BayesiaLab: Application Examples of Bayesian networks
Global Risk Analysis and Security Policy
Sophie Levionnois & Lionel Jouffe
Bayesia
During an accident, transport of passengers or industrial plants
can have harmful impacts on the environment or on the population. As this kind
of activities is exposed to various exogenous threats, the management of the
associated risk requires using a global approach
The example below illustrates the methodology that can be used
with BayesiaLab in the risk
management problematic. Modeling a prevention policy implies modeling the random
variables of the global system security and the "barriers" that have
been defined during the Defence in Depth analysis. (see: « Et si les risques
m’étaient comptés » - J. Valancogne, J.L. Nicolet,
G. Planchette). The following Bayesian network describes the barriers used to
prevent trains collisions.
When two barriers are added to the light system (automatic
braking system and sound signal), the accident probability is divided by 10
with a utility gain of 77%.
The Dynamic Bayesian Networks of BayesiaLab allow simulating
the temporal evolution of the security state of the system and to evaluate maintenance
policies. The graph below illustrates the evolution of the accident probabilities
and the probability evolution of the accidents with physical injuries. The maintenance
policy that is evaluated (the system is repaired every 5 time steps) allows
obtaining a probability of physical injuries lower than 0.43%
The maintenance policy can also be learned directly thanks
to the BayesiaLab's Decision nodes.
BayesiaLab then offers a complete and original risk management toolbox that
is both powerful and understandable to anyone. As it can be seen on the graph
below, the automatically learned maintenance policy reduces the accident and
physical injuries probabilities.
|
Global Risk Analysis and Security Policy
