Weakness Monitors for Fail-Aware Systems

Abstract

Fail-awareness is the ability of a system to detect an upcoming failure before it actually happens. In this paper, we propose a weakness monitoring approach for observing a complex system during its operation, identifying possible degradation of its behavior, and finally raising an alarm in case of an estimated upcoming failure before the system actually goes out of its specification. Our procedure uses online linear regression to monitor trends over time – it is used to optimize the system service. We evaluate our approach on three case studies from the automotive and avionics domains.