Cpk Measurement for Knee Implants on CMM Machine

[validationeng]

Hope you can help here in guys...

I need to do a cpk analysis in relation to knee implants we are making...a few details.

1. We have made 36 seperate parts (knee implants) on a CNC as part of
validation
2. One of the CTQ's is tray thickness. This is measured across 8 different
points along each of the 36 parts.

My question is should I average out all the 8 tray thickness values the CMM gives along each of the 8 points for one part and then do the same for the other 35 parts. ...than do the cpk on these 36 values.....or

Do I need to seperate the 8 points i.e Tray thickness 1..tray thickness 2.....right up to tray thickness 8 and treat them as 8 seperate CTQ's....effectively geting 8 seperate CTQ values...i.e for tray thickness 1 I would take the value for it on all 36 parts and run cpk analysis.....then do same for tray thickness 2,3,4,5,6,7 &8 in the same way and thus eventually ending up with 8 cpk analysis values

All vey complicated but any advice is appreciated !!

[Stijloor]

A Quick Bump!

Can someone help?

Thank you very much!!

[JAltmann]

If it was me looking at the data i would probably output 8 points individually. I might then teat them as a sub-group of 8 in the statistically calculations.

I am assuming these ribs are machined features?

[Pat McGhie]

In Reply to Parent Post by validationeng

Hope you can help here in guys...

My question is should I average out all the 8 tray thickness values the CMM gives along each of the 8 points for one part and then do the same for the other 35 parts. ...than do the cpk on these 36 values.....or

Do I need to seperate the 8 points i.e Tray thickness 1..tray thickness 2.....right up to tray thickness 8 and treat them as 8 seperate CTQ's....effectively geting 8 seperate CTQ values...i.e for tray thickness 1 I would take the value for it on all 36 parts and run cpk analysis.....then do same for tray thickness 2,3,4,5,6,7 &8 in the same way and thus eventually ending up with 8 cpk analysis values

All vey complicated but any advice is appreciated !!

Based on the data you presented and the fact that it is labeled a CTQ, I would most likley do a Cpk on each of the 8 points seperate, so yes, I would agree with the 8 individual Cpk analysis values.

It may be just as effective to do an average (and depending on the customer that may be ok) but as you will measure all 8 anyway, I would tend to just report them each as an individual.

Pat in NH

[Bev D]

it really depends on the variation between the 8 points and the 36 pieces. If there is a systematic difference the 8 locations the statistical analysis must take into account the 'fixed' effect of the 8 locations. If the variation between the 8 points appears random or is much smaller than the difference between the 36 parts then allof th edata can be essentially be treated as the same and you woudl use all 36*8 points in a single calculation.

If you would post your data we can help you determien the correct analysis to apply. It might also be helpful if you could describe the process by which the 'tray' dimension is created.

[bobdoering]

First of all, you are looking for process capability. In order to accurately do this, you need to compare the process over time – apples to apples. That means, you need to treat each point separately. Averaging just dilutes the value of the data.

What is the source of variation from part to part? Over those 36 parts, were the tools used changed? Were there any adjustments made? If so, why, and what triggered the change? If not, then the only variation you should see is tool wear. Now, how you handle the capability depends on how you handle the process.

When and how is the process adjusted over time? Are any of these points SPC charted for process control? If so, with what charting?

Sure, you might think what do all of these questions have to do with dumping data into a formula and calculating Cpk?

One answer: everything.