Quality control - Concrete Mix Design, Quality Control and Specification - page 240

Testing is normally one at 7 days and two at 28 days. Some operators
tend to keep the second of the two 28-day results for test at 56 days if the
first is below the anticipated strength. In contrast to this, our recommenda-
tion is to bring forward one of the 28-day specimens for test at 7 days if the
7-day specimen gives an unexplained low result. It is a matter of determin-
ing a course of action rather than making a second attempt to obtain an
acceptable result.
Some operators may have taken the precaution at an early stage of mak-
ing one or more additional specimens to establish a conversion factor to
predict 7-day results from 3-day results. If so, it is an even more satisfactory
solution to bring forward one or more intended subsequent 7-day results
for test at 3 days to see whether an apparent shortfall is continued.
The use of 7 days as an early-age test is a consequence of many laborato-
ries only operating 5 or 6 days per week. If 7-day laboratory operation is in
force, 3 days is an equally satisfactory early-age test (although more atten-
tion is required to it being at close to 72 hours and to curing temperature
having been under good control). Control can also be based on testing at
24 hours or less, but in this case it is strictly necessary to insert a thermal
probe to establish the exact Arrhenius “equivalent age” as explained in
Chapter 7.
There is only one effective answer to “detecting any change in the qual-
ity of the concrete at the earliest possible moment” and that is by cusum
analysis. As already noted, the strength of a particular grade of concrete
may not be the most important requirement, but strength is the most effec-
tive parameter for the detection of a change in quality.
10.5 NORMAL DISTRIBUTION
The assumption is normally made that the results for any individual grade
will form a normal distribution. We have found this to be a reasonable
assumption, except that the percent below mean -1.65 SD is often 3 or 4
rather than the anticipated 5. This may be due to control action on extreme
high slump.
What is of more importance is the occurrence of double-peaked or exces-
sively skew distributions. A double peak is usually a sign of two different
distributions being combined. When this occurs an attempt should be made
to separate the two sources as one may be of inadequate strength, which is
masked by combination with the other.
If a distribution is skewed on the high side, it may be that low results are
being withheld and this possibility should be followed up. If a distribution
is skewed to the low side, it may be that the coarse aggregate has limit-
ing either strength or bond characteristics. Alternatively it may be that the
operator is afraid of explosive failure or wishing to avoid cleaning up after