PFMEA Controls - Rating Prevention, Detection and Occurrance

[Jools]

Please can anyone help me with a question I have received from one of our customers.
"When creating controls for an PFMEA if you only have Preventative Controls and no Detection Controls what score do you give the Detection Controls and what do enter into the Detection Controls box"?

Thanks

Jools

 

[Bill Ryan - 2007]

Welcome to the Cove

If you have no means of detecting a Failure Mode, the Detection rating is "10".

I have seen PFMEAs where the Prevention Control value will "override" the Detection Control value. This will only "work" if you truly have Root Causes and the means to prevent the Failure Mode from occurring (a prevention value of "1"). That's a bit too risky of an approach for me. We are a die caster and there are just too many interactions of the process parameters that have a bearing on product quality.

 

[D.Scott]

Welcome to the Cove Jools.

If you had no way of ever detecting the nonconformance, you would have to put a 10 in the box. Under "Current Controls", you should put the process control you currently have for preventing the defect. If you do an inspection or test at a later stage is it possible it will show up there?

Review your process and think about the in-line and follow up inspections or tests. Maybe what you are doing has to match up with another part further down the process. Determine if there is some way to find a defect and if so, use that to determine your detection number from the chart in the FMEA book. If you really don't have a way to find a problem, then I guess you are stuck with a 10 until you can build a detection into the process.

Dave

 

[Howard Atkins]

Welcome to the Cove
As I see it this discusssion is a linguistic mistake, what has been said is that if you mistake proof something and it is impossible to make a mistake and you have no detection then the D=10, why mistake proof??
This could in fact be a result of a FMEA.

The handbook says page 49 and 50 (paraphrased)

Current process controls are divide into 2 prevention and detection to assist the team in clarity of thinking.

Detection rank is assosiated with the best control,

the prevention is one of these controls(HA)
Proof of this

see table of ranks page 53

Detection - Controls certain to detect - Discrepant parts cannot be made because item has been error proofed by process/product design-1

The validation of the mistake proofing belongs to the design stage.

If you had to detect every thing we would return to 100% inspection which is ranked 8

 

[Bill Ryan - 2007]

Howard
I've heard you're argument before (and even used to use it with my customers) but it doesn't hold water with my customer base. Reading the AIAG Detection Table listings, everything deals with your ability to detect the Failure Mode once it has occurred. You can have the best preventions around but if you have nothing in place to detect a Failure Mode (ANY Failure Mode), the rating needs to be a "10". The Prevention Controls have a direct influence on the Occurrence value, not the Detection value.

 

[Howard Atkins]

Bill,
I don't agree.
Whilst it is true that if it is mistaked proof then the detection should be 0 which AIAG don't beleive exists
but the handbook says that because ocurerence is low then the detection is not necessary low becuase you need to find the 1 in a million.

If this is the case then it should not be in the FMEA but the customer etc awants to see that you use Poka yoke
example, a failure mode had an occurence of 5 and a detection of 5
and to reduce this there was a design change which made the occurrence impossible
Do you leave the controls in place to detect nothing.
This is playing with words cause we know as manufacturers and not as SQE's that this is what we need in practice, they are making us play word games and not do the work.
I had a similar discussion with a supplyer about his FMEA in regard for a contamination issue, he said look I have a 100% inspection, he could not explain why he had not cgone to the source of the contamination.

The slogan of the automotive industry is prevention is better that detection, every where except in FMEA

 

[Marc]

made the occurrence impossible
Do you leave the controls in place to detect nothing.

That depends upon the severity of the potential effects (such as death). The key word here is 'impossible' which is a pretty strong word. I have seen error proofed processes go bad - for example an entirely automated process. It took over 6 weeks for the engineer investigating how a customer received a defective brake assist assembly. I remember my laughing at this guy - a guy who was new to automotive and he kept telling me that it was 'impossible' for a bad unit to pass. Turns out an unauthorized 'rework' procedure was being used on 3rd shift and there was no way to test the unit after the rework because it was an inline test which you could not put an assembly in.

I'd stay away from assuming any process is entirely error proofed without any 'possibility' of a failure somewhere.

 

[Bill Ryan - 2007]

Howard,

If you are dropping the Occurrence to "1" because of a process design change (there can't be an Occurrence of "0" or you have no Failure Mode to address on the PFMEA- OR - you have no cause for the Failure Mode which means an Occurrence rating of "10"), and you decide to "drop" the inspection, the Detection rating goes to "10". I don't see anything wrong with an RPN of "100" (worst case) with a Recommended Action of "None" as long as you have the documentation to back it up. In fact you could elaborate in the Recommended Action column as to the decision for no inspection needed.

Yes, automotive pushes "prevention" and you're example is a good one to illustrate an achievement to that end. However, I would think you would need to "inspect" for a time to get a confidence level that you could back off the inspection. That would be your documentation. I'm normally pretty leery of a "1" anywhere on a PFMEA (even a fully automated process) because there is always, somewhere, the "human factor(s)" involved.

 

[Jools]

Thanks

Thanks guys, you have all given me food for thought, but I think I will relay back to my customer that he ought to be putting a 10 in the Detection Controls, as he doesn't have any!

 

[mahoc]

This is another question that has an relationship with the thread.
One customer told us that Prevention controls DON'T must been ranked, only detection controls. What do you thing about this?

Martin