Integrating Model-Based and Data-Driven Approaches for DfX-Oriented Systems Engineering

The paper presents a hybrid approach that combines model-based and data-driven methods to assist the Systems Engineer in exploring the design space and performing trade-off analyses in accordance with the Design for $\mathrm{X}$ (DfX) paradigm. The proposed solution leverages data-driven models to generate design solutions, while iteratively refining them by incorporating physics-based evaluations. This integration not only improves the system's ability to generate viable design configurations, but also enhances the quality and consistency of the generated solutions. The methodology is formalized through an optimization-based neural network model, incorporating physical constraints and objective functions to balance multiple design criteria. A reference application scenario, focused on aircraft wing parameterization, illustrates the feasibility and effectiveness of the proposed approach.