A dynamic alarm management strategy for chemical process transitions

Abstract

Chemical processes frequently operate upon a multitude of steady states and transitions between these states are inevitable. Unfortunately, transitions are exactly where alarm floods often occur. Alarm floods cause critical alarms overwhelmed and thus increase the probability of larger safety issues. Existing techniques for the design of alarm systems mostly focus on one steady state of operation and yet cannot effectively deal with alarm floods during transitions. In this paper, a dynamic alarm management strategy is proposed for controlling alarm floods during transitions of chemical processes. In this strategy, the artificial immune system-based fault diagnosis (AISFD) method and a Bayesian estimation based dynamic alarm management (BEDAM) method are integrated. During transitions, dynamic alarm limits obtained by the BEDAM method can control alarm floods. However, if a process fault occurs during transitions, a flood of alarms could still be yielded. To generate useful alarms in fault situations, an artificial immune system based on dynamic time warping (DTW) is used for fault detection and diagnosis. Finally, in case studies, the dynamic alarm management strategy is applied to the startup stage and a throughput change transition in a pilot-scale distillation column.