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

FMEA Detection Ratings - Multiple layers of acceptance?

T

tsparks

#1
For a detection rating - if a product goes through multiple operations and at those operations the same dimensions are measured repeatedly throughout the process, does this mean I could use a detection rating of 3 or 4 depending? I'm confused as you can see. Any help is appreciated.
 

Miner

Forum Moderator
Staff member
Admin
#2
tsparks said:
For a detection rating - if a product goes through multiple operations and at those operations the same dimensions are measured repeatedly throughout the process, does this mean I could use a detection rating of 3 or 4 depending? I'm confused as you can see. Any help is appreciated.
It would depend on your particular controls. If you are using a sampling scheme at each operation, I would not reduce the detection rating at all. If you are 100% inspecting at each operation such that you end up doing a 300 - 400% inspection then you could justify reducing the rating slightly, but not to the level of 100% automated inspection or mistake-proofing. Recall that 100% inspection is about 85% effective. 200% will imrove it slightly, lets say 89%, 300% to 92%. The exact percentages are not important. The point is no matter how many time you 100% inspect, it is not 100% effective.
 
R

Rob Nix

#3
If inspection is 85% effective, then there is a 15% chance of missing the defect. If there are three "inspections", then there is a (.15 x .15 x .15 =) 00.3375% chance of missing the defect, or 3,375 out of a million (1 every 296). But all of that applies only if it is visual inspection, and has a margin of error of a football field. "Measured repeatedly throughout the process" could range from highly repeatable automatic gaging to a rough visual comparison.

Now, throw all of what was said above away. Ratings (such as for detection and occurence) are nearly useless anyway. Give it a 2, 3, 4, or a 5; whatever you want. In the end you will either choose to do something about that line item or you will not.

tsparks, please describe the failure mode line item information you are concerned about and I'll explain further. Thanks.
 

Jim Wynne

Super Moderator
#4
Miner said:
It would depend on your particular controls. If you are using a sampling scheme at each operation, I would not reduce the detection rating at all. If you are 100% inspecting at each operation such that you end up doing a 300 - 400% inspection then you could justify reducing the rating slightly, but not to the level of 100% automated inspection or mistake-proofing. Recall that 100% inspection is about 85% effective. 200% will imrove it slightly, lets say 89%, 300% to 92%. The exact percentages are not important. The point is no matter how many time you 100% inspect, it is not 100% effective.
I'm not sure I understand the reasoning here. First, there is no empirical basis that I know of for the 85%-effective bromide. I think the intent of it has always been to serve as a warning that 100% inspection is likely to be less than 100% effective, and that rational sampling might yield better results more econmically in the long run.

For the sake of discussion, though, let's assume that it's accurate (leaving aside the wealth of variables such as lot size, inspector acumen and the nature of the defect). If we do 100% inspection x times (where x is >1), and the 85% thing is statistically reliable and all else is equal, then the effectiveness is going to be 85% each time. But wait a second...if we have a finite lot size and some known quantity y defectives in the lot, and the first 100% inspection results in some of the defectives being discovered, the second and subsequent inspections should be less effective because there are fewer defects to detect. This assumes, of course, that the purpose of the inspection is to end up with a defective-free lot.

The point here is that we should always try to determine the method of inspection that's most likely to yield the desired results. If the lot is 20 pieces, the defect is glaring and the inspector is well-trained and experienced, 100% inspection might be just the ticket. On the other hand, with a comparatively large lot size, defects that are difficult to detect or evaluate, and inspectors who had seven beers for lunch, sampling is probably a better idea.
 

Jim Wynne

Super Moderator
#5
tsparks said:
For a detection rating - if a product goes through multiple operations and at those operations the same dimensions are measured repeatedly throughout the process, does this mean I could use a detection rating of 3 or 4 depending? I'm confused as you can see. Any help is appreciated.
Aside from Rob's good advice as to the significance (or lack thereof) of the actual number you choose, if your PFMEA covers all of the operations, each one should have its own detection rating. All but the last one should address the likelihood of value being added to undetected "bad" parts. The last one should reflect the likelihood of defective end product getting to the customer without being detected.
 
B

Brian Myers

#6
Placing focus where actions count

Rob Nix said:
Now, throw all of what was said above away. Ratings (such as for detection and occurence) are nearly useless anyway. Give it a 2, 3, 4, or a 5; whatever you want. In the end you will either choose to do something about that line item or you will not

I agree with Rob on this one. In all the DFMEA and PFMEA I have done, DETECTION rating was the least useful in the decision making process. Occurance and Severity were always the more important factors. The higher the Severity, no matter the detection or occurance, the more attention that line item would get. The Occurance was also usually focused on because there were mathematical "certainties" to the ranking given and there were usually identifiable (and sometimes even cost-effective!:rolleyes: ) ways to improve the Occurance Rate and therefore lower the ranking.

If you want to use the RPN as your "trigger" for action, set your threshold values assuming all Detection rankings are a 4 or 5 (sort of average). This way your thresholds for action will be more sensitive to the other factors.

I also always looked at it like this...

Occurance I can control (preventative/proactive approach)
Severity I can design down (DFM/DFA/Safety Factors)
Detection is "luck o the draw" (not to mention a waste of money - too little too late, you already made the scrap!:mad: )

Brian
 
R

ralphsulser

#7
We were taught in Quality 101 that 100% inspection is at best 75% effective, and have seen the same type results in actual applications. This applies to visual and Go/No-Go gauging.
 

Jim Wynne

Super Moderator
#8
Brian Myers said:
The Occurance was also usually focused on because there were mathematical "certainties" to the ranking given and there were usually identifiable (and sometimes even cost-effective!:rolleyes: ) ways to improve the Occurance Rate and therefore lower the ranking.
There is often history of the same or similar parts that will (should) contribute to the objectivity of the "detection" rating. It shouldn't be a very difficult thing to estimate the likelihood of some defect x going undetected.

Brian Myers said:
If you want to use the RPN as your "trigger" for action, set your threshold values assuming all Detection rankings are a 4 or 5 (sort of average). This way your thresholds for action will be more sensitive to the other factors.
I'm fairly liberal in my own review of supplier PFMEA documents; I would rather see evidence of defect-prevention efforts that make sense as opposed to a strictly orthodox PFMEA document that looks like it was put together by somnambulists. One thing I will never accept, however, is "trigger" RPNs. I've looked at literally thousands of these things, and one thing I can assure you is that if the trigger is RPN > Y, then the biggest RPN is likely to be Y - 5. Trigger values are self-fulfilling prophecies.
 
R

Rob Nix

#9
"Ralph, see there, 75% fits within the 85% +/- a football field tolerance zone," he said, tackling the issue acceptably.
 
Top Bottom