Anomaly Detection in Chemical Processes with Semantic Knowledge Graphs: An Approach to Reduce Cause-Effect Diagrams

Research output: Contribution to book/conference proceedings/anthology/reportConference contributionContributedpeer-review

Abstract

When operators and users face an unexpected problem regarding the process, it would be easier and less time-consuming for them to narrow down the possible causes that result in the final issue. In this study, a pilot modular plant with two Process Equipment Assemblies was considered as the use case. At first, a knowledge graph representing this process was developed in Protégé software, which semantically described not only the equipment type and connectivity but also the behavior of the process. Then, the knowledge graph was imported to Python, where the sensor data were placed in their particular position in the knowledge graph. Afterward, an algorithm was developed to query the knowledge graph and verify if the relevant equipment was functioning correctly or not. Our results indicate that this approach can reduce the cause-effect diagrams in almost all scenarios. Nonetheless, there are situations where further sensor data (e.g., the temperature in a tank) is required for the algorithm to decide.

Details

Original languageEnglish
Title of host publication33rd European Symposium on Computer Aided Process Engineering
Number of pages6
Publication statusPublished - Jan 2023
Peer-reviewedYes

External IDs

ORCID /0000-0003-3753-3778/work/142238508
ORCID /0000-0001-5165-4459/work/142248312
Scopus 85166933877
Mendeley 12f29a87-628f-3688-8742-9b4c65f483aa