Is capability applicable for a destructive test?

[nhonibru]

Dear all,

I'm a QE beginner, and I have some doubts concerning cabability process in a destructive test.

Background: I have a crankshaft that have a special characteristic about the supported torque. The minimum torque needs to be 21,5Nm, and when we do some tests with torquemeter, when the torque is exceed, the crankshaft breaks.


The question is: Is capability applicable for a destructive test?

Thanks in advance!

Bruno

 

[Bev D]

welcome nhonibru! are you asking about the capability of the torque measuring system or the capability of the strength of the crankshafts?

 

[nhonibru]

Hello Bev D, thanks for your fast answer!

It's about the capability of the strength of the crankshafts.

 

[Al Rosen]

I think we/you need more data. Did you incrementally apply the torque? in what increments? How much did you exceed the rating?

 

[nhonibru]

I think we/you need more data. Did you incrementally apply the torque? in what increments? How much did you exceed the rating?

Yes Al Rosen, we incrementally applied the torque. The values should be at a minimum of 21,5Nm and we found a lot of values in a range of 20,0 to 21,5Nm.

 

[Al Rosen]

Yes Al Rosen, we incrementally applied the torque. The values should be at a minimum of 21,5Nm and we found a lot of values in a range of 20,0 to 21,5Nm.

If you want some help, you post all your data, but out of the chute you are not meeting your spec minimum if you have a lot of values <21.5. You don't need a capability study to tell you this.

 

[nhonibru]

If you want some help, you post all your data, but out of the chute you are not meeting your spec minimum if you have a lot of values <21.5. You don't need a capability study to tell you this.

Attached the file with all the measures...

I agree with you, but the main question is if It's applicable the capability study in this case, as this is a critical characteristic , but to measure it you need to break all the samples you are measuring (destructive testing), I mean, It's not just a dimension that you can measure without breaking the samples (non-destructive testing).

I don't know if there is a standard saying something related to this, about non destructive tests and destructive tests and their capabilities...

I don't know if it's clear to you! I'm trying to explain my best...

Thanks one more time.

 

[Bev D]

Yes you can do a capability study on characteristics that are tested by destructive methods. Your samples should be spread out over your NORMAL testing frequency. Just plot the data in time series against the specification limits. If you are required to calculate the silly Cpk index do so after you have tested enough pieces. Remember that Cpk is NOT intended to be calculated for each lot...

 

[Miner]

As BevD said, you can definitely perform a capability analysis on destructive tests. However, there is a lot you can do with purely graphical analysis before you worry about quantifying the capability.

First question: Do you have a stable process? Assuming the data are in time order, the process is reasonably stable. The points flagged below may be because the process is not normally distributed (strength tests are usually skewed).


Second question: Is the process capable? Even an untrained eye can see that the process is not even close to being capable. No statistics necessary.


What next? Your problem is either in the design (i.e., the design never was nor ever will be capable), or in the processing (e.g., machining, tool geometry, etc.) that reduces the torque strength from the designed strength. I would recommend carefully analyzing and identifying if there are differences between parts that fail at 23.6 versus those that fail at 20.8. If there are differences, it may be a process issue. If there are no differences, it may be a design issue.

 

[nhonibru]

As BevD said, you can definitely perform a capability analysis on destructive tests. However, there is a lot you can do with purely graphical analysis before you worry about quantifying the capability.

First question: Do you have a stable process? Assuming the data are in time order, the process is reasonably stable. The points flagged below may be because the process is not normally distributed (strength tests are usually skewed).
View attachment 26621

Second question: Is the process capable? Even an untrained eye can see that the process is not even close to being capable. No statistics necessary.
View attachment 26622

What next? Your problem is either in the design (i.e., the design never was nor ever will be capable), or in the processing (e.g., machining, tool geometry, etc.) that reduces the torque strength from the designed strength. I would recommend carefully analyzing and identifying if there are differences between parts that fail at 23.6 versus those that fail at 20.8. If there are differences, it may be a process issue. If there are no differences, it may be a design issue.

Thank you so much! It really helped me a lot.