Unchanging concepts - Concrete Mix Design, Quality Control and Specification - page 260

exceed the specified strength by at least 3N/mm 2 (3 MPa). This investigation
has shown that unless such adjustments are expressed in terms of a multiple
of the standard deviation, the systems would give a substantially different
relative performance according to whether the production was at high or
low variability. Another aspect of system efficiency is the use of multiple
criteria. A system can be made to give a better ratio of correct detections
to false alarms by composing it of several subsystems running in parallel.
In this case the better performance is obtained at the cost of a more com-
plicated criterion, a larger program, and slower operation. With computer
assessment, these costs would be negligible compared to increasing physi-
cal testing frequency and it should be realised that a more efficient analysis
system has as much value as additional testing. For example it would be
possible to analyse results using a combination of all the systems and to
accept that a downturn had occurred when one was detected by any two,
or any three, of the nine systems shown. This would no doubt give both a
better detection rate and less frequent false alarms. However, the improve-
ment would probably be relatively small since false alarms are frequently
due to aberrations in the results affecting several systems rather than to
aberrations in one of the detection systems. (In this respect it would be of
value to people involved in concrete QC to examine a selection of the data
generated for this investigation in order for them to realise the extent to
which apparently convincing downturns in a set of results occur as a result
of normal statistical variation.)
The real reason militating against the multiple criteria approach is that
they must still be suitable for the average user. Complication must be
avoided as far as possible, both to ensure comprehension by all concerned
in their enforcement and to avoid the much greater effort of examining
compliance by manual calculation by persons not having computer
knowledge or facilities.
11.3.1 Relative performance of the systems
All the systems, except ACI 214, are nominally directed toward assuring
a characteristic strength that 95% of results will exceed. Therefore, that
characteristic strength is given by the mean minus 1.645 times standard
deviation, that is for this exercise, 35 - 1.645 SD.
In the case of ACI 214 the requirement is for only 90% of results to exceed
the specified strength. Therefore that strength in this exercise becomes
35 - 1.28 SD. However in the adjusted limit section, the ACI system is still
comparable as what is reported is the amount of adjustment required.
It can be seen that both the ACI and the UK systems, in their origi-
nal forms, give rapid detection of a downturn but also give a high rate
of false alarms. The Australian system on the other hand appears unduly
lenient. The numerical cusum was adjusted to comply with the 70/80 false