Alarm Deadband Design Based on Maximum Amplitude Deviations and Bayesian Estimation

Abstract

Alarm deadbands are commonly used in modern computerized monitoring systems to avoid false alarms. A new method is proposed to design an alarm deadband, to achieve a desired percentage of removed false alarms with respect to the counterparts without using an alarm deadband. An optimal value of the deadband width is determined based on the cumulative probability of maximum amplitude deviations between an alarm threshold and values of a process variable in the alarm state. The Bayesian estimation approach is used to evaluate whether the estimated cumulative probability is reliable and whether the designed deadband width is trustworthy. Existing methods are either limited to independently and identically distributed (IID) process variables or require complicated techniques of Kalman filters or particle filters for non-IID ones. By contrast, the proposed method is simple for implementation and has no discriminations to IID and non-IID process variables. Numerical and industrial examples are provided to support the proposed method.