Hello all

I am an industrial engineering student and I learned FMEA in school. My question is the ratings given to severity, occurance and detecion seems like qualitative, like something which one knows based on experience with process or product or design. Is FMEA primarily a brain storming session with persons from all levels sit together to discuss possible failures. This forum is really helping me understand things. Thanks for all your help and knowledge sharing.

Thank you
Raghav

My question is the ratings given to severity, occurance and detecion seems like qualitative, like something which one knows based on experience with process or product or design.
In short, yes. However, there are guidelines/standards available relating to the use of FMEA in specific areas, i.e. automotive and medical devices. Keep in mind, that it is a good idea to use the ratings in the guideline, since we have a tendancy to consider the risks etc. lower than they should be, if we have been involved with the production or development of a product/process.

Is FMEA primarily a brain storming session with persons from all levels sit together to discuss possible failures. Yes, it is basically a structured brainstorm - resulting in a list of possibilities for improvement. FMEA is a iterative process, and the benefit and the output of a FMEA session is entirely up to the team involved and the management's support of the process.

Thank you. I am sure my learning curve is going up through this forum, Thanks for your time and guidance.

In short, yes. However, there are guidelines/standards available relating to the use of FMEA in specific areas, i.e. automotive and medical devices. Keep in mind, that it is a good idea to use the ratings in the guideline, since we have a tendancy to consider the risks etc. lower than they should be, if we have been involved with the production or development of a product/process.

Yes, it is basically a structured brainstorm - resulting in a list of possibilities for improvement. FMEA is a iterative process, and the benefit and the output of a FMEA session is entirely up to the team involved and the management's support of the process.

Hello all

I am an industrial engineering student and I learned FMEA in school. My question is the ratings given to severity, occurance and detecion seems like qualitative, like something which one knows based on experience with process or product or design. Is FMEA primarily a brain storming session with persons from all levels sit together to discuss possible failures. This forum is really helping me understand things. Thanks for all your help and knowledge sharing.

Thank you
Raghav

You are correct. All the numbers generated are fairly arbitrary but should, with the right amount of analysis, be fairly accurate. Take the guidelines to heart.

The important thing to remember is to set a standard for re-evaluation of the process. Not being from the auto industry I'm not 100% positive but I think the generally accepted RPN number is 100. Anything higher (or if occurance and/or detection are high, such as >7) should prompt a reassessment of the process.

Good luck!

The important thing to remember is to set a standard for re-evaluation of the process. Not being from the auto industry I'm not 100% positive but I think the generally accepted RPN number is 100. Anything higher (or if occurance and/or detection are high, such as >7) should prompt a reassessment of the process.


No! RPN thresholds are not a good idea. RPNs should be viewed as relative, not absolute numbers, and there will be many times when something can be improved even though the RPN is less than the "trigger" number. There will also be cases where the RPN is higher than the threshold, but there will be nothing that can be done to improve the situation, and there might be several opportunities for improvement where the RPN is lower.

RPN is mostly useless, to tell the truth, and causes a lot more problems than it helps to solve.

there will be many times when something can be improved even though the RPN is less than the "trigger" number. There will also be cases where the RPN is higher than the threshold, but there will be nothing that can be done to improve the situation, and there might be several opportunities for improvement where the RPN is lower.

RPN is mostly useless, to tell the truth, and causes a lot more problems than it helps to solve.

Agreed that it should be taken as reference because of the possible situations you mentioned...but is there not typically a limit as which point it should be evaluated? The threshold I'm referring to is a generic number...not an absolute. Understood that sometimes things cannot be done...and understood that sometimes you can improve it when it "seems" there is no need to...but if there is no threshold then what's the point of putting a numbering system against it?

Agreed that it should be taken as reference because of the possible situations you mentioned...but is there not typically a limit as which point it should be evaluated? The threshold I'm referring to is a generic number...not an absolute. Understood that sometimes things cannot be done...and understood that sometimes you can improve it when it "seems" there is no need to...but if there is no threshold then what's the point of putting a numbering system against it?

Again, they're relative values. This means that if you look down the RPN column on a FMEA document, and see (for example) 20, 28, 30, 31, 25 and then 108, there's something going on with the 108 that needs to be considered. But if the threshold were 100, and that last number were 90, there would be no requirement for action. At the same time, and at the risk of repetition, there might well be something in one of those first five cases (or some combination of them) that can be improved.

In most instances, people who participate in FMEAs are already aware of what they need to look out for. The RPNs just provide a method, in many instances, of documenting relative risks. The fact that there are risks doesn't mean, however, that there is always going to be a reasonable strategy to prevent bad things from happening. In my view, the FMEA process, and the document that comes out of it, serve as evidence of due diligence, and not much more.

Thanks for all those who replied and started a discussion. Couple of them in the replies mentioned follow the GUIDELINES, are these AIAG guidelines or the guidelines of severity, occurence and detection with number 1,2,3...,10 mentioning for example hazardous with warning, hazardous with out warning in severity. Thank you all for your time.

raghav

I completely agree with Jim Wynne on this one.
You have to look at the overall picture, not just the RPN.

Which of the two scenarios has the greatest need to be addressed?
1. Severity 10 x Occurrence 1 x Detection 9 = RPN 90
or
2. Severity 1 x Occurrence 10 x Detection 10 = RPN 100

In scenario 1, although the failure doesn't occur very often, it is extremely severe and has a great probability of not being detected.
Scanario 2 the failure mode happens often, is undetectable, but has no real effect.
If a RPN threshold of 100 were set, you would spend all your time trying to reduce the amount of insignificant failures while potentially harmful product was beins shipped to your customer.

Thanks for all those who replied and started a discussion. Couple of them in the replies mentioned follow the GUIDELINES, are these AIAG guidelines or the guidelines of severity, occurence and detection with number 1,2,3...,10 mentioning for example hazardous with warning, hazardous with out warning in severity. Thank you all for your time.

raghav

Yes they are.

Ok, obviously something as been taken the wrong way.

Never did I mention that the RPN# is the bible and that there should be a hard and fast point at which something should or should not be evaluated. Yes it is relative, yes there are other parameters to review before making a decision (as there always should be). What I did mention is that both the RPN and the individual numbers (S/O/D) should be reviewed. If the RPN is higher than acceptable look at the individuals to see where the culprit is. If the RPN is low, review the individuals as well just in case there is one errantly high number. The question is...what number is acceptable in terms of both RPN and individual? That is something that should be decided from the get go.

Never did I mention that the RPN# is the bible and that there should be a hard and fast point at which something should or should not be evaluated.

Well, you said,
The important thing to remember is to set a standard for re-evaluation of the process. Not being from the auto industry I'm not 100% positive but I think the generally accepted RPN number is 100.
Yes it is relative, yes there are other parameters to review before making a decision (as there always should be). What I did mention is that both the RPN and the individual numbers (S/O/D) should be reviewed. If the RPN is higher than acceptable look at the individuals to see where the culprit is. If the RPN is low, review the individuals as well just in case there is one errantly high number.

Sorry, Pazuzu, but what you're saying is that the S-O-D numbers should be evaluated regardless of their RPN product, which is absolutely correct, but also absolutely contradicts your threshold premise.

The question is...what number is acceptable in terms of both RPN and individual? That is something that should be decided from the get go.

No, it's not. The purpose of performing the analysis is to discover and evaluate potential sources of risk, so the purpose is at least partially defeated if you make up your mind "from the get go" that any RPN with a value greater than x requires special action.

I never said that 100 is THE threshold. I never said that any number in paricular is THE threshold. In fact, I never even said there should be a "threshold" until you mentioned it. What I did say is that I have been told by a few people from inside the auto industry that 100 is generally used. Whether or not this is the same parameter for other industries I have no idea. I also stated that the SOD numbers are equally assessed and that 7 is generally used by those same individuals.

Use whatever numbers you want...dont use whatever numbers you want...just do whatever works best for your application.

Sorry if my initial choice of words were interpreted wrong...that's the difficulty with forums.

Just to pass along another twist on the FMEA philosophy, I was trained in an earlier life to a "FEMCA" philosophy (the Shanin Risk ReduXtion philosophy) and "D-FEMCA" for design projects. When you approach an FMEA with this different spin, you are looking first at the "Effects" then the "Modes" and potential "Causes" as opposed to the "Modes" then "Effects". In other words, you look at the "Y" (effect) before you look for the "X" (root cause).

Using this approach, you still use the Severity, Occurrence and Detection numbers; however, setting "Bogeys" (as they were called to us) by setting an arbitrary RPN threshold was frowned upon. Instead, the approach was to rank the failures and systematically analyze them from worst to least.

The other "key" to not setting bogeys was so you didn't "stop" analyzing the data when you reached that magic threshold. The plan was to set a reasonable length of time within the project framework and then knock down the potential failure effects one by one until you run out of time. Of course, a lot of thought needs to go into this approach as well since you don't want to give yourself a month to fix a year's worth of problems - but it's merely a different perspective on a time honored risk analysis tool.

Just my :2cents: ... :bigwave:
-Steve

Agreed that it should be taken as reference because of the possible situations you mentioned...but is there not typically a limit as which point it should be evaluated? The threshold I'm referring to is a generic number...not an absolute. Understood that sometimes things cannot be done...and understood that sometimes you can improve it when it "seems" there is no need to...but if there is no threshold then what's the point of putting a numbering system against it?

Stay away from Thresholds or else all of a sudden you will find all the RPNs will magically come out below this threshold. Instead work on top 5, 10 or 20 items until you feel all risk is at an acceptable level or keep doing the continuous improvment thing.

Engineers are great at selecting three numbers that will come in below your threshold. Whoever first came up with the idea of a threshold on a subjective ranking scheme should be locked away.