A unified model for metallic materials reliability data and its application in NIMS metallic materials database (Kinzoku)

A data model to organize metallic materials’ reliability properties are presented, along with data format variations to implement the conceptual model. By analysing the structures behind the Creep Data Sheet and Fatigue Data Sheet series from the National Institute for Materials Science (NIMS), commonalities were identified to establish a standardized list of entities and their relationships. The data model incorporates a multilayered approach that acknowledges both uniform materials and non-uniform materials such as welded joints. Standardized identifier system supports interlinking between the different entities. The model was shown to be applicable to data from various independent initiatives and is effective in supporting datasets for machine learning. The model was implemented in three different formats: spreadsheets, relational database, and key-value document store. Their characteristics are comparatively discussed. Plain-text spreadsheets provide low-barrier editing and are version controllable. Relational database management systems offer data integrity and fast querying, suitable as a backend to a web application; NIMS’s Metallic Materials Database (Kinzoku) system was renewed by taking advantage of these characteristics. Key-value document stores can be flexible, highly machine-readable, and self-describing. They also allow for linking with external ontology for heterogeneous data integration. Digitized management of materials reliability data using this model supports storage, federation, and accelerated utilization in data-driven methodologies.