What a Steam Separator Removes and Why

Why this matters

Two pieces of hardware on a steam line both take water out, and shops routinely buy one when the job needed the other. A trap removes water the flow has already dropped. A separator removes water the flow is still carrying. Those are different waters, held up by different physics, and neither device touches the other one's share.

The cost of confusing them is a customer who paid for a separator that solved nothing, or a device downstream that keeps eating trim while a perfectly good trap discharges every ninety seconds and the tech concludes the water problem is handled.

Everything below is diagnosed with non-contact instruments and eyes. A steam main and every fitting on it sits at supply temperature, a leak is invisible at the source, and nothing here calls for touching hot pipe or opening a live joint. Where a separator or trap has to be opened, the section is isolated at both ends, drained through its drip leg, vented to atmosphere, confirmed at zero on the upstream gauge, and locked and tagged under the written energy control procedure required by 29 CFR 1910.147 for stored pressure energy.

Two different waters

Steam leaving a boiler is never perfectly dry. Some water leaves with it as carryover, from priming, from a high water level, or from dissolved solids raising the surface tension in the drum. A companion group of articles owns boiler water chemistry and why carryover happens; cite those rather than re-deriving them. Take the carryover as given.

On top of that, the pipe itself makes water. Every foot of main loses heat, and the steam that gives up its latent heat at the pipe wall becomes condensate. On an insulated main this is a small continuous rate; on a cold startup it is a large slug, because the pipe metal itself has to be brought from room temperature to saturation temperature and every Btu for that comes out of condensing steam.

Where that water goes next depends entirely on how fast the steam is moving.

The flow regime is the gate

At low velocity, condensate on the wall runs down and along the bottom of the pipe as a stream. The steam is above it, the water is below it, and the two are separated by gravity. This is stratified flow, and geometry solves it: a drip leg at the low point, a takeoff from the top of the main rather than the side or bottom, a trap that keeps the leg drained.

Raise the velocity and that stops being true. The steam shears the surface of the water layer, tears droplets off it, and carries them. Past a certain point the water is a mist distributed through the flow and a film crawling around the whole inside of the pipe rather than a stream in the bottom of it. Gravity is no longer the dominant force on that water, so a top takeoff no longer excludes it and a drip leg no longer sees it, because it never reaches the bottom of the pipe long enough to fall out.

That is the gate for this entire subject. If the water arriving at your device is entrained by velocity rather than lying in the bottom of the pipe, only inertial separation removes it, and a drip leg and trap will keep discharging faithfully while the device downstream keeps drinking water.

Design velocities for saturated steam mains commonly land in the 4,000 to 6,000 feet per minute range, and mains at the top of that band are carrying entrained water as a matter of course. That is not a fault to be corrected; it is the trade a designer made to keep the pipe small.

What a separator does to a droplet

A separator forces the flow to change direction sharply. Steam, being about three orders of magnitude less dense than liquid water, follows the turn. A droplet has far more inertia per unit of drag area, keeps going, and hits a surface, where it joins a film that runs down into a collection bowl.

  steam plus droplets in
  ======================>
       \  \  \  \   baffle plates turn the flow
        \  \  \  \  sharply. Steam follows the
       \  \  \  \   turn, droplets carry straight
                    on and strike the plates.
                    ======================>
                    dry steam out
             bowl below collects the runoff
                    |
                  trap ---> condensate return

Baffle types do it with plates, as drawn. Cyclonic types do it by spinning the flow so centrifugal force throws the droplets to the wall. Coalescing types run the flow through a mesh so small droplets merge into large ones that drain. All three work on the same principle, which is that liquid and vapor have very different inertia, and all three publish their performance at a rated flow, never at a pipe size.

The velocity band cuts at both ends

This is the part that gets missed, and it is why a separator selected off the line size fails.

Too slow and there is no separation. Inertial separation needs inertia. Below the separator's rated band, droplets follow the turn along with the steam and pass straight through. A separator sitting on a main that runs at 20 percent of design load for most of the year is doing very little for most of the year.

Too fast and the separator re-entrains. Above the band, the film running down the collecting surfaces gets sheared off again and picked back up by the flow, and a cyclonic type will carry water straight out of the top. Manufacturers publish a maximum flow for a reason, and exceeding it does not degrade performance gently, it reverses it.

So a separator is selected on the flow range the line will actually see, at the pressure it will actually see, from the manufacturer's capacity data. That data is the authority; a pipe size is not a selection.

Case one: the humidifier that needed a pipe fitting

Air handler humidifier spitting liquid water into the duct, showing up as wet insulation downstream of the manifold. The steam source is a 4 in main at 100 psig.

The branch takeoff came out of the bottom of the main. That is the whole finding. Water lying in the bottom of the pipe drains straight down the branch by gravity, in preference to going anywhere else. At the flows this branch sees, this is stratified water and it is being fed directly into the device.

The fix is a takeoff from the top of the main, a full-size drip leg at the base of the branch drop, and a trap on that leg sized on the branch's own condensate load. No separator, because the water in question was never entrained; it was lying in the pipe waiting for the shortest way down.

A shop that answered this complaint with a separator would have installed it, seen improvement, and drawn the wrong conclusion. Improvement, because a separator's bowl will happily catch water that pours into it too. The wrong conclusion, because the pipe would still be feeding the branch a stream of water, the separator would be operating far outside what it was selected for, and the next similar complaint would get the same expensive answer.

Case two: the humidifier that needed a separator

Second air handler in the same building. Takeoff already from the top of the main, drip leg present and correctly sized, trap discharging on a normal cycle. Still spitting.

Run the velocity. The main is 4 in schedule 40, inside diameter about 4.026 in, cross-sectional area about 0.0884 square feet. Design flow is 6,000 lb/hr, and at 100 psig saturated steam occupies about 3.89 cubic feet per pound from the steam tables. That is about 23,300 cubic feet per hour, or roughly 389 cubic feet per minute, which through 0.0884 square feet is about 4,400 feet per minute.

At 4,400 feet per minute the flow in that main is not stratified. The takeoff geometry is correct and it is excluding nothing, because there is no bottom stream to exclude. The water is entrained, and it goes wherever the steam goes.

This is the separator case, and it is the same building, the same main, and a nearly identical symptom as case one. The two are told apart by one number that neither complaint contained.

There is a second finding buried in that arithmetic. At 20 percent load, 1,200 lb/hr, the same main runs at roughly 880 feet per minute. The flow at that velocity is stratified again, the entrainment problem largely disappears, and a separator selected only on the 6,000 lb/hr design flow is operating well below its band. If the humidifier only misbehaves in cold weather, that is consistent, and it is worth confirming before selecting, because the manufacturer's capacity data will tell you what that separator does across the whole range rather than at one point on it.

The separator's own trap is not optional, and it fails backwards

A separator collects water into a bowl. Something has to take it out, continuously, against the full line pressure. That is a trap, sized on the separated load at the full differential, with a full-size drop from the bowl and a dirt pocket below the outlet so scale settles somewhere other than the trap seat.

Here is the failure mode that makes this worth its own section. A separator whose trap has failed closed is worse than no separator at all. The bowl fills, the water level rises into the flow path, and the passing steam picks water back off the surface of a standing pool. You have installed a reservoir in the main and connected it to the flow. The device downstream gets wetter than it was before the separator existed, and every visible sign points away from the separator, because the separator is the thing that was supposed to fix this.

Confirm the separator's trap is discharging before you conclude anything about the separator. Read the trap outlet with a non-contact infrared thermometer from arm's length, or listen through the body with an ultrasonic instrument, both of which keep your hands off a component sitting at supply temperature. Where the trap has a test valve piped to a safe discharge point, it may be cracked open slowly with your body clear of the discharge path and wearing the eye, face, and body protection called for by your shop's hazard assessment under 29 CFR 1910.132, because what comes out is condensate that flashes to steam as it leaves.

Where it goes in the line

A separator belongs immediately ahead of whatever it is protecting, with as little pipe between the two as the layout allows, because pipe downstream of the separator makes fresh condensate that the separator cannot retroactively remove.

That places one on the inlet side of a pressure reducing station, ahead of the strainer, because wet steam cuts a reducing valve seat faster than dirt does. It places one ahead of a turbine, a culinary or humidification takeoff, and any process where product contacts the steam. It does not place one at the boiler outlet as a substitute for fixing carryover, because a separator downstream of a priming boiler is treating a symptom that chemistry and level control own.

References

• Separator manufacturer capacity data, which is the governing authority for the flow range a given body will separate at a given pressure, and for the minimum flow below which it stops working
• Saturated steam tables from any engineering handbook, for the specific volume used in the velocity calculation
• 29 CFR 1910.132, the OSHA general industry standard requiring a hazard assessment and appropriate personal protective equipment, and 29 CFR 1910.147, the energy control standard for isolating stored pressure energy before opening a separator or trap
• See related: What Wet Steam Costs the Equipment Downstream; What a Pressure Reducing Station Is Managing; What a Boiler Does to the Water in It