BTX: PROBLEM AND SOLUTION-Conclusion: CBUs eliminate BTX-induced catalyst deactivation

Nov. 5, 2007
The first part of this two-part series describes how Saudi Aramco faced chronic Claus catalyst deactivation for years as a result of benzene, toluene, and xylene (BTX) in lean acid gas feed to several of its sulfur-recovery units (SRUs).

The first part of this two-part series (OGJ, Oct. 22, 2007, p. 60) describes how Saudi Aramco faced chronic Claus catalyst deactivation for years as a result of benzene, toluene, and xylene (BTX) in lean acid gas feed to several of its sulfur-recovery units (SRUs). Construction of seven BTX removal units using regenerable activated carbon beds was completed in December 2005; commissioning took place in spring 2006.

This concluding article will discuss design issues, operating experience for the units, and their performance and effect on the downstream Claus catalyst. Fig. 1 shows one of the carbon-bed units.

This is one of the seven carbon-bed BTX-removal units (Fig. 1).
Click here to enlarge image

In brief, catalyst deactivation has been virtually eliminated. This has set the foundation to allow us to revamp the units to achieve higher recovery, which would not have been possible until catalyst deactivation had been resolved.

Design issues

A process selection study carried out in 2000 identified carbon beds as the most economical process solution. That evaluation was based on a five-vertical-vessel design.

Number of carbon beds

From the outset the question of disturbance to the SRU’s reaction furnace and air-demand analyzer during vessel switchover was a concern. For minimizing this impact, the greater the number of vessels the better because adding vessels reduces the proportion of the total feed being diverted.

Of course this is offset by matters of practicality. As the number of vessels increases, so does the plot area, complexity, and cost, driven to a great extent because of increasing switching valve count.

At the detailed design phase of the project, a design with three horizontal vessels in lieu of the five-bed configuration was proposed. In this arrangement, two beds would be online and one regenerating or in standby. Initially there was a concern that half of the flow being diverted to the regenerated bed during switchover might cause disturbances to the SRU. At the end of the regeneration step, the carbon bed retains a lot of water.

When the vessel is brought online, the acid gas has a capacity to absorb water. Being at 50% relative humidity, however, acid gas leaves the bed saturated. Over time the bed is striped of water and is completely dried long before the end of the adsorption cycle. This water is carried directly the SRU.

The three-vessel proposal had significantly more carbon than the five-bed design. This resulted in the number of cycles per day being less than half than for the five-vessel case. The water spikes would be greater in absolute terms because half the flow is diverted to a freshly regenerated bed with a three-vessel configuration. On a daily basis, though, the total amount of water sent to the SRU by the carbon-bed unit (CBU) is slightly less with the three-bed design.

It remained to be determined if the three-bed configuration would cause unacceptable disturbances during switchover. A simulation of the reaction furnace’s flame temperature indicated a 40° to 60° F. temperature drop assuming half the acid-gas feed was saturated with water at the bed conditions after regeneration. We thought this to be acceptable and proceeded with the three-bed design that had been recommended.

The recovery loss resulting from the addition of water was calculated to be 0.05%. (By the LeChatelier principle, water shifts the Claus reaction equilibrium unfavorably.) There is also a small amount of acid gas that is lost each time a vessel is regenerated.

As mentioned in the process description in Part 1, the acid gas in the vessel at switchover is displaced through the regeneration condenser to the pressure control drum. These noncondensables are sent to the flare system. With the net vessel volume and the number of cycles per day, the recovery loss from this amount of flared acid gas was calculated at 0.028%.

Design parameters

Designing a CBU requires several parameters be known. The first, capacity, defines how much heavy hydrocarbon can be retained per mass unit of carbon. Beyond that amount the mass-transfer zone moves up the bed. The capacity of typical industrial carbons expressed as a percentage is a single digit number.

The good news is that it is a function of concentration in the stream to be treated. This gives the process an inherent flexibility. If one designs for a given level of aromatics using the manufacture’s recommended capacity and the feed later has more contaminants, the capacity will also be greater at the new feed conditions. The capacity may not increase in direct proportion to the aromatics composition, but there is a mitigating effect.

Sizing of the vessels is done to achieve a target superficial velocity through the bed when in adsorption mode. Design guidelines have been reported in an earlier paper on this subject (OGJ, June 24, 1996, p. 31). The superficial velocity is calculated at flowing conditions as if there were no carbon loaded in the vessel.

Pressure drop through the bed is an important consideration because usually there is a limit on how much can be tolerated before the SRU capacity is reduced. The biggest sources of the pressure drop are the acid gas preheater and the beds themselves. The CBUs were designed for a total pressure drop of less than 2 psi.

The bed’s pressure drop is calculated from Ergun equation parameters provided by the carbon supplier. A design is developed by varying superficial velocity and bed height until a reasonable configuration is reached.

Removal efficiency refers to the percentage of each contaminant adsorbed from the treated stream. When the bed is dry the removal efficiency for BTX is 95-100%. Process conditions are maintained to enhance efficiency: lowest possible adsorption temperature balanced against relative humidity.

Bed regeneration is accomplished with low-pressure steam. For design purposes a fixed amount of steam per unit of carbon is used (OGJ, June 24, 1996, p. 31). There is a trade off in deciding how fast the regeneration is performed.

If the set quantity of steam is passed through the bed in a short interval, the bed can be brought online sooner, which means less total carbon need be installed because it is in service proportionately longer compared to the regeneration step. Conversely, this would require the downstream regeneration steam condenser and three-phase separator to be larger to accommodate the higher rate.

Vessel orientation

Vessel orientation-horizontal vs. vertical-is not a process issue: Either can be used effectively. It is a matter of providing an enclosure for the carbon bed whose size has already been defined from the process parameters discussed above. What determines the choice of orientation is the effective use of steel to hold the bed.

One can calculate and compare the “overhead volume”-the ratio of vessel volume not containing carbon to the total volume-between vertical and horizontal configurations. For given bed dimensions this will be different.

Related to this is the practical consideration of providing enough space for man ways and vessel entry. On a horizontal vessel, beyond a certain size, the curvature effects become negligible. In this case the vessel does not have to be enlarged just for entry purposes. On a vertical vessel of any size, a minimum free board must be allowed for above and below the bed.

Eliminating drying step

As noted, the adsorption efficiency is improved when the bed is dry. For benzene, this effect is more pronounced than toluene and xylene. Our pilot plant tests showed that we could still achieve substantially complete removal of the toluene and xylene without cooling and drying the bed after regeneration before bringing it back online. By comparison, we found that only about 80% overall benzene removal could be accomplished if the bed was left wet.

Uncertainty about how residual benzene slippage would affect Claus catalyst was the major issue in the early stages of the process selection study. Our contract research at Alberta Sulfur Research Ltd. determined the effect of benzene on Claus catalyst to be negligible.1 2 As a result we did not design for bed cooling and drying at the end of the regeneration step.

The carbon bed designers advised that we should expect a brief spike of BTX in the initial 3-5 min of the adsorption cycle because of the wet bed. For reasons completely analogous to the water spike discussion above, the three-bed design proved superior to the five-bed configuration. Although the BTX spike is greater than it would have been with a five-bed design, the total moles of aromatic slippage is less with the three-bed design because it cycles less frequently.

Eliminating the cooling and drying step from the regeneration sequence yielded substantial cost savings to the project. It was possible to reduce the scope significantly from what it would have been otherwise. There was no need for a drying gas cooler, separator, and recompressor. In addition, the amount of carbon installed was reduced because the regeneration step was shorter.

Materials

All of the vessels within the CBU are made of 316 L stainless steel. The initial specification called for carbon steel with internal stainless cladding, however. As detailed design and procurement proceeded, we determined that using a single material for vessel construction would in fact be simpler overall. Delivery times were shorter, and the cost difference was not a determining factor.

There was one concern with stainless steel construction that required mitigation: the bane of stainless steel-chlorides. After considerable deliberation we concluded that external coating of the vessels would protect against environmental chlorides. This was applied, and the vessels were insulated.

Water samples from the upstream knockout drum revealed chloride concentrations between 5 and 20 ppm. We wrestled with the question of to what extent, if any, chlorides in the water from the knockout drum would be able to migrate past the feed preheater and on to the carbon beds. To do this, free water would have to be entrained from the knockout drum. The Cl- ion cannot exist as a free atom; it must be in aqueous solution to exist as a chloride. In the preheater, because acid gas is heated to reduce the relative humidity to about 50%, there is no free water.

To guard against the remote possibility that free water containing chlorides could migrate to the carbon beds and to promote effective steam condensate drainage, we instructed the fabricator to flush-grind weld seams and the outlet nozzle at the center of the vessel in the 6 o’clock position. This is the nozzle from which the initial condensed steam is removed.

Our metallurgical engineer noted that, provided there was no build up of stagnant water within the vessel, there would be far less opportunity for chloride attack. Furthermore, he advised that stainless steel is far less susceptible to chloride attack if there is no oxygen present. For process reasons the carbon beds are always kept completely free of air during operation; this gave us added confidence there would not be a problem.

Commissioning

During initial commissioning, reaction-furnace flame instability was a problem during switchover of a regenerated carbon bed. Since then, however, very significant progress has been made to reduce the magnitude of disturbances to the SRU during switchover. We are continuing to refine the sequence logic further to minimize disturbances to the SRU from operation of the carbon beds.

CBU performance

We begin with a discussion of the disturbance to the reaction furnace and air-demand analyzer caused by bringing a regenerated vessel online. The process data discussed reflect operation with the sequence logic we have developed.

Fig. 2 presents plots of reaction-furnace temperature and air-demand analyzer output vs. time for an SRU during switchover. The reaction furnace’s temperature drops by about 60° F., in line with simulation results discussed earlier. Visual inspection of furnace flame when vessels are brought online shows no noticeable change in flame pattern.

Click here to enlarge image

Switching vessels causes a brief disruption to the tail gas H2S:SO2 ratio (Fig. 2b). The period of off-ratio operation is short-lived and corrected in less than half an hour. Because the vessels are switched fewer than four times per day, recovery loss due to the transitory period of off-ratio tail gas is insignificant.

To obtain a qualitative estimate, consider Fig. 2b which shows the air demand peaking at 3.25% excess air. Fig. VIII-1 in H. Paskall’s Capability of the Modified-Claus Process shows recovery loss due to off-ratio operation in terms of percent excess or deficient air.3 Taking as a basis 96% recovery for a very lean feed SRU with perfect control, the recovery drops by 1/3% for an excess air of 3.25%.

To be conservative we assume the time of decreased recovery due to off-ratio operation is 2 hr/day, which is definitely a worst-case assumption because it takes the peak off-ratio value over the entire period. Taking 1/3% recovery loss for 2 hr out of 24-hr results in a daily net recovery loss of (2/24) • 1/3% = 0.028%. Although this value is very small, we are working on eliminating these spikes in air demand altogether.

As far as CBU performance itself is concerned, the capacity and removal efficiency as defined earlier has generally met or exceeded expectations. A surprise has been the extent to which benzene removal has surpassed 80% even during the early part of the adsorption cycle despite the fact that the beds are not being dried and cooled as part of the regeneration step.

The adsorption time to breakthrough is greater than design, indicating that the capacity of the carbon used in design was conservative. (Breakthrough is defined as the mass-transfer zone reaching the top of the bed.) There have been some unexpected analytical results between the two gas plants that we are reviewing.

At this point we have not determined with certainty if there is a real process difference or an unresolved analytical issue. As one can appreciate when analyzing for aromatic components in the single-digit ppm range, a minor change in reported outlet composition will have a significant impact on calculated removal efficiency.

Click here to enlarge image

Deactivation of Claus catalyst due to BTX poisoning has been eliminated or reduced to such a small value that it has not been possible to measure over 20 months. Graphs in Fig. 3 show the impact. These graphs show first converter temperature profile vs. time for selected SRUs at Uthmaniyah and Shedgum gas plants that had a CBU installed to treat their feeds. The performance of the other units is similar and was reported in the original Laurance Reid Gas Conditioning Conference paper.

Feed variations

A few words of background are necessary to interpret these results. At Uthmaniyah and Shedgum, the trains receive acid-gas from headers that are fed from gas-treating units that process both associated and Khuff gas. The mix between the two varies from day to day; as a result acid-gas quality is not constant.

Variations in acid-gas composition will change the first converter ΔT for a given reaction furnace bypass amount and catalyst activity. That is why there is considerable variation in the converter ΔT shown in the figures. Still, one can observe a clear downward trend for converter ΔT before installation of the carbon beds and the stable temperature profile afterwards.

To illustrate the point, Fig. 3c for Train 4 at Shedgum is discussed in detail.

The temperature scale for this figure, greater than in the other figures, suppresses the day-to-day fluctuations caused by changes in sour-gas feed mix and allows the general trend to be seen more clearly. As well, this train underwent two catalyst regenerations, which shows how BTX was devastating the catalyst before installation of the CBU.

This is the train that processes a mix of acid gases from both associated and high-pressure Khuff gas treating and as a result has more BTX in its feed. From November 2005 to January 2006, the ΔT had dropped to about 35° F., indicating the first bed was only achieving 20-30% approach to equilibrium. A catalyst regeneration completed in early January restored a substantial degree of activity, bringing the ΔT up to around 165° F.

Immediately after regeneration, the activity started to drop dramatically again, so that by the middle of February the ΔT was only 80° F. At that time the CBU for Train 4 was commissioned and another catalyst regeneration performed. After that, the converter’s ΔT is seen to be essentially constant, only changing in response to feed composition variations.

Although Fig. 3 only reports first converter ΔT up until November 2006, the carbon beds continue to demonstrate excellent results. There has been no apparent decrease in activity in the first converters through to early August 2007.

The elimination of BTX-caused deactivation has already had tangible and important benefits. Whereas it had been the practice to replace large amounts of Claus catalyst every 2 years to maintain recovery, it has been determined that catalyst changes can be deferred, resulting in savings of hundreds of thousands of dollars because of the size and number of SRUs involved.

Perhaps even more significant is that our planned conversion of the SRUs at Shedgum and Uthmaniyah to the SuperClaus process can go ahead as planned. Without our first addressing the chronic deactivation that had been experienced for the last 25 years, SuperClaus would not have been a viable process retrofit.

First maintenance; inspection shutdown

As discussed earlier, there was a concern regarding chlorides and possible corrosion in stainless-steel vessels. We completed almost 1 year of operation on one of the trains and in November 2006, the CBU and SRU were shut down for routine maintenance. This afforded us the opportunity to open and inspect the carbon-bed vessels to look for signs of corrosion.

The bottom of the vessels showed signs of where the regeneration steam condensate had flowed to the center outlet nozzle. Whereas the vessel walls above and below the carbon bed were shinny, there was some discoloration at the ends of the vessel at the 6 o’clock position.

As well, the vessel wall behind the carbon bed was very mildly discolored. Red dye penetrant testing, however, showed no cracks or indication of incipient metal damage such as pitting. Our plan is to follow up with another inspection in a year’s time to monitor the condition of the vessels and ensure the results recently seen are maintained.

Installation of carbon beds to adsorb BTX from acid-gas feed to Saudi Aramco’s lean feed SRUs has successfully eliminated BTX-induced Claus catalyst deactivation.

The operational performance of the CBUs has exceeded expectations in as much as the adsorption cycles are longer than anticipated in the design. This has led to a lower-than-expected production of regeneration steam condensate.

Acknowledgments

The authors thank Graham R. Lobley in Saudi Aramco’s Consulting Services Department who advised the project on the metallurgical questions and Dean W. French in Gas Operations Continuing Excellence who advised on operational issues.

References

  1. Crevier, P.P., Dowling, N.I., Clark, P.D., and Huang, M., “Quantifying the Effect of Individual Aromatic Contaminants on Claus Catalyst,” 51st Laurance Reid Gas Conditioning Conference, Norman, Okla., Feb. 25-28, 2001.
  2. Crevier, P.P., Dowling, N.I., Clark, P.D., and Huang, M., “Performance of Commercial Titania and Titania Hybrid Catalysts in the Presence of Aromatic Contaminants,” 55th Laurance Reid Gas Conditioning Conference, Norman, Okla., Feb. 28-Mar. 1, 2005.
  3. Paskall, Harold G., “Capability of the Modified-Claus Process-A Final Report to the Department of Energy and Natural Resources of the Province of Alberta,” Western Research, March 1979.