I think I am trying to tune an impossible system

[allscott]

I am trying to tune a run out old water heating system for the screw of an extruder. I have come to the conclusion that the water flow has simply made extremely difficult if not next to impossible if you are not Peter Natcheway.

Here is the history.

The extruder is 25 years old. The original heating/cooling loop was controlled by a Eurother EM2 and it worked fine. The system consists of a water loop with an electric heater, this water runs through a heat exchanger that has a solenoid valve that is controlled for cooling. All heating/cooling is on/off control.

The EM2 was replaced with a clx utilizing a pide loop and a srtp instruction for the heating and cooling. The Engineer that started to design this system quit before he was done so it was never completed properly, however it worked.

So now I am thrust in to a system that wasn't completed and doesn't currently work with a limited knowledge of heating and cooling systems. My knowledge is limited simmply because every other one I have ever worked with has been simple to make work.

My issue is that whatever I do with the gains and how conservative I make them whenever the extruder starts to actually pump plastic the temperature starts to swing and it starts ringing and eventually gets out of control. This particular loop doesn't seem to ever require cooling so I have shut it right off with limited success.

I know that if a process is oscillating I should lower the P gain and that works to get the extruder up to temperature slowly but when it starts to run everything just starts oscillating.

Not understanding how the PIDE works totally and not understanding what it was doing when I tried to tune it manually (ziegler nichols) I decided to install a dedicated temperature controller (Eurotherm 2408). After an unsuccessful autotune I tried to tune it manually the best I know how with no success.

I'm convinced the problem is with the flow of water through the system. There just seems to be way much time between when the heater comes on and I see a reaction and when that happens things just swing, ring and get worse.

I know the first piece of advice I will get is to post some graphs but just with the setup I have that is hard to do. I am displaying graphs on a red lion but with the SRTP instruction they don't make a bunch of sense. I will post whatever I can if it helps.

 

[iant]

the water heating cooling process is designed to assist the constant control of the heating of the Screw and Product.
the increase of temperature by the zone heaters can be slow to be seen by the Process controllers.
so the need to cool the zones back to a known temperature is important.
for this reason the water temperature is critical.
the Zone controllers open the water valve for a predetermined time knowing that X seconds will cool the die by Y degrees.
If the water temperature is fluctuating greatly then the heating process will be out of control and the Parrison will vary greatly in length.

so when you want the water to be at 19 degrees it needs to be with in
+/- 0.05 or what ever the mfg specs are

 

[rdrast]

Are you saying that you only have electric heaters on the water loop itself? and not the barrel?
If so, have you tried putting at TC at the output of the heat exchanger, and just controlling that temperature (the water loop only)? I've done that with oil jacketed extruders with no issues.

If you have separate barrel heaters with an integral water jacket for cooling, you will have to play around mostly with the parameters on the SRTP instructions as the cooling and heating loops have wildly different characteristics. On extruders like that, I generally have the heating pretty linear from 51 to 100% from the SRTP, but the cooling (from 49 to 0 % SRTP) is actually scaled to hit maximum on time at about 35 to 40%.

For virtually all of our extruders, heating is only needed to initially get the barrel up to temperature. Once the screw starts, only cooling is generally needed since the heat comes entirely from the melt in the barrel.

 

[Peter Nachtwey]

I'm convinced the problem is with the flow of water through the system. There just seems to be way much time between when the heater comes on and I see a reaction and when that happens things just swing, ring and get worse.

Your instincts are right. What you are probably fighting is what is called dead time.

I know the first piece of advice I will get is to post some graphs but just with the setup I have that is hard to do.

You know what you have to do.

The problem with some systems is that the dead time is relatively long compared to the system time constants. You are right about reducing the P gain but even that may not help. These systems can still be control but it takes techniques like a Smith Predictor.
This can be programmed in PLCs but it won't be fun.

 

[allscott]

Are you saying that you only have electric heaters on the water loop itself? and not the barrel?
If so, have you tried putting at TC at the output of the heat exchanger, and just controlling that temperature (the water loop only)? I've done that with oil jacketed extruders with no issues.

If you have separate barrel heaters with an integral water jacket for cooling, you will have to play around mostly with the parameters on the SRTP instructions as the cooling and heating loops have wildly different characteristics. On extruders like that, I generally have the heating pretty linear from 51 to 100% from the SRTP, but the cooling (from 49 to 0 % SRTP) is actually scaled to hit maximum on time at about 35 to 40%.

For virtually all of our extruders, heating is only needed to initially get the barrel up to temperature. Once the screw starts, only cooling is generally needed since the heat comes entirely from the melt in the barrel.

Thanks for the replies.

There are heating and cooling loops on the barrel as well (5 of them), they are working fine. The problem I am having is on the "screw cooling" loop. The screw is hollow and there is a pipe through the middle of it that shoots water to the end of the screw and then it flows back around the outside of the pipe.

This particular extruder and product (XLPE) has a target screw temperature of 100C so in actuality there is never any cooling needed, just heat.

As for process variation I can live with control that is +/- 1deg The problem is what I am getting is about a 30 deg swing.

Peter I will get you some graphs. I'm not real familiar with the graphing functions in Logix 5000 but I'll figure that out and post.

 

[Peter Nachtwey]

Here is an example of how to tune a SOPDT system with a long dead time

SOPDT=second order plus dead time.
In this example the dead time is 6 minutes which is much longer than the two plant time constants of 0.5 and 2 minutes. This system would be very difficult to tune with just a PID. The P gain would need to be so low that the system wouldn't correct well for errors at all.

The Smith Predictor requires a model of the plant. You must first find K, tau1, tau2 and the dead time theta p. I have a program that can do that here:
http://deltamotion.com/peter/Scilab/AutoTune/

The idea is that the model is the same as the plant only without the dead time. The control signal from the PID goes to both the plant and the model at the same time. The model will react instantly to the control signals and it is the model PV that you use as feed back. The model PVs must be stuck in a queue that shifts every update. The queue must be long enough to hold the Model PV for each scan for as long as the dead time so if the PID updates every second and the dead time is 3 minutes there must be a queue of 180 entries.

The problem is that the PV estimated by the model and the actual PV will start to diverge so the estimated PV calculated 1 dead time ago must be compared with the current actual PV and the estimated PV is corrected.

Skip the gobbletygook and go directly to page 10/10. You can see how the PV responds about 180 seconds after the control output changes but the estimated value responds instantly.
http://deltamotion.com/peter/Mathcad/SOPDT/Mathcad - SOPDT SP.pdf

 

[allscott]

Problem solved

So as I suspected the problem was a flow issue. Unbeknownst to me the pipe that goes in the center of the screw was just replaced and they made it too long. The pipe was close to the end of the screw so the water pressure went up and the flow went down causing my system to be virtually uncontrollable with just a PID.

It was an interesting lesson although a very frustrating and extremely expensive one.

Thanks for the help and Peter I will read through the information you provided when I have a moment. I am kind of sick of temperature control at the moment.