Lessons / Separation / Gas–liquid separators

Two-phase gas–liquid separators

Flash drums, knockout drums and scrubbers: how to choose the type, size a vertical or horizontal vessel, pick the internals, read mist-eliminator and packaged-separator vendor data, apply ASME, API 12J and API 521, and keep liquid out of places it should never reach. Nine worked examples and sixteen problems to solve.

Figure 1. The gas sets a vertical separator’s diameter; the liquid sets its height

Mist eliminator
Presets
–m/s
Allowable gas velocity, K√((ρL−ρG)/ρG)
–mm
Inside diameter
–mm
Height, tangent to tangent
–
Height to diameter ratio
Band, bottom to topHeight, mm
Sizing by the method in sections 05 and 06: Souders–Brown with GPSA mesh-pad K values derated for pressure, half of that without a mist eliminator, 0.15 m/s for an upflow vane pack and 0.20 m/s for an axial-cyclone deck (both derated the same way); 100 mm added for the support ring and the diameter rounded up in 150 mm steps; the diameter is increased until height ÷ diameter is 4 or less. Nozzles: inlet ρv² ≤ 2,100 Pa with a half-open pipe, gas outlet ρv² ≤ 3,750 Pa, liquid outlet ≤ 1 m/s. Gas density from pV = ZnRT. Heads are 2:1 ellipsoidal and their volume is ignored. Illustrative screening only; the vendor and the vessel designer finish the job.

A separator is the plainest-looking vessel on a P&ID and one of the most consequential. When it fails, liquid ends up in a compressor, a fuel-gas header or a flare, and gas ends up in a vessel that was never designed for it.

This page is written for chemical and process engineers who size drums, review vendor proposals and sit in HAZOPs. Figure 1 is the whole method in one place: the gas flow and the K-factor fix the diameter, the liquid flow and the holdup and surge times fix the height, and the nozzles and internals fill in the rest. Try the Example 3 preset: there is so much liquid that the drum has to be made wider than the gas needs, and then the mesh pad is too big for the gas flowing through it.

The sections below explain every step behind that figure, then go further into horizontal vessels, internals, vendor data, mechanical design, standards and safety. If you can solve the problems at the end without opening the solutions, you can specify and check most two-phase separators you will meet.

01

What a separator does

A two-phase separator takes a stream that is part gas and part liquid and sends each phase out of its own nozzle, with as little of the other phase as the downstream equipment can tolerate. Whatever the orientation, it does this in four zones, which the GPSA Engineering Data Book labels A to D.1

  1. A, the inlet or primary section. An inlet device absorbs the momentum of the feed and drops out most of the liquid as bulk liquid. A good inlet device also spreads the gas evenly across the vessel. A bad one shatters liquid into fine mist that nothing downstream can catch.
  2. B, the gravity settling section. The gas slows down and droplets fall out under gravity. It is usually sized so that droplets of about 100–150 µm and larger settle before the gas reaches the mist eliminator.712
  3. C, the mist extraction section. A mesh pad, vane pack or bank of cyclones catches the fine droplets, down to around 10 µm or smaller, by impaction and coalescence.68
  4. D, the liquid collection section. The liquid gives up its entrained gas bubbles, and the volume between the level settings gives the control system and the operators time to react.

Most sizing arguments are about which of these zones runs out of room first. In a scrubber with a little condensate, it is the gas zones B and C. In a flash drum taking a large liquid flow, it is zone D.

Figure 2. Vertical and horizontal separators have the same four zones in a different arrangement

Schematic, not to scale. Left: a vertical separator with a half-open-pipe inlet and a horizontal mesh pad. Right: a horizontal separator with an inlet deflector and a vane pack in horizontal gas flow. Red dashed lines are the high-high and low-low trip levels. Zone letters follow the GPSA convention.1

Levels, holdup and surge

The liquid zone is defined by its level settings. From the bottom up they are usually the low-low trip (LALL), the low alarm (LAL), the normal level (NLL), the high alarm (LAH) and the high-high trip (LAHH). Svrcek and Monnery give the definitions most sizing procedures use:2

  • Holdup is the time for the level to fall from normal to the low-low level at the normal outlet flow with no feed. It protects whatever the liquid goes to next, such as a pump, a column or a heater.
  • Surge is the time for the level to rise from normal to the high-high level at the normal feed rate with the outlet closed. It absorbs slugs and upsets. Without a better basis, Svrcek and Monnery take surge as half the holdup time.

Some companies measure surge between the low and high levels instead. Which definition you use matters less than writing it on the data sheet so the vendor sizes the same volume you did.

02

Types and where they are used

Orientation

Vertical separators are the usual choice when the gas-to-liquid ratio is high or the total gas flow is modest, as in scrubbers and knockout drums.1 They need less plot space, which matters offshore, and they make solids easier to remove from the bottom head.15 A given change in liquid volume also moves the level further, which makes level control and trips more responsive. Their weakness is that droplets must fall against the rising gas, so the full cross-section is limited by the settling velocity.

Horizontal separators win when there is a lot of liquid, when the liquid foams or carries dissolved gas that needs a large surface to escape, when slugs arrive, or when a second liquid phase must also settle. Droplets fall across the gas flow rather than against it, so for the same flow area a horizontal drum can run at a higher gas velocity than a vertical one.33 The cost is plot space and a level that moves less for a given volume change.

Spherical separators are compact and strong, but they have little surge volume and are hard to fit with internals. They are rare in new plants and are outside the scope of API RP 12J.4

Services you will meet

Flash drum

A liquid enters below its bubble point downstream of a let-down valve or a heater and partly vaporizes. Examples are amine and glycol flash tanks, hot and cold high-pressure separators in hydrotreaters, and stage separators in gas plants.

Vertical or horizontal, usually with a mesh pad. The feed nozzle sees flashing two-phase flow.

Knockout drum

Removes free liquid and condensate from a gas: fuel-gas knockout drums, upstream of burners and gas turbines, ahead of meters and analysers.

Vertical, with a mesh pad or vane pack.

Compressor suction scrubber

The last line of defence for a compressor. Liquid slugs break valves in reciprocating machines and erode impellers in centrifugal ones, so the scrubber has a conservative K and a high-high level trip that stops the compressor.

Vertical, mesh pad, vane pack or cyclones. GPSA multiplies K by 0.7–0.8 for this service.1

Inlet, production and test separators

Separate well fluids into gas and liquid in oil and gas fields. Test separators measure the rates from one well at a time.

Usually horizontal. API RP 12J and API Spec 12J cover them.45

Reflux drum or overhead accumulator

Collects condensed distillation overhead, supplies reflux and product, and vents non-condensables. Often carries a water boot.

Horizontal; the holdup is usually set on the reflux flow.28

Flare knockout drum

Removes liquid from relief and blowdown streams so the flare does not rain burning liquid. Sized for droplets of 300–600 µm and 20–30 minutes of holdup for the largest liquid release.24

Usually horizontal, no internals. Covered by API 521, not API RP 12J.416

Slug catcher

Receives liquid slugs at the end of a multiphase pipeline. Above about 35 bar, a manifold of large pipes (a finger-type slug catcher) is often cheaper than one thick-walled vessel.34

Vessel type or finger type; sized on the largest expected slug volume.

Steam drum and evaporator separator

Separates steam from boiler water, or vapour from a boiling liquor. Steam purity sets the design, so high-efficiency mesh is common.

Horizontal steam drums; vertical evaporator bodies.8

Filter separator and coalescer

Removes submicron aerosols that a mesh pad cannot, ahead of membranes, amine contactors, turboexpanders and gas turbines.

Filter elements followed by a vane section. Outside API RP 12J.4

Compact and inline cyclonic separators

Use swirl instead of gravity to save weight and space, mostly offshore and in debottlenecks.

Proprietary designs, sized by the vendor. Outside API RP 12J.4

03

The flash: what arrives at the separator

A separator does not decide how much gas and liquid it receives. Thermodynamics does, and the separator only decides how cleanly they part. In design the vapour and liquid rates, compositions and densities come from a flash calculation, usually in a simulator. It is still worth being able to do one by hand, because a hand flash is how you catch a wrong thermodynamic package, a missing component or a feed entered on the wrong basis.

For an isothermal flash of F kmol/h of feed with mole fractions zi into V kmol/h of vapour and L of liquid, with equilibrium ratios Ki = yi/xi, the vapour fraction β = V/F is the root of the Rachford–Rice equation:

Σ zi(Ki − 1) / [1 + β(Ki − 1)] = 0(1)
xi = zi / [1 + β(Ki − 1)]    yi = Ki xi(2)

The left side of (1) falls steadily as β goes from 0 to 1. If it is positive at β = 1, the feed is all vapour; if it is negative at β = 0, it is all liquid. Otherwise there is exactly one root between them, which a few bisection steps will find.

For a quick estimate of K at low to moderate pressure, the Wilson correlation uses only critical properties and the acentric factor:

Ki ≈ (pc,i/p) exp[5.373 (1 + ωi)(1 − Tc,i/T)](3)

A drum downstream of a let-down valve is an adiabatic flash: the feed enthalpy is conserved, so the drum temperature is found by adjusting T until the vapour and liquid enthalpies, weighted by β, equal the feed enthalpy. That temperature drop is how separators downstream of choke and let-down valves end up cold enough to form hydrates or to need impact-tested steel.

Example 1 in section 12 runs a six-component flash by hand. Its vapour and liquid rates feed Examples 2 to 8.

04

How droplets settle

A droplet of diameter d in a rising gas feels its weight less buoyancy pulling it down, and drag pushing it up. It stops accelerating when they balance, at its terminal velocity:

(π d³/6)(ρL − ρG) g = CD (π d²/4) ρG vt²/2(4)
vt = √[ 4 g d (ρL − ρG) / (3 CD ρG) ](5)

The drag coefficient CD depends on the droplet Reynolds number Re = ρG vt d / μG, so (5) is solved by iteration. A smooth-sphere correlation that covers the whole range up to Re ≈ 2 × 10⁵ is Turton and Levenspiel’s:22

CD = (24/Re)(1 + 0.173 Re0.657) + 0.413 / (1 + 16,300 Re−1.09)(6)

Two limits are useful. At Re below about 1, CD = 24/Re and (5) becomes Stokes’ law, vt = g d² (ρL − ρG) / (18 μG): velocity rises with the square of the diameter. At Re above about 1,000, CD is close to 0.44 and velocity rises only with the square root of the diameter. Most droplets in process separators fall between the two.

Rearranging (5) shows where the Souders–Brown equation comes from:

vt = K √[(ρL − ρG)/ρG]    with    K = √[ 4 g d / (3 CD) ](7)

K has units of velocity and carries the droplet size and the drag. For a separator without a mist eliminator, K is a statement about which droplet size you are willing to let through. For a separator with a mist eliminator, K is an empirical limit on the device, set by re-entrainment and flooding, and droplet physics no longer sets it directly.

Figure 3. At 39 barg a 150 µm droplet falls at 0.16 m/s; in air at 1 atm it falls almost three times faster

–m/s
Terminal velocity
–
Droplet Reynolds number
–
Drag coefficient, equation (6)
–m/s
Equivalent K, equation (7)
Hydrocarbon case: ρG = 35.3 kg/m³, ρL = 620 kg/m³, μG = 0.012 mPa·s, the conditions of Example 2. Air and water: ρG = 1.20, ρL = 998 kg/m³, μG = 0.0181 mPa·s. Drag from Turton and Levenspiel.22 Stokes’ law overestimates the velocity of large droplets badly. Shaded bands: 100–150 µm (gravity section ahead of a mist eliminator) and 300–600 µm (flare knockout drums).

Two numbers from Figure 3 are worth remembering. At the hydrocarbon conditions, the GPSA K for a vertical drum without a mist eliminator (half of 0.093, or 0.047 m/s) corresponds to a droplet of about 180 µm. The mesh-pad K of 0.093 m/s corresponds to about 390 µm. So a vertical drum running at its mesh-pad limit carries droplets of several hundred microns up to the pad, and the pad, not gravity, does most of the work. That is why the mist eliminator is sized so carefully, and why it is the first thing to check when a drum carries over.

05

The K-factor and the Souders–Brown equation

The Souders–Brown equation gives the maximum allowable gas velocity, and with it the minimum flow area:

Vmax = K √[(ρL − ρG)/ρG]    Amin = QG / Vmax(8)

QG is the actual gas flow at separator conditions (m³/s), not a standard flow. A is the full cross-section of a vertical vessel, or the face area of the mist eliminator.

The equation is simple; choosing K is the engineering. The values below are the starting points used in industry. They are for clean, non-foaming systems and for the mist eliminators described, and they assume a reasonable inlet device and an even gas distribution.

Table 1. GPSA K for vertical drums with a horizontal mesh pad. K is 0.107 m/s up to 7 barg and falls by 0.003 m/s for every 7 bar above that.1
Pressure, barg07214263105
K, m/s0.1070.1070.1010.0920.0830.065

GPSA adds three corrections to Table 1: use half the value for a vertical drum without a mist eliminator, multiply by 0.6–0.8 for glycol and amine solutions, and multiply by 0.7–0.8 for compressor suction scrubbers, mole sieve scrubbers and expander inlet separators.12 Under vacuum, Svrcek and Monnery suggest K = 0.20 ft/s (0.061 m/s).2

Table 2. Typical K by device and arrangement, for air and water near ambient conditions unless stated. Derate for pressure as in Table 1, which GPSA and Campbell use as an approximation for all mist eliminator types.6
Device or vesselK, m/sNotes and source
No mist eliminator
Vertical drum, gravity only≈ 0.05Half the GPSA mesh value1
Horizontal drum, gravity only0.137 × (L/6)0.58NORSOK typical maximum, L = effective length in m14
Whole separators to API 12J, with a standard mesh pad
Vertical, 1.5 m tall0.037–0.073API 12J range6
Vertical, 3 m tall0.055–0.107API 12J range6
Horizontal, 3 m long0.122–0.152API 12J range6
Mesh pads, vertical upflow
Standard, 144 kg/m³0.107The traditional guideline710
High capacity, 80 kg/m³0.122More open, less efficient7
High efficiency co-knit0.076Fine fibres knitted with wire7
Mesh pad, horizontal gas flow0.1280.42 ft/s9
Vane packs and cyclones
Vane pack, vertical upflow, no pockets≈ 0.150.5 ft/s9
Vane pack, horizontal flow, pocketedup to 0.260.85 ft/s9
High-capacity vane profiles with drainage channelsup to 0.45At the vane face, not the vessel8
Axial demisting cyclones0.15–0.30Vessel K; vendor rated8

Why K falls with pressure

The Souders–Brown equation already lowers the velocity as gas density rises, so the extra derating in Table 1 can look like double counting. It is not. As pressure rises the gas-to-liquid density ratio rises and the surface tension falls, so droplets are smaller, break up more easily and are re-entrained from wet surfaces at lower velocities. Campbell attributes the decline of mesh pad K with pressure mainly to these two effects.7

When to lower K further

  • High liquid load to the mist eliminator. Above about 0.75 US gpm/ft² (1.8 m³/m²·h) for a standard mesh pad, Campbell reduces K by 10 % for each additional 1 gpm/ft² (2.4 m³/m²·h).7
  • Foaming or low surface tension: glycol, amine, some crudes and condensates near their critical point.
  • Fouling, wax or solids: use a lower K with an open mesh, or a vane pack, and plan for cleaning.
  • Slugs and surges: vendors apply a derating factor for these exceptional conditions.8

The opposite mistake is just as real. A mesh pad needs enough velocity for droplets to hit the wires. Below roughly a quarter of its design velocity it loses efficiency, and droplets pass straight through.8 A scrubber sized for a future flow can therefore carry over today.

06

Sizing a vertical separator

This is the procedure behind Figure 1. It follows Svrcek and Monnery’s method with GPSA K values, which is still the most common hand method in process design offices.21

Step 1. Gas and liquid at separator conditions

Convert every flow to actual volume at the separator’s pressure and temperature. For the gas,

ρG = p M / (Z R T)    QG = ṁG / ρG(9)

p in Pa absolute, M in kg/kmol, R = 8,314 J/(kmol·K), T in K. Take Z from the same equation of state as the flash.

Step 2. Diameter from the gas

Pick K from Tables 1 and 2, derate it for pressure and service, and find Vmax from equation (8). The mist-eliminator diameter follows, and the vessel is made slightly larger to leave room for the support ring:

DME = √(4 QG / (π Vmax))    D ≥ DME + 75 to 150 mm(10)

Svrcek and Monnery add 3–6 in and round up to the next 6 in (about 150 mm).3 Without a mist eliminator the whole cross-section is the gravity section and no ring allowance is needed.

Step 3. Liquid heights from holdup and surge

hH = tH QL / A    hS = tS QL / A(11)

Holdup and surge times are set by what the liquid goes to and how good the control is. Table 3 gives the usual starting points.

Table 3. Typical liquid holdup and surge. These are rules of thumb; use the owner’s specification where one exists.
ServiceHoldup or retentionSource
General gas–liquid separator or knockout drum3–5 min; surge about half the holdupBranan, Svrcek and Monnery282
Drum feeding a distillation column5–10 minRules of thumb28
Drum feeding a fired heater10 min or moreBranan28
Reflux drumAbout 5 min on the reflux flowBranan28
Oil–gas separator, oil above 35 °API1 min retentionAPI Spec 12J5
Oil–gas separator, 20–30 °API1–2 minAPI Spec 12J5
Oil–gas separator, 10–20 °API2–4 minAPI Spec 12J5
Flare knockout drum20–30 min of the largest liquid releaseAPI 521, as summarized by Prophet and Lea24
Spacing between alarm and trip levelsAt least 30 s of flow or 100 mm, whichever is greaterNORSOK P-10014

Step 4. Fixed heights

The rest of the height is set by spacing rules, not by flow. Table 4 lists the values used in Figure 1. They are minimums; instrument connections, manways and the inlet device itself often push them up.

Table 4. Height bands of a vertical separator with a mist eliminator, from the bottom tangent line up.
BandTypical minimumSource
Bottom tangent to LALL300 mm, or as the level instrument needsDesign guide adapted from GPSA35
LALL to NLLHoldup, equation (11)
NLL to LAHHSurge, equation (11)
LAHH to inlet nozzle centreline300 mm + ½ dNDesign guide35
Inlet centreline to underside of mist eliminatorLarger of 0.5 D and 610 mm + ½ dN (915 mm + ½ dN with no mist eliminator)Monnery and Svrcek3
Mesh pad thickness100–150 mm, 150 mm most commonKidnay and co-authors; Campbell2936
Top of mist eliminator to gas outlet nozzleAt least (D − dout)/2, so gas leaves the whole pad face evenly. Measured to the nozzle, so the depth of the top head (D/4 for a 2:1 head) counts towards itCheresources guideline30

Step 5. Add up and check the proportions

HTT = hLALL + hH + hS + hin + hD + hME + htop(12)

Vertical two-phase separators usually end up with a height-to-diameter ratio of about 3 to 5.12 If H/D is well above that, the liquid is controlling: increase the diameter, which shortens the liquid column, and repeat. If H/D is below about 2, the fixed spacings are controlling and the vessel is about as small as it can be.

Widening a vessel for liquid has a side effect that catches people out. The gas velocity in a full-diameter mesh pad falls, sometimes below its turndown limit. The fix is to keep the pad sized for the gas: either a full-diameter pad with an annular blanking plate, or a smaller pad in a box around the gas outlet.12 Figure 1 does this automatically when the velocity in a full-diameter device falls below about 25 % of design for mesh (30 % for cyclones, 40 % for vanes). Example 3 works through it.

Step 6. Nozzles, then mechanical design

Size the nozzles from section 08, check the inlet nozzle against the height you allowed for it, then pass the vessel to mechanical design (section 10).

07

Sizing a horizontal separator

In a horizontal drum the gas flows along the vessel above the liquid, and a droplet released at the top must fall through the gas space before the gas carries it to the outlet. The cross-section is shared, so every calculation needs the area of a circular segment. For a liquid depth h in a drum of diameter D, the fraction of the cross-section below the liquid surface is

f(h/D) = (θ − sin θ) / (2π)    with    θ = 2 arccos(1 − 2h/D)(13)

θ in radians. At h/D = 0.5, f = 0.5; at h/D = 0.25, f = 0.196. The gas area above a level is AT[1 − f(h/D)].

The procedure

Svrcek and Monnery’s method, in SI units:2

  1. Find the settling velocity with a no-mist-eliminator K (half the GPSA value), and use 75 % of it, UV = 0.75 UT, for a conservative design.
  2. Calculate the holdup and surge volumes, VH = tHQL and VS = tSQL.
  3. Guess L/D and get a first diameter, assuming holdup and surge fill about 60 % of the drum:
    D = [ 4(VH + VS) / (0.6 π (L/D)) ]1/3(14)
  4. Set the low-low level. Svrcek and Monnery use 0.5 D(ft) + 7 in, about 180 mm + 0.042 D(mm); many companies use a flat 300 mm minimum.
  5. Set the gas space above the high-high level to at least the larger of 0.2 D and 0.3 m, or 0.6 m with a mist eliminator.
  6. Find the length for the liquid, and the length for the droplets:
    Lliq = (VH + VS) / (AT − AV − ALLL)    Lgas = (QG/AV) × (hV/UV)(15)
  7. Take the larger length. Repeat with other diameters until L/D is between 1.5 and 6, and choose the lightest vessel that meets both. Svrcek and Monnery narrow the range by pressure: 1.5–3 up to 250 psig (about 17 barg), 3–4 from 250 to 500 psig (17–34 barg) and 4–6 above that.

Typical horizontal two-phase separators end up with L/D of 3 to 5.12 Low-pressure drums often end up shorter and fatter, because their walls are thin and the cost follows the shell area, not the wall thickness.

Two more checks are worth making. The gas velocity over the liquid must stay low enough not to whip liquid off the surface; NORSOK’s typical maximum, 0.137 (L/6)0.58 √[(ρL − ρG)/ρG], is a useful screen.14 And the effective length available to the droplets is shorter than the shell: Campbell uses Le = L − D to allow for the inlet and outlet zones.6

Figure 4. Rate a horizontal drum: the droplet must reach the liquid before the gas reaches the outlet

Inputs

m
m
kg/h
kg/m³
m³/h
kg/m³
mm
mm
min
m/s
  • Droplet design velocity UV = 0.75 K √[(ρL−ρG)/ρG], released at the top of the drum.
  • Effective length L − D; liquid volume in the heads ignored.
  • Gas velocity screen: NORSOK typical maximum for horizontal separators.
–m/s
Gas velocity above LAHH
–m
Length a droplet needs to settle
–min
Time from LAHH to LALL at the liquid flow
–
Length to diameter ratio
Open Example 4
Rating, not design: enter a vessel and see whether it does the duty. The defaults are Example 4. The droplet path is drawn from the top of the gas space at the inlet end and falls at UV while it travels at the gas velocity. Method after Svrcek and Monnery; gas velocity screen after NORSOK P-100.214

08

Nozzles and internals

A correctly sized vessel with the wrong inlet will carry over. Excess momentum in the feed breaks liquid into fine mist that the mist eliminator cannot catch, and excess momentum in the gas outlet sucks gas through one part of the mist eliminator and leaves the rest idle.37 Nozzles are therefore sized on momentum, ρv², rather than on velocity alone.

Table 5. Nozzle criteria from published company practice. ρm is the mean density of the feed mixture, ρm = (ṁG + ṁL)/(QG + QL). Confirm vendor devices with the vendor.
Nozzle and inlet deviceLimitSource
Inlet, no inlet deviceρmv² ≤ 1,000–1,400 PaTwo company guides13
Inlet, half-open pipeρmv² ≤ 1,500–2,100 PaTwo company guides13
Inlet, vane type (Schoepentoeter)ρmv² ≤ 6,000–8,000 PaCompany guides; Sulzer138
Inlet, single cycloneρmv² ≤ 10,000 PaCompany guide13
Inlet, multi-cycloneρmv² ≤ 35,000 PaCompany guide; inlet cyclones have run to 65,000 Pa1312
Gas outletρGv² ≤ 3,750 PaCompany guide; others use 4,500–5,400 Pa1338
Liquid outletv ≤ about 1 m/sTypical practice, 0.9–1.5 m/s38

Also check the inlet line against the erosional velocity of API RP 14E, ve = C/√ρm in US units, with C = 100 for continuous service and 125 for intermittent service.26 NORSOK asks for separator process nozzles at least one standard size larger than the connected pipe, and manways of at least 24 in.14

Inlet devices

  • None, or an open pipe end. Cheap, and prone to short-circuiting and jetting liquid onto the opposite wall.12
  • Deflector or diverter plate. A flat or dished plate in front of the nozzle. It stops the jet but tends to make small droplets and foam.12
  • Half-open pipe. A pipe stub with its lower half cut away, pointing the flow downward. A simple default for vertical drums with modest momentum.
  • Vane inlet (Schoepentoeter and similar). Rows of curved vanes slice the feed sideways and spread the gas. It is the normal choice at moderate to high momentum for feeds that are mostly gas, and it copes with slugs.8
  • Inlet cyclones. Use the feed momentum to spin out liquid, break foam and help coalescence, at momentums around ten times those of pipe inlets.12

Give the inlet a straight run of about 5–10 pipe diameters with no pockets, so slugs and stratified flow arrive as evenly as possible.38

Mist eliminators

Table 6. Mist eliminator types compared. Values are typical vendor figures; performance depends on pressure, liquid load and properties.89
TypeSmallest droplet removed wellPressure dropTurndownBest forWeak at
Knitted mesh pad2–3 µm in air and waterUnder 2.5 mbarTo about 25 % of designClean services, the default choiceFouling, wax, solids, high liquid load, very high pressure
Vane pack (chevron)About 20 µm; 10–15 µm with drainage channels1–10 mbarTo about 30–50 %Fouling services, high liquid load, horizontal flowFine mist
Axial demisting cyclonesAbout 10 µmUp to about 30 mbarTo about 30 %, more with a pre-conditionerHigh pressure, compact vessels, debottlenecksCost, pressure drop, vendor-specific rating
Mesh plus vane, or mesh plus cyclonesSet by the second stageSum of bothWide; the mesh works at low flowWide-range duties: the mesh coalesces at low flow and floods into larger drops at high flow, which the second stage catchesHeight and cost

Mesh pads work best horizontal with the gas flowing up through them and lose effectiveness if tilted.29 Vane packs are happiest vertical with the gas flowing horizontally through them.36 In two-stage arrangements the mesh pad can be run deliberately above its flooding point as an agglomerator, because the liquid that leaves a flooded pad comes off as droplets much larger than those that entered it.7

Other internals worth naming on the data sheet

  • Vortex breaker on every liquid outlet, usually crossed plates, to stop gas being drawn into the pump suction.26
  • Mist eliminator drains. Liquid from a vane box or cyclone deck drains through a pipe that must be sealed below the lowest liquid level. The pressure drop across the device lifts liquid in the drain pipe; if the level is too high the device will not drain, and liquid can be siphoned. A small hole near the top of the drain pipe is sometimes used as a siphon breaker.26
  • Perforated baffles or flow straighteners in horizontal drums to even out the gas and liquid velocity after the inlet.
  • Sand jets and drains where solids arrive with the feed.
  • Level bridles or chambers for the level instruments, with the trip transmitters on separate connections from the control transmitter.

09

Vendor data and how to read it

Most separator internals are bought, and most packaged separators are selected from a vendor’s range. Vendor data is good data, but it is quoted under the vendor’s conditions, which are rarely yours. Reading it well means finding the basis first and the number second.

Mesh pad styles: four names for the same pad

A mesh pad is described by four numbers: bulk density (kg/m³ of pad, which for stainless steel follows from the void fraction), void fraction, wire diameter and specific surface area (m² of wire per m³ of pad). Each vendor gives its styles its own names. Converting to the four numbers shows that the catalogues are much closer than the names suggest.

Table 7. Roughly equivalent stainless-steel mesh styles. Imperial figures converted at 1 lb/ft³ = 16.02 kg/m³ and 1 ft²/ft³ = 3.281 m²/m³. Koch-Glitsch lists newer high-capacity styles replacing its traditional ones with more than 20 % more capacity at the same efficiency.10
ClassDensity, kg/m³Surface, m²/m³Koch-Glitsch10Sulzer KnitMesh8AMACS9Okutani11
High capacity, dirty service80145–158931, now 7084540 (80, 145)7CA (80, 148)SK-80 (80, 158)
Standard, general purpose144258–280431, now 1729030 (144, 258)4CA (144, 279)SK-144 (144, 280)
Heavy duty, higher efficiency192350–377421, now 7099033 (192, 350)4BA (192, 377)SK-192 (192, 375)
Fine wire, very high efficiency115–192390–660326 (128 kg/m³, 0.15 mm wire)369001 (192, 643)3BF (115, 394), 3BA (192, 656)–

Okutani lists its SK-80, SK-144 and SK-192 styles as equivalent to York (now Koch-Glitsch) 931, 431 and 421, which is a useful check on the table.11 Standard wire is 0.011 in (0.28 mm) or 0.25 mm; fine wire is about 0.15 mm.

K-values: read the conditions

Vendor K-values are usually for air and water at ambient conditions: AMACS quotes 0.35 ft/s (0.107 m/s) for vertical flow and 0.42 ft/s for horizontal flow on that basis.9 Sulzer’s standard KnitMesh is quoted at K up to 0.107 m/s, its V-MISTER at up to 0.13 m/s and its 9797 design at up to 0.12 m/s.8 None of these numbers applies directly at 70 bar with a condensate. Derate for pressure (Table 1), liquid load and foaming, and ask the vendor to rate the device at your conditions and your liquid load.

Figure 5. The mesh pad that removes 99 % of 4 µm droplets in air and water removes only about four-fifths of 10 µm droplets in 69 barg gas

–m/s
Gas velocity through the pad
–µm
50 % removed (d50)
–µm
90 % removed (d90)
–µm
99 % removed (d99)
Illustrative model in the form vendors’ curves take, not a vendor curve. Pad efficiency 1 − exp(−0.238 S T ηw) after Carpenter and Othmer, as given by Campbell, with S the specific surface, T the thickness and ηw the single-wire impaction efficiency from a Langmuir–Blodgett-type fit ηw = [Stk/(Stk + 0.25)]², Stk = (ρL−ρG) d² V / (18 μG Dw).723 Wire 0.25 mm; surfaces 158, 280 and 375 m²/m³. Design velocity from K = 0.107 m/s (0.122 for 80 kg/m³), derated to 0.080 m/s at 69 barg. Air and water: ρG 1.20, ρL 998 kg/m³, μ 0.0181 mPa·s. Gas: ρG 60, ρL 500 kg/m³, μ 0.013 mPa·s. Re-entrainment above design velocity is not modelled.

Grade efficiency curves

A grade efficiency curve gives the fraction of droplets of each size that the device removes, at one velocity, one pressure and one fluid pair. Read three numbers from it: the size removed at 50 %, at 90 % and at 99 %. Then check four things before you use them:

  1. The test fluid and pressure. Figure 5 shows why: the same pad loses most of its fine-droplet efficiency at high pressure, because the gas is denser, the droplets slow less and the velocity through the pad is far lower.
  2. The velocity. Efficiency for small droplets falls as velocity falls. At turndown the pad you bought may be a coarse filter. Move the velocity slider to 25 % to see it.
  3. The liquid load. Curves are usually for light loads. Above the load limit, capacity falls (section 05).
  4. The droplet size reaching the device. A curve that removes everything above 10 µm is no use if the inlet device is making 5 µm mist.

Pressure drop curves

Pressure drop rises slowly with velocity until the device starts to load with liquid, then rises steeply near flooding. A clean, dry mesh pad is typically under 2.5 mbar, vane packs 1–10 mbar and cyclone decks up to about 30 mbar.8 Two practical uses: a rising ΔP in operation means fouling or flooding, and the device ΔP sets how far liquid rises in its drain pipe, which in turn sets how far below the low-low level the drain must be sealed.

Packaged separator capacity tables

Packaged separators, such as well-test units, are sold from capacity tables. Table 8 reproduces two rows of SLB’s published table for its conventional horizontal test separators.25

Table 8. Two rows from a published capacity table. Working pressure is at 38 °C; oil capacity is at 1 min retention. The sheet does not state the pressure, gas gravity or liquid density behind the gas figures.25
ModelVessel sizeWorking pressure, kPaOil, high level, m³/dOil, low level, m³/dGas, high level, m³/dGas, low level, m³/d
SEP-C42 in × 10 ft9,9282,2891,057708,0001,698,000
SEPL-E48 in × 18 ft9,9284,7702,8622,123,0002,973,000

How to read it:

  1. One vessel, two operating points. “High” and “low” are liquid level settings. At high level the vessel holds more oil for its minute of retention and leaves less area for gas, so the oil rating rises and the gas rating falls. You choose the level, and you get one pair, not the best of both columns.
  2. Find the basis. Oil capacity is tied to 1 min of retention, the API 12J figure for light oil. If your oil is heavier or foams, your capacity is lower in proportion to the retention time you need.5
  3. Find what is missing. The gas figures are in standard volume, but the sheet does not say at what pressure. Gas capacity at a fixed velocity rises roughly in proportion to pressure, so a figure quoted at full working pressure means much less capacity at half the pressure. Ask.
  4. Check it independently. Example 9 backs out the liquid fill and the K-value the SEP-C rating implies. If a vendor figure implies a K well above the ranges in Table 2, ask what internals justify it.

What to send, and what to check in the bid

  • Send every operating case with gas and liquid rates, densities, viscosities, surface tension, molar mass, Z, and the liquid load and droplet distribution if you have them. Include turndown.
  • State the inlet piping arrangement, expected slugs, foaming tendency, solids and wax.
  • Ask for the K-value and the area it refers to, the predicted outlet liquid carryover (for example in US gal/MMscf or mg/Sm³), the pressure drop and the turndown, for each case.
  • Check that the proposed vessel height leaves the spacings in Table 4, that the drain seals work at your level settings, and that the materials match the vessel’s corrosion and temperature basis.

10

Mechanical design and standards

Process design hands mechanical design a diameter, a height, nozzles, internals, a design pressure and temperature, a corrosion allowance and a material. Mechanical design returns wall thicknesses and a weight, and sometimes a reason to change the diameter. It is worth knowing the shell formulas well enough to check them.

Design pressure and temperature

Design pressure is set above the maximum operating pressure with enough margin for the relief valve to sit above normal fluctuations without simmering; Svrcek and Monnery’s example uses 110 % of operating pressure;2 a common company rule is the greater of 110 % and operating plus 1.7 bar (25 psi), which governs on low-pressure drums. Design temperature covers the highest operating temperature plus a margin, and the minimum design metal temperature (MDMT) covers the coldest case, including auto-refrigeration during depressuring.

Shell and head thickness to ASME VIII Division 1

For internal pressure P, inside radius R and diameter D in the corroded condition, allowable stress S and joint efficiency E, the long-standing Division 1 formulas are:17

Cylindrical shell, hoop stress (UG-27(c)(1)):   t = P R / (S E − 0.6 P)(16)
Cylindrical shell, longitudinal stress (UG-27(c)(2)):   t = P R / (2 S E + 0.4 P)(17)
2:1 ellipsoidal head (UG-32(d)):   t = P D / (2 S E − 0.2 P)    Hemispherical head:   t = P R / (2 S E − 0.2 P)(18)

Valid for the hoop formula when P ≤ 0.385 S E and t ≤ R/2. Add the corrosion allowance and the plate under-tolerance, and allow for thinning when heads are formed.

Rearranged, the same equations give the maximum allowable working pressure (MAWP) of the plate actually chosen, P = S E t / (R + 0.6 t) for the shell. The standard hydrostatic test pressure is 1.3 × MAWP × the ratio of allowable stress at test temperature to that at design temperature.17

The most common carbon-steel plate for separators is SA-516 Grade 70, with an allowable stress of 20,000 psi (138 MPa) at moderate temperatures.42 Check ASME Section II Part D at your design temperature, and the impact-test rules at your MDMT.

Table 9. Head selection used by Svrcek and Monnery.2
Diameter and design pressureUsual head
D below 15 ft (4.6 m), P above 100 psig (6.9 barg)2:1 ellipsoidal
D below 15 ft, P below 100 psigTorispherical (flanged and dished), knuckle radius 0.06 D
D above 15 ft, any pressureHemispherical

Relief device accumulation

Under Section VIII Division 1, a pressure relief system must keep the pressure from rising more than 10 % (or 20 kPa) above MAWP with a single device, 16 % with multiple devices, and 21 % in a fire.17 The overpressure protection rules have since moved into ASME Section XIII.18

In Canada

Pressure vessels designed for more than 15 psig (103 kPa), apart from the small vessels the code exempts, must have their design registered with the provincial safety authority under CSA B51, which gives the design a Canadian Registration Number (CRN) that appears on the nameplate. In British Columbia the authority is Technical Safety BC.19 The current edition of CSA B51 is the 2024 edition.19 Plan for registration in the schedule: a vessel built elsewhere without a CRN cannot be installed until it has one.

The standards a separator touches

Table 10. Standards and what each contributes. Editions as of November 2026.
StandardWhat it coversWhere it bites on a separator
Process design
API RP 12J, 9th ed. (2024)Process design of oil and gas separators and scrubbers, vertical and horizontal, with conventional inlet devices, mist eliminators and baffles; instrumentation guidance; floating facilities4K-factors, inlet and outlet sizing, level spacing. Excludes spherical separators, flare knockout drums, filter separators and inline separators
API Spec 12J, 8th ed. (2008, reaffirmed 2023)The earlier specification for oil and gas separators, listed as withdrawn in September 2024 when RP 12J was published5Retention times by oil gravity, still widely quoted and found on older data sheets
GPSA Engineering Data Book, Section 7Separation equipment: types, K-values, droplet settling, mist eliminators1The source of Table 1
NORSOK P-002:2023 with corrigendum AC:2024 (and earlier P-100)Offshore process system design: design pressure and temperature, safety systems, line sizing14Level spacing, horizontal K, nozzle sizing
IPS-E-PR-880 and IPS-E-PR-850Iranian Petroleum Standards for the process design of gas (vapour)–liquid separators and of separators in production plants15Selection criteria, design factors and data sheets on Iranian projects
Shell DEP 31.22.05.11 and other owner standardsCompany rules for type selection and design of gas–liquid separatorsInlet momentum limits in Table 5 come from documents like these13
Pressure vessel and piping
ASME BPVC Section VIII, Div. 1 and 2 (2025)Construction of pressure vessels17Thickness, nozzles, welding, inspection, hydrotest, nameplate
ASME BPVC Section II and Section XIIIMaterials and allowable stresses; overpressure protection18SA-516-70 stresses; relief device rules
ASME B16.5 and B31.3Flanges; process pipingNozzle flange class; connected piping
CSA B51:24Canadian boiler, pressure vessel and pressure piping code19CRN registration
EN 13445, PD 5500European and UK pressure vessel codes39Where ASME is not the contract code
Relief and safety
API Std 521, 7th ed. (2020, errata 2022)Pressure-relieving and depressuring systems16Overpressure scenarios, fire case, gas blowby, flare knockout drums
API Std 520 Part I, 10th ed. (2020) and Part II, 7th ed. (2020)Sizing, selection and installation of relief devices27PSV orifice, inlet loss, outlet piping
API RP 14C and ISO 10418:2019Safety analysis of offshore process components20Table 11 safety devices
API RP 14EOffshore production piping26Erosional velocity of inlet lines
IEC 61511Safety instrumented systems for the process industryIndependent LAHH, LALL and PAHH trips
NACE MR0175 / ISO 15156Materials for H₂S-containing serviceSour service hardness and material limits
API 510, API 579-1/ASME FFS-1In-service inspection; fitness for serviceInspection intervals, corrosion assessment, re-rating

11

Safety: hazards and safeguards

A separator sits between a high-pressure source and lower-pressure equipment, holds a flammable inventory, and protects machines that cannot take liquid. API RP 14C’s safety analysis table for pressure vessels lists what can go wrong and the sensor that should catch it.20

Table 11. Undesirable events and safety devices for a pressure vessel, after the API RP 14C safety analysis table.20 PSH, PSL: high and low pressure sensors; LSH, LSL: high and low level sensors; TSH: high temperature sensor; PSV: pressure safety valve; FSV: check valve; ESS: emergency support systems.
EventCausesDetectable conditionPrimary protectionFurther measures I would add
OverpressureBlocked or restricted outlet, upstream control failure, gas blowby from an upstream vessel, fireHigh pressurePSH and PSVSize the PSV for every credible case; independent PAHH trip on the inlet ESD valve
Liquid overflowInflow above outflow, slugs, blocked liquid outlet, level control failureHigh liquid levelLSHIndependent LAHH trip of the feed or of the downstream compressor; enough surge time to act
Gas blowbyLiquid outflow above inflow, liquid outlet left open, level control failureLow liquid levelLSLIndependent LALL closing a shutdown valve on the liquid outlet; downstream PSV sized for blowby
LeakCorrosion, erosion, impact, vibrationLow pressurePSL and FSVGas detection, corrosion monitoring, inspection to API 510
Excess temperatureHeater or temperature control failure (heated vessels)High temperatureTSHShutdown of the heat source

Fire

API 521 estimates the heat absorbed by the wetted wall of a vessel in a pool fire as

Q = 43,200 F Aw0.82   W, with adequate drainage and prompt firefighting(19)
Q = 70,900 F Aw0.82   W, without them(20)

Aw is the wetted area in m², counted only up to 7.6 m (25 ft) above grade or the fire source; F is the environment factor, 1.0 for a bare vessel.1640

The relief rate is the heat input divided by the latent heat at relieving conditions, W = Q/λ. Near the critical point, where the latent heat is small, API 521 suggests a minimum of 115 kJ/kg as an approximation.40 The gas-filled upper part of a vessel has no liquid to cool it; its wall can lose strength in a fire before the relief valve lifts, which is why depressuring systems exist alongside relief valves.

Gas blowby

If an upstream separator loses its liquid level and its level control valve is open, gas at upstream pressure flows through a valve sized for liquid into a downstream vessel. When the upstream pressure exceeds the downstream vessel’s MAWP, this is often the largest relief case on the downstream drum. API 521 does not allow credit for a favourable control response, so the valve is assumed to stay where the failure left it, and the bypass valve may have to be assumed open as well.3141 The gas flow through the valve follows the control-valve sizing equation of IEC 60534-2-1:21

W = N6 Cv Y √(x p1 ρ1),   Y = 1 − x/(3 Fγ xT),   x ≤ Fγ xT,   Fγ = k/1.40(21)

W in kg/h with N6 = 27.3 for p1 in bar absolute and ρ1 in kg/m³; x = Δp/p1; xT is the valve’s pressure-differential ratio factor. When x reaches FγxT the flow is choked and Y = 2/3.

The remedies are to size the downstream relief valve for the blowby flow, to limit the valve trim or add a restriction orifice so the blowby flow is smaller without affecting normal liquid flow, or to provide an independent low-low level trip, which some companies accept as a high-integrity protection system in place of relief capacity.41

Liquid carryover

Liquid reaching a compressor, a turbine fuel system or a flare can destroy equipment or start a fire. The defences are a conservative K for the service, an inlet device matched to the inlet momentum, a mist eliminator that works across the whole flow range, and an independent high-high level trip with enough surge volume between LAH and LAHH for the trip to act before liquid reaches the inlet nozzle.

Other hazards

  • Low temperature. Depressuring a vessel, or flashing liquid across its drain valve, can chill steel below its MDMT. Check auto-refrigeration cases and choose impact-tested steel where needed.
  • Hydrates and ice in drains, level bridles and instrument connections: heat tracing, insulation, methanol or glycol injection, and drains that can be cleared safely.
  • Sour service. H₂S is toxic and causes sulfide stress cracking: select materials to NACE MR0175/ISO 15156. Iron sulfide deposits can be pyrophoric when a vessel is opened, so keep them wet and purge with nitrogen.
  • Erosion from sand at the inlet nozzle and the opposite wall: wear plates, inlet devices, and velocities below the erosional limit.
  • Foaming that carries liquid into the gas outlet: a lower K, antifoam, or a cyclonic inlet.
  • Draining to atmosphere. Never drain a pressurized separator to an open drain through a single valve; liquid flashes, the valve can freeze open, and gas follows the liquid.
  • Vessel entry. Separators are confined spaces with residual hydrocarbons, H₂S and nitrogen. Isolation, purging, gas testing and an entry permit come before anyone goes through the manway.

Questions a HAZOP will ask

  • What is the highest pressure any upstream source can impose, and can the vessel see it?
  • What is the blowby flow if the upstream level valve and its bypass fail open?
  • How long between LAH and LAHH, and between LAHH and the inlet nozzle, at the maximum liquid rate?
  • Are the trip transmitters independent of the control transmitter, and tested?
  • What is the coldest the steel can get during depressuring?
  • Where does the liquid go when the drain is opened, and what does it do when it flashes?

12

Worked examples

Examples 1 to 8 follow one stream from a flash calculation through a vertical inlet separator, its nozzles, wall thickness and relief loads, to the low-pressure flash drum it feeds. Example 9 reads a vendor’s capacity table. All numbers were checked in code; small differences in the last digit are rounding.

Example 1. A six-component flash

Given
1,000 kmol/h of feed at 40.0 bar absolute and 15 °C: methane 0.60, ethane 0.10, propane 0.10, n-butane 0.08, n-pentane 0.07, n-hexane 0.05 (mole fractions). K from the Wilson correlation, equation (3).
Find
The vapour fraction, the phase compositions and the mass flows of each phase.
  1. Wilson K-values with Tc, pc and ω from standard tables: C1 7.245 C2 0.857 C3 0.184 nC4 0.044 nC5 0.012 nC6 0.004
  2. Check for two phases. At β = 0 the Rachford–Rice sum is Σz(K−1) = +3.456; at β = 1 it is Σz(K−1)/K = −20.87. One sign change, so there is one root between 0 and 1.
  3. Solve equation (1) by bisection or Newton’s method: β = V/F = 0.6554
  4. Compositions from equation (2):
    ComponentzKx (liquid)y (vapour)
    Methane0.607.2450.11780.8535
    Ethane0.100.8570.11040.0946
    Propane0.100.1840.21500.0395
    n-Butane0.080.0440.21410.0095
    n-Pentane0.070.0120.19860.0024
    n-Hexane0.050.0040.14410.0005
  5. Flows and molar masses: V = 655.4 kmol/h, M_V = 19.05 kg/kmol → 12,485 kg/h of gas L = 344.6 kmol/h, M_L = 53.88 kg/kmol → 18,565 kg/h of liquid

Two-thirds of the moles leave as gas, but 60 % of the mass leaves as liquid. The separator sees a modest gas flow and a large liquid flow, which Example 3 will show is a liquid-controlled vessel. Round to 12,500 kg/h of gas and 18,600 kg/h of liquid for the examples that follow.

Example 2. Terminal velocity of a 150 µm droplet

Given
Gas from Example 1 at 40.0 bar a and 15 °C with Z = 0.90; liquid density 620 kg/m³; gas viscosity 0.012 mPa·s.
Find
The terminal velocity of a 150 µm droplet and the K it corresponds to.
  1. Gas density, equation (9): ρ_G = 40.0e5 × 19.05 / (0.90 × 8,314 × 288.15) = 35.3 kg/m³
  2. Start with Stokes’ law and check Re: v = 9.81 × (150e-6)² × (620 − 35.3) / (18 × 1.2e-5) = 0.598 m/s Re = 35.3 × 0.598 × 150e-6 / 1.2e-5 = 264 → far outside Stokes’ range
  3. Iterate equations (5) and (6): Re 264 → C_D 0.715 → v 0.213 Re 94.1 → C_D 1.132 → v 0.169 Re 74.8 → C_D 1.269 → v 0.160 Re 70.7 → C_D 1.306 → v 0.158 converged: v_t = 0.157 m/s, Re = 69.3, C_D = 1.32
  4. Equivalent K, equation (7): K = 0.157 / √((620 − 35.3)/35.3) = 0.157 / 4.07 = 0.0386 m/s check: √(4 × 9.81 × 150e-6 / (3 × 1.32)) = 0.0386 m/s

Stokes’ law would have overestimated the velocity almost four times. And a gravity-only K of 0.0386 m/s is below the GPSA no-mist-eliminator value of 0.047 m/s at this pressure, which corresponds to about 180 µm. Figure 3 reproduces these numbers.

Example 3. A vertical inlet separator where the liquid controls

Given
12,500 kg/h of gas (ρG 35.3 kg/m³) and 18,600 kg/h of liquid (ρL 620 kg/m³, so 30.0 m³/h) at 39 barg. Standard mesh pad, half-open-pipe inlet, holdup 5 min, surge 2.5 min.
Find
Diameter, height, level settings and mist-eliminator size.
  1. Actual gas flow and K: Q_G = 12,500 / 3,600 / 35.3 = 0.0983 m³/s K = 0.107 − 0.003 × (39 − 7)/7 = 0.0933 m/s (Table 1) V_max = 0.0933 × 4.07 = 0.380 m/s
  2. Diameter from the gas: D_ME = √(4 × 0.0983 / (π × 0.380)) = 0.574 m D = 0.574 + 0.10 = 0.674 → round up to 0.75 m
  3. Liquid heights at 0.75 m: A = 0.442 m², Q_L = 0.500 m³/min h_H = 5 × 0.500 / 0.442 = 5.66 m, h_S = 2.83 m → H/D ≈ 14 A 14:1 vessel is not sensible. The liquid controls.
  4. Widen in 150 mm steps until H/D ≤ 4. At D = 1.20 m, H/D = 4.25; at D = 1.35 m: A = 1.431 m², h_H = 1.747 m, h_S = 0.873 m bottom to LALL 0.300 LAHH to inlet ℄ 0.300 + 0.203/2 = 0.401 inlet ℄ to pad = max(0.5 × 1.35, 0.61 + 0.101) = 0.711 pad 0.150 pad to top tangent = max(0.15, (1.35 − 0.154)/2 − 1.35/4) = 0.260 H_TT = 0.300 + 1.747 + 0.873 + 0.401 + 0.711 + 0.150 + 0.260 = 4.443 → 4.5 m, H/D = 3.3
  5. Check the mesh pad: full-diameter pad: v = 0.0983 / 1.431 = 0.069 m/s = 18 % of V_max → below turndown 650 mm pad in a box: v = 0.0983 / 0.332 = 0.296 m/s = 78 % of V_max
  6. Level settings above the bottom tangent: LALL 300 NLL 2,047 LAHH 2,920 inlet ℄ 3,321 pad 4,033–4,183 mm 30 s of liquid = 0.00833 × 30 / 1.431 = 0.175 m (> 100 mm), so LAL ≈ 475 mm and LAH ≈ 2,745 mm by the NORSOK spacing

Gas alone would fit in a 0.75 m drum; the liquid needs 1.35 m. Widening the vessel then left a full-diameter mesh pad running at 18 % of design, where it would barely work, so the pad is sized for the gas and boxed in. With this much liquid, a horizontal drum is worth pricing too. Select the Example 3 preset in Figure 1 to see the same vessel drawn to scale.

Example 4. A horizontal low-pressure flash drum

Given
Downstream flash drum at 5.0 bar a (4 barg) and 40 °C: 6,000 kg/h of gas with M = 28, Z = 0.98; 90 m³/h of liquid at 650 kg/m³. No mist eliminator. Holdup 10 min, surge 5 min.
Find
Diameter and length, and the level settings.
  1. Gas properties and settling velocity: ρ_G = 5.0e5 × 28 / (0.98 × 8,314 × 313.15) = 5.49 kg/m³, Q_G = 0.304 m³/s K = 0.107/2 = 0.0535 m/s (below 7 barg, no pressure derating) U_T = 0.0535 × √((650 − 5.49)/5.49) = 0.580 m/s, U_V = 0.75 U_T = 0.435 m/s
  2. Volumes and a first diameter, equation (14) with L/D = 4: V_H = 10 × 90/60 = 15.0 m³, V_S = 7.5 m³ D = [4 × 22.5 / (0.6 × π × 4)]^(1/3) = 2.29 m
  3. For D = 2.4 m: A_T = 4.524 m² LLL = 300 mm (Svrcek’s 0.5 D(ft) + 7 in gives 278 mm): h/D = 0.125, A_LLL = 0.326 m² h_V = max(0.2 × 2.4, 0.30) = 0.48 m: A_V = f(0.2) × A_T = 0.1424 × 4.524 = 0.644 m² L_liq = 22.5 / (4.524 − 0.644 − 0.326) = 6.33 m → 6.5 m L_gas = (0.304/0.644) × (0.48/0.435) = 0.472 × 1.10 = 0.52 m The liquid controls by a wide margin.
  4. Compare diameters: D 2.1 m → L 8.5 m, L/D 4.0, shell + heads 65.6 m² D 2.4 m → L 6.5 m, L/D 2.7, shell + heads 61.5 m² D 2.7 m → L 5.0 m, L/D 1.9, shell + heads 58.2 m² At 4 barg the walls are thin, so area tracks cost. Choose 2.4 × 6.5 m: an L/D of 2.7 is inside the 1.5–3 range Svrcek and Monnery give below about 17 barg, without being stubby.
  5. Levels on the chosen drum, from equation (13): NLL: A = 0.326 + 15.0/6.5 = 2.634 m² → h = 1,355 mm LAHH: A = 2.634 + 7.5/6.5 = 3.788 m² → h = 1,873 mm; gas space 527 mm at LAHH: gas velocity 0.413 m/s, droplet falls 0.527 m in 1.21 s, travelling 0.50 m NORSOK screen: 0.137 × 10.84 × ((6.5 − 2.4)/6)^0.58 = 1.19 m/s ≫ 0.41 m/s

A design droplet reaches the liquid within half a metre of the inlet, so the gas section is far bigger than it needs to be: this drum is a liquid surge vessel with a gas outlet. The Figure 4 calculator opens on this case.

Example 5. Nozzles for the vertical separator

Given
The Example 3 flows, a half-open-pipe inlet, and Table 5 limits. Schedule 40 pipe inside diameters.
Find
Inlet, gas outlet and liquid outlet nozzle sizes.
  1. Feed mixture: Q_m = 12,500/35.3 + 18,600/620 = 354.1 + 30.0 = 384.1 m³/h = 0.1067 m³/s ρ_m = 31,100 / 384.1 = 81.0 kg/m³
  2. Inlet, ρv² ≤ 2,100 Pa: v ≤ √(2,100/81.0) = 5.09 m/s → d ≥ √(4 × 0.1067/(π × 5.09)) = 0.163 m NPS 8 (ID 202.7 mm): v = 3.30 m/s, ρv² = 884 Pa
  3. Gas outlet, ρv² ≤ 3,750 Pa: v ≤ √(3,750/35.3) = 10.3 m/s → d ≥ 0.110 m NPS 6 (ID 154.1 mm): v = 5.27 m/s, ρv² = 981 Pa
  4. Liquid outlet, v ≤ 1 m/s: Q_L = 0.00833 m³/s → d ≥ 0.103 m NPS 4 (ID 102.3 mm) gives 1.01 m/s, just over; NPS 6 gives 0.45 m/s Take NPS 6 with a vortex breaker.
  5. Erosional velocity of the inlet line, API RP 14E with C = 100: ρ_m = 81.0 kg/m³ = 5.06 lb/ft³ → v_e = 100/√5.06 = 44.5 ft/s = 13.6 m/s ≫ 3.3 m/s

The inlet came out well under its momentum limit because the next size down, NPS 6, would give 5.7 m/s and about 2,650 Pa, over the limit. Always check the size you pick, not just the minimum.

Example 6. Shell and head thickness to ASME VIII Division 1

Given
The Example 3 vessel, 1,350 mm inside diameter, maximum operating pressure 39.0 barg. Design pressure 110 % of operating. SA-516-70, S = 138 MPa at design temperature; full radiography, E = 1.0; corrosion allowance 3 mm; 2:1 ellipsoidal heads.
Find
Shell and head thickness, the MAWP of the chosen shell plate, and the hydrotest pressure.
  1. Design pressure: P = 1.1 × 39.0 = 42.9 barg = 4.29 MPa
  2. Shell, equation (16), radius in the corroded condition: R = 675 + 3 = 678 mm t = 4.29 × 678 / (138 × 1.0 − 0.6 × 4.29) = 2,909 / 135.4 = 21.5 mm plus corrosion allowance: 24.5 mm → order 26 mm plate checks: P = 4.29 ≤ 0.385 SE = 53.1 MPa; t ≤ R/2
  3. Head, equation (18): t = 4.29 × 1,356 / (2 × 138 − 0.2 × 4.29) = 21.1 mm, plus 3 mm = 24.1 mm order 28 mm plate to allow for thinning when the head is formed
  4. MAWP of the 26 mm shell, corroded thickness 23 mm: P = 138 × 23 / (678 + 0.6 × 23) = 4.59 MPa = 45.9 barg
  5. Hydrotest at 1.3 × MAWP × stress ratio, taking the stress ratio as 1.0: nameplate MAWP 42.9 barg → test 55.8 barg nameplate MAWP 45.9 barg → test 59.6 barg

Rounding up the plate gave 7 % more pressure capacity than asked for. Stamping the vessel at the higher MAWP costs nothing now and can save a re-rating later, but it raises the test pressure and must agree with the relief valve set pressure.

Example 7. Fire-case relief load

Given
The Example 3 vessel on a skirt with its bottom tangent 1.2 m above grade, bare, with adequate drainage and firefighting. Wetted to LAHH (2.92 m above the bottom tangent), a conservative choice. Latent heat at relieving conditions taken as 280 kJ/kg.
Find
Heat input and relief rate.
  1. Height check: top of wetted zone = 1.2 + 2.92 = 4.12 m < 7.6 m → all of it counts
  2. Wetted area: 2:1 head ≈ 1.084 D² = 1.084 × 1.35² = 1.98 m² shell = π × 1.35 × 2.92 = 12.38 m² A_w = 14.36 m²
  3. Heat input, equation (19): Q = 43,200 × 1.0 × 14.36^0.82 = 43,200 × 8.89 = 384 kW
  4. Relief rate: W = 384,000 / 280,000 × 3,600 = 4,940 kg/h without drainage, equation (20): Q = 630 kW, W = 8,100 kg/h

The fire case is real but not large for this vessel. A blocked gas outlet with the upstream source still feeding would ask the valve to pass the full 12,500 kg/h of gas plus any vapour from the liquid, so fire is unlikely to set the orifice. It must still be documented, and it is the one case allowed 21 % accumulation instead of 10 %.

Example 8. Gas blowby into the low-pressure drum

Given
The Example 3 separator sends its liquid to the Example 4 drum through a level control valve with Cv = 25 and xT = 0.70. The downstream drum has a 10 barg design pressure and its relief valve relieves at 11 barg (12.0 bar a). Upstream gas: 40.0 bar a, 35.3 kg/m³, k = 1.27.
Find
The blowby gas flow if the separator loses its level with the valve fully open.
  1. Pressure ratio and choking, equation (21): x = (40.0 − 12.0)/40.0 = 0.700 F_γ = 1.27/1.40 = 0.907, F_γ x_T = 0.635 → x > 0.635, choked x = 0.635, Y = 1 − 0.635/(3 × 0.635) = 0.667
  2. Flow: W = 27.3 × 25 × 0.667 × √(0.635 × 40.0 × 35.3) = 455 × 29.9 = 13,600 kg/h
  3. Is the valve too big? The normal liquid duty, before allowing for flashing, needs only K_v = Q √(SG/Δp) = 30.0 × √(0.62/35) = 4.0, C_v = 1.156 K_v = 4.6 A valve with Cv = 10, still twice the normal requirement, would cut the blowby to 13,600 × 10/25 = 5,450 kg/h.

The blowby flow is more than twice the downstream drum’s normal gas flow of 6,000 kg/h, so it is likely to set that drum’s relief valve. The cheapest safeguard is often the one in step 3: do not oversize the level valve. An independent low-low level trip closing a shutdown valve is the other common answer.

Example 9. Reading a packaged separator capacity table

Given
The SEP-C row in Table 8: 42 in × 10 ft, 9,928 kPa working pressure; oil 2,289 m³/d at high level and 1,057 m³/d at low level for 1 min retention; gas 708,000 and 1,698,000 m³/d. Assume an inside diameter of 0.98 m, a 3.048 m shell, heads ignored, gas relative density 0.65 with Z = 0.85 at 38 °C, oil 800 kg/m³.
Find
The liquid fill behind each rating, and the K the gas ratings imply if they are quoted at full working pressure.
  1. Shell volume: A = π × 0.98²/4 = 0.754 m², V = 0.754 × 3.048 = 2.30 m³
  2. One minute of oil: high level: 2,289/1,440 = 1.590 m³ → 69 % of the shell (h/D ≈ 0.65) low level: 1,057/1,440 = 0.734 m³ → 32 % of the shell (h/D ≈ 0.36)
  3. Gas mass flow and actual volume at full working pressure: standard density = 0.65 × 1.223 = 0.795 kg/m³ low level: 1,698,000 × 0.795 / 86,400 = 15.6 kg/s ρ_G = 100.3e5 × 18.8 / (0.85 × 8,314 × 311.15) = 85.9 kg/m³ → Q = 0.182 m³/s
  4. Velocity and implied K: low level: gas area 0.68 × 0.754 = 0.513 m² → v = 0.355 m/s K = 0.355 / √((800 − 85.9)/85.9) = 0.123 m/s (0.40 ft/s) high level: gas area 0.233 m², v = 0.326 m/s, K = 0.113 m/s
  5. Consistency: the gas ratings differ by a factor of 2.40 and the gas areas by 2.21, so the two columns describe one vessel at two levels.

Read this way, the table is coherent: K of about 0.11–0.12 m/s is at or just below the bottom of API 12J’s horizontal range (0.122–0.152 m/s for a 3 m vessel). But that range is not derated for pressure the way GPSA’s mesh-pad values are, and at 99 barg a derated figure would be lower. The question to send the vendor is the one the sheet leaves open: at what pressure, gas density and liquid density are the gas ratings quoted, and with which internals?

13

Practice problems

Work each one before opening the solution. Answers are accepted within 2 % unless stated. Use g = 9.81 m/s² and R = 8,314 J/(kmol·K).

Problem 1. Mist-eliminator diameter

A vertical knockout drum with a mesh pad receives 2.0 m³/s of gas at actual conditions. Gas density 18 kg/m³, liquid density 700 kg/m³. Using K = 0.10 m/s, find the minimum mist-eliminator diameter.

m

Problem 2. Pressure derating

What K does the GPSA rule give for a vertical drum with a horizontal mesh pad at 56 barg?

m/s

Problem 3. Gas density

Find the density of a natural gas with M = 18.5 kg/kmol and Z = 0.86 at 70 bar absolute and 30 °C.

kg/m³

Problem 4. Holdup and surge height

A vertical drum of 1.5 m inside diameter receives 30 m³/h of liquid. The specification asks for 8 min of holdup and 4 min of surge. What height is needed between LALL and LAHH?

m

Problem 5. Stokes’ law, and whether it applies

Find the Stokes terminal velocity of a 50 µm water droplet (998 kg/m³) in air at 20 °C (1.20 kg/m³, 0.0181 mPa·s). Is Stokes’ law valid here?

m/s

Problem 6. Segment area

What fraction of a horizontal drum’s cross-section lies below a liquid level at one quarter of the diameter?

–

Problem 7. Inlet nozzle on momentum

A feed of 0.50 m³/s with a mean density of 45 kg/m³ enters through a vane-type inlet device limited to ρv² = 8,000 Pa. Find the minimum nozzle inside diameter. What would it be with no inlet device, limited to 1,400 Pa?

m

Problem 8. Shell thickness

Find the required shell thickness, including corrosion allowance, for a separator with design pressure 2.0 MPa, inside diameter 2,000 mm, SA-516-70 at 138 MPa, spot radiography (E = 0.85) and a 3 mm corrosion allowance.

mm

Problem 9. Fire-case relief rate

A bare vessel with adequate drainage and firefighting has 25 m² of wetted area below 7.6 m. The latent heat at relieving conditions is 300 kJ/kg. Find the relief rate.

kg/h

Problem 10. Mesh pad turndown

A scrubber’s mesh pad is designed for 30,000 kg/h of gas at its design K. If the pad works down to 25 % of its design velocity, what is the lowest gas flow at the same pressure and temperature at which it still separates well?

kg/h

Problem 11. A binary flash

A feed of 40 mol % light component (K = 3.2) and 60 mol % heavy component (K = 0.45) is flashed. Find the vapour fraction V/F.

–

Problem 12. Is this scrubber overloaded?

An existing vertical scrubber of 1.2 m inside diameter with a full-diameter mesh pad now handles 0.55 m³/s of gas at actual conditions, 50 barg, ρG = 30 kg/m³, ρL = 600 kg/m³. What K is it running at? Compare it with the GPSA value at this pressure.

m/s

Problem 13. Blowby flow

A level control valve with Cv = 40 and xT = 0.72 fails open on a separator at 30 bar a. Gas density upstream 25 kg/m³, k = 1.30. The downstream vessel relieves at 6 bar a. Find the blowby gas flow.

kg/h

Problem 14. Hydrostatic test pressure

A vessel has an MAWP of 25 barg at a design temperature where the allowable stress is 118 MPa. At test temperature the allowable stress is 138 MPa. What is the standard hydrostatic test pressure?

barg

Problem 15. Settling length in a horizontal drum

In a horizontal drum the gas space above LAHH is 0.9 m deep. The design droplet falls at 0.20 m/s and the gas moves at 0.8 m/s. What is the minimum effective length?

m

Problem 16. What a vendor rating implies

A packaged horizontal separator of 0.98 m inside diameter and 3.048 m shell length is rated at 2,289 m³/d of oil with 1 min retention. Ignoring the heads, what fraction of the shell volume is liquid at that rating?

–

14

What goes on the separator data sheet

The data sheet is where process design becomes a purchase. Everything a vendor needs to size, rate and guarantee the separator should be on it, for every operating case.

15

Sources

  1. Gas Processors Suppliers Association, Engineering Data Book, Section 7, Separation Equipment. K-values and corrections as reproduced in the Eng-Tips separator FAQ and Wikipedia, “Souders–Brown equation”.
  2. W. Y. Svrcek and W. D. Monnery, “Design two-phase separators within the right limits”, Chemical Engineering Progress, October 1993. Copy online.
  3. W. D. Monnery and W. Y. Svrcek, “Successfully specify three-phase separators”, Chemical Engineering Progress, September 1994; vertical height guidelines. Copy online.
  4. API RP 12J, Process Design of Oil and Gas Separators and Scrubbers, 9th edition, September 2024. Scope summary.
  5. API Specification 12J, Specification for Oil and Gas Separators, 8th edition, October 2008, reaffirmed June 2023 and listed as withdrawn on 19 September 2024 (BSI record). Preview; retention-time table as reproduced in this summary.
  6. M. Moshfeghian, “Gas-liquid separators sizing parameter”, PetroSkills / John M. Campbell Tip of the Month, September 2015. Article; tables in the PDF version.
  7. M. Bothamley, “Gas/liquids separators, Part 2: Quantifying separation performance”, Oil and Gas Facilities (Campbell series). Copy online; mesh pad K table, liquid-load derating, Carpenter–Othmer pad model.
  8. Sulzer, Gas/Liquid Separation Technology brochure; and Applying the right mist elimination physics.
  9. AMACS, Mesh mist eliminators and Mist eliminator selection.
  10. Koch-Glitsch, DEMISTER mist eliminators brochure; style improvements table.
  11. Okutani Kanaami, Wire mesh mist eliminator styles.
  12. PetroWiki, Separator sizing and Inlet.
  13. Company inlet-momentum guidelines as quoted on Cheresources and Eng-Tips, including Shell DEP values.
  14. NORSOK P-100, Process Systems, Rev. 2, November 2001 (copy online); NORSOK P-002:2023+AC:2024, Process system design (edition notice; 2024 corrigendum). The horizontal K expression 0.137 (L/6)0.58 is from the P-100 errata sheet.
  15. IPS-E-PR-850, Process Requirements of Vessels, Reactors and Separators, June 1999, which refers gas–liquid separator design to IPS-E-PR-880. Copy online.
  16. API Standard 521, Pressure-relieving and Depressuring Systems, 7th edition, 2020, with Errata 1, 2022. Publication notice; errata listing.
  17. ASME BPVC Section VIII, Division 1. Formulas of UG-27 and UG-32 as summarized on CR4; UG-99 test pressure on an API 510 formula sheet; UG-125 accumulation in the Wisconsin adoption text.
  18. M. Henry, overpressure protection by system design and the move to ASME Section XIII, TAPPI 2023; Emerson, Kunkle data supplement.
  19. CSA B51, Boiler, pressure vessel, and pressure piping code: SCC record (B51:19 superseded by B51:24); Canadian Registration Number; 15 psig threshold.
  20. API RP 14C safety analysis tables for pressure vessels, as reproduced here; ISO 10418:2019.
  21. IEC 60534-2-1 constants and gas sizing method as given in Fisher Catalog 12, Section 2 and CheCalc.
  22. R. Turton and O. Levenspiel, “A short note on the drag correlation for spheres”, Powder Technology 47 (1986) 83–86; equation as implemented in polykin.
  23. Carpenter and Othmer, “Entrainment removal by a wire-mesh separator”, AIChE Journal, 1955. Abstract.
  24. N. Prophet and C. Lea, “Is your knockout drum adequately sized?”, AIChE Global Congress on Process Safety, 2022.
  25. SLB, Conventional horizontal separator product sheet, 2011.
  26. “Separator sizing” notes covering nozzle sizing, API RP 14E erosional velocity, vortex breakers and demister drains. Copy online.
  27. API Standard 520 Part I, 10th edition and Part II, 7th edition, 2020. Publication notice.
  28. C. Branan, Rules of Thumb for Chemical Engineers, Chapter 8, Separators/Accumulators; process heuristics summary.
  29. A. J. Kidnay, W. R. Parrish and D. G. McCartney, Fundamentals of Natural Gas Processing, on mesh pads; excerpt.
  30. Cheresources, Distance between mist eliminator top and vapor outlet nozzle.
  31. A. Aldeeb and Z. Kumana, “Understanding gas blowby scenario calculations”, AIChE GCPS 2015; “When a control valve phase change results in unfavorable control response”, GCPS 2023.
  32. TWI, Changes in ASME BPVC 2025 edition, Section VIII Division 1; CEI, What’s new in ASME Section VIII 2025.
  33. Separator design guide, Rev B, 2000, citing GPSA Figure 7-9. Copy online.
  34. Seoul National University course notes on gas–liquid separation and slug catchers. PDF.
  35. Separator design guidelines adapted from GPSA, on low liquid level and high-level shutdown spacing. Copy online.
  36. Notes on mesh pads and vanes (type 431 and type 326 pads, vane orientation). Copy online.
  37. Eng-Tips, Separator nozzle energy.
  38. Process engineering design course, Chapter 4, gas–liquid separator nozzle and inlet piping guidelines. Copy online.
  39. CheCalc, Vessel thickness comparison of ASME VIII, EN 13445 and PD 5500.
  40. Fire-case relief per API 521: Chemical Engineering, January 2010 (copy online); BRE; 7.6 m rule.
  41. Cheresources, Relief capacity of degassing drums after an HP separator (API 521 credit rules, restriction orifices, HIPPS).
  42. Engineers Edge, ASME dished head formulas, SA-516-70 at 20,000 psi.