Multi-level architecture modelling and analysis: The case for model-based systems engineering of fusion diagnostics

Fusion reactors are complex systems of interconnected parts, many of which rely on novel technologies. Understanding the patterns of connectivity is essential for many design aspects: assessing technical risks, leveraging modularity, preventing rework and enabling functional integration. The enabling design artefact is the system's architecture, its conceptual model of functions, components and interfaces. In this paper, we present two architecture modelling and analysis methods based on Dependency Structure Matrices (DSM). Firstly, the bottom-up method builds an architecture model from decentralized documentation of all ITER diagnostics and their interfaces. The DSM highlights a modular system topology, integration challenges and opportunities for managing complexity. The limitations of this method, however, motivate the development of a top-down alternative. The top-down method is suited for early-stage design definition, therefore generating valuable insights throughout the development lifecycle. The architecture model serves as a central storage of design information, from which viewpoints can be automatically generated: DSMs, functional flow diagrams, cabling diagrams, interface reports and requirements propagation tables. This model-based information paradigm drastically improves design quality, project planning, change management and stakeholder communication. The top-down method is demonstrated for the Visible Spectroscopy Reference System, one of ITER's first diagnostics, but promises a more widespread application.