Capability Analysis - Help with Minitab results

[Licht]

Re: Capability Analysis - Help with results

Miner, thanks for the article!

I've done the Quality Trainer from Minitab to learn how to use their functions, however I need informations/guides that shows how to select parts correctly (quantity, mode and frequency of collection, etc) for the MSA study. Is there a tutorial or a practical textbook to indicate?

>> Below I describe how I made the parts collection for gage R&R, Linearity and Bias studies:

1) The machine is automatic with manual loading by one operator. The measurement is automatic and the manual

loading does not influence the measurements. So, I conclude that there is no reproducibility;
2) I selected 14 parts ramdomly;
3) I measured each part in a laboratory with a calibrated equipment to have a master (reference) value for the Bias and Linearity study;
4) I changed the parts (electronic components changed) to have values ​​distributed in all specification range, ie, values between LSL and USL:

- 02 parts modified to have values ​​close to LSL
- 02 parts modified to have values ​​close to USL
- 10 parts with values ​​between LSL and USL (unmodified)

5) Measurements were made (42 measurements = 14 parts * 3 trials) randomly;
6) Enter data in Minitab for analysis (gage R & R, Linearity and Bias studies) using only the collected data (42 values).

>> Below I describe how I made the parts collection for Capability study:

1) During the production I recorded the values ​​of 112 parts once sequentially;
2) Enter data (120 values) in Minitab for analysis (Capability study) using only the collected data (112 values).

Could you indicate where were my mistakes in this procedure adopted, please?

 

[Miner]

You went wrong in step 4. If you had used the 14 parts as-is, you would have been okay.

Step 4 is only valid if your interest in the gauge is strictly for inspection.

 

[Licht]

Miner, thank you for the patience...

I understood your comment, but it is not clear to me the concept of use of gage: part inspection to a tolerance, for process control or for statistical studies ! My thought was that gage R&R study is to evaluate the measurement system only, is not it?

Could you tell me more details about the difference of them and an application example ?

Greetings

 

[Miner]

The intended use of the gauge affects two things, the selection of parts for the study and the metric used to determine suitability.

• Inspection device: Parts selection does not matter. Your approach is acceptable. % Tolerance used to assess suitability.
• Process Control (i.e., SPC): Parts must be selected to represent actual process variation. % Study Variation, % Process Variation or ndc used to assess suitability.
• Statistical studies/Capability Studies: Same as Process control

Try reading my blog on Part 5b for more info on the metrics.

 

[Qualgal]

I have been taught that if the mean, mode, & median are pretty much the same, you can present this information and exempt the data from the normality requirement. These values are now a requirement on our capability analysis reports.

 

[Bev D]

QualGal - not sure who taught you, but that is a myth. any symmetrical distribution will have equal measures of the central tendency. That does not make them Normal. heavy tails, light tails or even a uniform distribution will result in erroneous Cpk calcualtions. A 'Normal' distribution that has a low resolution gauge that produces 'chunky' data will result in an erroneous calculation.

bottom line: there is no substitiute or shortcut for plotting your data in time sequence against the spec limits to truly understand your process capability.