2022 Fiscal Year Annual Research Report
Design of auxetic metamaterials using deep learning
| Project/Area Number |
22J11202
|
|---|
| Allocation Type | Single-year Grants |
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
ZHENG Xiaoyang 筑波大学, 理工情報生命学術院, 特別研究員(DC2)
|
|---|
| Project Period (FY) |
2022-04-22 – 2024-03-31
|
| Keywords | Mechanical metamaterial / Architected material / Materials and design / Deep learning / Finite element method / 3D printing / Inverse design |
|---|
| Outline of Annual Research Achievements |
I focus on designing mechanical metamaterials by both forward design and inverse design. In the forward design, I proposed a strategy to design three different reprogrammable flexible mechanical metamaterials, which have the potential for different applications, such as soft robots, actuation, adaptive safety, and sports equipment. The strategy can be easily extended to other structures and smart materials. More importantly, this strategy paves the way to change the mechanical responses for similar architectures.
In the inverse design, I developed a deep learning framework for the inverse design of 2D auxetic metamaterials and 3D architected materials, which have potential application for bone implants and biomedical stents. The inverse design framework will accelerate the revolution of designing such complicated architected materials by removing guesswork from material design in a variety of applications. In addition, other researchers will also get inspiration from the work with respect to materials discovery and development. For example, experimental researcher can get their targeted materials from the inverse design without any prior experience in materials design.
The code of the inverse design of auxetic metamaterials has been shared on MDR, and scientists are able to get access to the code for generating metamaterials with their desired mechanical properties for their targeted applications. Based on this study, I have published 5 peer-reviewed papers on international journals during this academic year.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
I have finished my research proposal. In my research plan, I aimed to develop next-generation flexible, tough auxetic metamaterials with extreme properties, taking advantage of deep generative networks. This work has been done on my published paper, in which I proposes an inverse design method for auxetic metamaterials using deep learning. The deep learning framework enables generating a batch of auxetic metamaterials with a user-defined Poisson’s ratio and Young’s modulus by a conditional generative adversarial network without prior knowledge. The network was trained based on supervised learning using a large number of geometrical patterns generated by Voronoi tessellation.
In addition, I also proposed reprogrammable mechanical metamaterials. These metamaterials are made from light-responsive shape-memory polydimethylsiloxane with reprogrammability. Moreover, I proposed a deep learning framework for the inverse design of 3D architected materials. This inverse design framework is a three-dimensional conditional generative adversarial network (3D-CGAN) trained based on supervised learning using a dataset consisting of voxelized Voronoi lattices and their corresponding relative densities and Young’s moduli. A well-trained 3D-CGAN adopts variational sampling to generate multiple distinct Voronoi lattices with the target relative density and Young’s modulus.
Overall, the study is going smooth, without any problem. The results are beyond expectation. I also did other academic activities, such as creating a university club, Research Support, University of Tsukuba.
|
| Strategy for Future Research Activity |
My future work aims to develop smart mechanical metamaterials that can take responses to their environments and external stimulus. The smart mechanical metamaterials will be generated in inverse-design approach using deep learning and fabricated using 4D printing technology. These designed smart mechanical metamaterials will be used in soft robotics and shape morphing.
The most unique and creative aspect of this research is that I focus on inverse design of mechanical metamaterials. Traditional forward design approaches rely heavily on intuition and experience, whereas inverse design employs deep learning models to autonomously generate a geometry that meets the desired mechanical properties. This allows for rapid production of optimal mechanical metamaterials based on functional requirements, eliminating the need for inefficient trial-and-error processes.
This work is expected to clarify the growth rules of irregular architected materials in nature, and build a universal dataset that can map structure-property-function relationships for mechanical metamaterials. I also believe this research can be valuable to not only researchers working on mechanical metamaterials but also researchers working in the field of materials informatics, by which researchers can take inspiration to revolutionize their traditional design approach of not only materials but also chemicals, electronics, devices, etc.
|
