iSixSigma

Kanban concern

Six Sigma – iSixSigma › Forums › Old Forums › General › Kanban concern

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • February 9, 2005 at 9:13 am #38358

    Chen
    ★ 10 Years ★
    Participant

    Dear colleague’s,
    Common in industry is a service level of 2 when you set the size of the Safety Stock. So, there is 98% chance that the parts are available.What is the change that all parts are available if there are, for example, 40 parts involved in a final assembly?
    Is it just as with TPY, 0.98 x 0.98 x 0.98 x………….. = 45!!!!???
    Hope some of you can help me.
     
    Bernard
     

    0
    February 9, 2005 at 9:29 am #114662

    Anonymous
    ★ 20 Years ★
    Guest

    Do you mean 0.45?
    Notwithstanding the calculation, I would advise a safety stock according to the historical availability performance of the component. You can use an np chart to work out the process performance.
    As I’m sure you are aware, the usual practice is to use small ‘economic runs’ so the key issue is how quickly you can pull parts from a supplier, if necessary. Now that many suppliers are in China, this is going to become more of a potential problem.
    Regards,
    Andy

    0
    February 9, 2005 at 10:45 am #114664

    Chen
    ★ 10 Years ★
    Participant

    Hi Andy,
    Safety stock is calculated as follow:
    Sdtdev (historcal orders) * service level * (manuf. cycle time)^0.7
    So, a service level of  2 wil give an availability of 98%But what happens when I have 40 parts in my assembly?
    45% chance that I can assemble my finished good??
    Best regards,Bernard
     

    0
    February 9, 2005 at 11:16 am #114665

    Anonymous
    ★ 20 Years ★
    Guest

    Hi Bernard,
    Agreed …
    What metric do you use for the stdev? Is it based on ‘on-time delivery,’ or losses.
    (By losses I mean you expected to receive 10 only got 9, and out of the 9 only 8 were usable, etc.) This is what I implied by plotting the number of losses on an np chart, and working out the process performance. Of course this all assumes that a process has long term stability.
    For complex processes such as vacuum braizing that have to have a cosmetic finish and seal fluids, the availability performance might be quite different to that predicted by the process performance. In other words, it could suddenly become unstable and unpredictable.
    Regards,
    Andy

    0
    February 9, 2005 at 11:22 am #114667

    Peppe
    ★ 10 Years ★
    Participant

    It could means that you are able to produce, at least, 98% of order on time, based on calculated historical order and lead time.
    The bins, of course must be full. 
    The parts you have isn’t important, because it is kanban re-order, based on bins numbers, to give you always material available.
    Rgs, Peppe

    0
    February 9, 2005 at 12:01 pm #114668

    Anonymous
    ★ 20 Years ★
    Guest

    Peppe,
    Thanks for the clarification ..
    I became confused between the ‘availability’ and the ‘stocking level.
    If instead of the ‘stocking level’ one calculated the availability performance of each component based on a historical availability performance then isn’t here cause for concern because it only take one missing component to bring the line to a grinding halt?
    Cheers,
    Andy

    0
    February 9, 2005 at 1:55 pm #114669

    Wilson
    ★ 10 Years ★
    Participant

    Could someone please define “service level”??

    0
    February 9, 2005 at 2:42 pm #114672

    Chen
    ★ 10 Years ★
    Participant

    Hi Andy,
    stddev is related to the part usage in the end assemblies over the last year .
    For examp:
    week 1  

    0
    February 9, 2005 at 2:42 pm #114673

    Chen
    ★ 10 Years ★
    Participant

    Hi Andy,
    stddev is related to the part usage in the end assemblies over the last year .
    For examp:
    week 1   

    0
    February 9, 2005 at 2:53 pm #114674

    Chen
    ★ 10 Years ★
    Participant

    Hi Andy,
    stddev is related to the part usage in the end assemblies over the last year.For examp.:
    week 1         45 partsweek 2        200 partsweek 3         20 partsect.stddev = 98assume MCT = 2 weeksSafety stock = 98 x 2 x  2^0.7 = 318With this 318 parts safety stock you make sure that for 98% (+/- 2 sdtdev under the normal curve) parts are available.
    The factor 0.7 is an empirical value and has to do with the reliability of the supplier (0.7 = standard, = wors) an thus indirect with the yield you mentioned.
    With best regards,
     
    Bernard
     

    0
    February 9, 2005 at 3:13 pm #114675

    Anonymous
    ★ 20 Years ★
    Guest

    Thanks Bernard … I believe Peppe’s answer is correct. I was just confused this morning!
     

    0
    February 10, 2005 at 7:33 am #114690

    Chen
    ★ 10 Years ★
    Participant

    Hi Andy and Peppe,
    Thanx for your replies.
    Maybe I have to formulate things in an other way:
    300 types end assemblies .Each type end assembly consist of a different combination of 40 parts out of a pool of 500 part numbers.Per part number is the chance that it is available 98%.So, what is the chance that I can assemble a random chosen end assembly.
    Thanx again,
    Bernard
     

    0
    February 10, 2005 at 10:07 am #114692

    Peppe
    ★ 10 Years ★
    Participant

    Based on your data, it is always 98%.
    But this is not a Kanban concern.
    Rgs, Peppe 

    0
    February 10, 2005 at 2:33 pm #114706

    Chen
    ★ 10 Years ★
    Participant

    Hi Pepe,
    Thank you for responding.When it’s indeed 98% we can live with that.Has it to do with paralel or serial flows, just as with currency?
    Is it possible to explain to me why the chance to make an end assembly is 98% and not less?I will be grateful when you could explain it to me.
    Best regards,
    Bernard
     

    0
    February 11, 2005 at 8:11 am #114742

    Peppe
    ★ 10 Years ★
    Participant

    Bernard, maybe we are misunderstanding about terms or definitions.
    If you talk about the product of probability, it is correct that 0.98*0.98.. ~0.45, but I think it isn’t the case. I think the basic question is about Kanban sizing, tipically based on 2 bins.   From your data seems be that over 500 single parts, have been selected 318 of them most  used, based on your historical data, lead time, etc..,  so that if the orders will follow the data considered and the calculation is correct, you’ll be always able to assembly 98% of 300 different end assembly, of 40 parts,  on time.
    If my understanding is wrong, please advise, so that I can correct it.
    Rgs, Peppe
     

    0
  • Author
    Posts
Viewing 15 posts - 1 through 15 (of 15 total)

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