iSixSigma

FMEA Template

Six Sigma – iSixSigma Forums Old Forums General FMEA Template

Viewing 26 posts - 1 through 26 (of 26 total)
  • Author
    Posts
  • #29730

    Prasad Malapaka
    Participant

    In the FMEA template, teh “effects” columns preceds the “Causes” column. But in practice we write the causes and then the effects. Due to this, we are filling up the Causes column then we are going back and fillup the effects column. This is a inconvenience.
     
    Can anybody say why it is designed like this ? How is this filled in general and how is this understood? (Probably I am filling it wrongly because I understood it wrongly ?)
     
    thanks
    Prasad

    0
    #76726

    Gabriel
    Participant

    Prasad:
    My opinion is that you are filling it wrongly, may be because of a wrong understanding.
    The logic steps are (it is done for a process FMEA, but can be easily extended to a deign FMEA):
    1) Define the objective(s) of the process: For example: From a strip of 1+/-0.1 thick, cut the washer with an external diameter of 10+/-0.5 and an internal diameter of 5+/-0.2, leaving it flat and with no burrs. Ussually all those objectives are written one in each row to analize them one by one.
    2) Failure modes: For each objective, what can go wrong? For example, if we take “internal diameter 5+/-0.2” we may imagine above USL, below LSL, too much ovality, no hole, etc. Again, each failure mode is written in a different rows.
    3) Effects: For each failure mode, how bad would that be for this process, for downstream processes, for the company, for the customer, for the end user, etc? The effects are only related with the correspondet failure mode. The causes has nothing to do here. For example, never mind why the hole diameter is below the LSL. If it is you may have a difficulties to locate the piece in the tooling in the next process, the customer may find impossible to put the washer in the screw, the company may need to rework or scrap pieces, etc. Then for each effect of a failure mode you assign a “severity” factor. You keep only the worst (highest) severity for each failure mode. But remember: The effects are the consequences of a failure mode, and have nothing to do with the causes for that failure. You don need nor want to know the cause of the failure mode to analyze the effect of that failure mode. Then there is no reason to fill the cause first and then the effect. However, it could be done like you do it because you dont need to know the effect of a failure mode to analyze its cause anyway.
    4) Causes: For each failure mode, why may this failure happen? Causes are also only related with the failure mode, and the effects have nothing to do here. For exmple for the same failure “hole diameter below LSL” you can have a worn out tooling, a strip too hard, a wrong set up, etc. Then for each cause you assign a “probability of occurrence” factor. In this case you must keep all idividual causes with its occurrence factor. Remember: Causes are generators of the failure mode, and have nothing to do with the effects of that failure. In a cause-effect flow we would have CAUSE ==> FAILURE MODE ==> EFFECT. We start from the failure mode (the middle of this chain) because we are in a “process FMEA” and the inputs of the FMEA are the process objectives, those that when they are not reached we have a “failure mode”. After analyzing the failure mode, which end of the chain you work on first makes no difference, as I see it.
    5) Controls: You have two kinds of controls. One type answers the question “How can be the failure mode be detected?”. You can have no, one or more of these controls for each failure mode. For example, for “hole diameter below LSL” you can have an iinitial control in the metrology room, a go-nogo gage on 5 pieces each hour, etc. For each control of a failure mode you assign a “probability of no detection” factor. You keep only the best control (the one with the lower factor). The other type of controls answers “Hou can the cause for the failure be detected?” You can have no, one or more of these controls for each cause. For the cause “worn out tooling” you can have a visual inspection after each batch before taking the tooling to the shelf. For each control of a cause of failure you assign a “probability of no detection” factor. You keep only the best control (the one with the lower factor).
    6) RPN (risk priority number): Now for each failure mode you have several effects with a “SEV” each but you keep only the highest. For each failure mode you also have several causes with a “OCC” each. Also for each failure mode you have several controls with a “DET” each but you keep oly the lower. And also for each cause you have several controls with a “DET” each but you keep only the lower. We know that NPR is SEVxOCCxDET. How do you combine all this in a NPR? You will have a NPR for each CAUSE. Just multiply the higher SEV of the effects of the failure mode caused by this cause, times the OCC of this cause and times the lower between the lower DET of the controls for failure mode caused by this cause and the lower DET for the controls for this cuse. Believe me, it is easier to do it that to write it.
    Just my opinion. Hope this helped.
    Gabriel

    0
    #76728

    Marc Richardson
    Participant

    Prasad,
    Grabiel has some interesting, if unorthodox, advice in the completing of a Process FMEA. I would highly recommend that you visit http://www.aiag.org and obtain a copy of the FMEA Manual. It is written primarily for use in the North American automotive industry but you will find much solid guidance there regardless of the industry you are in. As far as your question is concerned, the reason that effects preceed causes is that a failure mode may have more than one effect. Also, even though causes preceed effects, the effect is what one experiences first, then one goes looking for the cause.
    Marc Richardson
    Sr. Q.A. Eng.

    0
    #76735

    Gabriel
    Participant

    Marc
    What I said is not very orthodox, is it? I think you mean mainly because of the way you keep the worse of these, the better of those, and each of them, and then make the multiplication.
    What I said is based in the AIAG’s manual + a training material from SETEC, a training company sponsored by Ford.
    Before I got the SETEC’s I had serious problems to make the multiplication. I think that we all agree that both causes and effects are not directly related one to each other but they are both linked eith the failure mode. Once a failure mode has hapened any one (or more than one) of the effects can occur, regardless of which one was form all the probale causes was the one that cuased the failure mode this time. Then, if you have lets say 4 effects and 3 special causes for the same failure mode, how do you make the multiplication to cover all causes-effects possible combination. You should multiply each SEV with each OCC, then you have 12 results. And we didn’t start with the controls yet! If you have several controls, some of them attaining some of the causes and some of them attaining the failure mode itself, how do you combine the 12 previous products with all this controls to get the NPR?
    I saw many times FMEAs where for each failure mode there was one effect, one cause and one control. Sure, it is easy in this way. But for me it looks that the team has a big lack of imagination or that thy just wanted to complete the FMEA for the FMEA itself, because they “had” to do a FMEA, and not trying to use it as a preventive and improvement tool.
    I adopted the solution from SETEC because it is sound. If the method is wrong, I dont want to avid resposibility saying “it is SETEC’s” bacause I agree with it.
    The cause is what comes first. Any of the causes can cause the filrue mode. But the cause alone is not enough. If the cause were detected immediatly and the process stopped, no part with a failure would leave downstream. If the cause was not detected on time, but the filure mode was detected, the deffective parts could have been stopped. A failure mode has an effect when the cause actualy happens and every control controls (even the better control) fails. That is why, for each cause, we multiply the OCC with the lower SEV (better control) among all the controls for this cause and for the failure mode. Now, which effect from all the effects of the failure mode will happen? We don’t know, so we take the worse.
    Note that this is much easier that making, for a given failure mode, all the possible combinations of SEV, OCC and DET. And making all tha combinations would not be fair anyway, because your higher NPR will come from the worse DET, when you may have another control for the same with a better DET. For example, measuring 3 parts every hour to assess to make a follow up, but all parts go through an attribute poke-yoke. A filure in detecting with the sample does not mean that the failure will pass the poka-yoke.
    Comments?

    0
    #76954

    Padmanabh Kelkar
    Participant

    In FMEA, Effects come first, because with Worse effects we decide on severity rating. Worse effects are studied first because,most of them may affect the CUSTOMER directly. Causes and related actions are something internal to your company/product/process. Hence the sequence of columns. I hope I have been able to clarify your doubts.

    0
    #76956

    Mike Carnell
    Participant

    Gabriel,
    Do you actually believe you are an orthodox thinker?

    0
    #76957

    Mike Carnell
    Participant

    Prasad,
    Just an FYI. There are no Six Sigma Tool Police and nobody gets 5-10 for for not following the rules in Statistical Jail. If it makes more sense to you to do the analysis in a different order then do it that way.
    People think differently ie. inductive and deductive reasoning. The point of the tool is not to make you follow the way some else thinks. the point is to get a completed document that can help you evaluate risk (the RPN).
    The manuals are OK to give you some advice and guidelines but you have to make it work for you.
    Good luck.

    0
    #76961

    Gabriel
    Participant

    I hope I am not.

    0
    #76966

    Mike Carnell
    Participant

    I don’t have any data that says you are. Very admirable trait.

    0
    #76970

    Ravi Athalye
    Participant

    Mike,
    This is the only game where rules can be changed, provided you are consistent with some higher `rules’ to get the results..reduce variation.
    I like your spin on calling Statistics a jail.
    I come back to our earliar exchange under banner of Variation when you groped to analyze the entire population of issues with the suspicion that stratifying it would throw out exactly the  problem whose solution would lead to a breakthrough.
    There is a congruence in the approach of a child and an expert ; the former  asks the right questions out of purity of innocence and the latter out of the courage of  conviction. Though we all begin with the innocently inquiring mind only a few at some stage become the lions of challenge.
    I  am confronted by my 12 year on why authors that communicate effectively do not always stay withing the rules of grammer.
    Six Sigma appears to foster this single mindedness of  reducing variation and if is through an  un orthodox method …so be it. (Though it would be worthwhile to understand and analyze the un orthodox approach and formalize it as a tool for posterity.)
     

    0
    #76971

    Nitin Agrawal
    Participant

    The Effects in the FMEA Table are the Effects of the Failure Mode, filled up in the previous column and not the effect of the causes. So you do not have to fill the causes first and then come back to fill the Effects.
    I think the sequence of the columns are absolutely logical. First we see what are the Failure Modes, then what will happen due to these failures (the EFFECT): will this affect the customer or will this create some internal problem for the process. Next we try to find out on the causes of the Failures, a root cause analysis type. Lastly we try to evalute what are we currently doing to control the occurance of these causes and consequently the failures.

    0
    #76978

    Mike Carnell
    Participant

    Ravi,
    When we were teaching Wave II at the Allied Automotive deployment I wrote the formula for Standard deviation on the board and swaped the x and the x-bar. It drove a person (statitician) in the class crazy. The bottom line is the next step squares the value so the order is irrelevant.
    Some of the things we do in SS mean something and some things don’t. The problem is that we get people who want a step by step formula and they don’t deviate. Deming called it “Instant Pudding” and said it didn’t exist. When I first learn a tool I don’t wander around a lot – but once I understand how something works I make it work for me – not me work for it.
    As far as people making the judgement about what is logical and what is not. I got a recomendation to read a book “A Whack on the Side of the Head.” ISBN 0-446-67455-9. This is a great book. A quote “….the number of things that can be thought about in a logical manner is small, and an overemphasis on the logical method can inhibit your thinking.”
    If doing SS is who cam memorize the most tools and who can memeorize the steps then it will ultimately die as a methodology. It will turn into a “Night of the Living Dead” with a heard of zombies wandering the vast Serrangetti plane of the business world going “you must fill out your FMEA from left to right because those are the rules.” If this is where SS is headed I don’t want anything to do with it. You see it repeatedly when you read some strings and work with BB’s. They run a Normality test and then they freeze. Or they fail the test, try to transform it and freeze.
    The thing I saw at GE when we started rolling out SS in 96 was they have a complete irreverence for rules. They suck every bit of knowledge out of you they can, digest it, internalize it to fit GE and make it part of the culture. The “d” wasn’t part of MAIC before GE. Is it good or bad? Who cares. It was the change they felt they needed to make it work at GE. Do they care if it works for someone else? Not at all. They did it to make it work for GE.
    Making hard and fast rules around a soft tool like FMEA is just plain ignorance. Yes there is a general method you use when you get started but if you find a way that works best for you do it then you do it.
    As far as groping for types of variation. I’m not groping for anything. I threw out the labels a long time ago as useless. Like I told someone in a early post in the string if it works for them then do it (label). The relationship of Y = fx works for me and even this has to be fluid enough for some of the x’s to become Y’s.
    Next time you encourage someone to think outside the box. Figure out what rule were they taught that put them in the box to begin with. If you have access to a library take a look through the anthropology section. They do a thing called an ethnology. It is an interesting discipline. If they have one for your culture read it objectively. I read one on the Nacirema. I thought they were a pretty strange group. At the end they told you that Nacirema was American spelled backwards. They did it so we didn’t read it defensively. Just another box.
    Good luck.

    0
    #76985

    Ravi Athalye
    Participant

    What logic has done to mankind is decrease the ability to process information  concurrently in multibit stream making jumps from one node of understanding to the other without visible threads of connectivity (of logic)  to a sequential processing entity filtering out the infinite information until the processor is ready to consider it . When it does,…. the reason for evaluating it is lost in myriad of  circles being run around the main issue.
    I have had occasions to hear some experts converse with commonality of  background with evocative references to books (nodes) . I use nodes as pegs or addresses on which you can hang enough experience or memories as the interacting members can understand. The process speeds communication along at expense of exclusion of the observers. At that point I have thanked the ability to logic…..there was no other way for me to grasp the subject ……sacrificing the ability to create new subject matter…having lost the `enlightnent’ that must come from seeing infinity in harmony. (That is how some of the prophets were  able to undertand the `absolute’.)
    Yin and Yang must exist . Both the intuitive and the logic because humans too have variability in their IQs depending on the time of the day, subject matter and the status of emotion (emotion being a varying  concurrent process).
     
     
     

    0
    #76989

    Mike Carnell
    Participant

    Ravi,
    You have intellectualized the thought process. If that works  for you that is great. Doesn’t work for me.
    Good luck.

    0
    #76990

    Gabriel
    Participant

    Ravi
    It is clear (for me) that the effect is the effect of the failure mode, and not of the cause. Also that the cause is the cause of the failure mode, and not of the effect.
    Once you identified the failure mode(s), does it make any difference whether you work with the effects first and then with the causes, or with the causes first and then with the effects? I think it is the same (as long as you understand the first paragraph).

    0
    #76995

    Ravi Athalye
    Participant

    Mike,
    The point is that when one steps out  of  a box …one steps into a larger box. PG Wodehous would say `wheels within wheels’.
    SS too in its entirety exist in a box. If  you decide to step out of it one has to confront the  the larger issue. As you yourself  seem to promote that stepping out of the box can yield neat stream lined solutions that needed the logic of the lesser box.
    Ultimately we fail because we settle on being `Children of a lesser box’!

    0
    #76999

    Mike Carnell
    Participant

    Ravi,
    So we have come full circle Using Common, assignable, etc is children of a lesser box.
    As far as the SS box. This wasn’t called sixsigma when we were using it at Motorola (pre 88) it was just how we did our job. It got yurned into SS as a corporate wide initiative. It got comercialized and sold. Now it is being turned into an intellectual esoteric playtoy.
    You can reference some of the old posts with Withheld. I don’t consider myself restricted by perceptions of SS because I do whatever I am going to do which is called “operating results.”
    At the end of the day you can create the most perfect DOE, pontificate on the difference between SIPOC and a process map, etc. but until you go out and make a change you haven’t done anything of value that will enhance your businesses probability of survival.
     

    0
    #77001

    Ravi Athalye
    Participant

    You have just defined breakthrough in two parts:
    – break through in thinking……breaking the box paradigm
    – break through in effect……….survival of the corporate business. (Note not survival of the model..)
    The connection between thinking and doing is : skill, courage and willing to let go the `who I am’ (ego)…also called willingness to change. (I am reminded of your comment on middle management comfort zone.)
     

    0
    #77002

    Mike Carnell
    Participant

    Ravi,
    My definition of breakthrough comes from Juran.
    At this point my thinking hasn’t changed from anything I thought when this started.

    0
    #77004

    Ravi Athalye
    Participant

    Thinking may not change , but given the same set of questions would the arguments and conclusions change…?  A whole new set of notions not appropriate for this forum.
    We all part intact .
    Thanks
     

    0
    #77006

    Mike Carnell
    Participant

    Ravi,
    I am not sure what you thought changed. We have pushed the business link since we took it into Allied in 95. I dumped the common cause, assignable cause, etc. over a decade ago.
    Good Luck.

    0
    #77008

    Ravi Athalye
    Participant

    Are you serious? Obviously you are. Purely from analysis point of view you must be able to seperate the chance variable from the assignable cause. If you dumped these classifications do you consider the entire population  of variance to perform the statistics?  Should you not `brush off’ the definitely known cause of variance which could be the off lier (hope thats the right word).
    I am pretty new to the commercialized version of TQM called SS. Would like to know……..would appreciate your inputs.
    Thanks.
    Ravi

    0
    #77010

    Mike Carnell
    Participant

    Ravi,
    Yes I am serious. It is NVA. The tag adds nothing. It takes you from a larger wheel to a smaller wheel – you become a child of a smaller box – multiple sources of variation become only two catagories – you lose resolution.  You have turned catagorical data into a dichotomy. If I want to know the actual effect of a variable and I treat them as catagories (factors) I can run a DOE and get an actual measure on its % contribution. Tell you what you can run with your two catagories and I will run with a pareto based on the size of the effect and lets see who gets a bigger reduction faster. How could you possibly believe that less information adds any degree of leverage to resolving a problem.
    The idea of a “chance” variation is a losing strategy. They all have a cause – you may not understand the cause but it happens because of something.
    As far as what I “should” or should not do really isn’t up for a vote. Like I have said on several occasions you do what works for you. If you have some use for Common Cause and Assignable Cause have a ball. Your logic came back around on you and satisfied me that I was doing OK. If you see it differently thats fine with me.
    As far as your comment about SS being a comercialized version of TQM that is pure unadulterated bull shit. Can’t really think of an esoteric way of putting it.
    Good luck.

    0
    #77012

    Ravi Athalye
    Participant

    Mike,
    – By eliminating variation whose cause is known (with confidence) you are not lobotomizing information. You are reducing the haystack in which the pin fell. If the cause is not known by all means use it for analysis.
    – By creating pareto you are actually  stratifying more than the two version stratification. (Common and Assignable)
    – There is nothing that SS offers that is in contradiction of TQM. It is infact continuation of  TQM …up the evolutionary process. Ofcourse SS has better audience and customers …hence my remark on commercialization.
    – Variation , cost of quality , failure rate etc is but measuting an object with different system….FPS or CGS or ……the object is the same and will have the same dimensions independent on what system you use.
    I guess SS has a nerve going through it that is readily agravated !
    Cheers !
    Ravi
     
     

    0
    #77013

    Mike Carnell
    Participant

    Ravi,
    We have strayed considerably from the original FMEA question into this fecal maelstrom of esoteric philosophy. Just to do a little maintenance: Prasad – you can fill out the FMEA in any order you choose. It is the accuracy of the final document that matters (assuming you use it for something.)
    As far as Six Sigma having a nerve? Ravi, you will have to keep up. About 8 posts back I told you I do not consider myself as representing Six Sigma. I speak for myself. The Six Sigma industry involves a lot more than just me.
    As far as agravating the nerve. Yes, ignorant crass remarks that minimalize the work of a lot of people at Motorola agravate me. If you had done about a nano second of research before you made such an inane remark you would understand that it was an internal program at Motorola. One of several initiative to accomplish the vision of “Total Customer Satisfaction.” As far as it going outside Motorola – nobody back in the late 80’s cared if anyone outside the Motorola did it. You made the remark – so come up with a fact that substantiates your crude remark.
    For some reason you have placed so much importance on this concept of common cause it must reside somewhere in our limbic brain. I figured anything that important to understanding variation must permeate every fiber of every tool. Amazingly it seems indigenous to SPC.
    I looked at Ishikawa Diagrams – nothing told me to classify my various inputs by common and assignable cause. It amazed me so I pulled out an FMEA and it didn’t have a catagory for it either. I figured it must be an oversight since it was so important. It isn’t mentioned in the manual either.
    How could such an imortant concept be missing. I decided it must be a part of the more “advanced” tool sets. I looked through Launsby, Montgomery, Barker, Duncan and Box, Hunter and Hunter. It was amazing how all these people who have built such amazing careers as statisticians and purveyors of tools for improvement have not, at the very least, warned us that we need to comprehend the factors we use in our DOE’s and Hypothesis Tests, in terms of their being Common Cause or assignable cause.
    How is it possible that something so important seems to be relegated to one tool. The other concepts such as central tendancy and measures of dispertion seem to have at least managed to be recognized by more than one tool. How is it possible that this thing that is so important was missed?
    Becuse it isn’t that important.
    This string was interesting for a rainy day in the Texas Hill Country. The clouds are breaking so it is time to move on.
    Good luck.

    0
    #77015

    Ravi Athalye
    Participant

    Hello Mike,
    Don Quixote did not die. Phantom adversories make for a creative outburst.
    I do not be little SS. I am a BB too. I view it a natural evolution of the process of improvement. 
    Common and Assignable classifications was an attempt at breaking up variation into manageable bits. Your pareto extended the concept. I could not have agreed more.
    Appears that we have been in violent agreement . However agreeing is out of fashion, not news worthy and ……….ignored!
    Sorry for offending your extra ordinarily sensitive SS artistic self. That was an un intended main effect !.
    I will thank you again for the education.
    Ravi

    0
Viewing 26 posts - 1 through 26 (of 26 total)

The forum ‘General’ is closed to new topics and replies.