Greg: In this column, we look at distillation control with Vivek R. Dabholkar, an independent advanced process control (APC) consultant who previously offered his expertise in “Achieving plantwide, multivariable control”.
Vivek, what’s your assessment of distillation control to help prepare us for a deep dive?
Vivek: Automation/control systems engineers implementing controls in chemical plants and refineries, or those stuck with classroom vapor liquid equilibrium (VLE) thermodynamics and McCabe-Thiele method in distillation, may find this refreshing and useful. Simply put, distillation operation requires application of heat via reboiler duty to separate lighter components from the heavier. Separation is based on difference in boiling points, more accurately relative volatilities. An increase in reboiler duty, while driving lighter components toward the top, carries the lightest of the heavier components (heavy key) towards the top, leading to adverse effects on top-product composition.
Corrective action is taken by pushing lighter components via reflux flow to washdown heavier components carried upward by an increase in reboiler duty. Increased reflux introduces lighter material in the bottom, which is not desirable for maintaining bottom-product specs. Without the temperature controller in the bottoms section, reflux cleans the top but dirties the bottoms, while the reboiler does the opposite. You can see this tug-of-war between the reboiler and the reflux actions—one trying to clean up one end, while dirtying the other end.
This highlights the importance of the temperature controller in the bottom or top section of the distillation column. In the case of bottoms-section Tray-TC, any increase in reflux is automatically countered by just the right amount of increased reboiler duty to hold Tray-TC at its setpoint, preventing dirtying the bottoms to a large extent. If the tower pressure is controlled, increased reflux flow and reboiler duty cleans the bottoms product impurity instead of increasing it. Further refinement is possible for a change in operating pressure by using pressure-compensated temperature. If pressure goes up, then pressure-compensated temperature must go down, and calls for increased reboiler duty because volatilities get worse at higher pressure differences. An engineer expects reflux and reboiling requirements to increase, unless the column is in vapor-flooding regime. The setpoint of Tray-TC can be trimmed to maintain bottoms composition at the spec limit.
Greg: What are your thoughts on artificial intelligence (AI)-based control design for distillation columns?
Vivek: I am encouraged by advances in AI, but with respect to logic-based reasoning, AI is still in its infancy. Logic-based reasoning may be a reality in the future, however. If such reasoning can be auto-learned and recommendations provided by the AI system, then it will be revolutionary.
The know-how consists of logically piecing together relevant information. I envision a future where AI can come up with the best design for a regulatory process control scheme by using a process flow diagram (PFD) with flow rates, temperature profile and feed composition. I recommend every aspiring control engineer to mark PFDs with tag-names and operating conditions. P&IDs are too detailed for this effort.
Another case to consider is a C3-splitter with a high reflux-to-feed ratio and small bottoms flow. Knowing the feed contains propane and propylene with close boiling points and a nearly flat temperature profile across the column, an experienced engineer can immediately rule out implementing a temperature controller either in the top or the bottom section. With Tray-TC out of the picture, an experienced engineer can come up with bottoms-level control cascaded to reboiler duty as opposed to cascaded to the bottoms flow. Large moves in reflux flow keep top product composition in check and can’t possibly be counter balanced by bottoms flow (in the absence of a temperature controller) due to widely different relative magnitude of the bottoms flow in relation to the reflux flow.
Greg: What are the peculiarities of distillation column control dynamics where the bottoms level is cascaded to reboiler duty?
Vivek: In this configuration, the increase in feed regardless of its state (vapor, liquid or two-phase) ends up in the reflux drum. In case of vapor feed, a large portion ends up in the reflux drum, but any fraction that goes down the column increases reboiler duty to keep bottoms level constant.This uses the same reasoning as when net change in ramp rate of reflux drum level with respect to reflux is zero. It can only affect it dynamically during initial drop, until the time reflux makes it to the bottom-level transmitter. This is because all liquid reflux that goes down the column must come up as vapor due to increased reboiler duty to keep bottoms level constant. It’s similar for a change in bottoms flow. Any decrease (increase) in bottoms flow must result in an increase (decrease) of reflux drum level ramp rate because an increase (decrease) in reboiler duty or decrease (increase) in bottoms flow holds the bottoms level constant. This is an extreme case that should be compared to bottom section Tray-TC in control.
In the case of Tray-TC control, a portion of feed that boils at a temperature higher than the Tray-TC setpoint ends up in the bottoms product, even though the reflux comes up as a vapor stream due to temperature control in the bottoms section. The ultimate destination of this incremental vapor depends on the method of pressure control. If the pressure control is done by adjusting the vapor distillate flow (or an expander inlet valve), then a large portion may become vapor-distillate flow, leading to a net change in ramp rate for reflux drum level due to an increase/decrease of reflux flow. In the case of a total condenser, all vapor ends up in the reflux drum. Therefore, the net change of reflux drum level ramp rate (apart from initial dynamic response) on an increase of reflux will be zero, just like in the case of C3-splitter in the bottoms level-to-reboiler duty case study.
Greg: If feed increase ends up in the reflux drum, how is material balance control achieved in this configuration (bottoms level cascaded to reboiler duty)?
Vivek: You’re correct that there’s no direct material balance control. However, if the feed is increased and bottoms propane flow is held constant, overhead propane in the propylene product will increase and the bottoms will clean up (less propylene in propane) due to increased reboiler duty in case of liquid feed. The reflux drum level can be balanced by 1:1 change (with respect to feed) by changing either the bottoms flow or the side-draw. You can mentally simulate this based on what was explained earlier. However, in this exercise, both product compositions are altered. This is where reflux flow and one of the flows between the bottoms flow and the side-draw flow come to the rescue to keep both compositions constant, resulting in split of feed between the bottoms flow and the side-draw.
Greg: Are there any other cases where you don’t recommend using bottom section Tray-TC?
Vivek: Sometimes the top product spec is more important than the bottom product spec. If there’s sensitive tray-temp in the top section that correlates with the top impurity, use the top section Tray-TC instead of the bottom section Tray-TC. The tray location for Tray-TC is important to catch the composition profile at the correct point in the tower. Choosing the best tray often depends on the product specification at the top or bottom section. Where bottoms reboiler is limited and there’s no side-reboiler to offload the bottoms reboiler, stable Tray-TC is not feasible. If implemented, there’s a constant struggle to keep Tray-TC in control by adjusting its setpoint. In this case, you may consider cascading reflux drum level to reflux flow and manipulating the reboiler flow or valve directly.
I witnessed one site where the bottoms reboiler was limiting, and the side-reboiler wasn’t being used, even with the free-reboiling medium (cracked discharge gas flow in an olefins-unit). Multiple attempts by operators to increase side-reboiler duty led to off-spec ethylene product, which was clearly undesirable. The side-reboiler process outlet temperature ran higher than the tower bottoms temperature due to small process flows on the cold side, while the heating side was driven by hotter, large-quantity, cracked gas discharge flows.
The missing piece was the bottoms section Tray-TC. When Tray-TC was configured, any increase in side reboiler duty, and bottoms reboiler duty was automatically reduced by an appropriate amount, restoring the heat balance. They had to be careful not to make huge moves in side-reboiler flow during transfer of duty to avoid dynamic overshoot in the top-composition, until the bottoms section sensed an increase in the Tray-TC process value cutting the bottoms reboiler duty. This dynamic orchestration of duties is nearly impossible to achieve manually, which is why off-spec product was produced without the bottoms section temperature controller.
Greg: Are there any “gotcha” moments for practicing younger control engineers while implementing Tray-TC controller?
Vivek: Based on typical process control training, I encourage an open loop test of Tray-TC by changing reboiler flow/duty. Hopefully, the control engineer recognizes after interacting with the operator, or due to the process knowledge acquired, that increase (or decrease) poses little or no risk of sending product off-spec. Next, make a move in reboiler flow/duty and wait for Tray-TC’s response, holding feed and reflux constant. You must monitor the process’ real-time trends for important column variables, not only reboiler flow and affected temperature.
In my experience working in olefins units, Tray-TC never stops changing over 45-60 minutes. As a result, control engineers record a large process gain along with long response time.This is based on the tuning rules/spreadsheet, which let the engineer arrive at small controller gain along with large integral time. Everything practiced so far is consistent, except in this case, where control engineers (sometimes even experienced engineers) failed to recognize that temperature is an integrator and not a usual steady state variable. This means Tray-TC must have large gain and smaller (~10 mins) integral time to control temperature effectively.
In many columns with relatively close boiling points, or in cases where bottoms impurity is in ppm, Tray-TC often is an integrator. Many academics pull their hair because they believe it can’t possibly occur at steady state. However, over the operating range of interest, it occurs in practice without a shadow of doubt. As Nobel Laureate physicist Richard Feynman famously pointed out, “if theory doesn’t agree with experiment, it’s wrong.”
This is one area that also gives steady-state optimizer experts a headache when matching Tray-TC in a column temperature profile. Instead, they match reboiler duty and reflux flow (surrogate for condenser duty). In case of ppm quantity impurity, after a sharp increase in reboiler duty, incremental liquid from upper trays sequentially returns to control tray. This recovers heat from increased vapor flow, so it requires a dynamic large change in reboiler duty with near-zero, steady-state change in the reboiler duty to raise the control tray temperature to a higher setpoint.
An open-minded engineer must look for either corroborating or refuting evidence. If on a step change in Tray-TC setpoint you find that the net change in reboiler flow/duty is very close to zero (can’t distinguish from noise in flow/duty), then it would confirm that Tray-TC in open loop must be a ramp and vice versa. Next, this should be cross-confirmed by a net-zero change in ramp rate of bottoms level. For example, rate of change of bottoms level at steady state is zero (if open at regulatory control), or the zero change in the bottoms flow (if closed, i.e. level cascaded to bottoms flow). In case of total condenser, the effect of change in Tray-TC setpoint on the reflux drum level ramp-rate must also be zero. This doesn’t mean that the effect of Tray-TC on overhead composition is zero. In fact, it’s prominent. In this case, net transport of ppm-level material to the top is a “bucket” effect, as opposed to a “hose” effect, when a non-zero, incremental stream of material appears in the top due to a net change in reboiler duty. Conversely, if there’s a finite change in reboiler duty when a Tray-TC setpoint change is made, then it must have a non-zero, ramp-rate effect on both the bottoms level as well as on the reflux drum level.
Taking this analysis to another level, in a non-zero, ramp-rate case, change in bottoms flow to balance bottoms level must be equal and opposite to a change in distillate flow to balance the reflux drum level. Final reasoning is based on the stream of mass transported from the bottom stream to the top when increasing Tray-TC setpoint. In case of ppm level impurity at the bottom, this delta change in ppm (converted to delta weight-fraction) multiplied by the bottoms flow is very small and almost indiscernible within the noise level to cause any steady-state change in ramp-rate of the levels involved.
Another peculiarity in case of ramp-like/integrating is open-loop response of Tray-TC to change in reboiler duty. If you change bottoms flow holding the feed constant, then based on textbook incremental mass-balance, you’d expect incremental distillate to decrease (ΔD = - ΔB). But, since the bottoms level can’t be balanced by any finite increase of Tray-TC setpoint (with zero steady state gain to reboiler duty), the top distillate flow is unchanged. However, this (ΔD = - ΔB) holds if the Tray-TC has a steady-state response to change in reboiler duty, or if reboiler duty/flow is manipulated directly (not recommended).
Greg: Tell us more about the qualitative reasoning applied to distillation operation?
Vivek: As they say, there’s no free lunch. If the feed is increased and both product streams are kept at constant composition, then the reboiler duty and the reflux flow must increase. How can reboiler duty for vapor feed possibly increase in the presence of bottom-section Tray-TC control? For vapor feed with Tray-TC in the bottoms section, one finds that reboiler duty is reduced due to latent heat of condensation of the vapor feed. This will cause the bottoms level to increase, allowing it to draw more of bottoms flow to keep bottoms level in balance. Thinking along further, the vapor feed will rise in the column quickly and will increase top impurity. This would call for more reflux flow to maintain constant top composition of on-composition control. Increased reflux flow will tend to increase the reboiler duty at steady state to hold Tray-TC constant despite an initial cut in feed increase. Note the difference in change in reboiler duty due to increase in load (reflux) versus setpoint change of Control Tray-TC. In case of liquid feed, reboiler duty will increase due to Tray-TC increasing feed. Plus, the appropriate amount of feed that boils at a temperature higher than Tray-TC setpoint will end up in the bottoms resulting in increased bottom flow.
Greg: Do you have any thoughts on tower pressure manipulation?
Vivek: Reduced tower pressure is often desirable due to the favorable effect of relative volatilities. If a tower is vapor-flooding, then reducing the pressure will hurt due to increased vapor traffic in the column. You must also be mindful of a curved ball effect on the tower pressure setpoint change on the top composition. You may encounter inverse response to overhead composition on pressure increase. Some control engineers zero out this initial inverse dynamic portion due to lack of understanding. This effect is due to initial increased liquid traffic due to differential condensation in the upper part of the column. This cleans up overhead impurity due to increased internal reflux initially, but then the unfavorable effect on relative volatilities kicks-in along with the bottoms section temperature controller action, and the overhead impurity rises.
Often C3-splitter pressure floats on cooling water temperature (to operate at minimum pressure) in Gulf Coast chemical plants. You can see this large inverse response during thunderstorms, which causes drops in tower-pressure and resulting excursions in overhead impurity.
Greg: Since so many of my colleagues are of retirement age, I thought the ISA book I wrote Funnier Side of Retirement for Engineers and People of the Technical Persuasion with cameo appearances by the late Stan Weiner and cartoons by Ted Williams would be a good source of top ten lists for the next several months. The ISA price is just $15 in a Fire Sale of this book.
