Generative Artificial Intelligence in Smart Engineering Design Optimization: A Synthesis of Computational Creativity, Structural Performance, and Human-Centered Practice

Abstract

Generative artificial intelligence has moved from a peripheral computational curiosity to a central instrument of contemporary engineering design, reshaping how structural forms, materials configurations, and manufacturing strategies are conceived and refined. This paper undertakes an integrative, theory-driven synthesis of the scholarship surrounding generative AI in smart engineering design optimization, with particular attention to generative adversarial networks, variational autoencoders, diffusion-based models, and emerging large language model agents. Rather than cataloguing individual techniques, the study interrogates how these computational paradigms interact with established engineering design theory, including design space exploration, multi-objective optimization, and design for manufacturing. Employing a qualitative, meta-synthetic methodology grounded in a structured review of peer-reviewed literature published primarily between 2014 and 2026, the paper develops an analytical framework that maps generative model classes onto discrete phases of the engineering design lifecycle. The findings indicate a discernible trajectory from narrow, single-objective topology generation toward multimodal, constraint-aware, and increasingly autonomous design agents capable of negotiating competing structural, material, and manufacturability requirements. The analysis further reveals persistent gaps concerning interpretability, manufacturing feasibility validation, and the epistemic role of human designers within increasingly automated pipelines. The paper concludes that generative AI functions most productively not as a replacement for engineering judgment but as a computational collaborator that expands the exploratory reach of the design process while imposing new obligations of verification, accountability, and domain grounding. These insights contribute a consolidated conceptual map for researchers and practitioners seeking to situate emergent generative techniques within a coherent, rigorous engineering design paradigm.