Reactor Cooling Control Approach

[TiCl4]

We have a 8,500 gallon, semi-batch, emulsion polymerization reactor that has an external half-pipe jacket and internal coils, all used for cooling. This reactor may make 50+ different formulations. Leaving out a lot of detail, most batches will load in 10-40% water then pump reactants into the reactor over the course of 2-5 hours. Depending on formulation, pseudo-steady state BTU heat of reaction load (based on inlet reactant flow) can be anywhere from 15,000 BTU/min to 50,000 BTU/min. Process temp setpoint varies from 150-190 F, depending on the formulation. Additionally, cooling tower water varies from 85 F to 35 F, depending on the outside wet bulb temp. On top of that, reactor contents' viscosity may vary from 1-20,000 cP (shear-thinning/thixotropic materials).

Given the large liquid hold-up volume in the jackets/coils and the relatively large volume to heat transfer surface area, the process is very slow (temperature peak-to-peak times are 5-10 minutes when the process is controlling well). Additionally, more heat transfer area becomes available as reactor level increases, but inlet reactant flow rate remains constant. Process response also changes with formulation and with the season because the deltaT available varies dramatically.

To control all of this, one control valve is available, and is PID controlled by cascading off the temp measurement in the reactor. The CV Bias is a parameter that is set for individual batches (may be 30% for one batch and 45% for another batch). Batch quality parameters are set to keep reactants within a +/- 3 F range or so.

I currently have the PID loop (ISA) as a mostly proportional controller with a small amount of integral. I have tried adding derivative to reduce the steady oscillations our controller now has. However, the sensor noise makes the controller very jittery without providing any benefit. I can try having a filter put on the loop reading to reduce noise, but wanted to see if I am going down a rabbit trail here.

Has anyone seen similar processes? What was your approach to temperature control? Was derivative control helpful in that endeavor?

 

[Compositepro]

One way to control reactor temperature is to use a reflux condenser and a vacuum pump. Reduce the pressure in the reactor with the vacuum pump so that water boils at the desired set-point temperature. The jacket would be used only for heating to reaction temperature and the heat of reaction would be removed though boiling of the mixture and refluxing of the condensed water. The pressure in the reactor will be controlled by adding or removing small amounts of nitrogen downstream of the condenser. The temperature then pretty much self-regulates. You just have to make sure you have enough condenser capacity.

 

[Latexman]

I’ve seen similar. A few differences. Separate CVs for coils and jacket . . . and other. One split range controller for staged response. No derivative, but we do have an anticipatory feed forward control from the “heat meter” algorithm. That works wonderful.

I’ve used vacuum and condenser in other technology’s, like phenolic, alkyd, polyester resins and others. Works great. I’ve heard it works with emulsion polymerization because the monomers make good defoamers, but with up to several % of surfactant always present, we’ve never had the guts to try it.

Good Luck,
Latexman

 

[TiCl4]

The vacuum idea is interesting, but I'm not sure it would work very well. Many of our monomers boil at or below water's boiling point, so we'd be refluxing as much (or more) monomer as we would water. We do not have sub-surface reflux, so that would increase the potential for monomer pooling, especially in thicker batches. Also, as Latexman said , I have a feeling this would foam out the reactor.

Latexman - how did you account for varying heat transfer coefficients and fouling factors when using feed-forward control? Given the wide range of viscosities and the fact that the heat-transfer surfaces are very prone to polymer fouling, I'm not sure how a feed-forward controller would work.

Now that I think about it - we do have flow rate, inlet temp, and outlet temp measurements of the cooling water. Is there a way to compare the BTU gain in the cooling water to the expected heat of polymerization from incoming monomer and produce an expected exotherm value in the reactor? E.g. If the cooling water is gaining 40,000 BTU/min, but the incoming monomer stream has a heat of polymerization at 50,000 BTU/min, a feed-forward controller could anticipate the need to increase cooling tower water flow even if the PV is near the SP.

 

[Latexman]

We used atmospheric (no vacuum capability) condensers on vinyl acetate homopolymer and co-polymer emulsion reactors. If you pushed them too hard (refluxing a lot), they would foam up. I've seen that several times.

We don't. Early in the reaction, liquid level is low and hta is low but htc is high (water-like). Late in the reaction, liquid level is high and hta is high but htc is low (viscous). It all works out. Have you noticed when during a batch heat transfer is usually at a minimum? My experience has been it's about 50-75% of the way through the feed time. Fouling is more of a problem though. A steady degradation of htc. We calculate a fouling factor dynamically and predict when to clean so as to not get into trouble.

You might want to just keep track of accumulation on monomer. That's the hazard.

Good Luck,
Latexman

 

[TiCl4]

I did not mention that this reactor is currently being replaced, so I have added some instrumentation like cooling water temps/flows as well as a reactor level device in the project in order to have some better process visibility and control. The old reactor only had temperature measurement and control valve output, so I didn't have much to work with. With these new instruments I should have more options to do much of what you said.

Thank you both for the input.

 

[Latexman]

Cool. I've put in, or been involved in the installation of, over a dozen emulsion reactor lines in my 41 year career. A DCS or individual loops?

Good Luck,
Latexman

 

[TiCl4]

A single PLC controls the reactor train system. The site is small, so it does not warrant a full-time controls engineer. We have an electrical/controls engineer we contract out when we need controls work/programming done. I was glad when they finally got an instrument tech a few years ago...