Logo image
Back
Generative Deep Learning for Designing Printable Multifunctional Microstructural Materials: Application to Piezocomposites
Journal article   Peer reviewed

Generative Deep Learning for Designing Printable Multifunctional Microstructural Materials: Application to Piezocomposites

Mohammad Saber Hashemi, Fatemeh Delzendehrooy, Khiem Nguyen, Levi Kirby, Xuan Song and Azadeh Sheidaei
Journal of the mechanics and physics of solids, Vol.204, 106253
11/2025
DOI: 10.1016/j.jmps.2025.106253
url
https://eprints.gla.ac.uk/view/author/65576.html>View
Open Access

Abstract

This study presents a novel generative deep learning framework tailored to design printable, multifunctional microstructural materials. Our approach integrates a custom-developed voxelized microstructure generator, HetMiGen, with a new machine learning model, TransVNet, which together facilitates the rapid and accurate design of materials with desirable multifunctional properties, especially those showing competing properties with microstructural changes, such as being soft and piezoelectrically sensitive concurrently. Keys to our methodology are the efficient computational homogenization using fast Fourier transform (FFT) techniques and a bi-directional establishment of structure-property relationships that significantly condenses the design cycle. The effectiveness of our framework is validated through the experimental manufacture and testing of piezocomposite microstructures, confirming computational predictions. Results demonstrate the framework's capability to expedite the development of materials with tailored functionalities, offering significant implications for advancing material design technologies. [Display omitted]
 Ceramic–elastomer composites Discrete Fourier transform Generative deep learning Microstructure design Piezoelectricity

Details

Metrics

13 Record Views
Logo image