Emerson’s Mark Coughran has been busy sharing his process control expertise lately. His latest article, Improve Batch Reactor temperature control, appears in the June issue of Chemical Processing magazine.
Mark describes three batch reactor temperature control cases with split-range control configurations. The first case involves control valves to hot and cold headers on the reactor jacket. The second case involves control valves to steam and chilled-water heat exchangers and the final case involves a control valve on the chilled fluid and variable electric heater.
You’ll see common advice in the posts where Mark is featured. In this article, he summarizes this advice into five recommended steps on how you should approach loop tuning:
- Make the process dynamics as linear as possible.
- Minimize dead time.
- Measure the process dynamics.
- Choose the right controller algorithm to compensate for the process dynamics.
- Tune for the speed required, without oscillation.
Proper selection and sizing of control valves and minimizing non-linearities in control strategies such as dead zones in split-range control help to address the first point. For a batch reactor, the jacket heating and cooling responses may be very different. One way to mitigate this difference is to use a controller, which supports gain scheduling to provide separate tuning parameters for the cooling and heating steps.
Dead time (the time delay from an output change to a change in the process variable) can occur in the transport delay of heating/cooling media from the control valve into the jacket. Circulating pumps and jacket-temperature sensor location can help reduce this cause of dead time. Also, filters applied to the temperature transmitters will appear as dead time to the control loop. Mark counsels that you allow one overshoot on the jacket-temperature setpoint response to get the fastest linear response and to minimize dead time.
For measuring the process dynamics for integrating (those that ramp at various slopes on a change in output), processes like reactor temperatures are easily determined from step tests with the loop in manual mode. The proportional + integral + derivative (PID) controller compensates for these process dynamics. Proportional action is mainly used for integrating processes. Some derivative action may be needed on the reactor temperature controller but usually not for the jacket controller.
Mark recommends the Lambda tuning method to tune for the speed required without oscillation. Start with the jacket (slave) control loop first. It must be faster than the reactor (master) control loop per the rule of cascade tuning. For processes with significant nonlinearities, fuzzy logic control might work better.
As he concludes in the article, the benefits of getting this tuning right is improved product quality, reduced batch cycle time and reduced energy usage and waste.
