Alarm flood is a serious hazard for industrial processes, alarm management techniques such as delay timers and dead-bands often are incapable of suppressing alarm floods because of the existence of consequence alarms. However, this problem could be handled by alarm flood analysis, which could help find the root cause, locate badly designed part in alarm systems and predict incoming alarm floods. In this paper we propose a method for pattern mining in multiple alarm flood sequences by extending a time-stamp adapted Smith-Waterman algorithm to the case with multiple sequences, making up one of the missing steps in alarm flood analysis. The technique involves the following new elements: similarity scoring functions, a dynamic programming equation, a back tracking procedure, and an alignment generation method. A dataset from an actual petrochemical plant has been used to test the effectiveness of the proposed algorithm. (C) 2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.