The classical GR&R studies the variability of the measuring device separate from the variability from the human operator. With an automated test station, the influence of the human operator (the textbook definition of “reproducibility”) is meaningless if the operator does nothing but press the cycle start button.
However, the GR&R approach can still provide valuable information, for example, what portion of variability is due to the instrument (electronics may have drift or non-linearities, even with no moving parts) versus the part loading cycle (which often involves mechanical motion, and a single part is not positioned in precisely the same position with repeated loading). You can call the part loading cycle “reproducibility” even though untouched by human hands. Many times, the automatic test station in setup mode can retest the part 3 times without unclamping the part. Running repeat measurements without unclamping does not randomize the measurement order, so if there are bias effects over time such as changes in ambient temperature, your analysis might conclude spurious effects. Repeating the automatic test in a different order would be one way to detect confounding of bias-over-time.
If you wanted to study whether the automated station gives the same result on 3 separate days (or 3 different hours, if you suspect ambient temperature is a factor) you could run the same set of parts in pre-determined random order at different times, and call that “reproducibility”. What you hypothesize is each part will test the same way. If not, the difference will be highlighted once you examine the results graphically.
However, the GR&R approach can still provide valuable information, for example, what portion of variability is due to the instrument (electronics may have drift or non-linearities, even with no moving parts) versus the part loading cycle (which often involves mechanical motion, and a single part is not positioned in precisely the same position with repeated loading). You can call the part loading cycle “reproducibility” even though untouched by human hands. Many times, the automatic test station in setup mode can retest the part 3 times without unclamping the part. Running repeat measurements without unclamping does not randomize the measurement order, so if there are bias effects over time such as changes in ambient temperature, your analysis might conclude spurious effects. Repeating the automatic test in a different order would be one way to detect confounding of bias-over-time.
If you wanted to study whether the automated station gives the same result on 3 separate days (or 3 different hours, if you suspect ambient temperature is a factor) you could run the same set of parts in pre-determined random order at different times, and call that “reproducibility”. What you hypothesize is each part will test the same way. If not, the difference will be highlighted once you examine the results graphically.