June 2018

Process Control and Instrumentation

Advanced process control of debutanizers makes use of inferential predictions

Following good experience with a generalized cut-point calculation (GCC) inferential package, which models main fractionators with side streams, French energy major Total decided to experiment with a generalized distillation shortcut (GDS) inferential package, which models simpler, two-product distillation columns.

Following good experience with a generalized cut-point calculation (GCC) inferential package, which models main fractionators with side streams, French energy major Total decided to experiment with a generalized distillation shortcut (GDS) inferential package, which models simpler, two-product distillation columns.

Several GDS models have been implemented at Total’s Lindsey oil refinery in Lincolnshire, UK. Two debutanizers are studied in this article: the fluid catalytic cracking (FCC) gas plant debutanizer and the unifiner debutanizer. The precise separation of FCC liquefied petroleum gas (LPG) from naphtha is of value because olefinic FCC butane is an alkylation ingredient, whereas pentane is a gasoline component. Sending pentane to the alkylation unit takes capacity and creates undesirable reactions. For the unifiner, separation of LPG from light straight-run naphtha (LSRN) is of value because LSRN is a gasoline component, and high Reid vapor pressure makes it difficult to blend.

GCC model and discussion

GCC applies heat balance, mass balance and partial pressure-corrected temperatures to estimate the crude true boiling point (TBP) curve. From the TBP and internal reflux, GCC can be used to estimate product properties. Not only is GCC useful for providing an inference model, but these models are also robust enough to remain in closed-loop control during crude switches, which can be considered an advantage.

Total has successfully implemented GCC on multiple crude distillation units (CDUs) across the organization and, in particular, at the Lindsey refinery. Fig. 1 is one example of many, demonstrating GCC inferential performance for several days at Lindsey refinery’s Crude Unit 2. This chart trends 10 d of selected GCC inferences (blue lines) vs. lab results (magenta squares). The red line at the bottom of Fig. 1 is a trend of the slope of the crude TBP curve (i.e., the increase in temperature needed to increase evaporation by 1% of the crude run).

Two massive crude switches occurred during the 10-d period, one from intermediate to heavy, and then back to intermediate. On Day 9, another, less-massive switch was observed. In addition, several minor switches—likely tank-to-tank switches of the same crude—were observed. During the period, inferred properties showed good trends against the lab results, and advanced process control (APC) continued uninterrupted. GCC technology and related APC performance are documented in literature.1–5

GDS theory

While GCC makes use of boiling curves without defined composition, GDS works with distinct components and analyzes DCS measurements around a specific section of a distillation column to obtain an inferential model. The green areas of Fig. 2 show measurements around the lower part of a stripping section. The inputs are pressure, temperature and adequate measurements to permit vapor and liquid traffic calculation by heat balance. A minimum of two column temperature points are needed—one at the bottom and the other on a tray. That tray must be distant enough from the bottom to have a light key component content of 10% or higher.

The model makes use of Colburn’s method,6 which estimates the ratio of vapor composition on Tray N (the upper temperature measurement tray) to the bottom liquid composition. The convenient, closed-form calculation correctly takes into account the nonlinear effects of column loading and number of trays. For a simple stripping section, the Colburn ratio takes the form shown in Eq. 1:

Ri = 1 + (Zi – 1) × (Ki – 1) ÷ (Ui – 1)                                    (1)


Ri = Ytrayi ÷ Xboti is the Colburn ratio

Ki = Component i volatility, vapor to liquid composition

(V ÷ L) is the vapor to liquid molar flow ratio, calculated by heat balance

Ui = Ki × (V ÷ L) is the effect of column loading

N is the number of theoretical trays in the section; i.e., Zi = UiN.

Following the calculation of volatilities, column loading and Colburn ratios, GDS simplifies the composition into four components and solves a set of four equations with four unknowns. An example GDS equation set for a debutanizer stripping section follows. The problem setup has four unknowns to describe column bottom composition: NC4, C5, C6 and C7+:

  • NC4 is the light key component, to be kept at a level of 0.5%–1%
  • C5 is the heavy key component, volatile on all stripping trays
  • C6 is the heavy-heavy key component, volatile in the lower trays only
  • C7+ is the extra-heavy key component, assumed to be nonvolatile even inside the reboiler.

To estimate the bottom composition, GDS offers four equations, listed in their simplest forms in Eqs. 2–5:

Bottom mass balance:

Σ (Xboti) = 1 (Sum of bottom molar fractions = 1)                                (2)

Reboiler equilibrium: 

Σ (Kboti × Xboti) =

1 (Sum of reboiler vapor molar fraction = 1; indicates equilibrium)         (3)

Section separation: 

Σ (Ri × Xboti) = 1 (Sum of Tray N vapor molar fraction = 1)                  (4)

Ri is the Colburn ratio5 for component i, the ratio between vapor composition on Tray N to bottom liquid composition.

Tray equilibrium: 

Σ [(Ri ÷ KTrayi) × Xboti] = 1                                                               (5)

(Sum of Tray N liquid molar fraction = 1)

The ratio of vapor to bottom liquid composition on Tray N is 1 ÷ Ktrayi. Therefore, Eq. 4 is a tray equilibrium equation.

While the four-by-four matrix coefficient calculations are nonlinear and involved, the four resulting equations are linear, and solution is guaranteed. The calibration procedure for this model involves adjusting tray efficiency for the total section and for Tray N (affecting Eq. 4), to obtain a good fit between model and lab results.

FCC debutanizer description

Consider the debutanizer of Fig. 3, column C-4. The tray temperature controller (TC) manipulates the reboiler, whereas operators or APC manipulate reflux flow. Distillation control is a two-by-two problem, with the two degrees of freedom being “cut” and “fractionation.” Cut is distillate yield to be manipulated when one product is too pure and the other is too contaminated. Fractionation is column loading, to be manipulated when both products are too pure or too contaminated. The debutanizer tray TC is a cut control device. When the tray is too cold, the TC increases reboiling, sending more vapor to the overhead drum and, eventually, to the distillate, thereby increasing the cut. It helps that tray temperature is also a rudimentary inference of product purity.

FIG. 3. Lindsey refinery FCC debutanizer configuration.

Such a control structure is convenient because it avoids interactions between cut and fractionation. Changes of cut do not alter the reflux. Changes of reflux and column loading have some dynamic effect, but they do not alter the steady-state tray temperature and only minimally alter the cut.

Debutanizer pressure is controlled by manipulation of gas flow to condensers, a method commonly used with fully condensed overhead product. The hot bypass shown is always shut, however.

The DCS overhead drum level control is on the LPG product. However, due to downstream unit issues, LPG flow can be manipulated only slowly and up to a certain limit. Therefore, when APC is active, it takes over level control and, if necessary, sacrifices distillation purity economics. In the absence of constraints, economics call for maximizing the content of C5 in butane up to approximately 1%. GDS actually infers C5 in LPG, but it also calculates the complete LPG composition: C2, C3, C4 and C5. It can also estimate C5 impurity in total butane (two columns downstream of the debutanizer).

FCC debutanizer inferences

Fig. 4 shows trends in the content of C5 in butane over a 6-mos period. The dark blue line is the Total GDS model. The light blue line is an aging analyzer reading, noisy and not reliable enough to work in closed loop. Lab results are shown in the magenta squares.

FIG. 4. Trend of C5 in FCC LPG.

Due to sampling difficulty, lab samples are not normally taken; the few tests were carried out by special request to help calibrate the inference. These infrequent lab tests show how confident operations are regarding the accuracy of the GDS inference; there is no longer a need to sample this stream, and GDS calculation can be relied upon almost entirely. Calibration was influenced by a desire to match the lab, and not the noisy analyzer. While more lab tests will help ensure better calibration, the inference shown is presently used for controlling the column.

FIG. 5. Trend of C4 in FCC light naphtha.

The existence of a tray temperature point in the stripping section permits a bottom GDS model, thereby predicting slippage of butane into naphtha. The fit of the bottom C4 prediction vs. the lab and analyzer is shown in Fig. 5. Here, analyzer performance is again not overly impressive. Inference quality is decent, but it works in open loop only because the objective is to minimize C4 in naphtha, not to control it. Bottom C4 minimization is accomplished by maximizing the reboiler (and reflux) to pressure control constraints.

FIG. 6. Unifiner debutanizer configuration.

Unifiner debutanizer description

The unifiner debutanizer is shown in Fig. 6. As opposed to the FCC debutanizer, this column has a furnace reboiler, and reboiler heat is controlled to satisfy reboiler outlet temperature. While industry is in general agreement that a tray TC should manipulate reboiler heat duty, furnace reboilers are often equipped with a coil outlet temperature (COT) controller.

The tray TC has the advantage of cleaner inferential value, but the reboiler outlet TC is acceptable. A rectifying tray TC is located on the reflux, but it is preferentially kept open. With the cut being more or less determined by reboiler COT, the influence of reflux on the percentage of C5 in LPG is small, and closing the rectifying section TC would not have been successful. The authors’ experience confirms previous research7 showing that two temperature controllers can compete at times, thereby driving the column into limit cycling.

As in the FCC debutanizer, the LPG rundown valve has been a constraint, but for different reasons. The solution was to open the level, thereby manipulating the LPG flow valve directly while controlling the reflux drum level as a control variable (CV) in APC. Column pressure was controlled by the manipulation of offgas. Since the desire was to minimize offgas, the offgas flow is a CV, controlled to minimum by increasing pressure or decreasing column heat load. The pressure itself is not steady, which is not a problem for the inferential models.

As a side note, the percentage of C5 inference is controlled as an integrator, using primarily the reboiler COT.

Unifiner debutanizer inferences

The existence of tray temperatures in the rectifying section makes the inference of C5 in LPG possible. This is helpful because economics call for the minimization of the bottom C4 subject to a 1% C5 in LPG constraint. Lab samples are not normally taken here, but Total has requested several tests to help calibrate the inference. Calibration results are shown in Fig. 7. Fig. 8 shows the trend of a later 6-mos period when the column was under APC, which often kept the content of C5 in LPG up to the 1% limit. Unfortunately, only two lab tests were carried out during the period.

FIG. 7. Three-month trend of unifiner C5 in LPG.

FIG. 8. Six-month trend of unifiner C5 in LGP.


GDS is a simplified distillation model, reducing a multicomponent problem into just four components and estimating separation in a section of column without rigorous, tray-by-tray calculations. The inferential results are surprisingly accurate and robust. C5 can be controlled in LPG at a target of 1%, benefitting the refinery in key ways:

  • Increasing alkylation unit feed.
  • Reducing the RVP of light virgin naphtha (LVN), making it easier to blend into gasoline. This is also required to maintain the integrity of the LVN
    storage tank.

The GDS technology continues to provide robust and reliable service, both as an indication to the panel operator and as a CV input to APC applications. Each of the APC controllers discussed operates with a high service factor (> 85%) and has seen excellent acceptance by process operators since their commissioning. HP

Literature cited

  1. Y. Z. Friedman, “Asphalt DSR prediction and control,” ARTC conference, March 2014.
  2. Nam, S.-Y., Y. Z. Friedman, P. Khumar and A. B. Azahar, “Delayed coker advanced process control at Petronas Melaka refinery,” Hydrocarbon Processing International Refining and Petrochemical Conference (IRPC), June 2010.
  3. Haseloff, V., Y. Z. Friedman and S. Goodhart, “Coker advanced process control at BP Gelsenkirchen refinery,” Hydrocarbon Processing, July 2007.
  4. Fuentes, J. O., J. A. Sanchez, M. J. A. Lopez, A. A. Bascones, J. Hall and Y. Z. Friedman, “Implementation of APC on Repsol Poetollano CDU1,” ERTC Computer Conference, May 2007.
  5. Singh, P., T. Hiroshima, P. Williams and Y. Z. Friedman, “Multivariable controller implementation for a crude unit: A case study,” NPRA Computer Conference, October 2002.
  6. Colburn, A. P., Ind. Eng. Chem., Vol. 33, 1941.
  7. Friedman, Y. Z., “Prediction for APC in refinery FCC gas plant,” PTQ, 2Q 2016.

The Authors

Related Articles

From the Archive