Model-based fault diagnosis and prognosis of dynamic systems: a review

In maintenance of engineering systems, condition monitoring, fault diagnosis and fault prognosis constitute some of the principal tasks. With the increase of the number of machines within processing plants and their operational complexities, many engineers and researchers have started looking for automated solutions for these tasks. In most of the proposed solutions, these dynamic systems are modelled using tools like automata, Petri nets, bond graphs and Bayesian networks to diagnose and predict faults in those systems. This paper reviews these graphical model-based techniques related to fault diagnosis and prognosis and give suggestions for future research directions identifying research gaps in the field.