Model-Based Prognostic Methods Applied to Physical Dynamic Systems

In several engineering fields, especially in the recent years, the development of adequate diagnostic/prognostic methodologies able to provide a timely and reliable evaluation of the health status of a given system has become a strategic task in order to guarantee suitable levels of reliability, robustness and logistic availability. In particular, at this moment are in the spotlight some prognostic approaches that, on the basis of some representative parameters (measured directly or indirectly), are able to evaluate the health status of a physical system with a suitable (and quantifiable) level of accuracy and robustness; it must be noted that, especially in recent years, these methods are increasingly meeting interest and application in many technical fields and, nowadays, they represent an important task in various scientific disciplines. If considered failures are characterized to progressive evolutions, the health status of a given dynamic system (e.g. environmental, mechatronic, structural, etc.) and the related failure modes can be identified and quantified by means of different approaches widely described in the literature. In the last ten years more and more researchers studied and proposed new strategies aimed to design prognostic algorithms able to identify precursors of the progressive failures affecting a system: in fact, when a degradation pattern is correctly identified, it is possible to trigger an early warning and, if necessary, activate corrective actions (i.e. proper remedial or maintenance tasks, replacement of the damaged components, etc.). Typically these methods are strictly technology-oriented: they can result extremely effective for some specific applications whereas may fail for other purposes and technologies; therefore, it is necessary to "design" and calibrate the prognostic algorithm as a function of the considered problem, taking into account 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 an overview of the most common model-based diagnostic/prognostic strategies (derived from aerospace systems field), putting in evidence their applicability, strengths and eventual shortcomings.