基于深度学习和近场动力学的层合板冲击工况识别  

Impact working condition identification of composite laminate based on deep learning and peridynamics

作　　者：唐和生[1] 谢雅娟 陈豪[1] TANG He-sheng;XIE Ya-juan;CHEN Hao(College of Civil Engineering,Tongji University,Shanghai 200082,China)

机构地区：[1]同济大学土木工程学院,上海200082

出　　处：《计算力学学报》2023年第1期52-59,共8页Chinese Journal of Computational Mechanics

基　　金：上海市级科技重大专项(2021SHZDZX0100);科技部国家重点实验室基础研究项目(SLDRCE2019-B-02)资助项目。

摘　　要：在冲击荷载作用下复合材料会产生断裂和分层等损伤。基于损伤数据对冲击工况进行识别,对改善复合材料的设计和确保其安全使用具有重要意义。基于此,本文提出一种基于深度学习和近场动力学(PD)理论的层合板冲击工况识别方法。首先使用改进的表面修正系数PD理论建立复合材料层合板刚体冲击损伤演化分析PD模型,PD模型数值模拟结果结合噪声数据增强技术构建层合板的冲击工况数据库;基于深度学习-卷积神经网络(CNN),对不同工况下的冲击损伤演化数据进行训练,实现对未知冲击工况的识别。结果显示,对于钢球冲击速度和角度的识别准确率均高于90%。Under the action of impact load,damage such as fracture and delamination often occurs in composite materials.The identification of impact working conditions based on damage data is of great significance for improving the design of composite materials and ensuring their safety in use.For this purpose,a method for identifying impact working conditions of laminates based on deep learning and peridynamics(PD)theory is proposed in this paper.First,the improved“surface correction factor”PD theory was used to establish a PD model for the damage evolution analysis of composite laminates under the action of rigid body impact,then the numerical simulation results of the PD model and data enhancement technology were adopted to build the impact working condition datasets of the laminates.Deep learning-convolutional neural network(CNN)was adopted,which was trained using impact damage evolution data in different working conditions,and the identification of unknown impact working conditions was realized.The results show that the identification accuracy of initial falling speeds and angles of the steel balls in each working condition is higher than 90%.

关 键 词：深度学习 近场动力学 复合材料层合板 工况识别 冲击损伤 

分 类 号：O313.4[理学—一般力学与力学基础]