Performance Monitoring of Industrial Plant Alarm Systems Using Event Correlation Analysis

摘要

Alarm System is essential to ensure the plant safety and effective operation, and the management efforts aiming at maintaining and improving alarm systems have recently intensified in process industries. Alarm management for existing plants is basically CAPDo approach that begins by evaluating the alarm system performance, and the result is quite important in order to find issues to be maintained and improved effectively. Conventional methods which evaluate alarm rates, alarm and event distributions, standing alarm times, etc. in quantity, remain a long way from an effective evaluation of alarm systems performance, because they do not evaluate each alarm as a signal requiring operator attention. Engineering Equipment & Materials Users' Association (EEMUA,2007) says that every alarm presented to the operator should be useful and relevant to the operator. Thus, the relationship between alarm and the operator response is thought of as a new key performance indicator (KPI) of alarm system performance.Takai et al. (2010) proposed an evaluation method of the relationship between them using operator questionnaire. Questionnaire is reasonable for working-level at- plant operation but questionnaire-based evaluation is often subjective, and may contain bias arising from the format of the questionnaires or from individual respondents. In the paper, we propose a new KPI of evaluation alarm system performance and the calculation method by using an event correlation analysis (Nishiguchi and Takai, 2010) which is a data mining method to quantify the degree of similarity and time lag between two events by cross correlation function, from event log data composed of discrete alarms and operator actions with the time they occur. In the event correlation analysis, events pairs separated by consistent time intervals are considered to be related, since the length of the time lags is determined by factors such as process dynamics and operator reaction time. A similarity measure between all event pairs is calculated from the event log data along with the probability distribution of correlation regarding independent event pairs. For similarities and intervals in all the combinations between event pairs are calculated, and groups with highly related events are identified using pair-wise similarities by the hierarchical clustering method. Then, each alarm is assayed if the based alarm has the corrective actions after the alarm occurs in the group. The relevant alarm rate is defined as the ratio of number of relevant alarms in all alarms. The effectiveness of the proposed method was validated with actual plant event data and simulation results.