Residuals-based fault diagnosis of industrial automation systems using timed and untimed Interpreted Petri nets

The problem of detecting and isolating a fault using a discrete event model has received a lot of attention over the last two decades. Indeed, industrial automated systems can be usually modelled by a discrete event system. The problem is particularly significant if a fault-free model is used. Recently, residuals, well-known in continuous time systems context, have been proposed to obtain set of candidate faults from real time observations for discrete event system also. They have been obtained using finite state automata. In this paper, inspired by these works, residuals are formulated using Petri net models, both timed and un-timed ones. The computation of residuals for net models requires some efforts to update and predict the state, especially in timed models, but it returns to be highly efficient and scalable thanks to the local state representation and intrinsic distributed nature of the these models. A practical example, consisting of a plant simulated by using a 3D environment interfaced to a Programmable Logic Controller to simulate/emulate the closed-loop behaviour, is used to illustrate the results of the paper.

Automated summary

This paper focuses on detecting and isolating faults using a discrete event model, introducing a new approach of obtaining candidate faults in discrete event systems using residual methods, and exploring the formulation of residuals using Petri net models. The effectiveness of these methods is illustrated through a practical example.