We are looking at two different methods of an exothermic batch control for DeltaV. The plant likes the current heat control, but the way it's tuned, its pretty much on/off control.

I am looking at one method, another engineer has another preference... and I'm sure there's many more. Any books out there on Exothermic control? I can find minor details in other control books, but nothing real solid.

Can anyone help distinguish between the following?

1. Cascade/Split Range Control: Master controller looks at Reactor Temperature. Output of Master controller goes to splitter. One side goes to the slave controller for the cooling valve. The other side goes to the slave controller for the heating valve (hot oil). The split will not be 50/50 because the process gain for heating is much higher than cooling, so the plan is somewhere around 0-30% output for cooling and 31-100 for heating (but we can play with it).

The heating slave controller has the Hot Oil temperature input. The cooling valve controller has the cooling water outlet temperature input.

2. Separate controllers: Two seperate controllers. One for the heating valve with the reactor temperature input and one for the cooling valve with reactor temperature input. There is also a 'deadband' controller so when the setpoint is, say 300 F, and the temperature increases to 305, the cooling controller will start to come on, and when the temp drops to 295, the heating controller will come on.

This method, it is said, will be better because since the process gains are different, the tuning parameters of the heating controller and cooling controller can be different and better control can be had using two seperate controllers.

Any thoughts on the two options would greatly be appreciated. Any other ideas would be welcome, but I also like to live under the KISS method....

 

This is a very interesting topic. We built a pilot plant using a Siemens PCS7 and encountered the same issue with our reactor temperature controllers. Our solution was to change tuning constants using a sequential function chart when heating and cooling. This scheme provided an acceptable level of control but was by no means ideal.

 

I am working on a problem like this right now at a plant in Akron, OH. I think, at a minimum, the idea of two sets of control parameters is a good idea but I am going at it a bit differently in this plant and think it will work out well. What I do is to analyze the data from 40-50 batches run in a given set of reactors that are being run on a PID loop control scheme. The plant has been messing around trying to get good control this way for about 15 months, by the way, and the control still stinks. What I have done with the data is to model the values for UoAdT based on temperature rise in the reactor during the non-exothermic part of the temperature range and compare that with the temperature rise in the reactor, again during the non-exothermic part of the reactor heat up.

Basically, what I did was figure what UoA was for this reactor by looking at how fast the reactor heated up, knowing what the weight of the reactor shell and the composition and quanitiy of what was in the reactor and what the heat capacity of all that stuff was. Hopefully, you know what I'm talking about.

I fooled with the mCpdeltT in the reactor with the UoAdeltaT term for the reactor until the two compared fairly well, meaning that I had a pretty good idea what the jacket was capable of in terms of heat addition and removal for a given reactor temperature and jacket temperature.

With this data I am now in a position to calculate what I am calling the "apparent exothermic heat output" of the reaction contents in real time. The value of the exothermic heat output is then the difference between the mCpdT term and the UoAdT term. In words, if the reactor heated up 3 degrees in the last 5 minutes and that calculates to a heat input of 500,000 BTU/hr but the jacket temperature is actually COOLER than the reaction mass and calculates to having REMOVED 500,000 btu/hr, then there is about 1,000,000 btu/hr being evolved by the reaction mass.

These equations are not difficult and what I'm going to do is to keep a real time output on the control screen and in the control logic that is always telling the operations people as well as the controller, itself, what the current apparent exothermic heat output of the reaction mass is. I know the capability of the jacket and if the exotherm gets to a scary level, I'm going to put cooling on high and get us out of the problem. Also, I can use this data to improve control. I intend to apply it just before the reactor reaches temperature and the exotherm is starting to significantly kick in.

If the reactor is, say, just coming up to termperature and the temperature is still rising fairly fast, this means the jacket is still too warm. I am going to take control away from the PID loop at 10 C below the reactor setpoint and then start using the apparent exotherm to calculate the setpoint for the jacket that will slow the exotherm at an ideal rate as the setpoint is approached. I intend not to try to stop the heat up 10 C below the setpoint, obviously, because if I did, it wouldn't reach setpoint. Instead, I have written an equation to sort of linearly stop the heat up between that 10 C below setpoint temperature and the setpoint temperature, itself. After the temperature is stabilized, I kick the control back to the PID parameters which work well for maintenance but not for catching the exotherm on heatup.

What I've described is a simple model-based control that has worked well for me. Quality and safety can both be improved by catching the exotherm at the setpoint and keeping the temperature from badly overshooting.

I haven't gone into all the details on this (it got to be a longer answer than I thought it would be!) but suffice it to say that the more exothermic a reaction is, the harder it is to catch at the setpoint. Some charge recipes can be caught with a simple PID loop, some (as I've had the scary experience to witness) run away.

I hope this was helpful.

Wes Bullock
330-374-3408

 

Yes, you need separate gains. In the SMAR CD600 single station controller and the SYSTEM302 we either use the scheme of two controllers and a selector switch. Or, the simpler method is to use adaptive gains (a.k.a. gain schedule). Based on deviation a different set of gains is chosen. When deviation positive it uses on set of gains, when deviation is negative it uses another. You can also program the gains based on external variables. See the function block libraries here:
http://www.smar.com/products/ld301.asp
http://www.smar.com/system302/

Jonas Berge
SMAR
===========
[email protected]
www.smar.com
Learn fieldbus at your own pace: www.isa.org/fieldbuses

 

Thanks Novartis:

We do almost the same thing at an existing plant that is non-exothermic. However, for an exothermic reactor, it's more difficult to do a sequential function during exotherm. Heating/cooling control has to automatically 'flow' between heating and cooling to maintain the temperature. I'm not sure if we can change the tuning parameters while the controller is operating. For example, when the output is cooling (<30%) switch the tuning for cooling, and when the output is >30% switch the tuning for heating.

 

Wes: I'm out of Columbus, OH working on a new facility overseas - so don't have any historical data to analyze. Also, I refer you to my self-assigned user name to note that I am new to the control field (not fresh out of school though), so this UoAdT/mCpdeltT thing is Greek to me.

Oh, also, I found another website that has a better layout and is more easily managed. It is eng-tips.com. For control issues, you'll have to do a search on "control", but the format of that site is much better. I've posted my question their as well. I think you'd be an asset to that site.

 

Normally temp M /temp S Cascade loops are always a funny situations. We had similar problem but we solved it only on process point of view. we changed the split range for both the valves. 0-65/35 % rather than 50/50 which is of normal convention.

Look much on process point of view rather than conventional approach & decide the split tange.

Seperate gains for heat & cool services. We have done this conditionally & automatically.

If you can bypass Slave temp loop, please do the on top priority bases.

Jari