Q2. What future works have the authors mentioned in the paper "Reliability modelling with dynamic bayesian networks" ?
In a future works, in order to achieve to perform this modelling technique the authors have to define how the learning algorithms of BN can contribute to model the dynamics of the system reliability and how the parameters behaviour can be then modelled.
Q3. What are the problems considered in this paper?
The problems considered are those involving systems whose dynamics can be modelled as stochastic processes and where the decision maker’s actions influence the system behaviour.
Q4. What is the main challenge of the Extended Enterprise?
One of the main challenges of the Extended Enterprise is to dynamically maintain and optimise the quality of the services delivered by industrial objects along their life cycle.
Q5. What is the definition of a BN?
A BN is defined as a pair: G=((N, A),P), where (N,A) represents the graph; “N” is a set of nodes; “A” is a set of arcs; P represents the set of conditional probability distributions that quantify the probabilistic dependencies.
Q6. What is the purpose of this paper?
The purpose of this paper is to introduce Dynamic Bayesian Networks (DBNs) as an equivalent model to the Markov Chains (MCs) (Gertsbakh, 2000; Padhraic, 1997).
Q7. What is the simplest way to compute marginal probabilities?
Various inference algorithms can be used to compute marginal probabilities, the most classical one relying on the use of a junction tree (more explications can be found in (Jensen, 1996, pp. 76).
Q8. What is the definition of a CPT?
Defining these impacts as transition-probabilities between the states of the variable at time step k and time step k+1, these transition-probabilities lead to define CPTs relative to inter-time slices, equivalent to CPT defined in the previous section (eq. (5)).
Q9. What is the proposed method for the modelling of the temporal evolution of complex systems?
The proposed method, based on the Dynamic Bayesian Networks theory, easily allows constructing DBN structures for the modelling of the temporal evolution of complex systems.
Q10. What is the method of a low complexity component model?
This method allows to model the reliability of the system assuming the hypothesis of independence of the events (failures) affecting the entities.
Q11. What is the dependency between the failure modes?
The propagation through the Bayesian model allows taking into account the dependency between the failure modes for the computation of the system reliability.
Q12. Why is the MC model different from the DBN model?
the differences are due to the approximation made in the Markov model that assumes that simultaneous failures can not occurred, this hypothesis being not assumed in the DBN model.
Q13. What is the probability distribution of kiki nn?
MCkiki nn PP =+ )( ,1, (6)Starting from an observed situation at time step k=0, the probability distribution inkx over in states is computedby the DBN inference.
Q14. What is the methodology proposed in this paper?
The methodology proposed in this paper is an original formalisation of a system reliability model (section 4) by means of DBNs (section 3).
Q15. What is the probability of the system being available?
As it is shown by the figure, 25 states 1s … 25s arenecessary to model this system: states 1s to 11s are states for which the system is available in spite of thedegradation due to some failures; states 12s to 25s correspond to states where the system is unavailable due to the combination of failures.