sose2012 chaze presentation

20
Integration of a Bayesian network for response planning in a maritime piracy risk management system 1/20 Integration of a Bayesian network for response planning in a maritime piracy risk management system Genoa – Italy – July 16-19, 2012 Centre de recherche sur les Risques et les Crises - Xavier CHAZE - Amal BOUEJLA, Aldo NAPOLI, Franck GUARNIERI Mines ParisTech – CRC

Upload: xchaze

Post on 05-Dec-2014

449 views

Category:

Technology


0 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 1/20

Integration of a Bayesian network for response planning in a maritime piracy risk management system

Genoa – Italy – July 16-19, 2012 Centre de recherche sur les Risques et les Crises

- Xavier CHAZE -Amal BOUEJLA, Aldo NAPOLI, Franck GUARNIERI

Mines ParisTech – CRC

Page 2: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 2/24

Develop research, teaching activities, methods and tools.Contribute to strengthen organisations and territories

against disturbances.

47 people: 8 Researchers, 25 PhD Students,

3 Engineers + support

Science for engineers, Management science, Psychology,Computer science, Geography,Law.

Centre for research on

Risks and

Crisis

Research & training

Business creation

Industrial partnerships

Strong industrial contacts

Research centre (since 2008)

(Team created in 1998)

French engineering school

(since 1783)

French research association

(since 1967)

Page 3: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 3/20

Summary

Introduction Problem definition Method Results Conclusion

Problem definition

Issues and context

Operational needs

Contribution of the SARGOS project

Method

The Bayesian networks

Methodological approach

The SARGOS Bayesian network

Results

Attack scenario case study

Integration of the Bayesian Network into the SARGOS system

Page 4: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 4/20

Introduction Problem definition Method Results Conclusion

Page 5: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 5/20

Source : IFP Énergies Nouvelles

Introduction Problem definition Method Results Conclusion

Issues and context Operational needs Contribution of the SARGOS project

Mexico Gulf

Guinea Gulf

Brasil

• Offshore oil industry represents:

� 30% of the oil world production

� 27% of the gas world production

• Piracy cost is estimated between 7 and 12 billiards of US dollars per year. It is mainly due to:

� Ransom payments

� Insurance premiums

� Cost of trials and judiciary pursuits

� Installation of security equipment

• Political issues are also important. Serenity of a whole region can be disturbed:

� Conflicts between nations when the rig is located in a

one country while the company operating the platform islocated in another.

� The legal status of oil rig, the heterogeneity of applicable

regulations and the limits of laws and conventions

etablished for the fight against piracy.

• Offshore oil production over 1000 meters from coasts:

• Geographical expansion of pirate attacks: 2005-2011

Page 6: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 6/20

The solution is to develop a system that can manage the safety of oil fields and provideboth suitable protection and effective crisis management.

Threattreatmentprocess

Existing anti-piracytools

Benefits DIsadvantages

Detection of the threat

RADAR systems

Detection of medium and large cooperative vessels

• Poor performance against small targets in a sea clutter.• Relatively slow to scan a wide field.

Optronics surveillance system

Long-range detectionof small targets

• Disturbed by problems of solar reflectanceof the sea.• Sensitive to the meteorological conditions.

Responseagainst the

threat

Automatic Identification System (AIS)

Automatic exchange of messages

• Messages exchange in a restrictedgeographical area.

Surety and securityvessel

Intervention towardsattackers

• Uncertainty of the intervention dependingon the distance between threat and vessel.• Imbalance of arms between attackers and security officers.

Introduction Problem definition Method Results Conclusion

Issues and context Operational needs Contribution of the SARGOS project

Page 7: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 7/20

• The SARGOS system (Graduated OffShore Response Alert System) aims to design and develop a comprehensive system that takes into account the whole threat treatment process:

� Detection of a potential threat

� Edition of an alert report that lists the significant parameters of threat and target

� Definition of the response by the planning module of reactions

� Formalization and implementation of the reaction by the publication of a response plan

Consortium

ApprovementsFundings

Introduction Problem definition Method Results Conclusion

Issues and context Operational needs Contribution of the SARGOS project

Page 8: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 8/20

Introduction Problem definition Method Results Conclusion

Issues and context Operational needs Contribution of the SARGOS project

• Functional outline of the SARGOS system

The SARGOS system responds to an alert report with a response plan, which is the result of an intelligent analysis of the alert report.

Page 9: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 9/20

• The problem of the response planning against a threat to offshore oil fieldsexhibits strong constraints:

� Coordination between the different available counter-attack devices on the field

� Real-time gradation of the threat and the response adaptation depending on its increase

� Inherent uncertainty of threat parameters

� Automatization of the whole process

Choice of using the Bayesian networks

• A Bayesian network is a model that represents knowledge, and makes it possible to calculate conditional probabilities and provide solutions to various types of problems

Interest

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Page 10: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 10/20

• Bayesian networks are based on Thomas Bayes theorem (1702-1761) :

• Supposed that you live in London and according to your experience, during winter, it rains50% of the time and it is cloudy 80% of the time. You know, of course, that if it rains, so it isalso cloudy.

• What is the chance of rain knowing that there are clouds ?

• Where: Pl : it rains

N : it is cloudy

Thus, 62.5% of the time in London during the winter, if it is cloudy, then it is rainy

Definition & example

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Page 11: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 11/20

• The construction proceeds in 4 steps:

� Define variables of the problem (nodes)

� Set the modalities which describe all possible values for each variable

� Define the connections of the system (links between nodes)

� Specify the conditional probabilities resulting by the created links

How to construct a bayesian network?

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Page 12: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 12/20

The application to SARGOS

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Realisation of a Bayesian networkRealisation of a Bayesian network

Datamining learning

Database of the International Maritime Organisation

(IMO)

Brainstorming learning

Expert knowledge from the maritime and safety domains

Page 13: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 13/20

• Founded the 6th of march 1948, the International Maritime Organization is a specialized institution of United Nations.

• On the 15th of July 2011, the database contained 5502 recordings of piracyattacks or armed robbery.

• The database contains more information about:

� The name and the type of the attacked target,

� Longitude and latitude of the attack location,

� A textual description of the sequence of events

� …

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

The Bayesian network constructed from IMO data

Page 14: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 14/20

• The Bayesian network created from the IMO data made it possible to define:� The main tools and protection measures used by a crew, and their effectiveness

� The probability distributions of using the reactions

These results will be integrated into the Bayesian network

constructed from the marine community expert knowledge

• The construction of the Bayesian network was based on the expert knowledge duringmany brainstorming sessions.

• The prototype was tested and improved by an iterative process to refine the conditionalprobabilities of the nodes.

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Expert bayesian networkarchitecture

Page 15: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 15/20

SARGOS Bayesian network

Introduction Problem definition Method Results Conclusion

The Bayesian networks Methodological approach The SARGOS Bayesian network

Page 16: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 16/20

Introduction Problem definition Results Conclusion

time

Attack by an unknown vesselagainst a Floating Production Storage

and Offloading unit (FPSO).

T1

The high-manoeuvrability vessel is now identifiedas hostile. The threat is located less than 300

seconds and 50 metres from the target.

T1+t

Diagnosisimprovement

Increase in the leveloverall danger

Responsegraduation and adaptation

Attack scenario case study Integration of the Bayesian Network into the SARGOS system

Method

Page 17: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 17/20

• The SARGOS reactions planning results in the generation of a response planning report from the intelligent processing of the last issued alert report.

• The response plan gathers the necessary information for the physical and chronological execution of the reaction

• The interface between the bayesian module and the SARGOS system is completed thanks to JAVA scripts:

� Input

- Identification of a new alert report (XML file)

- Extraction of useful information

� Execution of bayesian module (API BayesiaEngine)

-Supply of the Bayesian network (set the observations of source nodes)

� Output

- Export of modalities and resulting probabilities of the Bayesian network

- Generation of the graduated suitable response planning report (XML file)

Introduction Problem definition Results Conclusion

Attack scenario case study Integration of the Bayesian Network into the SARGOS system

Method

PlanningReport(XML)

AlertReport(XML)

Bayesianmodule

Page 18: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 18/20

• Human-Computer interface of the SARGOS system: once the countermeasures have been selected, they are displayed in the response plan in a specific order.

Introduction Problem definition Results Conclusion

Attack scenario case study Integration of the Bayesian Network into the SARGOS system

Method

Page 19: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 19/20

• The use of a Bayesian network for the planning of the response is a major asset of the SARGOS system as this network can:

� Define a graduated response adaptated to the identified threat

� Take into account the uncertainty of some parameters

� Manage the real-time situation evolution

• Finally, the network is able to integrate feedback from attacks that has previously been used to administer and can therefore evolve. Consequently the planning module can be modified and improved iteratively.

Introduction Problem definition ConclusionMethod Results

Page 20: Sose2012 chaze presentation

Integration of a Bayesian network for response planning in a maritime piracy risk management system 20/20

Thank you for your attention

Questions ?

[email protected]