Notions of Diagnosability for Timed Failure Propagation Graphs

Abstract

Timed failure propagation graphs (TFPG) are causal models that capture the temporal aspects of failure propagation in dynamic systems. This paper presents several notions of diagnosability for timed failure propagation models. Diagnosability characteristics of TFPG models are defined based on three metrics; failure detectability, distinguishability, and predictability. The paper identifies the modeling and run-time parameters of the diagnosability metrics. The application of diagnosability analysis to the problem of alarm allocation is discussed.