process sigma calcs
Six Sigma – iSixSigma › Forums › Old Forums › General › process sigma calcs
 This topic has 20 replies, 13 voices, and was last updated 14 years, 5 months ago by V.Sankar.

AuthorPosts

December 20, 2005 at 3:35 pm #41782
Hi,
I am in the process of calculating the defect rate at my plant at a specific machine/process to eventually calcluate the process sigma level
The problem is this particular process has hundreds of product numbers being processed.
My question is if i have to calculate the defect rate for the process what should be the sample size?
each of these product numbers have different requirements.
how do i calculate the standard deviation & mean as well as the overall process sigma level for this particular process?
i hope somebody can help me with this
Thanks,
Vinod0December 20, 2005 at 3:48 pm #131398For small sample sizes (<30) use Rbar/d2, For larger sample size use the sum of square formula. Beware that you are not accounting for your nosies.
0December 20, 2005 at 4:21 pm #131401We use sigma calculators that can calculate overall sigma levels out of different processes, but I am afraid I cannot get you these because this copyrighted and confidential materials. Nevertheless I will try out what’s behind the calcs and try to update you on that one if I find out.
0December 20, 2005 at 4:25 pm #131402
Sean StephanMember@SeanStephan Include @SeanStephan in your post and this person will
be notified via email.Vinod,
I do not know what CB was referring to. If this Chuck Busack I would understand. A process with several different products running on one process line all you need to know is the number of total products ran for a minute, day, week, or whatever time interval you need. This is your opportunities and then the number of failures on the process line for all products. Divide your failures by your opportunities and this gives you your DPU (Defects per Unit). From this you multiply that by 1000000 and you get your DPMO(Defects per million opportunities).
In excel you can use the inverse of the normal distribution formula to get your sigma level.
Such as : = normsinv(1(DPU))+1.5
This should give you your Sigma. Minitab can also be used to determine your sigma as well. Reply and I will send that to you.
7epsilon0December 20, 2005 at 7:08 pm #131416RE: How to cal. sigma
0December 20, 2005 at 7:54 pm #131420Sean,
Thank you very much for your help.
Here are the details:
total number of units inspected at random=250
no of defects =89
Using the formula in MS Excel the sigma value turns out to be 1.86
The product is an Automotive Hose and has to be cut to a specific length.
I say there are hundreds of product numbers because there are hoses with different cut length requirements.For Example 32 hoses may be cut for the required cut length of 22.8 inches ,another 16 may be cut for 12.7 inches and so on.So the different cut length requirements refer to different product numbers.
Is this value of 1.86 sigma value good for the entire process at this cut machine??
How would I be able to calculate the mean,std deviation etc ?The problem I am facing is, when there are so many product numbers with varied cut length requirements how would i be able to find the process mean/std deviation?
I hope this information was helpful for you to understand the situation even better.
Thanks a million and I look forward to your solution
Vinod
0December 20, 2005 at 8:44 pm #131423Vinod,
Of the 89 defects you found, what are their length? Can you pareto the defects into categories by length i.e. (10 – 15, 1620, etc..) are your defects more likely with longer or shorter lengths? Are your defects to long or to short?
I would say you report your capality as a whole and look at your defects in a pareto format, you may find that over a certain length your accurracy tends to fall.
GDS
0December 20, 2005 at 9:40 pm #131425CB,
Sean is right – what are you talking about?
Go read the question.0December 20, 2005 at 9:46 pm #131426GDS,
Of the 89 defects most of it is excess(I mean more than the USL).This contributes to material wastage and ultimately lost$$$.
For cut lengths above 25 inches it starts getting worse(excess), and when it gets to cut lengths of 35 inches plus it is really bad(excess).But every once in a while we do come across a short one.
I still havent been able to figure out how to calculate the overall process capability.
Please note that we are looking at one particular machine and one particular process with a large number of different product numbers with respect to length.
Thanks, & I hope you can help me out here
Vinod0January 9, 2006 at 1:46 pm #132067Vinod,
This is what I would do if I were in your shoes. As I understand, you want to know the defect rate and all the associated statatistical values; sdt dev, mean, cpk..etc.
Since you are running multiple products on one machine and you are measuring a variable data, I would normalize my measurements and take the distribution of the normalized variable data with USL and LSL around them and calculate all your statistical values you are interested in.
ex: hose length of one type product 22.8 inches, another 12.7 inches. If these are target values, set them to zero. any variation from these treat them as + delta or – delta. and your USL and LSL also need to be normalized. after you get enough values representing your process then calculate your numbers as if you were using a normal distribution with real values…. from these you can get your DPMO, cpk and all the other good stuff…
let me know if you have any questions..
KE
0January 9, 2006 at 7:13 pm #132094Dear Vinod,
Since you know the target cut length of each product number you can subtract the cut length from the target cut length and generate a control chart based on the difference. I have seen this referred to as standard dviation about the target and will generate a standard deviation number.0January 9, 2006 at 7:35 pm #132095Hi KE,
Thanks a lot for your help. I shall try them and let you know
Thanks a million once again
Vinod
0January 10, 2006 at 8:30 am #132110KE,
I have a question for your suggestion. How we can normalized the USL/LSL. Usually, for exmaple, type product 22.8 inches, it tolerance is +/ 1 inch, but for type product 12.7, it’s tolerance is +/0.5 inch and so on. So, in this situation, this process can not normalize the USL/LSL, even we get +/ delta
Does anybody have some opinions?
Michael
0January 10, 2006 at 2:24 pm #132119Well, don’t you want to know the capability of your process? Now, you got your opportunity! You can report your capability numbers in the increments of ± 1, 2 3 , 4 inches or what ever your tolerances are! Your process capability is what it is unless you act on it! Therefore, you can report your process capability in the increments of your spec limits!!!! This way your management will not accept any product less than what the process can produce or will have the knowledge of % defectives and implement a screening inspection or do something about the process to improve its capability….
0January 10, 2006 at 2:27 pm #132121Hi Vinod,
I guess that you’ll need to do different things depending on the goal of this measurement.
If the goal is reduce the number of defects then it wold definitely make sense to do a Pareto on the different product types and go for the few big contributors.
If you only have to generate a report and check that nothing out of the ordinary happens then possibly measuring the total number of defects regardless of the product type and calculating the DPMO would be good first step.
Still, it is a known risk that you can have a good overall DPMO number with some subprocessess that are very poor, so I’d do the Pareto anyway and look at the poorest performers, maybe even track them separately.
Regards
Sandor0January 10, 2006 at 4:08 pm #132129Hi KE,
I somehow could not understand what you were suggesting about reporting the capability numbers in increments of +/ 1, 2, 3, 4 inches(or whatever the tolerances are).How can I possibly report the process capability in increments of the spec limits??
Heres an example:target/required cut length=22.8,USL=23.1,LSL=22.5(All units are in inches). So target would be set at 0 and any variation from this target is reported in terms of +/ delta.For example if the actual cut length of the hose turns out to be 23.3 inches we would assign it a value of +0.5.Is that right?.And USL would be +0.3 while LSL would be 0.3
Now again since we run a large variety of product numbers requiring different cut length requirements,we would have another product requiring cut length of 37.4 inches,with USL=37.9 & LSL=36.9.So the target =0 as before but how do we assign values for the USL&LSL??Let us assume that the actual cut length observed was 38.5 inches long(that is a value of 1.1 in terms of coded data)
How can we compute the overall process capability of the cut operation process regardless of the type/length of the product run??
I hope this would be useful,and sorry for not being able to understand
Thanks
Vinod
0January 10, 2006 at 4:19 pm #132130Vinod,
Send me your phone number in next 15 minutes. I will call you and walk you through the process.
KE0January 10, 2006 at 5:00 pm #132135KE,
Its 5737762114
thanks
Vinod0January 11, 2006 at 11:10 am #132189
Anshul JunejaParticipant@AnshulJuneja Include @AnshulJuneja in your post and this person will
be notified via email.Hi Vinod,
I know your intial query was on calculation of process sigma. But the ultimate goal is to improve the process. Taguchi online control is one way through which you can improve your process capability and its much better than going in for control charts and it provides a table for adjustments to the labour.
You can find lot material on taguchi online control, I know a way through which you can use nested anova and taguchi online control together, do let me know in case you are interested, along with your email id
Regards
Anshul0January 11, 2006 at 3:32 pm #132205
Chris SeiderParticipant@cseider Include @cseider in your post and this person will
be notified via email.KE,
I’m curious why did you state Vinod could calculate his DPMO? Did you tell him to calculate his value added opportunities or did I miss this string of post? The specification does not have anything to do with the opportunities.0April 12, 2006 at 6:58 am #136267Hi vinod
i am sankar working for manufacturing company in TN, i am going to do 6sigma practices in my small industry.so,kindly help me how to start with this .
NOTE:I have read ur problem regarding the populatin size of ur products, can u tell me,what is the figure u have got from the professionls,and do u have any idea by which they suggest.
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.