Re:Factors for Control Chart.
Six Sigma – iSixSigma › Forums › Old Forums › General › Re:Factors for Control Chart.
 This topic has 7 replies, 5 voices, and was last updated 14 years, 8 months ago by Erik L.

AuthorPosts

September 13, 2007 at 7:34 am #48103
Dear Dr. Scot and all. Thank you for your attention.
I post new message with details.
I’m working in glove factory. For 3 days, we run new line. Every single day, we use different parameters. For example, 1st day, the latex tank temperature is 160oC and the second day is 200oC. We doing that to get the optimum output. We assume, if the defect of the output is less, means the parameters for that day is appropiate for running line.
Our machine is running 24 hours. For each day, we take 125 samples from every 35,000 for 44 time.
I want to plot control chart for variables with data of defected glove. So, what i have understtod, the subgroup is 44 right because one day I take samples for 44 times.
My problem is, I cannot plot the chart because I cannot find the factors for control limits for xchart, schart and rchart for n=44. I mean the value of A2,A3,B3,B4,D3 and D4.All my reference book only stated n=1 until n=25. So how can I plot the control chart?
So am I wrong at any stage? please advice.
TQ so much. I’m not English. Sorry for not so good in writting.
0September 13, 2007 at 7:59 am #161076
drqualityParticipant@drquality Include @drquality in your post and this person will
be notified via email.Depp,
Can you post the data for each INDIVIDUAL value measured here? For the moment, forget about subgroup size, that is yet to be determined.
Also, why are you looking at temperature only? There are likely other factors that influence the quality of the gloves.
But for now, try and post the temperatures, and the quality measure you are using to “judge” the quality of the gloves. Preferably, the measure would be a variable measure (e.g., weight, thickness, strength, etc.) rather than an attribute measure (e.g., “good or bad” which sounds like what you have now). But either way, we can learn something from what you have.
I am happy to help if you wish.
Dr. Scott0September 13, 2007 at 8:19 am #161077Dear Dr. Scott.
For your info, we already fix the temperature for latex tank for each day. Means it is constant. Day 1,T=32oC, Day 2,T=42oC and Day 3, T=52oC.
T=Temperature.
The quality of measure I’m using to “judge” the quality of the gloves is counting the defect of the glove for every 125 sample from every 35,000 for 44 x.
I will post the data on next message.
0September 13, 2007 at 3:05 pm #161092deep,
If I understand your message correctly, 44 is the number of subgroups not the number of samples in each subgroup. So you would not look up n=44 to find the constants. It also appears that you’re charting the number of defects in your samples. If this is correct, you would not use a variables control charts (Xbar & R or s) so the values of A2, A3, B3, etc. don’t apply.
You should be using a cchart to chart the number of defects. Good luck.
Chris
0September 14, 2007 at 3:49 am #161108Dear Dr. Scott,
What I understood, to study capability process, I have to know the control chart for variable.That why I want to plot Xbar and R chart.
Below, I give result for Day 1 only.For Day 2 and Day 3, the result are similar but have bigger defect.
Day 1
QA lot Sample Defect
1 125 3
2 125 3
3 125 3
4 125 3
5 125 3
6 125 3
7 125 3
8 125 3
9 125 4
10 125 3
11 125 3
12 125 4
13 125 3
14 125 4
15 125 4
16 125 3
17 125 3
18 125 3
19 125 4
20 125 3
21 125 3
22 125 5
23 125 3
24 125 4
25 125 3
26 125 3
27 125 4
28 125 3
29 125 4
30 125 4
31 125 3
32 125 5
33 125 3
34 125 3
35 125 4
36 125 3
37 125 3
38 125 3
39 125 4
40 125 3
41 125 4
42 125 3
43 125 5
44 125 4
Do you have any idea how to process the day for study the process of capability ?
Tq for your attention.0September 14, 2007 at 3:53 am #161109Dear Chris,
Can I study the process of capability through Cchart?
TQ for your attention.0September 14, 2007 at 5:44 am #161113yes c’ chart may tell us about your process direction over time and will also your process movement on UBL and LBL
thanks
Ahsan0September 14, 2007 at 5:53 pm #161137Depp,Ive taken an initial look at the data that you have provided. There are some assumptions, up front, that Ill make based off of your description of the measurement process. First, Im assuming that within your sample of the 125, taken 44 times in a day, that a glove is deemed rejected if there are any defects observed. Im assuming that your tally of defects is not tallying scenarios where an individual glove can have multiple defects. If that is true, then you are dealing with defective data. Your attribute control charts of choice would be the np or p chart. You could also track the proportion defective via an IMR chart to assess signals that may be provided by that control chart.Before we would move into any capability assessment, we would want to ensure the accuracy of the measurement process that is being used by your inspectors to determine what is accepted and rejected. You could conduct an attribute MSA of the process and if you see Kappa scores in the .9 and higher arena you should feel pretty good about the data. If you havent conducted one of these, Id highly encourage it. Well assume that the process used to inspect is consistent across the various inspectors, shifts, days encompassed by your data.Three ppt slides that cover some of the analysis that you could conduct with this data.Slide 1 is an overall capability assessment of the data. Couple key points. Your 44 samples, per day, is sufficient to give a believable estimate of the average proportion defective for this process. Average proportion defective would work out to 0.02745 and a PPM level of 27455. That would work out to an estimate of the process sigma level of 3.4. Slide #2 show the data in an IMR format showing similar signals in the data. Additional insight that this chart provides is that we would typically expect a delta of 0.00577 sample to sample and could see a swing as high as .0188 and still be within the expected variation of the process data.If we assume that this is the baseline of the defective rate and that temperature is a key factor in mitigating the defects, how could we proceed? We could use that data to decide what kind of sample size would be required to capture a shift in the proportion defective. For instance I took a conservative view of your process data and took the lower 95% CI for proportion defective= 0.0233 and asked what minimum sample would need to be collected to show a shift to 0.01 with varying powers of .8, .85, .9, and .95. Since this is a yield scenario I force the data into a onetailed projection for required sample size. That output would be in the third slide.Of course, the more interesting analysis would be to test the assumption that temperature is truly a factor which impact the proportion defective in your data.Hope this has helped with some insight into your data and how you might proceed further.Regards,Erik
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.