Search the Elsmar Cove!
**Search ALL of** with DuckDuckGo including content not in the forum - Search results with No ads.

In-Process Production Test Stations in PFMEA (Process FMEA)


Involved In Discussions
In my company there is a rule (which is not extracted from any book/guideline as far as I'm concern) to split the testing stations in two categories:

1 - Non-touching test items: Camera/Sensors
2 - Touching test items: detection pins (feelers)

The first category is never described as a process step, under the assumption that the product is not being modified at that step (although one could argue that the information related to it has increased).

The second category is always included as a Process Step in order to assess the possibility that it damages the product while testing (ex: when you need to insert an electrical connector, or need to actuate the part a bit).

We do not use: Accept good part/Reject bad part. We take into account the reliability of this detection action in its D rating.

We have a general rule (and I think this is written in some of the guidelines) that we do not study "cumulated" failures, like two failures in a row or two simultaneous failures. As such studying the failure of a detection tool to detect a failure is not part of our typical scope.


Involved In Discussions
It is the case with us, all test items are always mentioned as Detection Actions.

We only add them also as Process Step when the test itself can physically harm/damage the product.

John Predmore

Involved In Discussions
The most common control for an inspection process [with a sensor] would be a calibration program. Others might include an R&R study, red rabbit testing, etc.
Thank you, Miner. Many people think of calibration strictly as an annual exercise and R&R as a one-time activity. I had not heard of the term red rabbit testing, but that means putting a known bad part into the process and making sure it's detected.

Taking that idea one step further, my company uses what we call "challenging" the automated test station. A green master part, typically spray painted Green, is created to pass every time, and a red master part is designed to fail every time. For a leak test station, for example, the green master was epoxied shut so it would never leak, and the red master, also epoxied shut but drilled with a single hole of precise diameter, provided a known, stable rate of leak to slightly exceed the leak specification limit. The red/green challenge test was done daily or at the beginning of every shift.

If the green master ever failed the leak test or the red master passed, the the test station was suspect, and manufacturing would halt while the engineers investigate. The advantage of using physical parts modified not to leak (as opposed to internal valves or disabling the vacuum source) was this tactic could discover random causes, such as a pinched tube in the equipment, a clogged filter restricting airflow, a damaged or dirty gasket, or a vacuum fitting which had loosened over time.

In the context of PFMEA, there are dozens of causes for a leak test to fail, too many to consider exhaustively, but the net consequence of failure is a good part might fail as a leak and a leaking part might pass as good. The challenge test I describe is a simple control method to improve the detection score on the FMEA, as an inexpensive way to reduce risk from a myriad of failure mechanisms.


Starting to get Involved
Guys, speaking of red rabbits for testing, some people say they should be classified in a PFMEA under the detection controls, and others like our OEM customer lead engineer that they should be listed under preventive controls columnm any thoughts? Also, how bad off limit should a bad part be made?



Involved In Discussions
I would not have thought to mention red rabbits in any of the two columns.

To me the red rabbits are used to verify that a detection system is working fine.

The testing of Red Rabbits, according to my understanding, is a form of Measurement System Analysis and doesn't belong to the PFMEA.

To answer the question: How bad off limit should the part be made? I think you would find your answers in the MSA handbook (from AIAG I think).

For continuous measurements like a height, it doesn't matter how off limit the NOK part, you just need to see if your machine is capable of measuring the real value (the acceptance/rejection is done later through Software controls).


Quality Manager
R&R and Capability is good, but glitches happen fast when sensors are used to check 24/7 for product type features or design intent, I would consider Running Red rabbit checks on known system aspects on a scheduled and Layered frequency, dependent on the severity of the inspection station. Also PM frequencies of the complete unit. If severity is on a life and limb application I would also consider redundancy Power circuit checks; checking that the checks worked.


Starting to get Involved
Fmea is used to find the parameters that cause defect, as long as your process just for test and not produce the products, you can use MSA to review your test
Top Bottom