Greg: A 19% decrease in cycle time from the application of model predictive control (MPC) to a batch distillation column motivated us enough to take a break from our retirement bliss to explore batch distillation control opportunities. Basically, the MPC setup used a mid column temperature as the controlled variable. The manipulated variables were steam flow and a ratio of take-off flow to total flow. The MPC could maximize steam flow as an optimization variable. The constraint variables were vent pressure and temperature.
Stan: We could have journeyed to the hills of Tennessee where the real expertise is still kicking in batch distillation but we might be a bit conspicuous in our sandals and control freak t-shirts. We thought about posing as the Dukes of Hazard but we felt any discussions on model predictive control might take unintended turns.
Greg: Instead we are going ask Wendy Kramer and Mark Sowell who have improved the control of batch distillation applications in their plant and Control Hall of Fame inductee Terry Tolliver who in my book is the worlds most proficient expert in industry on distillation column control.
Stan: In batch distillation, has the predominant opportunity been increasing capacity rather than reducing energy use?
Terry: Probably true if the capacity has value (i.e. sold out). This is done by maximizing steam against constraints such as reboiler heat transfer (steam valve position), tray flooding or entrainment (column differential pressure), or condenser duty (column pressure).
Wendy and Mark: Yes, but in one process the reaction step is limiting, not distillation.
Greg: Does one generally try to optimize the reflux to steam ratio?
Terry: Yes, although I'm more used to expressing it as a reflux to distillate ratio.
Wendy and Mark: Steam is quickly ramped up to a max flow. The temperature difference between key points in the column (Delta T) is used to set a reflux to distillate (take-off) flow ratio. Reflux starts out low to increase capacity. As the delta T lowers, the reflux ratio is increased to squeeze out the remaining product. The reflux ratio calculation we use is actually the reflux flow divided by the sum of the distillate and reflux flow.
Greg: Can a feedback temperature controller with a fixed or sequenced set point inherently do the optimization or does it need help?
Wendy and Mark: In an alcohol recovery distillation a cascade control system is used. Here we taking a lower side draw to waste. The temperature controller output is the set point of a reflux to distillate ratio controller to keep the appropriate alcohol concentration at the top of the column while we try to minimize the amount of alcohol in the lower side draw. In a product distillation, the batch sequence sets the reflux ratio. Here we separate the different product cuts overhead in a simple batch distillation, the temperature profile is changing as the batch progresses and the lighter boiling components are removed from the system. We increase the reflux ratio based on column delta T to maximize the pure product cuts (i.e. knock down the next boiling compound with additional reflux).
Terry: The time optimal path for a given cut always lies between constant overhead composition (as inferred from a fixed temperature) and a fixed reflux ratio (with the temperature rising to an end point). This can be achieved by sequencing the set point to either a temperature controller or a reflux ratio controller. The sequence may change from batch to batch and may be defined if the batch feed composition is known.
Stan: If you want to increase capacity, would the control system maximize steam and manipulate the reflux to steam ratio? For an overhead product as the batch goes from start to finish does the reflux to steam ratio generally go from a minimum to maximum?
Terry: Yes and yes.
Wendy and Mark: Yes, but we manipulate a reflux to distillate flow ratio.
Greg: When do weeping or flooding or other column constraints enter into the optimization?
Terry: Since you are maximizing steam, flooding may always be a consideration and may vary over multiple cuts due to the molecular weight of the batch material changing. Often adjusting the column pressure between cuts is used for this purpose, generally going to lower pressure as the batch proceeds.
Stan: How would you do multiple cuts?
Terry: Treat each cut with its own optimum path. Dropping the level in the reflux drum at the end of the cut may further help.
Wendy and Mark: A batch sequence is monitoring an online analyzer and when the overheads concentration reaches a set value, XV valves are closed and opened to change the cut to a different product tank.
Greg: Any things to watch out for?
Terry: Entrainment can sometimes be a problem which may be difficult to infer. Try calculating the vapor rate as the sum of the reflux flow plus distillate flow.
Wendy and Mark: We go to great lengths to minimize the volume in the reflux drum so we minimize the volume of the shoulder cuts. A shoulder cut is the material in between the main cuts that has to be blended off. We remember when we first developed a batch distillation process. Being continuous distillation guys we designed the reflux controls where the reflux drum level controller was cascaded to the distillate flow controller and the reflux flow controller set point was ratioed to the distillate flow. We quickly got into problems because during batch distillation the distillate flow would sometimes be zero and this would blow up the reflux to distillate ratio calculation. We had to switch everything over to a reflux to reflux plus distillate basis. This gives the sequence the latitude to adjust the ratio all the way up to 100% (total reflux). We take the output from the reflux drum level controller, send it through the desired ratio calc and then generate independent remote set points for the reflux and distillate flow controllers.
Greg: We conclude with a list of the Top Ten Signs you have a Dysfunctional MPC Team.
Top Ten Signs You Have
a Dysfunctional MPC Team
- The recommended sizes of controllers range from 0x0 to 100x100
- The models for the first controller fill up the hard drive
- The model after 4 months of testing looks suspiciously like the model from the first test
- The completion of the project is tied to the Second Coming
- Food fights break out in the cafeteria over matrix design
- Meetings kick off with kick boxing between consultants
- More than one consultant onsite at a time is ruled a health hazard
- A psychiatrist is chosen as the best possible project manager
- The project over runs its Prozac budget
- The creators of South Park request movie rights to the project