About Molecular Sieve Units for Natural Gas Dehydration

The widespread use of molecular sieve technology for simultaneous removal of water and mercaptans from both gas and liquid feed streams is increasing the need to better understand the design principles and operation of molecular sieve units. For economic reasons it is important to not overdesign the molecular sieve unit, while at the same time prevent it from becoming the bottleneck of the gas-processing plant at the sieve’s end-of-run condition.

This paper, split into two parts, discusses important details both in design and in operation of the molecular sieve units used for natural gas dehydration, and shows how applying expert know-how and practical recommendations can provide an effective way to maximize molecular sieve lifetime and performance. This first part of the paper focuses on the design of a molecular sieve unit.

Molecular Sieve Units for Natural Gas Dehydration - Part 1 explores design principles and practices

When having to treat a gas or liquid stream, such that it can be processed by a specific process unit, one of the treating units deployed in the industry is an adsorption unit. These units are most commonly used to remove water from a feed stream but adsorption units also can be used to remove additional contaminants (e.g. mercaptans).

Specifically, when deep removal is required (below 1 ppmv), molecular sieves, an adsorbent composed of a zeolite and typically a clay binder, are the preferred adsorbent. Capable of reaching extremely low specifications is one of the major reasons to use such adsorption units, as these tend to be the only viable option that can be incorporated in a process line-up. A major advantage of molecular sieves is that these can be regenerated which reduces the amount of molecular sieve required down to economical feasible quantities.

This paper will focus on the use of molecular sieve units for natural gas dehydration as these are a critical component in the operation of a Liquefied Natural Gas (LNG) or gas processing plant (typically combined with Natural Gas Liquids (NGL) extraction by cryogenic separation) and any limitation or loss in capacity of this unit can have a significant effect on the overall plant economics. 

An adsorption unit used for water removal is called a dehydration unit (DHU) that often consists of two or more vessels filled with molecular sieves that adsorb water during an adsorption period and are subsequently regenerated using a heated stream of treated gas. A sketch of a typical molecular sieve dehydration unit is shown in Figure 1 (Part 1).

The high temperature during regeneration causes water to desorb from the molecular sieve, and the process is thus called temperature-swing adsorption (TSA). Although TSA is a discontinuous process, the overall dehydration unit behaves like a continuous process because one or more vessels are always in adsorption mode while another vessel(s) is in regeneration mode.

In a typical DHU, the regeneration gas used is a side (slip) stream of the product stream (typically around 10%). Downstream the adsorber vessel, the wet regeneration gas is cooled down and water is condensed and subsequently knocked out in a knock-out (KO) drum. When a molecular sieve unit is used for dehydration it is possible, after compression to negate the pressure drop, to send the regeneration gas back to the feed thereby minimizing valuable product losses.

It should be noted that this type of line-up is only possible for dehydration units designed for only water removal as knocking out the water in the regeneration KO drum provides a “drain” from the system. Components that cannot be removed from the loop in such a manner (e.g. sulfur species such as mercaptans) will build up in the regeneration gas loop. Consequently, in such a line-up the regeneration gas will have to be treated first by an appropriate absorption unit (e.g. by a Sulfinol unit) before it can be “re-injected” in the feed gas.

Providing a cost effective and reliable design as well as optimizing the performance of the DHU unit requires a detailed understanding of its operation which will be discussed in the following sections.

Basic design rules and operational requirements of molecular sieve units

As mentioned previously, an adsorption unit is a discontinuous process, since adsorption is in essence a batch process. One of the most important design parameters of an adsorption vessel is the available water uptake capacity of the molecular sieve bed and there are a few factors that make this important design parameter less straightforward.

In adsorption, gas flows usually from top to bottom. During the adsorption period, the amount of water that can adsorb on the molecular sieve is dictated by both the capacity that can be reached in the limit of reaching equilibrium, and the capacity that results from the competition between adsorption kinetics and the flow of gas past a molecular sieve pellet. In the adsorbing bed two zones can therefore be identified.

The capacity limited part is situated at the top and as it is saturated with water under the feed conditions of temperature, pressure and water concentration is called the Saturation Zone (SZ).

The part of the bed below the saturation zone is engaged in dehydrating the gas from feed water concentration (wet) to effluent water concentration (dry), the so-called Mass Transfer Zone (MTZ) ‘- see Figure 2 (Part 1). During adsorption the MTZ migrates from the top of the bed to the bottom of the bed whereby the SZ lengthens. Ones the MTZ leaves the bed breakthrough occurs and the bed must be taken off-line for regeneration.

Another important parameter to take into account is that the capacity of the adsorbent decreases over time, as a function of the number of regeneration cycles, and thus the end-of-run (EOR) capacity must be used when considering the required amount of adsorbent. When the capacity of the adsorbent falls below the level where all water in the feed can be adsorbed during the minimum adsorption time, the adsorbent must be replaced, as graphically depicted in see Figure 3 (Part 1). 

Obviously another important design parameter is therefore being able to predict the rate of deactivation, for that it is important to understand the main factors that contribute to the deactivation of the molecular sieve. In general, these contributing factors are a consequence of the thermal regeneration process.

During regeneration of the water-saturated molecular sieve, hot dry gas is passed upwards through the molecular sieve bed and water desorbs. If the beginning of the regeneration cycle were started with dry gas at the maximum temperature, water would first desorb from the bottom portion of the bed, thus using much of the heat of the gas. Subsequently, water that desorbed from the bottom portion of the bed would be carried to the top portion, which would not yet have been heated by the regeneration gas.

 Because the top portion of the bed would still be cold and saturated with water, the desorbed water from the bottom portion of the bed could condense to form liquid water. The formation of hot liquid water in the bed can cause the clay binder of the molecular sieve to dissolve, subsequently forming a cake during the regeneration step.

Caking due to hot liquid water formation during regeneration can be avoided by using a well-designed heating profile during the regeneration cycle. If the temperature of the regeneration gas entering the bed is slowly increased, the top portion of the bed can be pre-heated before much water desorbs from the bottom portion of the bed.

Pre-heating of the top portion of the bed prevents the condensation of liquid water when the water from the bottom portion of the bed finally is desorbed. If a non-ideal temperature profile is used during regeneration, degradation of the molecular sieve might still occur without the formation of a large solid cake.

The characteristics of the temperature profile required to avoid liquid water formation depend on a number of factors. The flow rate and composition of the regeneration gas dictate how much heat is delivered to the molecular sieve bed.

The diameter and materials of the vessel determine how much heat, and at what rate, is transferred to the vessel, as opposed to the amount of heat that is transferred to the molecular sieve. Furthermore, the amount of water adsorbed on the molecular sieve, taking into account the mass transfer zone, dictates the gas-phase concentration profile of water moving upwards through the bed.

Another form of molecular sieve deactivation is through coke deposition on the molecular sieve. If the feed to the unit is heavy, the heavier hydrocarbons can be adsorbed onto the molecular sieve binder during adsorption. A portion will be desorbed during regeneration but the residual will remain on the binder, decompose and irreversibly form coke under hot regeneration conditions, which can lead to pore blocking that results in a lower water uptake capacity.

The least contributing factor is the effect that continual thermal cycling has on the molecular sieve. Some thermal degradation and the subsequent loss in water removal capacity of the molecular sieve will occur, the extent of which is dependent on the thermal stability of the molecular sieve. The regeneration process also results in the degradation of the binder material. The expansion and contraction experienced during thermal swings leads to dust production and a consequential increasing pressure drop that is detrimental to LNG production

As the capacity of an adsorbent is limited and the vessels become more expensive as the diameter increases (the required wall thickness increases) there is a significant economic driver to keep the vessels as small as possible. One way to achieve this is to operate at high pressures (typically 60 bara) as the mass flow to be treated is than compressed in a smaller volume.

Another way to achieve this is by reducing the amount of water that is deposited on the bed as much as possible by cooling down the feed stream and knocking out excess water in the feed KO drum. The temperature reduction is limited by the hydrate formation temperature and the operating temperature is chosen such that one can safely operate above that temperature. Another reason to operate at a low temperature is that the uptake capacity of the molecular sieve is higher.

Not discussed yet is the type of molecular sieve to be used. For dehydration, and especially for molecular sieve units used in LNG application, type 4A is used. As mentioned before, a molecular sieve is composed of a zeolite and a binder (typically clay). The binder is used as a glue i.e. to strengthen the particles. The type of zeolite typically employed for dehydration is de Linde Type A (LTA), whereby A refers to Angstrom, indicating the diameter of the zeolite channels where the water is adsorbed i.e. 4A refers to a zeolite with a channel diameter of 4 Angstrom. 

Apart from 4A there are also 3A and 5A LTA sieves, whereby the type of cation determines the channel diameter (K+ for 3A, Na+ for 4A and Ca2+ for 5A). The effectiveness of zeolites for water removal is based on two important characteristics, the channel diameter acting as a filter, thereby limiting the number of species that can co-adsorb, and the polar environment created in the channels, thereby creating an environment whereby preferentially polar molecules are adsorbed. Water is the molecule that is adsorbed strongest (consequently, water will displace all other adsorbed molecules).

As mentioned before, for LNG applications typically a 4A sieve is used as this sieve is more stable and has higher water uptake capacity than 3A sieve, and less co-adsorption will occur when compared to a 5A sieve. However, in gas processing plants that process pipeline gas typically 3A sieve is used.

The reason for that is that if methanol is added to the feed gas to prevent hydrate formation, it will pass the 3A sieve bed (methanol does not fit into the pores) and can be recovered downstream the beds in the part of the line-up where heavier hydrocarbons (e.g. propane, butane) are separated from the natural gas. Furthermore, co-adsorption of methanol on larger pore sieves will adversely affect the effectiveness of the molecular sieve w.r.t. water removal.

The question why molecular sieves are best suited for deep dehydration is illustrated in Figure 4 (Part 1), where typical isotherms of adsorbents used for dehydration are shown. Although molecular sieves have a lower saturation capacity than alumina or silica gel, that capacity is much less dependent on the partial water pressure which enables reaching deeper specifications (< 0.1 ppmv H2O) as well as a shorter Mass Transfer Zone enabling a more efficient use of the adsorbent bed. In this context it should be noted that the length of the MTZ is also determined by the particle size.

For that reason, molecular sieve beds are typically layered, with the larger particles (e.g. 1 /8”) used in the top layers to minimize the overall bed pressure drop and the smaller particles (e.g. 1/16”) in the bottom layer. Another reason to select molecular sieves for dehydration is that the water uptake capacity is less feed temperature dependent in comparison to silica gel and alumina. Furthermore, in contrast to molecular sieves, alumina and silica gel are large pore adsorbents and therefore more co-adsorption of hydrocarbons will occur.

Similar to molecular sieves, water will displace the co-adsorbed components as it adsorbs stronger to the adsorbent, but co-adsorption will also influence the length of the MTZ. Maybe even more importantly, when regenerating the adsorbents these co-adsorbed species will leave the adsorbent as peaks and if the regeneration gas has to be treated, the treating unit will have to be sized such that it can process these peaks. This significantly increases the cost of such a unit.

One could state that the majority of topics discussed so far are related to the chemistry involved in the process. However, there are other factors that are equally important from a design point of view, more specifically those determined by process engineering and economics.

In gas processing, typically a plant is designed and operated with one single purpose which is to process natural gas such that the maximum amount of condensates can be recovered and produced gas adheres to certain specifications. In general, the gas is produced for use in a countries grid or for liquefaction (LNG) so that it can be transported to markets.

The economic aim for such a line-up is to minimize the number and size of vessels, process steps as well as the plot space. From an engineer’s point of view that means to push as much mass flow as possible through the smallest vessels as possible. The total amount of gas and liquids to be processed is the economic basis of such a project. The design of the vessel has to adhere to maximum flow as well as minimum flow criteria.

In gas phase systems, gas flow during adsorption is downward in order to prevent fluidization of the bed under abnormal flow conditions. The diameter of the bed can be calculated once the superficial velocity is determined.[1] The superficial velocity must be chosen such that below criteria are met:

  • The maximum velocity is determined by a maximum absolute gas velocity criterion as such, typically around 0.2 m/s. In LNG/NGL extraction plants, there is significant price on each unit of pressure drop (∆P) consumed. In such cases it is the common practice to limit the ∆P to < 60 kPa. The end-of-run pressure drop is typically twice this value. For other cases, the designer should evaluate the economic benefit of a larger pressure drop. Note that a large pressure drop will also increase the risk of retrograde condensation occurring.
  • The minimum velocity is determined by several criteria. The pressure drop per unit length should be at least 230 Pa/m.[1] This criterion will ensure that the pressure drop over the packed bed is determining the gas flow and proper radial distribution of the flow is promoted. When the gas flow is divided into more than one bed during adsorption without individual flow control it is desirable to have an "appreciable" ∆P to ensure equitable flow distribution i.e. ∆P > 20 kPa. The minimum velocity is also limited by turbulent flow considerations as expressed by the Reynolds number (e.g. Re ≥ 40 > 100[1]).

Some of these criteria are mainly of importance to the main process flow but these should also apply to the regeneration gas flow (in each stage of the temperature profile), which for gas processing vessels is typically up flow. Apart from these criteria the regeneration gas flow has to carry enough heat into the system to ensure that the beds are regenerated properly. There should be sufficient flow to ensure stripping efficiency of the desorbed molecules and the beds should be cooled down within the time frame available.

Within these constraints the aim is to have a regeneration gas flow that is as small as possible, typically around 10% of the feed flow. Especially when the regeneration gas is fed back to the feed, a minimum regeneration flow ensures minimum size regeneration loop equipment and, even more important helps reducing the size of the adsorber vessels.

It is important to realize that for line-ups as sketched in Figure 1 (Part 1), the flow passing through the main process line units is the feed flow combined with the regeneration flow. As the regeneration flow is up-flow another important design constraint is to avoid fluidization of the bed.

The height (L) and the diameter (D) of the beds are mainly limited for practical and economic reasons. One important limitation is the strength of the adsorbent particles, especially those at the bottom of the bed. These should be sufficiently strong that they can withstand the sum of the pressure drop over the bed, the weight of the bed above them when saturated with water, as well as the weight of the pressurized gas. Apart from that there is the more practical observation that bed heights of more than 10 m are rarely observed.

In such cases the adsorber vessels tend to become among the highest units on the site. Minimizing the diameter is an important economic constraint. A larger diameter is associated with an increase in wall thickness, which will increase the cost of the vessel. A rule of thumb that is often applied in vessel sizing is the L/D > 2 criterion.

With lower ratio’s one tends to end up with so-called “pancake reactors” where most of the steel (i.e. cost) used for the construction of the vessel ends up in the top and bottom domes, where curvature is more or less fixed.

A sketch of a typical adsorber vessel is shown in Figure 5 (Part 1). As can be seen a certain distance between the bottom of the inlet distributor and the top of the bed is required to ensure pressure and flow equalization. The flow is reversed during regeneration and the same requirement applies to the bottom part of the bed. A careful observer will notice that the difference in size between the ceramic ball layers never exceeds a factor 2.

This is done to lower the probability that smaller particles will migrate between larger particles. In the worst case scenario such migration could lead to a small depression in the top of the bed thereby creating a preferential flow path (due to the lower pressure difference in that part of the bed) also referred to as a “channel”.

As the capacity of a molecular sieve bed is limited and declines as a function of the number of regeneration cycles a very important design criterion is the required life-time i.e. how long has the bed to last before a change out of the inventory is required. That depends somewhat on the application but in general molecular sieve beds are changed out during a major shut down of the plant. That happens every couple of years as various pieces of kit require maintenance.

For an LNG site, gas turbine and compressor maintenance is the critical factor determining when a major shut down occurs and at the moment, for most sites, that is every 4 years. Although this provides the process engineer with a time frame for designing a molecular sieve bed, a translation to the number of cycles that fits into that period is required because one has to be certain that at the end of these 4 years the beds still have sufficient water uptake capacity. 

If a channel is created in a bed the mole sieve in and around that path is depleted much faster than the surroundings (as a larger part of the flow moves through it) which will ultimately result in an early breakthrough of the bed.

[1]*Molecular sieve vendors do not always apply this criterion in the way it is formulated.

As the capacity of a molecular sieve bed is limited and declines as a function of the number of regeneration cycles a very important design criterion is the required life-time i.e. how long has the bed to last before a change out of the inventory is required. That depends somewhat on the application but in general molecular sieve beds are changed out during a major shut down of the plant.

That happens every couple of years as various pieces of kit require maintenance. For an LNG site, gas turbine and compressor maintenance is the critical factor determining when a major shut down occurs and at the moment, for most sites, that is every 4 years. Although this provides the process engineer with a time frame for designing a molecular sieve bed, a translation to the number of cycles that fits into that period is required because one has to be certain that at the end of these 4 years the beds still have sufficient water uptake capacity. 

An adsorption cycle is composed of several steps. The major steps are adsorption and regeneration whereby the regeneration step can be subdivided in a heating, a cooling and a standby step.

The highest temperature the molecular sieve can be exposed to during regeneration is determined by the thermal stability of the molecular sieve, for a 4A sieve that is 320 °C. Although one can regenerate at lower temperatures the consequence is that more water will stay behind on the molecular sieve thereby reducing the effective water uptake capacity. A typical regeneration profile is shown in Figure 6 (Part 1). The step during ramp up is inserted to prevent too rapid heating up as that can result in liquid hot water formation.

A typical adsorption time is 16 h. In this context it is worth mentioning again that although TSA is a discontinuous process, the overall dehydration unit behaves like a continuous process because one or more vessels are in adsorption mode while another vessel(s) is in regeneration mode. Consequently, there is a limit to the time available for regeneration.

With an adsorption time of 16 h, in a 1+1 line-up there is 16 h available for regeneration (cycle time 32 h), while in a 2+1 line-up as shown in Figure 1, there is only 8 h available for regeneration (cycle time 24 h). A cycle time sequence for a 2+1 line up is graphically depicted in Figure 7 (Part 1).

It is important to realize that the timing in a cycle time sequence is critical as the transfer of beds from adsorption to regeneration has to go flawlessly. With a mistake in the timing one could end up in a situation where one bed is coming out of the adsorption step while the bed being regenerated is still in the heating step thereby creating a forced shutdown

. In this context it is worth mentioning that most sites work with fixed cycle times, which means that when the beds are “fresh” there is excess capacity available as the bed heights are designed for end of run conditions.

It is possible to minimize this effect by applying a “variable cycling” design where the cycle time is adjusted at regular intervals in such a manner that these changes roughly follow the deactivation profile as shown in Figure 8. In this manner one can design for smaller beds or, when troubleshooting or debottlenecking, extend the lifetime of the beds somewhat by reducing the overall number of cycles, slowing the deactivation rate of the adsorbent.

As a consequence of this mode of operation one should realize that the minimum required capacity as illustrated in Figure 8 (Part 1) is in essence determined by the minimum time need for regeneration, i.e. the fastest time one can heat up and cool down while ensuring that the beds are fully regenerated. 


The first part of this paper focusses on the main design elements of a molecular sieve dehydration unit although one can already notice that it is virtually impossible to design such a process unit without taking operational aspects into account. Some additional operational, debottlenecking and troubleshooting options will be discussed in the second part of this paper.

Only the main design elements were discussed for a fairly typical molecular sieve unit. Not discussed are the various other permutations that are possible for example the number of vessels deployed (e.g. 1+1, 3+1, 3+2, 4+2 line-ups), the choice and the routing of the regeneration gas, regeneration gas heating options, regeneration at lower pressures, external or internal insulation of the vessels, etc. Although these options will influence vessel sizing and total cost of the unit, the basic design elements as discussed in this paper will be the same.


Nowadays the main design elements required for designing and operating a molecular sieve dehydration unit are well understood. When taking these elements into account it is possible to design molecular sieve units that are reliable, deliver on specification and required lifetime, and require therefore relatively little operational attention.

Molecular Sieve Units for Natural Gas Dehydration. Part 2 explores some operational problems and possible solutions.

In the first part of this paper the main design elements of a molecular sieve dehydration unit were discussed. However, once a molecular sieve unit is constructed and the gas processing plant has started up there are elements during its operation that require special attention. This second part of the paper focusses on the operation of a molecular sieve unit used for natural gas dehydration.

Common issues with molecular sieve units and how to address these

A contributing factor to molecular sieve deactivation is deposition of liquids such as amines, originating from the upstream amine unit being deposited on the molecular sieve. Such liquids can enter the macropores of the molecular sieve and, during regeneration, causes severe degradation of the molecular sieve by destroying the binder material. This can ultimately result in caking and coking, mechanisms that have already been discussed in Part 1.

The deposition of liquid droplets can arise from poor upstream liquid/vapor separation, retrograde condensation or a regeneration profile that results in the formation of liquid water. To minimize this effect, the knock out vessel upstream of the drier beds usually has advanced de-entrainment internal, such as Schoepentoeter, mist mat, swirl deck and mist mat (SMSM) installed. Whilst these should provide adequate separation, it is conceivable that they have not been installed properly which can lead to liquid carry over in the form of entrainment.

It is also possible that a slug from the upstream amine unit will enter the molecular sieve bed. This type of liquid carry over is hard to prevent and will cause serious damage to the molecular sieve unit which usually requires immediate change out unless it is noticed early. If there are operational indicators that a large volume of liquids (e.g. by puking or foaming of the upstream amine unit) was deposited on the bed than this can only be mitigated be a very slow and careful ramp up in the following regeneration step.

Additional protection of the molecular sieve against small droplets can be provided by the installation of a guard layer on top of the bed. This is a small layer of silica or alumina specifically installed to catch liquid droplets and protect the molecular sieve bed.

If capillary condensation of hydrocarbons takes place in the bed this can also lead to a rapid increase in the rate of deactivation. Liquid hydrocarbons wet the sieve, covering it in a film, which adds extra resistance to the mass transfer of water molecules from the bulk gas phase to the molecular sieve active surface reducing the water removal capacity of the sieve.

Wetting of the molecular sieve also increases the sieves susceptibility to attrition which will lead to dust formation, increased pressure drop and channeling. The only way to mitigate against this is to operate the molecular sieve beds a few degrees above the hydrocarbon dew point which can be achieved by installing a heater downstream the feed knock-out (KO) drum.

Capacity tests runs, which provide information on the amount of water adsorbed by the molecular sieve at the time of testing, are essential for evaluating the performance of the molecular sieve over its runlength and for estimating the remaining lifetime. Once the capacity has been determined from a test run this can be plotted against the number of regeneration cycles the bed has experienced.

If the results of the test runs are plotted against the expected deactivation curve one can evaluate whether it is possible to reach the planned change out time. When sufficient data points are collected over time the true deactivation profile can be determined and extrapolated out to estimate the remaining lifetime. The results of several of these test runs for three trains are summarized in Figure 1 (Part 2).

As can be seen in the graph, the results suggest that the molecular sieve beds are deactivating much faster than expected. Once there is some certainty that the beds performance is indeed lower than expected one will attempt to identify the cause. The main aim would be to avoid an unplanned shutdown for changing out the molecular sieves and to keep the units running until the planned shutdown. At that moment in time there is only a limited amount of data one can collect and analyze in an attempt to identify the cause of underperformance of the beds.

The first type of dataset to be collected is plant data, more specifically flow, feed temperature, pressure, temperature profiles during regeneration, regeneration gas flow, pressure drop over the beds and analyzer data. The analyzer data is the data collected by the moisture analyzers, generally located in the bottom of the bed, at the outlet of the bed and in the common outlet. Sometimes there is also a mid-bed probe. The dataset collected should be analyzed for anomalies by a comparison with historic data i.e. data from a period the plant was functioning well.

If not already part of the standard sampling and analysis scheme implemented to monitor plant performance one can take condensate samples from the feed gas and regeneration gas KO drums for analysis. Again the dataset collected should be analyzed for anomalies by a comparison with historic data as well as the current plant data. The sampling scheme is graphically presented in Figure 2 (Part 2).

Feed gas samples are collected and analyzed on a regular basis. However, one should realize that this type of analyses only will generate a bulk composition, data one needs for analysis but will most probably not generate data on trace components due to the limited amount of sample available. In general, when discussing the quality of gas samples, one will also end up in a discussion on sampling reliability.

As the molecular sieve dehydration unit (DHU) is still running, it will not be possible to extract molecular sieve samples from the vessels itself. However, if the cause for the beds underperformance is not easily identified, molecular sieve samples should be taken and analyzed at the first opportune moment.

Even when there are no problems with the DHU it is recommended to take and store some samples during the planned change out. In this manner samples are available for comparison if the need arises. When opening and unloading vessels one should always observe closely on whether there are signs of caking (clunks of molecular sieve particles), excessive dust formation, excessive coloration (darkening) of the sieve and channeling (discoloration on the wall).

Molecular sieve samples (fresh and spent) can be analyzed by a variety of techniques. The more commonly used ones are :

  • Bulk crushing strength and plate crushing strength analysis to determine whether particles maintained their strength.
  • Thermometric Gravimetric Analysis (TGA) adsorption measurements of capacity and mass transfer properties to determine if there is an excessive decrease in the water adsorption capacity of the adsorbent i.e. to measure the degree of deactivation of the adsorbent.
  • Pyrolysis Combustion Mass spectrometric Elemental (PCME) and/or Flash Combustion analysis can be used to determine the amount of coking,
  • Mercury porosity determinations provides information about the pore structure of a material and its degradation.

One of the more common failure with DHU’s is the failure of the bottom support structure in the vessel. The bottom support is an essential structure in the vessel as it has to ensure proper flow distribution during regeneration. The bed rests on a support such that the bottom-dome of the vessel has a void space. Note already that a bottom dome filled with ceramic balls will achieve the same goal.

The root cause of these problems is the continuous expansion and contraction of the bottom support during the thermal cycles. If there is a weakness in the bottom support structure that can create a gap, than one can almost be certain that the molecular sieve particles will find that hole. Once the molecular sieve starts flowing it behaves like water and most probably part of it will flow into the bottom dome. When opening a vessel a clear indicator would be the observation that there is a depression in the bed.

Such a depression would create a channel i.e. a path of least resistance manifesting itself as an early breakthrough of the bed. The molecular sieve in the bottom dome might also start swirling itself around, thereby grinding itself to dust which can create problems for downstream equipment. A sketch of a “classic” bottom support structure is shown in Figure 3 (Part 2).

The most common causes of bottom support failure for the classic bottom support are Mesh screens installation failures, incorrect sizes of ceramic balls used and incorrect installation of ceramic rope packing in the gap. An example of the latter is that the ceramic rope is too tightly packed whereby it loses its flexibility. Accumulation of dust between the support grid and the wall can result in a deformed support grid.

An alternative to the classic bottom support structure are the use of V-wire mesh screens although these also have problems. It is worth noting that one should be careful that the design used in hydrotreater vessels is not copied. Although these reactors operate at high temperatures these are not exposed to continuous temperature cycles. The other option one can implement is filling the bottom dome with ceramic balls as sketched in Figure 4 (Part 2).

The main advantage of filling the bottom dome with ceramic balls is eliminating issues with the bottom support as it is removed from the vessel. However, there are also a few disadvantages. Obviously, the weight of the vessel increases which can be a significant disadvantage for floating or off-shore structures were weight and plot space are major cost items.

Due to the larger inventory in the vessel, more heat is required for regeneration which when employing the same regeneration flow means more time is needed for regeneration (heating and cooling) i.e. the minimum required uptake capacity increases[1]. Obviously such a system also has a higher CO2 footprint, a characteristic that nowadays is also evaluated during design.

It cannot be emphasized enough that proper loading of the molecular sieve beds is a critical operation. Especially during installation of the ceramic rope, the mesh screens, the ceramic balls and the first layer of molecular sieve a thorough check should take place.

Unloading of the vessels is preferably done fast but safe in order to minimize down-time. Before unloading takes place the beds should be regenerated and purged to assure that co-adsorbed species such as hydrocarbons or sulfur species have been removed down to acceptable levels.

One issue that is easily forgotten when operating DHU’s is what to do with spent adsorbent. Several aspects have to be considered. The health, safety and environment (HSE) regulations during loading and unloading and during transport and storage basically demand that full personal protection equipment (PPE) is worn. When loading and unloading dust will be an issue. De-dusting before start-up (e.g. by piston purging) is generally applied.

Local regulations for handling and transport of possibly contaminated materials as well as regulations for the safe destruction and recovery of spent adsorbents will apply. The latter is usually not an issue with spent molecular sieve. In general, it is important to realize that public concerns w.r.t. handling and storage of possibly contaminated materials might endanger the license to operate. For spent molecular sieve several disposal options are used. Usually, the sieve will be stored on site which is not a permanent solution.

Mostly, spent molecular sieve is disposed of by land filling a procedure which, as mentioned previously, is completely controlled by local regulations. One of the minimum requirements for land-filling would be that a stabilization and leaching test would confirm the safety of the land filling option. Safe destruction is also an option but this tends to be the most expensive one. In this context it is worthwhile mentioning that vendors usually will assist in dealing with spent adsorbents. Some vendors even offer “cradle to grave” support services.

Another equipment item that creates problems on a regular basis are the valves surrounding the vessels, enabling switching between the beds and guiding the flow for adsorption and regeneration. The valves used in this environment have to adhere to very stringent specifications to provide a tight seal over a wide temperature range at a high pressure.

In a worst-case scenario a valve would start leaking and deposit hot and wet regeneration gas on a bed in adsorption, severely reducing the uptake capacity of that bed and maybe even induce caking. Only a few vendors can deliver these types of valves. In some designs (or applications e.g. Oxygen plants) regeneration is done at a lower pressure which provides an additional driving force for desorption thereby increasing the efficiency of the regeneration. Such an operation increases the pressure differential over the valves, increasing the risk of a leak occurring.

This is one of the reasons that for TSA units the regeneration is done at the same pressure, although not the major reason. The major reason is that regeneration at a similar pressure as the feed reduces the size of the compressor in the regeneration loop.

Design and operation of molecular sieve units: A summary

Despite careful design and operation of the molecular sieve dehydration unit, there could be a variety of reasons (e.g. different feed composition, upstream equipment not working properly, etc. etc.) that make it impossible to run the unit in the manner it was originally designed for. If “quick-fixes” are not possible there are a few options left that can be considered.

Assuming that the main purpose is to reach the next planned shutdown and there is still some capacity left in 2 or more beds (assuming 2+1 configuration) than there are a few scenarios’ one might consider. One option might be to continue operating at a reduced flow throughput.

If one bed is severely underperforming one can consider to combine reduced flow throughput with running in 1+1 mode. Unfortunately, these options will always carry a large price-tag. In the case the beds are structurally underperforming one could consider to use a denser molecular sieve (as supplied by some vendors [9.10]) possibly combined with variable cycling as debottlenecking options.

Summarizing some of the recommendations given in the previous section that help ensuring that the DHU will run in such a way that the product will be on spec and the planned shutdown data can be reached:

  • Conduct performance test-runs on a regular basis
  • Ensure a proper design and operation of the upstream separator
  • Use a guard-layer
  • Pay attention to the design and the integrity of vessel internals
  • Select good quality molecular sieves
  • Ensure proper loading of the molecular sieve
  • Implement a regeneration profile that prevents the formation of liquid water
  • Analyze spent molecular sieve samples

Based on the data presented in the previous sections one might get the impression that molecular sieve units are the source of many problems on a gas processing plant. In reality that is not the case, these units tend to be reliable requiring relatively little operational attention. That in itself can be problematic when problems arise, as experience with troubleshooting these units is sometimes not available.

*As discussed in Part 1 of this paper, the minimum required capacity is in essence determined by the minimum time need for regeneration, i.e. the fastest time one can heat up and cool down while ensuring that the beds are fully regenerated.

More in Global Solution

Molecular sieves

Molecular sieve technology for simultaneous removal of water and mercaptans from gas and liquid feed streams. 

Contact Gas Processing

Please contact the Shell Gas Treating and Sulphur Removal technologies team for further information and we will respond to your enquiry via email.