A performance evaluation indicator system and method for manufacturing-integrated generative design

Generative design can explore a broad design space under specified performance and manufacturing constraints, yet its optimization objectives often emphasize component-level properties while giving limited consideration to manufacturing cost, environmental impact, and design-manufacturing coordination. This paper proposes a manufacturing-integrated performance assessment framework that embeds these concerns in the generative design process. The framework comprises four dimensions: technical feasibility, economic rationality, green sustainability, and process coordination. It maps measurable indicators from these dimensions to design objectives, hard constraints, and post-generation evaluation criteria. A two-stage decision procedure is proposed: Pareto-based screening first retains non-dominated alternatives, and a subsequent multi-criteria decision method ranks the retained candidates according to stakeholder preferences and indicator information. A performance-driven workflow is then illustrated using two industrial cases. In the wheel-hub case, a multimodal generation and prediction pipeline combines a conditional generative adversarial network with two- and three-dimensional variational autoencoders and finite-element analysis. The reported test results include a mean absolute percentage error of 2.79% for equivalent-stress prediction and an inference time of approximately 2.5s. A separate illustrative Pareto exercise uses synthetic data and is therefore treated only as a demonstration of the decision procedure. In the aircraft-bracket case, the optimized additively manufactured design reduces component mass by 46.2% and lifecycle cost by 39.50%, although unit manufacturing cost increases by 39.47%. These cases demonstrate why generative design should be assessed across the design-to-manufacturing lifecycle rather than by a single performance metric.