Can someone help me with Control chart selection. I've found some very basic flowchart and guides but nothing with the detail I'm looking for.

To briefly describe our manufacturing process:

Tightly toleranced machined components
Continuous data
Manufacturing cycle time of 1 part per hour
Automated measurement (CMM)
Automated SPC software (QDAS)
Features are a mixture of unilateral and bilateral tolernaces - resulting in both normal and skewed distributions
Options
I MR tends to suit sample sizes of 1 but requires the dataset to be normal otherwise erroneous control limit violations may occur.

Xbar and R tends to be more suited to high volume manufacture

Mov Av / Mr seems to be less common but would appear to be a compromise between the above two and would appear to be the better option for us.

Any advice?

Thanks in advance
Steve

PS I have seen Bob's suggestion for precision machining and whilst I agree with him its not practical to implement as its not an option in the SPC software we use.

PS I have seen Bob's suggestion for precision machining and whilst I agree with him its not practical to implement as its not an option in the SPC software we use.

Yes, this is a leading reason for people having to resort to less appropriate control methodologies. Unfortunately, no SPC software packages are up to date enough to have picked up on this methodology - the claim seems to be that SPC for machining is too much of a niche to bother with.

X-bar-R should never be used for machining. It is the worst possible chart.

However, using I-MR, with control limits set at 75% of the tolerance (you should be able to manually set your control limits in your software), with the R chart calculated as usual is probably you next best bet. BUT - you should get some idea about any roundness or parallelism issues from a capability study. You should reduce the control limits by those errors (as a % of tolerance) to be effective. Be sure to control the process as a sawtooth curve - do not run to the mean! That may be a cultural change, but it is the correct way.

Sorry about being stuck with the software. Join the crowd!