Testing - Concrete Mix Design, Quality Control and Specification - page 140

7.3.4 Rounding results
It is extremely bad practice in any technical field to fail to recognise and
take account of the inaccuracies inherent in test results. One aspect of this
is to avoid expressing results to more significant figures than their accuracy
justifies. In accordance with this various authorities require that certain test
results be rounded. An example is the Australian NATA, which requires
that compression test results be rounded to the nearest 0.5 MPa (= 75 psi)
and densities to the nearest 20 kg/m 3 (= approximately 1 lb/cu ft). Ken
believes that this practice requires reconsideration.
Take compressive strength. Why should 0.5 MPa be selected? The answer
is not that this is the order of accuracy, because different (competent) labo-
ratories can easily differ by 2 MPa and average pair differences can exceed
1 MPa. Rather the answer is that in the days before computers were used,
results were worked out from tables and 0.5 MPa steps gave about as large
a table as was convenient. The tables would have been five times as large
had 0.1 MPa been selected.
The important question is what use is to be made of the test result.
Originally the answer was to accept it as totally accurate and reliable, and
compare it to the specified strength. From this viewpoint it should certainly
be taken as ±2 MPa and so labeled.
It is bad practice to round calculations before the very last step. The
strength of the individual specimen used to be the last step, but now we
have hopefully realised that this should no longer be the case. Action on
compressive strength results should always be based on the analysis of
groups of test results, effectively ignoring individual results. So it is the
mean and standard deviation of a number of results that has significance. It
would be better to use less rounded results, but it may not make a great deal
of difference. However, when analysing (as we should) such items as within
sample ranges (based on average pair differences) and 7- to 28-day strength
growth, rounding to 0.5 MPa is obviously unsatisfactory.
It is proposed for compressive strength that it be expressed to 0.1 MPa
and given the written qualification ±2 MPa where appropriate. This (apart
from the ±2 MPa) will not consume any more paper and will marginally
reduce the computer program.
For density, a similar situation exists. It is not so much the absolute
density of a single specimen that should be of interest, but the range of
densities of all specimens from a single sample of concrete (since this will
reveal the competence of the specimen casting and enable its variation to be
monitored). Detecting any change in the average density of concrete being
produced, that is, of a group of samples, is the major reason for the test.
The proposal for density is that it be expressed as a four-digit integer,
since again this takes marginally less computer effort and no more paper.
The accuracy limits in the case of density may be much different for different