摘要
【目的】针对微结构拓扑优化设计中灵敏度分析计算量大的问题,建立一种基于神经网络的高效拓扑优化设计框架,实现负泊松比微结构的精确构型设计。【方法】首先,提出了一种全连接前馈神经网络(Fully-connected Feedforward Neural Network, FFNN)模型,以设计域坐标作为输入,密度场作为输出,建立了二者的映射关系;其次,引入了神经网络反向传播算法进行灵敏度分析以降低计算量;最后,通过对优化后的高分辨率负泊松比结构进行拉伸仿真及试验,验证了所提方法的有效性。【结果】结果表明,基于FFNN模型的优化框架能显著提升灵敏度计算效率,优化得到的微结构表现出明显的负泊松比效应,仿真与试验结果具有较高的一致性,为复杂微结构的设计提供了新思路。
[Objective]To address the high computational cost of sensitivity analysis in microstructure topology optimization,this study aims to develop an efficient design framework based on neural networks for microstructures with negative Poisson ratio.[Methods]The fully-connected feedforward neural network(FFNN)model was established,mapping the coordinates of the design domain to the density field.The back-propagation algorithm was utilized for sensitivity analysis to reduce computational redundancy.Numerical simulations and tensile tests were conducted on the optimized high-resolution negative Poisson ratio microstructures to verify the proposed method.[Results]The results indicate that the FFNN model framework effectively improves the efficiency of sensitivity analysis.The optimized microstructures exhibit significant negative Poisson ratio effects,and the consistency between simulation and test results demonstrates the validity and robustness of the optimization approach,providing new insights for the design of complex microstructures.
作者
王怡
郭延定
古观成
冮铁强
陈立杰
WANG Yi;GUO Yanding;GU Guancheng;GANG Tieqiang;CHEN Lijie(School of Aerospace Engineering,Xiamen University,Xiamen 361102,China;College of Marine Equipment and Mechanical Engineering,Jimei University,Xiamen 361021,China)
出处
《机械强度》
北大核心
2026年第3期96-103,共8页
Journal of Mechanical Strength
关键词
神经网络
拓扑优化
负泊松比微结构
灵敏度分析
有限元方法
Neural network
Topology optimization
Negative Poisson ratio microstructure
Sensitivity analysis
Finite element method
