Need some advice on a slightly different control loop

[alvinstewart]

I've got a project coming up whe a proportional valve controld a hydraulic cylinder. The cylinder controld the control the window size of a slidind gate that allows varying amounts of material to discharge. The Output is a proportional valve with 12ma. closed, 12+m.a. increases flow through the valve in one direction and 12-ma. does the same in the opposite direction. The PV is a level sensor sensing the level of material in the hopper containing the material. The SP is the desired level to maintain in that hopper. The hydraulic cylinder will have no position feedback of the stroke. In the perfect world, the cylinder would open to create acertain size window for the discharge and then close at 12ma. The PV will vary based on the amount of material entering the hopper via a conveyor and being stratified to varying densities via vibration or pulsing air over the material in a water bath. Since the valve must output to a certain position, close, and settle for the next calculation and make the appropriate adjustment, the output will be going to a closed state 12ma. until it needs to make an output adjustment based on the PV. Can anyone offer some advice here. Any help or direction is greatly appreciated. A similar machine uses a rubber paddle drived at varying speeds controls the dischagre. The output is always at some level being sent to the VFD. That one is easy and easy to tune. Thanks again for any help here.

 

[Peter Nachtwey]

Loops within loops within loops

What is happening is that you must have a valve that takes 4-20ma that is used as a pilot for a bigger valve moving the cylinder. The difference between the control signal to the smaller valve and 12ma is effectively integrated. The magnitude of the difference is the rate at which the big valve will accelerate. This integration will make the control tricky. On top of that there is an outer loop for the level control.

First, there are simpler ways to do this using a valve that responds directly to the magnitude of the control signal instead of the extra integration stage. Also, feedback on the cylinder would be a big help.

Since this system has a built in integrator I doubt one will be needed for your outer loop control. However, you will need a derivative term on the level. The derivative term will be important to avoid overshooting.

One issue I see is how perfect is this 12ma offset. If the 12 ma offset isn't perfect then another integrator may be required and that is bad because the system already has too many.

What PLC are you using?

Whoever designed this did not do it with control in mind. I wouldn't wish this project on anyone that needs to ask. This would be tricky even for me, not because I couldn't work out the correct solution in theory, but because the real system $uck$. I would worry about getting a good derivative value from a level sensor.

 

[alvinstewart]

Thank you for your reply. I inherited the job as another version of an existing system (as mentioned with a vfd directly controlled by the loop), just with some subtle mechanical changes and a different PLC. The PLC is an Automationdirect P3000. I expected using a cascaded loop and realize having some position control on the cylinder travel itself would be of help. I'm curious to see what other replies come forth.

 

[baldone]

Without having a cascade loop with feedback on the cylinder position it's going to be very difficult to get things to work very well and to handle process upsets.

Going to have some anti-windup programming to do as well.

 

[shooter]

? are there 2 coils on the hydraulic valve
You should be able to use only one PID controller.
what are signals used for input and for output?
make a small drawing
P is very small all the action is done with the integrator:
if too low open valve
if too high close valve
output is -12 mA at 0 % and +12 at 100%.

 

[ndzied1]

I think a big question is how fast is all of this happening?
How much lag is there in the level sensor?
How much lag in the valve response?

 

[alvinstewart]

To clarify............
The process is somewhat proprietary per customer request.
However, it is not a fast response process. The hopper will contain material of varying densities. The feedback has proven reliable in measuring this. The clylinder is varying the size of the discharge window. The generic question (to safeguard the process description) is the fact that the PV is the level of material and using a variable window opening to keep the level within a fairly tight value, albiet not a fast process. the concerns I have is notknowing where the cylinder position is and that the hydraulic flow is zero in iether direction (flow off) at 12ma. due to the proportional valve. Below 12ma. is moving in one direction and above 12ma. is moving in the other direction. I was expecting to use a cascaded loop. My concern is I will not know where the cylinder is and once a position is output, where is the cylinder travel and getting the valve back to stop until the next sample/calculation. I realize it is challenging given the parameters I have to work with. That's why I threw it out there. One response indicated concern that I was having to ask about this if I tackling it. I was simply seeing what ideas I might get. I've been involved in industrial automation since 1979. I just have done less process than digital control and this is an unusual situation. thanks again for any constructive ideas and comments.

 

[Peter Nachtwey]

Yes, watch the lags.

Now it is clear how the valve woks.
Level control is an integrating system. That is 1 integrator
The velocity of the cylinder is integrated to get position. This adds another integrator
Each integrator adds 90 degrees of phase lag.
the system has 180 degrees of phase lag.
The valve will have some phase lag too. This makes it worse yet.
If the valve isn't perfectly null an integrator may be required to adjust for the valve null. This adds yet another integrator.
When the open loop system has more than 180 degrees of phase lag and the gain is above 1 the system will oscillate. I see too many lags (integrators ). Derivative gains will help add some lead back into the system but that assumes the level feedback isn't too noisy and needs a filter ( more lag ).

As it is Alvin may get by with just a little P gain but the response will not be as fast as what could be achieved if it were designed properly. If the level signal is clean then use the derivative gain to add some phase lead.

There isn't much one can do. The designers thought they would save a buck by removing the feedback from the cylinder. If there is feed back the inner loop can be programmed so that the actual position follows the reference position from the outer ( level ) pid very closely and the problem is gone.

 

[alvinstewart]

I have just found out thatthe hydraulic cylinder DOES have a position feedback device (Balluff)to indicate the shaft position. I have done the other piece of equipment that uses a variable rotating discharge. It is easy. thank you again for everyone's help and advice. I'm hoping the hydraulic position indication will make life easier. The porportional valve also has some feedback information I did not know about and still do not. Thanks again, everyone.

 

[alvinstewart]

I just receive the proportional valve from the customer today. It has a feedback signal but I haven't investigaged it. So I'm thinking between knowing where the hydraulic cyclinder is positioned and possibly the proportional feedback, I should be able to use a cascaded loop (the valivebeing the innerloop)and make this work.