Testing - Concrete Mix Design, Quality Control and Specification - page 137

at higher strengths. However, a further reason is that bleeding voids, which
are more likely at lower strengths, may have a greater effect on cubes than
cylinders owing to the different orientation during testing.
7.3.3 Bad concrete or bad testing?
Day was invited to give a paper on the aforementioned topic to the 1989
ACI San Diego Convention (Day, 1989). The paper has not been published
(it is however now on Day's website), but the conclusions presented and the
fact that an ACI session organiser requested a paper on this topic indicate
that the question merits close attention.
The first half of the paper presented factual data showing that it is unrea-
sonable to expect that a properly presented result from a reputable testing
laboratory will always be an accurate representation of the quality of the con-
crete. Pair differences exceeding 5 MPa were noted for apparently identical
test specimens from the same truck of concrete tested by the same laboratory.
Seven- to 28-day strength gains were also shown to be capable of ±50% from
sample to sample of concrete of the same mix design using the same materials.
The clear conclusion was that a strength test result was a totally unreliable
piece of information. The audience awaited Day's proposal of some more sat-
isfactory means of assessing concrete quality than a compression test.
The second half of the presentation showed that the very same data used
in the first half could be analysed to show quite accurately when a genuine
change in concrete quality occurred. Cusum graphs of 7- and 28-day
strength showed downturns and upturns on exactly the same dates in spite
of individual differences. The two laboratories showing the large differ-
ences on individual samples nevertheless agreed exactly as to when these
change points occurred.
The overall conclusion presented was that an appropriate analysis
of a series of test results can yield very reliable conclusions but that any
individual test result should be regarded with great suspicion. Some of
other conclusions presented were as follows:
1. Concrete producers are not so good that it is unnecessary to test
concrete nor testing labs so bad that it is ineffective to do so.
2. There is no better complete replacement for traditional cylinder/cube
testing because it is the only way in which the combined effects of
batch quantity variation, material quality variation, silt and dust
content variation, air content and temperature variations, delivery
delays, and added water effects can be integrated.
3. We must cease to think of a single test result as an invariably accurate
judgment as to whether a particular truck of concrete is acceptable.
First, it may well not be accurate, and, second, we should show as
much concern for those trucks we did not test as for those we did test.