Скачать или смотреть Bi-material 4D-printed facial shells designed by deep learning

This work has been published in Soft Science.
Journal link: https://www.oaepublish.com/articles/s...

This study develops a new 4D-printing technique for manufacturing 3D curved surfaces using deep learning. In 4D printing, a planar structure is first fabricated via 3D printing and then transformed, known as shape morphing, into the desired 3D shape through deformation over the fourth dimension—time—by applying heat. Since no support material is required for printing overhanging structures, both time and material consumption can be significantly reduced. Compared with traditional injection molding, this technique does not require expensive molds or equipment, making it suitable for high-mix, low-volume production. However, achieving precise transformation into the target geometry requires solving an inverse design problem. Using the human face as the model system, this study proposes a new method capable of generating designs directly from a single depth image, with potential applications to other 3D curved-surface structures in the future.

本研究以深度學習開發4D列印製造3D曲面的新技術。4D 列印係指以3D列印平面結構後，透過第四個維度「時間」，將其加熱變形成立體的目標結構，因不須列印懸空結構之支撐材，可大幅節省時間及材料；相較於傳統射出成型，本技術不須昂貴的模具與設備，適合少量多樣製造。然而，要精準變形成目標幾何，須解決反向設計問題。本研究以人臉為模型系統，提出能由單張深度照片直接生成設計的新方法，未來可應用於其他3D曲面結構。