Context-Specific Automated Quality Assessment for System Models in MBSE

abstract

In Model-based Systems Engineering, system models serve as central engineering artifacts that support design, verification, and validation throughout development. Model quality strongly affects the usefulness of these artifacts, yet existing assessment approaches are either too generic, require excessive manual effort, or are insufficiently tailored to project-specific quality expectations for practical use. This paper presents CoSAMoQA, a context-specific method for automated model quality assessment that combines metric-based extraction, structured expert review, and data-driven estimator generation. The method is designed to derive project-specific quality estimators from model repositories and expert knowledge while preserving traceability to review criteria. We further report on an initial evaluation in a case-based setting using versioned AutoFOCUS3 model repositories. The case study yields strong inter-rater agreement across five quality goals (Krippen- dorff’s α ≥ 0.87) and a comprehensibility estimator achieving 90.9% test accuracy on temporally held-out data, supporting the feasibility of context-specific quality assessment in MBSE.