A Generative AI Pipeline for Sustainable Product Design in Additive Manufacturing Applications

Abstract

Generative AI is rapidly transforming how products are conceived, enabling more intuitive, flexible, and automated design processes. In the context of sustainable manufacturing, these advances open new possibilities for creating customized products that are not only functional and appealing but also optimized for material and energy efficiency. This paper introduces a modular pipeline that combines text-to-image diffusion models with a Blender-based 3D modeling workflow to generate manufacturable objects suitable for additive manufacturing. The pipeline supports end-to-end generation and evaluation, integrating semantic guidance, geometric transformation, and performance assessment. Design outputs are evaluated across four key dimensions: semantic alignment (CLIP Score, CLIP-R-Precision), perceptual quality (HPS-v2), sustainability indicators (build time, filament volume, energy consumption, production cost, and carbon footprint), and a mesh-based printability check to ensure manufacturing feasibility. The pipeline is instantiated in the Eco-Case Architect tool, a demonstrator for the generation of customized 3D-printed phone cases from natural language prompts. Experimental results reveal the impact of prompt complexity and visual category on both aesthetic quality and eco-efficiency. The results demonstrate that AI-generated prompts can produce functional and eco-efficient designs, offering a novel approach to sustainable product customization within digital manufacturing workflows.