Short Run Capability - Batch quantity is 15,000 to 20,000

[Andrews]

We are a batch (small) production company manufacturing number of turned parts for the automotive industry.Usually the batch quantity is 15,000 to 20,000 nos. This batch quantity will be completed in 2- 3 shifts (8 hr shift).

For special characteristics, we have defined that products will be checked 5 nos. per hour. All SPC manuals require atleast 20 subgroups to be taken before calculating control limits. 20 subgroups will take atleast 2 - 2 1/2 shifts. So we are able to use the control limit calculated for only 1/2 shift in some cases.

 

[Rick Goodson]

Andrews,

By your own numbers you produce between 625 and 1,250 pieces per hour. You have selected a time period of one hour and a sample size of five pieces. The basic concepts of SPC do not dictate the sample size or time between samples. A sample of three pieces taken at 10 minute intervals would give you 20 subgroups in slightly over three hours. In addition, you could look at some of the short run techniques such as the nominal X chart for runs of parts with similar specifications. You would have to look at the short run requirements such as similar standard deviations to assure they would work. Go to www.i-q-i.com for more information on short run SPC.

 

[Andrews]

Rick said :

A sample of three pieces taken at 10 minute intervals would give you 20 subgroups in slightly over three hours

Problem is in addition to special characteristic , we have a min. of 15 characteristics on a product.We check these characteristics 20 nos. in an hour. Therefore it is difficult to check three pieces at 10 minute intervals for special characteristics.

 

[Sam]

Actually the requirement is 25 subgroups. Also, "the level of initial process capability shall be determined to be acceptable prior to submission for all special characteristics . . . . ". Sounds like you may need approval from your customer.

 

[Chemman]

Andrews,

I also have a problem similar to yours. Several thousand pound batch of material produces roughly 3000 parts over 2 - 3 shifts. There are several physical tests that are performed downstream after oven curing and machining. Parts are set out for QA and by the time the tests are complete, all the parts are sitting in the audit area. Due to the nature of the process, length and expense of the testing, it is impossible to sample parts at defined intervals to establish a control chart for that batch. The parts needed for testing are all pulled at one time to represent that batch. What we end up with is a chart where subgroup to subgroup is a different batch of material, with lots of points showing out of control, but after the parts are already made and obviously no way to then react and bring the process into control. So I end up with a chart of 25 subgroups, all from different batches of material. I do not believe this is proper use of SPC, but I have a hard time explaining this to the customer who only understands if a point is between the dotted lines or not. Between the customer and the auditor, it is driving me nuts and I have yet found a way to work around this. Is there any other way to do SPC in this situation?

Thanks,
Chemman

 

[Graeme]

I have recently discovered a book that may be useful for gaining insights into this topic.

Statistical Process Control Methods for Long and Short Runs, Second Edition by Gary K. Griffith. (ASQ Quality Press, 1996)

I have only recently started to study this book, but it has already refreshed one key thing I had pushed to the back of the brain -- focus on the PROCESS, not necessarily the parts. The goal is to determine (and monitor) the capability of the PROCESS.



Graeme