Dear Sirs,

I was hoping to get some advice on batch reactor temperature control using more conventional methods. A huge exothermic reaction is causing a significant and unpredictable disturbance in the reactor temperature profile which i would like to control quite tightly.

 

Hi all,

I suppose that what you mean by "conventional" refer to PID control. If that the case, you must first undestand that a PID controller is a second order abstraction of a non-linear system. This means that the capability of this kind of control for an exothermic process depend a lot on the nature of the chemical reaction itself (the level of non-linearity of the chemical reaction). If this the reaction is a first degree (following the Arh=E9nius law) and the reactor use a direct heating / cooling (or both) system, it could be possible to found the right parameters for the controler. If overall non-linearity of the system is higher than 3 (the chemical reaction and the reactor) which is often the case, a "standard" PID controler will be not enough (you will need more complex algorithm to control the process).

Terence, in all the cases, i need more information on the process / system to be more precise in my answer. Please feel free to contact me (directly or through the list).

Best regards

Steve Monnet ([email protected])

 

Hi Steve,

Thanks for your comments thus far.

I did in fact mean "PID" from the term conventional- ...extended to something more adaptive perhaps like gain scheduling etc.

Due to not having a strong process background this is my poor excuse for a description of the system:
Various reactants are loaded into a reactor and they are charged to a point. The catalyst is introduced and the exothermic reaction takes place raising the reactor temperature well above its desired setpoint. (the profile of this disturbance is impulse-like)

The outer walls of the reactor are heated by an electrical heater and cooled by a cooling medium inside cooling coils.
Temperature cooling is achieved manipulating the flow of the cooling medium and ideally this should negate the effects of the exothermic reaction and keep the temperature around its setpoint without overshooting or undershooting too far.
Since this is being done in a laboratory environment the exothermic reaction varies with each batch so ideally an adaptive strategy, which takes into account the profile of the reaction, should do the trick.

Kind Regards

Terence

 

Although not falling under the category of "conventional" control, you should check out Brainwave (www.brainwave.com). It is pretty simple to set up, and the control of batch reactors is one of its prime applications.

Roger

> Hi Steve,
>
> Thanks for your comments thus far.
>
> I did in fact mean "PID" from the term conventional- ...extended to something more adaptive perhaps like gain scheduling etc.
>
> Due to not having a strong process background this is my poor excuse for a description of the system:
> Various reactants are loaded into a reactor and they are charged to a point. The catalyst is introduced and the exothermic reaction takes place raising the reactor temperature well above its desired setpoint. (the profile of this disturbance is impulse-like)
>
> The outer walls of the reactor are heated by an electrical heater and cooled by a cooling medium inside cooling coils.
> Temperature cooling is achieved manipulating the flow of the cooling medium and ideally this should negate the effects of the exothermic reaction and keep the temperature around its setpoint without overshooting or undershooting too far.
> Since this is being done in a laboratory environment the exothermic reaction varies with each batch so ideally an adaptive strategy, which takes into account the profile of the reaction, should do the trick.
>
> Kind Regards
>
> Terence

 

The standard configuration is a cascade with two PID's. You can find it on almost every reactor - despite the huge differences in e.g. a pure batch or a fed batch operation and esp. in the key operating objectives - very tight or rather loose temperature control, easy or difficult to follow temperature profiles, need to minimize batch time or not etc. Besides, all process parameters (static and dynamic) change dramatically over the batch run.

Both factors contribute to the generally poor performance of the controllers. Furthermore, the results of poor operation can be dramatic, off-spec product being the minimal risk: We have seen cases where the material in the reactor was baked solid and had to be cut out piece by piece.

Therefore we recommend model based controllers like our AMC controller, that is a very easy to implement and runs in every DCS. In one case we have even developed and tuned a complete feedback-feedforward control scheme in one single day.

More info is on our Web site under "Technology"

H.E.
ACT
www.act-control.com

 

I have no idea if I am responding to this thread correctly or not, but here goes ...

The best solution is the use of a PID controller with batch preload capabilities. The preload is a setting that shifts the PBand so the proportional control action occurs before temperature reaches set point. This eliminates the overshoot. I don't know what DCS platform you are using, or if it has this batch preload feature or not, but it was specifically designed for this application. My experience is only with the Foxboro systems, and they always had it.

If you need further help, reply to me off-line ([email protected]) as I simply stumbled across this web site and will probably not return.

 

How fast (approximately) is everything happening?
- When you start the chemical reaction, how fast does the temperature increase? and (approximately) how much?
- If you make a step response on the cooling flow, how long time does it take until you see a obvious response? (ie I do not mean one small pixel of change on a huge scale)
- how fast is the temperature changing after that at maximum flow?

Let's look at the numbers...
If you have cooling enough to handle the rise in temperature (I suppose so, but I have after all seen too may examples of where obvious things like that is missed)

Are the cooling ciruit fast enough, ie can the temperature begin to change fast enough when
we start to see a difference? (chemical reaction slower than cooling system dead time by a factor 3-5 or more)

If you have a no on any of those questions you can not handle the situation with a PID, and would probably have problems with other controllers too.

If the step response from opening the cooling flow looks approximately the same in all cases you don't really have to use anything more advanced than a PID (assuming yes on the other questions of course)

You can of course use a feedforward signal realtively close to the typical rection curve to
lower the control error, even if it don't match it would probably match better than nothing.

/Johan Bengtsson

Do you need education in the area of automation?
----------------------------------------
P&L, Innovation in training
Box 252, S-281 23 H=E4ssleholm SWEDEN
Tel: +46 451 49 460, Fax: +46 451 89 833
E-mail: [email protected]
Internet: http://www.pol.se/
----------------------------------------