The paper presents a novel multiscale topology optimization framework for designing porous Voronoi structures. The key aspects are:
Design Space: The framework considers Voronoi microstructures with varying thickness, anisotropy, and orientation as the design parameters, enabling a broader design space.
Connectivity: The method promotes macroscale connectivity by considering the neighboring cell sites during the training of the surrogate neural network.
Positive Definiteness: The neural network is trained to predict the Cholesky factors of the homogenized elasticity matrix, ensuring the positive definiteness of the constitutive matrix.
Computational Efficiency: The offline training of the neural network and its integration into the multiscale optimization process significantly reduces the computational cost compared to direct homogenization-based approaches.
The proposed method is validated through several numerical examples, including a tensile bar and a mid-cantilever problem. The results demonstrate the ability of the framework to generate optimized porous structures that meet the desired performance and volume fraction constraints, while maintaining high computational efficiency.
إلى لغة أخرى
من محتوى المصدر
arxiv.org
الرؤى الأساسية المستخلصة من
by Rahul Kumar ... في arxiv.org 04-30-2024
https://arxiv.org/pdf/2404.18300.pdfاستفسارات أعمق