› Tunnel Fire CFD Ventilation Critical Velocity Simulator Back
Tunnel Fire Safety

Tunnel Fire CFD Ventilation Critical Velocity Simulator

Compute the critical velocity V_c — the minimum upstream airflow that prevents smoke backlayering in a road-tunnel fire — using the Heselden-Thomas / Kennedy correlation, and compare it with the actual airflow generated by jet fans. Change the fire scenario (car, bus, HGV, tanker) and the tunnel cross-section to evaluate longitudinal smoke control in real time.

Parameters
Tunnel length
m
Height H
m
Width W
m
Fire size HRR
Design heat release rate (NFPA 502)
Jet fan thrust F
N
Number of fans n
Fan spacing
m
Ambient temperature T_amb
°C
Results
—
Hydraulic dia. D_eq (m)
—
HRR (MW)
—
Critical velocity V_c (m/s)
—
Fan airflow (m/s)
—
Velocity ratio V_fan/V_c
—
Visibility (m)
—
Tunnel section & smoke dispersion animation

Smoke rising from the fire (red) is pushed downstream by the jet fans (blue arrows). When the airflow drops below V_c, smoke flows upstream (backlayering) and evacuees lose visibility.

Critical velocity V_c vs HRR
Per-scenario V_c comparison
Theory & Key Formulas

$$V_c = K_1\left[\frac{g\,H\,Q}{\rho\,c_p\,T\,A}\right]^{1/3},\qquad V_{fan} = \sqrt{\frac{2\,F_{total}}{\rho\,A\,(1+f\,L/D)}}$$

V_c: critical velocity (K1 = 0.61 to 0.85, this tool uses 0.85), H: tunnel height, Q: fire HRR (W), A: cross-section area, F_total: total jet-fan thrust.

$$D_{eq} = \frac{4A}{P_{wet}} = \frac{4HW}{2(H+W)},\qquad f\,L/D \approx 0.02\,L/D_{eq}$$

Hydraulic diameter D_eq and tunnel-length friction loss coefficient. L: tunnel length, f: Moody friction factor (about 0.02 for rough concrete).

Tunnel fires — ventilation, evacuation, critical velocity and smoke control

🙋
Are tunnel fires really that dangerous? If there's a fire, you just walk out the portal, right?
🎓
It is much worse than people think. In 1999 the Mont Blanc Tunnel fire (France-Italy) started from an HGV engine bay and killed 39. St. Gotthard 2001 killed 11 and Tauern 1999 killed 12. In every case the victims did not burn — they suffocated or got disoriented in the smoke and could not move. That is why modern tunnel safety is mainly about smoke control, not the flames themselves.
🙋
So the heart of smoke control is this "critical velocity V_c". The default setting gives 3.47 m/s — is that fast or slow?
🎓
About a noticeable breeze on your skin — 3.5 m/s is roughly 12 km/h. Below that, the buoyant smoke plume from the fire starts flowing upstream and covers the evacuation route. That is backlayering. The Heselden-Thomas (1977) and Kennedy (1996) form V_c = 0.85·(gHQ/(rho·cp·T·A))^(1/3) comes from balancing buoyancy against inertia, and every major code (NFPA 502, PIARC C3.3) still uses it as the design benchmark.
🙋
What is interesting is that going from a 30 MW bus to a 300 MW tanker barely raises V_c. Is that the cube-root effect?
🎓
Exactly — Q^(1/3) flattens it out. A 10x increase in HRR only multiplies V_c by 10^(1/3) ≈ 2.15. That is a fortunate property: even for worst-case HRR (HGV 100-300 MW), if you can hold 3-4 m/s you are practically fine. The caveat is that larger fires throttle the airflow themselves — the hot smoke plug adds resistance and the fans lose efficiency. We simplify that here, but a real FDS or ANSYS Fluent CFD model resolves it explicitly.
🙋
Even 12 fans at 1000 N each only push 4.35 m/s. The efficiency is surprisingly low.
🎓
Good catch. Most of that thrust is eaten by wall friction along the tunnel length. With the defaults — 5 km long and D_eq = 8.84 m — the f·L/D term is about 11.3, so the speed is only 1/sqrt(12) ≈ 0.29x of the frictionless case. That is why very long tunnels (10+ km) have intermediate ventilation shafts that split the tunnel into independent zones. Eurotunnel and the Seikan Tunnel are classic examples.
🙋
One last question — what is the biggest pitfall designers fall into?
🎓
Trusting the steady-state V_c alone. A real fire's HRR ramps up over 5 to 15 minutes (t-squared growth). During that ramp there is a window where smoke is still stratified along the ceiling and the situation is safe even at V_fan < V_c. If you blast the fans to full speed too early, you break the stratification and mix smoke down to floor level — visibility actually gets worse. That is the smoke-stratification problem and it has been the headline research topic since Mont Blanc. Today, the standard is to validate a staged "detect, then ramp up fans" control philosophy in transient CFD before commissioning.

Frequently asked questions

The critical velocity V_c is the minimum upstream ventilation speed required to completely prevent backlayering — the upstream propagation of hot smoke that occurs in a tunnel fire. The semi-empirical Heselden-Thomas (1977) and Kennedy (1996) correlation gives V_c = K1 · [g·H·Q / (rho·cp·T·A)]^(1/3). Typical road-tunnel values are 2.5 to 3.5 m/s. Below that, smoke covers the upstream evacuation path and visibility and breathability deteriorate quickly. NFPA 502 and PIARC C3.3 both use this correlation as the design benchmark for longitudinal ventilation.
Take the per-fan thrust F (typically 500 to 3000 N) and the required total thrust F_total, then invert F_total = rho·A·V^2·(1 + f·L/D) to get the fan count n = F_total / F. Set V to 1.2 to 1.3 times V_c to keep a safety margin. Then add allowances for fan-pitch interference loss (about 10 to 15%), wind pressure at the portals, and the throttling effect of the hot smoke plug (which grows with HRR). In practice, jet fans are spaced 100 to 150 m apart.
Longitudinal ventilation pushes air in a single direction with jet fans and is best for one-way urban tunnels and short road tunnels. Transverse ventilation uses dedicated supply and exhaust ducts that feed and extract uniformly along the whole tunnel length, and is preferred for long bi-directional tunnels (over 5 km). Semi-transverse is intermediate: uniform supply but exhaust only at the portals. Construction cost grows longitudinal < semi-transverse < transverse. Since Mont Blanc 1999, long bi-directional tunnels typically combine transverse ventilation with concentrated extraction directly above the fire.
Tenability limits from ISO 13571 and PIARC are: visibility 10 m or more, CO 300 ppm or less, temperature 60 C or less and radiant heat 2.5 kW/m^2 or less. Below 5 m visibility evacuees lose direction and panic and disorientation rise sharply. Evacuation time is estimated assuming a walking speed of 5 km/h (a typical emergency average) to reach the nearest cross-passage (350 m spacing is standard). This tool uses a simplified smoke density that scales with HRR and flags the situation as dangerous once the limit is reached.

Real-world applications

Ventilation design of long mountain road tunnels: for road tunnels of about 5 km or longer (Kanetsu 11 km, Tokyo Bay Aqua-Line 9.5 km, Shin-Nagasaki 6.2 km, Frejus, Brenner Base) the design uses the same Heselden-Thomas / Kennedy basis as this tool. In Japan the rules are in NEXCO's road-tunnel design standard (ventilation), in Europe PIARC Report 05.16.B, and in the United States NFPA 502. The design fire size is typically 30 MW (bus) or 100 MW (HGV), and the fan layout is sized to meet V_c for that fire.

Metro and railway tunnel evacuation planning: longitudinal ventilation is also used in Shinkansen tunnels, metro systems, the Seikan Tunnel and Eurotunnel (50 km under the English Channel). Since the Daegu Subway fire of 2003 (192 fatalities, South Korea), rolling-stock flammability has been reduced and tunnel ventilation plus safe-haven planning has become an international requirement. Eurotunnel's Service Tunnel keeps the central evacuation passage at positive pressure, so it is always higher than the smoke pressure and smoke cannot enter — an advanced design example.

Urban underground roads and tunnels: in the Tokyo Yamate Tunnel (18.2 km, the world's longest urban road tunnel), the Tokyo Metropolitan Expressway Central Circular, Boston's Big Dig and Stockholm's Northern Link, ventilation shafts are combined with zoned ventilation. In a fire all jet fans within the affected zone are activated in unison and smoke is exhausted through the nearest shaft. Modern design runs hundreds of CFD cases (FDS, ANSYS Fluent, STAR-CCM+) covering combinations of fire location, wind direction and fan-failure patterns to confirm V_c is met in every one.

Underground car parks and concourses: the same physics drives smoke-control systems in underground car parks and underground shopping streets. NFPA 92 sets the required exhaust flow per floor area from the smoke layer interface height, which essentially inverts a V_c equivalent. Tunnel know-how transfers directly to everyday building fire safety.

Common misconceptions and pitfalls

The most common pitfall is thinking "V_c met = safe". V_c only prevents upstream smoke backlayering — it does not guarantee safety for evacuees downstream. On the downstream side the smoke is pushed past evacuees at high speed, so people trapped downstream of the fire lose visibility and find evacuation extremely difficult. Many of the Mont Blanc 1999 victims were in fact trapped in their downstream vehicles. Modern design therefore couples "drive evacuees upstream of the fire" with "stop traffic at the portal" systems based on CCTV and gates.

The second pitfall is over-strong airflow that destroys the smoke layer. Counter-intuitively, the theoretical optimum is exactly 1.0 V_c — pushing harder can be worse. Vantelon et al. (1991) wind-tunnel experiments show that once the airflow exceeds 1.3 to 1.5 V_c, the hot smoke layer trapped against the ceiling above the fire is entrained downward — "smoke plunging" — and reaches head-height of evacuees. That is why the verdict in this tool labels v_ratio ≥ 1.2 as ok and below as warn: too much margin is also a hazard.

The third pitfall is relying on steady-state analysis alone. This tool, like every textbook V_c calc, is steady-state. A real fire's HRR grows in time (t-squared) and the time from ignition to full-speed fan operation includes (1) detection (30 seconds to several minutes), (2) operator decision, and (3) fan start-up sequencing (sequential, 60-120 s to full bank). Smoke has already spread upstream during that delay. NFPA 502 therefore recommends transient CFD (transient FDS) to optimize the start-up sequence. Leading designs also use Monte Carlo to account for single-fan failure (n-1 design), power outage (30-minute backup), and throttling-induced resistance growth.

How to Use

  1. Enter tunnel geometry: length (m), height (m), and width (m) to establish cross-sectional area and hydraulic diameter D_eq.
  2. Input jet fan thrust (N) and heat release rate (MW) from fire scenario; the simulator calculates induced airflow velocity and critical velocity V_c using empirical correlations (e.g., Thomas formula or CFD-derived coefficients).
  3. Review output metrics: critical velocity V_c (m/s) to prevent backlayering, fan velocity ratio V_fan/V_c, and resulting visibility distance accounting for smoke optical density.

Worked Example

Road tunnel 800 m long, 7.5 m high, 10 m wide (A = 75 m², D_eq ≈ 8.6 m). Fire releases 50 MW heat; jet fan produces 120,000 N thrust (~35 m/s airflow). Critical velocity V_c calculated at 2.8 m/s using V_c = K × (HRR / (ρ × g × D_eq))^0.33, where K ≈ 0.22. Fan delivers 35 m/s, yielding V_fan/V_c = 12.5 (excellent margin). Visibility extends to 40 m with smoke extinction coefficient 0.15 m⁻¹.

Practical Notes

  1. Critical velocity scales with HRR^(1/3): doubling fire size from 50 MW to 100 MW increases V_c by only 26%, but backlayering risk rises sharply if fan capacity drops below V_c threshold.
  2. Hydraulic diameter D_eq = 4A/P; rectangular tunnels with aspect ratio >3:1 require larger D_eq adjustment, reducing V_c and improving safety margins for same fan power.
  3. Visibility (m) = 3/K_e, where K_e includes flame radiation and soot; at 50 MW fires, visibility typically drops to 10–20 m without ventilation; maintain V_fan/V_c > 1.5 for operational buffer.