Univariate Model-Based Deadband Alarm Design for Nonlinear Processes

Abstract

Alarm design is an essential industrial problem with significant implications for safety and performance. Standard alarm design algorithms are based on the assumption that the data are uncorrelated and stationary. In this Article, we relax these assumptions and develop a novel approach to design alarms for processes modeled as a stochastic nonlinear time-series model. In particular, we develop an algorithm to design deadband alarms by minimizing the false and missed alarm rates. The resulting algorithm is illustrated through extensive simulations on a reactor system.