Association Rules Mining Methods For Industrial Process Alarms With Applications

The alarm is a direct means to detect abnormal conditions of industrial production processes. When an alarm is activated, the operator should find out the cause of the alarm, taking corrective actions to eliminate the abnormality quickly, and making the production process stay in the safe state. However, there is a widespread problem of excessive number of alarms in current production process, resulting in the situation that operators are unable to response in time. So, it is of great significance to reduce the amount of alarms, and highlight the important alarms.This thesis proposes an improved data mining Apriori algorithms based approach to deal with industrial process alarms, which can find out the associated alarms by data mining, before selecting one from each group to present to the operators. At the same time, the repeated alarms of the process are significantly reduced.Specifically, we firstly investigate the fundamentals of data mining, association rules mining and Apriori algorithms. Considering the availability of repeating alarms in industrial processes, an improvement of traditional Apriori algorithms is made accordingly. In response to the problem of Apriori algorithms that all data items are stored after scanning, a judgement is added before storing to detect whether the data is in the items or not. Finally, the TE process is employed as an application example. The proposed method is applied, leading to satisfactory results.