SPC help - how to call out "parallelism"?

Let's say I turn a disc which has to be very flat, and suppose the surfaces will have to be parallel within a tolerance of 0.001". If I
apply the concept of statistical process control, is the "parallelism" measure still a normal distribution? After all, parallelism isn't really a dimension.
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>

Parallelism would be considered a unilateral tolerance. That is, it has no sign, much like a suface profile call out in GD & T. So for SPC in your example, your SPC chart would have a 0 to 0.001" tolerance, but your Xbar would be located at 0. There is a specific case for unilateral tolerances in SPC, but I cannot recall what the terminology and specifics are right now.
--
Anthony

You can't 'idiot proof' anything....every time you try, they just make
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>

If it matters to you, this is a Chi (pronounced KI) squared distribution. The plot of this distribution rises abruptly on the left and has a long tail on the right. SPC limits for this process are calculated by transforming the data to something with a normal distribution, determine sigma and limits, and then transforming back. I can look all this up for you if you really need it.
Or, run a bunch or parts. Look at the data by eye. Compare to your spec. Choose a value (SPC limit) that will alert you when something has changed, but before you're out of spec. This is called a SWAG in engineering terms. Much faster, and generally works better too.
Karl
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
says...

Actually, you don't know what kind of distribution you have until you run a sample. Take the measurements and check them for normality using Minitab or a similar program. If they are not normal, take groups of samples (say 3 or 5), The distribution of the averages of those groups WILL be normal. (Central Limit Theorem) To do this sort of thing you will have to be able to measure close enough to get a variety of readings so that you can have a distribution of some kind, say .0001 If you can't do that then you can use an Attribute Control Chart. This is what you use when (say) you have GO and NOT-GO pins and these are your only two conditions, "good" and "bad". For reference materials on how to do this you might try contacting The American Society of Quality: http://www.asq.org / or perhaps SME. If you are an SME member you can use the "Ask the Librarian" feature on their web site. http://www.sme.org/cgi-bin/libhtml.pl?/library/library.htm&&&SME& Actually you can use this if you are not a member too, but you have to pay for it. Hope this helps. -plh
--
I keep hitting "Esc" -- but I'm still here!