Page 48 - 4150
P. 48
assumption may lead to choosing measuring instruments which are
technologically different or from different manufacturers.
Applications
This method is implemented, in particular, in relation reference
standards: standard rings, standard masses, etc. The reference standard
represents the first link of the calibration chain inside the firm; if it
happens to drift, this may entail serious errors of measurement and,
most importantly, these are undetectable if no duplicating item is
available in the firm. Let us mention, by way of example, reference-
standard rings whose 80 mm diameter has altered by 2 μm in a year;
this alteration is to be compared to the uncertainty on the known
diameter of the ring which was +/- 0.17 μm. The cause of the
alteration was probably a defect in the stabilization of the material.
Second method: checking the coherence of the results
Principle
This method is based on the examination of the measurement
results and the calculation of characteristic values such as the standard
deviation of repeatability, or on drafting graphs, and comparing them
to typical values or to standard graphs.
Hypothesis
This method assumes that instruments of the same nature, or the
same technology, perform more or less similarly. Take, for example,
platinum-resistance temperature sensors; the model of the variation of
the resistance as a function of the temperature is shown by a second
degree polynomial:
R = Ro (1 + at + bt 2)
It is widely accepted that the value Ro can vary between two
thermometers, but the general aspect of the curve remains parabolic,
and a discontinuity in the curve will act as a warning.
Applications
The two following examples illustrate this method:
– An electronic comparator made of a table fitted out with two
inductive sensors is generally used to calibrate standard gauges. A
method can be used to monitor these benches; it consists of testing the
repeatability of the measurements and to compare them to a “typical”
value. The operator is alerted if the values are not similar. Statistical
46