Statistical analysis - Concrete Mix Design, Quality Control and Specification - page 227

relative performance on different grades of concrete. The argument
resurfaces from time to time, even though in Day's opinion general agree-
ment that standard deviation should be used was reached in the 1950s. The
authors have personally monitored thousands of test results covering 20,
25, 30, 40, and 50 MPa grades of concrete from the same plant over long
periods of time. There has never been any question in Day's mind that stan-
dard deviation remains reasonably constant over the 20 to 40 MPa grades
(i.e., mean strengths from 25 to 45 MPa or 3600 to 6500 psi). This opinion
was formed in the early 1950s when he consistently achieved a standard
deviation of less than 250 psi (1.7 MPa) on very tightly controlled factory
production with a mean strength in excess of 9800 psi (6.7 MPa). This
was certainly abnormal concrete produced in tiny quantities and, being of
earth-dry consistency, visual water control was very easy. However if this
figure is expressed as a coefficient of variation of less than 3%, it would
represent a standard of uniformity impossible to achieve on concrete of
normal strength, even under laboratory conditions.
This firm opinion, even allowing for the quoted high-strength expe-
rience, must be tempered by an acknowledgement that a slightly higher
standard deviation is normally experienced on 50 MPa and higher grades.
This appears to be largely due to the greater difficulty in achieving accu-
rate testing, perhaps in turn due to the different mode of failure of higher
strength concrete (where bond failure, or even aggregate failure, rather
than matrix failure tends to be experienced). The increase in both average
pair difference of specimens and overall concrete standard deviation is of
the order of 0.5 to 1.0 MPa.
Since publication of the first edition interesting further evidence is on
hand. The Petronas Towers in Kuala Lumpur, Malaysia, project (at that
time the world's tallest building,) involved more than 40,000 cubic metres
of 80 MPa grade concrete. Being under a UK type specification, this
required a mean strength of approximately 100 MPa (cube, at 56 days). It
can be imagined that in view of the importance of the project, the initial
concrete supply was at a conservatively high mean strength of just over
110 MPa. This caused the overall standard deviation for the whole of the
632 samples tested at 56 days to be inflated to 4.7 MPa. However when
things had settled later in the project, a run of 237 consecutive results gave
a standard deviation of 2.8 MPa with a mean strength of 99.3 MPa.
An even lower SD value of 2.6 MPa on 80 MPa concrete for the
Chateaubriand bridge is reported (de Champs and Monachon, 1992).
Set against these figures are the decisions of ACI Committees 211
(Mixture Proportioning), 214 (Evaluation of Test Results), and 363 (High
Strength Concrete) to adopt coefficient of variation as the meaningful
index of variability. The leading advocate of this view was Jim Cook, but
of course the decision was from the committees as a whole. Day (1998) sug-
gests that high-strength concrete offers more scope for increased variability