The FMEA Mini-Series: Using an FMEA vs. SxO for Prioritizing

[MSAFAI]

Dear Ibrahm,

I think there are two methods to use:

1- Give a D number for the TOTAL control system (cause + failure detection)

2- Use the lowest D number.(this is what I prefer)

Colleagues, please comment which approach is better?

regards,
MSAFAI

 

[Al Dyer]

If there are multiple process controls for a given cause, each of those controls should have its own detection ranking.

Under failure mode "xyz" I could have two separate causes.

I could have cause "A" that is controlled by 3 different controls. Each of those controls would need a separate detection ranking.

Cause "B" is controlled by 1 control method and would require 1 detection ranking.

I guess this all leads to the fact that we can have 1 failure mode that has multiple causes, controls, and RPN's.

As with all FMEA's, there is room for discretionary use with the knowledge that we need to be able to back up our reasoning for using whatever ranking systems we develop. The ranking systems in the AIAG are only guidelines as long as we keep it as consistant as possible.

ASD...

 

[Ibrahm]

Thank you guys. Those makes me think about some ideas.

 

[Marc]

FMEAs - The Mini-Series

You're welcome!

Blast from the past. Anyone have anything to add?

 

[s-bell]

If I use the first current control, D will be 5, so RPN = 240. Need action (I use RPN 100 as a limit)
If I use the second one, D will be 2, so RPN = 96. No need action.
What should I choose?

I think that the RPN number has a lot to answer for, as a lot of companies use this as a limit (around the 100 mark). Whilst I agree that the RPN number as intended is a useful tool for action prioritisation, we ( a T1 Automotive Supplier) go a little further and assign FMEA actions based on the following criteria.

1: High Severity (a severity rating of 9 or 10) including those rated as OS (operator safety).
2: Those with Internal / External significant characteristics.
3: High RPN (a RPN rating of >=100)
4: Other items identified by the team that could reduce a high occurrence or detection rating.

This increases the resultant actions from the FMEA but also ensures that High Severity items are also covered (more often than not the RPN for these is > 100 anyway), if the severity cannot be reduced from a resultant design change then the team must make a decision as to the management of the risk in production.

I would therefore recommend expanding your limits to look beyond just the RPN, even if no action is required, state this on the FMEA as agreed with the team and at least it shows that you have specifically reviewed the issue.

 

[Bill Ryan - 2007]

Another "methodology" I have heard of is using SxO for prioritizing. Using that approach, I would think, would "lead" you to have better detection implementation - at least for new products.

Bill

 

[Rob Nix]

I have been doing FMEAs for design, equipment, and process for not an insignificant number of years, and the one conclusion I come to is this: variable data of any kind (whether a severity of 8 or 9 or 7.826, or an RPN or 99 or 101) is simply a feel good exercise with no real value.

Also, using severity if high, even though the RPN is fairly low, may mislead: since it is somewhat of a brainstorming exercise, nobody will be vetoed if they state the failure mode - "machine may spontaneously transmogrify into the likeness of Jabba the Hutt and suck all the air out of the shop"! It's severity is 10, but its probability of occurrence and undetectability is somewhere in the neighborhood of 0.000000001.

The REAL VALUE of the FMEA is the exercise itself - and the talent utilized to create it! It all comes down to one attribute. It is either an issue that will be addressed, or it will not be addressed. Period. So for each failure mode and probable cause, the group determines (naturally based on its S, O, D, or RPN - or gut feel) whether action is needed or not. If so, assignments, responsibilities, and target dates are established. The RPNs or S values MAY be used to prioritize tasks if there are many.

Allow me to curl up into the fetal position before the pummeling begins.

 

[Bill Ryan - 2007]

If there is any "pummeling", include me. I agree with Rob's statements.

The only issue I have is with the B3 and Tier 1s who have decided it's "their way (with their bogies) or the highway". IMO, they are the #1 CAUSE for FMEAs not being used in a manner useful to a company.

Bill
(Wow - that came out a bit stronger than I think I intended )

 

[Tim Folkerts]

I wasn't around the first time this thread was discussed and found it highly thought-provoking. My question is "why tie the values to an arbitrary 10 point scale?" A severity of 10 is generally more than 10 times worse that a severity of 1. I know it would a big challenge to get the numbers, but why not come up with a score that really matters to management - some sort of cost/benefit analysis.

importance = (cost per occurrence) x (probabilty of occurrence) / income


If you can come up with good estimates, this seems to work wonderfully.

So for example:
1) Paracute school: with no backup chute, 1/10,000 jumps has a fatal error, which would cost $5.1 million (lawsuits and loss of future income). You make $200 profit per jump.
importance of main chute failure
= ($5,100,000) x (1/10,000) / ($200) = 2.55 = 255%
ie the expected cost of the failures is 2.55x your income and your net profit is $200 - $200*2.55 = -$310 per jump

with backup chute, 1/1,000,000 has a fatal error, but added expenses cut $40 out of your profit. Also, the lawsuit is less because you took appropriate precautions
importance of both failing
=($1,100,000) x (1,000,000) / ($160) = 0.0069 = 0.69%
importance of main failing
= ($10,000 in lost income) x 1/10,000 / $160 = 0.62%

i.e cost of failure = 1.3% of income, and your net profit is $160 - 0.013x$160 = $158 per jump.

2) Lug nuts: with 1 lug nut, you make $1000 profit per car, but you have a 1/10,000 chance per car that the wheel falls off and you loose a $5 million lawsuit. With 5 lug nuts per wheel, you only make $995 per car, but the odds of a lawsuit drops to 1/1 billion. You do the math.

Basically, change the severity scale from 1-10 to cost per failure, and change the occurrence scale from 1-10 to 0-1 (odds of failure).

Tim Folkerts

P.S. Two additional points. 1) The cost is not necessarily proportional to the # of failures. It probably costs Goodyear ~$100 to deal with one bad tire, but I'm sure it costs a lot more than $1,000,000 to deal with 10,000 bad tires. 2) Putting a price on human life can be difficult and politically incorrect, as Ford learned with Pinto gas tanks.

 

[Bigfoot]

Bogies / Targets for FMEA

If there is any "pummeling", include me. I agree with Rob's statements.

The only issue I have is with the B3 and Tier 1s who have decided it's "their way (with their bogies) or the highway". IMO, they are the #1 CAUSE for FMEAs not being used in a manner useful to a company.

Bill
(Wow - that came out a bit stronger than I think I intended )

Interesting that you mentioned the bogies / targets for FMEA here. I revisited the GM Statement of Requirements that their WWP group issued in Dec. 2002 which has a PFMEA bogey rpn number of 40 in their list of items suppliers are reuired to utilize / demonstrate / provide or get a signed waiver if they exceed rpn value of 40.