ASET vs RSET: The Inequality Behind Every Evacuation Plan

Every performance-based fire design ends in one line of arithmetic. The time until smoke makes a room untenable must beat the time people need to get out. Engineers call the first term ASET and the second RSET, so they write the test as ASET > RSET. The line looks simple, yet that simplicity hides the trouble. Each side rests on choices that can swing the answer by a factor of two or more. So this post walks through both sides of ASET vs RSET, the numbers behind them, and the term that deserves most of your doubt.

TL;DR

What ASET vs RSET means

• ASET, the available safe egress time, runs from ignition to the moment a room fails a tenability limit. RSET, the required safe egress time, runs until the last person reaches safety.
• The design passes when ASET > RSET with a margin. Practice uses a factor of 1.5 to 2.0, while Italy’s fire code asks for a margin of 100% of RSET.
• RSET splits into detection, alarm, pre-movement and travel. Pre-movement dominates, since it ranges from seconds in a drilled office to an hour in a hotel at night.
• ASET hangs on the design fire and on the tenability limit you pick. The usual limits read 10 m visibility, 60 °C air, 2.5 kW/m² of radiant heat, or a toxic dose (FED) of 0.3.

Why ASET vs RSET needs care

• One fixed pass/fail number hides the spread in both terms. So Monte Carlo work now reports a probability of failure instead.
• Uncoupled models flatter the design. At the Station nightclub, NIST put ASET near 90 s. A coupled model then matched the death toll only after a 15 s shift in the fire timeline.
• EV fires in car parks add hydrogen fluoride. So toxic dose can cut tenability time by up to 47% against a small battery.

What do ASET and RSET mean?

ASET stands for available safe egress time, the time until conditions in a space turn untenable. RSET stands for required safe egress time, the time occupants need to reach a place of safety.

According to ISO/TR 16738 and PD 7974-6, both terms carry these definitions. Both documents also state the ASET vs RSET goal in one sentence. The available time must exceed the required time by an adequate margin of safety. Write it out and you get the ASET vs RSET inequality.

$\text{ASET} > \text{RSET}$

Two numbers then describe how well a design passes. The safety margin takes the difference, while the safety factor takes the ratio. CFPA-E Guideline 19 writes the margin as follows.

$t_{\text{safety}} = t_{\text{ASET}} - t_{\text{RSET}}$

Prescriptive codes never show this sum. They fix travel distances, exit widths and door counts, and they hide the inequality inside those tables. Performance-based design drags it into the open. So an engineer can trade a longer travel distance for a smoke control system, then prove the swap with numbers. NFPA 101 calls this the performance-based option, while Australia’s NCC calls it a Performance Solution.

Where did ASET vs RSET come from?

From Leonard Cooper at the US National Bureau of Standards, now NIST, in the early 1980s.

Cooper published “A concept for estimating available safe egress time in fires” in Fire Safety Journal in 1983. A year earlier, he and Stroup had released the ASET computer program. His model tracked a hot layer filling a room from the ceiling down, a two-zone approach that lives on in CFAST. Cooper focused on the available side, though. The required side stayed a concept, and the RSET timeline that engineers now use grew over the next two decades. Babrauskas traces the root further back still, to smoke detector work at NBS in the 1970s.

How much margin does ASET vs RSET need?

No universal value exists. Most designers adopt a safety factor of 1.5 to 2.0, though a few codes set their own floor.

Jurisdiction or guide	Required margin	Note
SFPE Engineering Guide (US)	None codified	Analyst agrees a value with the authority
Common practice	Factor 1.5–2.0	Applied to the ratio ASET/RSET
Italy, Codice di Prevenzione Incendi	Margin ≥ 100% of RSET	Drops to ≥ 10% only with drill data
Singapore fire engineering guidelines	Factor ≥ 1.2	Used in sensitivity checks
Some car park CFD studies	60 °C limit cut to 52 °C	A 15% input margin for model error

The Italian rule deserves a closer look. Its chapter M3 asks for an ASET at least double the RSET, so a 5 min evacuation needs a 10 min ASET. The rule relaxes to a 10% margin only with drill-quality input, namely evacuation parameters taken from a real egress exercise. That relaxation tells you where the code writers see the risk. They see it in the human inputs rather than in the fire model.

Many engineers prefer to bury conservatism in the inputs instead. First they pick a faster design fire and a longer pre-movement time. Then they report the margin that falls out. Either route works, as long as the report states which one it took.

What goes into RSET?

Four blocks in a row. Detection, alarm, pre-movement and travel add up to the required time.

$t_{\text{RSET}} = t_{\text{det}} + t_{\text{alarm}} + t_{\text{pre}} + t_{\text{travel}}$

The figure shows an illustrative retail unit under a fast fire. The smoke layer reaches the 10 m visibility limit at 420 s. Detection and alarm together take under a minute, while travel takes 90 s in both rows. Only the pre-movement block changes between the rows, yet the ASET vs RSET safety factor drops from 1.6 to 1.1. So that single block deserves most of the attention in any ASET vs RSET report.

How long does detection take?

Tens of seconds for a fast fire under a normal ceiling, though far longer for a smouldering one.

Heat detectors and sprinkler heads follow the response time index, or RTI. A hot ceiling jet with speed $u$ and temperature $T_g$ warms the element at the rate below.

$\frac{dT_d}{dt} = \frac{u^{1/2}}{\text{RTI}}\,(T_g - T_d)$

The fire itself grows as $Q = \alpha t^n$. Programs such as DETACT-QS, released by Evans and Stroup at NIST in 1988, solve that pair with Alpert’s ceiling jet correlations. Smoke detectors use an optical density limit or a critical gas speed at the head instead. Zone models and CFD also do the same job with more detail.

Alarm time then covers the gap between a detector trip and a signal that people hear. An automatic system adds close to nothing. A staff alarm with an investigation phase adds minutes, though, and PD 7974-6 lists default allowances for each case.

Why does pre-movement time dominate RSET?

Because it spans three orders of magnitude, from seconds in a drilled office to an hour in a sleeping hotel. Walking speed varies by a factor of two at most.

Pre-movement covers recognition, the time to notice and read the cue, and then response, the time to decide and act. People finish a phone call, gather a bag, look for a colleague, or go to check the smell. PD 7974-6 tracks the first few occupants at the 1st percentile and the last few at the 99th. It then ties both to four factors. Do people sleep, do they know the building, how dense does the crowd sit, and how complex does the space feel?

The standard sorts buildings into design behavioural scenarios:

• A: awake and familiar, such as offices and factories.
• B1: awake, unfamiliar, high density, such as shops and restaurants.
• B2: the same crowd in one room with a focal point, such as a cinema.
• Ci to Ciii: asleep and familiar, from a family home to halls of residence.
• D: asleep and unfamiliar, such as hotels.
• E: medical care, where staff must move patients.

Table C.1 then gives default pre-movement times for three levels of fire safety management. A well-drilled office empties in a few seconds to a few minutes. Retail with trained staff who sweep the floor runs about as fast. Hotels sit at the other end, though. The standard warns of pre-movement times “very long and variable.” Even with very good management, the default 1st-percentile figure for a sleeping occupancy runs near 15 minutes. Average management then pushes the range to 30–60 minutes.

Put those numbers beside a fire. A fast t² fire reaches 1 MW in 150 s. So a hotel corridor can lose visibility long before the first guest opens a door. That gap explains why hotels lean so hard on compartmentation and on alarm sounders in every room.

How does the hydraulic model compute travel time?

By treating people as a fluid. Speed falls with crowd density, and the flow through a door tops out near 1.3 persons per second per metre of width.

The SFPE Handbook chapter by Nelson and MacLennan, later revised by Gwynne and Rosenbaum, gives the rules. Below a density $D$ of 0.54 persons per square metre, people walk at their free speed of about 1.19 m/s. Above that, speed falls in a straight line.

$S = k - a\,k\,D$

Here $a = 0.266$, while $k = 1.40$ for corridors, ramps and level floor. Stairs take a lower $k$ that depends on riser and tread. Movement stops near 3.8 persons per square metre. Specific flow then multiplies speed by density.

$F_s = S\,D$

The flow through an element multiplies that by effective width.

$F_c = F_s\,W_e$

Effective width subtracts a boundary layer from each edge, about 150 mm per wall and more beside handrails or seats. Nobody walks with a shoulder on the plaster. Specific flow peaks at about 1.3 persons per second per metre of effective width on level routes, and near 1.0 on stairs.

That peak sets the queue. A theatre with 1,000 people and 3 m of effective exit width can push out about 3.9 people per second. So the last person clears the doors some 4 min after the first, no matter how fast anyone walks. In dense assembly rooms, flow rather than pre-movement rules RSET. Other frameworks reach similar sums by other routes. Predtechenskii and Milinskii work in square metres of person per square metre of floor, for instance. Fruin ranks crowding by level of service A to F, while Togawa’s formula folds the queue into one line.

What else shapes RSET?

Behaviour. People move toward the familiar exit and toward their own group, and they investigate before they leave.

Staff who sweep occupants to the exits cut pre-movement more than any other single measure. Phased evacuation, stay-put strategies, refuges and evacuation lifts also change the sum. Yet standard analyses often skip the people who cannot use a stair. ISO/TR 16738 even leaves lifts out of scope, which tells you how young that part of the field remains.

What sets ASET?

Whichever tenability limit fails first. Engineers track five, and visibility usually loses the race.

Limit	Common threshold	Where checked
Visibility	10 m in large rooms, 5 m in small ones	Head height, 1.8–2.0 m
Convected heat	60 °C, though a few guides allow 100 °C	Head height
Radiant heat	2.5 kW/m², about a 200 °C hot layer	On the occupant
Smoke layer	Clear layer above 2.0 m, or 2.5 m	Whole room
Toxic dose (FED)	0.3 for design, 1.0 at incapacitation	Breathing zone

New Zealand’s C/VM2 shows a typical set. Visibility must stay above 10 m in rooms over 100 m², radiation below 2.5 kW/m², and air below 60 °C. The fractional effective dose for both CO and heat must also stay below 0.3, all checked at 2.0 m. NFPA 130 uses a light attenuation coefficient of 0.267 per metre for rail stations instead. That keeps signs readable at 30 m and doors visible at 10 m. Sweden’s BBRAD ties the smoke layer to room height $H$ in metres.

$z_{\text{smoke}} > 1.6 + 0.1\,H$

How do the heat limits work?

Convected heat at 60 °C keeps a person walking through hot air for the length of a normal escape. Radiant heat at 2.5 kW/m² lets skin cope for 30 min or more.

Above that flux, time to pain falls steeply, with $q$ in kW/m² and $t$ in minutes.

$t_{I_{\text{rad}}} = 4\,q^{-1.35}$

The formula carries about 25% uncertainty either way. A 10 kW/m² exposure gives about 0.18 min, or 11 s. So a hot smoke layer at 200 °C marks the end of tenability even for a person walking under it. Our post on radiant heat covers the flux side in more detail.

How does the toxic dose limit work?

Through the fractional effective dose, or FED, from ISO 13571. The model adds up the share of an incapacitating dose taken in each time step.

Carbon monoxide and hydrogen cyanide count as asphyxiants, while a carbon dioxide term multiplies the dose for faster breathing. FED = 1.0 means half of a population passes out. Designers therefore use 0.3, which affects about 11% of people, or 0.1 for the more sensitive, which affects about 1%. A parallel FEC term handles irritants such as HF, HCl and acrolein. ISO 13571 also treats visibility as a side issue. Critics flag that gap, because loss of sight often ends tenability first. Our post on smoke inhalation covers the gas chemistry.

Which model calculates ASET?

A two-zone model for simple rooms, and CFD for anything with a complex shape.

Zone models such as CFAST from NIST and B-RISK from BRANZ split each room into a hot upper layer and a cool lower one. They then solve mass and energy balances with a plume correlation. Zukoski’s plume, for instance, gives the mass flow into the hot layer as follows.

$\dot m_p = 0.21\,(Q^*)^{1/3}\,\rho\,\sqrt{g\,Z_{\text{int}}}\,Z_{\text{int}}^2$

Field models such as FDS solve the 3-D flow instead, at a cost of hours to days per ASET vs RSET run. Hand methods from Karlsson and Quintiere can also bound ASET with the smoke layer height alone.

Why does the design fire swing ASET so much?

Because the growth rate sets when the hot layer arrives, and the four standard growth rates span a factor of 64.

The t² model writes the heat release rate as a parabola in time.

$Q(t) = \alpha\,t^2$

NFPA 92 and NFPA 204 fix $\alpha$ at four values. Slow takes 0.00293 kW/s², medium 0.01172, fast 0.0469 and ultra-fast 0.1876. Those fires reach 1 MW at 600, 300, 150 and 75 s. Offices, shops and car parks usually take fast growth, while seated assembly takes medium and storage runs fast to ultra-fast. Sprinklers then cap the peak, often near 5 MW in retail.

Sensitivity work by Tosolini and colleagues found the ASET for a slow fire running several times the ASET for an ultra-fast fire in the same room. The “safety coefficient” of simple estimates ran from 1.2 for ultra-fast up to about 2 for slow. So pick the wrong letter and the answer moves by a factor of two before any occupant takes a step. Our post on t² fire growth covers where each rate comes from.

Where does ASET vs RSET break down?

At three points. The single number hides the spread, the comparison ignores space, and the two models never talk to each other.

Why criticise the single number?

Because both terms come as distributions, and one pass/fail line throws that information away.

Babrauskas, Fleming and Russell put the case bluntly in a paper titled “RSET/ASET, a flawed concept for fire safety assessment.” The same building and fire can yield very different ASET vs RSET results. The answer shifts with the scenario, the limit and the model, so the method invites steering. Other researchers therefore propose a map view, with ASET and RSET plotted point by point across a floor.

The probabilistic answer treats every ASET vs RSET input as a distribution. Monte Carlo runs with Latin hypercube sampling draw fire load, growth rate and pre-movement time thousands of times. They then report a probability of failure, the share of runs in which RSET exceeds ASET. That links the safety factor to reliability theory and to a target failure probability, the same logic as our post on F-N curves. Notarianni, Hurley, Frantzich, Ronchi and Kuligowski have all pushed the field this way.

What did the Station nightclub show?

That an uncoupled ASET vs RSET comparison can flatter a room by a wide margin.

On 20 February 2003 pyrotechnics lit foam on the stage walls of the Station nightclub in Rhode Island. In the end, 100 people died. NIST’s NCSTAR 2 report, by Grosshandler, Bryner and Madrzykowski, modelled the room in FDS against a 120 °C limit. It found that many occupants had less than 90 s after ignition to get out. Crowding at the single front door then pushed RSET far past that. So NIST recommended a 90 s cap on evacuation time for similar clubs, and told designers to assume the main door blocked.

A later coupled study at an IAFSS symposium ran the fire and the crowd together, so smoke slowed and felled agents as they moved. That model matched about 84 deaths only when the fire timeline slipped by 15 s. With the base timeline it produced about 180. Uncoupled runs came out “considerably over-optimistic,” which settles the question for any crowded room with a fast fire.

What did Grenfell show?

That ASET means nothing once the compartment fails. The stay-put strategy assumed each flat would hold a fire, yet the cladding broke that assumption.

On 14 June 2017, 72 people died in the tower. The Inquiry’s Phase 2 report, published on 4 September 2024, made 58 recommendations. The UK then answered with the Fire Safety Act 2021 and the Building Safety Act 2022. It also brought in BS 8629:2019 for evacuation alert systems, required in new residential buildings over 18 m in England from December 2022. On 2 September 2024 the government also announced person-centred risk assessments and residential PEEPs for disabled residents in high-rise buildings. That reversed a 2022 decision not to mandate them.

What do EV fires change in ASET vs RSET?

The governing limit. Lithium-ion battery fires release hydrogen fluoride, so toxic dose can end tenability before smoke blocks the view.

Do EVs burn hotter than petrol cars?

No. Peak heat release runs about level with a modern petrol car. The differences show up in duration, re-ignition and the gases.

RISE in Sweden measured battery-electric cars at 5.7 ± 1 MW against 6.2 ± 2.5 MW for combustion cars. Total heat came to 6.1 GJ against 5.9 GJ, so the authors called the gap “no significant difference.” Korean full-scale tests by Kang and colleagues put electric cars at 6.51–7.25 MW against 7.66 MW for petrol. The ultra-fast label therefore belongs to battery storage racks in runaway rather than to one burning car. Our post on EV fire frequency covers the incident data.

Why does hydrogen fluoride govern?

Because a burning pack gives off 20–200 mg of HF per Wh of capacity, and the safe exposure limits sit in the tens of ppm.

Larsson, Andersson, Blomqvist and Mellander at Chalmers and RISE measured those yields across cell types and charge states. They also found 15–22 mg/Wh of phosphoryl fluoride on top. A 100 kWh pack can thus release 2–20 kg of HF. The US EPA’s 30 min AEGL-2 limit for HF sits at 34 ppm, while the 10 min AEGL-3 sits at 170 ppm. HCN yields still remain poorly measured, a real data gap.

Kim and colleagues modelled an underground car park in FDS with packs of 24, 53 and 99.8 kWh. Asphyxiant gas tenability fell from 20 min 35 s for the small pack to 15 min 22 s for the large one. Once proximity to the fire enters the sum, the cut reaches 47%. HF peaked at 488 ppm in the worst ventilated zone and stayed above AEGL-3 for over 17 min.

A visibility-only study by Brzezińska and Bryant took the other view. Using PD 7974-6 limits and a 52 °C cap, they concluded that an EV fire in a garage would pose no threat to people inside. Both papers can hold true at once, though. They asked different tenability questions, and only one of them asked about HF. In both, ventilation drove the result.

What complements the inequality?

Redundancy, compartmentation, suppression and intervention. ASET > RSET stays necessary and never sufficient.

Passenger ships show a different way to write the same ASET vs RSET test. IMO circular MSC.1/Circ.1533 applies to new passenger ships keel-laid from 1 January 2020. It sets the pass line as a weighted sum of four times.

$1.25\,(R + T) + \tfrac{2}{3}\,(E + L) \le n$

Here $R$ and $T$ cover response and travel, while $E$ and $L$ cover embarkation and launching. The limit $n$ equals 60 min for ships with up to three main vertical zones, or 80 min above that. The rule folds a 1.25 factor into the human terms and assumes no smoke effect at all. So it fixes the margin at the input rather than at the output.

Model choice needs its own check in any ASET vs RSET study. NIST Technical Note 1822 by Ronchi, Kuligowski, Reneke, Peacock and Nilsson sets out verification and validation tests for evacuation models. The tests run from pre-evacuation distributions to counterflow and exit blockage. Pathfinder, buildingEXODUS, STEPS, MassMotion and FDS+Evac all sit in the field, along with open-source codes such as JuPedSim and Vadere. Yet only FDS+Evac and buildingEXODUS couple the smoke to the crowd out of the box.

How should you run an ASET vs RSET analysis?

Treat pre-movement as the main uncertainty, report ASET per limit, and state the margin. Seven habits cover most jobs.

On the inputs

1. Put your doubt on pre-movement, not walking speed. Run the analysis across all three PD 7974-6 management levels. If the margin changes sign anywhere in that range, the design lacks robustness.
2. Report ASET for each tenability limit and name the one that governs. Never hand over a single ASET.
3. Compute FED and FEC for battery or high-soot fuels. If toxic ASET falls below visibility ASET, toxicity governs. The ventilation then needs a second look.
4. State a safety factor and defend it. Use 1.5–2.0 by default, or 1.2 with drill data plus a sensitivity study.

On the method

5. Go probabilistic where the stakes run high. Run Monte Carlo on fire load, growth rate and pre-movement. Report a probability of failure as the output.
6. Couple fire and crowd models for nightclubs, assembly rooms and car parks. The Station result leaves no room for an uncoupled shortcut there.
7. Check what the inequality skips. Compartment integrity, sprinkler reliability, a blocked main exit, and a plan for people who cannot use the stairs.

Key takeaways

ASET vs RSET reduces every evacuation plan to one comparison. The time until a space fails a tenability limit must beat the time people need to leave, with a margin near a factor of 1.5 to 2.0. On the available side, the answer hangs on the design fire and the limit you pick. On the required side, it hangs on pre-movement time, which spans seconds to an hour.

So the inequality earns its place as a frame, never as a verdict. Report the spread, couple the models where crowds meet fast fires, and add HF to the dose sum where batteries burn. Then check the assumptions the sum rests on. Grenfell showed what happens when the compartment behind ASET gives way.

Cite this article

Dinh, D. C. (2026, September 4). ASET vs RSET: The Inequality Behind Every Evacuation Plan. PyroRisk. https://pyrorisk.net/blog/aset-vs-rset-the-inequality-behind-every-evacuation-plan/

Get citation →

---