Giga-Voxel Multiscale Composite Architecture Mirrored Through a Data-to-Model Closed-Loop Digital Twin

Abstract

Giga-voxel digital models offer abundant geometric detail; however, no mainstream method currently exists to efficiently distribute individual voxels across massive image volumes, and designing complex anisotropic composite materials remains infeasible due to the absence of promising methods. Herein, we propose a systematic digital twin workflow tailored for generating high-fidelity virtual representations of anisotropic composite microstructures and giga-voxel meso-structural models, leveraging a harmonious integration of top-down image-based modeling and bottom-up data-driven generation. Our study demonstrates the efficacy of micro-digital representations as foundational building blocks within a continuum of digital assembly processes tailored for mesostructural models. Utilizing 3D image data, specifically X-ray tomography, our data-driven modeling meticulously characterizes the geometric attributes of the experimentally observed objects, thereby facilitating the creation of digital unit twins, each endowed with distinct identities assigned through a random seed generation. The closed-loop system provides feedback mechanism between data and model to ensure the 3D quality of the generated models. For hierarchical organization at the giga-voxel level, the digital unit twins are methodically expanded into cohesive 3D architectures based on assembly relationship at length scales of more than four orders of magnitude. Remarkably, this hierarchical model provides intricate insight into micro-to-macro geometrics while preserving the intrinsic microstructure.