From June's Chemical Processing:
We have a plug-flow reactor that is cooled by water circulated through a surrounding jacket. The temperature of the exothermic reaction is measured at several places within the reactor, and the highest measurement is used for control. The cooling water flow is co-current to the flow through the reactor. The cooling water inlet temperature is cascaded to the reactor temperature controller. The water inlet temperature is adjusted by an equal percentage valve that allows fresh coolant into the system. Both are proportional-integral-derivative (PID) controllers.
Until recently, we had no trouble controlling the reactor temperature, but the temperature has started to cycle and seems to be getting worse. We have tried tuning both controllers without any improvement. The temperature can't even be controlled manually. Only cutting our production rate seems to reduce the amplitude of the cycle, but this is too costly. We are not operating near the cooling limit because the temperature starts to nose-dive when the control valve is only half open, and we didn't have this cycling in the summer when the valve operated more fully open. Should we invest in model predictive control? Is there another solution to this problem?
Check cooling water quality
In addition to your controllers, you may want to check for simple "physical" things, such as scaling on your heat transfer surfaces and valves. What is the quality of the cooling water? Hard? High organics or bio-count?
Hans G. Schroeder, Ph.D.
International Consultants Association, Encinitas, Calif.
Too much heat extracted
Although model predictive control is all the rage, this problem should not be used to justify it. All the symptoms described are consistent with too much heat being extracted.
A temperature sensor may be reading high or otherwise is not operating properly. This would cause the coolant control valve to overreact and dump too much cooling water to the jacket, resulting in the temperature nose-diving. As the temperature drops, the valve closes and the cycle repeats. In addition, a lower production rate would lower the amplitude of the cycle. Check the calibration and operation of the sensor.
Another issue is that the half-open valve causes overcooling of the process, whereas in the summer the valve was more fully open and performed well. This suggests that more heat now is being removed from the process at a lower coolant flow rate. If the reactor or jacket were cleaned, it would increase the overall heat-transfer coefficient, making the heat removal more efficient. To be sure this is the case, you need to know the age of the system and when it was last cleaned. If the reactor were brought up to the original specifications, restoring the original tuning parameters might solve the problem.
Keith Dackson, engineering scientist
Henkel Technologies, Olean, N.Y.
Use different control scheme
Send fresh cooling water directly to the jacket to control the exotherm. There is no need to control the coolant inlet temperature. I would suggest two independent controls.
First, control the coolant flow based on the highest reactor temperature. The goal is to maintain maximum coolant flow (the control valve should be wide open at all times). To achieve this, control the reactant feed rate based on the reactor temperature. Use separate temperature measurement for both controls and ensure they are installed at the same location in the reactor. The temperature set point for the second controller should be slightly higher than the first one (2 Degrees F-4 Degrees F). This control philosophy maximizes the cooling capacity and optimizes the reactant feed rate for seasonal variation of reactant temperature. For added safety, use interrupt temperature (4 Degrees F-6 Degrees F higher than reaction temperature) to stop the feed if cooling fails.
Vinay Deshmane, senior process engineer
Akzo Nobel Polymer Chemicals LLC, Pasadena, Texas
Unstable reaction rate
The cooling system may not be at fault , reaction rates may be unstable. Make sure the reactor feed composition and flow rate are consistent. Good reactant mixing with a static mixer could help. As a final option, consider recycling product or adding an inert. The non-reactive mass will have a dampening effect on the temperature fluctuations.
Stephen D. Marcum, principal process engineer
Aker Kvaerner, Houston
Predictive control caveat
If your process normally would benefit from model predictive control, then you should purchase it. Do not purchase it only to fix your current problem. It seems you have a process/chemistry issue occurring. The predictive controller only will mask the problem until the model can no longer solve it.
Invest a large amount of resources investigating the basic process and chemistry of the reaction system. It would appear that something in the process has changed. If you were able to control the system a year ago, you should be able to do the same now.
You may want to use Six Sigma methodologies to attack your problem instead of placing a Band-Aid on your system.
Jerome Rister, technical and materials manager
Ashland Chemical, Calumet City, Ill.
Buildup of scale on sensors
The symptoms described may be explained by a buildup of thermally insulating contaminants on the temperature-sensing elements. Normally, these sensors (e.g. thermocouples or PRTs) are in intimate thermal contact with the process stream and have a response time constant. These constants may be ignored when compared with the response time constants of the control elements, in this case the equal percentage valves. However, when an insulating layer of contamination is allowed to build up on the temperature sensors, their time constants can become significant and prevent tuning the PID to stability.
A similar drift toward instability could be caused by a buildup of contaminants on the reactor wall or water jacket. However, this increased thermal impedance could be compensated for by tuning to limit the amount of time the controller spends in saturation.
Barry K. Norling, Ph.D., Division of Biomaterials
University of Texas Health Science Center, San Antonio, Texas
Recirculate cooling water
I have a similar problem with an old pickup truck that has no thermostat. I control the engine temperature with a cardboard air blocker in front of the radiator. In the winter, the engine temperature is much more difficult to control than in the summer since the high delta T makes the system very sensitive to small changes in flow, outside temperature, etc.
Using the lesson from the truck, I would approach the plug flow reactor by trying to recreate the summertime conditions. A recirculation loop in the coolant line could return warm coolant to the inlet, making the whole system slower and less sensitive to swings.
William M. Hall, reliability engineer
BP, Naperville, Ill.
Install an orifice
It sounds like the cooling water inlet valve is either fully open or fully closed. The PID controller adjustments may be inadequate. It also is possible that the cooling water inlet valve is inadequately sized, allowing in too much water or shutting it off completely with no graduation between open and close.
You may want to install a flow orifice in the cooling water inlet to adjust the flow. Alternatively, you can change the valve type to restrict cooling water inlet flow.
Michael K. Ryan, validation specialist
Merck & Co., Barceloneta, Puerto Rico