CEMENT and CONCRETE RESEARCH. Vol. 19, pp. 476-486, 1989. Printed in the USA.

0008-8846/89. $3.00+00. Copyright (c) 1989 Pergamon Press plc.

S T A T I S T I C A L ANALYSIS OF S T R E N G T H AND D U R A B I L I T Y OF CONCRETE

MADE WITH D I F F E R E N T C E M E N T S

K. Rose, B.B. Hope. and A.K.C. Ip

Department of Civil Engineering, Queen's University, Kingston. Canada

(Communicated by G.G. Litvan)

(Received Sept. 19, 1988)

ABSTRACT

This investigation examined variables influencing the 28-day compressive strength

and 50-cycle salt scaling loss of concretes made with eighteen Type 10 cements.
Statistical analysis was performed on test data consisting of chemical and physical
properties of cements; properties of fresh concretes; compressive strength, sak scaling
loss, and air void parameters of mature concretes. For a similar water-cement ratio
and cement content, results from the correlation analysis indicated that the 28-day
concrete strength and 50-cycle salt scaling loss were influenced significantly by the
chemical and physical properties of the cement used in the mix. The variables iden-
tified in the correlation analysis were analyzed and equations predicting strength and
salt scaling loss of concrete were developed using multiple linear regression.

INTRODUCTION

Cement manufacture has changed significantly in the past two decades mainly as a result of
environmental and energy concerns (1). It is now being reported that modern cements do not pro-
duce concrete as durable as that made with older cements. Concrete placed since about 1970 is
much more vulnerable to carbonation (2), and older parking structures appear to suffer less deteri-
oration than newer ones (3). It has been suggested that excessive expansion of concrete subjected
to wetting and drying cycles is related to incomplete kiln reactions and that the deleterious reactions
occur in the larger cement particles (4).

Independantly of the above it has been observed, for some time, that the quality of the con-
crete used in Ontario has shown considerable variation, even though all the cements used in the
manufacture of concrete met the Ontario Provincial Standard Specifications Form 1301 (CSA
standard CAN 3-A5-M83) requirements for Type 10 cements. As a result of the observed variation
in concrete quality in Ontario a program of testing was undertaken (5). Firstly, physical and
chemical properties of eighteen samples of cement representing fourteen cement plants were deter-
mined. These cements were then used to produce samples of concrete with nominal compressive
strengths of 20 and 30 MPa respectively. Within each of these strength categories two types of
coarse aggregate were used in the mix, a good quality crushed dolomitic limestone called 'A' and
a poorer quality partially crushed natural gravel called 'B'. Various properties of these four re-
sulting concrete types (20A, 20B, 30A, 30B) were also measured.

476
Vol. 19, No. 3 477
STRENGTH, DURABILITY, CONCRETE, STATISTICAL A.'LkLYSIS

[t must be emphasised that the experimental investigation was designed to examine the influence
of different cements on the strength and durability of concrete. Traditional factors such as cement
content and water-cement ratio were not intended to be variables. These two factors have a pro-
found influence on the strength and durability of concretes made from a single cement. The con-
cretes examined in this experiment all used the same aggregate, had the same cement content and
had virtually the same water-cement ratio. This was done to minimize these influences, so that the
effect of the cements alone could be studied.

This paper is based on a statistical analysis of the test data and was undertaken to identify
the factors responsible for the observed variations in the quality of the concrete specimens.

It was decided that the quality of the concrete should be assessed in terms of the compressive
strength of the concrete (CSA standard A23.2-3C,9C) and also the cumulative mass loss (mass of
material lost from surface) in a salt scaling test (ASTM C672-84, using 3% sodium chloride solution
as the de-icer), since these properties most closely reflect the concerns expressed about the dura-
bility of concrete. Strength was measured at 3, 7, 28 and 91 days, and salt scaling loss at 5. 10.
15, 25 and 50 cycles. It was decided that the 28-day compressive strength and the 50-cycle salt
scaling loss would be the most appropriate variables to use.

Therefore, for the purposes of this analysis, these two variables (for each of the four types
of concrete) were considered to be the dependent variables. The analysis predicted the value of these
dependent variables in terms of the other independent variables. A more detailed discussion of the
analyses reported in this paper is available (6).

THE DATA FILES

The M T O data have been reported (5) in the form of 12 tables labelled A through L. These
tables included measurements of chemical and physical properties of the cements, measurements
made on the fresh concrete and measurements made on the mature concretes. These are defined
in the Appendix.

Table C of the M T O data contained the grading curves for the cements. The original 13
variables showed the percent passing specified sieve sizes. It was decided to augrnent these data
by calculating the percentage of cement between two sieve sizes, that was passing the larger sieve
but retained by the smaller. This resulted in 42 more variables. It should be noted that because
the data are highly correlated only one or two were used in any models that were developed. Finally
it was decided to include the calculated total specificarea of the cements as a variable. This variable
was included because its calculation made use of all the data in the grading curve.

Table D contained the results of nineteen chemical tests performed on the cements. In addi-
tion, the data were used to calculate three more variables (the percentage of C3S , C2S, and C~AF)
thought to be of possible relevance.

PRELIMINARY ANALYSIS

The first stage in the analysis was to show that the observed variation in the measured
strength and salt scaling loss for the different concretes was not due to chance alone.

The 28-day compressive strength of each concrete sample was measured for four cylinders
and the 50-cycle salt scaling loss for each concrete sample was measured on two cylinders. From
these data an analysis of variance was undertaken to test if the variation in results was due to the
different cements. The results of this analysis are summarized in Table 1. These results show that
in all cases there is better than a 99.5 % confidence that the cements differ in their effects on the
observed compressive strength and salt scaling loss in the concretes.
478 Vol. 19, No. 3
K. Rose, et al.

TABLE 1. V A R I A N C E A N A L Y S I S FOR TEST DATA

CONCRETE F RATIO P R O B A L I L I T Y THAT F VALUE

TYPE FOR CEMENT SOURCE COULO OCCUR BY CHANCE

30A STRENGTH 75,31 < 0.001

SALT SCALING 4,57 0.003

308 STRENGTH 116.12 < 0.001

SALT SCALING 3.89 0.005

20A STRENGTH 60.09 < O.001

SALT SCALING 24.75 < O.001

208 STRENGTH 52.07 < 0.001

SALT SCALING 16.92 < 0.001

C O R R E L A T I O N ANALYSIS AND I N F L U E N C E CHARTS

The basic step in developing models, that can be used to predict concrete quality, is to identify
the degree of association between a dependent variable and the independent variables.

The SYSTAT (7) package of statistical programs was used to analyze the data. The Pearson
Correlation Coefficient provides a measure of the inter-relationships between pairs of variables. A
correlation coefficient close to -- 1 (or -l) indicates a strong association with a positive (or negative)
relationship. A coefficient close to zero indicates little or no association between the pair of vari-
ables.

The calculated correlation coefficient, between the dependent variables and each of the in-
dependent variables, was classified as:

1. potentially important
2. possibly important
3. not important

with the rules for making this classification were respectively as follows:

1. Correlation coefficient between -1.0 and -0.4 or +0.4 and + 1.0

2. Correlation coefficient between -0.4 and -0.2 or +0.2 and +0.4
3. Correlation coefficient between -0.2 and +0.2

The first break point ~of- 0.4) for making a decision is based on the fact that the 95 %
confidence level for the correlation coefficient of 18 pairs of observations is + 0.44 .

Tables 2 and 3 show which variables were potentially important for the 28-day strength and
the 50-cycle salt scaling loss, respectively, for each of the four concrete types. The number asso-
ciated with each variable is the calculated correlation coefficient.

Similar data have been reported for variables possibly important and not important (6). The
independent variables, that were found to be potentially important, were further analyzed to detect
any cross correlation between them. This was necessary because a good predictive model should
not contain closely related variables.

T H E G R A D I N G C U R V E DATA

The fifty-six variables derived from the cement grading curve data were treated separately
because of their strong inter-relationships. For each dependent variable, plots were made showing
the range of particle size under consideration, labelled with the corresponding correlation coefficient.
Vol. 19, No. 3 479
STRENGTH, DURABILITY, CONCRETE, STATISTICAL ANALYSIS

TABLE 2. VARIABLES P O T E N T I A L L Y I M P O R T A N T TO 28-DAY S T R E N G T H OF

CONCRETE SPECIMENS

INDEPENDENT VARIABLE COEFFICIENT INDEPENDENT VARIABLE COEFFICIENT

20A CONCRETE 3OA CONCRETE

Soundness -0.66 Alkali content -0.75

Alkali content -0.57 Ferric oxide 0.68
Silicon dixoide 0.53 Tetracalcium aluminoferrite 0.58
Spacing factor (total) -0,52 Potassium oxide -O.60
Free calcium oxide -0.48 S t r e n g t h 2 8 - d a y , cement 0.60
Chord length (total) -0.47 Relative density, cement 0.57
Potassium oxide -0.47 Set time O.55
Set time 0.46 F i n e n e s s (% p a s s i n g 50#m) 0.55
Relative density, cement 0.46 Carbonate addition -0.54
V o i d s p e r 25mm ( t o t a l ) 0.45 Soundness -0.48
Void specific surface (total) 0.44 Density, concrete 0.46
Sulphur trioxide -0.43 Tricalcium aluminate -O.46
F i n e n e s s (% p a s s i n g 40~m) 0.40 Cement c o n t e n t 0.45
CarDonate addition -0.40 False set 0.45
F i n e n e s s /% p a s s i n g ~5~m) 0,45
L o s s on i g n i t i o n -0.41

208 CONCRETE 3OB CONCRETE

Alkalt content -0.85 Alkali content -O.79

Cement c o n t e n t 0.77 Silicon dioxide 0.62
Potassium oxide -0.71 Potassium oxide -0.62
Free calcium oxide -0.68 Cement c o n t e n t 0.62
Density, concrete 0.64 Sulphur trioxide -0.62
Silicon dioxide 0.60 Soundness -0.58
Relative density, cement 0.59 Density, concrete 0.56
Soundness -0.59 Air void content (total) -0.53
Ferric oxide 0.58 Air content -O.53
Slump -0.55 Free calcium oxide -O.51
Set t|me 0.55 Tricalcium aluminate -0.49
Sulphur trioxide -0.53 Paste-air ratiO (total) 0.48
Strength 3-day, cement -0.49 Aluminum oxide -0.46
EffeCtive water-cement ratio= -0.46 Relative density, cement 0.44
Trtcalcium alumtnate -0.45 Ferric oxide 0.44
Effective water-cement ratio -0.43 Tetracalcium aluminoferrite 0.44
Strength 3-day, cement -O.43
Set time 0.43
Carbonate addition -O.43
Effective water-cement ratio= -O.41

From an inspection of these plots a representative variable was selected for inclusion in the
modelling process.

A consistent pattern emerged from these plots. The 28-day strength had few or no strong
correlations with grading. The salt scaling results on the other hand exhibited two common fea-
tures. First, the percentage of" very fine particles (less than 24~m) had a strong positive correlation.
Second, the percentage of particles in the variables around the range 4 - 20#m had very strong
negative correlations (the strongest of" a n y of all the variables measured), much stronger than the
commonly used percentage passing 45#m.

Table 4 lists the variables from the grading curve data, (and their correlation coefficients),
that were retained for the model building phase of the analysis.
480 Vol. 19, No. 3
K. Rose, et al.

T A B L E 3. V A R I A B L E S P O T E N T I A L L Y I M P O R T A N T TO 50-CYCLE S A L T S C A L I N G
LOSS OF C O N C R E T E S P E C I M E N S

INDEPENDENT VARIABLE COEFFICIENT INDEPENDENT VARIABLE COEFFICIENT

20A CONCRETE 30A CONCRETE

Fineness (% b e t w e e n 2 - 20~m) -0.79 Fineness [% b e t w e e n 2 - 2 0 ~ ) -O.75

Effective water-cement ratio* 0.68 E f f e c t i v e water-cement r a t i o * 0.67
Aluminum o x i d e 0.68 Aluminum oxide O.B6
Tricalcium aluminate 0.67 Tricalcium aluminate O.85
F i n e n e s s (% p a s s i n g aS~m) -0.66 Strength 91-day, concrete -0.63
Paste-air ratio (total) 0.61 Fineness (% p a s s i n g 45~m) -O.61
A l r void c o n t e n t (total) -O.59 Strength 7-day, concrete -O.60
Strength 2B-day, concrete -O.52 I n s o l u b l e residue 0.58
Ferric oxide -0.49 Potassium oxide 0.48
Tetraca|cium aluminoferrite -O.49 Tricalcium silicate -0.46
V o i d s p e r 25 mm ( t o t a l ) -0.49 Ferric oxide -0.44
Insoluble residue 0.49 Tetracalcium aluminoferrite -0.44
Strength 91-day, concrete -O.48 Strength 28-day, concrete -O.44
S e t time -0.42 Specific surface areas 0.43
Tricalcium silicate -O.41 Chord length (total) 0.43
Effective water-cement ratio 0.43
Cement c o n t e n t -O.40
Spacing factor (total) 0.40

20B CONCRETE 306 CONCRETE

Fineness (% b e t w e e n 4 - 20pm) -0.53 Effective water-cement r a t i o - 0.73

Air content -0.46 Fineness (% between 2 - 2 0 # m ) -0.72
Set time -O.46 F i n e n e s s (% b e t w e e n 45~m) -0.62
Strength 3-day, concrete -O.43 Tricalcium aluminate O.61
A]uminum oxide 0.57
Ferric oxide -0.57
Tetracalcium aluminoferrite -O.57
Potassium oxide O.50
Alkali content 0.48
Insoluble residue 0.46
Set time -0.46
Air content -O.46

T A B L E 4. POTENTIALLY IMPORTANT VARIABLES FROM THE GRADING

CURVE

CASE VARIABLE COEFFICIENT

2B-day strength 2OA concrete % p a s s i n g 40am 0.40

2B-day strength 2OB concrete none
2B-day strength 30A concrete % p a s s i n g 50~m 0.56
2B-day strength 308 concrete none

5D-cycle salt scaling 2OA concrete % between 2 - 20~m -0.79

5D-cycle salt scaling 2OB concrete % between a - 20#m -0.53
SO-cycle salt scaling 3OA concrete % between 2 - 20~m -0.76
SO-cycle salt scaling 3OB concrete % between 2 - 20~m -0.72

E N G I N E E R I N G E V A L U A T I O N OF T H E C O R R E L A T I O N A N A L Y S I S

Before proceeding to the development of the models it was decided to review the results of the
correlation analysis in terms of their e n ~ n e e r i n g significance. This exercise was not intended to
Vol. 19, No. 3 481
STRENGTH, DURABILITY, CONCRETE, STATISTICAL ANALYSIS

reject any variables from further analysis but rather to see if the conclusions to date were reason-
able in terms of conventional theories about concrete behavior.

It must be emphasized that the following discussion does not purport to imply or suggest that
the statistical analysis indicates that the different cement properties are a result of the manu-
facturing processes discussed. On the contrary the statistical analysis stands on its own merits and
the following discussion simply shows that the results of the analysis are not unreasonable.

The influence of various chemicals on the properties of both wet and hardened concrete has
been reported in numerous publications. An excellent in depth discussion of these phenomina can
be found in reference 8 and references 9 and 10 contains a condensed discussion of the effects of
some of the cement constituents.

Strength

A study of the statistical results for the strength of the four concrete mixes indicates that
relative density of the cement and set time, likely have a positive influence on strength; soundness.
potassium oxide and alkali content likely have a negative influence in all four mixes. There is
engineering evidence that this is reasonable.

S o u n d n e s s : Soundness is primarilly related to the presence of free CaO and MgO in the cement.
The CaO results from incomplete kiln reactions and the MgO results from impurities. It is quite
reasonable, therefore, that cements which exhibit larger unsoundness expansions produce lower
strengths and are perhaps those with less complete kiln reactions.

get T i m e : Set time is related to the chemistry of the cement and is largely a function of the hy-
dration of the aluminates. The reactivity of the C4AF phase increases with both increasing
aluminate content and a decrease in kiln temperature (9). Setting time also decreases with high
CaA and high SO3 where the SO 3 is a result of addition of gypsum to the clinker. It is also known
that the addition of gypsum to the raw materials lowers the required kiln burning temperature (9'*.

Sulphates from the fuel used to fire the kiln are known to influence the early reactions of the
cement (9). High fuel SO3 combines with the alkalis thus reducing their effect on the C3S (8).
Modern gas and oil have less SO3 than they used to have and thus this effect may gradually have
been reduced as permissible emission is reduced.
It is reasonable, therefore, that cements which exhibit shorter set times could be those fired at Lower
kiln temperatures using fuels low in SO~. However, actual kiln burning temperatures for the ce-
ments studied are not available.

It is well known that rapid initial hydration has a detrimental effect on strength (10). This
effect is particularly evident in hot weather concreting when it is not possible to obtain strengths
developed at lower pouring temperatures. The effect is probably caused by the rapid formation of
the initial calcium silicate hydrate structure and it seems that the longer this structure has to form
the stronger is the cement.

P o t a s s i u m Oxide: Potassium oxide is known to adversely influence the strength gain of cements
and it is known to enter into the C2S phase replacing some of the CaO (8). This either results in
an increase in C3S content or results in a greater amount of free CaO. Increased K~O is also known
to lower the required burning temperatures (9). Cement dust is rich in K20 and if the dust collected
from modern precipitators is mixed with the raw ingredients and fed back into the kiln, this will
increase the K~.O content.

A l k a l i C o n t e n t : It has been previously reported from various investigations that increased alkali
content decreases the final strength of cement (11) and it is not surprising, therefore, that this factor
was identified by the statistical analysis.

R e l a t i v e Density: It is not known how the relative density of the cement can influence strength
but it does not seem unreasonable that the higher the relative density the stronger the concrete.
482 Vol. 19, No. 3
K. Rose, et al.

Other Factors: It m a y seem surprising that the statistical analysis did not identify the traditional
factors such as water-cement ratio. This is not unreasonable, however, since the concretes were
all manufactured with a constant cement factor and more or less constant slump. The factors which
affect strength per se could quite likely be the same factors which affect water demand for a con-
stant slump thus eliminating the influence of water-cement ratio from the results. It must be re-
membered that the traditional influence of water-cement ratio refers to its effect on a single cement
and not on cements with different chemical and physical characteristics.

S u m m a r y : The statistical results indicate that those cements which perform better than others
from the strength point of view are likely low in K20, exhibit slow set times and smaller
unsoundness expansions, and have higher cement densities.

Salt Scaling Resistance

There is no single parameter which appears to influence the salt scaling resistance of all four mixes.
This is perhaps not too surprising since the 20 MPa mixes are not likely to be particularly resistant
with any cement. This is particularly true for mix 20B which was made with an inferior aggTegate.
It should be noted, however, that different cement fineness parameters occur in all four cases.

There are a number of parameters in the statistical analysis which are likely to influence the
scaling resistance of three of the four mixes. Cement fineness, set time, Fe20 ~ and C, AF all have
a positive influence on the scaling resistance whereas insoluble residue, C3A, A0 a and water-
cement ratio based on the cement fraction passing the 45 micron sieve all have a detrimental effect
on scaling. Care must be exercised in the interpretation of these results since insoluble residue,
which comes mainly from impurities in the gypsum added to the clinker, C3A , ALzO3, Fe,.O3 and
C, AF are all interrelated. The insoluble residue is a function of the gypsum content which is related
to the amount of C3A . The amount of C3A is related, in turn, to the A~O 3 in the cement. Fe203
on the other hand reduces the amount of C3A formed and increases the amount of C4AF. Thus the
effect of all these parameters may be simply a manifestation of the quantity of C~A or the amount
of gypsum added to the clinker.
C e m e n t F i n e n e s s : The most consistent parameter indicated by the statistical analysis for the salt
scaling resistance is the amount of the very fine particles in the cement. Indeed, the amount finer
than 10~m is consistently indicated as important. From the engineering point of view, however, it
seems more practical to consider the amount of cement finer than, for example, 20urn since this is
a larger fraction of the whole and experimental deviations in the measured quantities are likely to
be smaller. The importance of cement in the 3 - 30~m range on the strength of concrete has been
previously identified I 11).

This influence of the very fine particles on durability could possibly be attributed to the early
strength gain of the concrete. Certainly if the saturation of the water in the mix with silicates is
associated with the hardening process then the very fine particles will have an influence because
of their very large surface area.

Set Time: As for the strength analysis the statistical analysis indicates that the longer the set time
the more likely the cement will produce a concrete with superior salt scaling resistance. Again it is
probable that this could be a result of the superior structure formed by the hydrating cement if this
initial structure is formed slowly.

T r i c a l c i u m Aluminate Content and Related Parameters: Higher C~A contents result in faster
set times since, as explained above, it is primarily the aluminates which dictate set times. CaA is
also known to react with the chloride ion to form tricalcium chloroaluminate. Whether this has an
influence on salt scaling resistance is not known.

S u m m a r y : The statistical analysis indicates that a cement which produces concrete with superior
salt scaling resistance is one which contains a relatively high proportion of very fine particles, is low
in C3A content and exhibits a slow set time.
Vol. 19, No. 3 483
STRENGTH, DURABILITY, CONCRETE, STATISTICAL ANALYSIS

REGRESSION ANALYSIS

The SYSTAT program contains a set of routines called Multiple General Linear Hypothesis
that can calculate several types of regression equations and apply many tests to check their statis-
tical significance. One particular option in the SYSTAT package is stepwise regression. With this
method the program reviews all the possible independent variables and introduces them into the
equation one at a time in an attempt to maximize the coefficient of multiple regression. This
technique is very useful in reviewing candidate variables but the resulting equations must be
examined with care.

This approach was applied to the 28-day compressive strength and the 50-cycle salt scaling
loss for each types of concrete. Tables 5 and 6 show the final equations derived for the strength and
the salt scaling loss of each type of concrete (along with the most relevant statistics), respectively.
An explanation of the output is as follows:

VARIABLE lists the variables in the estimated regression equation. COEFFICIENT is the calcu-
lated value of the coefficient for each variable in the equation. For example, for the 28-day strength
of type 30A concrete

28-Day Strength = -163.23 - 14.86(Alkali Content) + 0.55 (Fineness, % passing 50~m) +

55.05(Relative Density, Cement), where the variables are defined in the Appendix.

t is the Student's t statistic for each coefficient. It is a test to see if the value of the coefficient is
different to zero. A value of t larger than + 2 for a sample size of eighteen indicates a 95 % confi-
dence that the calculated coefficient is significant. R: is a measure of the success of the regression
equation in explaining the variation in the data. The value is between 0 and ~- i, the larger the
value the better the equation.
Care must be taken not to overload the equation with variables just to improve the value of this
statistic. E is a n estimate of the variation about the regression. The smaller the value the more
precise will be the predictions.

Reviewing these equations reveals the following observation:

TABLE 5. R E G R E S S I O N ANALYSIS FOR 28-DAY S T R E N G T H OF C O N C R E T E SPEC-

IMENS

TYPE VARIABLE COEFFICIENT ~ R2 F

2OA Constant -128.49 -3,26 0.86 1.09

Soundness -18.23 -3,84
Chord length (total) -43.91 -4.13
Relative density, cement 32.39 4.19
F i n e n e s s (% p a s s i n g aO#m) 0.18 2.29

20B Constant 45.28 20.45 0.89 1.13

A|ka]i content -15.21 -6,35
Free calcium oxide -1,76 -2.45
Set time 0.03 3.20

30A Constant -163.23 -1.91 O.79 2.09

Alkali content -14.86 -3.31
F i n e n e s s (% p a s s i n g 50#m) 0.55 2,85
R e l a t i v e d e n s i t y , cement 35.03 2.13

3OB Constant 71.23 15.99 0.88 1.22

Alkal| content -16.12 -S.84
Air content -3.04 -4.33
CarDonate aadttion -O,33 -2.66
484 Vol. 19, No. 3
K. Rose, et al.

T A B L E 6. R E G R E S S I O N A N A L Y S I S F O R 50-CYCLE S A L T S C A L I N G LOSS OF CON-

CRETE
SPECIMENS

TYPE VARIABLE COEFFICIENT = R2

20A Constant 1858.58 1.39 0.74 215

Fineness (% l a e t w e e n 2 - 20~m) -26,59 -1.21
Aluminum oxide 238.08 2.39
Air void content -195,34 -1.74

20B Constant 10673.21 2.83 0.28 1160

FiruBness (% b e t w e e n a - 20~m) -203.88 -2.40

30A Constant 3705.41 3.99 O.68 245

Fineness (% b e t w e e n 2 - 2OUm) -66.95 -4.25
Potassium oxide 577.65 2.14

3OB Constant -748.56 -O.51 0.80 237

Effective wateP-cement Patio 8674.63 4.65
Air content -492.64 -3.64
Insoluble residue 622.65 2.00

Strength

Good equations were obtained for 20A, 20B. 30A, and 30B concretes. The R: values were large and
all of the coefficients were significant.

Salt s c a l i n g

2 0 A c o n c r e t e : The overall quality of the equation was getting worse. In particular the t value of
the coefficients was approaching +_ '2 indicating lower confidence in the calculated values of the
coefficients.

20B c o n c r e t e : The equation had a very low value for R: and would be of little practial use.

3 0 A c o n c r e t e : The R~- value in the equation was reasonably large. This, combined with the fact
that all the coefficients were significant means that the equation has some potential.

30B concrete: The R: value in the equation was large, but the t value of the constant term was
very small. This is probably due to the inclusion of a variable in the equation that has a large mean
value but a small standard deviation. To the regression equation this variable looks like a constant
term even though the variable is highly correlated to the dependent variable.

In order to give an overall impression of the effectiveness of the equations, Fig. 1 shows a
typical plot of the predicted values along with the corresponding measured values of 28-day strength
for 30A concrete.

CONCLUSION

The results of this analysis indicate that, for a similar water-cement ratio and cement con-
tent, the concrete strength and salt scaling loss are highly depended on the chemical and phyc, ical
properties of the cement used in the mix. The cement variables affecting the concrete strength and
salt scaling loss at.. shown in Tables 5 and 6, respectively.

No conclusions regarding different manufacturing techniques can be drawn from the analysis
but from the engineering point of view the analysis does not violate present day concrete theory.
Vol. 19, No. 3 485
STRENGTH, DURABILITY, CONCRETE, STATISTICAL ANALYSIS

fl,. 60 PREDICTED VALUE = MEASURED V A L U E /

A EQUATION (TABLE 5)
A"
IJJ 55
,_1
<
>
5O A
1J.I
I- 45 ZX A ZX/
O
u
a
i,fl
rr 40
I I
35 40 45 50 55 60
MEASURED VALUE (MPa)

FIG. 1. P R E D I C T E D AND M E A S U R E D V A L U E S OF 28-DAY STRENGTH

F O R 30A C O N C R E T E

The concrete samples and measurements were made under carefully controlled laboratory
conditions and the statistical analysis has shown that the differences in strengths and durabilities
are a function of the cement and are not a random effect of experimental inaccuracy. We therefore
concluded that the finding of this analysis should prove useful in identifying methods to improve the
strength and durability of concrete. To refine these results, we would ask researchers who have
undertaken similar measurements to forward copies of their results to us in order that the d a t a base
used to estimate these relationships can be broadened.

ACKNOWLEDGEMENTS

The authors wish to thank the Ontario Ministry of Transportation for providing funds in the
form of a research contract for this project.

REFERENCES

1. A.T. Corish and P.J. Jackson, Concrete 16, 16 (1982).

2. A.M. Neville, ACI SP-100, 21 (1987).
3. G. Litvan and J. Bickley, ACI SP-100, 1503 (1987).
4. H. Vivian, ACI SP-100, 1691 (1987).
5. R. Northwood, B. Chojnacki, and R. Newell, ACI SP-100, 973 (1987).
6. K. Rose, B.B. Hope, and A.K.C. Ip, Report No..~IAT-88-08, The Research and Development
Branch, Ontario Ministry of Transportation, (1988).
7. L. Wilkinson, SYSTAT: The System for Statistics, Evanston, IL: SYSTAT, Inc., (1986).
8. F.M. Lea, The Chemistry of Cement and Concrete, Edward Arnold Ltd., London (1970).
9. P.K. Mehta, Concrete Structure, Properties and Materials, Prentice Hall (1986).
10. A.M. Neville, Properties of Concrete, Pitman Publishing (1981).
11. F.W. Locher, Symposium on Process Technology of Cement Manufacture, 632, Plymouth.
England, MacDonald and Evans (1977).
486 Vol. 19, No. 3
K. Rose, et al.

APPENDIX

VARIABLE DEFINITION/TEST METHOD

Air content M e a s u r e d at time of casting ,%~

A i r v o i d c o n t e n t (total) M e a s u r e d on hardened concrete ,%,
Alkali content Na, O + 0.658,K~Oi i%,
A l u m i n u m oxide A~O 3 content of cement (C~c~
Carbonate addition [L.O.I. iat 1050"CI - L.O.I. qat .550C~I El00 x
limestone I L.O.[.)] ~%)
Cement content C e m e n t content per m ~ concrete ~kg m
Chord length (total) D e t e r m i n e d by A S T M C 4 5 7 - 8 2 a ~mm,
Density, Concrete CSA S t a n d a r d C A N 3-A23.2-M77 (kg m::
Effective water-cement ratio Mass of w a t e r ' m a s s of c e m e n t
E f f e c t i v e w a t e r . c e m e n t ratio::' Mass of w a t e r / m a s s of c e m e n t finer than 45~m
False set CSA S t a n d a r d C A N 3-AS-MS3 t%)
Ferric oxide Fe:O3 content of cement (%)
F i n e n e s s (% p a s s i n g xx~m) C e m e n t finer t h a n xx#m
F i n e n e s s (% b e t w e e n x-yygm) C e m e n t between x and yy sizes !%)
Free calcium oxide A S T M C l 1 4 - 8 3 a t%)
Insoluble residue CSA S t a n d a r d C A N 3-AS-MS3 (%)
L o s s on i g n i t i o n CSA S t a n d a r d C A N 3-AS-M83 (,at lO.50~C ~ ,%~
P a s t e . a i r r a t i o (total) M e a s u r e d on hardened concrete
Potassium oxide K~O content of c e m e n t [%)
Relative density, cement CSA S t a n d a r d C A N a-AS-M83
S p a c i n g f a c t o r (total) M e a s u r e d on h a r d e n e d concrete (mm~
Specific surface area Caculated from cement grading (mm--',
Set time CSA S t a n d a r d C A N 3-AS-MS3 (rain)
Silicon dioxide SiO:content of cement (co)
Slump CSA S t a n d a r d CAN 3-A23.3-M77 imm~
Soundness Le Chatelier T e s t (% expansion)
S t r e n g t h x-day, c e m e n t CSA S t a n d a r d C A N 3-AS-MS3 at x days !MPa,
Strength x.day, concrete CSA S t a n d a r d C A N 3-A23.2-M77 at x days <MPal
Sulphur trioxide SO 3 content of cement
Tetracalcium aluminoferrite C~AF content of c e m e n t (%)
Tricalcium aluminate Calculated C~A content of c e m e n t
Tricalcium silicate Calculated C3S content of c e m e n t
V o i d s p e r 25 m m (total) M e a s u r e d on h a r d e n e d concrete
V o i d s p e c i f i c s u r f a c e (total) M e a s u r e d on h a r d e n e d concrete (ram -~)