疲劳裂纹扩展的卷积神经网络辨识  

Convolutional Neural Network Based Fatigue Crack Growth Identification

作　　者：袁慎芳[1] 李晓泉 陈健[1] Yuan Shenfang;Li Xiaoquan;Chen Jian(State Key Laboratory of Mechanics and Control of Mechanical Structures,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学机械结构力学及控制国家重点实验室,江苏南京210016

出　　处：《航空科学技术》2020年第7期64-71,共8页Aeronautical Science & Technology

基　　金：国家自然科学基金(51921003,51635008,51635007);江苏省重点研发计划(BE2018123);江苏高校优势学科建设工程资助项目。

摘　　要：本文提出一种基于卷积神经网络(convolutional neural network,CNN)的疲劳裂纹扩展辨识方法,首先针对导波健康监测损伤特征分散性问题,建立二维多通道多频率损伤特征模式用于卷积神经网络辨识,设计具有复杂非线性运算能力的卷积神经网络解决疲劳裂纹扩展的定量辨识问题。所提方法在变幅载荷疲劳裂纹扩展试验中进行了验证,相比常规多项式拟合辨识方法最大误判长度由2.95mm减小至1.66mm,均方根误差由1.20mm减小至0.33mm,实现了疲劳裂纹扩展辨识准确率的有效提升。This paper proposes a convolutional neural network(CNN)based fatigue crack growth identification method.Firstly,for the dispersion problem of guided wave damage features,a two-dimensional multi-channel multifrequency damage features pattern is proposed as the CNN input.Then,a convolutional neural network is constructed for quantitative identification of fatigue crack growth,resorting to its capability of complex nonlinear computing.The proposed method was verified with the experiment of fatigue crack growth under variable amplitude loading.Compared with the conventional polynomial fitting identification method,the maximum identification error was reduced from 2.95mm to 1.66mm,and the root mean square error was reduced from 1.20mm to 0.33mm,showing great improvements on the accuracy of fatigue crack growth identification.

关 键 词：疲劳裂纹扩展辨识 卷积神经网络 导波结构健康监测 变幅疲劳载荷 

分 类 号：TB553[理学—物理]