FIELD DATA, NEW DESIGN CORRECT FAULTY FCC TOWER REVAMP

Scott W. Golden, Gary R. Martin
Glitsch Inc.
Dallas

Karl D. Schmidt
Lyondell Petrochemical Co.
Houston

In 1987, Lyondell Petrochemical Co. revamped a fluid catalytic cracking unit (FCCU) main fractionator by substituting structured packing for trays.

But because this revamp did not achieve its design objectives, a second revamp was performed in 1992.

Packing large-diameter main fractionators can increase unit capacity and decrease pressure drop while meeting fractionation objectives. A packed main-fractionator revamp, however, must take into consideration proper design and inspection procedures.

Lyondell's experience illustrates the procedures that produced a successful structured packing revamp.

BACKGROUND

The 79,000 b/d FCCU main fractionator at Lyondell Petrochemical Co.'s Houston refinery was revamped from trays to structured packing (Fig. 1). The justification for the revamp was a capacity increase to 92,000 b/d.

An ultimate capacity of 100,000 b/d was anticipated, at the same gasoline cut-point as the trayed column, and at reduced column pressure drop (Fig. 2). Before the revamp, unit capacity and conversion had been limited by a low catalyst-to-oil ratio. The proposed unit pressure-balance adjustments associated with the reduced column pressure drop were predicted to allow higher catalyst circulation and increased conversion. 1 2

The lower column pressure drop was used to offset changes in the design of the regenerator-air grid plate distributor. The new grid plate design was more modern and required a higher pressure drop, therefore the regenerator pressure was lowered to offset these changes.

At the 92,000 b/d charge rate, the gasoline true boiling point (TBP) endpoint was consistently 550 F. or greater,

The endpoint did not change with increased fractionator reflux or decreased unit feed rate.

This high gasoline end point resulted in 7,000 b/d of heavy gasoline being blended to the middle-distillate pool. Several modifications to the packed column internals did not improve gasoline quality.

The column internals eventually were modified by a second revamp. This revamp achieved not only the 92,000 b/d capacity, but also the design pressure drop and fractionation.

Any refinery packed main column internals must be designed, installed, and properly inspected before start-up to ensure proper operation. Many packed columns have failed because of poor design practices or faulty installation.

PROBLEM DEFINITION

The original project justification was increased unit charge rate and higher gasoline production, but the increased gasoline production was not realized. The main fractionator raw gasoline (liquid product from overhead receiver) had a high endpoint, and the light cycle oil (LCO) product contained 2,400 b/d of recoverable gasoline.

This unit fractionates FCC gasoline into light, middle (heart-cut), and heavy gasoline prior to blending or reprocessing (Fig. 3). The LCO product consistently had a TBP 20 vol % point of 430 F. or greater.

The net result was that 7,000 b/d of heavy gasoline had to be blended to the middle distillate pool year-round, and 2,400 b/d of gasoline was lost to the LCO product. The remainder of the heavy gasoline could be blended to the refinery gasoline pool because of the "masking" effects of the light refinery-gasoline blending components. ,

The approximate FCC gasoline material balance after the revamp was:

• Light gasoline (gasoline splitter), 17,600 b/d

• Light gasoline (cat. naphtha fractionator), 4,800 b/d

• FCC gasoline splitter bottoms, 6,000 b/d

• 2222 Heart-cut, 9,950 b/d

• Heavy gasoline, 12,000 b/d

• Total FCC gasoline 50,350 b/d.

After initial attempts to correct the problem failed, Lyondell and Glitsch began troubleshooting the column. Several field observations were made:

• The column overhead vapor temperature did not respond to increased reflux. (At times, the overhead temperature actually increased with higher reflux.)

• The gasoline endpoint and the front end of the LCO did not change materially with large reflux-rate changes.

• The fractionation between the unstabilized gasoline and the LCO product was the same at a 60,000 b/d or 92,000 b/d feed rate. At 60,000 b/d, large reflux-rate changes had little or no effect on fractionation.

• The measured column pressure drop was near the design value. The individual bed pressure drop could not be measured because no instrument taps were installed and low pressure drop is difficult to measure accurately.

The problem appeared to be poor liquid distribution throughout the column and little or no remixing of internal liquid. Poor liquid distribution causes significant variations in liquid-vapor ratios across the tower cross sectional area, which result in compositional gradients.

Fig. 4 is a schematic of the gasoline/LCO fractionation section. The structured-packing bed yielded approximately 0.5 theoretical stages, or an apparent height equivalent of a theoretical plate (HETP) of greater than 20 ft.

This may seem impossible, but it is typical of poor distribution in large refinery fractionators. The beds are short (3-8 theoretical stages) and the columns are typically large diameter. If one section of the bed has an IN ratio much lower than another section, the result will be high-endpoint material leaving that section of the column.

Lyondell concluded that there was poor liquid distribution, and the consensus was that structured packing does not work in large-diameter FCC main fractionators. But because the column was not well- equipped with instrumentation, good field data were not available. A test run was planned to gather additional field data.

INITIAL REVAMP DESIGN

In the original trayed column, the feed rate was limited to 79,000 b/d (Fig. 5). The measured main-fractionator pressure drop was 3.7 psi.

At 79,000 b/d, 75% conversion, and test-run heat and material balance conditions, the calculated pressure drop using the column internal loadings was 3.7 psi. The measured pressure drop during the test run was 3.6 psi.

The stated revamp objective was a 15% increase in feed rate while maintaining fractionation comparable to that of the trayed column. Minimizing column pressure drop consistent with these stated objectives had product selectivity benefits.

Reduced pressure drop allowed adjustments in the reactor/regenerator system pressure profile. These adjustments enabled Lyondell to increase the catalyst-to-oil ratio with a corresponding increase in unit conversion and selectivity.

The regenerator pressure was lowered, permitting higher air-grid pressure drop without reducing air-blower capacity.

Table 1 shows the raw gasoline and LCO product distillations during a test run of the trayed column. The top circulating reflux operated at about 126 MMBTU/hr heat removal, which equates to about 58,000 b/d of internal reflux.

Table 2 shows the calculated percent flood and observed fractionation-section efficiencies. The trayed column had approximately three theoretical stages between the top pumparound and the LCO product-draw stream.

The trayed fractionator was flooding in one section, although the exact section could not be determined. Lyondell conducted a plant test to determine whether the flooded section could be isolated.

The top pumparound had the highest calculated percent flood.

It was postulated that if this section of the column were flooding, the fractionation could be improved by reducing the loading in the raw gasoline/LCO fractionation section.

The column heat balance was shifted by moving about 17 MMBTU r from the top pumparound to a previously shutdown LCO pumparound system. The fractionation became worse, which was construed as proof that incipient flooding was not occurring.

Two theories prevailed. The first assumed the slurry-pumparound baffle trays were flooding, resulting in massive entrainment of liquid to the wash-zone trays. The second theory was that wash-zone trays 8-10 were flooded by high vapor rates caused by reduced slurry-pumparound heat removals.

The symptoms of the flooding were rapid loss of bottoms level and rapid buildup of pressure drop in Trays 9-17, where a differential-pressure recorder was installed.

The only tactics that eliminated the flooding were higher slurry-pumparound heat removal and higher column pressure. Adjusting the main-fractionator pressure was not a reasonable control method because of its effect on the pressure balance.

The operators monitored the tray differential pressure drop and maintained the slurry-pumparound duty at about 200 MMBTU/hr to avoid flooding. The column was operated with maximum slurry-pumparound duty consistent with the slurry-product gravity specification.

Heavy cycle oil (HCO) product was drawn off as a fuel oil blend component to control the slurry-product API gravity to less than - 1.0. The slurry is sold as carbon black feedstock. If the slurry-pumparound heat removal was too high, it was not possible to meet the slurry-product gravity specification.

The column internals had to be modified to achieve the new capacity objectives and lower the pressure drop. Lyondell decided to replace the trayed-column internals with structured packing. The baffle trays were replaced with grid and the remainder of the internals, with structured packing.

The packed-column hydraulic design was consistent with 100,000 b/d fresh feed at 75% conversion, although the unit would not run at more than 92,000 b/d because of environmental permit limitations. (A new wet flue gas scrubber planned for 1994 will allow operation at as much as 102,000 b/d.)

Table 3 summarizes the fractionation-bed depth and design packing performance. The column has two pumparound side draws and one product side draw (Fig. 6).

A combined liquid collector/redistributor was selected to increase the depth of each fractionating bed, the theory being that more packing is better. The column pressure drop was 0.8 psi. The packed column started up in early 1987.

COLUMN PERFORMANCE

The design performance objective was improved fractionation in all zones. When the column started up, some of the initial distillation data showed a naphtha-splitter bottoms gasoline endpoint as high as 575 F., with typical values as shown in Table 4.

The FCC gas plant initially separates the FCC gasoline in a naphtha--stripping column, producing light-gasoline overhead and a bottom product. Part of the bottom product is sent to a heart-cut splitter in the reformer and the remainder is sent to storage.

The heart-cut splitter separates this stream into light, heart-cut, and heavy gasoline streams. The light and heavy cuts are refinery gasoline blendstocks and the heart-cut stream is reformer feedstock.

A test run was conducted on Oct. 7, 1987 at 93,700 b/d charge. Table 5 shows the raw gasoline and LCO-product distillations.

The column was producing high-endpoint gasoline and a light front-end on the LCO product. The gasoline losses to LCO product were estimated to be 2,400 b/d, assuming the main fractionator operated correctly.

The main problem has the high-endpoint gasoline. Of the total FCC gasoline production, about 7,000 b/d had to be blended to the refinery middle-distillate pool year round (Fig. 7). (During winter months, the refinery operated in maximum middle distillate mode, and the 7,000 b/d was typically needed for middle distillate production.)

During gasoline season, 14% of this unit's gasoline production was being lost to the middle-distillate pool, which had a major impact on refinery economics.

The heaviest portion of the FCC gasoline has a low Rvp relative to average FCC gasoline. Its impact on the total refinery gasoline pool is thus greater than the volume effects (loss of 7,000 b/d). The two possible solutions were to replace the packing with trays or fix the packed column.

TROUBLESHOOTING

The three primary reasons packed refinery main fractionators do not perform well are poor liquid distribution, flooding, and poor vapor distribution.

Fig. 8 shows the combined top-pumparound liquid collector and gasoline fractionation section distributor. The top pumparound rate was about 6,000 gpm while the internal liquid rate to the fractionation section was about 3,500 gpm.

The use of a combined pumparound collector and liquid redistributor device like this has never been successful in a large-diameter refinery fractionator. To illustrate, Table 6 summarizes the observed HETPs of the fractionating beds.

Theoretically, the operation of this device can be shown to produce poor liquid distribution and no remixing. But how can data be gathered to support the poor liquid distribution theory, when the column lacked instrumentation and no temperature measurements were available?

The column head and top-pumparound draw temperatures before the revamp were available. Table 7 shows the recorded temperature at about the same top column pressure before and after the revamp.

The top pumparound draw temperature was much higher after the revamp, indicating high-endpoint material at this point in the column. The pumparound draw temperature is approximately its bubblepoint; therefore, the higher draw temperature implies high endpoint.

Previous distillation analyses of the top pumparound draw were not available.

The column had no thermowells, but skin temperatures measured with a portable thermocouple under the insulation can be used to infer column internal temperatures. If there is a thermocouple on the column, the offset between the vessel skin temperature and column internal temperature is known.

If skin temperatures are measured radially at a given elevation (either above or below a packed bed in the vapor space) the magnitude of maldistribution can be inferred from the temperature differences of the radial measurements.

Fig. 9 shows a radial survey taken below the top collector/redistributor. The temperatures varied from 495 F. near the draw nozzle to 360 F. directly opposite the pumparound draw nozzle. Fig. 10 is a radial skin temperature survey taken above the LCO product draw tray.

There was severe liquid maldistribution throughout the column and little liquid mixing at the collector trays, Compositional gradients caused by the maldistribution were never corrected at lower elevations. (A column that has one bad distributor and good mixing at the lower redistributors eliminates the propagation of the composition gradients.)

In an effort to reduce investment and increase packing bed depth, combined liquid collector/redistributors were used. (An orifice pan distributor should never be used in large-diameter refinery columns.)

The device had a wide sump with a draw nozzle at one end. There were also no drip points in the sump, leaving a large part of the packing without initial distribution. The pumparound collector/distributor probably had a significant liquid gradient from the side opposite the pumparound draw to the draw nozzle.

An attempt was made to modify the collector/distributors by installing drip tubes in the sump and using a second nozzle opposite the first. These modifications made no apparent improvement in column performance.

GLITSCH REVAMP

Lyondell decided to revamp the existing packed column. The operating data from the 1987 test run were used as the design basis. A new design-basis heat and material balance was used to establish product yields and column design internal loadings.

DESIGN/INSTALLATION

Table 8 shows the new, estimated design yields for the main fractionator. The raw-gasoline design endpoint was set at 455 F. The gasoline product yield improvements were based on increasing the fractionation in the gasoline/LCO fractionating bed to four theoretical stages.

The column heat balance is shown in 'Table 9. The estimated product-quality comparison between actual performance and Glitsch's design is shown in Table 10.

Fig. 11 illustrates the proposed Glitsch design. The new design uses separate liquid collectors and redistributors and less packing.

A major misconception is that collector-distributor spacing should be sacrificed to gain more packing. The assumption is that the packing HETP is important and more packing gives more efficiency, even when the distributor design is compromised.

Approximately 25% of all refinery large-diameter packed main fractionator revamps involving fractionation do not meet design objectives, primarily because of poor liquid collector and redistributor system designs. In one case, the liquid collector design caused a major product yield loss on a large-diameter vacuum column. 3

This case is not unusual. Packing HETP in large-diameter main fractionators is driven by the internals design; inherent packing efficiency in a 4-ft diameter column is of little relevance.

When discussing commercial HETPs in large-diameter refinery fractionators, actual column performance with a given collector/redistributor system is the important issue. Packing HETPs in a C6/C7 system are not relevant unless HETP claims are very low.

When designing a liquid distributor for a 24-ft diameter column, the following items need to be evaluated:

• Liquid rate

• Distributor feed method
  • Feed pipe

  • Liquid collector (such as a pumparound or product draw)

• Mechanical requirements
  • Support

  • Installation

  • Leveling.

The liquid rate is a critical design parameter on a trough-type distributor. (Glitsch does not know of a successful application of a pan distributor in a large-diameter main fractionator.)

In a lube vacuum column or an LCO/HCO fractionating bed in an FCC main fractionator, the liquid rates are approximately 2 gpm/sq ft of tower area. In the gasoline/ LCO section of an FCC main fractionator and the light/heavy naphtha section of an atmospheric pipe still, the liquid rates are 6-12 gpm/sq ft of tower area.'

These two types of distributors will not be exactly alike. The higher-liquid-rate distributors have very different momentum and horizontal-velocity considerations. At low liquid rates, these collector/redistributors are easier to design.

There are no fixed rules that apply to all distributors. Each collector/redistributor system must be evaluated separately because, among other factors, the number of draw nozzles, draw rates, available space, and internal liquid rates are all different.

For this column, the gasoline/LCO liquid distributor was designed for 3,500 gpm with turndown to 1,750 gpm. It is possible to achieve higher turndown with multiple levels of orifice holes or slots, but the quality of distribution cannot be maintained across the entire operating range.

Distributors are susceptible to maldistribution resulting from level differences in the trough. It is always true that higher-turndown distributors do not have the distribution quality of lower-turndown distributors.

The liquid distributors were designed with one parting box. Multiple parting boxes look good on paper, but there are many practical problems with designing a large-diameter multiple parting box distributor. Multiple parting boxes should be avoided except in high-liquid-rate, very large diameter columns.

The troughs and parting boxes were all leveled to within - I/s in. The complete distributors were shop-tested at the design flow rates.

The parting boxes were tested by feeding the liquid into the parting box in the same manner as the actual column design configuration. This is important because, although Glitsch has made many distributors, design modifications were necessary in several cases. The changes were relatively minor, but the quality of distribution improved greatly afterward.

Liquid entering parting boxes has momentum and the means of feeding the parting boxes should minimize horizontal velocity in the parting box. Aeration and horizontal velocity are major causes or poor liquid distribution; parting boxes should use some type of calming zone to reduce their effects.

The head in the parting box should be adequate over the entire operating range so that horizontal velocity is not a problem. Although short parting boxes are less costly and do not require as much column height (more packing), in large-diameter columns-especially at higher liquid rates-it is disastrous to use short parting boxes.

Higher liquid rates require more elaborate and more costly designs. There is no standard distributor design for column revamps. Each design uses the same basic principles, but each is different.

The number of draw nozzles on the collector above a distributor differ, which significantly impacts the liquid redistributor design. All the fractionation-bed liquid distributors in this column are fed internally from collector trays Fig. 12 . This is typical of refinery columns having multiple product draws and heat removals.

The column internals were designed to be installed level. Sections of the collector tray and the parting boxes and distributor troughs are all part of the liquid-distributor system. The weirs feeding the liquid from the collector trays, parting boxes, and troughs were installed with water levels.

INSPECTION

A checklist of all the items to be field inspected was made before the shutdown. An experienced engineer who knows what to look for should inspect the column internals before the man-ways are closed.

Many columns fail because there is no one person responsible for this activity. A column should not be inspected by committee or by an inexperienced engineer because the job is critical.5

Inspections also should not be made for the sole purpose of identifying that the equipment was manufactured per the drawings. Sometimes something is designed incorrectly, but built and installed correctly.

The inspection is an opportunity, to catch mistakes that cause shutdowns. Many columns have to be shut down for modifications, which is much more costly than correcting an error before start-up.

PERFORMANCE

The column internals were modified in early 1992, after several delays. A detailed test run was planned to determine the actual packed-bed efficiencies.

Before the test runs, the meters were zeroed and calibrated and a material balance was attempted. (Do not perform a test run unless the unit material-balances for a few dan,s, and never use data that do not material balance to do a computer simulation.)

Once the material balance was established, a heat balance around the fractionator was performed. The data heat-balanced well and a full test run was scheduled.

The unit was operated as stably as possible for 24 hr. A full set of stream samples was taken every 8 hr for laboratory analysis. Material and energy balances were performed on these three sets of data. The resulting data were then used to run the computer simulation.

This main fractionator has no metered reflux streams; therefore, consistent heat and material-balance data were needed to determine the efficiency of the gasoline/ LCO fractionation bed.

On a main fractionator with external reflux from the overhead receiver, it is possible to check the accuracy of the heat-balance data by letting the simulation close the heat balance on the reflux. If the wash oil rate, or another internal reflux stream, is measured because there is total draw, then a good check of the column heat balance is possible.

The top pumparound design basis was 204 MMBTU/hr, which resulted in an internal reflux rate of 3,200 gpm. The column was operated at approximately 170 MMBTU/hr with an internal reflux of about 2,500 gpm.

The column was operated with lower-than-design internal reflux because the top pumparound heat removal was limited by pump circulation and control problems. The top pumparound loop has three series heat exchangers with bypasses around each exchanger.

Table 11 shows the product qualities for the trayed column, initial revamp, Glitsch design, and actual performance of the most recent revamp. The column has met design objectives (Table 12).

The separation is less than design because the column reflux ratios are below design. The top pumparound pumps will be modified during a future shutdown so that column internal reflux can be increased.

The column now responds to reflux changes and when there is high heat removal in the slurry pumparound and the reflux drops below the distributor minimum, the gasoline endpoint increases beyond an acceptable level.

The column now responds to operational changes and the product qualities reflect these changes. After 4 years of running a column that did not respond to such changes, it has taken the operators some time to become comfortable with the column control.

REFERENCES

1. Golden, Scott Sloley, Andrew W., Martin, Gary R., "Revamping FCC Main Fractionators Effects on Unit Pressure Balance," Hydrocarbon Processing, March 1493, pp. 7,-81.

2. Golden, Scott W., "Integrating theoretical and practical aspects of refinery FCCU fractionator revamps," 1989 AlChE Spring National Meeting, Houston, Apr. 2, 1989.

3. Lieberman, Norman P., and Lieberman, Elizabeth T., "Design, installation pitfalls appear in vac tower retrofit," OGJ, Aug. 26, 1991, P. 57.

4. Golden, Scott IN'., and Costanzo, Stefano, "Commercial Performance Data for Structured Packing in Refinery Atmospheric Crude Columns," AlChE 1990 spring national meeting, Orlando, Fla., March 1990.

5. Lieberman, Norman P., and Lieberman, Elizabeth T., "Inadequate inspection cause of flawed vac tower re%-amp," OGJ, Dec. 14, 1992, p. 33.

6. Golden, Scott "'., and Sloley, Andrew W., "Simple methods solve vacuum column problems using plant data," OGJ, Sept. 14, 1992, P. 74.

7. Kister, Henry Z., Distillation Operations, McGraw Hill, New York, 1990.

Copyright 1993 Oil & Gas Journal. All Rights Reserved.