PID and PID Auto-Tune Issue

[russher]

I have a control system using a Do-More PLC that has an auger feeding grain to a disc mill that grinds the grain. The Do-More monitors the amperage on the disc mill motor (4-20 mA input) and also delivers an analog output (0-10 VDC) to a variable speed drive that is controlling the speed of the auger. I am attempting to use a PID to control the speed of the auger based on the disc mill motor amperage - higher amperage slows the auger; lower amperage speeds up the auger - to maintain a setpoint amperage. I have used this before with great success. On this system, however, there is at least 40 seconds delay between the auger and the flow of the product to the mill motor giving the amperage reaction. This is the first problem - a long delay in the speed control and reaction. The PID Auto-Tune doesn't seem to be very effective on such a long delay. I have tried to counter this with delaying the PID until the disc mill amperage was brought up very close to the setpoint using an initial speed on the auger. I'm finding, however, even if the motor amperage is very close - or even slightly above - the setpoint when the PID kicks in, the PID output goes high, the amperage soars and the mill motor often plugs and overloads. Any ideas would be welcome!

[Garyhlucas]

I think you need a second auger, the long one is a transport auger and the second short one is a feed regulator.  The first could be continuous speed on and off just keeping the second one supplied.

I face a similar issue with feeding waste water into a treatment plant.  When the feed pump kicks on the dissolved oxygen (DO) level drops like a rock and the PID controlling DO overshoots badly. The current project uses a positive displacement feed pump that is on a PID loop to keep the tank level constant, then DO regulates nicely.

[russher]

I don't think a second auger is an option in this application - but I understand what you're saying, and I appreciate your insight. Thanks.

[Controls Guy]

Yeah, that's a tough situation.  I assume the volume between the feed auger and the point where it feeds the mill is somewhat variable, or the feed rate would track the auger speed well.   Here would by my suggestions in order of preference.

1) Reduce or eliminate the dead time if possible.  Probably by moving the CV closer to the process, or getting rid of that feed system volume/rate variability.  One solution might be a cableveyor (tube with a cable in the center with discs moving your product through the tube) discharging near the entry point to the mill.  Then run the cableveyor on the VFD.

2) Put in an instrument to measure the feed rate at the point where the grain enters the mill.  Then you could do a cascade PID with the feed rate as the CV in the mill loop and the auger modulating to maintain feed rate.  This might require removal of some of that elasticity in the feed system between the auger and mill.

3) See if you can write a Smith predictor for your process.

4) Tune the loop slow enough so that it doesn't wind up faster than you get feedback.  Not very good response but if you can't do any of those other things, it's probably the best you can get.

[BobO]

When I developed the FOTD model for the Sim's process simulator, I quickly found that long delay times are nearly impossible to control. My semi-educated guess is that if the delay time is much longer than the response time, you have virtually zero chance of controlling well using PID. Might try using the process simulator in DMSIM to test alternative control schemes.

[russher]

Thanks for the suggestions - all are appreciated.
I did try the PID setup in the Do-More SIM with a .DeadTime of 45 seconds and a .TimeConstant of 60 seconds. After an Auto-Tune, I got values that are radically different than the on-site application Auto-Tune results. Armed with some of these results, I may try, next, a much slower loop and see where that gets me. As pointed out, the response time to transients will also be slow, but in this application that may not be a problem.

[ATU]

If the auger doesn't feed directly up to the mill, what is in-between? Can you post a photo?

[Controls Guy]

Another option: put a bigger motor on the mill so it doesn't stall so readily, and then use a rolling average of the amps as your PV for a slow loop.

[russher]

If the auger doesn't feed directly up to the mill, what is in-between? Can you post a photo?

Sorry, no photo, but we're talking a large (commercial size) feedmill with a number of ingredient bins each holding up to 100,000 kgs of grain. Each of 5 bins have an auger that feeds a grain cleaner. The grain cleaner feeds a transfer auger. The transfer auger feeds the disc mill. This is why the lengthy delay between the bin auger and the mill. Our initial control system used an Allen-Bradley SLC-500 with a PID controlling the speed of the ingredient auger. This was not completely successful either, as I had to cap the ingredient auger speed so as not to overload the mill. I changed out the AB for the Do-more (for a number of reasons) and was hoping for a better result with its PID. The goal here is to load the mill motor to its optimum amperage for 2 reasons - 1) consistent grind and 2) make the best and most efficient use of the mill motor. The mill motor is already 100 hp so increasing the size would not be an option.

[ATU]

Is there any accumulator after the grain cleaner? If there was, you could use level sensing to control your bin augers. Is the final feed auger controlled by the PID with the current sensing from the mill motor?

[russher]

There is no buffer past the bin ingredient auger. Ingredient auger --> grain cleaner --> transfer auger --> disc mill. The bin ingredient auger is on a VFD with speed controlled by PID with feedback from the mill motor amperage - hence the long delay between auger speed adjustment and mill motor amperage reaction. The grain cleaner and transfer auger run full speed all the time.

[plcnut]

I don't think you are going to be able to use an auto-tune. You will probably have to 'hand-tune' this one, or just use a very basic feed-forward method. ie: Have a preset for each product type that gets loaded when the run starts, and then: if PV=50HP, increase the Output by 200 counts, wait 40 seconds (or whatever the update time actually is), and then either increase, decrease, or hold our output.
What kind of accuracy are you looking for? 20amps? 5amps?
How long are your process runs? 20 minutes? 20 hours?

[ATU]

There is no buffer past the bin ingredient auger. Ingredient auger --> grain cleaner --> transfer auger --> disc mill. The bin ingredient auger is on a VFD with speed controlled by PID with feedback from the mill motor amperage - hence the long delay between auger speed adjustment and mill motor amperage reaction. The grain cleaner and transfer auger run full speed all the time.

I have no experience with that type of feed system, but I think if you had any buffer there, you could probably control the final feed auger to get a consistent feed rate to the mill. It wouldn't take much, but I have no concept on how fast the grain is moving through there.   

[russher]

I really appreciate all the feedback on this forum,and I am starting to realize the PID Auto-Tune may not be too effective in this application. The answer may be a "hand-tuned" - very slow - PID or perhaps some form of Ramp and Soak.
This application is a batching system, so a process run is 15 to 20 minutes. Goal accuracy would be +/- 3 amps on a 100 amp setpoint.
I hadn't used the Do-More simulator until now - testing PID loop parameters. In the simulator setup is a .DeadTime setting (the period in seconds before an Output change begins to effect the PV). I was thinking this would be handy in a real-time PID setup.
A buffer within the flow system would certainly be preferred, but would make for substantial hardware changes in the mill at this point. I will have to make the best of the existing system. I have a number of existing similar systems where the delay from auger to mill is substantially shorter, and the PID Auto-Tune works just fine. Typical system throughput is about 3-5 kgs per second.

[plcnut]

This is an important question for the situation you describe:
If you set the feed speed at a specific setpoint (no PID, just a constant speed on the feed auger), how much fluctuation is there in the amp load?