3 term control for temperature loop 2

[nedwardous]

I'm working on a gas fired heater using a fixed air flow, fixed process gas flow and variable fuel gas flow which is controlling the exit temperature of the process gas.

The default PID settings are Gain 50, Integral Time 5 Derivative rate 0. The manufacturers state that the heater is designed for a higher process flow than at present, however, these still seem pretty rare values. The fuel gas control valve is a ball valve. The exit temperature oscillates around 10 degrees centigrade each side of the set point (300), almost like 2 step control.

Could it be that what it says is the Gain is actually the Prop Band? Plus, how do I know if the Integral time is in repeats/min or mins/repeat?

Any ideas apart from trial and error on this one??

 

[Skogsgurra]

Isn't there a manual? Or a manufacturer?

Gunnar Englund

http://www.gke.org

 

[FrancisL]

It could be that the Ball valve is sticky, or oversized, but there are many possible reasons. I suggest tuning the loop manually, the defaults are probably wrong for your application.

 

[khan101]

I guess you should forget about the default parameters setting and try to retune the controller. Initially, reduce the gain, turnoff derivative, and increase the integration time setting. Then gradually decrease integral time, then slowly increase gain, this way you will learn more than any book can teach you.

You can also use Ziegler Nicholls, and can then adjust in a more scientific manner. If you don't know what it is, just do a web search and you will find a good description of it.

 

[ScottyUK]

From the tiny amount of information you've given it is impossible to give a definitive answer. We only have what you tell us, and you have told us virtually nothing about this controller. Assuming you didn't build it yourself, presumably it was made by one of the established manufacturers such as Moore, Eurotherm, Foxboro, etc. If you tell us the manufacturer and model we might be able to help rather than fill this thread with futile speculation. Even better, you could contact the manufacturer and ask if they have a manual.



----------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[FrancisL]

Who made the controller really makes not difference. It's the application that matters

 

[Skogsgurra]

FrancisL,

The OP asked: "Could it be that what it says is the Gain is actually the Prop Band? Plus, how do I know if the Integral time is in repeats/min or mins/repeat?"

To be able to anser these questions, it is necessary to know manufacturer - and perhaps also model. P-band or Gain? Reset or integration time? It's anyone's guess as long as you do not know what controller you are talking about.

It really does matter, you see.

Gunnar Englund

http://www.gke.org

 

[FrancisL]

Gunnar
I don't agree at all. Gain is the reciprocal of Proportional band, regardless of who made the controller, (likewise Reset versus integral).
True there are differences in PID algorithms used by differing suppliers, but these are relatively minor.
Anyway, if the loop is managing to control, albeit with a 10 Degree oscillation, then it is working to a degree and I think it would not do this if the gain was as drastically wrong as being the reciprocal of what it should be.
khan101 is right about needing to tune the controller.
However even if it were perfectly tuned, such problems as stiction in the valve could still cause oscillation.
Francis

http://www.controldraw.co.uk

 

[Skogsgurra]

Yes, you do agree. (Even if you do not realise it).

The OP specifically asked how he could know if it is Gain or Proportional Band etc that has been set on the controller. To know that, you have to know how this particular controller manufacturer usually does it.

The fact that PB = 1/Gain is text-book knowledge and the OP also shows that he understands that.

So, I still maintain that a direct contact with the manufaturer is the best way to go. Or find the manual. On the web or in a book-shelf.

Gunnar Englund

http://www.gke.org

 

[Skogsgurra]

I have to add this: I agree that the problem usually is in the plant and not in the controller. And the valve stiction is a very probable cause for the +/- ten degree deviations. Absolutely.

Gunnar Englund

http://www.gke.org

 

[FrancisL]

I don't agree that I agree
But I think I know what you mean, ie that it might be worth seeing the controller manual to know what they mean by gain - except if they mean PB it would normally say so on the physical (or software) controller Faceplate.
Cheers

 

[Skogsgurra]

;-)

Gunnar Englund

http://www.gke.org

 

[jsummerfield]

You can find PID tutorials on the web. I found this one

http://www.jashaw.com/pid/tutorial/pid6.html.

If the controller measures the temperature of a product exiting the furnace and controls the fuel valve you need to use some derivative. Other more complicated schemes accommodate differing fuel composition, varying flow rate for the product, etc. Check Harold Wade's site

http://www.wadeco.com/principa.htm.

He published a good book and has a great simulator.

If you have a characterized ball valve you would have much better control than possible with a standard API 6D type ball valve.



John

 

[CARF]

Dear all,

This is a good book on Parctical Process Control:

http://home.hccnet.nl/m.dijk/pressure_drop_calculator/ftp/practicalprocesscontrol.pdf

I suggest to create some step responses of the system and derive the proper P and I values from the curve. Yes that's famous Ziechler & Nichols approach.

Cheers,
CARF

http://www.carf-engineering.com

 

[JimCasey]

Nobody has mentioned that this is a temperature control loop and they are notoriously impossible to tune due to the time lags involved. There is a time lag between changing the process temperature and getting it to the temperature sensing probe, and there is also the thermal resistance of the thermowell and probe itself. Control overshoots in a temp loop are practically inevitable when using a simple feedback control loop.

Using a feedforward loop will resolve the problem: Measure the parameter that varies that causes you to have to modulate the fuel flow, then control the signal to the fuel flow valve as a function of this variable. In your description you said that the fuel flow is the ONLY variable, but if this was true your valve would never have to modulate unless the fuel composition was changing.