have the task to validate that the in-line inspection system for an automated machine functions properly. How can I perform an MSA for this case, since there is no operator interaction? I can only think about assessing repeatibility but don't know if we can verify reproducibility. Also, it is correct to measure a part with a less accurate instrument like a caliper and compare it to the automated system to verify accuracy or bias? If not how can I validate that the system output provides the correct result, the most accurate system in the plant right now is a CMM. Thanks in advance for your help
Jorge,
A number of things come to mind as I read your post.
The inline measuring system you are discussing is an instrument “important to quality”. Instruments important to quality are required to be calibrated. Has the calibration of this instrument been performed? If so, you should have answers to many of your questions. The calibration paperwork should answer your questions on bias, precision, linearity, and stability. The calibration lab is also where you will find the standard for the part, although you will not need it for what you need to do.
How was the calibration performed? Was it an in place calibration or was the instrument removed and taken to the calibration lab for calibration?
If the instrument was removed for calibration, or calibrated prior to installation, then you are probably going to have to remove it to perform your MSA. Your biggest problem when you remove the instrument for the MSA is you have to make sure the instrument is measuring the parts under the same conditions present during actual operation. For example:
1. Get the instrument in the same relative position.
2. Get the feed to the instrument as close as possible to the feed used during actual operations.
3. Get the environmental factors equivalent, such as:
a. Vibration caused by the machine or other nearby equipment.
b. Ambient temperature.
c. Temperature of the parts entering the device while the system is running.
Some of these factors may be unimportant to your MSA, but they need to be formally addressed by engineering and documented.
Hopefully, the calibration was done in place, because that means a method exists for inserting the parts into the line upstream of the measuring device, after all, the calibration standard had to go in there. You can use that same inlet to insert your parts into the system for performing the MSA.
If there is no method for inserting the test parts into the system, you need to get engineering to design a modification to put one in. There are going to be periodic calibrations and periodic MSAs to perform throughout the life of the system so you might as well get the capability to perform those operations installed now.
To perform the MSA, I would get 10 parts (clearly numbered 1 – 10) and run them through the measuring system 10 times in random order each time, taking data just like any other MSA. I think you will find that there is enough slop in the system that the part will not always be in the same position when the measuring device measures it which will cause some variation in the measurement. That variation will be read by the MSA as reproducibility. Accordingly, you will have reproducibility to deal with in your MSA, the reproducibility of the automated system for setting up the part correctly for measurement.
Finally, it is not okay to measure a part using instruments that are less accurate than the accuracy of the measurement system under test. If you do, in the end you cannot say anything about the accuracy of your instrument under test with any reasonable confidence.
That is how I would proceed if I were you.
I hope this helps a little.
Best regards,
Jim Shelor
PMP, SSBB
