A defect recognition model for cross-section profile of hot-rolled strip based on deep learning

作     者：Tian-lun Li Wen-quan Sun An-rui He Jian Shao Chao Liu Ai-bin Zhang Yi Qiang Xiang-hong Ma Tian-lun Li;Wen-quan Sun;An-rui He;Jian Shao;Chao Liu;Ai-bin Zhang;Yi Qiang;Xiang-hong Ma

作者机构：National Engineering Research Center for Advanced Rolling and Intelligent ManufacturingUniversity of Science and Technology BeijingBeijing100083China Academy of Machinery Science and TechnologyBeijing100044China School of Engineering and Applied ScienceAston UniversityBirminghamB47ETUK 

出 版 物：《Journal of Iron and Steel Research International》 (国际钢铁研究杂志)

年 卷 期：2023年第30卷第12期

页      面：2436-2447页

核心收录：

学科分类：080503[工学-材料加工工程] 08[工学] 0805[工学-材料科学与工程（可授工学、理学学位）] 0802[工学-机械工程] 080201[工学-机械制造及其自动化] 

基　　金：supported by the National Natural Science Foundation of China(No.52004029) the Joint Doctoral Program of China Scholarship Council(CSC)(202006460073) Liuzhou Science and Technology Plan Project,China(2021AAD0102) 

主　　题：Hot-rolled strip cross section:Curve recognition Deep learning-Stacked denoising autoencoder Support vector machine Imperfect data 

摘      要：The cross-section profile is a key signal for evaluating hot-rolled strip quality,and ignoring its defects can easily lead to a final *** characteristics of complex curve,significant irregular fluctuation and imperfect sample data make it a challenge of recognizing cross-section defects,and current industrial judgment methods rely excessively on human decision making.A novel stacked denoising autoencoders(SDAE)model optimized with support vector machine(SVM)theory was proposed for the recognition of cross-section ***,interpolation filtering and principal component analysis were employed to linearly reduce the data dimensionality of the profile ***,the deep learning algorithm SDAE was used layer by layer for greedy unsupervised feature learning,and its final layer of back-propagation neural network was replaced by SVM for supervised learning of the final features,and the final model SDAE_SVM was obtained by further optimizing the entire network parameters via error ***,the curve mirroring and combination stitching methods were used as data augmentation for the training set,which dealt with the problem of sample imbalance in the original data set,and the accuracy of cross-section defect prediction was further *** approach was applied in a 1780-mm hot rolling line of a steel mill to achieve the automatic diagnosis and classification of defects in cross-section profile of hot-rolled strip,which helps to reduce flatness quality concerns in downstream processes.