Niyazmand Tahereh, Izadi Iman
Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran.
ISA Trans. 2019 Jul;90:287-293. doi: 10.1016/j.isatra.2018.12.050. Epub 2019 Feb 4.
Proper monitoring of performance of an alarm system throughout its life cycle is an important factor in safety and reliability of industrial plants. Complexity and extent of modern industrial plants and poor design and management of alarm systems, have increased the importance of monitoring of alarm systems. Alarm floods, defined as a large number of alarms triggered in a short interval, is one of the problems that modern complexes are facing regularly. Many researchers have been focusing on this issue both in academia and industry. One approach to deal with alarm flood is analyzing alarms triggered in different floods and finding similar patterns. The identified patterns could help in locating the root cause of an alarm flood. In this paper a modified PrefixSpan sequential pattern recognition algorithm is used to find alarm patterns in different floods. The effectiveness of the algorithm is demonstrated with real alarm floods from a natural gas processing plant.
在整个生命周期内对报警系统的性能进行适当监测，是工业工厂安全与可靠性的一个重要因素。现代工业工厂的复杂性和规模，以及报警系统设计和管理不善，都增加了监测报警系统的重要性。报警洪泛，定义为在短时间间隔内触发大量报警，是现代工厂经常面临的问题之一。许多研究人员在学术界和工业界都在关注这个问题。处理报警洪泛的一种方法是分析不同洪泛中触发的报警，并找出相似模式。识别出的模式有助于定位报警洪泛的根本原因。本文使用一种改进的PrefixSpan序列模式识别算法来查找不同洪泛中的报警模式。通过来自天然气处理厂的实际报警洪泛证明了该算法的有效性。
