Investigation of crack segmentation and fast evaluation of crack propagation, based on deep learning

作     者：Than V.TRAN H.NGUYEN-XUAN Xiaoying ZHUANG Than V.TRAN;H.NGUYEN-XUAN;Xiaoying ZHUANG

作者机构：Institute of PhotonicsLeibniz University HannoverHannover 30167Germany CIRTech InstituteHUTECH UniversityHo Chi Minh City 700000Vietnam College of Civil EngineeringTongji UniversityShanghai 200092China 

出 版 物：《Frontiers of Structural and Civil Engineering》 (结构与土木工程前沿（英文版）)

年 卷 期：2024年第18卷第4期

页      面：516-535页

核心收录：

学科分类：12[管理学] 081406[工学-桥梁与隧道工程] 08[工学] 0810[工学-信息与通信工程] 0711[理学-系统科学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 080203[工学-机械设计及理论] 0837[工学-安全科学与工程] 0805[工学-材料科学与工程（可授工学、理学学位）] 0813[工学-建筑学] 0802[工学-机械工程] 0835[工学-软件工程] 0814[工学-土木工程] 0811[工学-控制科学与工程] 0701[理学-数学] 082301[工学-道路与铁道工程] 0823[工学-交通运输工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：The first author would like to thank European Commission H2020-MSCA-RISE BESTOFRAC project for research funding 

主　　题：deep learning crack segmentation crack propagation encoder−decoder recurrent neural network 

摘      要：Identifying crack and predicting crack propagation are critical processes for the risk assessment of engineering *** traditional approaches to crack modeling are faced with issues of high computational costs and excessive computing *** address this issue,we explore the potential of deep learning(DL)to increase the efficiency of crack detection and forecasting crack ***,there is no single algorithm that can fit all data sets well or can apply in all cases since specific tasks *** the paper,we present DL models for identifying cracks,especially on concrete surface images,and for predicting crack ***,SegNet and U-Net networks are used to identify concrete *** gradient descent(SGD)and adaptive moment estimation(Adam)algorithms are applied to minimize loss function during ***,time series algorithms including gated recurrent unit(GRU)and long short-term memory(LSTM)are used to predict crack *** experimental findings indicate that the U-Net is more robust and efficient than the SegNet for identifying crack segmentation and achieves the most outstanding *** evaluation of crack propagation,GRU and LSTM are used as DL models and results show good agreement with the experimental data.