Criteria-based alarm flood pattern recognition using historical data from automated production systems (aPS) |
| |
Institution: | Institute of Automation and Information Systems, Technische Universität München, Boltzmannstrasse 15, D-85748 Garching near Munich, Germany |
| |
Abstract: | The operation of industrial automated production systems (aPS) usually requires human operators that observe and, if necessary, intervene to keep aPS in steady operation. Inside the distributed control system (DCS) of an aPS, notifications are generated by an alarm management system (AMS) and visualized, informing operators about critical aPS situations, e.g. faults of a device. Since a huge number of notifications are usually configured inside the AMS, operators nowadays often face the problem of receiving more notifications than they can physically address. This paper proposes an approach, which allows automatic identification of alarm floods by using criteria-based search strategies. In order to address the problem statement, four hypotheses are stated. To evaluate the proposed algorithm regarding its ability to identify causally dependent notifications, historical notification logs of real industrial aPS are analyzed. For this purpose, notification logs of eight existing industrial aPS as well as the assessment of industrial experts are taken into account. |
| |
Keywords: | Alarm analysis Sequence detection Causality analysis Sequence pattern recognition Frequent pattern recognition |
本文献已被 ScienceDirect 等数据库收录! |
|