A local alignment approach to similarity analysis of industrial alarm flood sequences

Similar alarm sequence alignment algorithms have been used to find similar alarm floods in the historical database for the prediction and prevention of alarm floods. However, the existing modified Smith–Waterman (SW) algorithm has a high computation complexity, preventing its online applications within a tolerable computation time period. This paper proposes a new local alignment algorithm, based on the basic local alignment search tool (BLAST). The novelty of the proposed algorithm is three-fold. First, a priority-based similarity scoring strategy makes the proposed algorithm more sensitive to alarms having higher alarm priorities. Second, a set-based pre-matching mechanism avoids unnecessary computations by excluding all irrelevant alarm floods and alarm tags. Third, the seeding and extending steps of the conventional BLAST are adapted for alarm floods, which reduce the searching space significantly. Owing to the novelties, the proposed algorithm is much faster in computation and provides a higher alignment accuracy than the SW algorithm. The efficiency of the proposed algorithm is demonstrated by industrial case studies based on the historical alarm floods from an oil conversion plant.