Mechanical side, so bear with me. The Tank 3 level control valve on the mixer line has been chattering for a week now. You can stand next to it and watch the stem go up and down every few seconds. The pump bearings downstream are taking it too.
So I asked the controls guy to pull the trend, Ignition 8.1, and the level and valve output come up as a smooth line, near enough dead straight. He reckons the loop is fine because the trend says so. I showed him a video on my phone and he went and pulled the tag history for that same minute and it still shows nothing. The positioner was replaced last year, so I don't think that's it, though he does.
Is the trend lying or am I? Any ideas what to get him to check before the positioner carries the blame again?
What's the tag and its range, and how often is the history sampling it? Open the tag in the designer, History section, sample mode and the rate.
And the valve output. Is that trended from the PLC output, or from a position feedback?
Level tag is LT301, 0 to 100%. Valve output is the PLC's CV to the positioner, no feedback wired.
History on both says sample mode Periodic, rate 30 seconds, deadband 0.5. Whatever that means for what I'm seeing.
It cannot see it. Your valve goes up and down in about 3 seconds and the historian takes one number every 30 seconds, so it gets one point per ten cycles, lands somewhere different on the wave each time, and draws a straight line through those points. That is not a steady loop. That is a sample rate slower than the oscillation. The trend is not lying, it is blind.
For the tuning job set LT301 and the CV to Periodic at 1 second, or On Change with the deadband at 0, temporarily. The chatter shows inside a minute. Put it back after. 1 second on every tag fills a database fast.
The 0.5 deadband hides it too. Anything under half a percent of range is thrown away before it ever reaches the historian.
Controls guy set both to 1 second for the afternoon. There it is. A sawtooth on both pens, level swinging about 2% and the CV chasing it. He's stopped talking about the positioner.
Good. Before he touches the gain, prove the signal. Loop calibrator on LT301, or just watch the raw mA on the input card while it hunts, because a noisy transmitter with a tight gain will chase a valve about in exactly the same way bad tuning does and the trend looks identical either way. If the mA is steady it is the tuning.
mA was steady. So tuning. He backed the gain off and it settled inside the hour, and I'm not sure he expected that either. Stem doesnt move now, and the pump sounds better already.
So the trend was sampling every 30 seconds on a 3 second swing, with a 0.5 deadband throwing away the small moves on top of that. The loop had been hunting the whole time and the screen couldn't show it. What fixed it was 1 second sampling for long enough to see the thing, then the gain.
He's put the tag back to 30 seconds, which I guess means we won't see it next time either. That bit I don't love.
Thanks marekw and valh7.
Leave LT301 at 1 second at least. One tag won't hurt the database.