Proctologically Violated©® wrote:
Well, I'm just genuinely curious:

Spose I'm making a part.
I know, from general principle, that the *least* amount of a part requires
makes for a cheaper part.
So I determine just how much slop I can get away with.

Spose, through consultation, trial and error, etc., I determine that the ..25
dia.pin I'm making can "tolerate" a maximum tolerance of +/-.005, but
"feels" a little better w/ +/- .003 tolerance, altho production might be a
little slower at .003.

It is now a judgment call to decide whether to go w/ .005, or .003

How would statistics, in any way, have helped me in this
determination/judgment call?


Mr. P.V.'d
formerly Droll Troll


Well if you are really trying to decide, you probably use some
statistics to determine not just the tolerance of the pin, but also
the tolerance of the hole it fits into. Then assuming that the
distribution of the parts is Gausian, you can figure out how loose the
tolerances can be and still have an acceptable assembly rate. You
might set the tolerances so that once in say 3000 times the pin will
not work with a hole. It might be cheaper to throw away a pin once in
a while than to make closer tolerance pins.

If machinest don't need statistics, why does J & L sell so many
measuring tools with a spc output port? And why does J & L sell
statistical control software?

Engineering isn't just designing something so it works. Any damn fool
can do that. An engineer designs it so it works and costs less.

Dan