Is it correct to have a detection (error proofing) in subsequent operation?

[Jsalazarv30]

Hello

i'm new in this FMEA world and i am reading and watching videos to get more knowledge also i took some courses but i still have lots to learn. I know here there are many experts and i woud appreciate a lot your help in this doubt i have.

Checking a pfmea i noticed that in one operation we have 100% visual detection for an omission part, but in the next station we have a vision camera for that omission... as far as i know, we should have the error proofing where the failure mode can occur, thats my first doubt, the second one is the next:
we have like this the fmea

St FM FE FC PREVENTION DETECTION
10 Omission of bushing .......... .......... .......................... Visual by operator
Vision camera (St 20)

20
Installation Spring omission ....... .. ....... . . ..................... Vision camera
of other components
My concern is if we should put in the st 10 the error proofing of the st 20, or only the one that we have there (Visual by operator)?

 

[Jimmy123]

In St20 for me it’s not clear, if the camera checks existence of the bushing too a second time? The description in the FMEA should be always detailed, that others understood the process too.
Ask the team: 5 W's and H Questions
WHY is this inspection necessary, can it be remove by a prevention?
WHAT characteristic, cause or failure mode have to be check by the inspection?
HOW does the inspection work (conditions of the inspection, capability, 100% or sample check only, operator self check or from a independent person,…’ and what is the reaction in case of a failure (bad parts handling, re-adjustment machine, …)?


Sometimes a good visual inspection is better as a bad camera vision system?

 

[Sebastian]

First, setting sampling frequency within PFMEA is useless.
Look at detection rating table, frequency has no impact on rating.
Method used has and visual inspection is poorly rated.
It is control plan, where frequency has to be set

Second, all activities related to failure mode prevention or detection have to be recorded in section (operation), for which failure mode was identified. So, inspection in operation 10 and inspection in operation 20 recorded in operation 10 section.

Third, inspection in operation 20 is not error proofing against failure mode identified in operation 10.
It could be only error proofing against failure mode of "usage of nonconforming component" identified in operation 20.
But, there are few wise sentences in FMEA manual, related to not considering failure of usage of nonconforming components/materials, if i remember it correctly.
Summarising, again it deserves to be moved from 20 to 10.

 

[Bev D]

In the fmea documentation all detection - regardless of where the detection occurs - belongs in the section where the failure mode occurs. Remember the FMEA is not a flow chart of the process but a listing of processes or functions and how the failure mode will be prevented, controlled and detected. So even if the only detection method is at final inspection prior to boxing the thing up, that inspection is listed in the section where failure mode can be created…

 

[Bev D]

First, setting sampling frequency within PFMEA is useless.
Look at detection rating table, frequency has no impact on rating.
Method used has and visual inspection is poorly rated.

Here is where my approach to FMEA differs from the common dogma. The ability of any detection method to actually detect a defect or failure is directly related to the sample size and sampling frequency as well as the physics of the detection method (science and MSA). I also do not use guessing or rating for either detection or frequency of occurrence. A FMEA is supposed to be used as a development guide and is supposed to be a living breathing documentation of the results of actual analysis, experiments and testing results.

So depending on the failure mode and frequency of occurrence visual inspection can be very effective or it can be useless. The same can be said for any detection method.

 

[Bev D]

If I understand your post correctly you are stating that 100% visual inspection isn‘t 100% effective, I agree.
There are many many posts here regarding the effectiveness of 100% visual inspection. My statement was that “depending on the failure mode and its frequency of occurrence visual inspection can be very effective or it can be useless.” My point was that no single or narrow range of ‘detection’ ratings without regards to the facts can represent visual inspection effectiveness. Only engineering and testing can do this. Quality is not a pre-ordained, look it up in a table, cut and paste exercise.

Some defects are obvious even to the untrained eye. Some require good visual acuity, high magnification, lighting, specific sample prep and specialized training.

If the defect has a very low occurrence rate, humans will naturally become lax and distracted leading to misses. A relatively high occurrence rate will result in in a high detection rate - not a perfect one, but a high one.

Sample size and frequency of sampling also play a part in the effectiveness - this is basic probability. It too is related to the frequency of occurence and it’s homogeneity.

All of this means that no single or narrow range of ‘detection’ ratings without regards to the facts can represent visual inspection ( or any method) effectiveness. Only engineering and testing can do this. Quality is not a pre-ordained, look it up in a table, cut and paste exercise.

 

[Ron Rompen]

Going back to the original question (where should the detection rating be placed) I would place the detection rating for visual inspection at Station 10 which is where it occurs. (DET at this station for this failure mode = 7)
The 100% automated inspection at Stn 20 results in a 'cannot build' and a DET score of 3 IF it will lock the part in station to prevent further processing.

 

[Jimmy123]

Going back to the original question (where should the detection rating be placed) I would place the detection rating for visual inspection at Station 10 which is where it occurs. (DET at this station for this failure mode = 7)
The 100% automated inspection at Stn 20 results in a 'cannot build' and a DET score of 3 IF it will lock the part in station to prevent further processing.

Looks like you have standard ratings, visual inspection is worst at a camera system. I disagree such ideas, because I can tell you story’s from real live, where a camera is very bab and does’t detect a obviously failure, what a human detect it with D=3 for sure. Bye the way, I suppose to improve the newest AIAG&VDA evaluation tables, there are some weak description, which are dangerous.

 

[Ron Rompen]

I don't disagree, Jimmy, however the existing tables are good as a guideline. In the past, I have developed my OWN tables for detection, but many customers are not comfortable with that, and insist on using the AIAG tables.

 

[Bev D]

Science people, science. I know that AIAG asks for things that fly in the face of established science and mathematics. And Customers who are stuck in the past and are loathe to think for themselves perpetuate the nonsense. But WE are smarter than that. WE should give good science based advice, not just rote answers to cut and paste approaches that only check the box on a form…Detection and probability/frequency of occurence should never be a guess. We should have and can have substantiating data that tells us what the occurence rate is and how effective our detection methods are.

Perhaps the ’standard’ tables can provide a guideline when just starting but they should not dominate the final result. FMEA is not a check the box activity.