prof ed galea director fire safety engineering group [email protected] rmit melbourne 29 oct...

[email protected] http://fseg.gre.ac.ukRMIT Melbourne

29 Oct 2015

Engineering Safety, Security and Efficiency

through Fire and Evacuation Simulation

Prof Ed Galea

Director Fire Safety Engineering Group

University of Greenwich


29 Oct 2015

FSEG: Modelling safety and security• FSEG was Founded in 1986 by Prof Galea

in response to the Manchester Airport B737 fire.

• Today it consists of 30 researchers including:

– fire engineers, CFD specialists, psychologists, mathematicians and software engineers.

• Research interests include the mathematical modelling and experimental analysis of:

– evacuation dynamics in complex spaces,

– pedestrian dynamics in complex spaces,

– combustion and fire/smoke spread,

– fire suppression,

– homeland security

• Application areas include:

– aerospace, built environment, marine and rail.


29 Oct 2015

Applications of the EXODUS software

Rail Stations

Large PAX Ships Naval Ships

Royal Ascot Historic Buildings

A380 – Super Jumbo Millennium Dome Stadium AustraliaAirbus flying wing

Canary Wharf

Beijing Olympic Stadium

WTC 9/11 analysis Pentagon Shield

Statue of LibertyForensic analysis

Rhode Island


29 Oct 2015

Applications of the SMARTFIRE software


29 Oct 2015

FSEG Data Collection

Rail Car Evacuation

Cruise ship evacuation

at sea

Assist devices for PRMUnannounced Full

Building Evacuations

public eventComplex spaces


29 Oct 2015

EXODUS Applications


29 Oct 2015

Underground station evacuation• LuL station with fire.


29 Oct 2015

buildingEXODUS and SMARTFIRE simulation of Station Nightclub fire

•Link fire simulation directly with evacuation analysis

•Directly expose agents to developing hazard environment

•Predict fatalities and injury levels.

• Last survivor evacuates after approx 127 seconds.

• Simulation predicts :

•84 fatalities compared with 100 in actual incident.

•25 serious injuries, of which 6 are life threatening.


29 Oct 2015

Evacuation Times

How long to get everyone

out?


29 Oct 2015

WTC1: evacuation 5-25 min into simulation

• Precise evacuation time for WTC1 is not known – 102 min to collapse.

• buildingEXODUS prediction of WTC1 evacuation suggests that the

evacuation time for 8239 people was 87 min.

• If fully occupied with 25000 people, predicted evacuation time 138 min,

resulting in 7500 fatalities.


29 Oct 2015

buildingEXODUS V6.0 Lift Model

• Full building evacuation using stairs alone often

not practical.

• Use of lifts have been suggested as the panacea

for all of our evacuation problems – but how

good are lifts?

• Most lift models and design engineers do not

include human behaviour, they treat people as

compliant ball-bearings, if assigned to use the

lift, the agent will use the lift!


29 Oct 2015

Stairs or Lifts? How will people

actually behave in an emergency?


29 Oct 2015

Why is it important to consider human

behaviour associated with lift usage

• Under what conditions will people wait for the lift?

• WTC survivor from the 77th floor of North Tower who

evacuated on 9/11 said:

“Let me add too that, at the 44th floor there was what they call

an inter-zone elevator bank, we were led off the stairwell at

the 44th floor and shown to that elevator where there are

hundreds of people milling and I looked at that and I

turned around to my team and I said ‘no, I am not waiting

for an elevator in a building on fire. Let’s go’ and I walked

back to the stairwell and they did too and then we proceeded

down”


29 Oct 2015

• buildingEXODUS lift model includes kinematics of lift

operation AND human behaviour associated with lifts.

• The agents behaviour in utilising a lift is complex and

consists of:

– Deciding to utilise the lifts or the stairs.

– If using the lifts:

• Which lift bank to use

• When arriving at the lift bank, deciding whether or not to wait.

• If waiting, where to wait.

• How long to wait for.

• When lift arrives, boarding behaviour.

– Much of the behaviour used in the model is based on data collected

by FSEG from an international survey involving 468 people from 23

countries.

buildingEXODUS V6.0 Lift Model – Behavioural features


29 Oct 2015

International survey: 468 people from 23 countries

Direction

As wait time

increases, %

electing to wait for

lift decreases

exponentiallyAs floor height

increases, % that

would wait for the lift

for a given time

increases

Less than 10%

prepared to wait

more than 15 mins

across all floors

Majority only

prepared to wait

up to 5 mins


29 Oct 2015

WTC Evacuation using Lifts• So could the 25,500 people in the WTC have been evacuated

if they used lifts?

• The building had 20 express lifts (55 person capacity)

servicing the Sky Lobbies and 60 local lifts (12 person

capacity).• buildingEXODUS suggests:

• 1 hr 23 mins to clear

tower

• 40% faster than stairs

• 58% only use stairs

• 39% only use lifts

• it would have been

possible to get everyone

out before the building

collapsed.


29 Oct 2015

Wayfinding

How do people find their way

out?


29 Oct 2015

WTC WAYFINDING• WTC1/025/0002:

P “honestly I didn’t know where the evacuation stairwells where….. they say, … look for the exit signs when you go in a place, they really mean that because, y’know unless something’s happened before, you’re not go to be able to find it.

• WTC1/057/0002:

P “… we couldn’t at that point find the exit. Our stairwell had ended and there were no guide posts to go anywhere….so a number of people started searching for some place to go for another stairwell to go down from the 44th floor. Eventually someone found it so we continued down.”

• WTC1/087/BDAG

P “…we actually walked past the fire escape, kinda had to turn around and double back until we found the fire escape…

• Many people were unable to find the stairs, even though they had been in the building for months.

• Many people failed to see the emergency signage


29 Oct 2015

Sign Recognition Experiments• Experimentally examine how occupants interact with signage

in both normal movement and evacuation.

• Attempt to determine likelihood that those who can see a sign,

recognise the sign, correctly interpret the information and

follow the information.

•68 test subjects

•41 naïve subjects


29 Oct 2015

Sign Recognition - T intersection

• Female, naïve subject, 2 sec decision time, makes correct decision. From questionnaire subject said they saw and followed the sign.

• Female, naïve subject, 9 sec decision region, makes incorrect decision. From questionnaire did not see sign

• For T intersection, 61% of naïve subjects failed to “see” the sign.

• Of those who “registered presence of sign”, 100% followed instructions

• Average decision time for those who see sign 2.6 s, those who do not see sign 5.6s

•Correct choice •Incorrect choice


29 Oct 2015

Sign Recognition – Improved affordance• It is suggested that poor identification of signs is due to poor

affordance associated with signs.

• Can improve affordance of a sign in several ways:

• make larger – improve sensory and cognitive affordance

• introduce lights – improve sensory affordance

• introduce green lights – improve cognitive affordance

• introduce flashing lights - improve sensory affordance

• introduce running lights - improve cognitive affordance

• FSEG in collaboration with UK company EVACLITE

(www.evaclite.com) have come up with a sign addressing the

poor affordance issue by introducing running, flashing, green

lights to the sign.

• Sign is activated on alarm – Active Dynamic Signage System


29 Oct 2015

Sign Recognition – Improved affordance• Using the ADSS, the wayfinding experiments were repeated

in July 2012 with 48 unfamiliar participants

• 85% (41/48) of people ‘see’ the dynamic sign – an increase in

detection rate of 120%.

• 100% of people who see the dynamic sign follow the sign.

• The vast majority of people interpret the flashing arrow

correctly and find the new design useful.

Detected

sign?

DSS Conventional

sign

Yes 1.8 s 2.6 s

No 5.7 s 5.6 s


29 Oct 2015

Intelligent Signage Systems• As part of EU FP7 project GETAWAY, the ADSS concept is

being expanded to include Intelligent Active Dynamic Signage

System - IADSS.

• In addition to making the exit sign more noticeable:

• Signage system also indicates that an emergency exit

route is no longer considered viable

• ADSS is controlled via simulation and human intervention

to identify the optimal exit route given the current situation.

• Optimal route can be determined by faster than real time

buildingEXODUS simulation taking into consideration:

• Current population distribution and

• Spread of fire hazards (heat, smoke and toxic gases)

• Optimal route can also be determined by human operator


29 Oct 2015

EXODUS Evacuation simulation

resultsADSS control

commandSensor

readings

The GETAWAY Intelligent Active Dynamic Signage System

CAE

DE EXODUS

Station CCTV

Live video input

FDS & ADSS control unitSensors DSS

Real time people counting data

Simulation request

Station Supervisor

Situational information collection Intelligent decision making

ADSS


29 Oct 2015

Intelligent Signage Systems• ADSS not only has application for fire applications but also

for terrorist situations.

• Using CCTV security staff identify regions were the hazard

(gunmen) is located and direct people away from the region by

activating appropriate signs


29 Oct 2015

ADSS Showing Negated Route• ADSS extended to indicate that an evacuation route is no

longer considered viable.


29 Oct 2015

ADSS Negated Sign Concept• Survey involving 451 people from 10 countries

• 4 potential designs shown to each participant, without providing an

explanation of what the sign meant.

• Participants asked to write what they thought the sign meant.

Non-Fire

Correct

Fire

Correct

Total

Correct

93%

(193 total)

93%

(238 total)

93%

(431 total)

85%

(196 total)

83%

(240 total)

84%

(436 total)

72%

(182 total)

79%

(227 total)

76%

(409 total)

56%

(191 total)

63%

(239 total)

59%

(430 total)


29 Oct 2015

Exit A

Platform 2

Exit B

Exit CExit D

� T1 – 139 participants

Configuration and Signage Systems Tested

Exit A

Platform 2

Exit B

Exit CExit D

� Participants distributed across boxes 1-7 on platform

� T2 – 152 participants


29 Oct 2015

Trial 1 – 100% of participants

use their nearest exit. Signage

only accounted for 26.8% of

participant exit selection

Trial 2 – 63% of participants

use the indicated exit. 57% used

their nearest exit, compared with

100% in Trial1.

Comparing Trial 1 and Trial 2


29 Oct 2015

� TS3 – 64 participants, distributed in box 1 and left blank area

TS3 – Configuration and Signage Systems Tested

Exit A

Platform 2

Exit BExit C

Exit D


29 Oct 2015

• 66% of participants by-pass exits 1, 2 and 3 and utilise target exit.

• 34% of participants chose to use their nearest exit compared to 100%

TS3 2014, Trial 2


29 Oct 2015

IADSS

EXODUS required 33 sec to perform all 7 evacuation scenarios

Ranking algorithm identified

optimal exit strategy and was able

to activate signage prior to alarm

activation


29 Oct 2015

TS3 Questionnaire Analysis Results

Statement 1

“This sign assisted me in selecting an exit to

use/ which exit NOT to use/which exit not to

use and which exit to use.”

Level of agreementAgree /

Strongly Agree

Disagree /

Strongly DisagreeTotal

94% 3% 79

74% 19% 53

83% 13% 80

70% 25% 77

Weighted Average 81% 14% 289

� Participants’ level of agreement that the sign assisted them in

identifying an exit to use (TS3.2 and TS3.3).


29 Oct 2015

Exit Flows

Still Lots to Learn


29 Oct 2015

Exit Flows• The characterisation of exit flows is one of the

fundamental concepts in building design.

• Key parameter in building safety regulation all over the

world, values accepted and used for many years.

• In the UK for example, the unit flow through an exit is

defined as 1.33 people/m/sec.

• This is taken as a fundamental value to be applied in all

situations.

• But how fundamental is this parameter?

• How is it affected by:• Environmental conditions e.g. ambient temperature

• Occupant encumbrance e.g. people exiting a rail station or

airport with luggage?


29 Oct 2015

COLD v WARM

COLD WARMFlow (ppm)

247.1

[242.0 – 252.0]

Flow (ppm)

278.0

[268.0 – 286.0]


29 Oct 2015

Impact of EnvironmentPeak period flow (ppm) Average

(ppm)

Average

(p/m/s)Time interval (sec)

5-10 10-15 15-20 20-25 25-30 30-35

Cold (ppm) 278 247 242 259 226 216 246 1.71

Warm (ppm) 360 316 292 244 236 220 278 1.93

• COLD: Max temp 9oC, light

snow on occasions

• WARM: Max temp 19oC, sunny.

• Average WARM exit flows

13% faster than the COLD exit

flows.

• Trials repeated 3 times, data

represents average of the three.

• Different populations for COLD

and WARM trials, but very

similar demographics.

• Initial packing density identical.


29 Oct 2015

Impact of Encumbrance• Participants were given a range of different types of

encumbrance ranging from handbags to pushchairs and

bikes.

• Roller bags and baby buggy were weighted down with a

15kg sand bag.

• Smaller luggage items were packed with bubble wrap in

order to pack them out.

• Three levels of encumbrance were investigated involving

set proportions of the population carrying some form of

encumbrance. These were 0%, 40% and 60%.

• Each trial repeated 3 times.

• Efforts were made to ensure that the number and type of

encumbrance was the same through repeat trials.


29 Oct 2015

Types of Encumbrance

Type Available 40% 60%

Handbag 66 34 49

Satchel 36 15 29

briefcase 24 11 19

rucksack 22 9 13

Small

roller bag

3 2 3

Large

roller bag

3 2 3

bike 3 1 3

Baby

buggy

3 2 3

160 76 122


29 Oct 2015

0% v 40% v 60% Encumbrance

0% Baggage

Flow (ppm)

278.0

[268.0 – 286.0]

Flow (ppm) 233.7

[226.3 – 243.4]

40% Baggage 60% Baggage

Flow (ppm) 251.4

[246.9 – 255.4]


29 Oct 2015

Impact of Encumbrance

• As encumbrance level increases, exit flow decreases.

• Compared to the case where the participants have no encumbrance, exit flow

decreases by 10% when approx 40% of participants are encumbered and 16% when

60% are encumbered.

• Increasing encumbrance level from 40% to 60% results in a 7% decrease in flow.

• The natural variation in the average exit flow rate increases as the level of

encumbrance increases from 40% to 60%.

Flow (p/min)

Unit Flow (p/m/s)

0% Encumbrance 40% Encumbrance 60% Encumbrance

Flow (unit flow) 278 (1.93) 251.4 (1.75) 233.7 (1.62)

% Difference -- 9.6 15.9


29 Oct 2015

Use of Bollards• Bollard Arrays are intended to protect critical infrastructure from

hostile vehicles or “car bombs”

• Had a bollard array been present around the entrance to Glasgow

airport it would have prevented the vehicle from approaching the

airport terminal.

Glasgow airport

30 June 2007


29 Oct 2015

Use of Bollards• Today, security bollards are a common sight in London and

other cities around the world.


29 Oct 2015

4.5m exit, No BA, 1m, 2m and 3m BA

1m BA

Unit Flow (p/m/s)

1.76 (-9%)

2m BA

Unit Flow (p/m/s)

1.75 (-9%)

3m BA

Unit Flow (p/m/s)

1.79 (-6%)

No BA

Unit Flow (p/m/s)

1.92


29 Oct 2015

Modelling VS Experiment: 4.5m exit, 1m BA

Time (s) 3 Trial Average

(ppl/m/min)

100 Sim Average

(ppl/m/min)

Difference

(%)

5-10 112.9 107.7 -4.6%

10-15 104.9 109.2 +4.1%

15-20 110.2 106.8 -3.1%

20-25 98.7 108.3 +9.8%

25-30 100.4 100.0 -0.4%

Average 105.4 106.4 +0.9%


29 Oct 2015

Research: Urban Scale

Evacuation and Crowd

Dynamics


29 Oct 2015

Large crowd simulation and visualisation

• Trafalgar Square

demonstration:

125,000+ people

simulation

• Love Parade Disaster

reconstruction: 100,000

people simulation.


29 Oct 2015

Assembly and bus boarding process


29 Oct 2015

EXODUS Coupled to Flood Model• As part of a proof of concept project, EXODUS was coupled to a flood

model.

• Rising water levels predicted by the flood model could be passed onto

EXODUS and represented within the evacuation model.

• Agents within EXODUS are then given behaviours to move away from

rising water levels.

• When water level rises above head height, agent is assumed to have

drowned.

• Model is still in early phase of development but demonstrates that it is

possible to couple advanced egress models to advanced CFD based flood

prediction models.

• EXODUS capabilities to read map geometries, to utilise multiple

computers for rapid computation and to utilise hybrid space concept

enable it to be used to model vast regions in relatively short times.


29 Oct 2015

EXODUS Coupled to Flood Model• 10,000 agents, city section of 1 km x 1 km – 20 minutes of simulation.


29 Oct 2015

Large Scale Disaster Planning and Management• As part of EU FP7 project IDIRA EXODUS is being configured

for use in large scale urban disaster applications.

• This could be for applications in floods, Tsunami, earthquakes,

forest fires, etc.

• Software is used to assist in planning large-scale movement of

people and for use during an incident to assist in management.

• Models of urban regions can be pre-built and stored for use during

an incident or regions of interest can be built during an incident.

• During an incident the model can be reconfigured as new

information is made available as the scenario changes.

• e.g. loss of evacuation routes, changes in status of refuge areas, etc.

• Web Application – An easy to use GUI for clients to interact with

the EXODUS simulation tool.

– OpenLayers – Client application used to display base maps

(Googlemaps/OSM) and Overlays (Population density contours)


29 Oct 2015

Tokyo Japan

21 November 2014

FOREST FIRE• Swinley forest fire was the largest in Berkshire’s history

• 5 May 2011, 300 hectares of forest

• Very close to built up areas

• 1220 people directly affected: TRL - 800 , Business Estate - 200, Pub - 200,

Residential dwellings 20

• Close to the high-security Broadmoor Hospital

A3095

Broadmoor Hospital


29 Oct 2015

Tokyo Japan

21 November 2014

FOREST FIRE• Conditions were variable on the day.

• Concerned of repercussions if wind changed.

• Spread of fire modelled using Prometheus by Tom Smith KCL

• Considered what would have happened if wind changed direction.

• How long to evacuate threatened population?

• Actual region burnt• Simulated burn region given wind change

Business

Estate TRL

Residential

Dwellings

Pub

Assembly

Area


29 Oct 2015

Evacuation ScenarioEvacuation Scenario:

Assume evacuation initiated at the start of the fire.

TRL:

• Main exit, turn right onto B3430, go to Assembly Area

• Initiated by phone call at t = 0 s, RT = 1 – 2 min

Business Estate:

• Main exit, turn right and right onto B3430, go to

Assembly Area.

• Initiated by phone call at t = 0 s, RT = 1 – 2 min

Pub:

• Exit, turn left onto B3430, go to Assembly Area.

• Initiated by police visit at t = 0 s, RT = 0.5 – 1 min

Residential Dwellings:

• Follow B3348 and then A3095 towards the pub, turn

left onto B3430, go to Assembly Area.

• Initiated by police door to door, requires 5 min to reach

first house, 5 min to get ready, 1 min to next house.

Assembly Area

TRL

Pub

Business

Estate

Residential

Dwellings

House

B3348

B3430

A3095

1.5 km

1.7 km


29 Oct 2015

Assembly Simulation •Each dot represents an area of 6m x 6m, colour represents population density


29 Oct 2015

Safety MarginsSafety of Pub

• Pub clears point (B) at 4 min 28 sec.

• Pub assembles at 18 min 35 sec

• Fire approaches A3095 at 91 min.

• Pub has safety margin of 86 minutes.

Safety of TRL

• TRL clears point (A) in 10 min 45 sec.

• TRL assembles at 35 min 22 sec.

• Fire approaches TRL at 301 min,.

• TRL has a safety margin of 290 min.

Safety of Business Estate

• BE clears point (E) at 13 min 13 sec.

• BE assembles at 51 min 44 min

• Fire approaches BE at 392 min.

• BE has a safety margin of 378 min

Safety of Residential Dwellings

• RD clear safe point (C) at 53 min 37 sec.

• RD assemble at 1 hr 11 min 08 sec

• Fire approaches A3095 at 91min.

• RD have a safety margin of 31 min.

• Fire reaches A3095 at 126 min, a critical

point deeming that section of the road is

unusable.


29 Oct 2015

Assembly Times• Overall Average Assembly Time: 1 hour 11 min 08 sec

• Pub Assembly Time: 19:01; average travel distance: 819m

• TRL Assembly Time : 35:22; average travel distance: 1452m

• Business Estate Assembly Time: 52:11; average travel distance: 2338m

• Houses Assembly Time: 71:01; average travel distance: 2726m

Pub TRL Business

Estate

Houses


29 Oct 2015

Fire and Evacuation in

Domestic Dwellings


29 Oct 2015

Source: Fire Statistics Great Britain, 2011-2012, Department for Communities and Local Government

0

10

20

30

40

50

60

70

Per

centa

ge ADFs as a %age of all

fires

ADF deaths as a %age

of all fire deaths

• Proportion of fire-related deaths and injuries caused by ADFs remained between 58% and 70%.

• Despite reduction in number of fires, proportion of ADF-related injuries and deaths unchanged.

• While likelihood of having a domestic fire have decreased, when a domestic fire does occur, the likelihood of being injured or dying has not decreased.

HUMAN BEHAVIOUR IN DWELLING FIRES


29 Oct 2015

HUMAN BEHAVIOUR IN DWELLING FIRES• Why study human behaviour in dwelling fires?

– While the number of UK accidental dwelling fires is falling year on year, percentage of injuries and deaths incurred in these fires has not decreased and so the risk of harm when experiencing a dwelling fire remains unchanged.

– Must determine why so many injuries and fatalities in ADFs.

• FSEG, in collaboration with Kent Fire and Rescue Service through a KTP funded by the EPSRC and TSB, are undertaking a project to understand human behaviour in domestic fires – project LIFEBID aims to:

– Better understand how people respond to a developing emergency

– Inform emergency services how best to advise the public on what to expect and do in the case of a dwelling fire

– Reduce the number of negative outcomes from dwelling fires

– Advise on strategic investment of resources to achieve best effect

– Measure the effectiveness of interventions


29 Oct 2015

BEHAVIOURAL RESPONSE TO FIRE• As part of the LIFEBID project a survey was developed to assess

public response to a developing fire.

• Participants shown a segment of video of a developing kitchen fire and asked if they felt that they could:

– Safely enter the room to tackle the fire.

– Safely extinguish the fire.

• Participants shown video for a only a few seconds, as if they just briefly witnessed the fire and had to make a rapid decision as to what they should do.

• Participants split into three groups and shown the fire and smoke video representing:

– Early Fire Growth Stage, most people should be able to safely extinguish fire, room safe to enter

– Mid Fire Growth Stage, unlikely most people would be able to safely extinguish, room is verging on dangerous to enter.

– Late Fire Growth Stage, most people would not be to safely extinguish fire, room is not safe to enter.


29 Oct 2015

BEHAVIOURAL RESPONSE TO FIRE

•Early Stage: most people would be able to safely extinguish the fire.

•Mid Stage: some people would be able to safely

extinguish the fire.

•Late Stage: most people would not be to safely extinguish the fire.

Do you believe that you could

safely extinguish the fire?

Stage of fire

development

No Not Sure Yes

- Early/47 9% 11% 81%

- Mid/47 17% 36% 47%

- Late/47 47% 23% 30%

• Could people be reasonably expected to safely extinguish the fire in:


29 Oct 2015

LIFEBID FIRE BEHAVIOUR STUDY• Males more likely to tackle large fires

MALES Do you believe that you could safely extinguish

the fire?

Stage of fire

development

No Not Sure Yes

- Early/20 15% 20% 65%

- Mid/25 12% 32% 56%

- Late/32 34% 31% 34%

FEMALES Do you believe that you could safely extinguish

the fire?

Stage of fire

development

No Not Sure Yes

- Early/27 4% 4% 93%

- Mid/22 23% 41% 36%

- Late/15 73% 7% 20%


29 Oct 2015

• LIFEBID has moved onto the second stage of the study.

• Anyone who has experienced a fire in the past 12 months in

their home – however small – can take part in the online

survey.

• The aim of the survey is to understand people’s actions and

decisions during a fire in the home.

• Anyone who has experienced a fire in a domestic

environment can complete the survey, not just those living in

the UK.

• Details can be found on the FSEG web site

at: http://fseg.gre.ac.uk/fire/lifebid.html


29 Oct 2015

maritimeEXODUS


29 Oct 2015

• Royal Caribbean vessel Jewel of the Seas• Carries approx 2500 pax and 842 crew across 12 pax decks• The vessel has:

• 18 assembly stations in 4 areas spread over 2 decks• 7 MVZ

• The public spaces include:• Accommodation areas,

• Retail areas,• Restaurants, • Bars, • Theatre,• Cinema,

• The Assembly exercise was carried out on the first leg of its Baltic Sea trip, between Harwich (UK) and Copenhagen (Denmark)

• The assembly exercise took place on 31st July 2010 at 10:00

• 2292 paxs on board

• 1950 paxs with tags assembled

• Assembly required 29 mins

Full-Scale Cruise Ship Evacuation at Sea


29 Oct 2015

maritimeEXODUS predictions


29 Oct 2015

Simulated Assembly for 200 heel • 20o Heel results in a 23% increase in the average total assembly time for

the IMO day case scenario i.e. increase from 17.2 min to 20.7 min


29 Oct 2015

airEXODUS and

SMARTFIRE


29 Oct 2015

Rupture Scenarios with Wind

• Exit Scenario L1, R1, R3, R4

R4 not used due

to spread of fire

R3 not viable

after 90 s

�297 passengers and crew

�Local flashover occurs at 55 s, much

sooner than the required evacuation time;

� 55 fatalities occur

� Evacuation time 270.1s

Fuselage rupture

due to crash


29 Oct 2015

SMARTFIRE GPU


29 Oct 2015

SMARTFIRE – GPU IMPLEMENTATION

� Currently SMARTFIRE run times can be reduced using parallel version on multi-core CPUs and networked PCs.

� Present development of SMARTFIRE-CUDA produces solver speed-ups of 40-50 times faster than the present version of serial SMARTFIRE.

� These reduced time frames lead to greater possibilities in using CFD, e.g. fully 2 way coupled fire and evacuation modelling.

� In the future it is planned to extend the SMARTFIRE-CUDA to also run on networks or clusters of GPU equipped PCs to deliver vastly reduced simulation times.

� Next generation SMARTFIRE will be able to use Graphics Processing Units (GPUs), e.g. NVidia Graphics Cards and Tesla Cards.

� GPUs are orders of magnitude faster for certain computational operations (mostly SIMD).

� Are generally “cheap” (< $1000) and give high Gflops/$.

� Suitable card may already be present on your PC!


29 Oct 2015

Standard one-way coupling of

SMARTFIRE and EXODUS

is expanded to include two-

way coupling through

enhanced performance of

GPU version of SMARTFIRE


29 Oct 2015

Two-way Coupling of SMARTFIRE

GPU and buildingEXODUS


29 Oct 2015

VIRTUAL TRAINING

ENVIRONMENTS


29 Oct 2015

AUGGMED – EU HORIZON 2020 Programme� The AUGGMED platform will generate non-linear scenarios for special forces dealing with

terrorist situations in crowded public spaces

� Engine makes use of high fidelity simulation of crowd behaviour (EXODUS), fire/toxic gases

(SMARTFIRE), explosions, etc.

buildingEXODUS

SMARTFIRE


29 Oct 2015

AUGGMED ENVIRONMENT

SMARTFIRE

UNITY3D

buildingEXODUS


29 Oct 2015

AUGGMED ENVIRONMENT

AUGGMED environment will allow users to:• Select a particular simulated blue team member

and assume control of that agent

• All actions of selected agent dictated

by real player rather than EXODUS

• In VR user is within the VR environment

playing the scenario sitting at a desk

• In MR user is located in the targeted installation

viewing the real structure through head

mounted display and viewing virtual people

(civilians, red team members) as participants

in the scenario


29 Oct 2015

Damage Control Simulation Environment – DSTO

• FSEG in collaboration with DSTO are developing DCSE to:

– assess effectiveness of current procedures. – develop new procedures,– Train crew, and– On-board live damage control decision support.


29 Oct 2015

Modular View of Integrated Survivability Architecture

• Eventual architecture will link many key ship survivability and simulation capabilities into the DCSE.

• The Phase 1 prototype will include Blast Damage, Fire, Crew Behaviour within the proposed simulation environment.

maritimeEXODUS

SMARTFIRE


29 Oct 2015

Logical Architecture of the DCSE


29 Oct 2015

CONCLUDING COMMENTS• Safe evacuation is challenging and requires careful planning, it doesn't

just happen.

• Use of reliable modelling tools in conjunction with good dataenable fewer arbitrary assumptions to be imposed, allowingconditions to be modelled rather than assumed.

• Simulation can be used to assist in planning to ensure:

– efficient throughput,

– comfort,

– safety and

– security.

• Finally, while it may be appealing to make simplifying assumptions

concerning human behaviour it is important to remember:

There’s nowt so queer as folk

prof ed galea director fire safety engineering group [email protected] rmit melbourne 29 oct...

Documents