PID Control - Duty Assist pump

[TConnolly]

Peter,

I really like your solution.

I like Peter's solution as well. I've been watching this thread for ideas. I have a hydraulic press purchased from China. The iron is not bad but the hydraulics really are a POS (and leaked even when brand new), as were the original controls, which have been partially upgraded. A full upgrade is planned. In the meantime I'm trying to work with what I have and improve it without interrupting production.

The system has two fixed displacement hydraulic pumps. Both pumps are on VFDs. The main ram is stroked in and initially pressurized using the larger pump (ramp is PID controlled). When pressure reaches 1600 psi (110 bar) the main pump turns off abruptly and a very small pump on a VFD (also PID controlled) turns on and runs at low speed to hold pressure. I have the two loops tuned so that they perform very well (Thank you Doug Cooper), but the transition from the large pump to the small pump is rough. Anyways, there are some interesting ideas here that I've been considering.

 

[TConnolly]

Right you are. I would always come back with the question "so you want them both to fail at the same time so the whole plant will be down?". No problem.

Of couse, it should go without saying, Bubba still needs to repair the pump when it goes down, not put it off because he has a redundant pump.

Two weeks ago we went down hard, total production stop. The plant central chiller has redundant everything, pumps, compressors, controls, etc. The main compressor went out nearly a year ago. It was running on the backup compressor. Maintenance Boss Cletus knew it was down - but didn't schedule a repair. When the backup compressor went down the plant went down. Redundancy does no good without active maintenance.

 

[Peter Nachtwey]

The system has two fixed displacement hydraulic pumps. Both pumps are on VFDs. The main ram is stroked in and initially pressurized using the larger pump (ramp is PID controlled). When pressure reaches 1600 psi (110 bar) the main pump turns off abruptly and a very small pump on a VFD (also PID controlled) turns on and runs at low speed to hold pressure. I have the two loops tuned so that they perform very well (Thank you Doug Cooper), but the transition from the large pump to the small pump is rough. Anyways, there are some interesting ideas here that I've been considering.

Do you need a PID for the big pump? The pressure will increase at a rate proportional the flow into the cylinder ( VFD speed ). I would just use a simple P gain with a output limit proportional to the desired pressure rate. The pump will probably be running at the P gain output limit until it gets close to 1600 psi. When the proportional band is entered the big pump will slow down until 1600 psi is reached. Except for leakage a pressure system is an integrating process. It shouldn't over shoot either if just a P gain is used. Can the small pump be turned on while the big pump is on? If the small pump can be phased in while the big pump is phased out then the transition will be smooth.

A problem with PID blocks is that they constrain your thinking. You think there must be a SP and a PV. This isn't quite right. There only needs to be an error One can set the PV to 0 and use the SP as an error input. So why bother?
Well one can then calculate Error=SP-PV and then divide this error between two PIDs. As the PV approaches the SP the error is gradually shifted from the Big to the small pump PID. This way there is no discontinuity. There are lots of methods for shifting the error. Basically one starts with K=1

BigPumpError=K*Error
SmallPumpError(1-K)*Error

Find a method to calculate K. For starters K=Error/BigPumpPorportionalBand or
K=Error/10 psi
You get the idea.

The good part is that you don't need to change your tuning. The response to error each pumps sees doesn't change.

 

[danieluy]

nice algo peter...

and one thing good about this algo is that once the Error goes big then the BigPump will automatically compesate.. Cool!

But i think the BigPumpProportionalBand should be proportional to the BigPumpSize/SmallPump Size.

I know there are 2 types of implementation
1-1 pump control by PID others left at 100% or 0%
2-All pumps control by PID (hence all pumps change SAME speed..)

Which control is more effective in controlling constant pressure?
Theoretically, it should be the 2nd option but then again i might be wrong.

 

[Peter Nachtwey]

But i think the BigPumpProportionalBand should be proportional to the BigPumpSize/SmallPump Size.

]
I don't understand. How would you change the equation for K?

I know there are 2 types of implementation
1-1 pump control by PID others left at 100% or 0%
2-All pumps control by PID (hence all pumps change SAME speed..)

Which control is more effective in controlling constant pressure?
Theoretically, it should be the 2nd option but then again i might be wrong.

For fine pressure control you want to have only the small pump running when at the set point. There are other options.
One could run both pumps all the way up to the combined proportional band.

CombinedProportionalBand=SmallPumpProportionalBand+BigPumpProportionalBand
 If error > CombinedProportionalBand then
   SmallPumpError=100
    BigPumperror=100
 else if error < CombinedProportionalBand and error > SmallPumpProportionalBand then
   SmallPumpError=100
   BigPumpError=(Error-SmallPumpProportionalBand)/BigPumpProportionalBand
 else
   SmallPumpError=Error/SmallPumpProportionalBand
   BigPumpError=0
 endif

I think I like this better. This method has both bumps running all the way up to the SmallPumpProportionalBand. Near the set point only the small pump is running.

 

[TConnolly]

Thanks for the ideas Peter. You always seem to come through with really good solutions. I'll try it out tomorrow.

 

[danieluy]

I don't understand. How would you change the equation for K?

Opps.. sorry for the misunderstanding... I should have wrote it this way

But i think the BigPumpProportionalBand should be proportional to the SmallPump/BigPump Size RATIO.

this is relating to Peter's orig algo... because i think that the bigger the size diff of the 2 pumps the bigger the proportional band should be.. to avoid overshoot or BUMPS as you call it.

But i think your new solution is better...

cheers! Peter