Metadata and Interoperability in Materials Modelling

On Thursday, 10 March 2016, the European Materials Modelling community held a workshop to discuss metadata and interoperability in materials modelling. The following overview is based on an introduction provided by Adham Hashibon (Fraunhofer IWM).

The purpose of the meeting was to discuss a holistic view on materials modelling data, recognising the universal structure of all models ( Physics Equations (PE) and Material Relations (MR)). It was shown how all basic elements of materials modelling can be represented in a four chapter organisation, the so-called MODA. Such a universal structure will allow a more focused interpretation of modelling information.

The question addressed in the meeting  was how to represent knowledge and not just a collection of raw data (numbers). The metadata extracted by means of these MODA are used to establish interoperability between different types of models and between models and data.

The interoperability is achieved by a fundamental open metadata schema that is based on the elements of material modelling. Starting from this fundamental scheme, means to achieve both syntactic and semantic interoperability were discussed and how these can be further extended to achieve a more global, cross domain level of interoperability. This metadata schema is capable of providing a channel to link different specific domain standards. The schema is not intended to replace existing specific standards, but is rather intended to harmoniously integrate with, and augment existing domain and implementation specific standards of data. The schema is therefore providing for new fundamental interoperability avenues.

The proposed modelling element structures and metadata schema are neutral to any implementation in specific computer programming languages or formal mark-up schemes and also not bound to any specific data file format. Nevertheless, specific examples of implementations of the specification of the schema in both simple language (MODA) and the more formal mark-up languages such as YAML and JSON were presented. Additionally, it was shown that widely endorsed HDF5 based file formats, with their associated simple hierarchical data model can implement the data schema rapidly and efficiently.

The underlying fundamental open schema is further supported by a basic syntactic layer that provides common universal basic attributes (CUBA) defining a set of internally constrained materials modelling vocabulary. The semantics used for the CUBA are further elaborated in the schema allowing machine interpretations so that translations to other domain specific syntaxes and standards are seamless. This is achieved by incorporating a semantic level augmenting the e-CUBA with a common universal data structure (e-CUDS) that provides a neutral representation of the computational metadata including elements from the user case description. In essence, the e-CUDS provide the open semantic based metadata schema and the e-CUBA provide a common language to bridge the nomenclature gap between specific domains and communities.

It was shown that the MODA together with the e-CUDS and e-CUBA allow for a representation of the computational metadata of all models, including electronic, atomistic, mesoscopic and continuum models.

The workshop concluded with a series of challenges presented from the engineering, manufacturers and software owner view points. A particular case example of delamination was posed and answered by a formal representation within the schema presented.