Capability with an Upper Limit but a Lower Limit which is Nominal Value

[Ospif1]

Hi all,

I am required to perform a capability analysis for one of our parts. The customer spec requires a particular hole dimension to be 10mm +0.2/-0.0. How can I perform a capability for this? With the lower limit also being their specified nominal value, the capability will be terrible. Is there a method to measure capability when this scenario is encountered?

Regards,
Reece.

 

[bobdoering]

I am required to perform a capability analysis for one of our parts. The customer spec requires a particular hole dimension to be 10mm +0.2/-0.0. How can I perform a capability for this? With the lower limit also being their specified nominal value, the capability will be terrible. Is there a method to measure capability when this scenario is encountered?

You have a variety of issues to consider.
1. Are you certain 10mm +0.2/-0.0 means 10.0 is "nominal"? Sometimes in metric, the tolerance is written that way as a standard method without that specific meaning.

2. The USL and LSL for 10mm +0.2/-0.0 is 10.2 mm and 10.0 mm no matter how it is written.

3. Capability is about the process, not the specification. What process is this? Drilling? Honing? What is its distribution and what is the variation over time?

4. Is your measurement accurate enough to determine the capability? Do you have an ndc of at least 10? If not, the variation you see and its representative distribution may just be measurement error.

These are a good start. If you are drilling, that is not "precision machining", so the capability over time based on drill bit lots, material lots and how that particular drill behave when it wears can be very erratic. Capability on one lot - especially with one drill, is pretty much a pencil-whipping exercise.

 

[Ospif1]

You have a variety of issues to consider.
1. Are you certain 10mm +0.2/-0.0 means 10.0 is "nominal"? Sometimes in metric, the tolerance is written that way as a standard method without that specific meaning.

The 10mm is the nominal diameter as confirmed by the customer.

2. The USL and LSL for 10mm +0.2/-0.0 is 10.2 mm and 10.0 mm no matter how it is written.

3. Capability is about the process, not the specification. What process is this? Drilling? Honing? What is its distribution and what is the variation over time?

It's a steel punching operation within a large press. Initial sample inspection reporting suggests we have variation from 10-10.03mm measured over 125 parts taken from various lots out of a 300 bunch. It's a new part so it's an initial capability.

4. Is your measurement accurate enough to determine the capability? Do you have an ndc of at least 10? If not, the variation you see and its representative distribution may just be measurement error.

It appears to be very accurate and consistent.

Thanks for the reply, I am just confused as to how we can go about it since having the upper and lower of 10.2 and 10 would be well and good but since all measurements are at the LCL, the capability would be terrible.

 

[bobdoering]

It's a steel punching operation within a large press. Initial sample inspection reporting suggests we have variation from 10-10.03mm measured over 125 parts taken from various lots out of a 300 bunch. It's a new part so it's an initial capability.

Do you expect the main source of variation to be tool wear? Do you expect the hole to get smaller as the punch wears? If so, do you start with a large punch - and larger diameter - then allow the hole to get smaller until the punch needs changed? If so, Cpk is not applicable and long term you data will not be normal. Short term it may appear normal, but that is because measurement error (which is normally distributed) will be larger than the tool wear in the short term, making the data appear normal (masking the true process distribution).

In the long run if your major variation is tool wear, (which it should be - material variation and warm-up are special causes) capability is:

(USL-LCL)/(UCL-LCL), where UCL is the initial size of the hole with the new punch and LCL is the smallest size of the hole when you intend to change the punch. Both the UCL and LCL should be 75% of the tolerance. This is because your distribution is non-normal and uses different statistical analysis.

 

[Ospif1]

Thanks for the advice all. I had a discussion with the customer and they admitted error on their behalf and as a result I have suggested a spec change to +/- 0.2mm which they have accepted. Unfortunately as it always is with car companies it will probably take 20 weeks and 5000 man hours to get it sorted but that is how it is.

Cheers again.