I am dealing with an interesting situation at my job now and I would like people's opinion on the matter. We are currently utilizing a X-R chart to monitor power wattage for a laser operation. Since we introduce the X-bar chart, the power wattage collected tend to go out of the control limits. The process engineer had spent lots of time looking into the process to identify potential special causes but in vain. So we decided to conduct a gage R&R to see if the out of control condition has anything to do with the gage. The following are the results:

% study var for total rr = 45%
% tolerance for total rr = 12%
number of distinct categories = 2

based on the process engineer, the process is kind of "coarse" in nature so putting the power wattege into x-bar chart scrunity might have been overkilled. based on the results of rr, can i say that the meausrement system is only capable of distinguish the parts into high/low which is equivalent to pass/fail. judging from the fact that %tolerance = 12% so the measurement system is "good" to evaluate the parts against the specifications, do i have sufficient justification to collect pass/fail data and not to use the x-bar chart?
any comments are appreciated. thank you

:biglaugh:

Can anyone help with this issue?

at a % study variation of 47% and NDC = 2 combined with a % Tolerance of 12% I woudl infer that your study parts were well within tolerance...? I could further infer that your process is very capable (assuming that your study parts were representative of the total observed variation of teh process over soem significant period of time - NOT a very safe assumption as many people tend to take tehir R&R study parts from a very limited timeframe of production resulting in a very low study variation)

Without actually understanding how you're process varies over time I couldn't make much of a recommendation except to say that a pass-fail chart will ahve very little sensitivity to drifts and you wouldn't catch a change until it your capability had eroded to the point where you were making out of tolerance parts and you wouldn't even know when teh drift occured to assist in pinpointing the cause...I tend to use pass-fail type SPC charts only for management - rarely, if ever, for controling a process on the line. My recommendation - absent the data - is that you fix your measurement system adn redo the R&R study and recalculate the control limits.

If you improve your R&R but still have a lot of unexplained out-of-control points you shoud lthen look at your subgrouping scheme to determine if it is rational...statistically speaking, not logically thinking.