Professional Documents
Culture Documents
TECHNOLOGY
ELSEVIER P o w d e r T e c h n o l o g y 83 (1995) 245-252
Abstract
The fineness of a Portland cement and its chemical composition are the key factors in determining cement strength
characteristics. This paper describes a study of the relationship between cement particle size distribution and surface area,
and the effects of size distribution, surface area and chemical composition on concrete strength. A new correlation between
particle size distribution and surface area has been established and models which include the size distribution and chemical
composition as factors have been developed to predict cement strength based on a large data base generated from industrial
surveys. These models are useful in optimizing clinker grinding by simulation. The relationship between the cement compressive
strength determined by the concrete mix and mortar test methods is also discussed.
Keywords: Cement strength characteristics; Particle size distribution; Surface area; Chemical composition; Clinker grinding
2. Data collection The other model was proposed by Sumner et al. [6],
in which the cement product surface area, SSA, can
144 sets of particle size distribution (PSD) and as- be calculated by the equation:
sociated surface area data were collected from surveys
conducted at four industrial cement plants in Australia
6 (wiF~
SSA= z5 ~ \~-6] (m2/kg) (3)
[26]. F r u r types" of cement were involved, identified as i--1
Sm = 807 + 1.20,,
where Om is the measured surface area (cm2/g), and
Sm is the surface area calculated from Eq. (2) (cm2/g):
(2)
<
i
'00
3oo
200
I
pi-lXm
where p is the cement density (g/cm3),Xm is the geometric 1oo
mean size of xi and x~÷l (cm), AQ~(x~, x~+l) is the
difference of the cumulative mass distribution of the I I I ~ I
i,h and i + 1TM particle size (%), and n is the number 0 100 200 300 400 500
of size fractions. This model was based on three types Observed SSA (m2/kg)
of cement, 45F, 35F and 55, manufactured in Germany Fig. 2. Observed vs. predicted surface area (based on Eq. (3); shape
[17]. factor, F = 1.15).
Y.M. Zhang, T.J. Napier-Munn / Powder Technology 83 (1995) 245-252 247
instruments, or a limitation of the database used in indicates that it may have more general applicability
the original model development. than Eqs. (1) and (3). The model has been successfully
used in the simulation of clinker grinding processes
3.2. Development of a new model [12,13].
20
5.1. A descriptor for chemical composition 62 64
I
66
I I
68 70
I I
72 74
S
SD 0.08 5.537
R 2= 0.652; SE = 2.128; N = 50
8 R2=0.814; S E = 1.307; N = 5 0
30
•
35
,
40 45
P r e d i c t e d 2 8 - D a y S t r e n g t h (MPa)
R 2 = 0.798; SE = 1.639; N = 50
Fig. 12. Predicted vs. observed 28 d strength (based on Eq. (16)).
Eqs. (11)-(13) therefore provide predictions of strength
as a function of composition and particle size, and Eqs. 6. Correlation of strength determined by two testing
(14)-(16) as a function of composition and surface methods
area. The predictive capabilities of the two sets of
equations are not significantly different. Surface area Two standard methods are in common use for testing
and size can thus substitute for each other in terms the compressive strength of cement. They are the
of predictive power. concrete mix test (also called the cylinder test) and
the mortar test (also called the ISO/CEN mortar test).
Although the mortar test has the advantage of re-
24 quiring less labour and a smaller cement sample com-
pared with the concrete mix test, the concrete mix
22 .
testing values are, in general, more accepted by the
},8
20
..,y. ". customers than those obtained by the mortar test.
However, no correlation between the strength results
determined by the two standard methods has apparently
~ 16 yet been reported in the open literature. It was thought
to be useful to establish the correlation so as to permit
the prediction of cement strength to be made according
to either definition.
A total of 71 data sets recorded by the laboratory
8 10 I I I I I I at one cement plant over four types of cement (type
10 12 14 16 18 20 22 24 A, type B, type C and type ACSE) were therefore
P r e d i c t e d 3 - D a y S t r e n g t h (MPa) collected and used to develop a correlation between
Fig. 10. Predicted vs. observed 3 d strength (based on Eq. (14)). the strength results determined by the two testing
Y.M. Zhang, T.Z Napier-Munn / Powder Technology 83 (1995) 245-252 251
methods. The correlation is given in Eq. (18). The cessfully used to predict the cement strength determined
predicted concrete mix strength calculated from Eq. by the concrete mix test from the mortar test results
(18) is plotted against the regressed results in Fig. 13. for any of the four types of cement used in the study.
STNmtx=0.679 STNmo,t,r +0.114 D A Y - 4 . 4 9 5 (18)
SD 0.0269 0.03704 0.8617 7. Conclusions
R2=0.956; SE=2.137; N = 7 1
Previous models correlating cement particle size dis-
where STNmiX is the strength value determined by the tribution to the specific surface area developed by
concrete mix test (MPa), STNmort,r is the strength value Kuhlmann and Sumner et al. [6] were applied to a
determined by the mortar test (MPa), and DAY is the large data base (144 data sets) collected in four Aus-
time (d). tralian industrial sites, and it was found that these
A further 12 samples were collected over three types models were not adequate for the data base collected.
of cement (type A, type ACSE and type B), and these The reason for this may be the different methods of
samples were tested at the same laboratory using both determining the size distribution, or a limitation of the
concrete mix and mortar tests, at three different setting data base used in the original model developments. A
times, providing 36 new data points. These independent new correlation between the particle size distribution
results were used to evaluate the model. The plot of and surface area was then developed based on the
observed and predicted values using the model (Eq. present large data base (Eqs. (4)-(6)).
(18)) is shown in Fig. 14. It is obvious from Figs. 13 The effects of cement particle size distribution, size
and 14 that the test results are strongly correlated and fraction, surface area and chemical composition on the
that a simple linear expression (Eq. (18)) can be suc- strength have been investigated. A total of 50 sets of
data (forming part of the larger database) over three
50 types of cement were used in the investigation. The
multi-linear regression program, LREGMB, developed
40. by Kojovic [25], was used in the study to select the
fr
most significant correlations based on statistical criteria.
30- The correlation between the concentration of the
four major chemical components and 28 d strength
characteristic, F28, developed by Knofel [24] was found
20-
to be inadequate for the data collected in the study.
o
U ' A new descriptor of chemical composition, FCH, was
~ 10- defined (Eq. (9)) and found to be more strongly related
to the 28 d strength than F28 (Eq. (10)). Regression
0 | I I I I
models were then developed, expressing strength in
0 10 20 30 40 50 terms of chemical composition (FCH) and the pro-
Observed Concrete Strength (MPa) portion in certain size intervals (Eqs. (11)-(13)), and
composition and surface area (Eqs. (14)-(16)).
Fig. 13. Observed vs. predicted concrete strength (Eq. (18) regression).
These equations show that the early strengths (3 d
50
and 7 d) are significantly affected by the proportion
of - 8 ~m material, whilst the later strength (28 d) is
strongly related to the proportion of - 3 2 / ~ m material.
The cement strength, both early and later, was found
to increase with an increase of surface area. The
30 selection of a model to predict strength will depend
upon what information is available: PSD (Eqs. (11)-(13))
or surface area (Eqs. (14)-(16)). In view of the variety
6 of cement types and sources incorporated in the data
base, it is expected that these correlations will have
wide applicability.
A new relationship between the cement compressive
strength results determined by the concrete mix and
0 10 20 30 40 50
mortar test methods has been established based on an
Observed Concrete S~ength (MPa) extra 71 data sets logged from one industrial laboratory
Fig. 14. Observed vs. predicted concrete strength (based on Eq.(18)) (Eq. (18)). This model was then applied to the results
-- i n d e p e n d e n t data set. from an additional 36 independent data points obtained
252 Y.M. Zhang, T.J. Napier-Munn / Powder Technology 83 (1995) 245-252
with three types of cement. This demonstrated that Southern Cement Ltd, particularly Mr M.R. Boadle,
the cement strength determined by the concrete mix for their support and assistance during sampling surveys
test can be well predicted from the results determined and data analysis, and for financial support of the work.
by the mortar compressive test using the simple model.
These new relationships are helpful in optimizing
clinker grinding by simulation using grinding and clas-
sification models [12,13], since they permit performance References
to be expressed in terms of meaningful criteria (such
as strength) and commonly-used operating indicators [1] A. Kato and K. Hirose, Ass. Jpn. Portland Cem. Eng., Rev. Gen.
Meet. 23, (1969) 109.
(such as Blaine Index or surface area), as well as size
[2] F.W. Locher, S. Sprung and P. Korf, Zem. Kalk-Gips, 26 (1973)
distribution. 349.
[3] G. Frigione and S. Marra, Cem. Concr. Res., 6 (1976) 113.
[4] L.G. Austin, P.L. Luckie and H.M.V. Seebach, in H. Rumpf
8. List of symbols and K. Schonert (eds.), Proc. Fourth Eur. Syrup. ZerMeinem,
DECHEMA Monogr., 79 Nr. 1576-1588, Verlag Chemie, Wein-
3D 3 d compressive strength (MPa) helm, pp. 519-537.
7D 7 d compressive strength (MPa) [5] K. Kuhlmann, H.G. Ellerbrock and S. Sprung, Zem. Kalk-Gips,
38 (1985) 169 (Translation No. 4/85).
28D 28 d compressive strength (MPa)
[6] M.S. Sumner, N.M. Hepher and G.K. Moir, 1989, Cim. Bdtons,
c~s dicalcium silicate Pldtres, Chaux, 778 (1989) 164.
C~A tricalcium aluminate [7] G. Frohnsdoff and J.R. Clifton, IEEE Cement Industry Tech.
C3S tricalcium silicate Conf., Toronto, Canada, May 19-22, 1980.
C4AF tetracalcium aluminoferrite [8] H.F. Welles, Zem. Kalk-Gips, 35 (1982) 425.
[9] A.M. Neville and J.J. Brooks, Concrete Technology, Longman
DAY time (days)
Scientific and Technical, New York, 1987.
F surface shape factor [10] A.C. Davis, Portland Cement, London Concrete, 1934, p. 1-17.
F28 28 d concrete strength characteristic (%) [11] L.G. Austin, N.P. Weymount and O. Knobloch, Eur. Syrup.
FCH chemical characteristic for cement Particle Technology, Amsterdam, 1980, pp. 640-655.
N number of data sets used in regression [12] Y.M. Zhang, A. Kavetsky, T.J. Napier-Munn and D.S. Rapson,
Cement and Concrete Association of Australia Tech. Conf., Moss
Om measured surface area, Eq. (1) (cm2/g)
Vale, Aug., 1987.
Os measured surface area, Eq. (4) (m2/kg) [13] Y.M. Zhang, T.J. Napier-Munn and A. Kavetsky, Trans. Inst.
R wt.% retained in Eq. (7) Min. Metall. Section C, 97 (1988) C'207-C~14.
R2 coefficient of determination for regression [14] Anon 1980, Australia standard method of testing Portland and
equation blended cement, fineness index of Portland cement by air
permeability method, AS 2350.8, Standard Association of Aus-
SD standard deviation of regression coefficients
tralia.
$32 cumulative % passing 32 /zm [15] Anon 1988a, Australia standard method of testing Portland
$8 cumulative % passing 8 /zm and blended cement, compressive strength of Portland and
Sm surface area calculated from regression Eq. blended cement, AS 2350.11, Standard Association of Australia.
(1) (cm//g) [16] Anon 1988b, Australia standard method of testing Portland
and blended cement, compressive strength of Portland and
Ss surface area calculated f r o m Eq. (6)
blended cement, AS 1012.2, Standard Association of Australia.
(m2/kg) [17] K. Kuhlmann, 1984, Zem. Kalk-Gips, 37 (1984) 257: Translation
SSA cement surface area (m//kg) No. 9/84.
SE standard error of regression equations [18] G. Herdan, Small Particle Statistics, Elsevier, London, 1953,
STNmix concrete strength by mix test (MPa) p. 413.
[19] E.C. Higginson, 1970, ASTM, Spec. Tech. Publ., 473 (1970)
STNmo,t~ concrete strength by mortar test (MPa)
71-81.
w1 weight percent in size fraction i (Eq. (5)) [20] G.E. Troxell, H.E. Davis and J.W. Kelly, Composition and
x size (cm) Properties of Concrete, McGraw-Hill, New York, 2nd edn., 1968.
Xm geometric mean size (cm) [21] F.W. Locher, J. Wuhrer and K. Schweden, Tonind.-Zeitun~ 90
AQ, weight percent in size fraction, i (Eq. (2)) (12) (1966) 547-554.
(%) [22] W.H. Taylor, Concrete Technology and Practice, McGraw-Hill,
New York, 4th edn., 1977.
density (g/cm3 or kg/m3) [23] A.M. Neville and J.J. Brooks, Concrete Technology, Longman
Scientific and Technical, New York, 1987.
[241 D. Knofel, Zero. Kalk-Gips, 9 (1979) 448-454.
Acknowledgements [25] T. Kojovic, Ph.D. Thesis, University of Queensland, Australia
(1988).
The authors express their appreciation to the staff [26] Y.M. Zhang, Ph.D. Thesis, University of Queensland, Australia
of Adelaide Brighton Cement Ltd and Blue Circle (1992).