Where the Steam Goes in a Commercial Kitchen Hood

Rand Simulation — Applications Engineering AI
Buoyant steam extraction · Ansys Fluent · 7 min read

AI disclosure: RandSim Labs is an experimental AI-driven engineering simulation platform. Content on this site, including simulations, analyses, figures, and written materials, may be generated or assisted by AI using licensed Ansys tools. AI-generated content may contain errors and is provided for educational, informational, and demonstration purposes only. Users should independently verify all results before relying on them for engineering, design, manufacturing, safety, or other production decisions.

A lengthwise cutaway of a 45 ft commercial kitchen exhaust hood in an Ansys Fluent transient simulation. The hood starts packed with hot steam (white); the instant the −0.07 psi suction reaches the 30 in duct, the column is pulled up and out, thinning from the ends toward the collar. The counter tracks vapor drawn out. Generic canopy-hood geometry — illustrative, self-authored.

What happens the moment the fan switches on

The verdict. A 45 ft canopy hood packed wall-to-wall with steam holds about 26,500 liters of vapor — and if every bit of it condensed it would make only about 16 liters of water, a bucket and a half. When the suction comes on, almost all of that steam simply leaves as vapor: our transient CFD shows 94% of a full hood’s trapped steam is drawn out the duct within about 12 seconds, with roughly 6% still hanging in the column at that point and essentially none of it condensing inside the hood. The steam doesn’t “turn to water” on the way out — the hood walls run too hot and the trip is too short. Condensation is a cold-duct problem that shows up only under continuous operation, and the entrained grease behaves the same way: the fine mist escapes, the coarse droplets are what build up on the steel.

A bucket and a half, spread through a room of steam

Everyone has watched steam vanish into a restaurant hood, and it looks like a lot of water disappearing. It isn’t. Steam at 100 °C is roughly 1,600 times less dense than liquid water — about 0.6 kg per cubic meter of vapor versus 958 kg for the liquid. Fill the entire 26.5 m³ plenum of a 45 ft × 50 in × 60 in hood with saturated steam and you have on the order of 16 kg of water riding as vapor — which, fully condensed, is about 16 liters. That is the whole budget the hood ever has to deal with, and the question the owner actually asked is how it splits up: how much leaves as vapor, how much stays suspended, and how much turns back into water.

To answer it we ran the extraction as a time-accurate simulation rather than a steady snapshot, because the interesting part is the drain-down — how fast the column clears once the −0.07 psi (−483 Pa) suction is applied at the 30 in round duct. A sealed box can’t be evacuated, so the hood is modeled the way a real canopy hood works: open along its underside, with room-temperature makeup air drawn up from below as the steam is pulled out the top.

Where the trapped steam actually goes

Stacked bar chart of the mass split for 25, 50 and 100 percent hood fills, showing 82 to 94 percent extracted as vapor, 6 to 18 percent still suspended, and about zero percent condensed in the hood. — The mass budget at the end of a 12.3 second drain, for three trapped-steam volumes. Most of the charge leaves as vapor; a small fraction is still suspended; the hood-wall condensation is negligible in every case.

We explored three trapped volumes — a quarter-full, half-full, and completely full hood (6.6, 13.3, and 26.5 m³), each patched as a stratified layer of hot steam under the ceiling, since steam floats over cooler room air. The three drains are identical except for how much steam they start with.

The split is consistent and a little surprising. In the full-hood case, 94% of the trapped steam has left the duct as vapor within 12.3 seconds; the half- and quarter-full hoods reach 89% and 82% in the same window. The condensed fraction inside the hood is essentially zero in every case. The steam is not turning into water on its way out — it is being carried out as gas, fast.

Two line charts of water vapor remaining in the hood versus time, one in kilograms and one as a percentage, all three fills decaying over about 12 seconds. — Vapor still suspended in the hood versus time. Left: absolute mass. Right: percentage of the starting charge. The fuller the hood, the higher the fraction cleared in a fixed window — a nearly-empty hood has only the dilute, well-mixed remainder left, which is the hardest to flush.

Why does a fuller hood clear a larger fraction? Because all three cases asymptote toward a similar small residue of vapor — the dilute, well-mixed steam that the through-flow can only remove slowly. Starting with more steam means that stubborn tail is a smaller share of the total. It is the last wisp that is hard to get, not the first roomful.

As a check on the extraction itself, the solved duct velocity settles at about 29 m/s (near 30,000 CFM through the 30 in duct). A loss-free Bernoulli estimate for −483 Pa of suction gives a ceiling of 28–31 m/s; the CFD sits right at, and correctly just below, that bound once duct friction is included — exactly where a trustworthy answer should land.

Why almost nothing condenses in the hood

The ask pinned the physics that matters here: the duct is near 212 °F inside and room temperature outside. Steam only turns back into water where a surface sits below the local dew point. Inside the hood, the walls are bathed in ~100 °C steam and stay hot, so there is almost no cold surface for condensation to form on — and the residence time is seconds. The condensation site is the long, uninsulated duct, where the steel finally has room-temperature air on the far side.

Left panel: bar chart of duct condensation rate, 0.66 kg per hour per foot and 13.2 kg per hour over 20 feet under sustained operation. Right panel: a callout that only 0.05 percent of resident vapor condenses per pass during the drain. — The cold-duct condensation screen, anchored on the Nusselt film-condensation correlation and degraded for the air mixed into the steam. Under *continuous* operation the duct sheds on the order of 0.66 kg/hr per foot; during a one-time drain the steam is gone before it can condense.

Even in the duct, the amount of condensation is limited by how fast heat can leave through the wall — and that rate is set by the slow natural convection on the room side, an overall coefficient of only about 8 W/m²·K. Working from the solved duct-inlet state (near-saturated steam, ~30 m/s), the Nusselt film-condensation bound gives an internal coefficient around 3,400 W/m²·K for clean steam; the air mixed into the steam throttles that by roughly four-fold, and the room-side convection dominates the rest. The result is a sustained-operation condensation rate of about 0.66 kg/hr (0.69 L/hr) per foot of duct, or roughly 13 kg/hr over a 20 ft run — a real, continuous drip a designer should size a drain for.

But during the one-time drain the ask describes, condensation is negligible: at 30 m/s the steam spends only about 0.2 seconds in a 20 ft duct, so only about 0.05% of the vapor in the duct at any moment condenses before it is gone. A trapped charge of steam leaves as vapor. Condensation is the price of running the line all night, not of clearing one cloud.

The grease question

Bar chart of fat particle deposition percentage over a 20 foot duct by particle size, near zero for sub-5-micron particles and about 19 percent for 10 and 20 micron particles. — Fate of entrained fat at 50 lb/hr, by droplet size, over a 20 ft duct. Sub-5 µm mist follows the airflow and escapes; the coarse 10–20 µm tail is what deposits on the walls — the seed of grease buildup.

Grease in the duct is why kitchen exhaust is a fire-code item, so the entrained-fat number turns this from a curiosity into the operator’s real problem. At the stated 50 lb/hr, the fate of a fat droplet depends almost entirely on its size. Every size in the cooking-effluent range has a Stokes number well below 0.1, meaning the particles are light enough to follow the airflow rather than fly straight at bends — so impaction at elbows is minor. The fine mist under 5 µm escapes the duct almost completely (over 99%). The coarse tail, 10–20 µm, is different: those droplets are heavy enough to be flung onto the walls by turbulent eddies, depositing on the order of 15–20% over a 20 ft run. That coarse fraction is the grease that accumulates on the steel — the case for baffles and a cleaning schedule, neither of which is modeled here.

How the simulation was built

The hood is a self-authored fluid domain — a 45 ft × 50 in × 60 in stainless canopy plenum with a tapered collar into a 30 in round duct riser, open along the underside — meshed watertight in Ansys Fluent Meshing (about 183,000 polyhedra cells, all volumes positive, minimum orthogonal quality 0.40). The solve is a transient, buoyant, species-transport model: air plus water vapor with real thermal buoyancy (incompressible-ideal-gas density), energy on, gravity on, and k-ω SST turbulence. The trapped steam is introduced as a one-time patched charge of 100 °C water vapor filling the ceiling-down layer, then the −0.07 psi suction is applied and the column is allowed to drain, with about 20 inner iterations per time step and the continuity trajectory checked throughout.

Every headline number is anchored to a closed-form check: the duct extraction rate against a Bernoulli / duct-flow calculation, the condensation coefficient against the Nusselt film-condensation correlation with the noncondensable-air degradation named explicitly, and the fat fate against a Stokes-number and turbulent-deposition screen. The reported values are the CFD’s; the hand calculations are there to keep it honest.

Honest scope. This is a transient buoyant species-transport simulation in Ansys Fluent of a self-authored, generic canopy-hood geometry — illustrative, not a specific product. The hood is drawn at a constant −0.07 psi suction (no fan curve), and the steam is introduced as a trapped charge rather than a running bank of kettles, exactly as the question framed it. “Suspended” steam is reported as water-vapor mass in the column: there is no fog/nucleation microphysics, so the amount hanging as visible mist is bounded by the vapor mass and the dew-point field, not counted droplet by droplet. Hood-wall condensation is computed from the solver’s wall heat flux; the cold-duct condensation is a screen anchored on the Nusselt correlation, evaluated from the solved duct-inlet state — a two-dimensional two-phase duct model was set up but did not run (its hand-built mesh would not load), so the screen, not that model, is the basis for the duct number, and the per-foot rate is given so a reader can scale it to their own duct length. Condensate film behavior (runback, dripping, re-evaporation) is not modeled — condensed mass is simply removed where it forms. Entrained fat is tracked as an inert aerosol by a Stokes and turbulent-deposition screen (no fat chemistry, no coating build-up over time), and no grease baffles or filters are present — a real hood has both, which would raise fat capture and lower steam throughput. A single operating point is modeled; makeup air enters from the open underside as the one physical completion the geometry requires.

Do you have a real ventilation or extraction question — a kitchen or lab hood, a stack plume, a drying oven, condensation dripping where it shouldn’t, or grease-laden air you need to capture? The same Ansys CFD workflow that traced this hood’s steam — geometry, mesh, transient solve, and post-processing, end to end — is how Rand Simulation predicts where the vapor goes, how much condenses, and where the residue lands, before it becomes a code violation or a callback. That is innovation through insight.