Need Advice on Unique SPC Model
Six Sigma – iSixSigma › Forums › General Forums › Methodology › Need Advice on Unique SPC Model
- This topic has 3 replies, 4 voices, and was last updated 1 year, 6 months ago by
doe.
- AuthorPosts
- August 26, 2019 at 11:09 pm #241473
jmwlmsonParticipant@jmwlmsonInclude @jmwlmson in your post and this person will
be notified via email.I need advice on a project- I have been asked to create a potential SPC model. I will describe a hypothetical:
The company is making printed circuit boards at approximately 300 boards per 8 hour shift. We have the ability to pull any board at any time and place in a testing chamber that will analyze the board to determine it has been built to specification.
The company would like to develop a “SPC model” to determine how many boards should be pulled out for testing in order to meet a CPK <1.33, assuming all boards tested are within specification. However, the issue I have is that the specification has a lower limit, but not an upper limit- it is a pass/fail test. So, for example, if the results of a board pulled from production and placed in the testing chamber are >8 units, then the board is acceptable for release.
I am new to this, so please be kind. Could anyone provide some guidance on how or if this can be done?
0August 27, 2019 at 10:12 am #241478
Chris SeiderParticipant@cseiderInclude @cseider in your post and this person will
be notified via email.I hope you mean Cpk>1.33 :)
A pass fail test and using a sample of 8. Just say no and convert the attribute test to a variable value test.
Why not use SPC as designed and not do SPC on a Cpk value on 8 parts (the imprecision of the s.d. enough is to make this kind of ridiculous in my mind). Just do SPC as designed and work to raise the average to get parts to get your process capability greater than 1.33. You’ll have to balance the financial impact of getting too far above the minimum spec but that would be the value of doing SPC on your Xbar chart–cause action if the 8 part average is too high or one of the other Shewhart rules you use.
- This reply was modified 1 year, 6 months ago by
Chris Seider.
0August 28, 2019 at 12:42 pm #241582
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.@jmwlmson I am with Chris Seider on the ridiculousness of this at least if you are going to parade it around as SPC.
I cannot even imagine what test I could do after a board is complete that would tell me that the board had been built correctly particularly if it is a multilayer board.
0September 2, 2019 at 1:23 pm #241836
doeParticipant@[email protected]Include @[email protected] in your post and this person will
be notified via email.I’d use continuous run sampling.
Example:
Management wants to know if the process is at, or can demonstrate, a 1.33 Cpk (or a 4 sigma quality level). What is the sampling plan for a continuous running process? The desire is to not interrupt production, and want to be 95% confident.
Answer:
A) 4.0 sigma is equal to 31,671 PPM defective. To get to the level of 4.0 sigma (1.33 Cpk) quality is: 1.0 – .031671 = .968329. (The .968329 is the PPM equivalent of 4.0 sigma (1.33 Cpk-two tail).
B) 95% confidence = .05% risk
C) The formula is:
n = nL( .05) / nL(.968329)
= 94
Therefore, if your process makes 94 or more, defect-free parts, consecutively, you can be 95% confident that the process is showing a capability at least equal to 1.33 Cpk. Would require looking at 94 continuously made parts, however.
I’d sample the process and get the dispersion factor and location parameter.
For 95% confidence in a
0 - This reply was modified 1 year, 6 months ago by
- AuthorPosts
You must be logged in to reply to this topic.