tolerance analysis

[dvennard]

how simple is it to apply some six sigma analysis to a tolerance stackup? (ignoring GD&T stuff, just basic linear dims) so if i have 10 dimensions in a stack-up, what are the odds of every one of those dims being at the extreme point of their tolerance range?


i am hoping there is a way to determine a good estimate for a realistic error. for example, if my 10 parts have a stack up of plus/minus 2.5mm, applying some 6 sigma may tell me that realistically it will only be plus/minus 1.5mm (i just made those values up, just trying to get the point across)


i just remember a teacher talking about finding the 99 percentile tolerances, and basing a stack up off of that...any thoughts on the subject????


thanks


dan

 

[deedub777]

A method I've used in the past is the RMS method (or Root Square).

For a stack of (about) four or more dimensions, here's the method:

Add up the squares of the tolerances and then take the square root. This gives you a more realistic outcome as opposed to the worse case result.

Take care in using this methos as it depends on the manufacturing process. For example, plastic injection molding parts can have this applied but I would not use it for 'manual' manufacturing process (eg sheet metal , etc.)

I have a spreadsheet that works it all out for you if you are interested.

Hope this helps.

 

[dr_gallup]

If you want to use statistical analysis the best way is to measure the CPk of each dimension then you can accurately determine the the stackup tolerance. However, manufacturing must be using SPC and keeping the processes in control.

If you don't have any CPk data the RMS method deedub777 mentioned is a very simple down & dirty method. The more dimensions in the stackup the better it works because some of the variation in the dimensions will offset each other. If you have one or two tolerances that are bigger than the rest they will tend to dominate the calculation.

 

[dvennard]

thanks for the help guys!