No, I am not going to deliver any sermon on why R&R should be done or how it is different from calibration.
You are quite right. There is an issue with destructive measurement systems because parts cannot be tested repeatedly. The MSA manual also is rather ambiguous, esp. if you try to read the section IV of the third edition. However, there are accepted methods of calculating GR&R for such systems. As with any R&R study, it is very important that parts for the study are selected carefully. With certain assumptions about 'homogeneity' of sample parts, Nested ANOVA is usually recommended. Due to the nature of the test, some of the part-to-part variation is always confounded with other components of variation and there is no way you can avoid it. This risk can be minimzed (but not completely eliminated) with proper selection of parts. The basic purpose is to find out if there are any 'undesirable' errors creeping into the measurement due to factors such as appraisers and methods etc - apart from the equipment used. You can then make an attempt to minimize these errors and the risk associated with a fauly measurement system. A carefully done R&R study may well "throw up a surprise or two" and may provide opportunities for improvement. To that extent, it is a waste of efforts.
That said, here are two thoughts that may be worth discussing:
1. Have there been instances when the prduct you okayed was rejected by the customer based on the same test/measurement? Is there a chance of obtaining a fauly measurement even when using a well calibrated equipment? In such case what is the risk that you or your customer can accept? If the measurement system is such that the opertors, the method used or the setup required for testing do not influence the measurement, you can probably argue that calibration is enough. However, you will have to prove that there is no appraiser variation in the system by conducting at least a sample R&R study. Thereafter you can use the Stability Study to monitor the staus of the measurement system.
2. If your process has been stable, highly capable and centered over a long period of time (seen from historical SPC data/charts), then, according to AIAG, there may be no need to study your measurement error from a purely "acceptability" viewpoint. Many times it not economical to try to improve the measurement system beyond a certain point. You can use this measurement system for the purpose of monitoring the process - status quo - only. You will also have to ascertain the ability of the measurement system to detect any special cause of variation in the process.
The bottom line, as usual, is: discuss with your customer. There is no question of faith in the holy book, the standard. IMHO, you can try it out - there's nothing to lose.
One good reference is: "Evaluating the Measurement Process" by Donald J. Wheeler.
You may also try to investigate Shainin's ISOPlot method which some people recommend for destructive Testing (Sorry, I do not know much about it).
