Research Interests

1
B R E C entre for F ire Safety Engineering
FireGrid An integrated Emergency Response
System for the Built Environment Stephen Welch
EPSRC/TSB Pilot Projects meeting Thursday-Friday
17-18 July 2008 NeSC, Edinburgh
2
FireGrid project

• Technology Strategy Board
• Technology programme
• 2.3M, 2006-2009
• 6 RA/SRA
• 6 PhD
• EPSRC
• Network project
• 2005-2006
• Dalmarnock fire tests
• 2006

3
Partners

• BRE research and consultancy
• Arup research and consultancy
• SIMULIA multi-physics software
• ANSYS-CFX fire/structure software
• Xtralis detection technologies
• IHPC grid, HPC
• LFEPA fire rescue services
• UoE
• SEE, EPCC, NeSC, AIAI

4
Mont Blanc
Kings Cross
Kobe
Piper Alpha
5
Emergency services lack even very basic
information when responding to most emergencies
6
The costs of fire

• Life Loss
• Material Loss
• Loss of Business
• Insurance
• Fire Protection
• Fire Brigades

0.8-1 of GDP of industrialised economies
7
Sensor-rich built environments
8
The Grid
The Internet
The next generation of internet the Grid exists
for science and research use and should in due
course be available for commercial use
enables a DISTRIBUTED system to ensure perennial
availability
Dynamic, multi-domain virtual organisations
In 2003, Europe 267 institutes, 4603 users
Elsewhere 208 institutes, 1632 users
9
Vision of FireGrid

• Facilitate transformed emergency response
• Via information on incident evolution
• Real time status
• Prediction of future hazard
• Innovative simulation tools
• Grid-enabled
• Sensor-linked
• Command and control
• Intelligent systems for end users

10
Novel framework

• Computations and secure info transfer via grid

Emergency Responders/ Actuators
1000s of networked sensors
Data Maps, pre-runs, scenario matching
Super-real-time simulation (HPC)
11
Technologies

• Sensors, networking and communication
• Super real-time simulation egress,
  fire/smoke/heat/toxicity, structure (HPC)
• AI, KBS command, control, communication,
  intelligence (C3I)
• Distributed resources for robustness and
  reliability (Grid technologies)
• Fire fighting technologies actuators (auto),
  firefighters tools (manual)

12
Technology integrations
13
FireGrid architecture
14
Dalmarnock fire tests
Featured in a BBC Horizon Documentary (2007) See
www.see.ed.ac.uk/fire/news.html
15
Fire sensors
ENLARGE
16
Structural sensors
17
The fire
18
Aftermath
19
Controlling the Fire
20
Controlled fire
21
Controlled fire aftermath
22
Simulation tools

• A-priori simulation
• A big challenge
• Wide scatter in predictions
• A-posteriori simulations
• Also challenging!
• Complexity of fire phenomena
• Multi-fuel
• Wind effects
• Random aspects
• Model steering via sensors

23
Current limitations

• Disparate technologies
• Hardware and software
• Not fast enough!
• Particularly advanced tools
• Require
• Holistic approaches
• Hierarchical
• Redundancy
• Grid enablement
• Sensor-linking

24
Data handling

• Large volumes of data logged
• 25GB of results
• Dominated by video records
• Data storage and access via grid
• Instrumentation issues
• Wireless sensors
• Data reduction
• Attenuation

25
Command control

• Control
• Fire Test Two controlled fire
• Early intervention successful
• Assist fire fighting
• Command
• Human decision-makers quickly overwhelmed
• Understanding current conditions
• Making predictions
• What does the end user
• actually need?!

26
Model integration grid

• Simulation tools for
• Fire development
• Human behaviour
• Structural response
• Provide support for
• Early fire detection
• Guiding egress
• Hazard prediction, including collapse

27
Sensor linking
Modelling
Model state
Measurements
Actual state (reality)
Observed state (sensor data)
28
Data assimilation
Hypothesis (model)
Experiment
Reality
Model uncertainty
PDF(H)
Posterior Distribution
Prior Distribution
Data likelihood
29
(No Transcript)
30
Sensor-linked simulation
Courtesy S. H. Koo
31
Real-time steering
32
Sensor networks

• System requires
• Large numbers of sensors
• Large buildings
• Frequent updates
• Early detection
• gt Significant burden on
• communication protocols
• Wireless networks
• Redundancies
• Self-organising

33
Command and control

• Scope
• Automated responses
• Human decision makers
• C3I (Command, control,
• communications intelligence)
• Draws upon AI concepts
• Knowledge-based
• Planning techniques
• Requires support layers
• Abstract raw data
• Interpret simulation results

34
CC architecture
35
Mapping ontologies

• Computation/simulation (Quantitative inputs from
  sensors and simulation based forecasts)
• Temperatures (C)
• Smoke obscuration (m)
• Heat flux (W/m2)
• Deflections, strains
• Safe egress times (min/sec)

• Fire fighting (Qualitative decision making)
• When to turn on the alarm (to minimise false
  positives)?
• When and what to actuate?
• When to escalate to full scale response?
• Should fire fighters break a window for throwing
  water in
• Is it safe to enter the building?

36
CC panel

• Green
• All Clear!

• Amber
• Alert!
• User can check details

• Red
• EMERGENCY!
• User initiates actions

37
C3I process panel
38
e-Response C3I Panel
39
Grid/HPC

• Grid
• Dynamic discovery and co-ordination of
  distributed computing resources
• HPC
• High performance computing
• Parallel processing
• Issues
• On-demand access
• Priority scheduling, escalation
• Security
• Authentication authorisation

40
Grid interface

• Provision of job execution service to run fire
  simulation models on remote host
• Mechanism for staging input files of simulation
  models to remote host before job execution
• Mechanism for transferring (relevant parts of)
  output files back to client
• Support for monitoring status of jobs
• Provision of suitable security mechanisms.
• Grid links C3I layer (I-X Agent system) and HPC
  resources

41
Prototype interface

• Integrated the off-the-shelf Grid middleware
  the Globus Toolkit on the server side,
• Integrated the Java CoG Kit for the client side
  (with the latter being integrated into the Query
  Manager I-X agent)
• Such implementation ensures that the system is
  sufficiently flexible to support job submission
  from both Unix/Linux and Microsoft Windows clients

42
Technology demonstrators
43
Story

• A Fire Modeller wants to determine how good
  his/her model is in real time
• Pool fire ignited
• Fire detected by the Fire Alarm DIU
• A query is launched to predict the value of the
  smoke layer height N minutes into the future
• The query manager launches a workflow to
  implement the query
• Query manager returns the prediction
• At the predicted time a comparison is made
  between measured and predicted value

44
Architecture
45
Smoke box demo
46
Integrations achieved

• Communications and Networking
• Interactions between different components of
  architecture achieved via grid and internet
  protocols (e.g. GridFTP, JDBC, GRAM etc).
• Simulation output filtering and CC as a grid
  service
• Output from simulation filtered at the HPC site
  to reduce transmission bandwidth
• BC3I presented as a Grid service, accessible to
  all FireGrid components via the I-X interface.

47
Integrations achieved

• HPC-implemented coupled/stand-alone codes
• Smoke layer height model ported to
• IHPC (Singapore)
• ECDF (Edinburgh)
• Staging of input output files using grid file
  transfer mechanisms
• Sensor-computation integration
• Live temperature data from thermocouples used to
  determine current smoke layer height

48
Full-scale demo

• Smoke movement studies with a burner fire
• Fully flashed-over fire with real furniture

49
Full-scale demo
50
Demo architecture
51
Further work

• Explore application issues
• Resource appropriation
• Predictive capabilities
• Intelligent decision-making
• Consultations with end users
• Education
• Technology transfer

52
Spare
53
Contents

• Introduction
• Vision of FireGrid
• Technology integrations
• Dalmarnock fire tests
• Current status
• Demonstrators
• Conclusions

54
Data assimilation

• MODELLED STATE
• Simulation tool
• limited understanding
• numerical errors

• MEASURED STATE
• Sensors readings
• experimental errors
• indirect patchy

FUSION
Completeness v Cost
Completeness v Speed

• ANALYSIS STATE
• Forecasts capability
• lead time
• confidence limits

55
Technology integrations
56
CC panel
Screen shot of the BC3I showing the launching of
a query
Screen shot showing the results of the prediction
57
ARCHITECTURE FOR D7.2
Control Request
Data
58
CC (C3I)

• Command and Control systems provide an
  infrastructure for the management of information
  and resources in a complex dynamic environment.

• Command and Control system provides the glue
  that binds everything together.
• However, in order to build the right system, we
  have to understand the nature of this sort of
  decision-making