Materials and Structures/Mat6riaux et Constructions,Vol.

32, November1999, pp 679-686

Factorial design models for proportioning selfconsolidating concrete

K. H. Khayat, A. Ghezal and M. S. Hadriche
Universit~de Sherbrooke,Sherbrooke,QC, CanadaJ IK 2R1

Paperreceived:September29, 1998; Paperaccepted:April 19, 1999

ASS+RACT

RgSUME
Pour la formulation du b~ton autoplafant (BAP) plusieurs gdch~es s'imposent, ~tant donn~ qu'il faut ma~triser
tous lesfacteurs affectant les propri~t~s ~ l'~tatfrais et durci
du b~ton. Des modules statistiques ont ~t~ g~n&e~s~ partir
de la re~alisationd' un plan d'exp~rience. Ces modules identifient les param~tres importants de la formulation sur la
performance du b~ton autoplafant : la d~formabilit~ caract~ris~e par l'essai de l'~talement, les parametres rh~ologiques, la capacit~ de remplissage, et l'entonnoir; la stabilit~ traduite par le test du tassement et la r~sistance ~ la
compression. La mod~lisation a n~cessit~ un total de 32
gdch~es de b~ton. Neuf autres m~langes ont ~t~ ajout~s afin
de v~rifier la validation des modules ~tablis. Ce papier pr~sente les modules g~n&e's qui traduisent l'effet des param~tres principaux ainsi que leur interactions sur les
r~ponses mesur~es. L'utilit~ des modules ~ ~tablir une
meilleure comprehension entre les param~tres des m~langes
et de trouver des correlations entre les diff&ents tests r~alise's
est discut~e.

A factorial design was carried out to model the influence of key mixture parameters on properties affecting
the performance of self-consolidating concrete (SCC).
Such responses included slump flow and rheological
parameters, filling capacity and V-funnel flow to assess
restrained deformability, surface settlement to evaluate
stability after casting, and compressive strength. Thirty
two mixtures were prepared to derive the statistical
models and nine others to evaluate their accuracy. The
models are valid for a wide range of mixture proportioning. The paper presents the derived models that unable
the identification of underlying primary factors and their
interactions that influence the modelled responses of
interest for self-consolidating concrete. Such parameters
can be useful to reduce the test protocol needed for the
proportioning of self-consolidating concrete. The usefulness of the models to better understand trade-offs
between mixture parameters and compare the responses
obtained from various test methods are highlighted.

|
1. I N T R O D U C T I O N

Self-consolidating concrete (SCC) is a highly flowable concrete that can spread easily through restricted
sections under its own weight without segregation and
blockage. Such concrete is used to ensure the filling of
congested sections and areas with restricted access to
vibration. It is also employed to improve the productivity of concrete placement and site working conditions
resulting from noise reduction due to the elimination of
vibration consolidation.
The proportioning of SCC is complicated because of
the various contradictory requirements needed to ensure
excellent flow characteristics and proper mechanical
properties. For example, a highly flowable SCC should

1359-5997/99 9 1LILEM

679

have a relatively low yield value to ensure good deformability but an adequate resistance to segregation and
bleeding until the onset of hardening. An increase in
water-to-cementitious materials ratio (w/cm) can secure
high deformability, however, it can reduce the cohesiveness and cause segregation of aggregate that can lead to
blockage of the flow. Inter-particle friction between
coarse aggregate, sand, and fines increases the internal
resistance to flow, hence limiting the deformability and
speed of flow of the fresh concrete. Such friction is
especially high when the concrete flows through a
restricted spacing because of the greater collision
between the various solids that increase viscosity. A local
increase in aggregate density in a poorly viscous system
can lead to coagulation and arching of the aggregate and

Materials and Structures/Mat~riaux et Constructions, Vol.32, November1999

an interference with the deformability of the concrete in

a restricted area [1-4]. Inter-particle friction between
cement grains can be reduced by using a high-range
water reducer (HRWR) to disperse the cement grains. A
high dosage of H R W R can however lead to segregation
and blockage of the flow. The combined use of H R W R
and viscosity-enhancing agent (VEA) or a H R W R and a
low w/cm can reduce the free water content necessary to
ensure adequate viscosity and maintain good suspension
of coarse aggregate and reduce inter-particle collision
and coagulation of solid particles during the flow.
In addition to providing adequate stabiiity during
placement, the concrete should have a proper stability in
the formwork until hardening to minimize bleeding and
segregation. This is important to secure homogeneous
properties of the hardened concrete. Ensuring adequate
stability is critical in deep sections where highly flowable
concrete can exhibit segregation and bleeding and a non
uniform distribution of mechanical properties, bond to
reinforcing steel, and microstructure [5-7].
The contradicting workability requirements needed
for successful placement of SCC necessitate tailoring a
concrete mixture to ensure good balance between
deformability and stability to prevent blockage during
the flow and ensure a homogeneous suspension of the
concrete constituents. Such homogeneous distribution
is necessary to ensure adequate structural performance
and durability. Engineers are faced with the complex
task of manipulating several variables to enhance concrete performance and reduce cost. Some guidelines
exist for mixture proportioning of SCC to reduce the
extent of trial mixtures required to strike a balance
between the various contradicting mixture requirements. The recommendations are mainly based on
increasing the paste volume, reducing the coarse aggregate volume and sand to powder ratio, optimizing the
granular squelette of all solids, etc. For the most part
they treat a specific class of concrete and may require
special test equipment or software [8-11]. The majority
of mix design guidelines do not consider the specific
effect of mixture parameters and their interactions on
concrete performance. The objective of this paper is to
illustrate the feasibility of using a statistical experimental
design approach to identify the relative significance of
primary mixture parameters and their coupled effects on
relevant properties of SCC. The models can be used to
evaluate the potential influence of adjusting mixture
variables on concrete properties required to ensure successful development of SCC. Such simulation can help
identify potential mixtures with a given set of performance criteria that can be tried in the laboratory, hence
simplifying the test protocol needed to optimize SCC.

2. FACTORIAL DESIGN APPROACH

Five key mixture parameters that can have significant
influence on mixture characteristics of SCC were
selected to derive mathematical models for evaluating
relevant properties of SCC. The five variables included

the concentrations of VEA and HRWR, the w/cm, the

content ofcementitious materials (CM), and the volume
of coarse aggregate (Vca). The concrete responses that
were modelled were the slump flow, and rheological
parameters to evaluate the deformability of concrete in a
non-restrained area, as well as the filling capacity and Vfunnel flow time to evaluate the deformability in a
restrained area that reflect its deformability and resistance to blocking. The other modelled responses
included the surface settlement, segregation resistance,
and compressive strength (fc) after 7 and 28 days.
The underlying factors that influence fresh concrete
properties and strength development are too complicated
to permit the development of an exact mathematical
model. Therefore, an empirical statistical model was
derived over a wide working range of mixture proportioning. A 25-1 statistical experimental design was used
to evaluate the influence of two different levels for each
of the five mixture variables on the relevant concrete
properties. Such a two-level factorial design requires a
minimum number of tests for each variable. The initial
levels of the five selected mixture variables were carefully
chosen after reviewing the demand constraints imposed
by the targeted concrete properties. Given the fact that
the expected responses do not vary in a linear manner
with the selected variables and to enable the quantification of the prediction of the responses, a central composite plan was selected where the response can be modelled
in a quadratic manner. Such a plan enables the evaluation of the five selected mixture parameters with each
studied in five distinguished levels: codified values of-a,
-1, 0, 1, and ct. The ot value is chosen so that the variance of the response predicted by the model would
depend only on the distance from the center of the
modelled region. The value ~t is equal to NF 1/4 where
N F is the number of fractional factorial points 25-1 = 16
(or = 161/4 = 2).
The 32 mixture combinations used in the factorial
design consisted first of 16 mixtures for the fractional factorial plan where the mixtures were set at coded values of
-1 and +1. The 25-1 fractional factorial design was
expanded to include 10 additional mixtures where each
variable was adjusted separately at the extreme 0t value of
-2 and +2 with the other variables maintained at the
0 central points. This was done to consider extreme values
of the five principal variables on the measured responses.
Six replicate central points were prepared to estimate the
degree of experimental error for the modelled responses.
The coded variables are calculated as follows:
coded w/cm = (absolute w / c m - 0.435) / 0.0325
coded CM = (absolute CM - 480) / 60
coded VEA = (absolute VEA - 0.125) / 0.0375
coded HRW1K = (absolute H R W R - 0.7) / 0.2
coded Vca = (absolute Vca - 320) / 40
The experimental region modelled in this study is
illustrated in Table 1. Mthough the models are valid for
mixtures between the -2 and +2 regions, it is recommended to limit their use to the area bound by coded
values corresponding to -1.5 to +1.5. This can eliminate
the outer regions approaching the edges of the modeled

680

Khayat, Ghezal, Hadriche

Table 1 - Grain-size distribution of sand and coarse aggregate

Factors

-2

-1.5

Central
point

1.5

Table 2 - Chemical and physical properties

of cementitious materials

w/cm

0.370

0.386

0.435

0.484

0.50

CM (kg/m 3)

360

390

480

570

600

VEA (% water)

0.050

0.069

0.125

0.181

0.20

HRWR (% CM)

0.30

0.40

0.70

1.00

Vca (I/m 3)

240

260

320

380

Silica
fume Fly ash

Cement

SiO2

93.6

41.9

20.7

AI203

0.3

23.43

4.0

C3S = 59.6

0.5
0.3

18.89
7.11

2.6

C2S = 14.5

1.10

Fe203
CaO

62.9

C3A = 6.4

400

MgO

0.5

0.86

1.4

1.57

2.3
0.76

C4AF= 7.9

Na20 eq.
C

1.9

3 days = 20
7 days = 27

region since the prediction error increases with the distance from the center.

3. MATERIAL PROPERTIES

fc (MPa)

LOI

2.8

2.97

3.0

Specific gravity

2.22

2.53

3.14

Bulk unit weight (kg/m 3)

280
410

345

Blaine(m2/kg)
20250
Surface area B.E.T. (m2/kg) 17 500

A proven ternary binder was used in this study to

88.9
% passing45 p m
100
79.5
enhance rheological properties and strength. The binder
contains 3% silica fume and 20% Class F fly ash.
The chemical analysis of the Type 10 cement, silica
Table 3 - Grain-size distribution of sand and coarse aggregate
fume, and fly ash are given in Table 2. A continuSieve size (mm)
ously graded, crushed limestone coarse aggregate
28 20 14 10
5 2.5 1.25 0.630.31 0.16 0.08
made of two sizes of 14-5 and 20-5 m m to enhance
particle packing was used. The bulk specific gravity
20-5mm 100 98 55 25
2
1
0
and absorption of the combined aggregate are 2.74
14-5rnm 100 100 93 45 0.6 0
and 0.39%, respectively. A well-graded natural
Sand
100 100 100 100 98 87 74 56 34
8 0.3
siliceous sand with a fineness modulus, bulk specific
gravity, and absorption of 2.36, 2.69, and 0.63%,
ter [12]. The test involves recording the torque required
respectively, was used. The grain-size distributions of
to maintain a four-finger impeller rotating in a planetary
the two coarse aggregate types and the sand are given in
motion at an angular speed of 0 to 1.2 revolutions per
Table 3.
second. The descending flow curve was used for linear
A naphthalene-based HRW1K conforming to the
regression analysis to determine g and h according to the
Canadian Standard CSA3-A266.6-M85. A hydroxyl
Bingham flow model.
carboxylic acid-based set retarder c o n f o r m i n g to
The facility of aggregate particles and mortar to
Canadian Standard CSA3-A266.2.M78 was used at a set
change their flow paths and spread through a restricted
dosage of 100 ml/100 kg of binder to enhance fluidity
area without blockage was evaluated using the V-funnel
retention. Kelco-Crete welan gum was selected for the
test shown in Fig. 1 similar to that suggested by Ozawa et
VEA to enhance stability of the fresh concrete. The
al., 1994 [13]. The flow of the concrete is noted as the
Kelco-Crete was premixed with part of the H R W R to
time between the removal of the outlet and the seizure
facilitate dispersion.
of flow. The filling capacity test shown in Fig. 2 was
used to determine the facility of the concrete to deform
4. EXPERIMENTAL PROCEDURES
readily among closely spaced obstacles [14]. The test
involves the casting of concrete in the non-reinforced
All mixtures were prepared in 60-L batches with a
rotating drum mixer. The batching sequence consisted
of homogenizing the sand and aggregate for 30 s, then
adding 75% of the mixing water and all of the H1LWR
that is not present in the VEA-H1LW1K dispersion.
Following 30 s of mixing, the CM was added, and the
mixing was resumed for one minute. The remaining
water was then added and followed by the VEA-HRWR
Fig. 1 - Schematic
dispersion and set retarder. The concrete was mixed for
of the V-funnel
apparatus.
three minutes, and after two minutes of rest, the mixing
was resumed for two additional minutes.
For each mixture, the slump flow was measured, and
the relative flow resistance (g in Nm) and torque viscos7 x 7 mm
ity (h in Nm.s) were determined using the IBB rheome'l

681

Materials and Structures/Mat~riaux et Constructions, Vol.3,2, November1999

Fig. 2 - Schematic o f the filling

of measured values can be accounted for with the proposed models.

The presentation in Table 4 enables the comparison
of various parameters and the interactions of modelled
responses. For the majority of the parameters the probabilities, it can be noted that there is less than 5% chance,
or 95% confidence limit, that the contribution of a
given parameter to the tested response exceeds the value
of the specified coefficient. A negative estimate signifies
that an increase of the given parameter results in a reduction of the measured response. For example, an increase
in w/cm increases the slump flow and reduces the Vfunnel flow time. For any given response, the presence
of parameters with coupled terms, such as w/cm.w/cm,
indicates that the influence of this term on the modelled
response is quadratic. The models in Table 4 give an
indication of the relative significance of various mixture
parameters on each response. For example, the 28-day
fc value is found to be affected mainly by the changes in
w/cm followed by a reduction of the coupled effect of
Vca.CM, and in a third level by the reduction in CM,
VEA.VEA, and CM.CM. The segregation model is not
given as it had relatively low R 2 and repeatability.
Nine mixtures were selected to verify the ability of
the proposed models to predict the measured responses.
The duplicate mixtures were selected to cover a wide
range of proportioning. Table 5 shows the mean measured responses of the six replicate mixtures, coefficients
of variation (C.O.V.), as well as the standard errors with
95% confidence limit for each of the measured properties. The relative experimental errors for the slump
flow, filling capacity, surface settlement, and 7- and 28day fc are shown to be limited to 3 to 7%. On the other
hand, the relative error for the V-funnel flow time and
the h rheological parameter was approximately 13%,
while that of the g parameter was 34%. The latter value
is expected to decrease with the increase in relative yield
value, since the mean g value of the highly flowable
SCC corresponding to the central points was a very low
value of 0.7 Nm.
The six central point mixtures used to establish the
repeatability of the results (Table 5) were used along

capacity apparatus.

section at a constant rate up to a height of 220 m m and

observing the facility of the concrete to flow in the
restricted region. The maximum theoretical filling
capacity is 100%.
A simple test that can be used to evaluate the stability
of concrete following casting and until hardening was
used [15-16]. An 800-ram high PVC column measuring
200 m m in diameter is filled with approximately 700
m m of concrete to monitor surface settlement. An
LVDT fixed on top of a thin Acrylic plate anchored to
the top surface of the concrete column was used to
monitor surface settlement.
For the segregation test, a fresh concrete sample was
gently poured from a 2-L container over a 5-mm mesh to
observe the quantity of mortar passing through the screen
after 5 minutes [17]. The mass of the mortar passing
through the screen was compared to the theoretical volume of mortar in the 2-L sample to determine the segregation index. A stable concrete should exhibit an index
lower than 5% [17]. Six 100 x 200 m m cylinders were
cast and moist-cured to determine fc after 7 and 28 days.

5. DERIVED STATISTICAL MODELS

The derived models along with correlation coefficients and Prob. >Jt[ values are given in Table 4. The
estimates for each parameter refer to the coefficients of
the model found by a least square approach. The Prob.
>lt[ term is the probability of getting an even greater t
statistic, in absolute value, that tests whether the true
parameter is zero. Probabilities less than 0.05 are often
considered as significant evidence that the parameter is
not zero, i.e. that the contribution of the proposed parameter has a highly significant influence on the measured
response.
The correlation coefficient (R 2) values of the proposed models for slump flow, filling capacity, flow time,
h, g, settlement, and 7- and 28-day fc are 0.95, 0.90,
0.90, 0.98, 0.83, 0.76, 0.91, and 0.83, respectively. For
the majority of the models, the high 1Z2 values demonstrate excellent correlation considering that at least 90%

682

Table 5 - Repeatability of test results

Filling
capacity (%)

Flow
time (s)

h (Nm.s)

(ram)
735

83.1

4.18

7.46

20.3

6.7

0.51

1.11

2.7

7.2

12.0

14.8

g (Nm)

Settlement

(%)

7-dayfc
(Mea)

28-dayfc
(Mea)

0.71

0.40

25.36

39.08

0.24

0.022

0.94

1.47

34.0

6.0

3.9

4.1

Slump flow
Mean (N = 6)
Standard
deviation
Relative error
95% confidence
limit (%)

Mean (N = 6)
Standard
deviation
Relative error
95% confidence
limit(%)

Khayat, Ghezal, Hadriche

Table 4 - Parameter estimates of all derived models

Parameter

Slump Flow

Fillingcapacity

R2 = 0.95

R2 = 0.90

h
R2 = 0.98

Estimate prob Estimate Prob Estimate Prob Estimate Prob

>ltl
Intercept

Ln (Flowtime)
R2 = 0.90

744

>ltl
81

>ltl

>It]

1.41

7.71

w/cm

74.38

14.25

-0.527

-5.99

CM

136.6

25.4

-0.914

-9.95

VEA

-35.13 0.001

-3.74

0.13

NS

NS

1.05

0.023

HRWR

69.46

10.8

NS

NS

NS

NS

Vca

21.88

0.026

NS

NS

0.184

0.028

-1.37

0.004

w/cm.w/cm

-15.8

0.069

NS

NS

NS

NS

1.64

CM.CM

-49.4

-7.6

0.001

0.344

0.001

4.17

HRWR.HRWR

-41.4

-7.2

0.002

0.151

0.039

NS

NS

0.026

0.417

4.49

NS

NS

2.53

NS

NS

NS

NS

NS

NS

CM.w/cm

-27.7

0 . 0 2 2 -6.98

VEA.w/cm

-38.94

0.002

HRWR.w/cm

-28.94

0.017

NS

NS

HRWR.VEA

NS

NS

5.26

0.086

HRWR.CM

NS

NS

NS

NS

- 0 . 1 7 1 0.094

Vca.CM

NS

NS

NS

NS

0.168

Vca.HRWR

NS

NS

NS

NS

VEA.CM

NS

NS

NS

Vca.Vca

NS

NS

NS

SQ-RTg
R2 = 0.83

Parameter

- 4 . 0 1 0.184

NS
0.024

0.099

1.88

0.001

NS

NS

-2.42

NS

-0.21

0.037

NS

NS

NS

0.173

0.02

NS

NS

Settlement
R2 = 0.76

0.92

>ltl
25.9

NS

-4.14

-1.47

0.002

-0.340 0.0004

CM

-0.679

0.0709

VEA

NS

NS

- 0 . 0 5 2 0.004

HRWR

NS

NS

w/cm.w/cm
CM.CM

NS

0
NS

-0.159 0.0662 -0.053 0.009

NS

NS

-1.07

0.017

-0.64

0.129

NS

NS

NS

NS

0.55

0.167

0.405

NS

NS

-1.45

0.001

VEA.VEA

NS

NS

CM.w/cm

0.324

0.003~

VEA.w/cm

NS

NS

HRWR.w/cm

NS

NS

HRWR.VEA

NS

NS

HRWR.CM

>ltl

0.28

w/cm

Vca

7-dayfc
R2 = 0.91

Estimate prob Estimate Prob Estimate Prob

>ltl

Intercept

NS
-1.28

0.0269 0.062i

-1.75

NS

NS

NS

NS

NS

-1.08

0.044

0.057

0.030

NS

NS

0.066

0.014

NS

NS

-0.102

NS

NS

NS

0.18

0.088

Vca.w/cm

NS

NS

0.093

0.001

NS

NS

Vca.VEA

NS

NS

0.044

0.082

0.73

0.161

Vca.Vca

NS

NS

NS

NS

-1.79

Vca.HRWR

NS

NS

1.18

0.029

Vca.CM

-0.071 0.009

0 . 0 3 1 0.135

NS

NS

NS

NS

with the duplicate nine mixtures to compare the measured-to-predicted values of the eight reported
responses. As shown in Fig. 3, for the measured-to-predicted value comparison, the estimated errors corresponding to 95% confidence limits are indicated. These
values were determined for the highly flowable mixtures
corresponding to the central point that exhibited a mean
slump flow of 735 mm. The estimated errors for the

slump flow, filling capacity, flow time,

and g and h parameters were + 20 mm, +
6%, 0.5 s, 0.24 Nm, and 1.1 Nm.s,
respectively. These values were 0.024%,
1 MPa, and 1.6 MPa for the surface settlement, 7 and 28-day fc values, respectively. Except for the surface settlement
model, the comparison of the measured
responses of the 15 mixtures to predicted
values is good, as the measured values lie
close to the predicted ones. On the average, for the 15 duplicative mixtures (6
central points + 9 simulation), the mean
ratio of predicted-to-measured slump
flow, filling capacity, flow time, g, h, surface settlement, 7 and 28-day fc were
0.99, 0.99, 1.04, 1.42, 1.20, 1.40, 0.98,
1.01, respectively.

6. LIMITATIONS AND
EXTENSION OF EXISTING
MODELS

Although the statistical models were

developed to cover a wide range of mix28-day fc
ture proportioning, the precision in preR2 = 0.83
diction of the responses will change with
Estimate Prob
the deviation from the set of materials
>ltl
used in deriving these models, however,
38.5
the models can still be used for mixture
optimization and simulation when pre-5.63
0
sented with a different set of materials
-1.4
0.021
providing
that such materials have limited
NS
NS
effect
on
the
prediction accuracy of the
NS
NS
modelled
responses.
For example, a
NS
NS
future mixture proportioning could
NS
NS
involve different combinations and types
-1.32 0.017
of CM but the same aggregate, VEA, and
-1.38 0.013
HRWR types used in the original factorNS
NS
ial experimental design. Such new
NS
NS
binder is likely to influence the predicNS
NS
tion models in Table 4, but to what
NS
NS
degree remains to be seen. A logical
design approach would be to use the
NS
NS
existing model to predict the optimal
-2.53 0.001
design, then carry out selected tests to
NS
NS
quantify the influence of the new binder
NS
NS
on
the model. A minimal estimate is
NS
NS
obtained by repeating selected tests at levNS
NS
els within the desired range for the new
job specifications, for example, a specified slump flow of 600 to 675 mm, a filling capacity
greater than 70%, and minimum 28-day fc of 35 MPa.
These mixtures are selected because the predicted properties from the existing model lie within the specified
range of fluidity, filling capacity, and strength. The data
obtained from the duplicated mixtures are then compared with the predicted values from the existing models. This approach was carried out using seven SCC

683

'

Materials and Structures/Mat&iaux et Constructions,

85~
I Slump
.

Flow

.
I ~ , ~''"

o.s

I ,~'~-""

250

450

650

~ ~ .-i~5-")

"~"

15 "1--- ----

4-------

~'~"

4" ---

.,~t

0.2

0.3

0.4

o W"

10

15

0.5

(mm)

--q--

j.:,~

_ ..a _ _ J. ~ - - ) ~ . - ' ~ _ - ~ - -

55

_ .~.~ ~'y_-:. ~_ .a__ .~__

_t-~V~.~"" ~ 1 I

40

. . . .

I
I
I
.j.~.flr,
- .-I -- -- -I.. -- ..xt-.~, ~
~1- -- -

70

I.~,U"

. . . .

0.1

85

--~iP~ -- --"
.

"" - ~ . . . .

100

.-~

II Mow ume I !

The derived models are useful t o understand interactions between mixture parameters affecting important
characteristics of SCC. This understanding can simplify
the test protocol because the models identify the relative
significance of each variable, thus providing key information required to optimize the design. It is important
to note that the statistical approach used here can be
applied easily to mixtures made using other materials to
modify the existing models.
The utility of such models to assist in the selection of
trial mixtures is illustrated through a few examples. The
proposed models can also be used to test the effects of a
group of variables on properties affecting the quality of
SCC. For example, the effect of increasing the w/cm vs.
the dosage of HR.W1L on slump flow and filling capacity
is shown in Figs. 5 and 6, respectively, for mixtures with
relatively low and high CM contents and fixed contents
of coarse aggregate and VEA. For any w/cm value and
HtLWR content, the concrete made with the higher
CM content of 540 kg/m 3 had clearly greater deformability and filling capacity than that containing 420 kg/m 3
CM content. The decrease in w/cm necessitates an
increase in the dosage of H R W R to maintain a fixed
slump flow, especially in the case of concrete with a

.'/,1"

o.~ ~ ' "

850
Measured

~ ~

25
10
10

20
Measured

as '

'

l .-'i.':.,'"

1----

""

40

.-/
4o.I-

-~ I

55

70

85

100

(s)

. . . .

,i

* -

- _~.~-,,~~.',~.~"

.-'"
-

.+ yl . . . .
35If--.t,..-.."---I-

2o 1-

L.'/.-"

25

Is, Lf-'"

15

20

25

I...-'"/.~
9
.o" I

i
30

soV;..'"l
35
Measured

80

35

lI

40

1999

7. EXPLOITATION OF THE DERIVED MODELS

i "~

0a t - - t @ ~ k ~ I
t-~..--t

November

l'
~'~.'"~
0.4 J~ - - - "-'1-- - - J 4 - _ ;,.'~1/-~:. _ .

. . . . .

.'~.-"

"i

II S e t t l e m e n t

Vol. 32,

45

(MPa)

Fig. 3 - Examples of measured properties vs. values predicted

from statistical models.

mixtures prepared with the same sand, coarse aggregate,

HRWR, and VEA used in the development of the models, except for the CM combinations that was different
than the 3% SF + 20% FA original binder. An eighth
mixture was also used and incorporated a binder with
the same CM but a different cement. All eight mixtures
had w/cm of 0.41, 185 1/m3 of CM, and 300 l/m 3 of
coarse aggregate. Although the CM content has the
highest influence on the slump flow and filling
responses, the change in CM type that can alter several
properties, such as the grain-size distribution, granular
porosity, water and admixture demand, and kinetics of
hydration, did not exhibit significant effect on the
degree of prediction. As can be seen in Fig. 4, the
majority of the predicted slump flow and filling capacity
values were within 20 mm and 6%, respectively, from
the measured values that constitute the experimental
errors for the repetition of the slump flow and filling
capacity tests.
The above approach can be used to verify the reliability of other models of interest to proportioning SCC.
The variation between the new tests and predicted values is expected to increase with the deviation from the
materials used in the original study. Depending on the
level of deviation, a limited number of mixtures can be
prepared to adjust the existing models to reflect the
influence of the new materials on relevant concrete
properties.

684

8650 l

6oo

r 550 ,If
550
....

j
600

650
700
Measured slump flow (mm)

95

85

,~ 75

~3
i/

65
65

75

85

95

Measured filling capacity (%)

10% FA

9 30% FA

20% FA

X 20% FA + 40% Slag o 3% SF + 30% FA

9 3% SF + 30% LF

9 3% SF + 40% Slag
4- 3% SF + 20% F A

Fig. 4 - Measured slump flow and fiUing capacity vs. predicted

values of mixtures made with various binders (all nfixtures had
same aggregate, H R W R , and VEA).

Khayat, Ghezal, Hadriche

Fig. 6 - Variations of filling capacity with w/cm and H R W R

Fig. 5 - Variations of slump flow with w/cm and H R W R .

Table 6 - Comparison of deformability, filling capacity,

and cost of stable SCC mixtures
Mixture

w/cm
CM (kg/m3)
VEA(% water)
HRWR(%CM)
Vca (l/m3)
Slump flow (mm)
Fillingcapacity(%)
Material c~st
(CAN $/m ~)

1
0.46
444
0.09
0.42
376
658
63

25
4

20

0.44 0.45 0.41 0.42 0.43 0.44

468 480 480 480 480 480
0.07 0.10 0.18 0.18 0.18 0.18

73
80

,T

78 83 85 87
100 104 105 107

.45//

/ /

~ Filling capacity (%)

/. / /

//

15

//%2////

lO

0.46 0.46 0.82 0.9 0.94 0.98

336 264 280 272 280 280
663 652 687 698 699 689

71
73 I 78

(57)x/

/ /

o
5oo

65/

/x/-/

550

/
/

7'x/

H".Y

/,o.,x/
600

650

/
7I)0

x./
750

800

Slump flow (mm)

lower content of CM. At high contents of HRWR, a

curvature in the responses is observed that can be either
due to a direct consequence of the mixture or an artificial consequence of the quadratic model. In Fig. 5 and
6, the central regions highlighted in white correspond to
the area where prediction error is the lowest, as the error
of prediction increase with the distance from the central
point of the modelled region. The curvature of the filling capacity response in the high precision central region
is an actual characteristic of the material. An increase in
HRWI< in a mixture with a fixed w/cm can lower viscosity and increase filling capacity. However, with the
increase in HP,.WK content, the filling capacity begins
to drop as a result of partial blockage of the aggregate as
the viscosity decreases with the increase in HRWR.
The models can be used to select optimized economical mixtures. This can reduce the effort often required
in carrying out trial batches for mixture optimization.
For example, Table 6 compares the deformability, filling
capacity and material cost of seven SCC mixtures that
exhibit good resistance to segregation (maximum segregation index of 10%). The unit cost of the concrete was
calculated to reflect only material cost. For the seven
mixtures made with a moderate CM of 440 to 460
kg/m 3 and 0.41 to 0.46 w/cm, the H R W R dosages were
adjusted to yield mixtures with slump flow of 650 to 700
m m and minimum filling capacity of 60%. In general, it
685

Fig. 7 - Relationship between filling capacity, slump flow, and

flow time.

can be seen that for the highly flowable concrete, the

increase in filling capacity necessitates an increase in
concrete cost. For example, mixtures 4 and 7 have the
same CM, VEA, and Vca values and resulted in similar
slump flow but different filling capacity results. The
predicted filling capacity of mixture 4 is 78% compared
to 87% for the other SCC that is expected to cost 7
CAN $/m 3 more than mixture 4.
One of the most useful applications of the existing
models is the establishment of relationships between the
responses of various test methods, which fairly independent on the materials in use. For example, a relationship
between the filling capacity, slump flow, and flow time
values can be derived for a given set of highly stable SCC
(Fig. 7). Such a relationship derived from 275 virtual
SCC mixtures selected to ensure a m i n i m u m filling
capacity and slump flow of 40% and 550 mm, respectively, a maximum flow time of 20 sec and segregation
index of 60%, and 28-day compressive strength greater
than 28 MPa. The multiple regression equation is
expressed as follows:
Filling capacity (%) =
8.1 + 0.107 slump flow (ram) - 1.107 flow time (s)

Materials and Structures/Mat6riaux et Constructions,Vol. 32, November 1999

[2] Ozawa, K., Maekawa, K., Kunishima, M. and Okamura, H.,
'High-performance concrete based on the durability of concrete
structures', Proceedings, 2nd East Asia Pacific Conference on
Structural Engineering and Construction, Chiang-Mai (1989).
[3] Ozawa, K., Maekawa, K. and Okamura, H., 'Development of
high performance concrete', Journal of the Faculty of Engineering,
the University of Tokyo (B) XLI (3) (1992) 381-439.
[4] Nagataki, S. and Fujiwara, H., 'Self-compacting property of
highly flowable concrete', ACI SP 154 (1995) 301-314.
[5] Khayat, K. H., Manai, K. and Trudel, A., 'In-situ mechanical
properties of wall elements cast using self-consolidating concrete', A CI MaterialsJournal 94 (6) (1997) 491-500.
[6] Khayat, K. H., 'Use of viscosity-modifying admixture to reduce
top-bar effect of anchored bars in fluid concrete,' Ibid. 95 (2)
(1998) 158-167.
[7] Petrov, N., 'On the bond and corrosion resistance of steel reinforcement embedded in self-consolidating concrete', (only available in French) Masters Thesis, Universitd de Sherbrooke,
Canada (1995).
[8] Okamura, H. and Ozawa, K., 'Mix design for self-compacting
concrete', Concrete Library of the Japan Society of Civil
Engineering, (25) (1995).
[9] Sedran, T., de Larrard, F., Hourst, F. and Contamines, C., 'Mix
design of self-compacting concrete (SCC)', Proceedings, RILEM
International Conference on Production Methods and
Workability of Concrete, Ed. Bartos, P.J.M., Marrs, D. L. and
Cleland, D. J., E&FN Sport, London (1996) 339-450.
[10] Sedran, T. and de Larrard, F., 'Rene-LCPC: software to optimize the mix design of high performance concrete', Proceedings,
BHP 96, 4th International Symposium on Utilization of High
Strength/High Performance Concrete, Ed. de Larrard, F.,
Lacroix, R., Paris (1996) 169-178.
[11] Petersson, O., Billberg, P. and Van, B. K., 'A model for self
compacting concrete', Proceedings, RILEM International
Conference on Production Methods and Workability of
Concrete, Ed. Bartos, P. J. M, Marrs, D. L., and Cleland, D. J.,
E&FN Spon, London, (1996) 483-492.
[12] Beaupr~, D., 'Rheology of high performance concrete', Ph.D.
Thesis, University of British Columbia, Canada (1994).
[13] Ozawa, K., Sakata, N. and Okamura, H., 'Evaluation of selfcompactability of fresh concrete using the funnel test',
Proceedings, Japan Society of Civil Engineering (25) (June 1995)
59-75.
[14] Yurugi, M., Sakata, N., Iwai, M. and Sakai, G. 'Mix proportion
for highly workable concrete', Proceedings, Concrete 2000,
Dundee (1993).
[15] Manai, K., 'Evaluation of the effect of chemical and mineral
admixtures on the workability, stability, and performance of selfcompacting concrete', (only available in French) Masters Thesis,
Universitd de Sherbrooke, Canada (1995).
[16] Trudel, A., 'Workability, uniformity, and structural behavior of
high-performance self-compacting concrete', (only available in
French) Masters Thesis, Universitd de Sherbrooke, Canada
(1996).
[17] Study on reducing unit powder content of high fluidity concrete
by controlling powder particle size distribution, Concrete Library
ofJSCE N-28 (December 1996).

with 1k2 of 0.80 and a standard deviation of 6%. The

contour diagrams of filling capacity shown in Fig. 7
illustrate that for a given slump flow value the decrease
in filling capacity is accompanied by a drop in the V-funnel flow time. The six values noted on Fig. 7 correspond
to actual results of mixtures prepared to derive the original statistical models for which the values of filling
capacity were predicted given measured slump flow and
flow values. These results are in close agreement with
the measured values reported in parenthesis. While the
derived relationship is not exclusive for all materials, it
shows that for the plotted region corresponding to stable
SCC, the V-funnel flow test does not provide good
indication of deformability through restricted spacing.
Concrete with a flow time of 5 s can have a slump flow
of 580 mm and 65% filling capacity but also a slump
flow of 700 mm and 77% filling capacity. This trend
confirms the findings in reference 13 that recommend
combining the V-funnel flow time with slump flow to
reflect the filling capacity level of the concrete.

8. CONCLUSION
The models established using a factorial design
approach are valid for a wide range of mixture proportioning and provide an efficient means to determine the
influence of key variables on SCC properties. Such
understanding can facilitate the test protocol required to
optimize SCC, hence reducing the effort necessary to
optimize specified concrete to secure balance between
various variables affecting flowability, deformability, stability, and strength. Although the models are based on a
given set of materials, they can be easily used as a building block to augment future studies involving other
materials. The existing models enable the comparison of
the possible responses of the different test methods to
identify trends useful for quality control (for example,
relationships between yield value and settlement for different mixtures).

REFERENCES
[1] Nanayakkara, A., Ozawa, K. and Maekawa, K., 'Flow and segregation of fresh concrete in tapered pipes', Proceedings, 3rd
International Symposium on Liquid-Solid Flows, ASME, FED75 (1988) 139-144.

686