Evacuation alarm system

An evacuation alarm system warns building occupants of a fire-related emergency so that they can begin the intended evacuation response. It communicates a cue through alarm signals or spoken information, often supported by staff instructions. Its effectiveness depends on people receiving, understanding and acting on that warning.

Detection and evacuation are separate stages. A detector may identify a fire before occupants know about it, and people may take further time to interpret the warning. The notification period, preparation period and movement period therefore have different meanings in an evacuation calculation. Combining them into one unexplained delay can hide the assumptions that determine the result.

The appropriate warning depends on the occupants, the building and the planned evacuation strategy. Familiarity, sleep, assistance needs, staff behaviour and previous false alarms can affect the response. Engineering assessments connect these human factors to the warning equipment and identify the evidence behind the selected times.

From detection to a useful warning

An evacuation warning begins with information that a fire may be present; an automatic detector can respond to smoke, heat or another measurable fire characteristic. A Manual call point introduces a human decision into this chain. Someone must first perceive the fire, recognise its significance and operate the alarm.

The Fire alarm system processes the initiating signal and activates the required outputs, and its warning function is one part of a wider system that may also operate other protective equipment. Schifiliti, Meacham and Custer distinguish detection, processing and signalling when analysing this sequence. Separating these functions makes their individual delays and purposes easier to identify.

A warning is a cue: information that may prompt a person to interpret the situation as an emergency. The International Fire Engineering Guidelines (IFEG, 2005) distinguish the occurrence of a cue from its recognition. Seeing flames, smelling smoke or receiving a warning from another person can also provide cues. These alternatives depend on the fire scenario and the people who can perceive them.

The analysis therefore identifies which cue is expected to reach each occupant group; a detector signal at a control panel is not automatically a warning received by every occupant. Equally, a person in the fire room may recognise smoke before the building warning operates. The selected cue and its timing need to be consistent throughout the evacuation calculation.

An evacuation alarm works through a complete response chain. Based on Australian Building Codes Board et al. (2005), p. 1.8-7 and Proulx (2002), pp. 3-350–3-351.

Alarm tones and spoken information

An alarm tone attracts attention but may provide little information about the event. Occupants may seek confirmation, observe other people or wait for instructions before moving. Proulx describes warning effectiveness in terms of the information available and the response it produces. The assessment extends beyond whether a signalling device has operated.

A Voice alarm system can communicate the nature and location of an incident and the action expected from occupants. Proulx reports that informative live messages can prompt a faster response than an unexplained tone; the same chapter distinguishes live instructions from general prerecorded messages. Their effectiveness depends on the message content and the conditions in which people receive it.

Audibility means that occupants can hear the warning. Intelligibility means that they can understand spoken information. Proulx identifies both as matters to assess when evaluating a voice communication system. An audible message offers limited benefit if the recipient cannot understand its meaning.

The information may also need to differ between parts of a building. People near the fire and people on a remote floor do not necessarily face the same conditions. The SFPE Handbook describes the need for information about fire location and extent in partial evacuation or stay-in-place arrangements. Such strategies connect warning design to the planned protective action.

Live voice and staff instructions can prompt faster response. Based on Proulx (2002), pp. 3-350–3-351.

Why occupants do not all move immediately

Pre-movement time includes the period between receiving a warning and starting evacuation travel. During this period, people may interpret the warning, seek information or prepare to leave; it is distinct from the time spent walking towards an exit. Leaving it out can produce a misleadingly short estimate of the required escape time.

Familiarity affects how readily people understand the building and its evacuation arrangements. Office workers may know the alarm and exits, whereas visitors may depend on staff. Being awake and alert also differs from being asleep when the warning occurs. Proulx treats these as characteristics of the occupants and their situation, rather than a single universal response time.

People with family members or other close companions may try to gather them before leaving. Those committed to an activity may be reluctant to interrupt it without clearer information. Proulx gives examples involving restaurants, travel and entertainment. These activities affect the interpretation of the warning and the preparation needed before movement starts.

The surroundings influence this response as well: an open space can allow occupants to observe other people and notice smoke or staff instructions. Separate rooms may restrict access to that information. The warning system and the building layout therefore affect the same human decision process.

Separating the time components

Required safe egress time is the calculated time needed for the specified occupants to leave the assessed location. In New Zealand, C/VM2, second edition (2025), separates detection, notification, pre-travel activity and movement. Notification begins after detection and ends when occupants are notified. Pre-travel activity then precedes the start of evacuation movement.

For its simple hydraulic calculation, C/VM2 uses the greater of travel time and flow time for evenly distributed occupants leaving the room of origin. Flow time represents congestion controlled by the capacity of the escape route; the expression below states that particular calculation, rather than a universal formula for every evacuation strategy. Additional provisions apply to delayed evacuation arrangements.

$$\mathrm{R}\mathrm{S}\mathrm{E}\mathrm{T}=t_d+t_n+t_{\mathrm{p}\mathrm{r}\mathrm{e}}+\max (t_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{v}},t_{\mathrm{f}\mathrm{l}\mathrm{o}\mathrm{w}})$$

Symbol	Description	Unit
$\mathrm{R}\mathrm{S}\mathrm{E}\mathrm{T}$	Required safe egress time for the assessed location	s
$t_d$	Detection time from ignition	s
$t_n$	Time from detection to occupant notification	s
$t_{\mathrm{p}\mathrm{r}\mathrm{e}}$	Pre-travel activity time following notification	s
$t_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{v}}$	Travel time towards safety	s
$t_{\mathrm{f}\mathrm{l}\mathrm{o}\mathrm{w}}$	Flow time associated with congestion	s

For standard evacuation strategies, this New Zealand method assigns a notification time of 30 s. Non-standard arrangements must account for any extended notification time, and the document gives management investigation of a single activation as an example. The 30 s value is a modelling provision within this method, not a measured delay for every alarm installation.

Notification time must match the evacuation strategy. Based on Ministry of Business, Innovation and Employment (2025), p. 29.

Worked example of warning and preparation

C/VM2 assigns different pre-travel activity times according to occupant characteristics, location and warning type. Consider awake, alert occupants unfamiliar with the building, such as visitors in retail or exhibition spaces. For those remote from the enclosure of fire origin, the table assigns 120 s following a standard alarm signal. A voice alarm signal has an assigned pre-travel time of 60 s for the same group.

Using the method's standard notification allowance of 30 s, the notification and pre-travel subtotal becomes 150 s for the standard signal. For the voice signal, the corresponding subtotal is 90 s; the arithmetic uses the same group, location and notification assumption in both cases. Only the table's warning-dependent pre-travel value changes.

$$(t_n+t_{\mathrm{p}\mathrm{r}\mathrm{e}}{)}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}}=30+120=150\mathrm{s}$$ $$(t_n+t_{\mathrm{p}\mathrm{r}\mathrm{e}}{)}_{\mathrm{v}\mathrm{o}\mathrm{i}\mathrm{c}\mathrm{e}}=30+60=90\mathrm{s}$$

The difference is 60 s within this calculation. Neither subtotal includes detection time or the movement component, so neither is a complete evacuation time. The result also does not establish that every real voice message saves exactly 60 s. It demonstrates the consequences of selecting the prescribed inputs for this particular New Zealand verification method.

Occupants in the enclosure of origin have different entries in the same table. For the awake, unfamiliar group, these are 60 s for a standard signal and 30 s for a voice signal. This distinction reflects the method's separate treatment of location. Applying the remote-room value to everyone would change the prescribed calculation.

Matching the warning to the evacuation strategy

Evacuation may involve all occupants leaving or different groups receiving different instructions. The SFPE Handbook identifies partial evacuation and staying in place as strategies requiring information about the fire. Warning arrangements must support the intended response of each group. A common tone alone does not communicate the same detail as a location-specific instruction.

Phased evacuation makes the timing and recipient of each warning relevant to the analysis. People instructed to remain temporarily cannot be treated as if they were already travelling towards an exit. The calculation needs to represent the adopted sequence and any associated delay. IFEG requires the cue and response analysis to be applied to the relevant enclosure or group of enclosures.

In New Zealand's C/VM2 (2025), delayed evacuation includes situations where occupants remain in place or move to safety inside the building. The spaces concerned must satisfy that document's place-of-safety provisions, and its commentary also identifies smoke detection and air handling as matters requiring particular attention in these environments. The warning assumption cannot be separated from the conditions in the place where occupants remain.

Occupants needing assistance introduce another dependence on staff response. C/VM2 distinguishes these populations from independently mobile, awake occupants when assigning pre-travel activity times. Proulx likewise identifies perceptual, physical and intellectual limitations as factors in starting evacuation. A suitable assessment describes the occupant group before selecting its timing assumptions.

Staff behaviour, training and alarm credibility

Staff instructions can translate a warning into an understandable action. In public buildings, visitors may expect someone responsible for the premises to explain what is happening. Proulx describes responsibility and staff behaviour as influences on the time taken to start moving. The presence of an alarm therefore does not remove the human role in the warning process.

Training needs to relate to the actual building and its evacuation procedure. Proulx notes that familiarity with one building's response arrangements does not necessarily transfer to another location. Public occupants may have received no training for the premises they are visiting. In those circumstances, the competence and availability of staff affect how the warning is interpreted.

Repeated alarms without a fire can weaken the credibility of later warnings. Proulx discusses false alarms, tests and prank alarms as experiences that may encourage occupants to wait for confirmation. The chapter gives three or more false alarms in six months as an illustrative high frequency. This is a behavioural observation in a 2002 source, not a universal regulatory threshold.

The response to a tone and the response to an informative instruction may therefore differ substantially. Precise messages can identify the incident and the action expected, while staff can reinforce the warning. Proulx reports these effects through studies of drills and incidents. Those observations support consideration of communication quality without supplying one guaranteed delay for every building.

Evidence and uncertainty in an assessment

IFEG recognises that validated methods for predicting pre-movement behaviour are limited, and it identifies documented literature, field studies and simulated evacuations as possible evidence. Where engineering judgement is used, its basis needs to be explained. A numerical input is more useful when its origin and applicability to the occupant group are clear.

Proulx reports an unannounced drill in a large retail store where staff quickly prompted customers to leave. Customers started moving after a mean of 25 s, with a reported maximum of 55 s. Staff closed cash counters and assisted people in changing rooms. These observations describe that exercise and its staff response, not an intrinsic property of the alarm signal.

The same chapter reports drills in three Canadian government office buildings involving more than a thousand recorded occupants. The mean time to start was 50 s despite occupant training and familiarity with the evacuation procedure. Some people completed tasks or gathered belongings before leaving their desks. Local wardens also prompted movement, showing why preparation and staff involvement remain relevant even among familiar occupants.

These examples illustrate the range of evidence that can inform an assessment; their interpretation requires attention to occupancy, training, observation method and the warning actually received. Transferring the reported mean alone would discard those conditions. The engineering task is to judge whether the underlying situation resembles the proposed scenario.

Drill observations and accounts of actual incidents describe different conditions. Proulx discusses uncertainty in people's retrospective estimates of the time taken to leave. The chapter also notes that initial fire cues can be ambiguous. A real fire cannot automatically be assumed to produce a shorter response than an exercise.

Different groups may receive different cues at different times. Some occupants may hear a message, while others notice smoke or receive instructions from staff. IFEG requires the selected cue to be justified when more than one is possible. This avoids combining an early detection assumption with an unrelated, more favourable human response assumption.

Sensitivity studies examine how the result changes when uncertain inputs vary. IFEG also distinguishes uncertainty studies and checks for failures affecting apparently redundant systems. Applied to warning arrangements, these checks expose dependence on a particular signal or human response. Their purpose is to understand the assessment's limits and the conditions under which the proposed sequence remains credible.

See also

• Evacuation strategy
• Protective action decision model
• Required safe egress time

References

• Australian Building Codes Board, National Research Council of Canada, International Code Council, & Department of Building and Housing New Zealand. (2005). International fire engineering guidelines. Australian Building Codes Board.
• Ministry of Business, Innovation and Employment. (2025). C protection from fire: Verification method C/VM2 (2nd ed.).
• Proulx, G. (2002). Movement of people: The evacuation timing. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., pp. 3-342–3-366). National Fire Protection Association.
• Schifiliti, R. P., Meacham, B. J., & Custer, R. L. P. (2002). Design of detection systems. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., pp. 4-1–4-43). National Fire Protection Association.