基于深度学习的工程结构损伤识别研究进展

为避免或减轻工程结构在建造和运营期间因结构振动产生不同程度损伤，造成安全隐患危及人们生命财产安全，针对结构振动损伤识别技术展开研究，探讨不同深度学习方法发展情况及其利弊，寻找更具可行性的损伤识别方法，并对其最新研究及应用现状进行全面综述。研究结果表明:应用深度学习开发新的结构损伤识别技术，无需冗余的数据预处理以及手工提取损伤特征，实现以较高精度实现损伤识别任务；一维卷积神经网络（1D-CNN）以其独特的应用优势，在数据样本有限条件下较二维卷积神经网络（2D-CNN）表现更为出色。研究结果可为数据驱动的结构损伤识别问题提供新思路，进一步完善土木结构健康监测研究体系。

Research progress in damage identification of engineering structure based on deep learning

In order to avoid or reduce different degrees of damage caused by structural vibration during the construction and operation of engineering structure,resulting in potential safety hazards and endangering the safety of people’s lives and property,the damage identification technology of structural vibration was studied.The development of different deep learning methods and their advantages and disadvantages were explored,then the more feasible damage identification methods were searched,and their latest research and application status were comprehensively reviewed.The results showed that the new structural damage identification technology developed by applying the deep learning could achieve the damage identification tasks with high accuracy without redundant data preprocessing and manual extraction of damage features.With its unique application advantages,the compact 1D-CNN performed better under the condition of limited data samples.The research results can provide new ideas for data-driven structural damage identification and further improve the research system of civil structure health monitoring.