Dynamic Alarm Design Using Bayesian Theory

摘要

Alarms are essential to be configured in every chemical plant and alarm systems play a critical role in industrial plant operations. However, poor alarm management has become one of the leading causes of many major industry incidents, such as the one in Texas City. Although alarm rationalization as a useful method can reduce average number of alarms, alarm floods are still not controlled. Generally, alarm floods occur upon a change of states in the process. In order to eliminate alarm floods, especially in the cases of the transition between normal states, dynamic alarm design based on Bayesian analysis is proposed. The Bayesian analysis method utilizes prior information and real-time process data to calculate the variance of variables. Then alarm limits of variables are obtained with the corresponding variance based on some rule. A two-tank system is here to show its performance.