION AND
horizontal force of 25 t. Compressed air was used as the jetting LOADING
medium. In accordance with normal practice (BRE, 2000), the Details of foundation construction and instrumentation have
vibrator penetrated to the design depth and was then fully been described by Watts et al. (1992), and only a brief
withdrawn to allow a charge of stone to be placed in the hole. summary follows. Precise levelling, as described by Cheney
The poker was replaced in the hole and the charge of stone (1973), was used to obtain a record of total foundation move-
compacted to a dense state. This process was repeated with ment. Six push-in spade-shaped pressure cells were installed
stone fed in small charges and the column was compacted in from small-diameter boreholes to measure changes in lateral
stages back to the ground surface. Initially, three isolated (radial) total earth pressure due to column installation and
columns (8, 9 and 10 in Fig. 1) were installed to establish an foundation loading, and four more were installed after founda-
appropriate treatment procedure for the trial footings. The tion construction by jacking horizontally from a slit trench
degree to which lateral densi®cation of soil takes place during between the trial foundations to measure vertical stress immedi-
penetration of the poker and compaction of stone is dependent ately beneath the strips in the ®ll between columns. Magnet
on particle grading, initial soil density, the mechanical speci®ca- extensometers were installed into the underlying stiff clay in
tion of the poker and workmanship. It is usually not desirable boreholes 4 and 5 to measure settlement with depth midway
to operate vibro equipment to the limit of its capacity. In between columns 2 and 3 of the treated foundation strip and in
granular soils this can produce a column with a higher load- a similar position under the untreated strip. Pneumatically
carrying capability than required, while in cohesive soils over- operated 0´3 m diameter ¯at jack pressure cells (Wood, 1990b)
working may reduce soil strength. For the purposes of this trial, were installed in the top of column 2 and column 3 to measure
column 8 was installed to the maximum capability of the poker the load carried by the stone columns under the treated founda-
at each stage of the column construction. Columns 9 and 10 tion. Similar 0´2 m diameter cells were installed in the ®ll below
were installed in accordance with generally adopted practice for each end of the two foundations. The foundations were cast in-
these ground conditions. Five columns (1±5) were similarly situ after excavation and blinding a prepared surface in accor-
installed at 1´8 m centres along the centreline of the treated dance with normal building practice. Mesh reinforcement
foundation strip as shown. Columns 6 and 7 were installed to (B283) with a total cross-sectional area of 216 mm2 was
densify ®ll replaced in trial pits that formed part of the original incorporated into the bottom element of the foundation. The
site investigation. In January 1992 four additional columns (15± footings were cast in shuttering that was reused to thicken the
18) were installed using the wet top-feed method. The wet foundations to study the effect of changes in foundation stiff-
`vibroreplacement' technique uses water jets to maintain a ness on the bending strains induced in the strip foundations
circulating ¯ow in the cylindrical void to remove loose material (Wood et al., 1996). Load was applied using kentledge compris-
and maintain stability while stone is compacted to form the ing 1´2 m concrete cubes, each weighing an average 4 t, and
column. This technique minimizes disturbance of the in-situ with the bottom row supported at intervals on steel spreader
soils. The purpose of these additional columns was to compare beams placed across the strip so as not to increase foundation
the effect of installing stone columns using the wet and dry stiffness. Each load increment comprising a row of blocks
methods in this ®ll. applied a total bearing pressure of 41 kN=m2 to the foundation

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 704 WATTS, JOHNSON, WOOD AND SAADI
strips. A large-capacity crane was utilized, with adequate reach Increase in horizontal earth pressure at depth of cell: kN/m2
to avoid any effect on the foundations or supporting soil. 0 10 20 30 40 50 60
Immediately after applying the ®rst load increment levels were 0
taken on both strips and the instrumentation was monitored Cells G4 G1
before a second layer of blocks was added, increasing the Distance from
0·9 1·5
foundation pressure to 82 kN=m2 . Both foundation strips were 0·5
column centre: m
then monitored at increasing time intervals for a period of one Change in stress:

Depth of poker tip below original ground level during treatment: m

month to observe creep settlement, after which the total average poker penetration
Change in stress:
applied foundation pressure was increased to 123 kN=m2 . The column compaction
foundations were monitored for a further 6±7 months before the 1·0
®rst phase of the trial was concluded by fully unloading both
strips.
1·5
DISCUSSION: FILL BEHAVIOUR AND FOUNDATION
PERFORMANCE
Soil response during treatment 2·0
Pressure cells at 0´9 and 1´5 m from the column axis and at
depths of approximately 0´4 m, 0´9 m and 1´8 m below founda-
tion level measured changes in lateral (radial) total earth Depth of cells
pressure during initial poker penetration and subsequent com- 2·5
paction of the stone column material. In Fig. 8 the increase in
total earth pressure measured by cells G1 and G4 during
treatment is plotted against poker depth, during both initial
3·0
penetration and subsequent stone compaction back to the
ground surface. During initial penetration no stress increase was
measured until the poker reached the level of the cells, but up
to 60 kN=m2 was registered by cell G4 during further penetra- 3·5
tion to design depth. The diminishing effect with increasing
distance from the column centre is indicated by cell G1. With-
drawing the poker to place charges of stone resulted in an
almost immediate return to the pre-penetration stress levels, but 4·0
similar elevated values were again measured as the stone was
Fig. 8. Lateral stress changes due to poker penetration and stone
compacted. The increases in earth pressures disappeared imme-
compaction
diately upon completion of the column, including those meas-
ured by the deepest cells G1 and G4, which may have been
located in more cohesive ®ll. Each column took an average
12 min to complete. In contrast, it took 27 min to complete the were measured at a distance equivalent to 2´5 times the column
high-energy column 8. Pore water pressure measurements were diameter from the centre of column 2. In granular soils densi®-
not made in the ®ll, but particle size analyses, which indicate cation would be expected around the periphery of the stone
essentially granular soils at the level of the pressure cells, columns owing to the rearrangement of soil particles in re-
combined with the rapid dissipation of excess pressure observed sponse to the vibratory action of the poker and stone compac-
following pressure cell installations, suggest a high permeability tion. In order to investigate this, dynamic probing was carried
for the ®ll at the level of these cells. out at a range of distances around stone columns. In Fig. 9
The measured increase in stress resulting from radial displa- dynamic probings at increasing distances from test column 9
cement of the granular soil, although not sustained, would result indicate a signi®cant increase in blow counts close to the edge
in some drained loading, compression and hence densi®cation of the column, with a more modest increase at 0´6 m from the
of the soil surrounding the columns. Modest stress increases column centre. Greater increases in blow count are apparent in

Blows/0·1 m Blows/0·1 m
0 5 0 5 10 0 5 10 15
0 0
Column 9 Column dia. = 0·7 m Column 16 Column 8
Column dia. = 0·8m
Column dia. = 0·7 m
Granular
1 fill 1
Granular Granular
fill fill
Cohesive
2 fill 2
Cohesive
Depth: m

Depth: m

fill Cohesive fill

3 Clay 3
Base of fill Base of fill

Firm
4 Firm 4
clay
clay

5 5
Pre-treatment (5) Untreated Pre-treatment (5)
0·58 m from centre of column 0·6 m from centre of column 1·0 m from centre of column
Close to edge of column Close to edge of column Close to edge of column
6 6
(a) (b)

Fig. 9. Radial densi®cation of ®ll around stone columns: dynamic probing

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 VIBRO GROUND TREATMENT SUPPORTING STRIP FOUNDATIONS 705
granular ®ll, but signi®cant increases above pre-treatment values 0 1 2 3 4
m
5 6 7 8 9
were also measured in the more cohesive ®ll. Much larger 0
increases, averaging ten times pre-treatment values, were meas- 41 kPa
5
ured in the granular ®ll immediately surrounding test column 8,

Settlement: mm
which received more compactive energy during treatment and 10 82 kPa
resulted in a larger column diameter. An average ®ve-fold
increase was measured in the granular ®ll at a distance of 1´25 15
times the column diameter from the column centre. By contrast, 20 123 kPa
probing around column 16, installed using the wet top-feed 25
Treated strip
process, which is designed to minimize disturbance of the kPa
Untreated strip
Foundation pressure
surrounding soil, showed no change in resistance in the sur- 30
East West
rounding soil. 0
These observations of changes due to column installation in Loose
the granular soils surrounding the columns, in particular the ash,
gravel,

Depth below foundation: m

direct measurement of increased penetration resistance, con®rm 1 boulders
densi®cation to at least 0´6 m (1:7 3 ro ) from the column
fill

centre, and suggest that the use of a diameter ratio n ˆ 1:5 in 2

Soft clay,
brick,
equation (6) is appropriate. No instrumentation was installed 1 2 boulders 3 4 5
fill
around column 8 to measure soil response during treatment, but
3
a much greater effect in terms of post-treatment penetration
resistance was measured at a distance equivalent to n ˆ 2:5. Stone columns
The increased stress measurements at cell G1 outside column 2 4
Firm silty clay
would translate into n ˆ 5:0. While it would not be prudent to
use this very high ®gure, the adopted value of n ˆ 1:5 is 5
judged to be amply justi®ed in these soil conditions. Both types
of measurement con®rm the effect of radial modi®cation of the Fig. 11. Differential settlement along foundation strips
surrounding granular soil as a result of column installation.
These observations suggest some validity in the use of K s rather
than K 0 for the calculation of the total pressure in the top of
the column for granular soils. The bene®ts of the stone column treatment in this case are
evident, but it is useful to examine the distribution of settlement
with depth for the treated and untreated conditions. Results
Foundation and soil settlement obtained from the borehole extensometers positioned under the
The maximum settlement of the 250 mm thick strip founda- centre of each strip are given in Fig. 12 for the two average
tions on treated and untreated ground is shown in Fig. 10, loading conditions of 82 kN=m2 and 123 kN=m2 . As in Fig. 11,
plotted against time since load was applied. The immediate it is apparent that the effect of the stone columns in signi®-
settlement resulting from the ®rst foundation pressure of cantly limiting surface settlement did not occur until the load
41 kN=m2 was similar for both strips. Increasing the foundation was increased to 123 kN=m2 . However, at that load the treated
pressure to 82 kN=m2 produced larger settlement of the un- foundation performed slightly better than predicted, and this
treated strip than the treated strip, and the rate of creep for the was due principally to lower compression in the upper granular
untreated strip was also greater after one month. The third material. This may be due to an underestimate of the diameter
foundation pressure of 123 kN=m2 induced much larger settle- ratio, n, and laboratory tests also indicate signi®cantly higher
ment of the untreated strip. At the end of a further 7 months' strength parameters for the ash ®ll and stone column material
monitoring, the maximum settlements of the treated and un-
treated strips were 16 mm and 26 mm respectively. Both founda-
tions exhibited sagging along their length. Fig. 11 shows the
development and distribution of settlement along each founda-
tion measured at the end of each loading period, with a long-
itudinal section indicating the average change in depth of ®ll
and the position and depth of the stone columns. The plot
shows greater differential settlement in the untreated strip after
the application of the highest foundation load and greater total
settlement where the ®ll is deepest.

Fig. 10. Maximum settlement along strip foundations Fig. 12. Settlement pro®le within foundation soils

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 706 WATTS, JOHNSON, WOOD AND SAADI
(Table 1) than those used to predict foundation performance cantly as load is applied. Hughes & Withers and Greenwood
(Fig. 7). Fig. 12 also shows that the presence of the columns draw the analogy between stone column behaviour and a
gives rise to much more deep-seated settlements occurring with- pressure meter.
in the cohesive ®ll and the underlying stiff clay below the base Consideration of all of the evidence from this trial would
of the columns. The percentage contribution to surface settle- tend to suggest that the deeper-seated settlements associated
ment resulting from the maximum applied load of 123 kN=m2 with the treated strip in this trial are due to the presence of the
for the treated and untreated foundations is given in Table 2. columns, which encourages load transfer to a greater depth: that
Examination of Fig. 12 also shows that there is an increase in is, concentrating the applied load within the column/soil mass
the compressibility of the cohesive ®ll, which may be due to and limiting the normal spread that has occurred under the
disturbance caused during column installation, load transfer untreated strip. When the observed settlements with depth are
down the stone columns, or a combination of both effects. Load compared with the predictions given in Fig. 7 several signi®cant
transfer down the stone columns is supported by the fact that points become apparent. First, for the untreated strip it is clear
there is some 3 mm of settlement observed in the stiff clay that the contribution to the surface settlement from the cohesive
beneath the base of the columns, which is absent for the ®ll has been overestimated, as illustrated in Fig. 13, and it is
untreated strip. Interpretation of the observed settlements is apparent the stiffness of this stratum has been underestimated.
further complicated by the apparent heave recorded at depth Second, for the treated strip the agreement between the ob-
below the untreated strip at an applied load of 82 kN=m2 but, served measurements and the prediction for n ˆ 1:5 is good.
as shown in Fig. 13, the pattern of settlement with depth is These comparisons are in contradiction to each other, and
similar even if the observations at an applied pressure of suggest one of two conclusions. Either the stiffness of the
82 kN=m2 are taken as the datum. cohesive ®ll is signi®cantly different under each of the strips,
The principal objective of vibro stone column treatment is to perhaps owing to the effects of the installation of the columns
reinforce the existing soil mass to produce a composite structure themselves, or the behaviour of the column/soil mass is differ-
that, overall, has load-carrying characteristics that are superior ent from that assumed in the method of settlement prediction,
to those of similar untreated ground. During the construction of albeit that the prediction of the surface settlement of the treated
columns stone is densely compacted and fully interlocked with strip was in good agreement with the observed value.
the surrounding soil. The mechanisms that govern stone column Although not conclusive it is felt that the balance of the
behaviour are well documented. Hughes & Withers (1974) evidence suggests that the presence of the columns leads to an
describe the similarities with pile behaviour in that, when increase in the contribution to the surface settlement from the
loaded, they settle and develop end bearing and cohesive resist- underlying deeper strata due to load transfer down the stone
ing stresses up the side. Unlike piles, they comprise compacted columns. It is thus important to distinguish between the use of
cohesionless granular material, which bulges under load and stone columns to increase the bearing capacity of poor soils
must be supported by lateral stresses exerted by the soil. from their use to limit settlements. Where stone columns fully
Greenwood (1991) goes on to explain the differing behaviour of penetrate soft and more compressible soils, and are toed into an
piles and stone columns arising from the hugely different ratio underlying stiffer stratum as in this trial, both the bearing
of their stiffness to that of the soil, and also points out that the capacity and the settlement performance of the foundation
relative stiffness of stone columns to soil can change signi®- should be improved. Where this is not the situation, as may
occur on thick deposits of soft soils when the stone columns are
limited to penetrating a partial thickness of the soft soil, the
Table 2. Contribution to surface settlement for treated and un-
bearing capacity may be improved but a signi®cant proportion
treated foundations of the predicted reduction in surface settlement may not occur.
The latter is dependent on the thickness of soft soil remaining
Strata Untreated: % Treated: % below the toe level of the stone columns and the load concen-
Granular ®ll 85 30 trating effect of the column/soil mass above. To ameliorate this
Cohesive ®ll 15 40 situation the depth of partial penetration treatment should be
Underlying clay 0 (5) 30 critically examined. Hughes & Withers (1974) proposed a
 There is an apparent movement of 1 mm, which is the limit of simple procedure for calculating the minimum length of stone
accuracy of the subsurface settlement measurement, and is judged to be column required to prevent end bearing failure at the toe of the
zero. column occurring before bulging failure at a critical depth near
the top of the column. This method can be used to estimate the
depth at which the vertical stress in the column is zero. Watts
& Serridge (2000) describe a ®eld trial of vibro stone columns
in soft clay soil in which the ef®cacy of this procedure is
examined and present some evidence to support the theory.

Stress distribution
It has generally been assumed that the ratio of stress in the
vibro stone columns to that in the intervening ground is at its
maximum when load is ®rst applied: that is, the columns
initially carry a high proportion of the total foundation load
(Greenwood, 1974). Greenwood (1991) also reported a maxi-
mum ratio of column to soil stress of 25 being measured in
very soft clay and a ratio of 4 measured in drained silt over
applied pressure ranges similar to this trial. In both cases ratios
were seen to decrease signi®cantly as the applied load in-
creased. In Fig. 14 stresses measured in the columns and the
intervening soil supporting the treated foundation strip are
plotted for the three load increments applied. Stresses measured
at similar locations under the untreated foundation strip are also
plotted for comparison. Changes occurring during periods of
constant applied load are indicated by arrows, and re¯ect a
redistribution of stresses that occurred while foundation settle-
Fig. 13. Percentage contribution to surface settlement ment took place. The ratio of average column stress to stress in

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 VIBRO GROUND TREATMENT SUPPORTING STRIP FOUNDATIONS 707
160

Changes over
six months
140
Average stress at top of columns 2 and 3
Average stress between columns
120
Average stress below untreated strip
Measured stress: kN/m2

100

80

Changes over
one month
60
Immediately after
load application

40

20 Fig. 15. Column/soil stress ratios with applied loads

0
0 20 40 60 80 100 120 140 160
locations were signi®cantly in excess of those calculated from
Average applied foundation pressure: kN/m2
normal stress distribution in an elastic medium, with the high-
est, as a proportion of the applied vertical load, being recorded
at 0´9 m from the column centre and 1´8 m below formation
Fig. 14. Vertical stresses beneath the foundation strips
level by cell G4. This indicates that the column may be bulging
at a depth equivalent to three times its diameter, just below the
the soil between columns is also plotted in Fig. 15, and shows a granular ash ®ll, and con®rms that there was signi®cant stress
consistent increase over the range of applied foundation loads, transfer down the column. This also con®rms the necessity,
as well as a signi®cant increase as foundation settlement took where cohesive soil is located at or close to the formation level,
place at constant load. On ®rst loading the ratio was about 0´5, for a check on ultimate column capacity using, for example, the
indicating that a high proportion of the applied load was carried method suggested by Hughes & Withers (1974).
by the intervening soil. At the highest applied load, the ratio
increased to a maximum value of 2´5. The increasing concentra-
tion of stress in the column resulted in the measured increase in CONCLUSIONS
vertical strain in the deeper soils (Fig. 12). (a) The trial has demonstrated the ability of vibro stone
Changes in lateral stress were also measured by the pressure columns to reduce total and differential settlement of
cells alongside the central columns during static loading of the conventional strip foundations constructed on a weak
foundations (Fig. 16). The stresses measured at all the cell variable ®ll.

Fig. 16. Lateral stresses measured at applied foundation load of 123 kN=m2

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.
 708 WATTS, JOHNSON, WOOD AND SAADI
(b) Predicted surface settlements for the treated strip are in Ki coef®cient of earth pressure for stone column/
close agreement with actual measured performance but improved annulus (modi®ed from Baumann &
settlement measurements with depth suggest that there was Bauer, 1974)
signi®cant compression in the lower clay soils, probably as K pc coef®cient of passive earth pressure for stone
column
a result of load transfer down the columns.
K p(ave) average coef®cient of passive earth pressure for
(c) Partial depth treatment of deep deposits of weak material stone column/improved annulus
should be critically examined to ensure settlements will be K 0 coef®cient of earth pressure at rest for unimproved
acceptable. soil
(d ) Stress measurements indicate a much lower proportion of P average imposed stress on foundation with area Ao
the applied load was carried by the stone columns than Pc average stress in stone column at foundation level
predicted by standard analysis. Pi average stress in stone column/improved annulus at
(e) The ratio of stress in the column/improved zone and foundation level
intervening soil, Pi =Ps rose with applied load and increased Ps average stress in unmodi®ed soil between columns
further as settlement of the foundations continued under (Baumann & Bauer, 1974)
ö9s angle of internal friction of unimproved soil
constant applied load. ö9c angle of internal friction of column material
( f ) Poker penetration and stone compaction affected the ground ö9a angle of internal friction of improved annulus
up to 1´5 m from the centre of the columns that supported ãs bulk unit weight of unimproved soil
the strip foundation. Densi®cation occurred in the granular ãc bulk unit weight of stone column material
®ll up to a distance equivalent to a diameter ratio, n ˆ 2:5 ó r9c radial effective stress in stone column
for normally constructed stone columns. ó ro initial radial total stress in soil prior to column
( g) A modi®ed Baumann and Bauer analysis, utilising a value construction
of n ˆ 1:5, gave a reasonable prediction of settlement of ó rs radial total stress on column boundary
ó v9c maximum vertical effective stress capacity of stone
the treated strip foundation but substantially overestimated
column
the stress ratio Pi =Ps .
(h) The radial effect of column installation was related to the
nature of the material, to the level of compaction (work-
REFERENCES
manship) and to the technique (dry or wet) employed. Abbiss, C. P. (1981). Shear wave measurements of the elasticity of the
ground. GeÂotechnique 31, No. 1, 91±104.
Baumann, V. and Bauer, G. E. A. (1974). The performance of founda-
tions on various soils stabilized by the vibro-compaction method.
ACKNOWLEDGEMENTS
Can. Geotech. Journal 3, No. 2, 509±530.
The authors appreciate the assistance provided by Hilary BRE (2000). Specifying vibro stone columns, Report BR391. Construc-
Skinner of BRE in the ®nal drafting of this paper. The BRE tion Research Communication, London.
research work formed part of a research programme for the British Standards Institution (1990). British standard methods of test for
Construction Directorate of the Department of the Environment, soils for civil engineering purposes: In-situ tests, BS 1377: Part 9.
Transport and the Regions. Bauer Foundations (UK) Ltd carried Milton Keynes: BSI.
out the ground treatment and were responsible for construction Cheney, J. E. (1973). Techniques and equipment using the surveyor's
of the foundations and logistical support. Further ®nancial level for accurate measurement of building movement. In Field
support was provided by the Faculty of the Built Environment, instrumentation in geotechnical engineering, pp. 85±99. London:
Butterworth.
South Bank University, London. Ahmed Saadi was in receipt of DIN (1974). Subsoil dynamic and static penetrometers: Dimensions of
a South Bank University research scholarship. Permission to apparatus and method of operation, DIN 4094: Part 1. Berlin:
carry out the trial was given by the site owners, Lancashire Deutsches Institut fuÈr Normung eV.
County Council. Greenwood, D. A. (1974). Vibro¯otation: rationale for design and prac-
tice. In Methods of treatment of unstable ground (ed. F. G. Bell),
London: Butterworth. 148±171.
Greenwood, D. A. (1991). Load tests on stone columns. In Deep founda-
NOTATIONp tion improvements: design, construction and testing, ASTM SPT 1089,
a (Ao =ð), equivalent radius of foundation area per 189±209. Philadelphia: American Society for Testing and Materials.
compaction Hughes, J. M. & Withers, N. J. (1974). Reinforcing of soft cohesive
cu undrained shear strength of soil surrounding the soils with stone columns. Ground Engineering, 7, No. 3, 42±49.
stone column National House Building Council (1995). Vibratory ground improvement
h `critical depth' at which bulging failure of stone techniques. NHBS Standards, Ch. 4.6. Amersham: NHBC.
column occurs NHBC (1988). In text but not in reference list.
mv coef®cient of volume compressibility Priebe, H. (1995). The design of vibro replacement. Ground Engineer-
n diameter ratio of improved zone to stone column ing, December, 31±37.
r radius of improved zone St John H. D., Hunt R. J. & Charles J. A. (1989). The use of `vibro'
ro original radius of stone column ground improvement techniques in the United Kingdom. BRE Infor-
u pore water pressure mation Paper IP 5/89. Garston: Building Research Establishment.
uso initial pore pressure in soil at depth h Watts, K. S. & Serridge, C. J. (2000). A trial of vibro bottom-feed stone
us pore pressure in soil at yield at depth h column treatment in soft clay soil. Proc. 4th Int. Conf. Ground
uc pore pressure in column at yield at depth h Improvement Geosystems, Helsinki, 549±556.
Ac area of column Watts, K. S., Saadi, A., Wood, L. A. & Johnson, D. (1992). A ®eld trial
Ai area of column/improved annulus to assess the design and performance of vibro ground treatment in
Ao total foundation area per stone column ®ll and strip foundations on vibrated stone columns, Proc. 2nd Int.
As area of soil between improved zones Conf. Polluted and Marginal Land, London, 215±221, Engineering
Dc stone column diameter Techniques Press.
Di diameter of improved annulus Wood, L. A. (1990a). The design of foundations on stone columns.
EAve: average modulus of deformation for column/ Proceedings of the international conference on construction on
improved annulus/unmodi®ed soil polluted and marginal land, (ed. M. C. Forde), Engineering Techni-
Ec modulus of deformation for stone column ques Press.
Ei modulus of deformation for stone column/ Wood, L. A. (1990b). The performance of a raft foundation in London
improved annulus clay. In Geotechnical instrumentation in practice, 323±340. London:
Es modulus of deformation for soil Thomas Telford.
H thickness of soil layer Wood, L. A., Johnson, D., Watts, K. S. & Saadi, A. (1996) Performance
K s coef®cient of earth pressure for unimproved soil of strip footings on ®ll materials reinforced by stone columns.
(after Baumann & Bauer, 1974) Struct. Engr 74, No. 16, 265±271.

Downloaded by [ Syracuse University] on [16/09/16]. Copyright © ICE Publishing, all rights reserved.