Transformation of fault tree into Bayesian Network Methodology for Fault Diagnosis
In this article, we have shown an application of a decision support tool which is the FTBN, The combination of Bayesian Network (BN) with Fault Tree (FT) is an interesting approach to diagnose mechanical systems. Bayesian networks are tools provide robust probabilistic methods of reasoning under uncertainty, widely used in the field of reliability and fault diagnosis. While fault tree is a method of deductive analysis based on the realization of a tree that is used to identify combinations of failures, since both tools have a probabilistic aspect, the main purpose of this works is to give a methodological approach based on the transformation method of fault tree into bayesian network to model a mechanical systems, And more specifically the fault diagnosis.
Fault tree construction allows building a Bayesians network. This step allows deriving the graphical structure of the bayesian network that represents the causal relationship between the different events, and exploits the mass of existing data (experience feedback database) of the system under study.
In this paper a methodology approach is used to conduct quantification of conditionals probabilities of this Network, and performed a diagnosis on the out of balance trough modeled scenarios.The proposed methodology in our paper is centred on the presence or absence of the out of balance of the motor pump. Knowing that the source of this unbalance is caused by tows essentially events in the fault tree: Bending rotor and Break of vanes.