You are on page 1of 11

CEMENT STRENGTH VARIABILITY: WHAT CAN THE

CUSTOMER EXPECT AND HOW MUCH DOES IT


AFFECT HIM?

Steve Crosswell
PPC Cement, PO Box 268, Milnerton, 7435.
Tel: +27 (0) 21 550 2183. E-mail: scrosswell@ppc.co.za

ABSTRACT

Many cement users have unrealistic expectations regarding the consistency of cement
strength and have drawn invalid conclusions regarding the effects of cement strength
variability on their concrete strengths. This in turn can lead to unnecessary customer
complaints and claims, which in turn can damage supplier/customer relationships.

In order to assess what the user may realistically expect, the compressive strength
results of a CEM I 42,5N cement from one factory over a three-year period from 2001 to
2003 were analyzed.

Strengths were measured using the ISO 196 mortar prism method and a modified version
of the concrete cube test described in the now superseded BS 12 standard.

Daily despatch samples were tested at ages of 2 and 28 days using the mortar prism
method and companion samples were tested at 7 and 28 days using the concrete cube
method (SANS 50863). In addition mortar prism tests were carried out at 2, 7 and 28 days
from grind samples from both mills.

The strength results are compared and an attempt is made to assess the effects of the
cement strength variability on the variability of the customers’ concrete strengths.

It was concluded that the ISO mortar prism test is not a good predictor of cement’s
performance in concrete and that test data produced by the factory could not simply be
transferred to the concrete strength variance problem. The variance inherited by the
concrete producer from the cement manufacturer averaged 1,6 MPa2 for the period under
review.

1. INTRODUCTION

A number of cement users have unrealistic expectations regarding the consistency of cement
strength and draw invalid conclusions regarding the effects of cement strength variability on
their concrete strengths. This can lead to unnecessary customer complaints and claims, which
in turn can damage supplier/customer relationships.

Paper presented at the Conference: Developing Concrete to Serve Practical Needs Hosted by:
ISBN Number: 1-920-01717-8 The Industrial Development Engineers Association (IDEA)
13 – 14th October 2004, Midrand, South Africa. The Cement and Concrete Institute (C&CI)
CD produced by: Document Transformation Technologies 163 The Concrete Society of Southern Africa (CSSA)
In order to assess what the user may realistically expect, and how cement strength variability
affects him, the compressive strength test results of one cement, from one factory, over a
three-year period were analyzed.

This paper discusses:


• Cement plant characteristics (plant specific)
• Cement plant operating and quality control parameters
• Test program description
• Test results and analysis
• Discussion
• Conclusions

2. CEMENT PLANT CHARACTERISTICS

The cement plant is located at Riebeeck West some 80 km north of Cape Town. The plant has
two long coal-fired dry process kilns, RK 1 (Riebeeck Kiln 1) and RK 2 (Riebeeck Kiln 2),
installed in 1959 and 1968 respectively. RK 1 has no pre-heater while RK 2 has a single stage
pre-heater. Both kilns were supplied by FLS and have clinker outputs of 760 and 850 tpd
respectively. Clinker from both kilns is stored in a common clinker store.

The plant has two finish mills, FM 1 and FM 2, installed at the same time as RK 1 and RK 2
respectively. Both mills produce CEM I 42,5N while mill 2 also produces CEM I 42,5R. Mill
capacities (for CEM I 42,5N) are 35 and 55 tph respectively. Two different processing or grinding
aids are used, one for each type of cement.

Figure 1. General view of plant, Kiln 2 in foreground.

This plant supplies all of the CEM I 42,5N (trade name “OPC”) in the Western Cape and the
western parts of the Northern Cape. Some CEM I 42,5 R (trade name “RAPO”) is supplied in
bulk to alkali-sensitive customers, and the plant also supplies all of the bagged RAPO in the
Western Cape.

In contrast to most plants, despatch samples of OPC are taken daily rather than the normal
twice weekly, and, also in contrast to most plants, concrete strength tests are also carried out on
these samples.
164
3. CEMENT PLANT OPERATING AND QUALITY CONTROL PARAMETERS

As well as the requirements of SANS 50197-1[1] (SABS EN 197-1), the plant had the following
output targets:
• Target 28-day mortar prism strength window = 50 ± 1 MPa.
• Standard deviation (previous 104 results) ≤ 2,5 MPa (subsequently lowered to 1,5 MPa)
• Sodium oxide equivalent ≤ 0,6%
• C3A < 8% (for the latter part of the period under review)

4. TEST PROGRAMME

This particular concrete test programme was started in February 1998, partly as a result of
complaints from customers that the ISO 196 –1[2] mortar prism tests did not reflect the
performance of the cement in concrete. There were also complaints regarding consistency of
strength.

It was decided that the sampling frequency would be increased from twice weekly to daily, and
that concrete tests would also be carried out on the same daily despatch samples. The daily
samples were split at the factory and half sent to the concrete laboratory in Cape Town. The ISO
196 strength tests were carried out at the factory by multiple operators, while the concrete
strength tests were carried by a single operator at the PPC Sales and Marketing Laboratory in
Montague Gardens, Cape Town. This is statistically important as one has to account for
multi-operator variance, or test reproducibility as well as test repeatability. The ISO 196 tests
were carried out at 2 and 28 days, with 7-day results being determined once per week. (Mortar
prism tests are also carried out on a daily basis on grind samples from both mills and these
tests are carried out at ages of 2, 7 and 28 days – these results are not considered in this
paper).

The concrete mixes were made with commercially available 19-mm greywacke and
Klipheuwel/Malmesbury pit sand, both supplied by Lafarge Aggregates. These aggregates were
selected as they are representative of the aggregates used in the greater Cape Town area.
Three water/cement ratios were used initially (0,5 0,6 and 0,75) but by 1999 it was decided to
standardize on a water/cement ratio of 0,60 as it was, and is, the same water/cement ratio used
by Lafarge Readymix (a major customer) for their routine cement tests. The concrete strengths
were measured at 7 and 28 days using 100-mm cubes [3].

Concrete cubes were made, cured and tested in accordance with the relevant SANS standards
and the Montague Gardens laboratory was accredited by SANAS to ISO 17025 in 2003.

The Riebeeck West factory laboratory operates in accordance with ISO 9001 and is audited by
the Cement and Concrete Institute and the PPC Group Laboratory Services laboratory, both of
which are also SANAS accredited.

For the purposes of this paper it was decided to focus on the period from January 2001 to
December 2003 as this period had the most complete data.

165
5. TEST RESULTS AND ANALYSIS

5.1 Compressive Strength Results

The compressive strength results are plotted in Figure 2 below.

ISO and concrete 28-day compressive strengths

70

60

50

40
MPa

30

20

10

0
03/01/2001
03/02/2001
03/03/2001
03/04/2001
03/05/2001
03/06/2001
03/07/2001
03/08/2001
03/09/2001
03/10/2001
03/11/2001
03/12/2001
03/01/2002
03/02/2002
03/03/2002
03/04/2002
03/05/2002
03/06/2002
03/07/2002
03/08/2002
03/09/2002
03/10/2002
03/11/2002
03/12/2002
03/01/2003
03/02/2003
03/03/2003
03/04/2003
03/05/2003
03/06/2003
03/07/2003
03/08/2003
03/09/2003
03/10/2003
03/11/2003
03/12/2003
28 day ISO strength 28 day concrete strength corrected (moulds)
30 per. Mov. Avg. (28 day ISO strength) 30 per. Mov. Avg. (28 day concrete strength corrected (moulds))

Figure 2. Concrete and ISO 28 day test results.

Figure 2 shows the mortar and concrete results converging from Jan 2001 until mid-March
2002. After which they started to diverge until December 2002/ January 2003 when there was a
drop in strength of some 4 MPa in the mortar results and nearly 5 MPa in the concrete results
over a three-month period. The strengths then increased and converged slightly. In both cases
the average strengths varied over a range of approximately 5 MPa over the three-year period.

Figure 3 below is the Cusum [4] chart of the same data plotted about the running average of 30
results. With the exception of an anomalous period in late 2001/early 2002 the correlation
between the two trends is extremely good. This shows that both test regimes detected the same
changes in strength trends. The value of plotting the trends in this way is that it is possible to
identify the exact time when significant changes occurred, for example at mid-March and
mid-November 2002, and the magnitude of the changes. For example the change in strength at
mid-March 2002 was an increase of approximately 1,5MPa.

166
Cusum - ISO and concrete 28 day strengths

20

Anomalous period
0
03/01/2001
03/02/2001
03/03/2001
03/04/2001
03/05/2001
03/06/2001
03/07/2001
03/08/2001
03/09/2001
03/10/2001
03/11/2001
03/12/2001
03/01/2002
03/02/2002
03/03/2002
03/04/2002
03/05/2002
03/06/2002
03/07/2002
03/08/2002
03/09/2002
03/10/2002
03/11/2002
03/12/2002
03/01/2003
03/02/2003
03/03/2003
03/04/2003
03/05/2003
03/06/2003
03/07/2003
03/08/2003
03/09/2003
03/10/2003
03/11/2003
03/12/2003
-20
MPa

-40

-60

-80

-100

Cusum concrete Cusum ISO

Figure 3. Cusums, concrete and ISO 28-day strengths.

5.2 Predictability of Concrete Strength

In view of the similarity of the cusum curves it was decided to check whether ISO prism test
results were good predictors of concrete cube strength. Figures 4 and 5 below indicate that they
are not particularly good. Figure 4 shows the concrete strength divided by the mortar strength
expressed as a percentage. Figure 5 shows the difference between the observed concrete
strength and notional concrete strength predicted from the observed mortar strength and the
average strength ratio of the two sets of test results over the previous 30 results.

Analysis of the differences in the predicted strength results gave a standard deviation of the
difference of 2 MPa, which means that the predicted concrete strength would be within 3.3 MPa
of the measured strength 90% of the time. A similar calculation for the relationship between
7 and 28-day concrete strength results gave predictions within 2.2 MPa for 90% of the time.

167
%
MPa

-8
-6
-4
-2
0
2
4
6
8
60.0
70.0
80.0
90.0
100.0
110.0
03/01/2001
03/01/2001
03/02/2001
03/02/2001
03/03/2001
03/03/2001
03/04/2001
03/04/2001
03/05/2001
03/05/2001
03/06/2001
03/06/2001
03/07/2001
03/07/2001
03/08/2001
03/08/2001
03/09/2001
03/09/2001
03/10/2001
03/10/2001
03/11/2001
03/11/2001
03/12/2001
03/12/2001
03/01/2002
03/01/2002
03/02/2002
03/02/2002

28 day concrete/28 day ISO


03/03/2002
03/03/2002
03/04/2002
03/04/2002
03/05/2002

168
03/05/2002
03/06/2002
03/06/2002
03/07/2002
03/07/2002
03/08/2002
03/08/2002
03/09/2002
03/09/2002
03/10/2002
03/10/2002
03/11/2002
03/11/2002
03/12/2002

Difference in predicted concrete strength


03/12/2002
03/01/2003

Difference in predicted concrete strength


03/01/2003
03/02/2003
Concrete strength/ISO strength (%)

03/02/2003
03/03/2003
03/03/2003
03/04/2003
03/04/2003
03/05/2003
Figure 4. 28-day concrete strength as a % of ISO strength.

03/05/2003
03/06/2003
03/06/2003
30 per. Mov. Avg. (28 day concrete/28 day ISO)

03/07/2003
03/07/2003

the running average of the ratio between mortar and concrete strengths.
03/08/2003
03/08/2003
03/09/2003
03/09/2003
03/10/2003
03/10/2003
03/11/2003
03/11/2003

Figure 5. Difference between observed concrete 28-day strength and strength predicted from
03/12/2003
03/12/2003
MPa
MPa

0
0.5
1
1.5
2
2.5
3
3.5
0
0.5
1
1.5
2
2.5
3
3.5
03/01/2001
03/01/2001
03/02/2001
03/02/2001
03/03/2001
03/03/2001
03/04/2001
03/04/2001
03/05/2001
03/05/2001
03/06/2001 03/06/2001
03/07/2001 03/07/2001
03/08/2001 03/08/2001
03/09/2001 03/09/2001
03/10/2001 03/10/2001
03/11/2001 03/11/2001
03/12/2001 03/12/2001
03/01/2002 03/01/2002
03/02/2002 03/02/2002
1.3 Consistency - Standard Deviations

03/03/2002 03/03/2002
03/04/2002 03/04/2002
03/05/2002 03/05/2002

169
03/06/2002 03/06/2002
03/07/2002 03/07/2002

SD (30) Concrete
SD (30) ISO
03/08/2002 03/08/2002
03/09/2002 03/09/2002
03/10/2002 03/10/2002

03/11/2002 03/11/2002

03/12/2002 03/12/2002

03/01/2003 03/01/2003
03/02/2003

SD (104) ISO
03/02/2003
results over 30 successive results and over 104 successive results.

SD (104) Concrete
03/03/2003
Standard Deviations of Compressive Strength Test Results

03/03/2003

Standard Deviations of Compressive Strength Test Results


03/04/2003 03/04/2003

03/05/2003 03/05/2003

03/06/2003 03/06/2003
03/07/2003
03/07/2003
03/08/2003
03/08/2003

Figure 7. Standard deviations of concrete results over 30 and 104 results.


03/09/2003
03/09/2003
03/10/2003
03/10/2003
Figure 6. Standard deviations of ISO mortar prism results over 30 and 104 results.

03/11/2003
03/11/2003
03/12/2003
03/12/2003
The following four figures show the standard deviations (SD’s) of the mortar and concrete test
MPa
MPa

0
0.5
1
1.5
2
2.5
3
0
0.5
1
1.5
2
2.5
3
3.5
03/01/2001 03/01/2001
03/02/2001 03/02/2001
03/03/2001 03/03/2001
03/04/2001 03/04/2001
03/05/2001 03/05/2001
03/06/2001 03/06/2001
03/07/2001 03/07/2001
03/08/2001 03/08/2001
03/09/2001 03/09/2001
03/10/2001 03/10/2001
03/11/2001 03/11/2001
03/12/2001 03/12/2001
03/01/2002 03/01/2002
03/02/2002 03/02/2002
03/03/2002 03/03/2002
03/04/2002 03/04/2002

SD (104) ISO
03/05/2002 03/05/2002

170
03/06/2002 03/06/2002

SD (30) ISO
03/07/2002 03/07/2002
03/08/2002 03/08/2002
03/09/2002 03/09/2002
03/10/2002 03/10/2002

03/11/2002 03/11/2002
03/12/2002 03/12/2002

SD (104) Concrete
03/01/2003 03/01/2003
SD (30) Concrete

03/02/2003 03/02/2003
03/03/2003
Standard Deviations of Compressive Strength Test Results

03/03/2003
03/04/2003 Standard Deviations of Compressive Strength Test Results 03/04/2003

03/05/2003 03/05/2003

03/06/2003 03/06/2003
Figure 8. Standard deviations, ISO and concrete over 30 results.

Figure 9. Standard deviations ISO and concrete over 104 results.


03/07/2003 03/07/2003

03/08/2003 03/08/2003

03/09/2003 03/09/2003

03/10/2003 03/10/2003

03/11/2003 03/11/2003
03/12/2003
03/12/2003
Two distinct trends are apparent, neither of which are intuitive:
• The standard deviation calculated over 30 results is lower than that calculated over 104
results for most of the time for both mortar and concrete, but the variability of the standard
deviation is higher.
• The standard deviation for the concrete results is generally lower than that for the mortar
results for both 30 and 104 results.

The reason for the first is believed to be that the longer-term analysis detects long term
seasonal changes.

The reason for the second is thought to be that, despite the added complication of more
variable aggregates in the concrete test, those tests were carried out by a single operator while
the mortar prism tests were carried out by multiple operators.

Further statistical analysis of the standard deviations is shown in table 1 below which confirms
the above-mentioned trends:

Table 1. Summary of standard deviation data.


ISO mortar results: moving Concrete results: moving
average of average of
104 results 30 results 104 results 30 results
Average standard deviation (MPa) 2.06 1.93 1.95 1.74
Standard deviation of the SD (MPa) 0.31 0.44 0.29 0.39
Upper 95% limit (MPa) 2.68 2.81 2.53 2.52
Lower 95% limit (MPa) 1.44 1.05 1.37 0.96

6. DISCUSSION

The differences in the standard deviations tabulated above are important as they clearly show
that both the test method and the period over which the SD is calculated affects the result.
Generally speaking a cement producer calculates the standard deviation of his results over 104
successive mortar test results whereas a concrete producer determines his standard deviation
over 30 successive concrete tests using a different test method.

From the results in section 5 above it is worthwhile attempting to quantify the variability in
strength that the concrete producer inherits from the cement producer. It is therefore necessary
to introduce the term variance [5]. Statistically variance is equal to the square of the standard
deviation and in this case has units of MPa2. The advantage of using variance is that variances
are additive, while standard deviations are not, and the total variance is equal to the sum of the
independent variances which contribute to it.

In this case the cement producer would be reporting an average standard deviation of 2.06 MPa
while the concrete user would only detect an average standard deviation of 1.74 MPa if his
standard of testing were as good as the cement producer’s.

If the observed SD is 1,74 MPa, the variance is therefore1.74 2 or 3.03 MPa2. The testing
variance has to be subtracted from this and this is estimated to be of the order of 1.22 or
1.44 MPa2. The inherited variance is therefore 3.03 – 1,44 = 1.6 MPa2, equivalent to a standard
deviation of 1,26 MPa.

171
To illustrate the effect of variance, if a concrete producer has a standard deviation of 4 MPa then
his variance is 16 MPa2, of which 1.6 MPa2 is inherited from the cement, the rest being made up
from materials, batching, sampling and testing variances.

7. CONCLUSIONS

The following conclusions were reached:


• Cusum analysis is a very useful method for detecting the time and magnitude of strength
changes. (It can also be used to analyze standard deviation values and other parameters).
• ISO 196 mortar prism test results are not a particularly good predictor of a cement’s
performance in concrete. Seven-day concrete test results are a better predictor of 28-day
concrete strengths.
• The strength and variability information supplied by the cement plant cannot be simply
transplanted into the concrete variance problem. Account has to be taken of sampling and
testing variance, operator variance, and analysis period.
• On average the concrete producer inherited approximately 1.6 MPa2 of his variance from the
cement over the period under review. If his standard deviation were 4 MPa, this would
account for only 10% of his total variance.

8. REFERENCES

[1] SABS EN197-1, Cement. Part 1: Composition, specifications and conformity criteria
for common cements, Pretoria: South African Bureau of Standards, 2000.
[2] SABS 196 -1, Methods of testing cement. Part 1: Determination of strength, Pretoria:
South African Bureau of Standards, 1994.
[3] SABS method 863, Concrete tests – compressive strength of hardened concrete,
Pretoria: South African Bureau of Standards, 1994.
[4] Cement and Concrete Institute, Monitoring concrete strength by the cusum system,
Midrand: The Institute, 1996.
[5] Underhill, L. and Bradfield, D., Introstat, 2nd ed. Cape Town: Juta & Co Ltd, 1996, pp.29-33.

172
CEMENT STRENGTH VARIABILITY: WHAT CAN THE
CUSTOMER EXPECT AND HOW MUCH DOES IT
AFFECT HIM?

Steve Crosswell
PPC Cement, PO Box 268, Milnerton, 7435.
Tel: +27 (0) 21 550 2183. E-mail: scrosswell@ppc.co.za

Biography

Steve Crosswell
Pr Eng MICT

Steve Crosswell was educated at St Mary’s Grammar School, Basingstoke, England, and SACS
in Cape Town. He graduated in Civil Engineering from the University of Cape Town in 1975 and
gained a Graduate Diploma in Engineering, also from UCT, in 1984.

He worked for the Roads Branch of the Cape Town City Council, from 1975 until 1981, where he
gained experience in the design and construction of major roads and the construction of
concrete canals and bridges.

In 1981 he joined the Portland Cement Institute (now the Cement and Concrete Institute) as
Regional Engineer for the Western Cape where he spent the next 15 years, leaving as Regional
Manager in 1996.

In 1996 he joined PPC Cement as Technical Support Manager for the Western Cape region.

He has been a member of the Concrete Society of Southern Africa since 1981 serving as
Honorary Branch Treasurer in the Western Cape for a number of years. He is a registered
Professional Engineer and a Member of the South African Institution of Civil Engineering and
the Institute of Concrete Technology.

173

You might also like