Abstract
The MaterialDigital or BWMD dataset is an RDF data repository generated by the Use Case Metals within the framework of the MaterialDigital project. It showcases the digitalization of the aluminum permanent mold casting process, followed by a two-stage heat treatment, involving solution annealing and artificial aging by means of ontology based semantic data structures. For the purposes of the study, the casting alloy AlSi10Mg was used. In two casting campaigns, project-specific databases were established using test bars, which were subjected to mechanical and analytical material characterization. A demonstrator casting was also cast, and subjected to static bending stress on a laboratory test rig. For both casting campaigns, variations in chemical composition of the AlSi10Mg alloy were introduced with respect to the silicon and magnesium content. Accompanying the casting campaigns were casting simulations to refine model parameters through temperature measurements in the test bar mold. There were two primary objectives of the project. The first aim was to leverage the digital workflow to structure the data from the test bar characterization campaign. Individual data sets from each process step were linked together to create a comprehensive and coherent knowledge graph of the process chain, which was then transferred to a graph database. The development of a domain ontology for the process chain allowed the extraction of expert knowledge on the impact of chemical composition and heat treatment parameters on various mechanical properties from the material data space. Beyond querying metadata, the heterogeneous raw data sets could also be accessed by machines, as evidenced by tensile tests. This technology is thus transformative in its ability to capture material- and process-specific expert knowledge, serving as the basis for further data-based analyses. In practical terms, this can guide decision-making regarding the ideal heat treatment parameters given the chemical composition that will ensure the attainment of specific material strength.
Technical Information
In order to work with this dataset download and unzip the folder BWMD_Dataset.zip. The dataset includes the BWMD Ontology file (BWMD_Ontologie_2020-08-12.owl) in non modularized version (as legacy of the MaterialDigital project), which can be opened e.g. with the free software Protégé (https://protege.stanford.edu/about.php). The process semantic data model or generic process graph template (Graph-ProcessTemplate-bwmd-2020-05-15.kdb) is included as well and can be visualized and edited with the Inforapid KnowledgeBase Builder software (https://www.inforapid.com/en-us/). However, the main feature of the uploaded dataset is the complete BWMD RDF graph database (BWMD_full_DB_Anonymized.ttl) generated based on the generic graph template pattern for all the semantically modeled processes. Note that this file contains as well the BWMD Ontology and all the inferred generated statements with the reasoner OWL2-2RL (Optimized) from the Graph DB Free Software. To work with the BWMD RDF graph database import the file BWMD_full_DB_Anonymized.ttl in a graph instance (e.g. Graph DB Free). Once the RDF graph is imported in a local repository it is possible to further work with it and to run SPARQL queries to extract information. To facilitate the user the knowledge extraction process, two Jupyter Notebooks are included (query_mdbw.ipynb and plot_tensile_tests.ipynb) with integrated SPARQL queries. To run these Jupyter Notebooks locally, please keep in mind to download the necessary Python packages and to update the SPARQL endpoint with your own one a the name of the repository you created to import the RDF BWMD database (BWMD_full_DB_Anonymized.ttl).
