Generative Artificial Intelligence for Industrial Design and Manufacturing: Generative Design, Synthetic Data Generation, and Diffusion-Based Optimization

Abstract

The growing complexity of modern manufacturing—encompassing additive manufacturing, precision assembly, and high-mix low-volume production—demands design and optimization tools that can explore vast design spaces, generate high-quality training data for data-driven models, and accelerate the transition from conceptual design to validated product. Generative artificial intelligence (Generative AI)—comprising generative adversarial networks (GANs), variational autoencoders (VAEs), diffusion models, and large language models—has emerged as a transformative toolkit for industrial design and manufacturing, enabling the automatic generation of novel designs, the synthesis of training data for defect detection and quality inspection, and the optimization of manufacturing processes under complex physical constraints. This review provides a comprehensive and critical synthesis of the application of generative AI in industrial design and manufacturing. We examine three major application domains: generative design enabled by deep learning and reinforcement learning for topology optimization of additively manufactured structures; synthetic data generation using GANs and diffusion models to address data scarcity in industrial defect detection; and diffusion-based 3D shape generation and microstructure design for materials discovery and manufacturing process optimization. We further connect these generative AI methods to advances in industrial sensing—precision 3D optical metrology, four-dimensional thermal imaging, and collaborative robotic inspection—demonstrating their complementary roles in the intelligent manufacturing ecosystem. A central contribution of this review is the articulation of a unified Generative-to-Physical (G2P) pipeline framework that connects generative AI design exploration, physics-based validation, and manufacturing process control, charting a course toward AI-augmented creative engineering.