How to Effectively Apply Short Run SPC

E

elmabawaan

can somebody help me or give me a little advice about short run spc...how to effectively apply this technique...

as of now we are using this technique, but im not sure if we really are using it effectively coz our production run is based on job order...please some advice...
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: short run spc...

We need to know what kind of process to give you a hand. :cool:
 
E

elmabawaan

ok, here's how it goes...we are into stamping of small parts for motorcycle, based on the control plan, there are some products that require SPC monitoring, the problem is we are producing parts based on job order or by batch. we have 22 parts for one customer & this parts are delivered on a weekly basis.
for products that require SPC we used DNOM chart since the normal SPC cant be applied because these products do not run 24 hrs a day.

mostly, these products are set in the production within an hour or two

example:
product A has JO of 5000 pcs in 1 month, 5000/4 since it is delivered weekly,
1250 pcs only requires 2 hrs to produce, so the data for our SPC is based on a 2 hrs production run, that is why we considered this as short run SPC, every production run we take 25 samples & this data is being computed based on DNOM chart...
 

Miner

Forum Moderator
Leader
Admin
What specific concerns do you have? That is what makes you think that you are not using it effectively? Or are you concerned whether DNOM is the correct approach?

Attached is a summary of Short Run SPC methods with the advantages and disadvantages of each.
 

Attachments

  • Short Run SPC.pdf
    75.8 KB · Views: 1,675

Steve Prevette

Deming Disciple
Leader
Super Moderator
we have 22 parts for one customer & this parts are delivered on a weekly basis.
for products that require SPC we used DNOM chart since the normal SPC cant be applied because these products do not run 24 hrs a day.

What is your basis for the assumption you can't do SPC? I would suggest keeping a running chart for each part, and just connect the various batches of that part together. I would think it would be worthwhile to be able to detect changes from batch to batch. Just update each part chart as you happen to create that part.
 
E

elmabawaan

What is your basis for the assumption you can't do SPC? I would suggest keeping a running chart for each part, and just connect the various batches of that part together. I would think it would be worthwhile to be able to detect changes from batch to batch. Just update each part chart as you happen to create that part.


meaning the moving range chart? i dont know if i understand it right, moving range is used when you have a small volume to produce & is impossible to get at least 25 samples? that is why we use DNOM chart...

by the way my concerns about this issue is that... is DNOM chart appropriate?
how can we compute for the Ppk or Cpk using this data, coz this is what we have as of now...this is always the issue every time we have an audit....

can we use the computed std dev for the computation?

sorry i have lots of concerns...but i really need some inputs...

thank you for your time responding to this query...
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I certainly understand the situation you have with short run stampings. Stamping itself is an unusual process, in that material changes and setup changes (which are special causes) have more influence on the output than the tool wear (common cause). But, proof is in the data, and what would be handy is some capability data to really see what variation you are seeing within a run (always the first step to preparing SPC). I suggest measuring 100 pcs in order - perhaps every 10th part to capture more variation - from the same cavity if multi-cavity and preferably at the same location on the part. Prepare a run chart of the data (or provide it to us and we can). We also need to see the gage R&R to determine the contribution of the measurement error to the total variation. It could be critical, in that the part to part variation is likely very small, especially if the process is controlled to the point where just tool wear is evident. I think it is one of those cases where we need to understand the actual distribution before tossing darts out for picking control charts and evaluating capability. You may have this data, and that is good, be we are at a loss to help you without it. :cool:
 
E

elmabawaan

I certainly understand the situation you have with short run stampings. Stamping itself is an unusual process, in that material changes and setup changes (which are special causes) have more influence on the output than the tool wear (common cause). But, proof is in the data, and what would be handy is some capability data to really see what variation you are seeing within a run (always the first step to preparing SPC). I suggest measuring 100 pcs in order - perhaps every 10th part to capture more variation - from the same cavity if multi-cavity and preferably at the same location on the part. Prepare a run chart of the data (or provide it to us and we can). We also need to see the gage R&R to determine the contribution of the measurement error to the total variation. It could be critical, in that the part to part variation is likely very small, especially if the process is controlled to the point where just tool wear is evident. I think it is one of those cases where we need to understand the actual distribution before tossing darts out for picking control charts and evaluating capability. You may have this data, and that is good, be we are at a loss to help you without it. :cool:

this is our GR&R result:

% EV = 0.9%
% AV = 0.19%
% GR&R = 0.92% looks good
% PV = 100% ???
ndc = 153 isnt this too much

this is the sample data for our SPC:

46.4 46.44 46.4 46.5 46.33 46.4 46.49 46.6 46.55 46.54 46.43 46.48 46.4 46.5 46.22 46.5 46.54 46.6 46.57 46.51 46.46 46.48 46.4 46.4 46.47 46.6 46.5 46.6 46.52 46.46 46.44 46.41 46.4 46.4 46.43 46.3 46.45 46.5 46.54 46.61 46.44 46.48 46.4 46.4 46.42 46.6 46.45 46.6 46.52 46.63
46.37 46.4 46.43 46.4 46.44 46.45 46.47 46.5 46.52 46.54 46.39 46.4 46.36 46.39 46.43 46.48 46.48 46.51 46.44 46.54 46.43 46.37 46.41 46.44 46.43 46.47 46.46 46.51 46.58 46.54 46.38 46.37 46.44 46.4 46.44 46.38 46.49 46.5 46.53 46.53 46.42 46.35 46.44 46.45 46.43 46.47 46.48 46.51 46.55 46.52
please can you provide me a sample of running chart based on this sample?
this data came from one of the products we are monitoring...

thank you so much for your help...
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I have attached the run chart, a histogram (did not have the specification, so just dropped in the autocalculated ones), as well as the raw data in case others want to try their hand at it.

There is something about the data versus the GR&R, especially with the resolution I would expect to see with an ndc of 153. What is the tolerance of this part, and what is the tolerance of the part used in the GR&R (or was the GR&R done on this part)? What is the shape of the dimension you measured (length or diameter)?

Thank you very much for sharing your data and allowing us to review it.

Hang with us, this will get forensic.:cool:
 

Attachments

  • stamping run chart.jpg
    stamping run chart.jpg
    38.7 KB · Views: 418
  • histogram 2.jpg
    histogram 2.jpg
    44.1 KB · Views: 404
  • Stamping Data.xls
    20.5 KB · Views: 571

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I decided to try an analysis to see if there was some underlying competing distributions. From 10,000 feet, this appears normal, but really one still needs to be able to understand why that may be. If it was tool wear alone, it would not have a normal distribution. In order to see an underlying trend, I separated all of the data above the mean from below the mean, and performed linear regressions on those data sets. The attached graph shows the results. It shows the variation getting smaller over time. One might explain that from warm-up of the tools. The outside of the slopes, the next thing to ponder is the distance between the trend lines - where is that variation coming from? Is it measurement error (normal distribution)? Are there cycles, in that the die settles in at a couple favorite places on the lead pins?

Again, this is hard to analyze from a distance...but fun nonetheless. Still would be handy to know the specs - we may be picking nits on an elephant, for all we know. :cool:
 

Attachments

  • Stamping Data Hi Lo Trend sm.jpg
    Stamping Data Hi Lo Trend sm.jpg
    27.1 KB · Views: 320
Last edited:
Top Bottom