Performance analysis for stochastic anomaly detectors with on/off-delay timers☆

A delay timer is a simple yet effective technique that has found widespread application in alarm management to reduce nuisance alarms. However, in most cases, the performance of an alarm system with delay timers can only be empirically verified after their application in real industrial processes. This post-deployment verification process is often costly, demanding substantial financial investment and time. For stochastic anomaly detectors that monitor the status of temporally correlated process variables, this paper provides explicit expressions for alarm performance in terms of false alarm rates and fault detection rates with different applications of on/off-delay timers. When both on and off-delay timers are used together, the performance metrics can be ascertained by analyzing the steadystate distribution of a finite-state Markov chain. The proposed methods facilitate pre-implementation evaluation of delay timers, aiming to significantly reduce the validation cost. The effectiveness of the theoretical results is confirmed with numerical examples. (c) 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.