Entrainment Control in LNG Pre-Cooling Kettle Reboilers: K Values, Disengagement and Shell Diameter

Separator Problems That Start Outside the Vessel: Drain Seals, Supports, Transients and Properties
11/15/2026
Entrainment Prediction in Kettle Reboilers, Why the Element Is Not Free to Size, and What CFD Is Actually For
11/15/2026
Separator Problems That Start Outside the Vessel: Drain Seals, Supports, Transients and Properties
11/15/2026
Entrainment Prediction in Kettle Reboilers, Why the Element Is Not Free to Size, and What CFD Is Actually For
11/15/2026

In an LNG pre-cooling cycle, the kettle reboiler is usually described as a heat exchanger. It is not, in the sense that matters: the tube bundle is fixed by the heat duty, but the shell diameter is almost always fixed by entrainment. Get the vapour-liquid separation wrong and the compressor sees liquid it cannot compress, whatever the heat transfer coefficient is. This page sets out how the vessel is actually sized, and why the mist eliminator is the item that decides whether a chiller is large and heavy or compact.

The reference point for this discussion is the propane pre-cooling section of a C3/MR liquefaction train. Pre-cooling represents roughly 30 to 35 per cent of the total refrigerant compressor shaft power, so a small improvement in the propane cycle is worth more than a large improvement anywhere else in the unit.

Entrainment is a mass ratio, not a droplet size

Kettle entrainment is specified as the liquid carried out per unit mass of vapour: 1 per cent means 0.01 kg of liquid per kg of vapour. That specification is about total liquid quantity. It says nothing about droplet size, which is why two vessels meeting the same entrainment number can have very different compressor outcomes: a given mass of liquid in a few large droplets is thrown out in the first stage of the compressor, while the same mass as fine droplets follows the gas much further.

The split matters because it decides which technology can help. Large droplets are gravity’s business and settle back into the pool. Droplets in the few-micron range are not, and no amount of freeboard will remove them: the vessel would have to grow until the rising velocity falls below their terminal velocity, which on a large chiller means a vessel several times the size.

Two mechanisms set the limit: flooding and re-entrainment

Throughput of a wire mesh or vane eliminator is bounded by two phenomena sitting on either side of the same velocity:

  • Flooding, or choking. Above a critical gas velocity the liquid cannot drain down through the element against the rising gas. The element fills, the passages close, and pressure drop climbs until liquid is pushed out of the top.
  • Re-entrainment. Even below flooding, shear at the liquid surface and at the element’s drainage edges can tear droplets off and carry them away, so an element that captured liquid can return it to the gas.

The design velocity is therefore a balance between the two, and the Souders-Brown form is the standard way to express it:

VG = K √( (ρL − ρG) / ρG )

where VG is the allowable gas velocity in m/s, K is the load velocity factor in m/s, and the densities are in kg/m3. The factor K is where technologies differ, and it is also where direction of flow matters:

Element type K, vertical flow (m/s) K, horizontal flow (m/s) Typical cut at 99% efficiency
Wire mesh pad 0.107 0.128 3 to 10 microns
Vane pack 0.152 0.198 10 to 40 microns
Vane, double pocket 0.305 0.305 as above, higher liquid load capacity

These are standard values for the element types, not for a specific vendor’s geometry. A supplier quoting a capacity factor outside this range is either using a different reference area or claiming a geometry that needs to be verified against its own test data.

Worked example: what the K value does to the shell diameter

Take a propane stage boiling at 17 C, where the liquid density is about 500 kg/m3 and the vapour density about 17 kg/m3, and a vapour flow of 1 m3/s through the disengagement space.

Element Allowable velocity Required area Equivalent diameter
Wire mesh, K 0.107 0.57 m/s 1.75 m2 1.49 m
Vane pack, K 0.152 0.81 m/s 1.23 m2 1.25 m
Vane, double pocket, K 0.305 1.63 m/s 0.61 m2 0.88 m

The same duty needs roughly three times the flow area with a mesh pad as with a double-pocket vane. Since the shell diameter follows the area, that is a factor of about 1.7 in diameter and something closer to three in shell steel and inventory. This is the arithmetic behind every claim about compact chillers: the vessel is sized by the element, not by the bundle.

The real problem is that vapour load is not uniform along the bundle

A kettle reboiler is a single-pass design with a large shell acting as both reservoir and disengagement space. The tube-side stream runs in series through the stages, and the temperature difference between the process stream and the boiling propane falls along the bundle as the process stream cools.

On a high-pressure stage in a warm climate, the numbers can look like this: propane boiling at 17 C, gas entering the bundle at 33 C and leaving at 20 C. The temperature difference is 16 K at the inlet and 3 K at the outlet, so the vapour is generated roughly five times faster at the inlet end than at the outlet end. A vessel with a uniform diameter is therefore over-sized at one end and locally overloaded at the other, and the carryover that reaches the compressor comes from the overloaded end.

Three ways to deal with it, in increasing order of effectiveness:

  1. Segment the kettle. Divide the length into sections and size the diameter on the worst section. Simple, gravity-based, and conservative: it produces a large vessel because every section is penalised for the inlet end.
  2. Adapt the nozzles. Size and place the vapour nozzles according to the local vapour production. This removes the congestion, but it produces asymmetric piping, a more complex stress analysis, and a vessel that is no longer a catalogue item.
  3. Put the elimination inside the vessel. A horizontal mist eliminator installed perpendicular to the ascending vapour forces the vapour to pass through the element before it reaches the outlet nozzles, so the velocity is set by the element area rather than by where the vapour happens to be generated.

The third route is the one that changes the design, and it has three requirements that are easy to miss:

  • Flow straightening below the element. A perforated plate under the mesh distributes the rising vapour across the full section; without it the element sees the local plume and fouls or floods locally while the average numbers look acceptable.
  • Drainage back into the pool, not through the ascending flow. Coalesced liquid is directed downwards onto a deflector plate and returned to the pool, so it does not have to fall counter-current through the gas it was just removed from. That is what removes the re-entrainment risk that a plain horizontal mesh carries.
  • Disengagement space above the element to the nozzle. Enough to keep the pressure drop low and let the element work at its design velocity. This distance is what a CFD study is normally used to confirm.

Why enhanced boiling surfaces change the level rule

With plain or low-fin tubes, the bundle has to stay flooded to keep the heat transfer up, which fixes a minimum liquid level and therefore a minimum distance between the pool surface and the element. Enhanced nucleate boiling surfaces change that: the enhancement is specific to low wall superheat, and nucleate boiling stays active at a wall superheat as low as 0.16 K. The tube does not need to be flooded to perform; it needs to stay wetted, which is satisfied by the foam region above the pool.

The consequence is a design degree of freedom that is rarely used: the liquid level can be set lower than a conventional kettle allows, which buys freeboard for the eliminator, which allows a higher element velocity, which reduces the diameter. On a large chiller that is the difference between a gravity-sized shell and a compact one.

What the compactness is worth

Reported results for a 5 Mtpa single-train C3/MR case, comparing a conventional low-fin design at 3 K temperature approach with an enhanced-tube design using in-vessel elimination:

Metric Conventional, 3 K approach Compact design, 3 K Compact design, 1 K
Chilling train length reference about 40% shorter about 20% shorter
Vaporiser weight reference about 45% lower lower still
Propane compressor power reference same 4.4% lower than 3 K
LNG production at fixed power reference same 1 to 2% higher

The efficiency figures scale with the temperature approach: moving from 3 K to 2 K gains about 2.2 per cent in propane cycle efficiency, and 3 K to 1 K about 4.4 per cent, with fuel gas consumption and CO2 falling by the same amount. Where the propane compressor is 33 per cent of the liquefaction unit’s energy, that 4.4 per cent becomes about 1.1 per cent at the unit level. Manufacturing emissions fall as well, because they track material weight: around 40 per cent lower CO2 for the vaporiser when the tube type is changed at constant approach.

There is a second, less obvious benefit. Propane inventory is a safety and permits issue, and inventory scales with shell volume. A design that reaches the same duty in a smaller shell is not only cheaper to build, it holds less flammable inventory and sits in a smaller pressure envelope.

What to specify to an internals supplier

Item Why it decides the answer
Entrainment limit in kg liquid per kg vapour Sets the element cut point and, for fine limits, whether gravity sizing is even possible
Vapour load profile along the bundle Decides whether one uniform element is adequate or the vessel must be zoned
Liquid level range and alarms Defines freeboard and the minimum distance to the element
Nozzle size and position Determines whether the element can set the velocity or whether local cones bypass it
Tube type and whether flooding is required Decides the minimum level, and therefore the available disengagement height
Requirement for CFD and for physical validation A horizontal element in a large shell is a flow-distribution problem, not only a media selection

For the entrainment correlation itself, the geometric limits that stop an element from being sized freely inside a reboiler, and what a CFD study can and cannot settle, see entrainment prediction and element sizing in kettle reboilers.

For why the same tube change makes some stages shorter and wider while others shrink, see why a shorter kettle bundle can need a larger shell.

The same rule in other reboiler services, where fouling or approach temperature rather than the bundle decides the element, is set out in reboiler mist eliminators in ethylene, C3 splitter heat pumps and refrigeration chillers.

Reference

The operating data, capacity factors and case results discussed here follow Knoepfler, Buffet, El Hajal, Lang, Provost, Rambure and Redjem Saad, “3, 2, 1 … K, nucleation! Or recent improvements in enhanced heat transfer is further reducing temperature approaches in LNG pre-cooling heat exchangers”, LNG2023, Vancouver, paper 3.2.1. The construction details of any specific element remain the supplier’s responsibility; the sizing logic above is the part that belongs in the process specification.

Frequently Asked Questions

Why is the shell diameter of a kettle reboiler set by entrainment rather than by heat duty?

Because the tube bundle is fixed by the heat duty, but the shell has to provide enough disengagement area for the vapour leaving the pool. Once the vapour load is known, the diameter follows from the allowable velocity of the mist eliminator, and the eliminator type rather than the heat transfer coefficient decides how large the shell must be. That is why a change of element can shrink a chiller without touching the duty.

What does an entrainment limit of 1 per cent actually mean?

It means 0.01 kg of liquid per kg of vapour leaving the vessel. The specification is a mass ratio and says nothing about droplet size, which is why two vessels meeting the same number can behave differently at the compressor: the same mass in a few large droplets is thrown out early, while fine droplets travel much further into the machine.

Why does the vapour load vary along the length of a kettle?

Because the process stream and the boiling refrigerant are in series and the temperature difference falls as the process stream cools. On a high-pressure stage with gas entering at 33 C and leaving at 20 C against propane boiling at 17 C, the difference is 16 K at the inlet and 3 K at the outlet, so vapour is generated about five times faster at the inlet end.

What is the difference between flooding and re-entrainment?

Flooding is the element holding more liquid than it can drain against the rising gas: passages close, pressure drop climbs and liquid is pushed out of the top. Re-entrainment happens below flooding, where shear at the liquid surface or at the drainage edges tears captured liquid back into the gas. Both bound the same design velocity from opposite sides.

What does a horizontal mist eliminator need in a large kettle to work properly?

Three things: a perforated plate below it to distribute the rising vapour across the full section, drainage that returns coalesced liquid downward onto a deflector plate instead of counter-current through the gas, and enough disengagement space between the element and the outlet nozzle to keep the pressure drop low. The position is normally confirmed by CFD.

Leave a Reply