Fusion of Complementary 2D and 3D Mesostructural Datasets Using Generative Adversarial Networks

Abstract

Modelling the impact of a material's mesostructure on device level performance typically requires access to 3D image data containing all the relevant information to define the geometry of the simulation domain. This image data must include sufficient contrast between phases, be of high enough resolution to capture the key details, but also have a large enough 3D field‐of‐view to be representative of the material in general. It is rarely possible to obtain data with all of these properties from a single imaging technique. In this paper, we present a method for combining information from pairs of distinct but complementary imaging techniques in order to accurately reconstruct the desired multi‐phase, high‐resolution, representative, 3D images. Specifically, the deep convolutional generative adversarial networks to implement super‐resolution, style‐transfer and dimensionality expansion. It is believed that this data‐driven approach is superior to previously reported statistical material reconstruction methods, both in terms of its fidelity and ease of use. Furthermore, much of the data required to train this algorithm already exists in the literature, waiting to be combined. As such, our open‐source code could precipitate a step change in the materials sciences by generating the desired high quality image volumes necessary to simulate behaviour at the mesoscale.

Minireview on Porous Media and Microstructure Reconstruction Using Machine Learning Techniques: Recent Advances and Outlook

“…Receiving low-resolution micro-CT images along with high-resolution SEM images, this CNN-based architecture incorporates image segmentation techniques in super-resolution to reconstruct high-resolution micro-CT images. A similar data fusion process has also been done using a GAN-based framework proposed by Dahari et al 37 3.2. Unsupervised Machine Learning.…”

TauFactor 2: A GPU accelerated python tool for microstructural analysis

“…Materials characterisation techniques are constantly improving, allowing the collection of larger field-of-view images with higher resolutions (Withers et al, 2021). Alongside these developments, machine learning algorithms have enabled the generation of arbitrarily large volumes, and can further enhance image quality through super-resolution techniques (Dahari et al, 2023). The resulting high fidelity microstructural datasets can be used to extract statistically representative metrics of a materials composition and performance.…”

True 2D-to-3D Reconstruction of Heterogeneous Porous Media via Deep Generative Adversarial Networks (GANs)

“…Leveraging their high-resolution imaging capabilities, 2D imaging techniques excel in the detection of intricate features at the micron to sub-micron scale across more expansive and consequently more representative areas. This proficiency makes 2D methods a viable alternative to 3D imaging techniques in effectively capturing both essential fine-scale features and the inherent heterogeneity within the sample (Fu et al, 2022;Dahari et al, 2023). Moreover, 2D images are more easily obtained and can be promptly utilized to quantify statistical spatial morphologies and microstructural characteristics (e.g., porosity, specific surface area, and pore sizes) within porous media (Torquato & Stell, 1982;Torquato & Haslach Jr, 2002).…”