Scott W. Golden
Glitsch Inc.
Dallas
Norm Lieberman
Process Improvement Engineering Inc.
Metairie, La.
Gary R. Martin
Consultant
Euless, Tex.
Coking in petroleum refinery main fractionators causes unit shutdowns and significant production losses. High temperatures and long residence times are the causes of coking in these units-which include crude vacuum columns, fluid catalytic cracking (FCC) and delayed coker main fractionators, and visbreaker and hydrocracker vacuum columns.
Coking, however, is not inevitable. It is a function of design and operating errors. If the equipment is not designed properly, the unit will coke, regardless of how it is operated.
As operating severities of refinery primary fractionators increase to improve operating economics, the incidence of coking also increases. If the consequences of more severe operation are ignored, the units will coke up and have to be shutdown (or very poor product quality must be accepted).
But higher unit distillate yields can be achieved through better design practices, and unit reliability can be improved while meeting higher distillate recoveries.
PACKING
Poorly designed vacuum columns can produce heavy vacuum gas oil (HVGO) products containing as much as 20 ppm vanadium or 2,000 ppm asphaltenes. Designs that work with low temperature, high overflash, low conversion, and high recycle simply do not work when the unit is operated at higher temperature or lower liquid rates.
The use of packing in refinery main fractionators and low-liquid-rate service is common. Using packing in the desuperheating section of an FCC main fractionator has process, efficiency, and mechanical advantages over shed or disc-and-donut trays. But while packing has the advantage of inherently low hold-up or residence time, it is much less forgiving. Engineers must customize equipment design to the specifics of each system.
The use of pacing-sizing features from flow-sheet models does not address the realities of these sorts of revamps. Packing efficiency and distributor drip-point density do not cause column shutdowns.
Several packed column case studies are presented in this article which detail the cause of several unit outages. Understanding the causes of coking allows the designer to properly design the equipment and avoid unplanned shutdowns.
COKING
Coke formation in petroleum refinery packed columns is caused by several design errors, including:
- Adequate vapor distribution-localized "dry-out"
- Ultralow local liquid rate-high residence time
- High entrainment coke fines or visbroken vacuum residue
- Inadequate liquid mixing-thermal and composition gradients
- High residence time on collector trays.
Coke formation is a function of temperature and residence time. The design issues surrounding refinery main fractionators, however, dramatically affect both residence time and localized temperature.
The steady-state or ideal temperature at any elevation in a column is controlled by the unit operation. Oil cornposition and pressure set average bulk temperatures, and localized temperature is a function of unit operation and equipment design.
Coking can be reduced or eliminated, but ignoring design issues is not the solution.
Revamping low-recycle delayed cokers to control heavy coker gas oil (HCGO) quality is an example that requires some thought. A "standard" design that works well at 11-D-20% recycle does not have any chance of working at 4% recycle. Using only a spray chamber to meet the low recycle rate does not address the low efficiency and relatively poor HCGO quality that result.
Packing can be used as-an alternative; however, an appropriate design must be selected. Packing has very low intrinsic residence time compared to other fractionating devices used in these services (such as bubble-cap, sieve, shed, disc-and-donut, or turbo-grid trays).
All these devices (as will packing) will coke, but none of them will meet the requirements imposed by the more-severe operating conditions being discussed.
By definition, packing is the only viable solution when the liquid rate is very low and product quality must be controlled. And although the use of spray chambers is common, product quality suffers.
VAPOR MALDISTRIBUTION
In a coker, coke-drum temperature, drum vaporline quench, and recycle determine the average temperature in the wash zone. But liquid and vapor distribution affect localized temperatures.
These real, albeit nonideal, relationships are manifested in operating units. But steady-state computer simulations do not reveal this. Vessel radial skin-temperature measurements, however, will show gross vapor maidistribution in many columns.
Wash sections of all refine main fractionators are characterized by very low liquid/vapor ratios. These wash zones are subject to significant revaporization of wash oil.
The objective of all wash sections is to minimize total overflash. This maximizes distillate product yield, whether on a delayed coker, atmospheric crude, crude vacuum, or visbreaker vacuum column.
If the vapor feed is maldistributed, the areas where the liquid/vapor rate is lowest will have the highest revaporization of the wash oil. And vapor maldistribution causes coking.
ULTRALOW LIQUID RATE
The coking in many vacuum columns occurs in the middle of the wash bed, not at the bottom. Computer calculations tell us that the highest temperature and lowest liquid rate are in the bottom of the wash bed. The temperature is not highest in the middle of the bed; but the liquid rate is lowest in the middle.
In this instance, the time temperature relationship of coking is the issue. The bottom of any wash section is partially wetted from entrainment. The packing liquid rate, therefore, is relatively high.
On many unit designs, the middle of the bed has very low flow because the design wash-oil flow rate is incorrect. The resulting high residence time in the middle of the packed bed causes coking.
INADEQUATE MIXING
FCC main fractionators have relatively high liquid rates (8-20 gpm/sq ft column cross-sectional area). The feed to the column is superheated vapor (960-1,040 F.), and the feed velocity to the column is 125-200 fps.
Typically the slurry section has no vapor distributor because these vapor distributors tend to coke. The heat removal in the slurry section comes from the pumparound circulation and the liquid entering the slury section from the wash section.
The liquid from the wash zone has a much lower endpoint than the slurry product (pumparound liquid), and a large portion of this stream revaporizes in the slurry pumparound. If the two liquid streams entering the slurry section are not distributed adequately, sections of the bed will have a much lower liquid rate and the bed will coke.
HIGH ENTRAINMENT
Visbreaker vacuum column feeds are thermally unstable. The degree of instability is a function of the unit conversion; however, the design of the flash and wash zones in these columns must be correct or unit run lengths of 6 weeks or less will result.
Designing these systems requires good vapor/liquid disengagement in the flash zone, good vapor distribution, and reliable liquid distribution. All of the following result in coking:
- Deep beds of packing that can dry out
- Poor vapor distribution causing localized devout
- High entrainment of unstable residue
- High collector-tray residence time
- Recycle of unstable material to the furnace.
CASE I
The disc-and-donut trays of a large FCC main fractionator slurry zone were replaced with grid to improve the efficiency of the pumparound section (Fig. 1). The objective was to decrease the light cycle oil (LCO) content of the bottom stream.
The unit ran well for about 3 months, then the pressure drop across this section of the column began to increase. The pressure-drop increase affected the reactor/regenerator pressure balance and caused operating difficulties in the unit.
The cause of the problem was stated to be "too high a heat transfer coefficient." The initial solution was to reduce the packing volume.
One analysis that was performed assumed the bottom section was a pure heat transfer section. The calculated bottom temperature was 875 F. The fact that the bottom stream of an FCC is slurry product and the stream above this is LCO which is blended to the middle distillate pool, was not considered.
The slurry bottom product composition is set by the reactor effluent (conversion) and the fractionation between the LCO and slurry. The calculated initial boiling point required to meet a bottom product composition equal to an 875 F. bottom temperature was about 790 F.
Composition and pressure determine temperature. The bottom temperature (composition) is related to the column's heat and mass balance, not to the heat-transfer coefficient of the packing.
Packing efficiency affects composition; therefore, it affects bottom temperature. The range of temperature impact may be as much as 30 F. Until the LCO product endpoint is increased to 900 F., it is not possible to have an 875 F. bottom temperature.
CAUSES OF COKING
FCC main fractionator slurry sections have relatively high liquid rates. Although the feed temperature to the column is about 980 F., the packing should be well wetted.
Liquid distribution can affect localized temperatures. This unit's distributor design has been used in several units. The quality of liquid distribution did not, however, cause the problem.
The wash liquid from the trays above the packing is fed into the distributor from the one-pass trays, as shown in Fig. 2. Fig. 3 is a schematic of the bottom section of the fractionator.
The bottom liquid temperature is controlled by the quench stream fed directly into the bottom section of the column. The bottom temperature is set at some maximum temperature to avoid coking (
The slurry pumparound flow can be varied to change the column heat balance. For a given slurry-section efficiency, the liquid temperature leaving the slurry packed bed is controlled by the heat balance.
The assumption here is that the refiner is maximizing the LCO draw to an endpoint constraint. The temperature of the liquid leaving the bottom section of the column is controlled by the heat (and mass) balance, not by the packing heat-transfer coefficient.
When the packing was installed in the column, the amount of quench required to maintain the bottom temperature, for the same heat balance, increased. The packing was more efficient and the slurry bottom product contained less light material.
LIQUID DISTRIBUTION
The distribution of liquid to a packed column is important. Without good distribution quality, the inherent packing efficiency for any given service decreases.
The slurry pumparound section desuperheats the reactor effluent vapors so that the oil can be fractionated. The zone's major purpose is desuperheating and, to this end, liquid distribution is important.
The slurry zone, however, is a severe operation. The vapor entering the column contains catalyst fines and the liquid distributor must operate for 3-5 years without problems. The distributor must be reliable, first, and high-quality, second.
The commercial results over the past 5 years have been very good in this service. With the exception of the case presented, slurry pumparound distribution quality has not been a problem.
LIQUID-MIXING
Compositional gradients resulting from inadequate internal liquid-flow redistribution have been known to cause columns with high theoretical stage counts to have poor apparent efficienCY.
Analogous to that is compositional gradient in a slurry pumparound, which results in coking. Varying liquid rates across the cross sectional area of the column can cause localized "hotSpots."
The wash oil and the slurry pumparound liquid TBP distillations for a typical unit are shown in Table 1.
The wash oil has a lower boiling range than the slurry pumparound return liquid. The liquid leaving the wash section is also bubble point liquid at 640 F. (Fig. 4).
The slurry pumparound is liquid subcooled by 200-300 F., depending on the heat sinks on the pumparound. The impact of the thermal and composition gradients on the liquid rate in the bed is significant.
COKED PACKING
The bed coked because the liquid from the wash section and the slurry pumparound were not mixed (Fig. 5). The liquid distributor in this column preferentially fed the wash liquid to one section of the packing and the slurry pumparound liquid to another.
Simplification of the mechanical design resulted in a fundamental process design error. Inspection of the packing showed localized coking in the area where only wash liquid was distributed.
Modifications to the liquid distributor allowed uniform mixing of the wash liquid and slurry pumparound. The unit has operated for 2 years without coking.
CASE 2
A revamp of a delayed coker main fractionator wash section to increase unit charge and reduce recycle rate caused a unit outage after less than 6 months of operation. Four sieve trays were replaced with an 8-ft bed grid. Fig. 6 shows the column before and after the revamp.
To reduce the initial revamp costs, the existing collector tray under the packed bed was left in the column and the shed trays were not removed. The packed-bed pressure drop increased from 6 in. of water to 32 in. over the 6-month period.
The asphaltene level in the HCGO product increased, resulting in high intermittent carry-over of coke fines to the HCGO. The coke fines plugged the downstream hydrotreater feed-filter system and caused chronic operating problems.
The initial shutdown replaced the grid and focused on the wash oil quantity. During the initial run with the grid, the wash oil was lost briefly on several occasions. Loss of wash oil was thought to be the problem.
TROUBLESHOOTING
During the first run, no field data were gathered to evaluate the cause of the problem. During the second run, the bed pressure drop began to increase. At this point, the cause of the problem was determined to be the packing.
A field troubleshooting effort was launched to evaluate the cause of the coking. The reason for using the grid was to reduce the recycle rate on the unit, Trays require 15-20% recycle and, even at these rates, trays have coked on other units.
During the field tests, the process stream flows were evaluated. Temperature and pressure surveys were conducted to see what was happening.
Many times, "office" engineering approaches are used to troubleshoot, and the unit operators are not consulted. Process equipment and process flow interact and, without understanding both, it is impossible to troubleshoot a unit.
The stream flows for this unit are shown in Fig. 7.
Cold HCGO was fed to both the wash bed and to the shed trays. The cold HCGO feed to the shed trays, in theory, cooled the rising vapor prior to it entering the wash section.
The column had three separate 4-in. nozzles feeding the six-pass shed trays. The cold HCGO feed to the wash bed was 9,000 b/d. Cold HCGO feed to the shed trays was 5,000 b/d.
PRESSURE SURVEY
A pressure measurement on the cold feed to the wash zone spray header showed that flow and pressure-drop were consistent. For a given number and size of nozzle, if the pressure drop is not close to the manufacturer's data, there is a problem in the header.
The pressure drop across the cold HCGO feed header to the shed trays measured 120 psig. The control valve was wide open. Two of the three feed lines to the column were cold.
The calculated pressure drop across one feed pipe was 5 psig. The only operating line was badly plugged.
The cold HCGO feed system was designed by the coker-technology company to cool the hot drum vapor before it entered the sieve trays. All 5,000 b/d were being "dumped" in one section of the column. The calculated feed-line velocity, assuming all three feed lines operated, was less than 1 fps.
TEMPERATURE SURVEY
A radial temperature survey on the vessel shell at an elevation just - below the packed bed showed the temperature variation to be 40 F. (Fig. 8). The wash-zone liquid rate at the bottom of the bed was only 0.5 gpm/sq ft.
The liquid rate to the bed is twice this number. In other words, half the liquid revaporizes. This assumes that the gas flow and temperature to the packed bed are uniform.
But what happens to the dry-out rate of the bed when the vapor is maldistributed and the temperature is not uniform? Localized sections of the packing have little or no liquid, and coking begins.
SOLUTION
Assuming only the vapor is maldistributed because of poor collector-tray design (Fig. 9) addresses only part of the problem. The liquid flow to the sheds consists of several streams, cold HCGO being the largest.
Why would one put wash liquid across the shed trays unless they assumed temperature caused coking?
Vapor temperatures in an FCC main fractionator are 1,000 F. Packing does not coke at the temperatures in a coker fractionator unless the oil has a high residence time. To get a high residence time in the grid, given its inherent low hold-up, requires either flooding or extremely low liquid rates.
If the vapor is very poorly distributed, it will locally flood the packing. If the vapor is maldistributed to a low-liquid-rate section of any refinery wash section, it causes high dry-out rates and very high local residence time. The oil sits and cooks and cokes.
The first recommendation was to remove the cold HCGO from the shed-tray section (Fig. 10). The shed trays served no useful purpose; then" were a "left-over" from the era of high recycle and sieve-tray wash sections. Most refiners remove these trays because they constantly coke and often are mechanically damaged.
Interestingly enough, the weir length (picket weirs) on a multiple-pass shed is so large that the shed does not distribute the liquid anyway. Delayed coker main fractionators should never have shed trans.
The history of this column is quite long and involved and the story is not over N,et. Although the grid coking was solved by eliminating the shed-tray section cold HCGO (all of it goes to the wash section) and installing a well-designed collector tray, the collector tray still cokes.
The decision to leave multiple-pass shed trays in the column results in a collector design that must distribute liquid to the trays belong. Vapor temperature variation, resulting from maldistributed liquid, was one of the major causes of the observed temperature gradient. The other was the inherent poor distribution of the shed.
With the sheds in the column, the collector tray below the packed bed must distribute liquid or it will cause temperature variation. Once the shed trays are removed, the collector tray can be desrned to drain any coke fines carry-over during drum switches or normal operation.
CASE 3
Deep-cut, vacuum-column revamps are common today but many units shut down to replace coked wash sections. Often these columns are shut down more than once because the cause of the problem is not determined correctly. The problem can be solved, however, with little or no loss of yield and good product quality.
Understanding that higher wash-oil rates do not result in higher overflash or, conversely that lower wash rate does not materially affect gas oil yield, makes little sense in light of computer design methods. Evaluating plant operating data, clearly shows that wash-zone design and operation on many units is wrong.
The cause of wash zone coking is:
- Incorrect design washoil flow rate
- Incorrect overflash rates.
WASH-ZONE COKING
Wash-oil rates twice the design flow are needed on many deep-cut revamps to meet the design overflash rates. Wash-zone coking on many of these vacuum columns is uniform across the tower cross-sectional area (Fig. 11).
The coking is caused by inadequate wash-oil rate to the packing. Distributed wash-oil is essentially all revaporized. The small quantity of oil not revaporized has very high residence time and it cokes.
Pressure drop increases of 2 mm Hg indicate the onset of coking, assuming the column vapor rate remains constant. Gas oil quality does not deteriorate until the pressure drop increases by 4-5 mm Hg. At this point, the gas oil contaminants will increase significantly.
When the pressure drop increases to 8-10 mm Hg, gas oil quality is very poor. Pressure drops as high as 40 mm Hg have been measured across some wash sections.
Wash-oil spray headers are generally sized for 15 psig pressure drop at design conditions. Nozzle type will dictate the onset of severe atomization; however, pressure drops greater than 30 psig will cause a significant amount of atomization. Wash-oil rate increases greater than this result in little increase in the liquid rate to the packing.
The wash-oil spray header design flow rate must be correct, otherwise the bed will coke. Incorrect wash-oil rate calculations result from poor feed characterization and incorrect flow-sheet modeling.
Typical models assume the vapor/liquid in the flash zone are in equilibrium. But this is an incorrect assumption. Depending on the transfer line design, the impact on the wash--oil rate varies.
Vapor entering the wash zone of a vacuum column is superheated. The quantity of superheat can be substantial, and this causes the large dry-out rates observed.
Correct feed characterization and flow-sheet modeling will yield better wash-oil rate estimates. Assuming the wash-oil spray-header design is correct, determining the correct wash-oil rate to meet an overflash target is very important.
Slop wax contains entrainment and overflash. The metered flow, however, is not overflash. In fact, on many units, entrainment is greater than 5017, of the slop-wax flow rate. Assuming all this liquid is overflash will result in determining the wrong wash-oil rate.
Controlling wash-oil rate based on gas oil quality is not a good idea. Good gas oil quality can be achieved with very low wash-oil rates; however the wash bed will eventually begin to coke.
OVERFLASH RATE
Estimating the overflash and entrainment rates should be used to control the wash rate. Samples of the liquid on the slop-wax collector tray and residue should be analyzed for contamination.
Material-balancing the flow rates and a contaminant will yield two equations and two unknowns from which to determine an estimate. An example of such a contaminant is sodium, which is nonvolatile. Low-level analysis, however, is difficult.
Conradson carbon or microcarbon is the easiest analytical test, but these are volatile contaminants and an estimate must be made of the volatile content in the overflash.
For typical crudes (not high-vanadium crudes like Mava and Venezuelan crudes), assuming the overflash contains 10-15 times the HVGO vanadium content will yield conservative results.
Asphaltenes are not volatilized except in very extreme vacuum-column operation. An asphalt balance also can be used.
VACUUM TOWERS
Both visbreakers and hydrocrackers convert vacuum residue to lighter components. A typical percent conversion for each (based on 1,000 F. and lighter products) is:
- Visbreakers, 40%
- Hydrocrackers, 80%.
To achieve maximum field of clean liquid products, many refineries have installed vacuum towers to reprocess the thermally degraded tar. Two case histories illustrate the central problem with such vacuum towers: coke formation of the wash oil grid.
CASE 4
This Gulf Coast unit had a long history of producing black distillates from the hydrocracker vacuum tower. Even though the flash zone entrainment velocity was moderate (about 0.3 fps), the HGO contained 4 wt % Concarbon and the LGO, 1.5 wt % Concarbon. The HGO was so poor in quality that it was redistilled in the plant's virgin vacuum column.
When the column was opened, the wash-oil grid was coked to the consistency of concrete. The authors' analysis had previously indicatea that the wash-oil section design could be improved by correcting several design flaws:
- Nonoptimum design of vapor horn
- Oversized chimneys on overflash pan
- Too high a residence time on overflash pan
- Too much wash-oil liquid hitting the walls of the tower above the wash-oil grid
- Too low a wash-oil rate during initial unit operation.
When the recommended modifications were installed, the initial results were excellent. The HGO Concarbon dropped to 0.4 ivt %. After 3 months of operation, however, the wash-oil grid delta P started to increase and the HGO Concarbon also began to escalate.
Why? The overflash recycle pump had never been commissioned. While the overflash pan did have a provision for internal overflow, the authors had made a mistake in design (Fig. 12): other overflash draw-off sump had a high residence time before it would overflow.
- If the overflash pump was running, the liquid level in the sump was low, and coking would be retarded.
- When the pump was not operating, the draw-off sump coked. With a virgin vacuum column, the temperature of this sump would not have resulted in a high rate of coke formation. But when processing thermally degraded feed, coking proceeds much more rapidly.
- Once the draw-off sump coked, access to the overflash pan's internal overflow pipe was blocked.
- The overflash pan began to overflow through the vapor chimneys, thus promoting re-entrainment of the overflash liquid.
- The re-entrainment caused the wash-oil grid to coke-off.
When the tower was shutdown, the wash-oil grid was replaced. Since that time, the overflash pump has been run continuously and the HGO quality has remained excellent.
The lesson from this incident is that a 100% reliable provision for internal overflow of overflash liquid is critical for vacuum towers processing thermally degraded feedstocks.
CASE 5
An overseas refinery constructed a new visbreaker vacuum column. The design was based on tower internal drawings from an operating visbreaker vacuum tower in a nearby plant. The new vacuum tower copied the radial feed-entry design, combination grid, and structured packing of the older unit.
When the new tower was commissioned, the HGO was black and contained 2 wt % Concarbon. Two months later, the unit was shut down because of a 25 mm Hg delta P through the wash-oil grid, and three times the initial HCO concarbon (6 wt %).
The grid was removed and the wash oil nozzles were modified. Initial operations have produced a black HGO with 2 wt % Concarbon. It is the authors' opinion that the central problem with this design is the radial feed entry (Fig. 13). This type of entry is believed to have promoted resid entrainment and, as a consequence, wash-oil grid coking. The authors' standard recommendation is a tangential entry.
Unfortunately, the prototype used for this design also produced a black, high Concarbon HGO. While using an existing unit as a basis for a new design is not proper, one should make sure that the unit being copied is actually meeting expectations.
Copyright 1994 Oil & Gas Journal. All Rights Reserved.
Issue date: 11/21/94
