Separator Problems That Start Outside the Vessel: Drain Seals, Supports, Transients and Properties

Which Mist Eliminator or Coalescer Do You Need? A Selection Map by Duty
11/15/2026
Entrainment Control in LNG Pre-Cooling Kettle Reboilers: K Values, Disengagement and Shell Diameter
11/15/2026
Which Mist Eliminator or Coalescer Do You Need? A Selection Map by Duty
11/15/2026
Entrainment Control in LNG Pre-Cooling Kettle Reboilers: K Values, Disengagement and Shell Diameter
11/15/2026

When a separator carries liquid over, the internals are the first thing blamed and usually the first thing replaced. In practice the mechanism that failed often sits outside the element: an unsealed drain, a support that let the pad lift, a slug that was never in the design basis, or a property in the datasheet that does not exist in the field. This page sets out the checks we run before agreeing that an element has failed, because replacing a correct element with another correct element fixes nothing.

1. The drain seal, which is the most common single cause

Vane packs, mesh pads with collection troughs, and demisting cyclones all depend on liquid leaving through a path that is sealed against gas. If the drain is not sealed, the device does not separate at all: gas flows up the drain and out through the device, and the liquid that was captured is pushed back into the gas stream.

The requirements are mechanical, not process, and they are easy to get wrong in a revamp:

  • The drain pipe must end below the vessel’s low-low or trip level, with margin. A common rule is at least 100 mm below the trip level, so the seal survives a level excursion before a shutdown.
  • The available static head between the device and the liquid surface must be more than the pressure drop across the device. Where the liquid foams, the effective density is lower than the measured density, so the pipe should be sized to use well under half the available static head.
  • The pipe should be arranged for self-venting flow. A drain that runs full and slugs cannot pass the required liquid rate, and the device floods from below without any obvious symptom at the element.
  • The seal has to exist at start-up, not only at steady state. A unit commissioned dry with a sealed drain that was never prefilled behaves exactly like a unit with no drain at all.

The diagnostic signature is distinctive: pressure drop across the element is lower than expected, the element looks clean on inspection, and carryover appears as soon as the unit is loaded. That is gas bypassing through the drain, not a media problem.

2. Support and installation, where a correct element fails mechanically

A mesh pad that is supported only in the middle of the vessel will lift at the perimeter, and the gap at the shell is a bypass path. Every section of the pad needs support top and bottom, not only the central span. The same applies to vane packs, where a missing hold-down lets the pack move under a pressure transient and open a path at the wall.

Two further points come from operating experience rather than from design codes:

  • In fouling service, a segmented pack that has to be cut out for cleaning costs more downtime than the element is worth. Individually removable vanes and larger drain passages cost a little at the design stage and repay it at the first wash.
  • Large inlet devices are structures, not fittings. Above roughly 36 in nominal pipe size, an inlet vane diffuser is a plate structure that sees flow-induced vibration, and the design has to be checked for it rather than assumed from a smaller unit.

3. Transients, including on duties that look steady

A wet-gas field with low liquid rates still slugs. A small suction scrubber sized on steady-state rates will overfill on the first slug and trip on high level. The design basis has to include a liquid holdup for the transient, and where the pipeline can generate slugs, the slug volume should come from a transient simulation rather than an assumed percentage.

Inlet piping belongs in the same check. A control valve of smaller diameter than the vessel nozzle, placed close to the nozzle, produces a high-velocity jet that does not have room to dissipate. Where the inlet device was sized on the slug alone and not on that jet, the device sees a force it was never designed for and can be dislodged.

4. Fluid properties that do not exist in the field

The most expensive error in a datasheet is a liquid phase described with “combined” properties of oil and water. The combined density is higher than the oil density and the interfacial tension is not the real one, so the vessel and the mist eliminator are both sized for a fluid that never occurs. The unit is then undersized for the real oil phase, and the symptom appears as persistent carryover with everything apparently correct.

The check is simple and should be applied to every datasheet: each phase listed separately, with density, viscosity and interfacial tension at the operating temperature, and the values sanity-checked against the field. A density that does not match a sample from the separator is a datasheet error, not a process change.

5. Droplet size is decided upstream of the separator

Separation performance is a function of the droplet size entering the element, and that size is decided by valves, pumps, fittings and the pressure drop across them. A high-shear path produces a fine dispersion that a vane pack cannot capture whatever its area, while the same fluid through a low-shear path separates easily.

This is why the selection question is not only “which element” but “what droplet size will actually arrive”. Where the upstream shear is high and cannot be changed, the answer is either a finer technology or a coalescing stage ahead of the separator, not a larger version of the same element.

6. Cyclonic devices carry a liquid-level constraint

An inlet cyclone or a demisting cyclone has a pressure balance that depends on the vessel liquid level. Two failure modes bound the operating window:

  • Gas blowby: at high gas rate the pressure inside the cyclone pushes the liquid out of the bottom, gas escapes through the liquid outlet into the vessel, and the result is foaming, loss of level control and carry-under to the next vessel.
  • Liquid carryover: at high liquid rate the level inside the cyclone rises until liquid leaves through the gas outlet, loading the mist eliminator beyond its design.

The operating level range, including alarms and trips, has to sit between the minimum level curve set by blowby and the maximum level curve set by carryover, across the whole gas and liquid rate envelope. Cyclones with internal blades can lose on the order of a bar, against inches of water for a simple cyclone, so the level window is correspondingly tighter. Multiple cyclones add a second requirement: the incoming two-phase flow has to be distributed evenly, or some cyclones blow by while others carry over.

7. What to check first, by symptom

Symptom Check before replacing the element
Carryover with low pressure drop and a clean element Drain seal: pipe below trip level, self-venting, prefilled at start-up
Carryover that started after a level or rate change Operating level against the cyclone or device level window
Element found displaced or damaged Slug and jet loads, hold-down and support, inlet piping geometry
Carryover on a duty that was sized correctly Droplet size entering the element, and upstream shear
Undersized vessel found on review Datasheet properties: separate phases, real density and interfacial tension
Fouling that recurs after cleaning Cleanability of the installed configuration, drain passages, wash coverage

Further reading

The SPE Separations Technology Technical Section publishes the Savvy Separator series in Oil and Gas Facilities, which covers these mechanisms in more depth, including worked case histories of separators blamed for failures caused elsewhere:

Frequently Asked Questions

Why does a mist eliminator stop separating even though it looks clean?

The usual cause is the drain. A vane pack, collection trough or demisting cyclone only works if the liquid outlet is sealed against gas. If the drain pipe ends above the vessel trip level, or was never prefilled at start-up, gas flows up the drain and out through the device, and the liquid it captured is pushed back into the gas. The signature is low pressure drop across an element that inspects clean, with carryover as soon as the unit is loaded.

How deep below the liquid level should a drain pipe end?

Below the low-low or trip level with margin, commonly at least 100 mm, so the seal survives a level excursion. The pipe should also be arranged for self-venting flow, and where the liquid foams the available static head should be more than the pressure drop by a wide margin, well under half being used, because foam has a lower effective density than the measured liquid.

What causes a mesh pad to lift or a vane pack to move?

Support and hold-down, not velocity. A pad supported only on the central span lifts at the perimeter and opens a bypass path at the shell; a pack without a hold-down moves under a pressure transient. Slug loads and inlet jets are the usual forces, which is why the support has to be designed for the worst transient rather than the steady state.

Can a separator be sized correctly and still be too small?

Yes, if the datasheet described the liquid phase with combined oil and water properties. The combined density and interfacial tension are not those of the real oil phase, so the vessel and the mist eliminator are sized for a fluid that does not exist in the field. Listing each phase separately with its own density, viscosity and interfacial tension is the check that catches this.

Why does droplet size upstream matter more than the element area?

Because capture is a function of the droplet size arriving at the element. A high-shear path through a valve or pump produces a fine dispersion that a vane pack cannot capture whatever its area, while a low-shear path separates easily in the same device. Where shear is high and cannot be changed, the answer is a finer technology or a coalescing stage, not a larger version of the same element.

Leave a Reply