基于电学成像与深度学习的蜂窝结构冲击损伤识别研究
Impact Damage Identification for Honeycomb Sandwich Structure by Using Electrical Tomography and Deep Learning
周登 1李雪峰 1严刚 1黄再兴1
作者信息
- 1. 南京航空航天大学航空航天结构力学及控制全国重点实验室,南京 210016
- 折叠
摘要
针对蜂窝夹层结构遭受外物冲击的情况,提出结合电学成像和深度学习的方法,对冲击损伤进行在线监测和识别,为结构完整性评估和决策提供准确信息.首先通过丝网印刷技术,使用碳油墨和银浆油墨在结构表面分别制备感应层和导电线路;然后对不同数量、位置和尺寸的冲击损伤进行数值仿真,获取对应的感应层电导率变化与边界电压变化数据样本,由残差神经网络进行深度学习,建立二者的映射关系;最后在结构遭受冲击前后分别测量感应层的边界电压数据,通过训练好的残差神经网络重建感应层电导率变化分布的图像,实现损伤信息的识别.通过对蜂窝夹层结构进行低速冲击试验,验证了所提出技术和方法的可行性和有效性.
Abstract
Aiming at the situation of honeycomb sandwich structure impacted by external objects,this study proposes a method to detect and identify the impact damage with electrical tomography and deep learning,and provide precise information for structural integrity evaluation and decision-making.The sensing layer and corresponding circuits are first printed on the surface of honeycomb sandwich structure with carbon ink and silver ink through silk-screen printing technique.Then numerical simulation is performed by considering impact damage with different quantities,positions and sizes to obtain training data of conductivity change and boundary voltage change of the corresponding sensing layer.Deep learning is carried out by a residual neural network to establish the mapping relationship between conductivity change and boundary voltage change of the sensing layer.Finally,the boundary voltage data of the sensing layer is measured before and after impact,and tomographic image of the conductivity change caused by impact damage is reconstructed by a trained residual neural network,identifying the locations and sizes of the damage.Low velocity impact test for a honeycomb sandwich structure is conducted to demonstrate the feasibility and effectiveness of the proposed method.
关键词
蜂窝夹层结构/冲击损伤识别/印刷感应层/电学成像/深度学习Key words
Honeycomb sandwich structure/Impact damage identification/Printed sensing layer/Electrical tomography/Deep learning引用本文复制引用
基金项目
航空科学基金(2017ZA52005)
航空航天结构力学及控制全国重点实验室开放课题(MCMS-E-0423G02)
出版年
2024
NETL
NSTL
万方数据