This article proposes a deep convolutional neural network with perceptible generalization ability for structural topology optimization. The architecture of the neural network is made up of encoding and decoding parts, which provide down- and up-sampling operations. The popular U-Net was adopted to improve the performance of the proposed neural network. To train the neural network, a large dataset is generated by Simplified Isotropic Material with Penalization (SIMP). The performance of the proposed method was evaluated by comparing its efficiency and accuracy with SIMP on a series of typical optimization problems. Results show that a significant reduction in computation cost was achieved with little sacrifice to the performance of design solutions. Furthermore, the generalization ability of the proposed method is discussed. This ability enables the model to obtain a solution to a problem when a boundary condition is not included in the training dataset with a certain accuracy.

本文提出了一种具有可感知泛化能力的深度卷积神经网络，用于结构拓扑优化。神经网络的架构由编码和解码部分组成，它们提供下采样和上采样操作。采用流行的 U-Net 来提高所提出的神经网络的性能。为了训练神经网络，使用带惩罚的简化各向同性材料 (SIMP) 生成了一个大型数据集。通过在一系列典型优化问题上与 SIMP 比较其效率和准确性来评估所提出方法的性能。结果表明，计算成本显着降低，而设计解决方案的性能几乎没有牺牲。此外，讨论了该方法的泛化能力。