Diagnostic/prognostics strategies applied to physical dynamic systems: A critical analysis of several model-based fault identification methods

The development of adequate diagnostic/prognostic methodologies, suitable to provide a timely and reliable evaluation of the health status of a given system on the basis of some representative parameters (measured in a direct or indirect way), is fundamentally started in engineering fields, but, especially in recent years, it is encountering more and more interest and application in many technical fields and nowadays it represents an important task in various scientific disciplines. The health status of a given dynamic system (e.g. environmental, mechatronic, structural, etc.) and the eventual incipient failures that concern it, especially if related to progressive evolutions, can be identified and quantified by means of different approaches widely described in the literature. It must be noted that, particularly in recent years, there has been a strong impulse in the development of strategies aimed to design prognostic algorithms able to identify precursors of the progressive failures affecting a system: in fact, if it is correctly identified the degradation pattern, an early warning can be triggered, leading to proper corrective actions (i.e. proper remedial or maintenance tasks, replacement of the damaged components, etc.). Since these algorithms are strictly technology-oriented, they can show great effectiveness for some specific applications, while they may fail for other applications and technologies: therefore, it is necessary to properly conceive the specific prognostic method as a function of several parameters such as the given (dynamic) system, the available sensors (physical or virtual), the considered progressive failures and the related boundary conditions. This work proposes a critical comparison between several diagnostic/prognostic strategies in order to put in evidence their strengths and the eventual shortcomings.