HI All,

Being relatively new to the position I am currently in I would be greatful if somone could advise on the following:

We manufacture parts for automotive customers in relatively small batch sizes of around 1000 - 5000 parts. All SPC requirements that we have used in the past have been Initial process studies at the start of production which used a sample size of N10 followed by a sample size of N50.

This has not, as far as I am aware, followed any recognised sampling procedures. I am not too clued up on any sampling procedures but we do use MIL-STD-105E a lot in our control plan inspections!

What kind of procedures are expected by automotive OEM's? Is there one standard for intitial sampling to try and establish process capability and one standard for ongoing inspection?

Please can somone please clarify my obvious confusion?

:thanx:

I may not be understanding you correctly on this but it sounds like you are mixing two situations together. Are you using your sampling for accept/reject decisions or for ongoing process control (control charts)?

If you want to sample for process control and record the data on a control chart, the sample size would typically be n=1 on an Individual Moving Range chart or from 2 to 5 samples on an X-Bar & Range chart. This is where you would normally establish process capability data.

MIL-STD-105 is usually used for accept/reject decisions and the sample size is based on the size of the lot you are making a decision on. Typical sample sizes would be n=10 to n=250 depending on the size of the lot.

There are some very good threads on both these topics in the SPC Monitoring & Statistical Tools forum. I hope this helps a bit.

Dave

If you want to sample for process control and record the data on a control chart, the sample size would typically be n=1 on an Individual Moving Range chart or from 2 to 5 samples on an X-Bar & Range chart. This is where you would normally establish process capability data.

AIAG mandates X-Bar & Range as the default expectation, with a minimum of 25 subgroups and 100 data points. I/MR should be used only in cases where X-bar/R is clearly inappropriate.

MIL-STD-105 is usually used for accept/reject decisions and the sample size is based on the size of the lot you are making a decision on. Typical sample sizes would be n=10 to n=250 depending on the size of the lot.
Actually, the MIL standard is used more for determining sample sizes than it is for accept/reject criteria. C=1 has been the universal standard for quite some time now. It's interesting to note that the statistical integrity of the MIL standard (or any OC curve) is wholly dependent upon random sampling, which is almost never done.

AIAG mandates X-Bar & Range as the default expectation, with a minimum of 25 subgroups and 100 data points. I/MR should be used only in cases where X-bar/R is clearly inappropriate.

Absolutely agree when submitting PPAP and original capability studies. My answer is for ongoing process monitoring on control charts. Sorry if I was unclear.

Dave

I do have on my bookshelf a Ford publication titled "Continuing Process Control and Process Capability Improvement" published by the "Corporate Quality Education and Training Center Corporate Quality Office Ford Motor Company" December 1987. My father sold castings to Ford in the late 80's and was required to learn the book. For xbar-R, it recommends "These data are reported in small subgroups of constant size, usually including from 2 to 5 consecutive pices, with subgroups taken periodically (e.g., once every 15 minutes, twice per shift, etc)."

It is a good publication it appears, with many practical examples.

HI all,

Thanks for the info. I did confuse things by talking about continual process monitoring and determining the Initial Process Study.

So what I should do for my Initial Process Study, is to take 100 samples in 25 subgroups and plot this on a X-bar R chart? Each subgroup obviously consiting of 4 samples!

Using this you can establish that you have a stable/predictable process or not and determine that this meets the required expectations and will continue to do so.

Longer term data should be collected to determine longer term capability and also facilitate monitoring the process to ensure no special causes of variation become apparent.

One more thing, can anyone suggest the best way of approaching the longer term study. Up until now we have not done this. Data collection intervals, length of time it should be conducted over, samples sizes etc....

:thanx:

HI all,

Thanks for the info. I did confuse things by talking about continual process monitoring and determining the Initial Process Study.

I may be wrong, but given the amount of data you have (which in the general scheme of things is a lot), I don't see much mechanical difference between the two. Perhaps you may narrow down the number of parameters (and thus the number of charts) when you shift to continual process monitoring, but the techniques are fundamentally the same.