In this study, we will propose a deep learning-driven inverse design framework for the generation of 3D architected materials. The expected result is that we input the target properties (e.g., porosity and Young’s modulus), and the framework will generate a batch of 3D architected materials with corresponding properties. Volumetric convolutional neural network and generative adversarial net will be combined for the implementation of inverse design. Finite element method simulations and mechanical testing will be conducted for the validation.

This work is an extended work based on our previous paper.

Reference:

Zheng, Xiaoyang, et al. "Controllable inverse design of auxetic metamaterials using deep learning." Materials & Design 211 (2021): 110178.