Prognosis is increasingly used in industrial applications, particularly for predictive maintenance. Diagnosis and prognosis are highly correlated: diagnosis determines the set of faulty components explaining the observed malfunctions, while prognosis determines the future state of the components based on aging/degradation models of the monitored system. These techniques can be used to determine the state of health and assess the residual lifetime of system components. They can use physics-based models, statistical data analysis, learning and/or estimation methods. DISCO studies the coupling of system diagnostics and prognostics with a view to refining possible strategies integrating corrective and predictive maintenance.

FOCUS

Diagnostic/prognostic prerequisite model

The team has developed a generic model of equipment aging that establishes a set of prerequisites for setting up a prognostic process on a complex, heterogeneous system, in conjunction with a process for diagnosing degradation and stress factors. This type of model has been applied to a stator in permanent magnet synchronous machines (landing gear) (Safran CIFRE collaboration).

Interval analysis and computation for diagnosis/prognosis

DISCO has developed a method based on the theory of interval calculus to solve the problem of equipment health management, which has been applied to the health management of a control valve (A320 Neo) for which a demonstrator has been implemented (Prognospice).

Formalism for particulate hybrid Petri nets

DISCO is developing a formalism which aims to model the dynamics of a hybrid system, both from a functional and behavioral point of view, and from the point of view of degradation and future aging. This unique formalism can be used as the basis for a diagnostic function (fault diagnosis, failure diagnosis) consistent with a prognostic function.