Notes_on_data_accuracy
Notes_on_data_accuracy
This brief will attempt to give guidance on the accuracy of material data reported in various sources.
There are three main sources for the scatter in reported property values.
The variation due to measurement error and sample preparation can be seen in round-robin studies.
These studies typically involve a large amount of material from one batch being divided up and sent
to several different laboratories for testing. All laboratories should follow the same testing standard.
The magnitude of this error should be the smallest possible. Also note that laboratories which take
part in these types of studies are probably the most careful, not all labs will be as careful or as well
maintained.
The variation due to composition variation will depend on the specifics of the material being tested.
Some metals have wide composition ranges, some pure metals are very sensitive to impurities,
some are stable at RT but age or anneal at high temperatures. If the material is not stable over the
measurement range, then a repeat measurement will give different results. The glass phase in some
ceramics is unstable at high temperatures. Many polymers are sensitive to moisture and to the mold
geometry. Glasses and polymers are sensitive to the cooling rate. The thermal, electrical, and
magnetic properties of metal at low temperatures are very sensitive to composition and impurities,
especially for the purer alloys, and the spread in values will be correspondingly larger than at RT.
The size of the test sample can also have an effect due to texture or different cooling rates for the
inner and outer material.
Heat treatments are frequently ill-defined. They may say “annealed” or “aged” where the exact time
and temperature are not specified, or a range of temperature/times may be given. Annealing
conditions can change from manufacturer to manufacture or from year to year but the material just
described as “annealed”.
All values given for the spread in the individual properties are approximate and are meant to guide
your expectations. I’ve never seen error bars on an FEA simulation.