iSixSigma

GE Turbine Engines – 12 Sigma?

Six Sigma – iSixSigma Forums Old Forums General GE Turbine Engines – 12 Sigma?

Viewing 38 posts - 1 through 38 (of 38 total)
  • Author
    Posts
  • #30652

    john beaudoin
    Participant

    There are many including Stan that do not think it is possible for GE to have a 12 Sigma level in their aircraft engine production.  Like any good belted individual, he is questioning the data because there has been a lot of talk about such things, but no one has shared any evidence that this kind of stuff exists.  If anyone out there knows of some processes where they have achieved better than 6-Sigma, please share that information so that the non-believers will see the light.

    0
    #80116

    Withheld
    Member

    John,
    Provided we can agree on some basic ground rules, I’ll play. I propose we define a defect as in-flight failure of a GE-manufactured engine. Opportunity is one per engine. Agreed?
     

    0
    #80117

    john beaudoin
    Participant

    Many in-Flight defects are prevented through strict maintenence/overhaul procedures.  Since these are not always performed by GE, should we be talking about defects found immediately after production of the engine, or by defects found in-flight?  Is GE preventing errors in engine manufacture so there is little or no rework within the plant, or is quality inspected into the product at 12 sigma, which shouldn’t count?

    0
    #80121

    Withheld
    Member

    As long as the opportunity is one per engine it matters little to me if we waive the in-flight definition from my other post in favor of zero defects after final inspection sign-off (or delivery to the customer – whichever you prefer). Either way all I need to do to win the game is find one documented instance of a problem with an engine. That assumes, of course that GE has not produced 300 gazillion engines (the sigma calculator at the top of this forum maxed out at 7.97 – one defect per 20,000,000,000 opportunities).
    Don’t get me wrong – I think GE is a wold class engine manufacturer. It’s the outlandish 12S claim and the associated impossibility of defending such a claim that holds my interest.

    0
    #80123

    Gabriel
    Participant

    I actually have a process that has a standard variation of 0.02 and the specification is +/-0.3, and it is centerd. Would that be a 15 sigma process (provided that there are no outliers)? How to prove it? Of course, we didn’t make gazillions of parts (whatever a gazillion is), but in the few millions we made we never had (or realized we had) a single defect or complaint (becuse of this process, I mean). So, how many sigmas is 0 PPM?

    0
    #80124

    billybob
    Participant

    Hello folks,You guys are so funny.Later, Billybob

    0
    #80125

    john beaudoin
    Participant

    Why would you want to put an artificial limit of one per engine? Sigma is always calculated from total opportunities.  At 12 Sigma, You would have 1 error in 1E+20 opportunities.  I do not know how GE is defining their opportunities, but based on the numbers above it would have to be based on each part in the engine as well as some unit of time that the engine is running.
    If you look at overall opportunity of and engine failing in flight for just one aircraft with 2 engines making 10 2-hour flights a day for a year, you would have 10 x 365 x 2 opportunities for just one aircraft or 7,300 opportunities.  If you consider the number of things in each engine that could go wrong, you would have possibly 7,300,000 opportunities for that aircraft with the 1000 parts in an engine.  This changes even more if the opportunities are measured in hours of operation vs flights, etc.
    The question is now asking to hear from someone at GE that knows how they get to 12 sigma.
     

    0
    #80126

    RT
    Member

    Whether you are measuring DPMO’s or Standard Dev.  It all equates to some probability of a defect.  Sigma level is a convienient way to catagorize quality level.  If I show Z values at a staff review of 2.5 starting and 4.0 improvement people know that is a successful project with quality output.  If I put up probabilities .00621 – .0000317 they would not understand the significance of the change.  A process  Z value can go to infinity but going much beyond 6 doesn’t really effect a part or process.  In fact too many times people test their final product and come up with Z values of 15 or 20  and walk away.  What you should be doing is looking at failure probability over time.  Now that gets into Weibull analysis  and DFR.  A different subject but a good way to predict failures over long periods of time.
    When GE says they are a Six Sigma process or company they are stating that they use the SS methodology to push their process to be the best in class.
     

    0
    #80129

    Withheld
    Member

    You define true zero defects as infinity sigma, I guess. As an aside, I wonder if it may not be in your company’s economic interests to take advantage of the robustness?

    0
    #80130

    RT
    Member

    I’m not sure I understand your response, if infact, it was to my post.
    I didn’t define defects at all.  I talked about probabilities of a defect and what Z values mean.  There is always some probability of a defect and thats why “infinity” comes up.

    0
    #80131

    Withheld
    Member

    What makes one finished unit “artificial” to the customer? If my plane falls from the sky due to engine failure my loved ones would care very little that the fatal defect was only one out of a billion opportunities.
    Defining every possible defect for every component may be useful in problem solving procedures, but it means nothing to the end user.

    0
    #80134

    Withheld
    Member

    I responded to Gabriel’s post, Rt. I don’t disagree with your comments.

    0
    #80136

    john beaudoin
    Participant

    I agree with you 100% from the customer perspective, but GE is the manufacturer in this case.  If GE’s manufacturing process was only 5 sigma and not 12 sigma, they would have 233 defects per million opportunities.  If they only had 1,000 opportunities per engine, they would have .233 defects in each engine produced.  As a result, almost 1 in 5 engines would have some kind of malfunction, thus giving the end-user very bad quality.  From the design and production standpoint, you may have to have a 12 sigma process to have the final unit perform at a 6 sigma level to the customer.
    If GE is not using the opportunities to get to 12 sigma, then they have had no engine failures due to production.  If you have a defect, you need the opportunities to get to 12 sigma, but if you don’t have any defects, the opportunities do not have to be nearly so high.  Have you had any luck finding a crash do specifically to a GE engine that was not caused by poor maintenance and was blamed on the manufacturer of the engine?

    0
    #80137

    Withheld
    Member

    John,
    I didn’t look to document engine failures. In fact, I viewed it as rhetorical since I’m positive that any search will prove that nobody’s perfect. Moreover, it’s hard enough trying to keep up my end of the conversation while I’m working! It probaly won’t be hard. If you get a second, do a google search for “engine failure” “ge” “faa” or whatever parameters you want.

    0
    #80140

    Withheld
    Member

    “I agree with you 100% from the customer perspective…”
    I wonder – when it comes to quality is there any other perspective that matters?

    0
    #80141

    RT
    Member

    $$$$$$

    0
    #80145

    Raju
    Participant

    I was thinking about this problem, generally, conventional application of statistical methods may not be appropriate when the process is operating at near zero defects level and there has been some work in this area and I found a book which has been published recently on this issue (attached the link). I haven’t gone through this yet hence I can’t comment on the application of it to this situation but I guess this may be a source to get some insight. I guess you can write to the author to get some articles on it for more information. 
     http://www.ise.nus.edu.sg/staff/xiemin/book/Announce-HQ-SPC.pdf
    Raju

    0
    #80153

    Gabriel
    Participant

    I have a great idea to increase the sigma level, from a design perspective:
    Add compexity! Add more components, but be sure that the filure rate of each added component is less than what was the original DPMO of the whole product. You will get more PPM, but fewer DPMO, and everybody happy! (except the customer, I guess).

    0
    #80162

    billybob
    Participant

    Hello folks,I used to work at a GE plant where jet engines were bulit ( I was in Steam Turbine) let me state a few known facts.  The plant I worked in was unionized and parts of it a piecework shop. Imagine paying piecework wages to make ject engine component as fast as you can! I know first hand of zyglo ( crack dected) operations that ocurred over and over again as cracks were removed. Was each rework a defect, or each part only a defective? The pieceworkes love it as they were paid well to find and repair, and repair, and repair defects. And this went for many other operations also. So put this in your 12Sigma formular and enjoy your flight. And to make everyone rest comfortably my department, Steam Turbine, made propulsion units ( turbine and gear sets) for every nuclear submarine in the fleet, we did plently of rework of turbine parts too.  We were piece work also and plenty of money was spent fixing mistakes.  But thats why we had test stands and test cells. Everything was well tested before it left the shop it was defect free reaching the customer, despite what had to be done to get it to that point!  You can ask but I am sure GE will never tell you its “real” internal defect rate……if they really know it unless their systems have changed greatly in the last 12 yrs.Later, Billybob

    0
    #80163

    Withheld
    Member

    Looks like you want a sip of that beer… ;-)

    0
    #80166

    Ronald
    Participant

    What probably should be consider here is two points, manufacturing defects and engine performance failures.  The manufacturing defects would be based on Rolled Throughput Yield and be a function of the complexity of the system/# of parts (opportunities).  This takes into account hidden factories, rework, etc and gives a truer view of the pass rate for the system (will probably never be 12 Sigma).  To be correct, the opportunities counted should only be those that relate to the CCR or CTQ identified.  From an engine failure perspective, the classification of engine failures could be the opportunities counted. 
    A process could have many different Sigma Levels depending on how the opportunities are defined.  As a customer, I wouldn’t care if a company was manufacturing was at 3 Sigma if they 100% inspected multiple times to ensure the engine performance failures in reality were at 6-7 Sigma or higher.  Of course the inspections would add cost and if I was buying engines I would not buy from a 3 Sigma manufacturing company.  I would buy from the company that is most capable of providing my targets both performance and cost.

    0
    #80167

    Withheld
    Member

    “A process could have many different Sigma Levels depending on how the opportunities are defined.”
    That’s why we need some structured honesty injected into the quality industry. Perfection is a laudable goal, but quite difficult – dare I say, impossible to obtain.
    Show me a company or individual who claims perfection and I’ll show you a liar. Show me a company or individual who stands by their claim to perfection when challenged and I’ll show you a fool.

    0
    #80171

    Swaggerty
    Participant

    How can anyone make a 12 sigma claim.  What about all the re-work done at the plant…as Billybob has pointed out.  What about all the maintenance done after every flight and at sceduled intervals.  The engines are still in service so how do we know that a defect will not happen at a later point.  6 sigma is a real name used by companies who use the methodology to improve processes.  12 sigma is crap!  I’ve never achieved any better than 4.5 sigma on projects i’ve worked, but still generated big savings for my company and satisfied customers along the way.  This is the real world of six sigma.  These claims are not real world…do GE employ humans?  I refuse to believe any of it.  Of course the Ge BB’s love all this, makes their market value soar and everyone here is promoting them.   Give it all a rest until one of them gets some real data on the table that we can discuss in a professional manner.Rant Over

    0
    #80175

    Ashman
    Member

    what’s not possible about it?  12 sigma just implies a defect rate measured in parts per billion instead of parts per million.  Do I believe that they are really operating at a 12 sigma level?  No, but that’s just advertising.  You can’t tell me there is a company in the world that doesn’t embellish their accomplishments for investors.  It’s just something to put the name in the paper and get people to talk about it.

    0
    #80177

    Withheld
    Member

    Steve,
    It appears there is at least a majority if not 100% agreement that 12S BS.
    Following your lead maybe it is appropriate to switch focus to the marketing potential or pitfall associated with such an outlandish claim?
    Are we (the masses) so gullible that we would not challenge that which cannot be true? Doesn’t the very fact that such a claim is made call credibility into question on other aspects of the company’s reporting?
    It does for me. For that reason and since I share your view that such claims would be marketing tripe, I would fire those responsible on the spot. In these days of corporate book cooking, calling your own credibility into question is a marketing nightmare.

    0
    #80178

    John J. Flaig
    Participant

    While it is possible to have a 12 sigma process it is just virtually impossible to prove it. If the number 10^20 is correct for the number of oppertunities, then I’d like to see GE’s supporting data.
    Dr. Deming always joked about the “zero defects” programs for the same reason. Having zero defects in the sample does NOT prove the process has zero defects. The sample size required to prove this is infinite. Of course GE may have been in touch with some higher metaphysical power that allows them to make such claims but I suspect the marketing guys got their inspiration at the local pub.
    John J. Flaig, Ph.D.
    Applied Technology (e-AT-USA.com)

    0
    #80183

    Jim Athon
    Participant

    Having not seen the original post, I don’t know what data was used,  whether this is for a single CTQ or for the entire product, etc.  Nor do I know what was considered and “Opportunity”, as this could make your sigma be whatever you want, ie MTBF per run hour will be 3600 times higher than MTBF per run second (I’ve seen that game played) Nevertheless, here goes…
    If you go with the premise that the only way to measure Sigma is through attribute measures, then several of the arguments posted are quite valid, and if 12 sigma existed, you could not “prove” it statistically, incredible sample size requirement, etc.
    However, if one is using continuous (aka variable, quantitative) data, all you have to do is be able to fit 12 std dev’s between the mean and the closest spec limit, hence 12 sigma!  I have literally seen 15 sigma processes, e.g. Cpk of >5, in some automotive applications.  Of course, that was for only a few of the CTQ’s, and others were significantly lower, so total product Sigma would have been dramatically lower, esp. if you had to resort to DPMO calculations to gain a composite capability.
    Consequently, it is possible to have a 12 sigma process or product.
    Never say Never, it all “depends” on how you answer the question. :-)
     

    0
    #80185

    RT
    Member

    I’ve been reading a lot of GE bashing going on here but has anyone actually looked for the claim that GE has an engine at 12 sigma?  I would be interested in seeing that article so I know if it’s based on anything specific.  Let me know if someone has a link to such an article.

    0
    #80187

    john beaudoin
    Participant

    Thanks Jim!  Your input makes a lot of sense and is a refreshing response.  Basic statistics says that the sample sizes for continuous data are a lot less than pass/fail data.  In manufacturing, continuous data of measurements within engineering specifications is very common.

    0
    #80194

    Mike Carnell
    Participant

    John,
    I haven’t been on in a while and you seem to have wound a lot of people up. I did the support a GEAE in 1996. We were not manufacturing to 6 sigma (let alone 12) then and I don’t believe they claimed to. The quality level they did deliver to was the result of a lot of rigorous test and inspection. Performance at the customer is a different story.
    Good luck.

    0
    #80195

    Withheld
    Member

    John,
    I can certainly understand the need to look for a port in the storm, but I think if you reread Jim’s post you’ll see that it may not be the port you think it is. ;-)
    Anyway – how about if we call our game a draw and agree to disagree? I just jumped in for a diversion and enjoyed the conversation.
    By the way, I find the great majority of your postings interesting, informative and always well-intended.

    0
    #80197

    billybob
    Participant

    Hello folks,
    You two guys were fun to watch bash it out.  You know how once in a while you drive down the street and see 2 one-tooth women fighting over the last beer and you got to pull over and watch’em go at it.  This 12 sigma stuff  between John and Witheld was to be just as interesting. 
    Thanks,
    Billybob

    0
    #80203

    Mike Carnell
    Participant

    Billybob,
    12 sigma is actually something that can be achieved relatively easy. You move the spec limits.
    A more common practise out there is to play with the opportunity counting. I am not sure if you have ever seen a in-circuit (bed of nails) type test. You can test hundreds to thousands of componets in the blink of an eye and with the high sigma levels of the passive componet manufacturers the test are relatively defect free (provided you don’t have knuckle-dragging Neaderathals handling your boards). Now for those lost souls who bought into the active/passive opportunity counting the active count went up and the probability of a defect went down. The sigma level goes up and removes any motivation that was ever present to remove a nonvalue add process. Instant Six Sigma.
    There were masters of the opportunity count quality improvement program at Motorola (not the majority) and some are consultants today.
    Good luck.

    0
    #80205

    Efrain
    Participant

    Hello, 12s statistically is possible, but in the real life is almost impossible. I work in a turbine repair shop, and the components of the turbines (blades, vanes) have a lot of rework.
    I know which is the work that have the turbine components.
    Just that GE use robots to make the engines, may can be 12s in it’s process.
    Say 12s is easy but in the floor is other thing.
    All processes have a lot of variables and a lot of data; if we can control of them in all.
     
     

    0
    #80206

    john beaudoin
    Participant

    Not so sure we disagree here….
    1) 1 error in 1E20 is ludicrous.
    2) There may be processes that stay within spec limits 100% of the time, however small. (I always get my car into the garage without hitting the sides of the garage, although others might not, or it is just a matter of time before they don’t)
    3) Opportunities can be over inflated to make up a Sigma Level and may not reflect what the customer experiences.
    4) Inspecting in quality shouldn’t count toward Sigma Level.
    You may still disagree with number 2, but I think we see eye to eye on the others.

    0
    #80208

    Mike Carnell
    Participant

    Efrain,
    I have been in a lot of bucket shops and I have never seen a blade that didn’t get hand finished. I may not hold the majority opinion on this but the hand finish after machining or forging is a rework. Probably not an average machining guy even on a good day but the shapes are fairly complex and I am not sure there are options. If you consider it rework you will stay focused on driving it out.
    I think part of the issue here is are engines built to some sigma level or do they operate to some sigma level. I don’t know anyone at GEAE who has ever professed to building to even a Six Sigma level. Delivery is another story.
    I would watch the comments like “all processes have a lot of variables… That isn’t any different than anyone else and you don’t need to control them all. They are not all leverage variables.
    Good luck.

    0
    #80221

    Mikel
    Member

    John,
    You wrote  “1 error in 1E20 is ludicrous”.
    I am proud of you, that was my original point to you.

    0
    #80223

    Withheld
    Member

    John,
    You’re right – it doesn’t seem that we disagree. It sure seemed otherwise for a while, though. It appears I’ve got some work to do on my comprehension skills.
    BTW, in your point #2, let’s assume that you can park your car without hitting the wall 100% of the time (I wouldn’t know – my garage is too cluttered to try).
    Is damage from the specific step of parking the car the failure mode? How about if you’re a little too far to one side and your kid scratches the paint while trying to get her bike out? If the Customer is a leasing agent and you return the car with a scratch, the 6+ sigma process you defined means absolutely nothing.

    0
Viewing 38 posts - 1 through 38 (of 38 total)

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