Bruce writes:

On Sun, 28 Mar 2010 12:43:51 +0200, (Martin
Schöön) wrote:

snip

And the accuracy was stated as? All material data I come across
at work is within +/- something. It is hugely important to make sure
material data used for engineering calculations are for the stuff
coming out of production and not from some lab. Material data should
be for relevant ambient conditions, temperature, humidity or whatever
applies for the intended application.

It is?
Here is a section of a specification sheet: "The standard requirements
for ASTM A516 physical and chemical characteristics are given in the
tables below."

Mechanical Properties:


A516 Grade 60 A 516 Grade 65 A16 Grade 70
Tensile Strength (ksi) 60-80 65-85 70-90
Tensile Strength (MPa) 415-550 450-585 485-620
Yield Strength (ksi) 32 35 38
Yield Strength (MPa) 220 240 260
Elongation in 200mm (%) 21 19 17
Elongation in 50mm (%) 25 23 21
Max Thickness (mm) 205 205 205


The supplier is Oakley Steel,
" a specialist steel supplier focusing in boiler and chrome moly steel
plates. We stock pressure vessel quality plate in ASME and ASTM grades
principally for use in the oil and gas industry."

Hard to find a mention any tolerance, any +/-.

No, in your example I find that right away for tensile strength. If
you find it hard to get the information it is either because you use
the wrong suppliers or because you are not important to them.

Having written that I think I have to point out that published
data sheets seldom are really helpful because the tolerances found
there are cooked up by the sales department... you have to go to the
next level in most cases.

snip

I have sent quite a number of coupons for testing and processed quite
a few materials certificates, furnished by steel makers, and I have
never seen a tolerance, never; "Tensile strength 50,000 psi +/- 10%".
Every materials certificate or test coupon has listed only the results
of the test as "tensile strength 50,000 psi" without a tolerance.

If I got such an answer I would start looking for another tester.
Someone who knows and acknowledges that test equipment and test
procedures have limited accuracy. Several samples should be tested
since all manufacturing processes have variations.

Trying to insert some boating content:
The other day I revisited a report on the structural modelling of a
racing yacht. One chapter is dedicated to material testing. They
manufactured laminate samples and had them tested at a test institute
to get data for the modelling. The data they got came with a mean
value and a standard deviation number. The standard deviation
for the material parameters of those carbon pre-preg laminates were
roughly 3%. (You see were my skepticism regarding that 90% figure
comes from.)

Back to my rant:
It is *very* important to realize that real world materials and
production are subjected to random variations and modern engineering
must acknowledge that and take it into account. "If you haven't
done a proper yield analysis you aren't done." Yield in this case
is not yield as in yield strength, but you realized that for sure.

http://en.wikipedia.org/wiki/Six_sigma
http://en.wikipedia.org/wiki/Cpk_Index
http://en.wikipedia.org/wiki/Statist...rocess_control
http://en.wikipedia.org/wiki/Design_of_experiments

http://en.wikipedia.org/wiki/W._Edwards_Deming

/Martin (sorry for the late reply, it has been busy times. I have
been studying the impact of mechanical tolerances on a proposed
design, among other things.)