Methodology and Application of Pattern Mining in Multiple Alarm Flood Sequences

Alarm floods have always been serious hazards in industrial process monitoring since they overwhelm operators with large amount of alarm messages raised within a short period of time. In this paper, we propose an algorithm to find an optimal alignment of multiple alarm flood sequences so that based on this alignment an alarm sequence pattern can be easily found. The pattern could reveal the correlation between alarm messages in the alarm floods, which cannot be obtained by applying other alarm management techniques such as delay timers and dead-bands. The pattern could also help find the root cause, locate badly designed part in alarm systems, and predict incoming alarm floods. A dataset from an actual chemical plant has been used to test the effectiveness of the proposed algorithm.