Research on Alarm Classification Methods Based on Multivariate Condition Data

Abstract

The primary focus of alarm classification is to classify alarms according to equipment condition and operator actions, which is the equivalent of classifying multivariate alarm-relevant condition data. This paper proposed an alarm classification model which aims to appropriately cluster values of alarm-relevant condition data through cluster analysis. We proposed an alarm classification model, and implemented a clustering tool for multivariate alarm-relevant condition data and a statistical tool for alarm classification. This model formulates alarm classification rules by conducting cluster analysis on condition data. These rules in turn can be jointly employed by the model to classify alarm events according to equipment working conditions and operator actions. The resulting alarm categories with a high frequency of occurrence are then displayed.