A survey and classification of incipient fault diagnosis approaches

Abstract

Incipient faults almost occur gradually at a low rate in systems and usually are unnoticeable during their early stages. If diagnostic tools or proper monitoring systems ignore them, they could not be detectable until their effects become severe and cause catastrophic damages to systems. This paper presents a survey on model-based (incipient) fault diagnosis approaches to show the significance of the incipient faults diagnosis in nonlinear closed-loop systems and, by taking a glance through data-based incipient fault diagnosis advancements, a picture of their present state of the art is also briefly discussed for completeness. Moreover, a classification of the most used state estimation filters is also provided. Consequently, recent works on incipient fault diagnosis approaches are reviewed, and an incipient fault diagnosis case study is investigated for a discrete-time nonlinear open-loop system affected by stochastic noise and disturbances. Specifically, a numerical example of a closed-loop three-tank system is considered, and simulations are accomplished, to demonstrate the inability of open-loop incipient fault diagnosis approaches in detecting incipient faults in the proposed closed-loop system.