Quantifying Human Performance During Passenger Ship Evacuation
Despite continual improvements in the shipping industry related to structural design, operational practices, onboard technology and regulations,
accidents still occur that result in sinking or capsize. When this happens to passenger ships, the results are often shocking and devastating with loss
of life, sometimes numbering in the thousands. For this reason, the International Maritime Organisation (IMO), which regulates the maritime industry,
provides guidelines for evacuation analyses for passenger ships1. The scope of work presented in this dissertation fills gaps in our understanding of
human performance during the evacuation process on passenger ships, particularly as it relates to passengers’ alarm response time, the influence of response
time distributions on total assembly time predicted by the evacuation model maritimeEXODUS, passenger movement onboard during the assembly process, and
methods for validation of evacuation models in general.
The research carried-out was experimental in nature and involved a total of 5582 passengers onboard three large passengers ships – a ferry without cabins,
a ferry with cabins and a cruise ship. All passengers had paid for their voyage and, prior to boarding, had no knowledge that an experiment was being conducted.
The experiment was carried-out as a typical assembly exercise, which started with sounding of an alarm and ended when all passengers had been assembled.
Passenger response to the alarm was recorded using digital video cameras and routes to assembly stations and associated times were captured using a novel
infrared light detection tracking system. The dataset collected represents the most comprehensive collected to date for passenger response and movement during
assembly trials at sea.
Analysis of the data has provided important insights into the nature of response time distributions for passengers on ferries and cruise ships. It was found
that response time distributions generally took a lognormal shape, which is consistent with response time distributions measured in the built environment.
Response time distributions generated from repeat trials on the same ship were statistically similar and could be combined to produce a single distribution
each ship - a powerful result suggesting that if the response times and demographics of a sufficiently large number of people are characterised for a given
type of structure, an assembly exercise repeated under similar notification conditions should result in a similar distribution. Another key finding was that
response time in cabin areas was not similar to that in public areas on the same ship. In addition, response times for passengers in public areas on ferries
was found to be statistically similar, while public space results for the cruise ship were different. This suggests that different response time distributions
should be used for different ship types.
Passenger movement results have enabled the development of two unique datasets for use in validating ship evacuation models – one validation dataset which is
relevant for ferries without cabins and the other for cruise ships. The validation method developed enables a clear, yet objective means by which ship
evacuation models can be assessed using the experimental data collected. It is felt that the suggested validation protocol and acceptance criteria developed
form a reliable basis for validating ship evacuation simulation tools.
This research has resulted in the submission of two information papers to the IMO suggesting credible response time distributions relevant for different ships
and different areas onboard, as well as a detailed method for conducting validation of ship evacuation models. The recommendations being made from this work
are significant since, if accepted by the IMO for inclusion in the regulations, will influence the design and construction of all new passenger ships worldwide.