The advance of distributed control systems (DCSs) in the chemical industry has made it possible to install many alarms cheaply and easily. While most alarms help operators detect an abnormality and identify its cause, some are unnecessary. A poor alarm system might cause alarm floods and nuisance alarms, which reduces the ability of operators to cope with plant abnormalities because critical alarms are buried under a lot of unnecessary ones (EEMUA, 2007).
Nishiguchi and Takai (2010) developed an event correlation analysis method for reducing unnecessary alarms. Their analysis is a knowledge extraction method that detects statistical similarities among discrete events of alarms and operations. The method uses operation data from the plant to quantify the degree of similarity with its time lag between two events by evaluating the cross correlation function. By grouping correlated alarms and operations in accordance with the degree of similarities, the following types of nuisance alarms and operations can be found.
(1) Sequential alarms: When a group contains multiple alarms that occur in sequence, changing the alarm settings might effectively reduce sequential alarms.
(2) Routine operations: When many operations are included in a group and operations in the same group appear frequently in the event log data, they might be routine operations. These alarms and operations can be reduced by automating routine operations using a programmable logic controller.
(3) Unnecessary alarms: Alarms might be unnecessary when operations are not carried out in response to them.
The event correlation analysis method was applied to the operation data of an ethylene plant operated by Idemitsu Kosan in Japan. Unnecessary alarms and operations were accurately identified within a large amount of event log data by using the method (Higuchi et al., 2010). However, the method sometimes failed to detect similarities between two physically related events when the variance in time lag between them was too big. Therefore, in this paper, we propose a new event correlation analysis method that is able to correctly identify similarities between two physically related events.
The proposed method first converts the plant event log data, which consists of occurrence times and tag names of alarms and operations, into sequential binary data using a certain size of time widow. Then, similarities between all combinations of any two events in the plant event log data are evaluated (Nishiguchi and Takai, 2010). If high similarity between two events is not detected, the time window size is doubled and the log data of two events are reconverted into sequential binary data using the new time window size. The expansion of the time window size and recalculation of the similarity continues until either high similarity is detected or the time window size becomes larger than the maximum pre-determined size.
We applied the proposed method to simulation data of an azeotropic distillation column. Results showed it was able to correctly identify similarities between two physically related events, even when the conventional method using a constant time window size failed due to the large variance in time lag. Using the proposed method, we can effectively identify unnecessary alarms and operations within a large amount of event data and reduce the number of unnecessary alarms and operations in chemical plants.
References
(1) Engineering and Equipment Materials Users’ Association (EEMUA), Alarm Systems – A Guide to Design, Management and Procurement Publication No. 191 2nd Edition, EEMUA, London (2007)
(2) Junya Nishiguchi and Tsutomu Takai, IPL2&3 Performance Improvement Method for Process Safety Using the Event Correlation Analysis, Computers & Chemical Engineering, 34, pp. 2007-2013 (2010)
(3) Fumitaka Higuchi, Ichizo Yamamoto, Tsutomu Takai, Masaru Noda, Hirokazu Nishitani, Rationalization of Plant Alarm System using Event Correlation Analysis, Proceedings of PSE Asia 2010, pp. 1395-1401, Singapore (2010)
See more of this Group/Topical: Computing and Systems Technology Division