Abstract
Hot-rolled strip steel is widely used in automotive manufacturing, chemical and home appliance industries, and its surface quality has a great impact on the quality of the final product. In the manufacturing process of strip steel, due to the rolling process and many other reasons, the surface of hot rolled strip steel will inevitably produce slag, scratches and other surface defects. These defects not only affect the quality of the product, but may even lead to broken strips in the subsequent process, seriously affecting the continuation of production. Therefore, it is important to study the surface defects of strip steel and identify the types of defects in strip steel. In this paper, a scheme based on ResNet50 with the addition of FcaNet and Convolutional Block Attention Module (CBAM) is proposed for strip defect classification and validated on the X-SDD strip defect dataset. Our solution achieves a classification accuracy of 94.11%, higher than more than a dozen other compared deep learning models. Moreover, to adress the problem of low accuracy of the algorithm in classifying individual defects, we use ensemble learning to optimize. By integrating the original solution with VGG16 and SqueezeNet, the recognition rate of oxide scale of plate system defects improved by 21.05 percentage points, and the overall defect classification accuracy improved to 94.85%.
Highlights
Hot-rolled strip steel is produced by rolling the billet at a temperature higher than the recrystallization temperature and going through a series of processes such as phosphorus removal, finishing, polishing, edge cutting and straightening
The hardware of the current strip surface defect detection system is sufficient to meet the use of detection, while the algorithm in the server determines the final accuracy of the strip defect classification
In order to obtain sufficient information about the diversity of each channel, Qin et al [29] proved that Global Average Pooling (GAP) is a special form of discrete cosine transform (DCT), and based on this proof, generalized channel attention to the frequency domain and proposed FcaNet, a channel attention network using multiple frequencies
Summary
Hot-rolled strip steel is produced by rolling the billet at a temperature higher than the recrystallization temperature and going through a series of processes such as phosphorus removal, finishing, polishing, edge cutting and straightening. The system takes high speed images of the top and bottom surfaces of the strip steel and determines the images that may have surface defects and passes them to the quality inspector. The main reason for the current use of manual further testing on the basis of the strip surface defect detection system is that the accuracy of the existing system is not yet as good as that of the quality inspectors. The hardware of the current strip surface defect detection system is sufficient to meet the use of detection, while the algorithm in the server determines the final accuracy of the strip defect classification. We validate the proposed algorithm on the X-SDD dataset [7], compare it with several deep learning models, and design ablation experiments to verify the effectiveness of the algorithm
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
