The GRU model, trained to predict stress-strain response and energy absorption, uses eight discrete parameters to characterize the design space. It efficiently predicts new design responses in 0.16 milliseconds, enabling the rapid performance evaluation of 128,000 designs any given strain rate and final strain.
Kushwaha, S., He, J., Abueidda, D., & Jasiuk, I. (2023). Designing impact-resistant bio-inspired low-porosity structures using neural networks. Journal of Materials Research and Technology, 27, 767-779. https://doi.org/10.1016/j.jmrt.2023.09.240
@article{kushwaha2023designing, title={Designing impact-resistant bio-inspired low-porosity structures using neural networks}, author={Kushwaha, Shashank and He, Junyan and Abueidda, Diab and Jasiuk, Iwona}, journal={Journal of Materials Research and Technology}, volume={27}, pages={767--779}, year={2023}, publisher={Elsevier} }
The training data is large in size and can be downloaded through the following UIUC Box link: https://uofi.box.com/s/pzypr2rs6il7ro2jxmsbsixhki3an4mz