Shape-functional fusion metamaterials for vibration isolation by integrating forward and reverse deep leaning-driven design

The work presents a deep learning-driven approach integrating forward and reverse design to generate families of mechanical metamaterials fusing different shapes and functionalities. A deep learning model of a double neural network integrates forward and reverse design provides a novel strategy for general metamaterial design by generating and optimizing the functional and structural properties of architected materials. We propose an application of the method to design, model and test a new class of mechanical metamaterials characterized by the presence of dual-mass local resonators inspired by the Chinese Taiji diagram. A shape-functional fusion metamaterial with customizable vibration isolation bands has been designed and prototyped. This innovative structure integrates various metamaterial shapes within diverse functional frameworks to achieve a cohesive design of form and function. It is suitable for applications in structural vibration control, noise reduction, vibration energy harvesting, and lightweight design across fields such as industrial equipment, construction, and transportation.