Topology Optimization via Neural Reparameterization and Automatic Differentiation

Introduction

The rapid development of deep learning has brought new opportunities for the exploration of topology optimization methods. The combination of deep learning and topology optimization has become one of the hottest research fields at the moment. In this work, the neural network is directly used for topology optimization. The update of the design variables (pseudo-density) in the conventional topology optimization method is transformed into the update of the neural network's parameters, which is called Neural Reparameterization. The sensitivity analysis in the conventional topology optimization method is realized by the powerful Automatic Differentiation technology.

Procedure

Fig 1. The procedure of AuTONR, taking the stiffness optimization of the linear elastic model as an example.

Fig 2. The architecture of the neural network.

Codes

An educational code for structural topology optimization based on AuTONR.

Results

The benchmark MBB example and cantilever beam are presented.

Running Requirement

• Python
• Jax
• Jaxlib
• Optax
• Flax

Citation

**Please contact to ** zhangzeyu_work@outlook.com
Disclaimer: The author reserves all rights but does not guarantee that the code is free from errors. Furthermore, we shall not be liable in any event.

@article{Zhang_TONRExploration_2021,
    title = {{{TONR}}: {{An}} Exploration for a Novel Way Combining Neural Network with Topology Optimization},
    shorttitle = {{{TONR}}},
    author = {Zhang, Zeyu and Li, Yu and Zhou, Weien and Chen, Xiaoqian and Yao, Wen and Zhao, Yong},
    year = {2021},
    month = dec,
    journal = {Computer Methods in Applied Mechanics and Engineering},
    volume = {386},
    pages = {114083},
    issn = {00457825},
    doi = {10.1016/j.cma.2021.114083},
    langid = {english},
    }