Reprinted From... CEP CHEMICAL ENGINEERING PROGRESS FRACTIONATION COLUMN CONTROL D. M. Boyd UOP Process Division Des Plaines, Illinois Vou. 71, No. 6 JUNE 1975 Paces 55-60Heavy Key in Overhead Light Key in Bottoms. Figure 1. Performance curve. Trays Temperature Figure 2. Typical temperature profile. Fractionation Column Control By taking advantage of the natural composition profile in a fractionation column, this methed enables control of the tower at the feed tray instead of at either end of the column. D.M. Boyd, UOP Process Div., Des Plaines, I. ‘The natural composition profile in a fractionating col: tumn can be used to achieve tower contro} more accu rately than most plant laboratory analyses by using un- complicated, readily available temperature measuring devices. Before discussing this type of process control, hhowever, itis best to review just how much control can bbe imposed, and the limits placed upon such regulation by the design of the colurnn, Once the number of trays, the reflux ratio, the feed composition, and the feed tray location all have been fixed, the column must operate along the curve of Fig ure I, which shoves a fixed relationship between the im- purities in the overhead and the impurities in the bot toms. Better purity of the overhead (light key) product CHEMICAL ENGINEERING PROGRESS Vol.71, No.6) can be obtained by allowing some of the ight key product, toappear in the bottoms product or vice versa, However, fixing the purity at one end of the column fixes the purity of the other product. The desired operating point 1s "X," which gives the maximum separation of both ends under the conditions chosen. ‘Suppose we now consider the effect that moving along the curve in Figure 1 has upon the temperature profile {or composition profile) of the column, shown in Fig ture 2. Curve X corresponds to point X on Figure 1, which gives the maximum separation at both ends of the cal- tinn, Note that it starts down from the top of the column ith very Little temperature change from tray to ray. when the rate of temperature change increases as the feed June 197555;tray is approached, and this rate diminishes again as the bottom of the column is approsched. However, curve 0 is quite a different matter. Because ‘there are some bottoms in the overhead, the temperature lightly higher at the top tray of the column. Thus, because it does not take as many trays to make this pur rity as it does for curve X, the temperature increases ‘much more rapidly as one goes down the column. ‘At point 1 on curve 0, by vapor liquid equilibrium, the same composition has been reached as is on the feed tray. Therefore, there is very little composition, or tem- perature, change between this point and the feed tray. ‘This is sometimes referred to as the “pinch point.” ‘Asone moves down the column from the feed tray, the ‘temperature approaches the bottoms temperature, giving. very little temperature change in the bottom section. ‘This, then, will give a higher bottom product purity than the operation shown on curve X. ‘Curve ¥ is the opposite of curve 0. The bottom tem- perature is slightly lower than that of curve X and the “pinch point” occurs below the feed deck. Above the feed deck the temperature profile rapidly approaches the top temperature, giving higher overhead purity than either ceurves X or 0. Tt would appear, therefore, that the area in the im- mediate vicinity of the feed tray, where such dramatic changes in temperature occur with changes in overhead fr bottom composition, should hold much promise as a ‘means of control instead of either the top oF bottom of the column as is traditionally used. Deseribing such a ‘means is the purpose ofthis artile Figure 3. Computer calculated performance curve. & BENZENE COLUMN aT ‘CONSTANT REBOILER OUTY PLOTNO. TOL. OVHD. /82. BOTT. pm Tope Aaases Ma a8 me aa fas ne Figure 4. Temperature profile vs. tray number. 56 June 1975 CHEMICAL ENGINEERING PROGRESS {Vol.71,No.6)a Figure 5. A-A\ temperature vs. top and bottom compesition. in actual fractionator TThe column chosen for our tests was an actual col ‘umm, feeding @ solvent-extracted aromatic hydrocarbon mixture and taking benzene as its overhead product, The bottom product is 50% toluene and the remaincer mixed xxylenes and ethylbenzene. It had 70 trays, with the over hhead product draw-off on tray 5 (numbering from the top). It therefore had 65 real fractionating trays, the top five being mainly for drying. Feed was on tray 39. Using an assumed tray efficiency of 80", a series of tray: to-tray calculations were run on a digital computer The resulting performance curve is shown in Figure 3. Plotting the tray temperature vs. tray mumber for the 10 points shown results in the family of curves shown in Figure 4. Note that each point in Figure 3 defines a unique eurve in Figure 4 "The control system used on this operating colutnn con- sisted of a differential temperature measurement. be- tween theoretical tray 5 and theoretical tray 15. (We ‘would now use theoretical tray 5 and theoretical tray 20.) A second differential temperature was taken in the bot: tom section of the column, between theoretical tray 30 fand theoretical tray 45. Thus, we used two differential temperature measuring circuits in each section of the column, ‘The two differential temperature signals were then fed into a subtraction relay and the difference between the two was the control signal. That is, we used the dif: ference between the two dif erential temperature mea: surements far the control signs! Figure 6 is a dovble plot, showing the effect of signal obtained by subtracting the lower AT from the upper 47" ‘upon the amount of toluene in the overhead product and the amount of benzene in the lower product, Several reflux-to-feed ratios are shown. Note that each point represents one given column operation. For example, 10 parts/million of toluene in the overhead will give a control signal of +10°F., while 10 parts/million of bem- zene in the bottoms will give a control signal of ~15°F. Certainly this is a very large signal for such a small ‘change in the purity ofthe products. In fact, the sensitiv- ity was so great that some concern was felt for the stabil ity ofthe control loop, Te can he seen in Figure 6 that going from 100.000 mols of overhead product to 100.030 mols of overhead CHEMICAL ENGINEERING PROGRESS (Vol.71,No.6) (Benzene Column) 1.35 R/F 4°1,20R/F AAT Predicted, °F 100,000 100.010 100.020 100.030 100.040 Mols Overhead Product, % of Base Number Figure 6. Effect of overhead flow upon - temperature. Figure 7. Benzene column. product the A-A7° goes from ~30°F. to +10°F., ‘That is a O'R, change in temperature for three parts in 10,000 change in overhead product! Te was apparent that an instrument with an extremely broad proportional band would be required to achieve stable control. Inasmuch as most controllers have a maxi ‘mum proportional band of 500%, a feedforward system ‘was devised by which, biasing the ratio setting on the June 197557feedforward system, we could multiply the proportional band setting on the controller by another factor of 4, This ‘would give a proportional band of 2,000% However, again examining Figuee 4, there is some hope that it would be stable. Note that going from curve 2 to ‘curve 10 would require a large change in the inventory on theoretical trays 15 to 25, Curve 10 has almost pure hen- zene on these trays, whereas curve 2 has a large amount ‘of toluene to give the increase in temperature. Naturally, it will take time to change this inventory because it has toe aceumulated from the feed to the column, Feedforward method not to be used The field tests proved this concern about stability to be unfounded, and the feedforward system was discon nected. The controllers were stable at a proportionsl ‘band setting of 20%. The reset rate was five minutes and the derivative was one minute. ‘Therefore, in the future, the feedforward feature will not be used. However, the square root extractor will be used on the controlled product draw-off to avoid the drastic change in gain produced by the non-linear flow signal from a DP cell, Figures and 8. The first installation of this system was made on benzene-toluene columns connected in series in the Trieste, Italy, refinery of Total Societia Italiana per Azoni (Aquila). The results were favorable, as shown by the chromatographic analyses in Table 1. The refinery chromatographic analyses were made on a nev, spec high-sensitivity chromatograph the refiner had in his search Dept,, in order that & high sensitivity could be used. The same samples that were analyzed on the refiner’s research chromatograph were analyzed again at the UOP laboratory in Des Plaines, Il, using « chromatograph equipped with a flame ionization detector. The toluene in benzene was found to range from 6 to 14 parts/million instead of zero (as suspected), However, the benzene in toluene was about 1 parts/ million instead of the 40 parts a | ee (ioe Figure 8, Toluene column. Benzene Column: Ay = 164; Th = 4690 { nm La Figure 9. Correlation of plant data with calculated data. Table 1. Effect of double differential temperature signal on concentrations of the key components. Basis: benzene-tolvene splitter ot Tieste* Tolvene in Benzenein _Xylenesin Toluene in Benzene in benzene col. telvene col. toluene ol. tlvenecal. benzene co. dd TRC, Feed rate ovhd.,ppm. _ovhd. ppm. ovhd,, ppm. ims. ppm. btm, ppm. Tet °C | design UoP Ref. UOP Ref. UOP Ref. UOP Ref. UOP 1 0 qian Vee 50s Oe DY Bas 42a On 6s Sen aaaine OLN? en Caeays ee Se 178 33 VO nae 8 Su 45 0.47 SoS 086 Shae $B tO is Oo. 06. 4a OBL 28 rs 62 dda ST “Analytical determinations were made at both the refinery and the UOP Research Laboratories. Col. column Btms, = bottoms Ovhd. = overhead ppm = parts/milion 58 June 1975 CHEMICAL ENGINEERING PROGRESS (Vol.71, No. 6)Increased Set Pol From +510 10°C | §8¢ and Increased 2 Feedio te! 'f Design Figure 10. Instrument records from benzene column. million reported by the refiner. (Our design specifications permitted up to 500) parts/ million.) Field data matched original computations What is even more interesting is that when the plant data, using UOP's flame ionization data, are plotted on the original computer ealeulation from this column, the plant data fall on the original curve, as shown in Figure 9. ‘The original curve was fora reflux ratio of 1,35:1 on feed! ‘The plant was run at 1-to-1. It must be concluded that, with the closer control, much higher tray efficiency was boeing obtained, due to the fact that the composition gra dient was being held better than we have been able to in the past. For high purity feactionation, it appears that by run ning the computer calculations and then measuring the double differential temperature in the column, the ity purities can be predicted more closely than the chro 'matographs available in the average plant conteol Iahors: tory ean determine. ‘The plant had been shut down by a power failure prior ‘to the author's arrival and had just been brought back fon stream. ‘The benzene in the feed varied from 6% to CHEMICAL ENGINEERING PROGRESS (Vol. 71, No.6) ure 11. Instrument records from toluene column. aout 14% over the next two day pero and then difted back to about 7.5% the rest of the time. (The design had bbe predicted on 20% benzene in the fee.) ‘The unit had been running at about 80% feed. Howe ator feed tank and was then increased to the design rate of design er, charge was accumulated in the fraction: ‘Therefore, the samples for the last three days were at 100% of design rate. These are identified in Table 1 as tests 2-6. Different settings of the control point of the double ditterenial (a) TRC wore tied. Note tht inially i ‘was set at zero and the benzene in the tolurne was 27) parts/million, This was changed to +2°C. and the cor~ responding samples (test 2) were 45 parts/million ben- zene in the toluene. The set point was next changed to +5°C. and the corresponding sample (test 3) showed 41 parts/million benzene in the toluene. Therefore, the dd TRC setting, was raised to +10°C. ‘The corresponding sample (test 4) showed 128 parts/million toluene in the benzene (UOP data, 33 parts/ million). Itshould be noted that tests 3 and 4 were run on the same day. Naturally, the plant could not be left running in this condition. It was intended to, run the column at 20% above the design rate, and the charge tank was becoming depleted. ‘The setpoint was lowered from 410°C. to “+5°C., the charge set at 120% of design and the steam. rate to the reboiler was inereased, This produced a big upset in the column, of course. However, the dd TRC cut the benzene to storage which, in effect, put the column in total reflux. Then when the proper A-A7' was estab- lished, it again increased the Mow to storage, The sam. June 197559AT Between Trays 388 ‘Sum of Composition and Pressure Drop Z Moot ‘Temperature Differential ‘Due to Pressure Drop, In Tray Temperature Differential —— due to Composition Column Load Figure 12. Effect of column loading on temperature differential. ples taken one hour later were on specification. These are shown as test 5. Figures 10 and 11 show the effect of the changes made aswell as the stability. They also show how the two com: {rollers handled a partial steam failure the night of the third dav. Noto the bad swing in column temperature and thatit cut the products going to storage until the column hhad lined out again. The samples taken the next morning, were again on specification, the column operating at 120% of design. vis also interesting to note the effect of the 20% in crease in throughput on the upper and lower differential temperatures of the toluene column. Both increased due (o the higher traffic, However, the difference between the two remained the same. This increase in differential temperature occurred because increased traffic increased the pressure drop through the column. At constant over: hhead pressure, the absolute pressure on each tray would increase, This would sllow the products to go “off specs,” sometimes seriously enough to cause a reversal in the action of the controller, thus completely upsetting the column. How the controller is adjusted In tuning the controllers on an installation of this system, itmust be kept in mind that the double differen: tial Lemperatare controller is controlling the product to storage, However, in order for the temperature to be af- fected, the reflux at the point of measurement has to be changed Tn the case of the benzene column, this created no problems because the product to storage came from a trap tray in the column, However, in the ease of the tol gene column the produet came from the overhead receiver. ‘Therefore, to affect the temperature it was first neces sary to change the level in the receiver, which would then change the reflux flow. ‘The level controller originally had a 50% proportional band setting, which was much too slow. It was changed to 3% proportional band and the performance improved tremendously. The double differential temperature con- trollers were each left with a proportional band of 20: five minute reset, and one minute derivative. ‘There-has been a gradual evolution in the control of fractionating columns, ‘Temperature has been the prime 60 June 1975 ‘means of control. However, temperature will only indi- ceate composition at a given pressure. When control for high purities is desired, oven barometric pressure changes ccan throw the product off specification. ‘A simple way to control product purity is to use differ- ential temperature, One thermocouple is located in the relatively pure material and the other at the desired con- ‘rol point. Therefore, as the pressure changes, both tem- peratures are affected, but the difference between the two remains the same. ‘This aytem has ono rather serious drawback; the differential temperature measured is com- posed of two factors. One is the result of the difference in composition and the other is the pressure drop in the column, Figure 12. ‘As the columa is loaded to improve composition, the differential temperature tends to decrease. However, the differential vemperature tends to inerease due to the pres: sure drop in the (rays, The net result is that the measured differential temperature passes through a minimum and then inereases again. ‘This phenomenon generally con- fuses the man operating the column and very definitely limits the operating range. ‘With the new aystem, the differential temperature re- sulting from the pressure drop appears in the top differ- ential temperature as well as in the bottom differential temperature, The bottom reading is subtracted from the top so the effect on the reading is eancelled out, In conclusion It would appear that this method of control is @ radical departure from the conventional method of controlling a fractionating column. Conventional methods control composition at one end of the column or the other by ‘means of temperature ot some analytical device This method, by taking advantage of the natural com: position profile that exists in a fractionating column, con- trols near the feed tray instead of at either end of the column. By keeping the column out of “pinch” conditions around the feed tray, compositions can be maintained at either end ofthe column to an accuracy greater than that fof most plant laboratory analyses. ‘The measurements "used are very simple temperature measurements that are well within the eapabilities of available equipment and are very easy to maintain and keep in operation, Field tests described here show the system to be ex- tremely stable as compared to conventional systems and certainly well within the capabilities of the average oper- ator to handle. Acknowledgment | would like to thank Silvano Maizzan of Aquila for permitting me to use and identify data obtained from the ‘company’s refinery in Trieste, Italy. ' D. M. Boyd, Jr. is manager, instrument design and service, Engineering Dept. UOP Process Div. His work ix concerned with the development and application of instru ‘mentation 10 refinery ond petrochemical processes, He received his BS.ChE. trom the Univ. of Colorado, and in 1950 wor ‘ewarded the honorary degree of chemical = engineer by the university for developing the grophictype control panel. He has been awarded 160 patents in the field of instrumentation and is the uthor of ‘numerous articles in thie fil CHEMICAL ENGINEERING PROGRESS (Vol. 71, No.6) Y