Using MSA for Situations Requiring Destructive Testing

[#1]

See how to:

• Address key assumptions for MSA (Parts do not change form over time and are not changed by measuring them)
• Overcome inability to repeat measurements when key MSA assumptions are violated
• Perform and interpret desctructive test that assumes perfect material and likely overestimates part-to-part variation
  

  • Assume material within a lot or a batch is uniform

  • Gather multiple lots where each lot represents a part

  • Subdivide lots so that each subdivision will represent a repeat measurement when tested

  • Test all parts in a random order.

  • Analyze using EMP method.

• Perform and interpret desctuctive test that assumes perfect test and likely underestimates part-to-part variation

  • Use a nondestructive test that correlates well with the destructive test

  • Build a calibration curve from destructive and nondestructive test data

  • Test parts nondestructively in a random order

  • Test parts destructively in a random order

  • Use the formula from the calibration curve to scale the nondestructive data

  • Combine results

  • Analyze using the EMP method
     

Notes: Questions are included at the end of the demo, beginning at ~ Time 32:30.

Resources

• Mastering JMP video and resources: Using the EMP Method for Measurement System Analysis
• Gauge R&R Studies for Destructive Measurements by Jeroen De Mast and Albert Trip 
• Measurement Systems Analysis for Curve Data (JMP Discovery Summit presentation)

A reminder that to contact Jason Wiggins in comments, include @Jason_Wiggins. in your comment. Measurement Systems Analysis.zip is a zipped version of Jason's JMP 16 Project used in the session.

[PatrickGiuliano]

Also linked above but I am including it here since this post is related:

Using the EMP Method for Measurement System Analysis - JMP User Community

[PatrickGiuliano]

For those who are interested (including myself), I'm linking a Discovery presentation from Autoliv where they covered a Destructive Gauge R&R measurement example for curved data (  ) that follows the third example covered in Demast's seminal destructive Gauge R&R paper linked in this post, and recopied here: https://ibisuva.nl/assets/publicaties/artikelen/2005-demast-gauge.pdf

Thank you @Jason_Wiggins  for curating these resources and providing this wonderful presentation!

[Jason_Wiggins]

Thank you, Patrick! I am happy that you enjoyed the presentations and resources. The MSA for curve data presentation is really exciting. Destructive MSA is a solid use case for Functional Data Explorer. Enjoy!

[Mauro_Gerber]

@Jason_Wiggins Thank you,

What you can add is to produce parts with the variation you need for the MSA (like precise milled parts with varying cross-section for a tension test).

If the variation is big enough, we also pick sample within a batch or a single part in our case.

We produce for example a 10km long cable and take samples that are 2m long and further split those in 20cm sub-samples with the same "part ID" for the repeatability.

@PatrickGiuliano 

The article from the ibisuva.nl page gives me an 404 error.
I found the same title under the following link: 2005-demast-gauge.pdf (ibisuva.nl)

[PatrickGiuliano]

@Mauro_Gerber Thank very much for correcting the broken link to the Demast Paper. 

[Jason_Wiggins]

Thanks for the recommendations @Mauro_Gerber. Enjoy!