overview on diagnosability and prognosability for system monitoring

Abstract

The complexification of systems has brought the emergence of a new field of study: system health monitoring. This field is deemed necessary because it improves system availability and it avoids unnecessary maintenance costs. System health monitoring is performed through diagnosis and prognosis methods. Diagnosis consists in detecting and identifying faults that may lead to system failures. Prognosis is related to the prediction of the system Remaining Useful Life (RUL) that corresponds to the remaining time until the system failure. This paper aims at giving an overview on the properties related to diagnosis and prognosis on different types of systems. We will focus on the diagnosability and prognosability properties. This paper will first briefly present the different types of systems of interest for the system health monitoring community. We will consider Discrete Event Systems (DES), Continuous Systems (CS), Hybrid Systems (HS) or Heterogeneous Systems (HtS). The rest of this paper will present the definitions given in the literature for the concepts of diagnosability and prognosability. The similarities and differences in these definitions for the different types of systems will be highlighted. Some metrics associated with the prognosability property will also be discussed.