FMEA Ranking and Prioritization - Failure Modes with higher severity first?

[YKT]

Correct me if I'm wrong, but I understand that the approach for FMEA prioritization is to go for Failure Modes with higher severity first, rather than the RPN calculation.

If I read the Detection ranking table correctly, the ranking '1' is where a mistake proof, or error proof is in place, thus it is impossible for the failure mode to happen. Thus, if we have identified a Failue Mode with a severity = 7, occurence = 5, and detection = 1, it is for sure that we are not going to do anything even though the RPN = 35, as with the detection = 1, no way the failure mode will happen again.

Am I right for the 2 occasion above ?

 

[Marc]

I'm sure others will have some comments, but for a quickie, see http://Elsmar.com/FMEA/

 

[Geoff Cotton]

thus it is impossible for the failure mode to happen. Thus, if we have identified a Failue Mode with a severity = 7, occurence = 5,

YKT,

If the process has been mistake proofed and it is impossible to happen (occur), should your occurrance ranking be more like 1?

 

[Randy Stewart]

YKT,
We teach our people to look at the Severity x Occurance scores to pick out the identified problem areas. In our fixture design operations first time quality matrix you would see that the highest occurance is spelling errors (8), detection is low (8) so this part of the RPN is a 64. However when you put in the severity score (1) it should become a non-issue. IMO you need to look at both severity and occurance, you don't want to spend resources on a high severity when the occurance is 1 in a million nor for something that may happen a lot but doesn't mean much to the customer (i.e. spelling errors).
An added benefit is that it's another way to show how customer requirements and COPs are passed down through the entire operation.

 

[Bill Ryan - 2007]

I agree with Geoff. If you can't produce the nonconformance your Occurence rating should be 1. If your Occurence rating is "truly" a 5 and you are "detecting" the nonconformance, you are "good to go" with the Dectection rating of 1 (just to clarify - you are not preventing the nonconformance but have a means to sort it out so it can't be shipped).

Randy's methodology of "prioritizing" action plans is a good one but there is no "fast rule" other than at least looking at any Severitys of 9 or 10 (Safety issues), possibly even 7 or 8 (Warranty issues) and making sure they are addressed as necessary.

Bill

 

[Randy Stewart]

the ranking '1' is where a mistake proof, or error proof is in place, thus it is impossible for the failure mode to happen.

A low detection rating means that the "Failure Mode" happens I am just great at catching it. I can't produce the part etc. But it is after the failure occurs. The customer doesn't see it but I still have rework or sort or whatever. I just can't pass a bad part. At least that's how I read it.
Geoff said:

If the process has been mistake proofed

This is proactive and not based on Detection. This is the prevention of occurance not detection of the occurance.
Bill said:

you are not preventing the nonconformance but have a means to sort it out so it can't be shipped.

 

[Al Dyer]

Not being disparaging, but an occurance rating of 1 could raise a real red flag. Is any system truly controlled to a point where mistakes will never happen?

Does this go against the thought of "process variation"?

Can we really have a straight line on a variable chart showing no variation in our product or process?

Can topics on a contingency plan truely address that all possible variables to the process are covered?
----------------------------------------------------------------------------

A small power surge occurs which affects the reliability of a gage.

Water pressure drops and the CNC's are not replentished to the proper levels.

Someone introduces a similar, but incorrect wrong part into the process.

I guess my overall point is that a process can never be truly error-proofed and therefore in the FMEA there should never be a rating of 1.

And as was said, occurance needs to be matched with severity to get a true picture for continuous improvement activities.

MHO

Al...

 

[Randy Stewart]

Not the Intent

Al,
I think I understand the point you are trying to make. However, I don't agree with the examples given.

A small power surge occurs which affects the reliability of a gage.

Water pressure drops and the CNC's are not replentished to the proper levels.

Both of these are outside the scope of a Process FMEA and should be addressed under different documents (#1 - MSA, #2 - Maintenance).

Someone introduces a similar, but incorrect wrong part into the process.

On pg. 31 of the FMEA book, the last sentence under 10) Potential Failure Mode reads " . . . the assumption should be made that the incoming part(s)/material(s) are correct.
So, this too, falls outside the scope of the FMEA.
When developing the PFMEA we have to be very careful that we do not get into "theoreticals" and make it so deep it is no longer a useful document.

 

[YKT]

thanks for the feedbacks.

Yes, let say the process has been mistake-proofed, so can i put the detection as '1' ?
In doing FMEA, first you identify the failure mode and then the effect to the customers, thus you assign a Severity ranking.
Then, with fishbone diagram, we come up with the causes, and assign occurence rating to each of them, and lastly see the current process control we have.
If at the last stage, we realised that the process has been mistaked proofed, do we take out the failure mode (erase it) ? or do we put Det =1.

If we still put Det=1, we may have Sev=8, Occ = 5, (based on old historical data), and Det = 1, which the RPN = 40, do we still have to take action ? Basically, no action, right, asthe process has been mistake-proofed.

Any comments ?

 

[Al Dyer]

Re: Not the Intent

Originally posted by Randy Stewart
Al,

...When developing the PFMEA we have to be very careful that we do not get into "theoreticals" and make it so deep it is no longer a useful document.

Very true, it is easy to get too theoretical and make the FMEA tool useless in real-time application.

And you are right to disagree with my examples, they should me addressed by a process that "could" feed into a possible mention on a FMEA.

My concern is stating an "Occurrance" rating of 1 in an initial FMEA leaves no room for improving the occurance rating. I can see starting with a 2 then after monitoring the process for a period of time and collecting imperical data, revising the the rating to a 1.

I believe my opinion is the result having many FMEA's rejected for occurance ratings of 1, whether or not based on actual data.:ca:

Al...