X bar and R chart calculation for within part variat

Just a suggestion : why do you want to use x bar chart. why dont u use individual chart instead if the average as you say is misleading.

Ramkumar,
If the data is not independent you should not use an X-bar and R chart. Instead, you should consider using the two multivariate control charts provided in Minitab 14. (But you’ll need to perform some preprocessing!)
If you can show the data is independent you could consider using two X-bar and R charts – one for the process mean ( average of all the data) and one for the uniformity. I would not advise you to disregard the process mean even though you’ve stated you’re only interested in the uniformity.
Remember, all the elements in a subgroup should be independent and each element should correpond to a ‘run’ and in the order of manufacture. Do not form a subgroup from measurements taken from one piece!
Good luck!
Andy

Seems like you have 2 characteristics of interest. Part Average, and part range.  A good method of monitoring the process is the use IMR charts on each characteristic. 
Chart 1 : IMR for part average
Chart 2: IMR for part range (I have used this in past jobs. TTV, or total thickness variation is a “within-piece characteristic” that is independent of average part thickness. Different “knobs” control this characteristic).
If you don’t like IMR charts, use rational subgrouping and monitor the same characterisitics.
Part 1, Readings 1, 2, 3    Compute Part Avg, and part range
Part 2, Readings 1, 2, 3    Compute Part Avg, and part range
Part 3, Readings 1, 2, 3    Compute Part Avg, and part range
Part 4, Readings 1, 2, 3    Compute Part Avg, and part range
Part 5, Readings 1, 2, 3   Compute Part Avg, and part range
Now you have 5 averages and 5 ranges for your subgroup.
Do and X-bar/R for part average and an XBar/R for part Range.
Seems confusing at first, but it is effective. 
 
 

Just wondering why he should use multi-vari if he already knows the major source of variability? Within-piece variation is a concern. He should be monitoring the variation appropriately and doing DOEs in parallel to identify the key input factors. Just my humble opinion!

It is amazing how little people understand the basic six sigma tools!
An XBar&R chart represents the voice of the process. Measure the actual dimension in question, sub-group at a scheme and frequency that will gather a representative sample of all the variation in the process (shift to shift, machine to machine, operator to operator etc.) plot the data for a minimum of 25-30 samples and see what the process is telling you!
The individual that mentioned the specifications is correct! Specifications are only a consideration after you have monitored the process and determined how this process is performing. When you have this (UCL & LCL) compare it to the spec to see where you are.
Remeber XBar&R charts are robust to distribution so this is not a worry. The XBar chart will show you group to group variation the RBar chart will show you the within group variation.
Do this and see what the nature of your problem is.

R.M.,
I see your point about exploring all sources of variation, even though “within-piece” has been identified as the most predominant. Many of my variation reduction plans have had differing critical X’s for within-piece, piece-to-piece, etc. variation. Multi-vari is probably one of the most under-utlilized tools in the industry (by people of all sizes, perhaps in Munchkin land too!) :)
We used to preach “Characterize, Optimize, Control” in my early days. Seems like the characterization piece should get a little more focus in this case.
Find out the extent of variation for each source, run DOES to determine critical X’s
Run optimization studies
Control process.
Just my 2 cents!