Cabin design and procedures for safe emergency evacuation, may be changed by computer modeling.

Martin Hindley/LONDON

AIRCRAFT EMERGENCY evacuations are designed as far as possible to work no matter what the nature of the emergency, but passenger behaviour is inherently difficult to define and predict. Present rules governing cabin emergency evacuation do not pretend to compensate for human error, panic or simple bad luck.

Future cabin safety procedures may now be improved, however, by a computer evacuation simulation model under development at the University of Greenwich, London, and funded by the UK Civil Aviation Authority.

Although there is evidence that evacuations are becoming safer and more effective, there is concern about the way predictable behaviour in these events, can rapidly give way to disorder in the cabin. As the scale and level of competition between escaping passengers escalates, so does the risk of injury or death.

According to cabin-safety researcher Professor Helen Muir of Cranfield University, UK, the passenger fatality rate - when there is an accident - has not changed appreciably over the last 20 years despite the "dramatic" reduction in the number of accidents themselves (Flight International, 24-30 May 1995). In her report Cabin Staff Behaviour in Emergency Evacuations, Muir says that the forces, which make accidents non-survivable are usually of such magnitude that changes to cabin-safety regulations would probably not make any difference to survival rates. It is those factors that improve the chances of living through a technically survivable accident that should be addressed by any cabin-safety research programme, Muir concludes.

To gain a better insight into what happens during an emergency, researchers at Greenwich are developing a computer model, capable of simulating passenger behaviour in an aircraft evacuation. The project is part of a wider cabin-safety research programme run by the CAA and the US Federal Aviation Administration, which ranges in scope from investigating passenger dynamics to studies in cabin crew assertiveness and behaviour.

The system could make a vital contribution to the planning of cabin safety and help save lives, says Edwin Galea, who is leading the Greenwich team. Working in the University's Fire Safety Engineering Group, part of the Centre for Numerical Modeling and Process Analysis, Galea has studied how humans respond to the effects of fire in confined spaces. His research has focused mainly on evacuation from buildings. However, the 1985 Manchester air disaster in which 55 passengers died on board a burning Boeing 737-200 made him consider an aircraft evacuation study.

"For reasons of cost and safety, real-life evacuation simulations are rarely carried out on aircraft. A full-scale trial can cost several million dollars, while the risk of personal injury has led to government pressure to reduce the number of tests, particularly in the USA," he says.


Galea is strongly critical of the existing evacuation test for aircraft certification. Aircraft manufacturers are required to demonstrate that their cabin layouts allow full passenger and crew evacuation in 90s or less.

In the test, volunteers representing an "average" cross-section of the travelling public are required to leave the aircraft via half the total number of exits normally available. Old people, however, may not take part in the test because of their vulnerability to injury. According to Galea, while the test is good for judging relative evacuation performance between aircraft types, it provides little useful information regarding the suitability of the cabin layout and design in the event of a real emergency, particularly one involving fire.

The 90s test lacks realism, he explains, because volunteers cannot be subjected to the hazard, trauma and panic of real damage. "The test is expensive, but not demonstrative," he says. "An average 6% of participants receive some form of injury in the simulation."

In the 1985 Manchester 737 disaster, the last passenger to escape alive emerged 5min 30s after the crew aborted take-off following an engine failure. Of the 137 people on board, 55 died, most of them from inhaling toxic fumes.

Galea and his team have produced a prototype mathematical egress model called Exodus, which is designed to track individuals as they make their way through the cabin towards the exit doors. It charts their progress on a computer screen, compiling statistics on escape rates and deaths caused by hazards on board.

The model is intended to help resolve "what if" questions, such as: if a problem develops with an aircraft door during a stop-over, how many passengers could be carried on the next leg "safely" with that door inoperable? Is there a preferred seating arrangement, which should be enforced in situations like these?

Galea claims, that the model could also be used by accident investigators, to throw light on what determines fatalities in an evacuation related aircraft accident. The evacuation performance of a proposed cabin layout can also be evaluated and optimised at the design stage, eliminating costly changes later on, says Galea.

The Exodus software is expert system based, with the motion of each individual determined by a set of parameters, called attributes. Using the latest generation of PC-based processors, the simulation can be run in around 90s - almost real-time.


Galea explains, that the governing principles of the software lie in five basic interacting components, or submodels: movement, behaviour, passenger, hazard and toxicity:

the movement model governs the motion of individual passengers from their existing positions in the cabin to the most suitable neighbouring locations as the evacuation unfolds;

the behaviour model is used to determine a passenger's response to his or her immediate situation, which is determined by the person's characteristics. It then passes decisions to the movement model;

personal characteristics, including age, sex and agility - some of which remain constant, some of which change in response to inputs from other submodels - are described by the "passenger" model, which recognises an individual as a collection of "attributes";

the hazard model controls the aircraft's atmospheric and physical environment, distributing fire hazards such as heat and toxic products throughout the cabin. This model also governs the operation or non-operation of exits;

the effects of toxic products on an individual are determined by the fifth sub-model: toxicity, which reports information back to the movement model via "behaviour".

In addition to the sub-models, Exodus has built-in tools which allow interactive construction of cabin geometry, passenger profiles and simulation interrogation. The progress of the simulated evacuation can be watched on a colour display.

The system, says Galea, depicts the cabin as a two-dimensional plan with seats represented as nodes. Each node is assigned with "attributes" peculiar to each passenger, which are updated by the hazard submodel during the evacuation. The nodes are connected by "traversable arcs", according to the particular layout of the cabin. Impassable obstacles such as walls, bulkheads and internal compartments, are unconnected. Each node is weighted with two additional attributes: "potential" and "obstacle".

The "potential" attribute, which can be modified by the hazard model, represents the distance of the node from the nearest exit, while the "obstacle" attribute is a measure of the difficulty in passing over that node, ranging in value from one to five. The initial obstacle attribute for all aisle nodes is set to One, whereas for motion over seat backs, the node value is Five, reflecting its difficulty. When a passenger is overcome by fumes and collapses the obstacle value of that person's node increases by one.


In the simulated evacuation, individuals assume the "potential" of the node they occupy. The movement submodel then attempts to move them in a manner that minimises their current potential. The movement of passengers, is restricted by their personal physical and behaviourial attributes and the fact that only one "conscious" passenger, may occupy each node.

In the Exodus prototype, the escape strategy adopted by all passengers is to leave via the nearest serviceable or assigned exit, even though, Galea points out, " may prove better to travel a greater distance to another exit to avoid crowds and queues". Passenger's phobias also play a part in the route-finding process, decisions depend on factors such as crowd density, distance, fire hazards and prior knowledge.

This approach should offer a reasonable approximation of the minimum evacuation times in hazardous conditions. To ensure that passengers will head for their nearest exit, node potentials for an exit are increased the further away a node is from that exit.

Increasing or decreasing a door's "exit potential" can ensure passengers move towards "assigned" rather than nearest exits. Tradeoffs between waiting to get into a queue and moving off towards an exit further away are handled using passengers' "patience" attributes.

If they are trapped between seat rows but their agility exceeds the obstacle value of the seats, they will be able to jump over.

In a simulated Exodus evacuation of a wide-bodied aircraft carrying 112 passengers with a toxic gas environment, there were 41 fatalities, despite an average evacuation time of just 17s. The majority of the fatalities occurred in the high temperature/toxicity zone. Five passengers were overcome in the immediate vicinity of the exits, while the 39th casualty, a male aged 36-57, traveled over 9.2m and survived for 16.3sec before passing out in the right wing exit.


Another source of information relating to historical trends in aircraft evacuations is the CAA-supported Accident Statistics Knowledge (ASK) project. Started by Galea as an offshoot to Exodus, the ASK project aims to extract evacuation statistics from the NTSB's accident report database, and use it to validate results.

Galea is pressing Airbus Industrie, McDonnell Douglas and Boeing for details of their evacuation certification data for comparison with the Exodus model. Future plans for the software include using virtual reality.

Source: Flight International