Multi-PID Interactions.. Clues requested

Gentle People.. Imagine...

You have a Very Long Pipe (oh, 800 feet), of relatively large diameter (say 6 inches ID) following what is basically a pure catenary....

Now, imagine you are going to charge that pipe with steam... up to oh, 400 PSI live steam.

Now, pretend you are going to back that steam up with some high pressure intert gas (at the top of the tube)

Now, pretend that you also have high-pressure water being injected into that pipe at the very bottom of it....

Now, pretend you want to control the actual level of the water in the pipe (which is on the catenary) *this pre-supposes you can actually MEASURE the water level* For water, we have pumps that can develop 650 to 800 PSI.

Now, pretend that there is a high-pressure seal on the other (top) end of the catenary... which in all cases must always be exactly 50 PSI higher then the steam pressure in the tube...

Now, pretend that the high pressure water pumps can develop 600 to 800 PSI, and must have a bypass system to keep the absolute output pressure (delivered to the tube) at 50 to 75 PSI over the steam pressure, but need to have a controllable bypass valve to dump the excess water back to the cooling towers....

Now, pretend that there are valves, near the steam/water interface in the tube that bleed off water to keep it at a certain temperature, but there MUST be actual WATER (NOT steam) flow...

How does one create a PID/PI/P/Anything system that can properly regulate the WATER LEVEL (And, for bonus points, how does one accuratly measure water level) in that kind of system?

Primary Setpoint is Steam Pressure.
Secondary Setpoint is (Top-Of-Tube-equizing pressure, 6 feet of 4 inch ID tube)
Secondary Setpoint is (Badly Sensed Water level In (6 inch Tube)
Tertiary Setpoint is (Water Pressure Delivered by high pressure pumps) (ed: does anyone else feel that 'pressure' should only have one 's'?)

The primary control, is on Steam pressure, however, the entire process will fail if the 1st secondary setpoint fails, or the tertiary setpoint fails.

The water level is critical. Steam PSI is the basis for everything... Top PSI is dependent on steam, but doesn't bleed off well. Water level is dependent on Water Pumps PSI > Steam PSI (to a point).

I see 4, interactive PID loops... Steam being the 'master control' of all the others. How would you make them all respond in a predictable, non-oscillatory fashion?

umm...how about a fuzzy solution?

And to measure the water level, can you weigh the entire pipe?

Define Cantenary

This is what I found.

http://mathworld.wolfram.com/Catenary.html

How is it oriented? Is the curve pointing down with the ends up or the other way around? Could it be sideways?

This is a good one. Before knowing all this does not sound like a application for four PIDs. You need a model.

Who ya gonna call ...? Smoke and Mirrors Inc!!!

Can't we just pretend it all works perfectly?

umm...how about a fuzzy solution?

Click to expand...

VERY!!!!!

A PID + fuzzy solution is what is in place now, and it is workable, but I know there is a better approach.

Basic picture is here:


The Steam, backer gas, water, and water bypass are all controlled by proportional valves (2" on steam/water, 1/4" in backer).

The biggest problem is that a very slight change in the water injection valve sends a shockwave up the tube, which interferes with the level sensors, and is also reflected in the steam and backer pressure controls. Severe dampening (averaging) on the level feedback is used because of this, which causes a long lag in achieving (and holding) a level.

The backer gas is also very critical, and while it's a small volume, it must be maintained at 25 PSI below the steam pressure.

Every time I look at these controls, I can't help but feel I'm staring at a bunch of springs.

Current methods of control include full PLC control, as well as mechanical, pneumatic, and discrete electronic controllers. Only the PLC controllers have setpoint coordination.

Setpoints:
Water Level (Operator Entered).
Steam Pressure (Operator Entered).
Water Supply Pressure (Actual Steam PSI in Tube + 50)
Backer Gas Pressure (Actual Steam PSI in Tube - 25)

Not shown in the picture, but present, is an adjustable bleed valve right at the steam/water interface level which drains out superheated water to keep the water leg cool.

Setpoints:
Water Level (Operator Entered).
Steam Pressure (Operator Entered).
Water Supply Pressure (Actual Steam PSI in Tube + 50)
Backer Gas Pressure (Actual Steam PSI in Tube - 25)

Click to expand...

It looks to me as if there are really three separate loops here.

The two most important items being controlled are Level and Steam Pressure.

The Steam pressure transmitter will feed the steam controller which will manipulate the steam valve.

The water level transmitter (which I have no idea on how to get a more accurate measure on) will feed a loop which will cascade to set the injection pressure of the water.

I think the setting of the injection pressure at fixed differential to the steam pressure is causing some of the oscillation. The injection pressure should be the manipulated variable in the level control loop.

A differential pressure between the steam and backer gas will feed a controller to manipulate the backer gas pressure. Depending on the seal between the gas and the steam this loop will have to be kind of loose to keep it from oscillating with the steam pressure.

Thanks for the picture – I was having trouble visualizing from your description. I still don’t have any clue what the utility of this process is.



Anyway, I zeroed in on your comments about how a minor change in water valve introduces a big disturbance in the process. If you don’t understand why this is the case, you should focus some effort here. Here are some thoughts/questions along those lines.

1) Is the water preheated to near the saturation temperature of the steam? If not, when additional water is introduced, the steam will condense to heat it and the steam pressure will fall; this disturbance will propagate through the system. So, if you don’t have it, consider a water preheater.

2) A water flowmeter used in a level/flow control cascade might eliminate many flow disturbances caused by pressure fluctuations as well as valve nonlinearites.

3) Where many control loops interact, it can be the seemingly well-behaved one that upsets the rest of the system. You probably have already, but rethink the control objectives and prioritize them as necessary. For example, your water line probably has a check valve in it – try detuning the water pressure loop and see if things quiet down.



Well, that’s it for now.



Good luck and keep us informed of your progress

Obviously a mass and energy balance

Tom Jenkins posted a link to an article on how he controls air content in a sewage system. I think this is the kind of approach needed where has a model.

Has anyone looked at this using steam tables, equations etc?

I think jamesau is right about
1. the effect of adding cold water. Do you know the temperature of the injected water?
2 how well are you controlling the injection of water?
3 what do you know about the steam introduced? Temperature, pressure?
4. Ditto, Jamesau comments about the loops interacting.
5. If the backer gas pressure is always less than the steam pressure then the seal or boundry between them would always be moving unless there something else that keeps this boundry from moving.
6. What is the purpose of the backer pressure?
7. Additionally, you indicate the operator sets the set points. I can see where the operator can set set points at where it may be impossible to control.

I think you need to come clean and tell us what and why you are doing this. There may be a better way. You don't want to put a lot of effort into solving the wrong problem. A long time ago I was asked if I were screwing around with the forum. I was but this is the kind the question I would make up. I would then roll on the floor reading the answers.

Since you haven't provided all the info we need I can tell that you don't really understand the problem and are patching or getting by the best you can. If the problem is legit then you are doing well to get this working at all. I think you need to get the forumlas that are used to generate the steam tables. You will need temperature sensors and flow meters. What you are adding to the system is water mass and energy. The ratio of the steam to water regulates the amount of energy per unit of mass. The second concern is keeping the water mass constant. The mass entering the system must equal the mass leaking out. I don't think a normal PID is up to this. This is a non linear multiple input and multiple output system (MIMO). This is master or Phd level stuff on the order of controlling hydraulic cylinder with a position feedback, two pressure feedbacks on either side of the piston and two 3 way valves on either side of the piston. I know I can't have a PID for every sensor, the inputs must be combined some how.

This is truely a thermo-dynamics, mass tranfer and energy flow problem. Once the process is understood then the control will be more obvious.

Where do we go from here? First I don't believe there are many things that are done for the first time. ( Why do rookies think they are the first to have the traffic light problem? ) If this is a legitimate process there will be information on the internet or in books on how this process should be done. That is why we need to know the process. Second, if this is a first then you might ask yourself why? If it is a first it could be that others have found a another ( better, simpler ) way.

Thanks for the plug, Peter!

rdast, I think you are putting some unnecessary constraints on the problem, and introducing some superfluous issues.

I don't see how the gas pressure can be maintained at less than the steam pressure - it seems to me that in a closed vessel the pressure of the two gasses is going to be essentially equal. I also don't see how you can inject the gas into a vessel if the pressure is lower than the steam pressure. First law of thermodynamics issues are rampant.

Measuring water level should be possible with a differential pressure unit. Level = (pressure at bottom - steam pressure) / water density at temperature. The height of the steam column will be a factor, but may be neglected if the height isn't significant. Gauge dampers or digital filtering should help with the sloshing.

I assume this is not a closed system, and that at least one fluid is continuously flowing in and out.

Don't try to control water pressure. If your pump has adequate capacity water will flow in when the valve is opened or the pump started.

It is a bunch of springs. You might consider a surge chamber for the water.

I agree that there are details missing. You really need to develop governing equation(s) for the process and define the really critical parameters to obtain the desired process result.

To answer some questions, yes, this is an actual process, which is curing rubber insulation, and cooling it before exiting the seals at the end of the tube.

The water injection is done at two ports directly in the side of the tube at the bottom. Steam injection is done at two ports at the very top of the tube.

The steam temperature (at running pressure) is approximately 450 degrees F. The cooling water is approximately 45 to 65 degrees F.

The backing gas is used to keep the steam from blowing back through the head of an extruder, so there is a seal of sorts there.

Other notes: RF/Ultrasonic probes are far worse than the Differential cells, at least when mounted at the lower end of the tube.

All control valves, as said, are 2" Leslie types, except for the backing gas, which are 1/4".

The reason the bypass water has to track the steam is so that massive pressure pressure surges do not get introduced into the water leg of the steam tube. It is also required, to always circulate water through the high pressure pump so it doesn't cavitate and self destruct.

Any other questions?

"Now, pretend that there are valves, near the steam/water interface in the tube that bleed off water to keep it at a certain t"

...and what is this? This is more than just a pipe with fittings and seals. I really can't tell what's going on here.

I've found that roughly 80% of problem solving is a proper and complete specification of exactly what the problem is. This is often a difficult and time-consuming undertaking. Sorry I can't offer more than this general experience. At this point, any other advice I could give would be a 'woof and a scratch' at a closed door.

Tom Jenkins said:

It is a bunch of springs.

Click to expand...

I agree on both counts. But I personally see a bunch of resistors, capacitors & inductors.

Tom Jenkins said:

You might consider a surge chamber for the water.

Click to expand...

Absolutely. Or softer modulating valves. Garbage in, garbage out. You can't control what you can't see, and water hammer always kills the readings.

The biggest problem is that a very slight change in the water injection valve sends a shockwave up the tube, which interferes with the level sensors,

Click to expand...

Is there any way you can put an accumulator in the circuit and let it act like a capacitor = shock absorber?

I also don't see how you can inject the gas into a vessel if the pressure is lower than the steam pressure. First law of thermodynamics issues are rampant.

Click to expand...

I agree Tom this ain`t gonna happen!