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

Complex industrial processes such as nuclear power plants, chemical plants and petroleum refineries are usually equipped with alarm systems capable of monitoring thousands of process variables and generating tens of thousands of alarms which are used as mechanisms for alerting operators to take actions to alleviate or prevent an abnormal situation. Overalarming and a lack of configuration management practices have often led to the degradation of these alarm systems, resulting in operational problems such as the Three-Mile Island accident. In order to aid alarm rationalization, this paper proposed an approach that incorporates a context-based segmentation approach with a data mining technique to find a set of correlated alarms from historical alarm event logs. Before the set of extracted results from this automated technique are used they can be evaluated by a process engineer with process understanding. The proposed approach is evaluated initially using simulation data from a Vinyl Acetate model. The approach is cost effective as any manual alarm analysis of the event logs for identifying primary and consequential alarms could be very time and labour intensive.

The productivity of chemical plants and petroleum refineries depends on the performance of alarm systems. Alarm history collected from distributed control systems (DCS) provides useful information about past plant alarm system performance. However, the discovery of patterns and relationships from such data can be very difficult and costly. Due to various factors such as a high volume of alarm data (especially during plant upsets), huge amounts of nuisance alarms, and very large numbers of individual alarm tags, manual identification and analysis of alarm logs is usually a labor-intensive and time-consuming task. This chapter describes a data mining approach for analyzing alarm logs in a chemical plant. The main idea of the approach is to investigate dependencies between alarms effectively by considering the temporal context and time intervals between different alarm types, and then employing a data mining technique capable of discovering patterns associated with these time intervals. A prototype has been implemented to allow an active exploration of the alarm grouping data space relevant to the tags of interest.

A typical chemical alarm database is characterized by a large search space with skewed frequency distribution. Thus in practice, discovery of alarm patterns and interesting associations from such data can be exceptionally difficult and costly. To overcome this problem we propose a data-driven approach to optimally derive the pruning thresholds which are relevant to the temporal data context of the particular tag of interest