Process Discovery of Operator Actions in Response to Alarms in Industrial Facilities

• Author / Creator

  Li,David S

• The propagation of alarms which results in multiple alarm annunciations can lead to what are known as alarm floods, where operators are unable to resolve all alarms at the same time. It is highly unlikely for operators to read through an alarm response manual during an alarm flood that requires immediate attention. In general, incorrect operator actions are not due to a lack of experience, but rather caused by the overload of incoming alarms. If theses alarms are not addressed in a timely manner, this abnormal operation can lead to unexpected plant shut-down, economic loss and even loss of life. As a result, there is a high demand by the industry to develop tools to assist operators with decision making. Alarm and event (A&E) logs contain historical information about the process and can be used to capture the correct operator actions to clear each unique alarm. On the contrary, A&E logs can also be used to capture actions that should not be taken if an alarm is not cleared after several attempts. This thesis attempts to adopt process mining algorithms into the field of alarm management. A framework for the process discovery of operator actions in response to both univariate and multivariate alarms will be developed. Two algorithms are studied in this thesis and their effectiveness is tested against industrial alarm and event logs. The first algorithm attempts to pre-process commonly found imperfection patterns in A&E logs and assign case identification based on temporal ordering rules. The second algorithm attempts to mine multivariate relationships between alarm events as well as operator actions. The mined results are visualized using color map representation as well as a new form of graphical representation specifically designed for A&E logs.

• Subjects / Keywords

  • Process Discovery
  • Porcess Mining

• Graduation date

  Spring 2018

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R39Z90T1Z

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.