PID control jacketed tank - tunning

Hi everyone,

Im having some questions regarding how to tune the PID for this application.

It is a jacketed tank, lets say 1000 liters, and its going to be heated with saturated steam. The liquid inside will be similar to water.

This is the first time tunning and implementing a PID.

I was thinking of implementing the ZN method in open loop to obtain initial values to the PID.

However I came with a question/problem. As the tank is closed and no liquid is going out, cant see how to stabilize the system with the valve opened at 50% lets say, so then I step the valve to 55% and see how the system reacts, and then get the constants for PID.

I assume if I have the valve opened at 50% and no liquid is going out, the water temperature will start to raise, until it evapores..

Cant figure out another situation, where a steady scenario might happen

If I open the steam valve, and there is no liquid going out of the tank, the temperature will raise until reaches 100°C, so steping the valve from 50% to 55% wont give me any information..

Am I missing something ??

Thanks!

What are you trying to control? The temperature of a tank of water that starts cold and you warm it up? A Tank that is warm and you are trying to keep it warm? A tank that will have water added to it? At what rate? How close to do have to keep the temperature to set point? Do you have a flow meter on your steam flow? What pressure (or temperature) steam are you using? Is there an agitator in the tank?

If the temperature is stable, and no water is going in or coming out, you won't use much steam. Only enough to match the heat loss from the tank. This tuning is much different from a tank that has water flowing through the tank where heat must be added to heat the incoming water.

Assuming you want to bring the tank contents from a cold to "hot" state and hold it at a particular temperature, you may be well served to simply use an on-off valve for the steam control.The reaction time is likely going to be pretty slow. I experimented using a PID control on a hot water tank equipped with an immersed steam coil. Ultimately, on-off control was the easiest way to control the temperature.

However, maybe I'm assuming incorrectly. Do you have an agitator in the tank or is the contents circulated?

proof said:

What are you trying to control? The temperature of a tank of water that starts cold and you warm it up? A Tank that is warm and you are trying to keep it warm? A tank that will have water added to it? At what rate? How close to do have to keep the temperature to set point? Do you have a flow meter on your steam flow? What pressure (or temperature) steam are you using? Is there an agitator in the tank?

If the temperature is stable, and no water is going in or coming out, you won't use much steam. Only enough to match the heat loss from the tank. This tuning is much different from a tank that has water flowing through the tank where heat must be added to heat the incoming water.

Click to expand...

Im going to control the temperature of the water inside the tank. A mixture will be made inside the tank and once its done, no more water is added. Then I have to control the temperature to a desired SP, which may vary. The temperature must be really close to SP. Dont know yet, but there might be an agitator. The Steam pressure will be around 3 bar.

Once it reaches the SP if the isolation is good, not much steam should be added, I agree, but how do I perform the ZN method. I mean i cant get the system to a steady state, once I open the valve to 50% lets say, the temperature will go up and up, until it reaches 100°C, then if I step to 55% the valve, there wont be any variation.. right ??

How it is suposed to do the tunning here ?? Cant figure it out.


Rguimong: I know on-off control would be much easier, but chief wants a PID control with a modulating valve.

I'm no fan of Ziegler-Nichols but that method doesn't require you to set a steady state controller output and determine the process time constant and gain. With Z-N you simply set the integral and derivative gains to zero and increase the proportional gain until you achieve a steady oscillation. Gain values are calculated from the proportional gain value that causes steady oscillation and the period of the resulting oscillation.

If you want to use one of the model based tuning methods you will need to determine the plant gain, time constant and deadtime. However, there is nothing magic about a 50% output. There must be some output level that results in a steady temperature that is not boiling. Step from something less than this output to this output and determine your plant parameters from that. Also keep in mind that at least your time constant is going to change depending on how full the vessel is. If you are not always going to be controlling a full vessel you may need to operate with more than one set of gains.

Keith

Keith thanks for answering,

The ZN method you are describing is the one in closed loop, the one I refer is the one in open loop, and after knowing the system time constant and other parameters, I can calculate the PID constants.

Which method do you use when tunning a PID ??

ZN is the one i have been taught.. not saying is the best, is the only one I know.

joaco1993
You can use the open loop zn method, just find open loop commands that will allow the temp to stabilize at less than 100C.

Instead of going from 50 to 55%, you may need to go from 25 to 35%.

I am a bit confused as to the exact nature of your process. A basic P&ID would help. That said there are a number of things which immediately come to mind when considering this type of system.
i)The rates of heating and cooling will obviously be very different so you are unlikely to get adequate performance from a single simple PID control.
ii) I would consider having two seperate control valves. One for bringing the temperature up to set point and a smaller valve for maintaining temperature once you are there.
iii) Assuming that the temperature will just climb to 100C is wrong. As the valve opening decreases toward 0% the steam side of the system will fill with condensate until there is very little heat transfer.
iv) Depending upon the design of the system you may need to fit vacuum breakers to permit operation at very small valve openings.
v)Removing condensate may be a problem at small valve openings.

I don't think that you can use Ziegler Nichols to tune this system. Your have two different problems that your are trying to solve, and neither the kind of system Ziegler Nichols will apply. The first problem, raising the temperature to a set point (which uses lots of steam), by definition, you have will not have a swing about a set point, as once you reach it, you are done, so you won't have a 'ultimate gain'. The second problem, maintaining a temperature (which uses little steam) might work, but you'll probably have too much dead time for a good number. Having tank agitation will help reduce this and keep the tank a much more uniform temperature.

joaco1993 said:

I mean i cant get the system to a steady state, once I open the valve to 50% lets say, the temperature will go up and up, until it reaches 100°C, then if I step to 55% the valve, there wont be any variation.. right ??

Click to expand...

what the estimate time opening this valve?

Are this valve run on manual or auto control?

maybe you can deactivate the PID system during valve opening

Proof, then how would you tune this system ??

Suppose there is only water inside the tank, no water coming in, no water going out, and agitator running.

If not with ZN, which other method would you recomend ??

Thanks!

Osmanjdt, the valve opens pretty quick, and runs in auto mode by the PLC.

The reason of the steping from 50 to 55 is because i wanted to use the ZN method open loop to tune the system. But before steping, i must have the system in a steady state, and cant figure out how to do this.

Originally posted by joaco1993:

The reason of the steping from 50 to 55 is because i wanted to use the ZN method open loop to tune the system. But before steping, i must have the system in a steady state, and cant figure out how to do this.

Click to expand...

There is no requirement in any of the step test based tuning methods that says the step needs to be between 50% and 55%. You seem to be hung on those numbers. Is this because the actuator being controlled is a dual function valve that will heat above 50% and cool below 50%? If not, don't get hung up with 50%. The only requirements are that, given a specific manual command, the system is "stable" and is near the design level of operation. Furthermore, the output step size needs to be big enough that you can easily discern the process response from the general system "noise". For our purposes stable means the temperature has reach equilibrium given the output you are sending. If this steady state output is 5% and the step size is 2%, so be it. You will need to determine what the starting output and step size are by experimentation or by experience.

Originally posted by joaco1993:

If not with ZN, which other method would you recomend ??

Click to expand...

Many of us don't like ZN because is results in more aggressive tuning that we want. In addition the gain correlations were arrived at largely through trial and error. That tends to make the correlations very system specific, which is what proof was getting at. But in the end, all the tuning rules are really just calculations or correlations based on information you pull from a step test.

I would tend to recommend looking here for information:

http://controlguru.com/

Specifically you might want to look here:

http://controlguru.com/pi-control-of-the-heat-exchanger/

The tuning correlations are based on a control concept referred to as Internal Model Control. These correlations are based on the process model parameters yo get from a step test but tend to be somewhat more forgiving of the actual process model deviating from the expected model that is the basis of the tuning correlations.

Finally, if the steady state output of your controller is VERY low when the temperature is at the normal setpoint you will probably have better luck treating this as an integrating process as opposed to a self-regulating process. That will result in different gain correlations again.

Keith

joaco1993 said:

Proof, then how would you tune this system ??

Click to expand...

I would probably tune it through trial and error, or if I had the time, money, and information needed, I'd model the system and calculate the tuning. My first step would be to figure out how I would control it if I had to control it by hand, then try to create a system that would automate that control.

The first step is to heat the tank. For this I'd open the steam valve to a preset position, try to determine the dead time between when the valve opens and when the tank temperature starts to climb. I'd wait until the temperature got close to the desired temperature, and shut the steam valve. I'd then see how much it continues to rise, and how long it takes, and see if it falls back down (which it will probably do if the tank is agitated). On subsequent runs after reaching a pre target temperature based on the initial overshoot results, I'd switch from a large valve opening to a smaller valve opening so I could creep up to the target temperature.

Once this first step gives reasonable results I'd switch to control over to a maintain temperature tuning. This should only need to provide enough steam to maintain the temperature, so I'd run it with a slow acting loop with upper and (and maybe lower) limits on the steam valve using PI control that starts from a value that matches to normal average steam needs. You might be able to use ZN for this. I would probably adjust the P until I see it swinging then cut in half, and then set the I reset time to 3 times the period of the swing. If this wasn't good enough, then I'd try to model the system.

One problem with this setup, if the same valve is used for the big steam needed to heat the tank is used for holding temp, the small valve opening for that control is out of the 20-80% range and gives poor results. In this case, the PID might have to use some sort of On/Off control, i.e., opening the valve 10% for 10 seconds, then close for 90 seconds instead of opening 1% continuously.

Thanks everyone for answering,

I will first try to stabilize the system by opnening the valve at low %, so as the temperature raise to 70°C. Then I will give a step open to the valve and see how it acts, and start playing with parameters of the PID,

Proof, there wont be 2 valves, there will be just one modulant valve and a PT100 for temperature,

If this is a batch process where the liquid is not continuously leaving the tank, then I believe this is an integrating system. If the liquid is not leaving the tank, then the temperature depends on the rate of heat transfer going into the tank versus the rate going out of the tank. If the rate coming in is greater than that going out, the temperature of the tank should rise at a near constant slope until the liquid boils.

These two articles should tell you everything you need to know how to perform the bump test, calculate the system parameters, and then calculate the PID gains.

http://controlguru.com/analyzing-pumped-tank-dynamics-with-a-fopdt-integrating-model/

http://controlguru.com/a-design-and-tuning-recipe-for-integrating-processes/

I recently went through these steps to tune seven of my plant's integrating processes and had a lot of success with it. We went with the PI controller due to the noise in the input signal. Note the PID gains assume a dependent PID.