Root cause analysis in the industrial domain using knowledge graphs: A case study on power transformers

Abstract

In the industrial domain, developing solutions that allow the identification, understanding, and correction of faults is essential due to the cost of handling such situations. However, to date, there are not many solutions capable of facilitating the human operator to discern the causes and possible solutions for a specific fault. In this work, we present knowledge graph-driven root cause analysis for working with faults in the industrial domain, based on three points of action: reasoning from the current state of machines or processes, failure classification using rules, and advanced querying using graph-query languages. We have conducted a power transformer case study that revealed that our proposed approach could be considered competitive as it has outperformed several alternative machine learning classifiers.