• The samples(subgroup) used to calculate control limit need come from different shift /operator , at least 20 subgroups is need, so, one day ‘s test-data is not enough.
    It is need to remind the control limit need to modify (CL may advise) if process capability is changed.
    Best regards

  • Hi:
    I believe we need to know the process capability of the control chart establishment.
    If the control chart is generated based on a capable process (e.g. Cpk>1.33), the in-control plot-point can ensure good shipment, because the possible escaped rate under this situation is less than the escaped rate of performing normal sample plan i…[Read more]

  • marklamfu replied to the topic Bug in the forum Suggestion Box 15 years, 8 months ago

    Hi 2 sigma:
    We can category CPK as “process CPK” and “product CPK”, for the multi-assemble process, we only need to know individual process CPK ( generally, not include inspection stations) as to improve,  a overall process Cpk seem no meaning.
    If we want to know overall qulity of final product which passed all assemble process, we also can us…[Read more]

  • marklamfu replied to the topic EOQ Practice in the forum General 16 years, 2 months ago

    Dear sir:
    The problem what you met is common issue, in fact, in some situations, it is very difficulty to directly measure a product by quantified indexs, my opinion is:
    1. According to SixSigmg basic rules, the defect means the defect of CTQ(critical to quality), it does not directly equal to defect of  product. so, your performance metric mus…[Read more]

  • marklamfu replied to the topic HRM in the forum General 16 years, 2 months ago

    The direct DMMP chart is not a variable control chart, generally, for process or product control , we perfer to use variable chart ( e.g. X_bar &R chart , IX_MR chart,etc.), DPPM chart is used to review the product performance achieved,
    If no chance to do variable control due to product nature (only attribute parameter) we also can use DPPM to…[Read more]

  • What you described seem be Champion rather than MBB.

  • The attributed chart(e.g. P-chart, np-chart,etc) do not fit the so low ppm, please use variable method to control the process, X_bar chart is OK, CPK trend is OK yet.

  • If the data is from a normal distribution, I think the minimum sample size is 30, >50 is perfer

  • Do not worry!I am QA manager yet, and ever handled ISO9001:2000/ISO14001:1996 certification and many similar projects, I would like to share my a few opinions here for your reference:1. Before start the project, the justfication for the project should be confirmed and promoted within senior meanagement team, and get their consensus and…[Read more]

  • marklamfu replied to the topic sample size in the forum General 16 years, 6 months ago

    It is a interested case!
    One operator to perform R&R stduy is not adequate, as you said, only one R (Repeatability) is done.
    You known, the aim of perform R&R study is to control the measurement as a suitable condition to meet customer requirement. Lack of another R, the status of this measurment can not be confirmed. 
    In fact, the persons in…[Read more]

  • The selection criteria of  BB is different because the product and culture of  organization are  different , Now,  I would like to share my company’s consideration for BB selection:
    1. To show out following behaviours: 
    Teamwork,  Information seeking and Analytical Thinking
    2. Skills:
    SPC knowledge, Project management,  risk analisis s…[Read more]

  • No way to 100% ensure right implementation all time. I recommend you to follow the process below to implement new process:
    1. Document the new process
    2. Train up & quaify the every user with the document
    3. Monitor and audit the implement
    4. Identify and correct the deviation
    5. Review and enhance the step1~step4.

  • For electrical test, usually, the variable sampling plan(e.g. Mil-std-414) is used to control outgoing parameter’s distribution.If you used attribute sampling plan, in order to smooth your process to reduce the WIP , you can reduce lot size, or apply CSP-1 sampling plan. 

  • marklamfu replied to the topic Falsified data in the forum General 16 years, 8 months ago

    It is for striving for Zreo defect and enhacing the sensitivity to defect, 0.1% defect rate =1000 ppm, 0.01%=100 ppm, 0.001% =10 ppm, when quality achieved a high level, using % to measure defect seem not sensitive, e.g. six sigma process defect is 3.4 ppm, (0.00034% seem ….), as if  to buy gold, we say to buy 10 gram not say to buy 0.001 kg ,…[Read more]

  • Yield is typical attribute data , t-test is used to comparing 2 sets variable data, if you want to use T-test or Poisson for attribute data, I believe, it is not accurate, Generally, we can plot unit(daily, weekly) yield as a trend for comparing, or transfer attribute as a variable data and then use t-test or other tool(s) for comparision,

  • marklamfu replied to the topic FMEA prioritization in the forum General 16 years, 8 months ago

    The high CP(2.29) means the process variation is small (the Spec. have cover about 13.8 std.), for 6 sigma process, only CP>2.0 is not enough, at same time, the CPK >1.5( with 1.5 sigma mean shift) and DPPM<=3.4ppm.
    Th cause of CP high but process out of control is big mean shift. the fianl index to measure process capability is CPK not CP, CPK…[Read more]

  • marklamfu replied to the topic Six Sigma in Montana in the forum General 16 years, 8 months ago

    Answer your question:
    1. Small std. compare to product spec. and  expected process capability , if  expected CPk>1, small std means that 6std 1.33, small std means that 8std<(USL-LSL).
    2. For grinding process you mentioned, if machine setting have limited change, the parameter mean of product will be shifted, but, the std. may be  no change. in…[Read more]

  • marklamfu replied to the topic Six Sigma in Montana in the forum General 16 years, 8 months ago

    Firstly, the most important condition of a stable process is to have small std, the high variation means process data can not follow statistical rules, at that time, we can not to precisely control process & predict what will be happen for the process because the population of process can not determined through samples analysis.
    Secondly, if proc…[Read more]

  • I do not think Champion equal to BU Head or Sponsor, I think Champion is only a honorary title for the BBs who have successfully completed serveal big six sigma projects. Champion may be a project leader , perhaps, he/she also may be a sponsor  in another project.

  • For getting a reliable control limit of X_bar chart, generally, number of subgroups is 20 at least , each subgroup include a few samples ( >3), total samples 60~100, for individual chart, every subgroup only have 1 sample, considering the nature of normal distribution, 30 samples is recommended.

  • Load More