Integral windup?????

I have a PLC in a pump station modulating a valve via PID. When the pump station shuts off the valve is supposed to shut. And it does for 3 months without problem, until suddenly I get a huge error number out of nowhere and the valve remains open....

The PLC is a Micro850. It was programmed using connected components workbench. According to the manual the PID block is supposed actively prevent integral windup... I have no idea why its doing this. It works for several months with no problem, then error suddenly jumps to a couple thousand. The units are L/s, typically around 40L/s flow rate. The error is in L/s aswell. The valve is modulated to reflect this error until desired flow rate setput is achieved. I don't get it how it can just jump to several thousand.
I tried phoning Allen Bradley tech support, and of course they gave me some guy who didn't care, was not helpful at all, regurgitated what I already read in the manual and just requested payment for further services.

This happened before and I tried implementing two of the following at the same time:
- Reset the PID block on pump shutdown
- Moving 0 into the error at PLC shutdown

Obviously this didn't do anything.

When the PLC is power cycled, it goes back to normal.

Maybe I can implement a solution to power cycle the PLC in the event that error gets insanely high? lol

Could it be that a slight negative (signed) value is being interpreted as very large because it is expecting an unsigned value?

Possibly, I have had that problem before with other parts of the program, but ive gotten around that by actually having the program itself intervene before that happens. It's really stupid because some function blocks don't let you change the data type.

Except with this one, I don't know for sure because it happens every 3 months and im not there to watch it happen. Also, the error is a REAL data type, so I am not sure going below zero could be the cause. Or maybe it could. This is pretty weird.

It sounds like a possible blip in the feedback to me....

ndzied1 said:

It sounds like a possible blip in the feedback to me....

Click to expand...

A blip in the feed back wouldn't cause the integrator to saturate.

When the valve is told to shut does the valve indicate it gets to the set point? If not there will be a small error and the integrator will windup slowly over time.
It is possible the integrator winds up enough to have an overflow and switch signs but I doubt it because Rockwell seems to be paranoid about overflow errors. Still the integrator can wind up.

Whenever a customer wants to move to a hard stop I recommend that once the actuator gets to the hard stop the controller is placed in open loop or manual control and driven to the hard stop using a small control signal. This way the integrator doesn't windup just because the actuator doesn't get to the set point.

Metal expands and contracts so when moving against a hard stop there can be variation is the stopped position by a few thousandths or micron.

Peter Nachtwey said:

A blip in the feed back wouldn't cause the integrator to saturate.

Click to expand...

I agree but I don't think it has been established that the integrator is winding up. I believe the OP is asking if integrator windup could cause his huge error out of nowhere.

A blip in the feedback would make the error jump instantaneously though.

blackbird307 said:

When the pump station shuts off the valve is supposed to shut. And it does for 3 months without problem, until suddenly I get a huge error number out of nowhere and the valve remains open....

...
When the PLC is power cycled, it goes back to normal.

Maybe I can implement a solution to power cycle the PLC in the event that error gets insanely high? lol

Click to expand...

I would lean more toward adding some code to put the PID in manual mode (or whatever controls and terminology apply to the Micro850 series) and then manipulate the output to make the valve shut.

In the PLC-5 and SLC world we used a method of "bumpless transfer" to and from manual mode to get a PID controlled variable to behave in predictable manner when there was a need to do something outside of its normal operation.

No idea about CCW or anything AB for the most part, but would it be possible to zero out an error when the valve is supposed to shut by switching to a huge negative Setpoint? You could then disable the PID when the error approaches 0...

Just a wild idea.

Another strange thing. When it is working normally the PID is enabled when the pumps starts, and disabled when the pump stops. When it gets disabled the error immediately resets to zero. However, for some reason, this time it has a massive value, enabling it and disabling the PID does nothing, it retains this value. The only way to get rid of it is to power cycle the PLC.

Can you write to the error value? If so, just manually zero it every time the pump stops, or as suggested earlier, whenever it gets way out of range (higher than would ever occur normally)

If you have a negative number and are using integers then the negative number will be read as the MSB being a very large number if you are not careful in your transformations in math.
You may have to put limits on the signal to keep within 15 bits so that the overflow (or negative bit) doesn't come into play.
I have seen similar things when there is an overflow when feeding command signals to VFDs.
The AB (and others) will shut down with the negative signal from an over range calculation.