# Capability for CMM Scanned Profiles on a feature with a Bilateral Profile Tolerance

#### Rand T

##### Inactive Registered Visitor
I have been struggling with this for awhile and never really resolved it.

I have parts that get scanned via CMM with hundreds of points. The PCDMIS CMM output is the min and the max along with a value called "measured" which is the absolute difference between the two if different signs and the greater magnitude of the two if the same sign.

example; a) min = 0.05 mm & max 0.10 mm. "measured" = 0.10 mm
b) min = -0.1 mm & max = 0.05 mm. "measured" is 0.15 mm

I am trying to calculate capability on a feature with a bilateral profile tolerance of 0.25 mm. I have thought about a couple different methods, but both seem to have pros and cons. Of course, I want to determine the "worst case" condition and apply it against the tolerance.

1) Use the "measured" value. The downside is that this "measured" value gives very different values depending upon if the signs of the min and max are opposite or the same which may end up overstating the amount of variation from nominal. In exp. b above, my largest variation from nominal is still only -0.1 mm, but the value I have is 0.15 mm.

2) Select either the min or max depending on which one is larger. This option was the one that made the most sense, but has one big drawback. The min and the max can be very close in magnitude, but with opposite signs. When I select the larger of the two, I often end up with values the bounce back & forth between + & -, again now overstating the variation.

exp; part #1 is min.= -.032 & max = .034 use .034 since it is larger
part #2 is min= -.033 & max = .031 use - .033 since it is larger

Now it looks like I have a range of .067 when in actuality it is only .003

Anybody have any suggestions?

I could do min and max capability seperately, but that seems like a lot more work and I'm not sure there is value there either.

Last edited:

#### bobdoering

Trusted
Re: Capability for scanned profiles

What is the process that generates this feature?

Typically, trying to use a single number (capability index) to describe a relationship of many points within a part to a single tolerance is a knotted up mess of a concept - more so than just using one measurement. Not sure if you are going to end up with a meaningful analysis.

But, that depends on the process.

#### Rand T

##### Inactive Registered Visitor
Re: Capability for scanned profiles

Machining of titanium. Customer is requiring 100% CMM inspection until we can show capability.

Unfortunately, don't have a good method of outputting the individual point data on the entire cloud of scan points. Logically, it makes intuitive sense to calculate based upon the single worst case point, however there are pitfalls as I noted above.

#### bobdoering

Trusted
Re: Capability for scanned profiles

Well, this may help. Is this profile made with one tool? If so, how many parts are made from a tool before an offset is required? How many before a tool change?

If not, then the analysis gets complicated further...

As I read through the algorithm you describe - if it is correct - I do not see how you can correctly calculate capability. You have apples (positive/positive) and oranges (positive/negative) - and capability is not a fruit salad.

Does PCDMIS allow any other analysis technique of the same data set? How many parts are you using as your sample set? Can you get the min and max for each part, and plot those values in a time order throughout the run? There might be hope there, if you can....

Last edited:

#### Rand T

##### Inactive Registered Visitor
Re: Capability for scanned profiles

Not the machining expert, but I believe it is one tool. The runs are fairly short, (60 pc) so few changes once it is set up and running.

I do have the min and max for each part and could plot in a time series. Right now, I am using approx. 30-40 pcs as a sample set.

The 100% CMM is killing me from a time standpoint, and I have a customer who is requiring capability before relaxing that requirement. As far as PCDMIS is concerned, I have to rely on what my programmer is telling me, so I'm not sure if there is anything else I can get. So far he hasn't indicated there are other options, but I haven't given up yet.

Last edited:

#### True Position

##### Inactive Registered Visitor
Re: Capability for scanned profiles

Are you applying any type of automated filtering to the data? Scanning or individual points?

#### Rand T

##### Inactive Registered Visitor
Re: Capability for scanned profiles

No All I have is the min and max from the cloud of scanned points. CMM measures 100+ scanned points and reports two values.

These mins and maxes are likely not even the same exact point from part to part, but whichever of the entire scan cloud were the min and max on each piece.

Ultimately, I am thinking I may need to do capability on both the min and max seperately and just ignore the one that has the highest indice.

#### True Position

##### Inactive Registered Visitor
Re: Capability for scanned profiles

No All I have is the min and max from the cloud of scanned points. CMM measures 100+ scanned points and reports two values.

These mins and maxes are likely not even the same exact point from part to part, but whichever of the entire scan cloud were the min and max on each piece.
Is there outlier filtering or no? Do the points tend to move?

#### Rand T

##### Inactive Registered Visitor
Re: Capability for scanned profiles

No filtering as far as I know. Not sure what you mean by "Do the points tend to move?"

The values are fairly consistent, (doesn't seem to really be any outliers) but tolerance on a couple is a profile total of .08 mm, so not a lot of tolerance to begin with.

#### True Position

##### Inactive Registered Visitor
Re: Capability for scanned profiles

No filtering as far as I know. Not sure what you mean by "Do the points tend to move?"

The values are fairly consistent, (doesn't seem to really be any outliers)
Are the min and max points always at the same place on the scanned parts? (Point #1 vs #60)