Optimal alarm trippoints and timers for avoiding false alarms *

Process disturbances, noise and unreasonable settings of alarm trippoints usually generate false alarms. A novel approach is hereby proposed to calculate optimal values of alarm trippoints and timers based on statistics of alarm durations to avoid too many false alarms. Most conventional approaches separately or simultaneously calculate alarm trippoints and timers based on statistics of process variables that are strictly pressumed to be independent and identically distributed (IID). However, by exploiting the statistics of alarm durations the applicability of the proposed approach is extended to process variables that are non-IID. Bayesian estimation is used to reliably estimate false alarm rates which is a major technical challenge. Numerical and industrial examples are given to demonstrate the proposed approach.