FAIRification of Additive Manufacturing Process Through Ontology Development and Linked Data

In order to improve current knowledge on Additive Manufacturing process the use of computer science principles is of interest. In order to optimize datasets for data analysis is a difficult process, due in part to the multiple data sources and the different output formats from each data source. The FAIRification process represents an attractive alternative to overcome the challenges associated with the storage, integration and reusability of these datasets. FAIR principles describe a set of guidelines applied to data and metadata to ensure a proper and efficient data management system. When applied correctly, it ensures that data is properly stored, organized, accessible and reusable. A fundamental component in the development of a successful FAIR framework and the integration of datasets is the development of ontologies. Domain ontologies are standardized representations of properties and relationships in a field and provide a sharing and common understanding within the research community. Using FAIR principles and domain ontologies as the backbone of organizing and processing data, the use of automated extraction methods are more feasible and able to process data regardless of the original capturing method. Data from disparate sources need to be processed and extracted in a method that can capture the feature of interest as well as relate the feature in a linked manner to the original data output. Information in the Additive Manufacturing process of Laser Powder Bed Fusion is sparse and separated among multiple instrument data sources that both capture different features and represent it using different encoding methods. These different encoding methods have the same object in reference while having different material characteristics extractable from the data source. In order to process through large data spaces of disperite features, data must be organized in a machine-compatible method. Furthermore, data must be retrieved with an automated pipeline for the ability to extract measurement of specific variables in a verifiably reproducible manner. Many datasets are reliant on operator specific inputs and often unreproducible evaluation methods. By using a more data scientific approach and organization scheme of linking datasets, information on Additively Manufactured parts, in spite of its sparse data, can be used and expanded further data visualization and trend relationships through specific software development and application. Organizing and streamlining data outputs to a common data framework allows for automation of analysis, quick retrieval of digital representations of variables of interest and the framework for decision making using data driven methods as opposed to physics informed simulations.