Toolbox for Automated Failure Diagnosis Using Ontology-Based Knowledge Graphs (TAFO-KG)

Machine data often lacks standardization due to variations in component configurations, sensor configurations, and signal nomenclature, leading to significant differences in data across machine types. This hampers the analysis of data and modeling over different machine types. Furthermore, improved availability of machine data has also increased the complexity of manual analysis and pattern detection, making it excessively time-consuming, necessitating automation to handle high-dimensional data effectively. In this project, a toolbox will be implemented, that solves these two problems by (1) automatically generating a machine ontology (based on expert knowledge and 
pattern mining techniques); and (2) knowledge graph generation and methods for interpretation for failure diagnosis for heterogenous machines (see Figure 1 below).

By using the ontology, failure diagnosis of different machine types becomes feasible and transparent. This toolbox is based on extensive research that has been done in the context of FAIR, supported by datasets and research efforts in linked ICON and EU projects. The synergy between the fundamental research in FAIR and the use of real-life fleet data in the linked projects necessitated the emphasis on usability of the toolbox in realistic scenarios with complex, heterogeneous data polluted by disturbances. To this end, the code is made robust and scalable. The toolbox will be tested on data from several wind farms consisting of geared wind turbines.

Downloads

owilab-vub-cm / tafo-kg · GitLab

Contacts

Xavier Chesterman - Xavier.chesterman@vub.be

Researcher

Jan Helsen - Jan.Helsen@vub.be

Research group

VUB AVRG

VUB AI Lab