Skyscrapers Need Solid Foundations: Pattern-Based Validation of SysML v2 Models with Refinery
TL;DR
The need for custom validation capabilities to ensure that models produced by both humans and LLMs are trustworthy enough to serve as the foundation of the MBSE ecosystem is addressed.
Abstract
The release of SysML v2, coinciding with rapid advances in Large Language Models (LLMs), marks a turning point for Model-Based Systems Engineering (MBSE). SysML v2 introduces a textual notation, a standard API suited for the Digital Thread, and a semantics grounded in a formal kernel, aiming to improve both syntactic and semantic interoperability between tools. These capabilities make it more important than ever to build valid and semantically consistent models. Although the SysML specification defines a large number of well-formedness constraints, it is still surprisingly easy to create semantically inconsistent models that can silently undermine the MBSE ecosystem built upon them. Emerging methodologies are also expected to introduce additional patterns, best practices, and modeling rules, which must be checked in addition to the built-in constraints. This paper addresses the need for custom validation capabilities to ensure that models produced by both humans and LLMs are trustworthy enough to serve as the foundation of the MBSE ecosystem. We present an approach that extracts the latest SysML metamodel from the official language sources in a separate preparation step, leverages the Systems Modeling API to retrieve models, and then uses the Refinery graph solver to detect undesired patterns defined in its declarative pattern language. The approach is illustrated using patterns from the "Common SysML v2 Pitfalls" blog series accompanying "The SysML v2 Book".
