A decision-centered framework for simulation configuration in knowledge-intensive engineering design

Engineering design, including generative design, relies on simulation to evaluate design candidates, yet the associated configuration choices are often treated as predefined. Choices about physical models, discretization, solver settings, and validation affect numerical stability, the reliability of results, and the comparability of design alternatives. This paper proposes a decision-centered framework for simulation configuration based on a Concept-Decision-Knowledge (CDK) meta-model. The framework connects design requirements, configuration choices, and supporting knowledge. It represents dependencies between configuration choices, supports adjustments based on validation results, and incorporates stage-dependent human review and AI-assisted support. Domain-specific instantiation specifies the physical entities, rules, solver interfaces, and acceptance criteria within the CDK structure. The framework is demonstrated through a NACA66 hydrofoil case examining configuration generation and comparison, and a coupled plate–cavity finite-element case examining knowledge reuse, human review, and configuration adjustment. In two compatible plate–cavity target cases, reusing previously validated configuration records reduces the number of harmonic-response frequency solves by 45.5% and 51.4% under the same numerical acceptance criteria. The cases demonstrate configuration and evaluation mechanisms under the tested conditions and motivate broader validation of efficiency, user trust calibration, and interaction outcomes across multiple participants.