MSA - How do I perform a Stability Study as a part of Measurement Systems Analysis?

We are a company producing metal pressings and plastic injected moulded components to 2nd tier Automotive customers. As part of our QS9K audit in May I have been asked to implement stability studies alongside R&R studies, linearity studies (calibration) which are already in place.
I have started by taking three components produced in production, measuring each one ten times and recording the results (on a control plan feature). I have then calculated the sum, average, bias and range for each subgroup.
I have then plotted two graphs, the first being a linearity plot control chart & range chart. From this I have calculated the estimated Standard Deviation using a constant (d2) for sample size of 3. In the Ford MSA book it states that you must look for trends on the control chart as there is no need to calculate a measurement system stability number. Does anyone know if what I am doing will suffice ?? Your comments will be appreciated.

Neilm,

It sounds similar to a method we have successfully implemented.

For a given fixture we have a calibrated master. Each Monday we take the master and verify (reset) the gage. We then measure the master five times and plot it on an X-R chart with a subgroup of five. Over time we have a picture of both the stability of the gage and also correlate it with any wear on the master, which is used to determine the calibration frequency of the master.

Short and sweet stability with added benefits.

ASD...

My scenario is different as you have a calibrated master to work to and I only have a part that could measure differently depending on what method used

[D.Scott]

Without a calibrated master, the MSA recommends the following (paraphrased): use a production part that falls in the mid-range of the production requirements. Further, they recommend 3 samples from the low end, high end and mid-range and do seperate control charts for each. Then on a periodic basis (daily, weekly), measure the sample 3 to 5 times. Plot the data on an X-bar & R control chart. Establish your controls and evaluate for out of control or unstable conditions per the normal chart. Compute the S/D for the measurements and compare it with the process to determine if the system repeatability is suitable for the application. - Reading through your method I don't see measurements being taken over a period of time. I am a little confused when you add linerarity and bias to the stability study. The stability can stand on it's own while the others are best calculated on the 10 parts used in the actual R&R (just my opinion here - certainly not an expert on this).
Dave

Thanks for your comments D.Scott
It is not possible for me to get 3 samples from the low end, high end and mid range of the tolerance as we produce these parts from a tool which has been cut to a fixed size. And therefore will measure what it measures (if thast makes sense!) Also is it important that we check parts so often as weekly?

Also the parts we produce are often in metal which will only creep or shrink minimal amounts, which wouldnt even be picked up by the accuracy of normal measuring instruments. I.e vernier, caliper etc

[D.Scott]

You are losing me Neil - every process has variability and that variability can be measured if you use the right equipment. If your current gages don't pick it up, you are probably using the wrong gage. If the variance is minimal and insignifficant, why are you measuring at all?
You should be able to identify hi low mid parts just by isolating them when you find them through normal measurements.
Yes, the stability should be tracked at the outside weekly, although I don't know of any "rule" that says so. A longer time would introduce too many other possible variables to make me feel happy about it. We normally do daily stability checks on our Tensometer & Micro-Vu.
From what you have said, you may have to reconsider the measurement system you are using. Again - not my field so only an opinion.
Dave