Modelling Human Behaviour and Evacuation in Fires
Building geometry can either be described by a coarse network (each node ~ 1 room, corridor or stair section) or a fine network (each node ~ 0.5m "pixel"). Fine networks are necessary to calculate complex flow rates from first principles, whereas coarse networks would require empirical equations. The more "homogeneous" the population behaviour (ie all flows heading to exits, no contra-flows & limited merging), the less distinct the difference between coarse and fine networks.
People's behaviours may depend on many factors, for example a defined role within the building population, the influence of training, familiarity with the building, age, gender, and other factors such as disabilities which may restrict the possible activities that may be attempted.
People's behaviour is firstly of significance in responding to the initial fire cues. The information content of the cues is of greatest importance in determining people's reaction to them. An uninformative cue is likely to be ignored, until reinforced by a stronger cue. The person's current activity at the time of the first cue will also influence how likely they are to change this activity and respond to the cue. A person with a defined role or training would behave more positively. If the initial cue is ambiguous, the most common positive response is to seek more information. People in groups will decide what to do depending on their perceptions of others' reaction; without a dominant personality providing an early lead, people may take much longer to respond than if they were alone.
Once the existence of a fire has been recognised, people may undertake a wide range of possible behaviours. These depend on role and training, and also people's perception of the developing fire situation (which may of course differ considerably from the truth). People may eg. continue working, go to collect belongings, attempt to fight the fire, seek to warn others (role/training may dictate a search of the building; on the other hand, people may simply warn others they encounter), rejoin family groups, rescue / assist others, seek refuge or escape. As the perceived situation worsens, the range of viable options reduces.
The evacuation time is conventionally split into two components, travel time and ?pre-movement? time. The pre-movement time incorporates the time to recognise that an alarm has been given, that some action needs to be taken, and a "response time" for all other activities performed prior to evacuation. "Pre-movement time" is thus a misnomer, since it may include activities involving movement prior to evacuation.
A modelling shortcut involves representing all pre-movement activities by a single time delay before a person moves; different people will move at different times, however the probability distribution for this delay is hard to quantify without considering the activity explicitly. Few evacuation models predict the pre-movement phase (most use an empirical distribution of delay times).
Pre-movement time can be a significant portion of the total evacuation time, so should not be neglected.
A key to understanding human behaviour is the observation that people rarely panic in fires. They choose from a range of behaviour options depending on their assessment of the prevailing conditions. However it must be recognised that in some cases, their assessment may be incorrect, based on inadequate or ambiguous information. With the benefit of hindsight, it is easy to dismiss the person's actions as "irrational" or "panic".
People may move (or continue to move) through smoke, or choose an alternative action / route leading away from the smoke. Generally it is assumed that when the tenability limits are exceeded, further movement through smoke ceases. Some models allow people to choose a different route if smoke is encountered. The fire environment will usually be calculated a-priori, although in the CRISP model all sub-models run simultaneously, which (among other things) allows the smoke movement sub-model to respond when people open and close doors.
A design approach that treats any encounter with smoke as a failure of the system may be too conservative. On the other hand, allowing people perfect knowledge to anticipate and avoid the movement of smoke will be too optimistic.
Such research as has been performed on this topic has concentrated on when people are escaping. The probabilities of people moving through smoke (and / or associated tenability limits) for different actions such as investigation, first-aid fire fighting, warning or rescuing others, etc, have not been determined.
The second issue is the speed of movement; as visibility decreases, people will move more slowly. Irritant smoke has a greater effect in this regard than non-irritant smoke.
The exit choices made by different people mean that all doors are not used by the optimum number of people to minimise the evacuation time. In other words, the number of people using each door is not necessarily proportional to the door?s width.
People will not necessarily choose their nearest exit either. The factors which influence exit choice are complex, and quantitative data is scarce. In qualitative terms, people tend to stick to familiar routes. A distinction can therefore be made between buildings where the occupants are familiar with the geometry (eg. office workers), and buildings where they are not (eg. public assembly). Even in the former types of buildings, it is good design to ensure the fire exits are part of the normal circulation routes. A clearly-signed ?Fire Exit? may otherwise be ignored in preference to the familiar route. However, building visitors given directions by building staff are more likely to choose unfamiliar exits.
The location of the fire may make (at least) one exit inaccessible. Alternative fire locations may need to be tried to identify the ?worst case?. It is customary to consider that the widest door is unusable ? but implicit in this approach is that this door would normally be used by the most people, which may not be correct.
In most evacuation simulation models, the length of the shortest route to the ?outside? (or place of refuge) is calculated for every possible location a person may occupy. The direction each person moves in is then chosen to continually minimise the distance they still have to go. Exit choice is built in to this modelling mechanism by adding additional distance for some possible routes from each location, but not for other routes. The effect is to prevent unfamiliar exits being chosen until a person is much closer to an unfamiliar exit than a familiar one. Since quantitative data is scarce, it is likely that the additional distances are chosen to determine the modeller?s ?best guess? of what the probabilities of exit choice are. Sensitivity studies should be performed.
The movement speed of people, in crowded situations as well as unimpeded, is well understood for most common situations. Empirical relationships have been derived for walking speed as a function of crowd density (people per square metre), flow rates in corridors or up / down stairs, and the rate at which queues of people pass through doorways or other similar constrictions as a function of their width. These relations enable the movement time portion of the total evacuation time to be estimated with a fair degree of accuracy. For further details, the SFPE handbook is a good starting point.
It needs to be remembered that travel time is often not the dominant component of the overall time required to escape.
There are many types of disability, physical, sensory or mental, and different degrees of severity. There are also many adverse impacts that disabilities may impose on a person's ability to be aware of, react to and escape from a developing fire situation. A number of models do allow a "disabled sub-population" to be generated with slower movement speeds than normal. At present no egress model takes account of other effects of disabilities explicitly.
As people move / remain within the building, they may be exposed to smoke and the toxic products of combustion. The exposure is usually quantified in terms of a fractional effective dose (FED) which depends on the concentration of particular toxins within the "cocktail" of fire gases, and the duration of exposure. When the FED for a particular person reaches unity, they are overcome by the smoke. There is considerable variation in the FED uptake rates for different individuals.
For some combustion products (eg irritants) the effect is almost instantaneous rather than cumulative, so just the concentration is the key parameter. Tenability levels for different rooms may be expressed in a number of ways, either the time for a person's FED to reach unity (if they remain in the room for the duration), or the value of the concentration of a given product that is sufficient to effectively prevent escape.
In the case of optical density affecting loss of visibility, the limit partly depends on occupants? familiarity with the building, ie represent willingness to move through smoke, rather than inability to see. Better visibility is required for larger rooms.
If the smoke layer is stratified above a clear layer, the layer interface height should also be one of the tenability criteria. Some safety margin should be allowed, ie the limit should not be set at "head height" or below. Layer height per se is not sufficient to cause loss of tenability, one of the other limits relating to smoke properties must also be exceeded.
Although there are no standard values for tenability limits for particular combustion products, the following (from the SFPE Handbook) may be taken as rules of thumb for the common parameters affecting building occupants:
Carbon Monoxide
6000 ppm
Carbon Dioxide
7%
Oxygen Deficiency
13%
Hydrogen Cyanide
150 ppm
Irritants 1ppm (acreolin)
75ppm HCl
Smoke Optical Density
0.1 m-1
Temperature
80°C
Radiant Heat
Flux 2.5 kW.m-2
Clear Layer Height
1m above head (for small rooms / corridors)
The total time for evacuation is termed the Required Safe Egress Time (RSET). This value is often compared with the Available Safe Egress Time (ASET) to determine whether or not a particular scenario poses an unacceptable hazard.
The Available Safe Egress Time (ASET) is the time between the start of the fire and the onset of conditions that create a hazard to the occupants. The Required Safe Egress Time (RSET) is the time required after the fire has started, for the last person to have reached a place of safety. It includes the time for the fire to be detected, the alarm raised, people to recognise the alarm for what it is, respond to the alarm, and finally evacuate. Clearly ASET and RSET both depend on the fire scenario.
The definition of ASET is in terms of time since ignition, but it may be more helpful to calculate the time difference between the first warning being given, and the onset of untenable conditions. Whatever definition is followed, it must be ensured that ASET and RSET are expressed on a common time basis.
The concept of ASET was originally intended for an effectively closed compartment containing a fire. The concept has been broadened to include rooms not containing a fire, in these cases calculations must account for the rate of smoke flow between rooms as well as the filling time of each room. It is important to be clear which portions of the building are being considered.
RSET and ASET are not single deterministic values, but will have probability distributions associated with them. Analysis must consider the safety factor, not just whether the mean value of ASET exceeds that of RSET.
The CRISP model, which has been developed by BRE staff, is a Monte-Carlo simulation for fire risk assessment. The basis of the risk assessment is the fractional effective dose (FED) acquired by the building occupants. The risk is expressed simply in terms of the fraction of people originally present who end up ?dead?, averaged over a sufficiently large Monte-Carlo sample. CRISP incorporates a detailed behaviour model, rather than something simpler which would run faster. Why? The justification is that accurate FED estimates require accurate exposure times. Therefore the behaviour model needs to predict where people will go, and how long they will spend in different areas (rooms) of the building.
BRE fully understand the impact on physical means of escape provisions, which is one of the greatest constraints in building design and construction.
References
? SFPE Handbook of Fire Protection Engineering, Third Edition, Society of Fire Protection Engineers, 2002. IBSN 087765 451 4. Chapters 2-4, 2-6, 3-10, 3-12 and 3-14 are most relevant.
? BS7974 Application of fire safety engineering principles to the design of buildings ? Part 0: Guide to design framework and fire safety engineering procedures.
? CIBSE guide E, chapter 4
? Canter, DA, ' Fires and Human Behaviour: Psychological Aspects of the Experience of Fires', David Fulton Publishers, 1990, ISBN: 1853461393
? Shields TJ, Fire & Disabled People in Buildings, BRE report BR 231 (1993), ISBN 851255469
? Shields TJ, Dunlop KE and Silcock GW, Escape of Disabled People from Fire: a Measurement and Classification of Capability for Assessing Escape Risk, BRE Report BR 301 (1996), ISBN 186081675
? S.Gwynne, ER.Galea, M.Owen, PJ.Lawrence, L.Filippidis, "A review of the methodologies used in the computer simulation of evacuation from the built environment", 1st Human Behaviour in Fire conference, 1998, p.681
? Fraser-Mitchell,JN, "Modelling Human Behaviour within the Fire Risk Assessment Tool CRISP", Fire & Materials 23, 349-355 (1999)
Full references to computer models have not been given, since new models are continually being developed.
Check the website http://www.firemodelsurvey.com/ for further details.
© Copyright 2011 Means Of Escape Terms & Conditions Privacy Policy
Copyright © 2011 Think Agency - Website Designers Kent