Fault Diagnosis in an Industrial Process Using Bayesian Networks: Application of the Junction Tree Algorithm

摘要

In this paper we present a Bayesian Network for fault diagnosis used in an industrial tanks system. We obtain the Bayesian Network first and later based on this, we build a defined structure as Junction Tree. This tree is where we spread the probabilities with the algorithm known as LAZY-AR (also Junction Tree). Nowadays the state of the art in inference algorithms in Bayesian Networks is the Junction Tree algorithm. We prove empirically through a case study as the Junction Tree algorithm has better performance with regard to the traditional algorithms as the Polytree.