How to do FMEA properly - A tutorial

[@donalddobert8652]

My goodness, Paul, thank you so much for your detailed reply. I am now 72 years old and have been in quality management for 44 years. I had a forced retirement in 2019 due to poor health. As President of a medical device manufacturing company our risk analysis sometimes had to be in compliance with IS0-14791:2012 and/ or MDD 93/42/EEC Medical Devices, which expressed RPN’s in terms of “harm to the patient”. I found that MDD 93/42/EEC was much stricter than automotive FMEA or ISO-14791:2012. Risks had to be mitigated, regardless of cost, so our severity-occurrence-detection was tighter than automotive. The following was required (to resolve ISO-14791:2012 and MDD Gaps): All risks, regardless of their dimension, need to be reduced as much as possible and need to be balanced, together with all other risks, against the benefit of the device (MDD Annex I, Sections 1 and 2). All risks, regardless of any “acceptability” assessment, must be reduced as far as possible and must be balanced, together with all other risks, against the benefit of the device (MDD Annex I, Sections 1 and 2). Risks are required to be reduced “as far as possible” without room for economic considerations (Annex I, Section 2). An overall risk-benefit analysis must take place, regardless of the application of criteria established in the management plan of the manufacturer (MDD Annex I, Section 1). Undesirable side effects must “constitute an acceptable risk when weighed against the performance intended.” (MDD Annex I, Section 6). These gaps shall be addressed unless they are not applicable (because they are needed to be addressed by the design function). Getting away from FMEA’s, I introduced your videos to my son (age 33), who is an engineer, and he became fascinated with your work. He bought your book, “Drink Tea and Read the Paper”. Currently, I am teaching him about six-sigma, MSA, and DOE. He recently purchased a machine vision system for a medical device project. As for the MSA gage R&R validation for the machine vision system, I told him if the high speed vision system is used for “pass/ fail”, without documented measurements, then the quality group will have to do a gage R&R study per accepted norms (MSA 4th Edition). This would be part of an accepted P.Q. The company must establish that gage error is less than 10%. Actually, I have found that external auditors like to see some work-around solutions when problems arise. Since the vision system is automated, and the company would perform an MSA via other means, the company should establish a correlation between the vision systems pass/ fail attributes versus variables measurements and MSA (which will allow establishing a valid Cpk and Ppk). I told him attributes cannot establish Cpk or Ppk. If you know of a better way to establish an MSA for a machine vision system, I would appreciate your opinion. About me: I was President and COO of a medical device manufacturing company (1999 - 2019). I did a lot of things you feature in your videos. I feel your work is excellent. Thank you again for your time.

[@runalong]

What a great and concrete example on how Rolls Royce *doesn't* allow "Operator Error" as a potential cause in the FMEAs!! I appreciate how you fleshed out the failure mode (inputs) of how Operator Error is a weak and inadequate cause.

[@paulallen5321]

Glad you liked it John...

[@David_Hoskins]

The descriptions in the Detection Table are PERFECT, and make 100% complete sense!! Thank you Paul for this excellent insight.

[@paulallen5321]

Hi David - thanks for the comments, funny though everytime I look it I think I could change it to make it better, i'll let you know if I come up with a better list...

[@cuongnguyentan246]

BANG! I have finally found my favorite Lean Six Sigma chanel.

[@paulallen5321]

Curong - good to hear from you, if you think there is a subject missing, drop me a line and ill make that video for you...

[@ninonino8332]

Thank you Paul for all the videos especially FMEA.

[@Mr190093]

Holy shit! A guy talking about FMEA and Six Sigma that doesn't speak in businessese. Thanks for the straightforward explanation! I'm going to check out your book now.

[@paulallen5321]

Ciaran - many thanks for the positive comments, hope you enjoy the book, please spread the word...

[@maedehdibadin9813]

Best video about the FMEA ever, thanks Paul

[@paulallen5321]

Cheers....

[@evandroa.m.martins1977]

Let's see Paul Allen's Control Plan...

[@jackiemedina5511]

Woah! excellent explanation. These are important for my FMEA for Occupational Safety. Thanks

[@paulallen5321]

Glad it was helpful!

[@arjunjagadeesh4290]

Thank you Paul for such a wonderful and crisp explanation of FMEA. Learned a lot in a short span!

[@paulallen5321]

Arjun - no problem...

[@tpthomas28]

Very confident presentation. Clearly understood the concept. Thanks Paul.

[@paulallen5321]

Paul - Thanks for the positive comments. if my video works for you and you want to more about all the tools, look up my book 'drink tea and read the paper' on LULU.COM

[@tpthomas28]

Thanks Paul.

[@Bo23481]

The best so far explained. I like input and output approach 👌🏻

[@paulallen5321]

Cheers Boris - Glad it was helpful!

[@andreivyrypaev5636]

Brilliant approach, brilliant rating grades and brilliant explanation.

[@PaulAllenAPL]

Andrei - thanks for that, my channel has been hacked and at the moment I can't load anymore video's. But if i can help any further drop me a line on [e-mail] or buy my book 'drink tea and read the paper' from lulu.com

[@thituyetnhuhuynh9084]

@@PaulAllenAPL HI, could you give me a template of FMEA? thanks a lot

[@adriancabrera3519]

Thank you Paul for breaking this down to my level, several out there that could not keep me focused, nothing like your breakdown.

[@paulallen5321]

Adrian - Thanks for the positive comments, if you like the video's you might like my book 'drink tea and read the paper' you can get it direct from LULU.COM...Its written in the same style as the video's...

[@san1940rst]

Very Nice I was looking simple video which you did very Nicely thanks Sir

[@VINOD600CC]

Hello Paul. You got right explained with examples. That really helpful.

[@paulallen5321]

No Problem Vinod, thanks for the positive comments. Take a look at my book 'drink tea and read the paper' from LULU>COM. All my video ideas are in there...

[@vishnauty]

Many thanks Paul! , very detailed explanation about FMEA , expecting one more great explanation on FTA

[@paulallen5321]

Cheers Vishwanath - Although I have to say FTA is not one of my tools currently in my toolbox...maybe I'll take a look.

[@inovitcmm7899]

Hi Paul, Thank you very much your explanations are exactly to what we come across in our daily work and and very relatable.

[@paulallen5321]

You are very welcome Sunny...

[@marutimishra3223]

Thanks Paul....Wonderful tutorial on FMEA....Never found before....Love you man

[@paulallen5321]

Maruti - Thanks for the positive comments. if my video works for you and you want to more about all the tools, look up book 'drink tea and read the paper' on LULU.COM

[@atekka1]

Perfect video! Thx a lot ! Extrem perfect explanation. Absolute crystal clear. Have video such of this more ? A

[@paulallen5321]

Oden - thanks for your positive comments...

[@geraintjones7192]

Thanks Paul for the video. One thing I think is very important to point out here is the limitation of the Risk Priority Number (RPN) as used here and used as the product of the multiplication of SEV x OCC x DET. The RPN is being used here to rank the risk, the implication is that the higher the ranking the more important the risk is to concentrate on and potentially mitigate in comparison to the other risks. If you have 10 ratings on each scale you have a range of 1000 RPN values. However, it turns out that there are only 120 possible values and that these values are not uniformly spread out between 1 and 1000, this means that the RPN values sort the 1000 problem descriptions into 120 artificial grouping of different sizes. So, the problem lies with the notion that RPN's can be used to rank between the problems. The reality is that RPN values are what's called Ordinal-Scale Data. Before operations of multiplication and division are meaningful the data must be ratio-scale data, these data posses ordering, distance, and an absolute zero point. The lack of a distance function and zero point can result in serious and trivial problems having the same RPN and some trivial problems ending up with larger RPN values than other, more serious, problems. The way forward? - Use 1 through to 5 for the rankings of each aspect -Create a 3 digit code for each failure mode - 1st is Sev, 2nd is Occ, 3rd is Det -Signify the ordering as the digits on the SOD code The SOD code based upon thee rankings of 1 to 5 will result in 125 values for 125 situations. When the SOD codes are placed in descending numerical order they will prioritize the situation firstly by Severity, secondly by Occurrence and thirdly by Detection within each combination of severity and occurrence. You can then choose the problems that need to be addressed. I hope this is helpful. Thanks Geraint

[@paulallen5321]

It's a nice alternative, but not what the rpn is for. It's much simpler than that. Every risk must fixed over a certain value. The only the thing the rpn gives you is more time on some issues than others. it doesn't give you a cop out...

[@geraintjones7192]

Until I saw your video I was always led to act on the RPN values as they are meant to be rankings. I don’t see how it would be possible to set ranges that you would be happy to act upon because you would surely have to take into consideration not only what you would feel comfortable with but crucially the limitations of the numbers themselves. I don’t see that as a simple task, there is no escaping the fact that the ordinals are being treated as ratio data ,therefore the scores you are providing a limit against are not valid.

[@paulallen5321]

Hi Geraint - if this was measurement system I would agree, but it isn't. (Used as a H&S risk assessment, I might agree as well). It's a tool to help you to design a process, make a decision and then take improvement action. If the RPN is 500 or 800 the value is immaterial the decision in both cases is the same reduce it. And in both cases the area of attack is the same. Improve the control. If you drive the control to mistake proofing both the rpn values will fall to less than 100.....FMEA is a practical tool to help the engineer to communicate their thoughts and to create a great control plan and please the customer that is what it's for

[@geraintjones7192]

@@paulallen5321 Ok, I take it from your reply that absolute value of the score doesn't matter, it's the reduction to some value that demonstrates the due diligence and hopefully control. So, thinking as an engineer having established and assembled the risks really everything is game, so I could spend a lot of time trying to control many different risks that perhaps in some cases don't impact the process. I can see that trying to get everything under lets say 100 is honourable. Which ever way it's looked at we are acting on scores, to reduce if we need to. The "if we need to" aspect really is just a judgement based the engineer or the teams opinion. This is the key aspect for me, you effectively don't use the scores you advocate working on all or the ones you feel are impacted the most. In which case I don't understand why the scores are there except to demonstrate that you started at X and finished at Y. Why not just use Red amber and green.

[@paulallen5321]

Hi Geraint - You're missing a step, the Design FMEA includes the voice of the customer which tells us which parameters, features and performace are important to the customer. That way only high risk issues get included in this part of the process (high risk means important to the customer) and the DFMEA output tells the PFMEA which tolerances etc are important to the customer. The PFMEA is then about controlling those important process inputs that deliver important results. Nothing on a PFMEA should be unimportant to the customer otherwise you've done the FMEA process wrong or disconnected the Design from the Process FMEA...

[@l_t_truth_seeker]

Thank you sir, Great video ... I finally bought your book

[@paulallen5321]

Truth Seeker - thanks for the support hope you enjoy it...

[@shubhamkaviskar]

thank you. it was so simple to understand.

[@paulallen5321]

Glad it helped!

[@ndzalamangobeni6631]

Thanks Paul

[@satishtalikota]

Excellent delivery !!

[@paulallen5321]

Satish - thanks for the positive comment, if you like my video content take a look at my book 'Drink tea and read the paper' you can get it from LULU.COM...

[@gejambazo]

An awesome learning material...thank you.

[@paulallen5321]

Thankyou - please share the links with your colleagues...

[@harrie205]

great video

[@paulallen5321]

cheers Harrie

[@cucheshire]

Hi Paul, thanks for the valuable information on FMEA. FYI, the Lulu site doesn't have very much information about the book. For example, it's missing the subtitle, it could use the back cover text, and the description is only one sentences. But it does look like a very good deal if it contains the same information that you presented in this youtube video!

[@paulallen5321]

Aaron - thanks for the info, i'll take a look see what I can improve...

[@ScottXe]

Thanks for a good tutorial!

[@paulallen5321]

Cheers Scott

[@louise4976]

Thank you Paul! You make this sound so easy, I hope you will be doing a video of the new AIAG/VDA FMEA soon! I have a couple of questions about making my own severity/occurrence/detection tables: 1. How much am I allowed to change? The standard doesn't say anything about this, so I am unsure what I can do and still pass an inspection. I am trying to maintain the same "levels", but I am making it completely my own. 2. In the AIAG/VDA version of the severity table they talk about "Loss of primary function" for the end customer. So if I am delivering parts that do not affect the car primary function (drive), then I can never get an 8 in severity (for customer)? Or should I consider loss of primary function for the specific part that I deliver?

[@paulallen5321]

Louise - good to hear from you and good questions. I'm not an expert on how AIAG/VDA might audit your design processes, but here;s my simple suggestion. Just line up an alternative scale next to the AIAG version so that your engineers can read across from your company scale to the AIAG scale. That way you can claim 'your' scale is additional information. If you end up with only 6 categories on your scale compared to AIAG scale, just miss numbers out. So your 6 (worst case) lines up with AIAG 10. Your 5, lines with AIAG 8 and there is no use of 9. Hope this makes sense. To be honest you're ranking risk on an FMEA, it doesn't really matter what the score system is, It matters that you use it consistently and what the response has to be be. When you get high risk results how do you behave or respond... I would say loss of primary function relates to the part you are making rather than narrowly looking at the drivability of the car.

[@paulallen5321]

Louise - Personally I would say you can change it all. It's your business, it's up to you to design the3 best systems for you, not try to implement a standard system. AIAG is for car making supply chain, if that is not your business then it's not the best standard to use. Loss of primary function would be linked to the part/system. So if it's light switch primary function is the lights not driving the car. I'll try to make a video about suggesting ways to keep the VDA process with your own instructions linked to it....

[@khalidnawaz6536]

Excellent Demonstration...

[@paulallen5321]

Khalid - thanks for the positive comments

[@khalidnawaz6536]

@@paulallen5321 You welcome paul

[@ndzalamangobeni6631]

Id be happy to see how fmea becomes input to control plan. Many people have different takes on it, i need your help there

[@paulallen5321]

Ndzalama - I think if you look up more of my FMEA video's you should see what you want. If you can't find it let me know....

[@be2ful1]

the video was so helpful. thank you, great explanation.

[@paulallen5321]

Thanks for the positive comments. if you like my video's you can get my book 'drink tea and read the paper' from LULU.COM. You'll find its written in the same common sense style on all things to do with quality...

[@kolbytober3102]

Hey Paul, do you have any tips and tricks for doing decision matrices well and making them useful?! I would love a video like this for doing a weighted decision matrix... btw, this is an awesome video!

[@paulallen5321]

Kolby - thanks for the positive comments. If you like to video try my book 'Drink tea and read the paper' from LULU.COM. There is a technique known as decision analysis it comes from Kepner Trago. I'll try to explain it and post a video for you...

[@stuartcooper1416]

Hi Paul, can the new AIAG and VDA FMEA be applied to Process FMEA?

[@paulallen5321]

Yes my understanding is that the new std is much better to use and closer to the process that I recommend...

[@andrejm9729]

Hello. Which section (Occurrence or Detection) will be re-evaluaded in case of implemented DOE for DFMEA and PFMEA? Thank you

[@paulallen5321]

Andrej - you'd have to give a fuller example of FMEA/DOE for me to comment further...

[@vivicristina5304]

amazing class..

[@paulallen5321]

Viviane - glad you found it helpful. If you ever need a question clarifying, drop me a line and I'll make a new video. I'm always looking for reasons to post something new...

[@a.h.m.shakilaktar6781]

Thank you so much

[@paulallen5321]

my pleasure...

[@Ri-Action]

I searched for Paul Allen, I wanted to see how Patric Bateman killed him, I did not expect this 😶

[@inovitcmm7899]

I have a Question, What is the method or process to reduce severity of an occurrence while performing PFMEA

[@MrMtkrulz]

Try understand the difference: - Severity Rating is how bad the Effects of the failure are if a process/product does not fulfill its intended funct. You cannot reduce/increase the severity. - Occurrance rating is how often will this failure occur. U can reduce occurance by optimizing the process or make changes to design. Hope that helps.

[@paulallen5321]

Sunny - yes i agree with MTK, if the failure is 'a fire breaks out' how serious is that? It doesn't matter how much work I do, I can't really change the fact that a fire will have a very high severity. I can reduce the likleyhood though or put controls in place like all material in the building is fire retardant to control the situation....

[@gatehieu1985]

Hi Paul. How can i buyer your E-book. I am in Vietnam

[@paulallen5321]

Cheers gatehieu..

[@LaShaundaJohnson]

Can you send the root cause analysis diagram?

[@paulallen5321]

Hi LaShaunda - Drop me a line at [e-mail] and I'll send it over...

[@AliAli-br6dx]

Good work .. continuous

[@paulallen5321]

Ali - thanks for the positive comment. If you ever want a subject covering drop me a line. Otherwise look up my book 'drink tea and read the paper' on LULU.COM

[@hectoro.2363]

You are mixing up thing. You are jumping into conclussions and actions in the diagnostic part of FMEA (severity, occurence and detection).

[@paulallen5321]

Hector good to hear from you.... I am making an assumption that a moulding company knows the physics of injection moulding. I think that is a good assumption to make....and I'm not looking at detection I'm looking at prevention, the proper way to control a process. I want to FMEA to produce a control plan not an inspection plan...

[@agentorange26]

Thank you for this video. I'm new to this tool and I've seen a lot of different opinions on how to do it. I got a question for you: should a FMEA contain all the characteristics of the product rather than only the critical ones? Some people in the company where I work think that makes a very robust FMEA, but I don't agree with them. You end up with a huge document and nobody will want to take a look at that as the tool it is meant to be. I hope you can help me clear this out. Cheers!

[@paulallen5321]

Wildp - thanks for the question, It is important to separate the DesignFMEA from the ProcessFMEA. In the design FMEA you should have all the performance charatersitics that the customer wants and the designer should deliberately address each one in the design. Then the process FMEA will use this information to quantify risk of a process control missing and how important it is. What maybe missing at the start of your process is a prioritised list of CTQ's that has been created from a Voice of the customer exercise. With a prioritised list it's easy to over design or control the design when it's not needed.

[@agentorange26]

@@paulallen5321 thanks a lot! This is a really helpful answer.

[@antariksafata]

You always make become very simple

[@vic1159]

Fantastic explanations for each of the steps, nice job and thank you...

[@paulallen5321]

Vic - If you like the video's maybe my book 'drink tea and read the paper' might be of interest. You can get it from LULU.COM

[@paulallen5321]

Fata - If you like the video's maybe my book 'drink tea and read the paper' might be of interest. You can get it from LULU.COM

[@java222100]

what is the difference between control planned and action plans

[@paulallen5321]

Javed - good to hear from you. I would say action plan is a one off plan to implement controls, control plans are something that happens regularly. For example: lets say we decide we want to control air pressure, but there is no gauge on the machine. The action plan would say install gauge and agree a standard for air pressure. The control plan would say check air pressure every 2 hours and make sure it is at 1.3 bar...

[@java222100]

In this video... FMEA is done for the new process... Hence we are using planned controls hence forth there is no action plan in this case.... Planned controls equal to action planned. Hope I am right?

[@paulallen5321]

Javed - It could still be even with a new process that we have an action to implement the new process. Then regular controls to keep the defects away...

[@java222100]

So you have not mentioned any action plan for this example..?

[@paulallen5321]

Javed - you are correct there should be an implementation plan and a control plan that maintains it...

[@abdulrhmanmohammed881]

thanks,everything is great

[@paulallen5321]

Abdul - thanks for the positive comment. If you ever want a subject covering drop me a line. Otherwise look up my book 'drink tea and read the paper' on LULU.COM

[@janzaibmbaloch5484]

You sound like Tommy Shelby. Great Tutorial

[@paulallen5321]

Cheers Janzaib - My accent is so much more 'black country' than Tommy Shelby...If you like the tutorial look up my book 'drink tea and read the paper' you can get it from LULU.COM...

[@donalddobert8652]

On January 28, 1986, the Space Shuttle Challenger exploded 73 seconds after liftoff, killing seven crew members and traumatizing a nation. The cause of the disaster was traced to an O-ring, a circular gasket that sealed the right rocket booster. This had failed due to the low temperature (31°F / -0.5°C) at launch time - a risk that several engineers noted, but that NASA management dismissed. NASA’s own pre-launch estimates were that there was a 1 in 100,000 chance of shuttle failure for any given launch - and poor statistical reasoning was a key reason the launch went through. In 1989, a research team wrote a paper (“Risk Analysis of the Space Shuttle: Pre-Challenger Prediction of Failure”) analyzing the data available before launch to determine if the failure could have been better predicted before launch. Using standard statistical techniques on previous launch data, they determined that the evidence was overwhelming that launching at 31°F would lead to substantial risk of failure. They measured a ~13% likelihood of O-ring failure at 31°F, compared to NASA’s general shuttle failure estimate of 0.001%, and a 1983 US Air Force study of failure probability at 3-6%. Think about if you would let the shuttle launch if you knew there was a 1 in 8 chance it would fail? 1 in 100,000 chance? I have a tough question: “In the risk analysis for a multiple cavity sequence of production, would you recommend changing the severity, likelihood and detection considerations if there were 8, 12, or 16 O-rings? Would once in a week or once in a month be appropriate for likelihood? Is there a statistical formula that you would recommend to determine your severity-likelihood-detection value rankings? Would you recommend a mandatory Poke-yoke setting for a critical FMEA (such as the one for O-rings) for all of the detection cavities?” By the way, you always give a great presentation and I admire your work. Thank you!

[@paulallen5321]

Donald - good to hear from you....sorry for the slow reply. It's a hell of a question you're asking. so I needed a bit of spare time to answer. I use this data in my classes and work out the probability of failure to be 90% on the day of the launch. However, anyone can be a monday morning quarter back! I think there is a number of points to make. Firstly don't follow car industry guidelines write your own based on oyur own industry. 3rd - Probability can easily make a fool of anyone and in this case it certainly did. FMEA is never going to be perfect. But the real failure in this case was that they broke their own guidelines. Once the shuttle started flying, the could see the fails so at this point there is real world data, They knew the problem existed, they knew the system failed to danger which is against NASA rules and still they launched. They could easily have put a launch rule in to say there was a minimum launch temp. After all there was rule to say you can't launch in cloud. Which is why they never took off on 22nd Jan, the orginal launch date. If they genuinely believed it to be 1/100,000 event. It would have been a rule they never used and wouldn't have been an issue. FMEA is a best guess, you can tie yourself in knots trying to make that guess perfect, but it's always a guess, the more complex you make it, the more likely you are to bend the rules and come up the 'wrong' guess. I do think though small production quantities are the most difficult to assess, making luxury yachts for example, the likelyhood of 1/10000 would raise concerns about everything, but would be once in 10 or 20 years...perhaps. so make your own FMEA rules that relate to low volumes etc that's as complicated as i would make it...hope this helps Donald....

[@CARLOSGARCIA-qy2mm]

tHE AUDIO IS VERY LOW

[@paulallen5321]

I'll see what I can do...

[@salvatoreshiggerino6810]

Let's see this guy's business card.

[@paulallen5321]

Just watched American pycho, so just got your message....

[@erickcontreraslopez5009]

people do not understand between ISO9001, Prev vs Detection, Six Sigma and Why implement Core tools, it is hard, because that means they are not standarize, that is why I always refer to reg. standards, and also to lessons learn structure, lessons learned is the basis for FMEA and all then engineering. fix that, you fix FMEA by default!

[@RavichandranNarasimalu]

Excellent video.really appreciable. Finally you connect this with money.bigbang.when a process is a foolproof method that it itself shows a happiness in system .of course it is all ends up money. Your profound words are nice. I really enjoyed it.