eduzhai > Applied Sciences > Engineering >

Microstructure Generation via Generative Adversarial Network for Heterogeneous Topologically Complex 3D Materials

  • Save

... pages left unread,continue reading

Document pages: 20 pages

Abstract: Using a large-scale, experimentally captured 3D microstructure dataset, weimplement the generative adversarial network (GAN) framework to learn andgenerate 3D microstructures of solid oxide fuel cell electrodes. The generatedmicrostructures are visually, statistically, and topologically realistic, withdistributions of microstructural parameters, including volume fraction,particle size, surface area, tortuosity, and triple phase boundary density,being highly similar to those of the original microstructure. These results arecompared and contrasted with those from an established, grain-based generationalgorithm (DREAM.3D). Importantly, simulations of electrochemical performance,using a locally resolved finite element model, demonstrate that the GANgenerated microstructures closely match the performance distribution of theoriginal, while DREAM.3D leads to significant differences. The ability of thegenerative machine learning model to recreate microstructures with highfidelity suggests that the essence of complex microstructures may be capturedand represented in a compact and manipulatable form.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...