Thesis Defence: Alarm Prediction for Manufacturing Process Automation Using Similarity-Based Scenario Matching
October 22 at 9:00 am - 1:00 pm
Sahar Sowdagar, supervised by Dr. Ahmad W. Al-Dabbagh, will defend their thesis titled “Alarm Prediction for Manufacturing Process Automation Using Similarity-Based Scenario Matching” in partial fulfillment of the requirements for the degree of Master of Applied Science in Electrical Engineering.
An abstract for Sahar Sowdagar’s thesis is included below.
Defences are open to all members of the campus community as well as the general public. Please email ahmad.aldabbagh@ubc.ca to receive the Zoom link for this defence.
Abstract
The ongoing shift toward Industry 4.0 has led to manufacturing systems of increasing scale and automation, in which a fault in one part of the process can propagate faster than an operator can manually track. Alarm systems are the primary mechanism through which industrial control systems communicate abnormal conditions to human operators. Conventional alarm systems are, however, inherently reactive: an alarm is annunciated only once a monitored variable has already violated its configured limit. This leaves operators little time to intervene before the fault escalates.
This thesis presents a framework for online alarm prediction using tag-level data obtained from programmable logic controllers (PLCs) in order to provide decision support to human operators for enhanced process operation. Specifically, an offline method is proposed for construction of a library of historical process operational scenarios, labelled using the corresponding alarm log. Based on this library, a similarity-based procedure is proposed for online alarm prediction, by comparing incoming PLC tag data against the constructed scenarios.
Moreover, the proposed framework is experimentally validated using the FESTO MPS 403-1 platform, a lab-scale production line. As part of this work, alarms were configured for the platform, with alarm conditions programmed in the PLC and linked to the human-machine interface for real-time annunciation and logging. The results demonstrate satisfactory performance and suitability for practical deployment.
