Analysis of Alarm Sequences in a Chemical Plant

Abstract

Oil and gas industries need secure and cost-effective alarm systems to meet safety requirements and to avoid problems that lead to plant shutdowns, production losses, accidents and associated lawsuit costs. Although most current distributed control systems (DCS) collect and archive alarm event logs, the extensive quantity and complexity of such data make identification of the problem a very labour-intensive and time-consuming task. This paper proposes a data mining approach that is designed to support alarm rationalization by discovering correlated sets of alarm tags. The proposed approach was initially evaluated using simulation data from a Vinyl Acetate model. Experimental results show that our novel approach, using an event segmentation and data filtering strategy based on a cross-effect test is significant because of its practicality. It has the potential to perform meaningful and efficient extraction of alarm patterns from a sequence of alarm events

A method for pattern mining in multiple alarm flood sequences

“…The difficulties encountered in manual analysis of alarm floods could be overcome by resorting to data mining techniques. In (Kordic et al, 2008), a context based segmentation was carried out and alarm messages during the segmented periods were filtered to be the patterns. The method requires to pinpoint a target tag beforehand to set the starting point of segmentation.…”

T-Patterns Revisited: Mining for Temporal Patterns in Sensor Data

“…The application of sequential pattern recognition in sensor networks includes long-term environmental monitoring [ 4 ], alarm event detection and propagation [ 5 , 6 ], localization and tracking of objects [ 7 – 9 ], recognition of human behaviour and interactions [ 10 – 16 ], and intelligent resource management [ 17 ] among others. The temporal data are in the form of a sequence of events, derived from a number of sensors that may have different modalities.…”

Patterns Relevant to the Temporal Data-Context of an Alarm of Interest

“…In order to correctly mine the alarm data using the proposed approach, a group of associated alarms should all return well before any new activation of the alarms within the group. Further research (Kordic et al, 2008) has been carried out to investigate this issue.…”

“…Figure 6 shows rules with respect to the first ten alarm tags. In terms of checking whether the group of associated alarms is correct, the results sets associated with fault TAG 1 (loss of %O2) can be checked against the Vinyl Acetate process flowsheet chart illustrated in Kordic et al (2008). Based on the Vinyl Acetate flowsheet analysis and the alarm status display, the loss of Oxygen feed (TAG 1) in the reactor could significantly affect temperature change in the Reactor (TAG 7), Heat Exchanger (TAG 6), Separator (TAG 9), and Vaporizer Level (TAG 4).…”