Surface Profile: Process Capability 1

[weisusu]

Surface profile can be used to simultaneously control different geometric tolerances. For instance 0.002"|A|B can be used to control form, location, and position. How would you address process performance/capability of such GDT callout? How would you know that you need to improve location but not the position or vice versa?

 

[KENAT]

Aren't location and position pretty much the same thing. Do you mean orientation?

Fundamentally the surface needs to fall in the specified zone.

If your orientation is off I'd think you'll be able to tell by the part being at angle to the specified datum (or more or of an angle for a feature not parallel to datum.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[weisusu]

Kenat:

I agree that fundamentally the surface needs to fall in the specified zone. But my question pertains to measurement of process capability/performance given use of surface profile to simultaneously control location, form, and orientation. I may be able to use all + dev and - dev and compare them to the +/- limits (for surface profile). But if the Ppk = 0.90, what would be a good way to know if it is the location that gives me issue or orientation?

 

[PaulJackson]

Surface profile can be used to simultaneously control different geometric tolerances. For instance 0.002"|A|B can be used to control form, location, and position. How would you address process performance/capability of such GDT callout?

Depending upon what kind of feature the profile is applied to predicting process performance/capability can be as easy as it is for a simple limit dimension or as complicated as one can imagine with an "all over" modification applied to a multi-surfaced non-uniform structure with no datum reference specified.

The simple answer to addressing process performance/capability is to figure out the extreme encroachment or breech of the measured surface profile on its tolerance limits for each sample and predict the likelihood of such for the remainder of the lot or the future of the process.

How would you know that you need to improve location but not the position or vice versa?

The answer to this question is far more valuable than knowing how to predict whether the lot conforms or not to the specs. By knowing what each characteristic of a feature contributes to its overall profile error can help one to make an informed choice on what needs to change in the process to improve its capability.

Take a flat surface for example. It has a profile tolerance on it. Is it flat? Of course not! How much?... Is the median plane of that of that not so flat surface square to the DRF? Of course not! How much? Relative to which axes X, Y, Z?
Is that not so flat…tilted median plane centered at its basic dimensions relative to the DRF? Of course not! How much? Again relative to which axes X, Y, Z?

The point is that unless you can identify what is causing the variation you are ill equipped to make process adjustments to improve it. I said earlier that figuring the process performance ‘can be as easy as it is for a simple limit dimension’ and that’s true but the simple limit dimensions typically have all the same parameters as the profile of the flat surface it is just that the limit dimensions DRF is arbitrary which further complicates its possible parameters.

We need to regress for a moment… In order for any capability prediction (process performance and/or process potential) to be trustworthy the data used to generate the prediction needs to exhibit random behavior free of special influences. It also helps to simplify process adjustment if the individual process data parameters are oriented the adjustable process parameters rather than those specified on the drawing if different. You see monitoring, controlling, and reducing variation in a process and predicting the process’ capabilities are two separate things. Process capability predictions require specification limits process control predictions do not.

Many people believe that process potential cannot be predicted on “one sided” geometric tolerances like profile or position but I disagree. If one was to take that individual parameter data for the surface profile and while retaining all the individual variation for form, orientation, and location yet make mean error adjustments to the data for orientation and location one could see the process potential if those adjustments could be realized. Isn’t that exactly what is happening in a one parameter two sided spec. process potential prediction!

It takes a little extra thought to figure what parameters can or should be monitored but even the most complex profile specifications can be simplified to single point two sided capability predictions where the limits are basic + and – material as specified. Believe it or not capability predictions can be evaluated with “variable limit” or “MMC” geometric tolerances where individual feature sizes control all or a portion of the specified tolerance. Process potential for those specifications involves finding the target for feature size that minimizes the probability of a defect for the size and geometric tolerance simultaneously.

Paul