Part 3 of the saga of verifying the measurement of steel strip width on a continuous pickle line. (See the preceding two posts)
Once the first Gage R&R study results were in, we reported to our management. Not wanting to leave any stone unturned, they suggested that we repeat the Gage R&R study using only one tape measure.
The idea was to separate out the individual measuring devices as a source of significant variation. It seemed a little like overkill and unnecessary to me, and could produce friction among the plant staff. But the order was given.
Back to the plant guys and a new Gage R&R test was arranged.
It was conducted like the previous one and the only difference was that only one tape measure, selected at random from the available devices, was used. The “campers” were unhappy, but never took it personally and cooperated fully.
(We like to think that won them over with a positive attitude, big smiles and the two dozen doughnuts that we brought to the test!)
The results were statistically indistinguishable from the first study.
Thank goodness for the power of the Gage R&R method! (W. Edwards Deming would have been proud.) Of course, in both studies, it was easy to find and eliminate the “fliers” due to either misreading of the tape or writing down a different value. They occurred in both sets of tests, and were obvious.
One obvious improvement in the verification process would be to reduce the likelihood of such errors.
That resulted in a flurry of meetings: What to do? A problem had been found that showed a significant contribution to our inability to know if we were in statistical control of the width trimming process.
(This, BTW, turned out to be a much smaller width uncertainty problem than one detected a little later in slabs coming from the Caster to the Hot Strip Mill – but that’s another story….perhaps later)
We needed a better measurement method to verify that the online, continuous width monitor was correct!
Practices that had been standard since mill operation began decades earlier were now inadequate. They were due, of course, to the tighter tolerance that were being put on the process monitoring and final product specs by the suppliers and competitors.
At least we knew what wouldn’t work.
The hunt was on for something that would. You can’t use a micrometer or vernier caliper to measure the width of something three to five feet wide, or can you?
Turns out the search was short lived. We quickly found a company that made a vernier caliper with digital output that measured to the required width with precision and resolution, typically , Â±0.005″ (Â±0.13 mm) that could be adapted to the plant operating conditions.
It was a custom variant of a standard Mitutoyo Digimatic (TM) vernier caliper. Mitutoyo manufactures vernier calipers with measuring capabilities of up to 80″ (~2000 mm).
It also turned out that many of our competitors and customers were already using the same devices from the same company. One of its obvious features was direct digital readout of the measured value that could be input directly to a Mill computer logging program.
The Mill staff took over. They bought, installed and set up the new device in the plant.
We were not asked to test it or validate its results; but that’s the way things happen in large corporations. Guess there were more unhappy campers, some with red faces, than we knew about.
Their staff, however, now knew how to read and follow the recommended Corporate Gage R&R practices. They could now easily set up and do their own tests, with or without doughnuts and “outsiders”.
It can be hard sometimes to be the bearer of bad news or discoverer of problems.
We helped solve a problem and show first hand how the Mill staff could better use some basic statistical methods to evaluate measurement uncertainty in this and other situations. That was enough of a reward for me.
Long term it didn’t help the corporate profits much because there were other, bigger QA problems in this and other Mills and Plants that helped eat into them. Management’s commitment to SPC seemed to be only lip-service and ultimately contributed, I think, in the failure of the company.
There were market problems, too, without a doubt. Some competitors, however, in similar or even more tenuous financial situations at that time, survived and today are prosperous. They didn’t falter in doing things correctly in measurement QA, SPC and management.