Control Charting a Feature with Bonus Tolerance

Does it make sense to use an X Bar and R Chart to chart a feature with bonus tolerance? Would there even be a LCL? I just came off the shop floor and seen where they are using and X Bar & R chart for a true position with a maximum material condition call out and there are upper and lower control limits. Isn't that like trying to aim for a moving target?

Re: Charting a feature with bonus tolerance

Welcome to The Cove, useridpain!

Since there is no relationship between control limits and specification limits, I don't see a problem. The control charts are looking at the process variation without regard to the specification.

Re: Charting a feature with bonus tolerance

Howste is correct.

Sometimes, however, people DO want to see data to a specification. And you are correct, MMC is then a moving target. The answer is to scale it. Call full tolerance (whatever it may be based on the diameter) a "1". Then if they are halfway to this one, you plot at 0.5. Three quarters of the way, you plot 0.75

Datum modifier bonuses make the statistics harder. (They make TOTAL sense when you consider you may be getting mating components from two different sources). Don't get me started on true position and capability ....

Thank you for the information ncwalker and howste.

What type of GDT dimension are you trying to chart? The complicating factor is not the bonus tolerance, which as noted earlier, is irrelevant in SPC. The problem is that GDT dimensions are typically absolute values, and some are partial polar coordinates.

If you are trying to do SPC on Position, you would be better off using the X/Y coordinates than the polar coordinates. Others such as flatness or parallelism are absolute values and may be highly skewed the closer you get to zero (see folded Normal distribution). This can cause nonsensical negative LCLs and more false alarms above the UCL.

It is a true position of a hole, .003, at maximum.

useridpain said:

It is a true position of a hole, .003, at maximum.

Click to expand...

I would use SPC on the x and y axes. If the position was out of control on position, it could be out in any direction around 360 degrees. You would have to look at the x and y coordinates to discover where the problem lies.

OK, thanks for your help.

what Miner said!

SPC should tell you what the process is doing. and to do this we need to use the raw data as the process creates it. bonus tolerances and other manipulations take us away from visualizing the process...

the old adage to keep it simple will almost never fail you. The point of SPC is so the operators know what is happening so they properly adjust things.


the issue with trying to do SPC and Spec limits is like using a spork. it really isn't very good at either...