You are on page 1of 13

EVACUATION ASSESSMENT IN TRANSPORT PLANNING MODELS

Ewa Prawdzik
Paul Clifford
Sonja Nightingale
Mott MacDonald Group

1. INTRODUCTION

Transport planning models operate in a wide range of macro and micro levels.
At the micro level pedestrian simulation models are an important tool in the
capacity assessment of transportation hubs and urban spaces. These models
are used to define the movement and interaction of people between one
another and the surrounding environment. The analysis helps architects,
transport planners and project teams to optimise the design and operation of a
building.

As the utilisation of transportation systems increases, globally, greater focus is


being applied to safety and this includes evacuation assessment. A pedestrian
simulation model is a tool that assists in the planning and design of an assets’
emergency route and capacity provision.

In recent years, significant resources have been applied to research,


operational standards and fire prevention, but the prediction of outcome is still
developing. Safety control measures for evacuation in underground, railway
stations or during sport and cultural events are constantly under review and
the benefit of assessing how the design and crowd control measures could
assist evacuation, is extremely important to support the effectiveness of the
evacuation systems for transport interchanges, worldwide.

This paper focuses on the methods available to assess the performance of a


design under emergency conditions, through a number of studies. First, it
presents a comparison of the results from a real-life office building evacuation
experiment with the outputs derived from a dynamic pedestrian simulation
tool. Secondly, it considers the relative performance of the static calculation
method in relation to a dynamic pedestrian simulation model. Finally, the
study presents a series of transport studies where dynamic crowd modelling
simulations have been used to support the findings from the static
assessment. Through these studies, this paper highlights the benefits and
drawbacks for each approach, and demonstrates how different methods can
complement each other.

© AET 2013 and contributors


1
2. EVACUATION ASSESSMENT METHODOLOGIES

Human behaviour is difficult to predict and individuals’ reactions during an


emergency even more so. There has been a consensus that people tend to
choose to evacuate by the same route they have entered a building, station or
other transport interchange. During times of confusion they may be unaware
of the nearest emergency exit, or be looking for their friends or family. Panic
may also occur, causing safer routes to be missed, as people follow others.
Despite these behavioural factors, evacuation during an emergency situation
is an action which can be greatly assisted by the provision of appropriately
located and sized means of escape, usage of relevant building materials and
construction methods, smoke control systems and competent personnel to
assist the evacuation process.

2.1. Evacuation assessment approaches

Evacuation time calculation methods have been widely used in the design of
means of escape. These can be divided into two main categories:

• Static calculations - using methodologies widely applied in the transport


sector, with examples covered in the National Fire Protection Association
(NFPA) 130 Standard for Fixed Guideway Transit and Passenger Rails
System [1], the London Underground Station Planning Standard [2] and the
Guide to Safety at Sports Grounds [3], known as the “Green Guide”
• Dynamic evacuation models, most frequently using agent-based approaches
In recent years, the latter has been used more frequently, either to support the
findings of static calculations or to replace these with more detailed results.

2.2. Evacuation dynamic modelling approach

Dynamic passenger simulation models have been widely used to assist in the
development of safe designs and space-proofing of a building or place during
normal operation. By undertaking this type of analysis, designers seek to
ensure that the levels of congestion and journey times are acceptable.

With respect to the simulation under emergency conditions, engineers


commonly assess clearance times to meet the required minimum evacuation
times, so as to minimise the risk of injuries.

In an evacuation computer simulation, virtual people are instructed to make


their way to the nearest available exit they are aware of, with their movement
modified according to a range of individual behavioural principles. The models

© AET 2013 and contributors


2
calculate evacuation times, exit usage and other criteria essential for fire-
engineering design.

Some key assumptions used within dynamic modelling include:

• Walking speed
• Flow rates through horizontal and vertical circulation elements, such as
passageways, corridors, doors, ramps and stairs
• Usage of available exit routes
• Familiarity with the environment
With respect to the above key assumptions, there are two principal
approaches that can be undertaken using the dynamic modelling
methodology, including:

• Fixed approach: key assumptions are set in the model to represent the
requirements of a particular industry standard. Examples of fixed assumptions
are outlined in Table 1.

• Flexible approach: a number of sensitivity tests can be carried out to assess


the impact of additional parameters, such as:

o Varied walking speeds: Fruin [4] distribution, locally observed speed


distribution, different speed for up and down stairs
o Taking into account the impact of smoke and therefore reduction of
visibility leading to decreased walking speeds
o Testing the impact of different capacities for each circulation element
o Variations in the usage of available exit routes
Table 1. Key parameters for evacuation assessments, as defined in industry
standards

Assumption NFPA 130 [1] LU [2]


38m/min on platform level
Walking Horizontal 38m/min
61m/min on concourse level
Speed
Vertical 15m/min 12m/min
Flat surfaces 81.2p/m/min 80p/m/min
Flow
Stairs 55.55p/m/min 56p/m/min
Rates
Escalators 55.55p/m/min 120p/m/min

© AET 2013 and contributors


3
3. DYNAMIC MODELLING

A number of specialist simulation software packages have been widely used


for evacuation simulation. The results of the dynamic models presented in this
study have been calculated using the Mott MacDonald agent-based
pedestrian modelling software, STEPS [5].

This paper considers the following three aspects of evacuation dynamic


modelling:

• Comparison between the static assessment and dynamic modelling


approaches [6]
• Comparison of a real-life experiment with dynamic modelling outputs [7]
• Examples of dynamic modelling studies demonstrating the benefits of
undertaking sensitivity tests

3.1. Comparison study – Static and dynamic approaches

A comparison of the results between the two approaches, for the same
evacuation scenario, has been undertaken [6]. Two hypothetical station
layouts have been considered.

The first case, ‘Case 1’, is a station with a single platform island, raised above
the concourse. The second case, ‘Case 2’, is a station with two platforms,
below the concourse.

The following has been considered:

• Static assessment, in accordance with NFPA 130 [1] standard, which


assumes:
o One escalator per level is unavailable
o Full use of available exit routes from the start of the evacuation
o Optimal use of all exit routes based on egress capacity of each egress
element
• Dynamic modelling (Fixed approach) – Model set to replicate the NFPA 130
[1] assumptions as closely as possible
• Dynamic modelling (Flexible approach) – Model set to represent more realistic
assumptions for walking speed, available evacuation routes, choice of
evacuation route and initial location of people
It should be noted that, in the flexible dynamic modelling for Case 1, two
sensitivity tests have been analysed:

© AET 2013 and contributors


4
• Option 1 – The escalator on the shorter end of the platform is assumed to be
unavailable
• Option 2 – The escalator on the longer, and therefore busier, end of the
platform is assumed to be unavailable

The summary of the results is set down in Table 2.

Table 2. Comparison of the results of the static and dynamic approaches [6]

Dynamic
Static
Fixed Flexible

Time (s) Time (s) Difference Time (s) Difference

CASE 1
Option 1 280 23%
Platform 228 227 0%
Option 2 263 15%
Option 1 294 2%
Station 287 278 -3%
Option 2 284 -1%
CASE 2
Platform 203 212 4% 229 13%
Station 405 412 2% 312 -23%

With respect to the results of the fixed dynamic approach, it was found that
the dynamic simulation provided similar evacuation time results to the static
approach. This demonstrated that dynamic models can closely replicate the
NFPA 130 behaviour, when set up to do so.

However, differences were observed between the static results and the
dynamic simulation results using the flexible approach: dynamic platform
evacuation times were found to be longer than the static predictions for Cases
1 and 2, while dynamic station clearance times were comparable to the static
assessment results for Case 1 and shorter for Case 2.

On the platform level, longer platform evacuation times were explained by the
following: while the static approach assumed full use of all egress elements
from the start of the evacuation, the dynamic model took into account the
initial location of people spread across the platform level. This led to a delay
between the start of the evacuation and the time all egress elements were
used to their maximum capacity. These results demonstrated how the static

© AET 2013 and contributors


5
approach provided optimistic platform evacuation times in comparison with the
dynamic modelling approach.

On the concourse level, the key driver of the concourse clearance time was
the pedestrian walking speeds, which led to shorter walking times from the
concourse level to the station exit. This explained the comparable or shorter
station clearance times between the static assessment and the flexible
dynamic approach. These results demonstrated how the static approach
provided conservative station evacuation times in comparison with the
dynamic modelling approach.

The differences observed between Options 1 and 2 were due to the layout of
the station in Case 1: taking out an escalator on the busier section of the
platform (i.e. Option 2) had a greater impact on platform evacuation times
than taking out an escalator on the less crowded section of the platform.

It was noted that the choice of escalator made unavailable did not have any
impact on the static assessment results. This demonstrated the benefits of the
dynamic modelling analyses, which could assist in a better understanding of
the possible scenarios and identify the potential risks of a proposed design.

3.2. Real-life evacuation and dynamic modelling

A validation exercise was carried out for the evacuation of a multi-level office
building with two evacuation routes [7].

A total of 356 people evacuated the building and this evacuation took 5
minutes to clear the building.

A STEPS model was developed to simulate the evacuation of the office


building. The walking speed was adjusted to represent the values from NFPA
130. The flow rates of stairs were aligned to replicate the observed flow rate.
Figure 2 presents the snapshot taken from the model and Figure 3 shows a
comparison between the predicted and observed evacuation flow rates [7].

© AET 2013 and contributors


6
Figure 2 Snapshot of the dynamic model replicating real-life evacuation drill
[7]

Figure 3 Comparison between the observed and predicted flow rates through
the front (left) and back (right) exits [7]

A comparison of the observed data with the NFPA 130 [1] and LU [2] flow
rates showed that the observed flow rates and speeds down the stairs were
significantly lower than those defined in the industry standards. This identified
that a static assessment of the building would have provided optimistic
evacuation times.

The results presented in Figure 3 demonstrated that the dynamic model was
able to replicate the observed conditions quite accurately, resulting in
evacuation times similar to the observed times.

© AET 2013 and contributors


7
The experiment highlighted the value of a dynamic modelling analysis for a
multi-level building evacuation assessment. Such modelling can assist in the
understanding of pedestrian movements on each floor. In particular,
simulations can allow the user to see how the people flows from each floor
interact with each other when merging onto the same evacuation route. This
can demonstrate how these can cause delays in the building evacuation
times, which cannot be predicted by a static analysis of a building.

3.3. Dynamic modelling – Project applications

Two project cases have been considered in this paper as follows:

• Manchester Piccadilly Tram Stop for Greater Manchester Passenger


Transport Executive (GMPTE)
• Silvertown tunnel for Transport for London (TfL)

Manchester Piccadilly Tram Stop

The tram stop at Manchester Piccadilly train station has a unique layout, with
separate arrival and departure platforms. The purpose of the dynamic
modelling of the evacuation conditions at the station was to support a Fire
Safety Review of the proposed infrastructure modifications.

Detailed dynamic modelling was undertaken, using the STEPS software, to


assess a ‘Tram on Fire’ scenario. A fixed dynamic approach was undertaken
with the assumption that all passengers were randomly spread across the
mezzanine and platform levels, as opposed to the static assessment which
assumed that, as a worst case, all passengers were initially located as far
away as possible from any exit. Figure 4 presents the evacuation routes
available for the Manchester Piccadilly Tram Stop.

The static assessment provided an evacuation time of 3.6 minutes, and this
was in accordance with the industry standard evacuation time of 4 minutes,
while the dynamic simulation predicted an evacuation time of 2.9 minutes.

Due to the assumption on the initial location of passengers, the minimum


walking distance to an exit on the mezzanine level was much greater in the
static assessment than in the dynamic approach, leading to longer total
evacuation times in the static assessment.

© AET 2013 and contributors


8
Figure 4 Manchester Piccadilly Tram Stop – evacuation routes

This example demonstrated how critical the assumptions in a static or


dynamic assessment are, highlighting, once again, how a dynamic modelling
analysis can complement the findings of a static assessment.

Silvertown tunnel

This project was an example of using STEPS pedestrian dynamic model as a


supplement to the evacuation static calculations for a road tunnel.

The Transport for London project has considered the development of options
for a bored tunnel road link across the river Thames in London. The project
was at a feasibility stage, with further development work required. The project
set down the requirements for emergency escape for tunnel users.

Both a static and a dynamic modelling assessment were undertaken. The


approaches used the same assumptions and took into account the following:

• Specific vehicle mix and occupancy


• Detection and pre-movement time
• Bus/coach exit rate
• Walking speed.

© AET 2013 and contributors


9
The worst case evacuation was assumed to involve a coach, carrying a large
number of people, located adjacent to the fire location.

The results of the dynamic modelling were comparable with those obtained
using the static approach. It demonstrated that people evacuated along the
walkway and into the cross passage without a significant delay due to
queuing. The last person reached the emergency cross passage after 10.3
minutes from the start of the fire. Figure 5 presents snapshots of the simulated
evacuation at 1 minute intervals.

Figure 5 Snapshots of the simulated evacuation at 1 minute intervals

This example demonstrated how the dynamic model can verify the findings of
the static analysis. In addition, the modelling provided stakeholders with a
visual representation of people movements during the evacuation process, as
well as further information on the interaction of evacuees between each other
and with their environment.

© AET 2013 and contributors


10
Other applications:

Dynamic pedestrian simulation has also been used in support of CFD


(Computational Fluid Dynamics) modelling analysis, which can predict
visibility conditions, temperatures and toxic elements’ concentrations for the
optimisation of ventilation systems.

The STEPS software can import this CFD data to provide the dynamic model
with a visual representation of smoke spreading in the environment and
assess the impact of reduced visibility on people behaviour and, therefore,
predicted evacuation times. For example, the pedestrian simulation software
can be set up to reduce people’s walking speeds as visibility conditions
deteriorate. In addition to this, the model can also extract the accumulated
exposure dose data for any toxic element for each person, enabling the
assessment of casualties for different scenarios.

Figure 6 illustrates the simulation example of a tunnel evacuation merged with


the visibility plots.

Figure 6 Snapshots of the simulated evacuation at 1 minute intervals

© AET 2013 and contributors


11
This example illustrates how dynamic models can provide more detailed and
valuable information than static assessments, as well as complement the
findings of CFD analyses, providing stakeholders with predicted results for
various scenarios.

4. CONCLUSIONS

This paper has given an overview of evacuation calculation methodologies,


outlined the capability of pedestrian dynamic models in evacuation and fire
safety analysis, and presented project examples where those type of analysis
have been used.

A static analysis is a quick, consistent and relatively straightforward way of


assessing a design. However, the examples in this paper demonstrate that it
can be too conservative in some cases and too optimistic in others, depending
on the assumptions made.

Dynamic models can provide results similar to those obtained using a static
approach and hence, can be assessed against traditional industry standards.

However, dynamic modelling can also supplement the findings of a static


assessment, and provide a better understanding of the potential risks of a
proposed design.

The dynamic modelling approach enables designers to model more realistic


people behaviour, take into account a number of additional factors, including

© AET 2013 and contributors


12
impact of smoke data on evacuation times (such as decreased walking
speeds as visibility levels drop) and choice of evacuation routes by people.

In addition, a dynamic analysis can provide stakeholders with a visual


representation of people movements during the evacuation process, as well
as further information on the interaction of evacuees with each other and with
their environment. This can highlight potential areas of localised congestion.

Evacuation assessments use a number of parameters. The need to consider


the sensitivity of any proposed design underlines the value of undertaking a
dynamic simulation.

BIBLIOGRAPHY

[1] NFPA 130 (2010) Standard for Fixed Guideway Transit and Passenger
Rail Systems

[2] London Underground Limited (2011) Station Planning, S1371 Standard,


Issue A5

[3] Guide to Safety at Sports Grounds (2008) , Fifth Edition

[4] Fruin, J.J. (1987) Pedestrian Planning and Design

[5] Mott MacDonald STEPS software: http://www.steps.mottmac.com

[6] Waterson, N. P., Le Bail, S., Castle, C. J. E. (2010) A comparison of


evacuation predictions made using agent based simulation and code-based
approaches, 8th International Conference on Performance-Based Codes and
Fire Safety Design Methods

[7] Waterson, N. P., Mecca, A., Wall, J. M. (2004) Evacuation of a multi-level


office building: Comparison of predicted results using agent-based model with
measured data, Interflam paper

ACKNOWLEDGMENT

The authors are grateful for the permission given by both TfL and GMPTE for
the use of their evacuation study material.

© AET 2013 and contributors


13

You might also like