Quality control - Concrete Mix Design, Quality Control and Specification - page 234

and reporting rather than the quality of his concrete. Exercises of this type
are reported on Day's website, where it is also reported that the laboratory
of another producer can be used as the “independent” laboratory. Where
the early-age results of the two labs differ, it is easy to arrange to witness
the actual 28-day tests.
10.2 MULTIGRADE CUSUM
Another concept introduced by Day is that of a multigrade cusum. Any
target can be chosen for a cusum, but if the target chosen is the continu-
ously updated mean of the item being cusummed, this places the emphasis
on the detection of change rather than adherence to a target. It is perhaps
not obvious that it is reasonable to combine a strength divergence from the
current mean of a 20 MPa mix with that from a 100 MPa mix, but Day
tried it 30-plus years ago and it works. This concept goes far beyond the
“families” concept of EN206 (and is much easier to apply) and enables all
test data on every mix in use (including lightweight aggregate [LWA] and
normal concrete, self-compacting, and no slump) to appear on the same
cusum graph. So long as the cusum graphs remain essentially horizontal,
all concrete is under control. When a change point is seen, the other cusums
are examined to see if they provide an explanation. If they do not, then a
further concept is implemented, that of “groups”.
10.2.1 Groups
The mixes in a multigrade cusum need to be collected into groups with some-
thing in common and not shared by other grades of concrete. One grade can
be a member of several groups. It may be that some use a different sand,
a different cement replacement material, or a different admixture. It could
also be that some are only supplied to a distant site or to a particular pump
known to give trouble from time to time (Pareto). The analysis program must
make it easy to call up each group in turn; even the simple free program does
this. The technique when a change is seen without a clear explanation is to
try selecting only the results from particular groups to see which are affected
and which not. This technique is particularly useful with the free program
KensQC, which has only a very limited range of variables in the basic system.
10.3 MONTHLY PRINTOUT
At the end of each month, or at any other selected time, even free programs
print out an analysis of all the mixes in use. Desirably this should be in
order not of grade number but of departure from the target strength.