Statistical analysis - Concrete Mix Design, Quality Control and Specification - page 229

current structure. Therefore a method of closely monitoring the situation
and taking early action to revert to the desired quality is very desirable.
This used to mean keeping a graph known as a Shewhart QC chart, how-
ever these have been superseded by cusum control charts, such as the
ConAd system.
As we have seen, a substantial error is possible in assessing the stan-
dard deviation, mean, and 5% minimum of a small group of results, so
that they cannot be used with any degree of fairness to reject or penalise.
Nevertheless more than 50%, perhaps as much as 70% or 80%, of such
assessments are quite realistic. They are therefore very useful as a guide to
the state of affairs provided they are used only as a warning that the situa-
tion should be carefully considered and not as a basis for precipitate action.
Having isolated the rigid legal requirement as based on an unquestionably
accurate assessment of a large quantity of results, it is then possible to
informally consider a large number of factors in deciding when a small
mix adjustment may be desirable. There will be scope for a small difference
of opinion between concrete producer and supervisor from time to time,
but the latter can afford to concede graciously and wait for the fullness of
time to bring retribution if it was merited, secure in the knowledge that the
quality shortfall will be minor and the retribution precise, inevitable, and
indisputable.
A very interesting matter is a comparison of the standard deviations con-
sidered normal in Australia and the United Kingdom. Day has for many
years considered 3 MPa (say 450 psi) to be a normal figure for an average
ready-mix plant in Melbourne. Of recent years the better practitioners are
attaining 2 MPa or even fractionally less. In the United Kingdom, a figure
of 4 to 6 MPa is considered normal. It is not likely that physical control
of production is genuinely twice as good in Australia and an explanation
is likely in the statistical concepts applied. In the United Kingdom, results
are corrected or normalised according to cement content so as to provide
a basis for combining results from different grades. It would appear that
this does not work very well. Having created an artificially higher vari-
ability in this or some other manner, the task of detecting change becomes
more difficult. When a rigid mathematical requirement (in the form of a
V mask) is applied to determine whether an adjustment should be made,
the difficulty is compounded. When adjustment is delayed in this manner, a
genuinely higher variability is created or allowed to continue. This question
is further examined in Chapter 11.