Morphology detection of converter mouth based on Mask-RCNN
DAI Zhang-jie1, HUANG Cheng-yong1,2, LIU Wei1, XIA Jian-chao3, YANG Shu-feng1, LI Jing-she1
1. School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing, Beijing 100083, China; 2. Equipment Department, Shanghai Meishan Iron and Steel Co., Ltd., Nanjing 210039, Jiangsu, China; 3. Steelmaking Plant, Shanghai Meishan Iron and Steel Co., Ltd., Nanjing 210039, Jiangsu, China
Abstract:With the further improvement of the automation requirements of the steelmaking process, it is of great significance to realize the automatic feeding of the process. Because the converter smelting is often accompanied by the phenomenon of slag overflow, the furnace mouth is easy to form nodules, resulting in the reduction of the inner diameter, which affects the next furnace feeding process. At present, the site mainly observes the morphology of the furnace mouth manually to determine whether it is necessary to repair the furnace mouth. This traditional manual detection is inefficient, and the detection results are unstable due to factors such as the light intensity in the furnace and the subjective judgment of the technicians. With the rapid development of deep learning methods in the field of artificial intelligence and its important role in various industries, a converter mouth shape detection method based on Mask-RCNN is proposed. Based on the output results of the network model, the method calculates the perimeter and area of the furnace mouth contour through image processing, and combines the least squares circle fitting method (LSCM) to characterize the morphology more reasonably and quantitatively through the roundness index. The experiment shows that with the continuous smelting, the area and perimeter of the furnace mouth will continue to decrease, and the corresponding roundness error value will continue to increase, which indicates that the morphology of the furnace mouth will continue to change due to the phenomenon of slag adhesion. The automatic detection method can perform real-time stable detection on the furnace mouth with different rotation angles when feeding or slag discharge. The recognition rate is as high as 99%, which has high detection performance. Compared with manual detection, it has the advantages of stability, accuracy and efficiency, and greatly improves the process stability.
戴张杰, 黄成永, 刘威, 夏建超, 杨树峰, 李京社. 基于Mask-RCNN的转炉炉口形貌检测[J]. 钢铁, 2023, 58(3): 73-78.
DAI Zhang-jie, HUANG Cheng-yong, LIU Wei, XIA Jian-chao, YANG Shu-feng, LI Jing-she. Morphology detection of converter mouth based on Mask-RCNN[J]. Iron and Steel, 2023, 58(3): 73-78.
[1] 马湧, 刘新建, 蒋胜龙. 基于机器学习的炼钢-连铸生产调度方法[J]. 冶金自动化, 2022, 46(2): 96.(MA Yong, LIU Xin-jian, JIANG Sheng-long. Machine learning-based scheduling approach for steelmaking continuous casting production [J]. Metallurgical Industry Automation, 2022, 46(2): 96.) [2] 甄景燕, 尹国强, 王舒军, 等 炼钢动态调度系统及其应用 [J]. 冶金自动化, 2017, 41(3): 23.(ZHEN Jing-yan, YIN Guo-qiang, WANG Shu-jun, et al. Steelmaking dynamic scheduling system and its application[J]. Metallurgical Industry Automation, 2017, 41(3): 23.) [3] PENG X Y,LI L,WANG F Y. Accelerating minibatch stochastic gradient descent using typicality sampling[J]. IEEE Transactions on Neural Networks and Learning Systems,2020,31(11):4649. [4] 何止戈. 基于深度学习方法的PCB图像缺陷检测[D]. 成都:电子科技大学, 2020.(HE Zhi-ge. Defect of PCB Image Detection Based on Deep Learning Method[D]. Chengdu: University of Electronic Scienceand Technology of China,2020.) [5] 李江昀, 杨志方, 郑俊锋, 等 深度学习技术在钢铁工业中的应用[J]. 钢铁, 2021, 56(9): 43.(LI Jiang-yun, YANG Zhi-fang, ZHENG Jun-feng, et al. Applications of iron and steel industry with deep learning technologies[J]. Iron and Steel, 2021, 56(9): 43.) [6] 于加学, 孙杰, 张殿华. 基于深度学习的热轧带钢头部厚度的命中预测[J]. 钢铁, 2021, 56(9): 19.(YU Jia-xue, SUN Jie, ZHANG Dian-hua. Accurate prediction of head thickness of hot-rolled strip based on deep learing[J]. Iron and Steel, 2021, 56(9): 19.) [7] Girshick R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision. Chile: IEEE Computer Society,2015: 1440. [8] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. Advances in Neural Information Processing Systems, 2017,39(6):1137. [9] He K, Gkioxari G, Dollár P, et al. Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE Computer Society,2017: 2961. [10] 孙志青. 基于深度学习的铝土矿分选方法研究[D]. 太原:中北大学, 2021.(SUN Zhi-qing. Research on Bauxite Ore Detection Method Based on Deep Learning[D]. Taiyuan: North University of China,2021.) [11] 李大军, 何维龙, 郭丙轩, 等. 基于Mask-RCNN的建筑物目标检测算法[J]. 测绘科学, 2019, 44(10): 172.(LI Da-jun, HE Wei-long, GUO Bing-xuan, et al. Building target detection algorithm based on Mask-RCNN[J]. Science of Surveying and Mapping, 2019, 44(10): 172.) [12] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI Conference on Artificial Intelligence. California: AAAI Press, 2017: 236. [13] 汤鹏, 刘毅, 魏宏光, 等. 基于Mask-RCNN海上升压站数字式仪表读数的自动识别算法[J]. 红外与激光工程, 2021, 50(s2): 163.(TANG Peng, LIU Yi, WEI Hong-guang, et al. Automatic recognition algorithm of digital instrument reading in offshore booster station based on Mask-RCNN[J]. Infrared and Laser Engineering, 2021, 50(s2): 163.) [14] 党顺峰, 熊锐, 李继辉, 等. 基于改进Mask-RCNN算法的车位检测研究[J]. 现代制造工程, 2021, 43(1): 91.(DANG Shun-feng, XIONG Rui, LI Ji-hui, et al. A research on parking slot detection based on improved Mask-RCNN algorithm[J]. Modern Manufacturing Engineering, 2021, 43(1): 91.) [15] 林娜, 黄韬, 孙鹏林, 等. 优化Mask-RCNN的高分遥感影像建筑物提取[J]. 遥感信息, 2022, 37(3): 1.(LIN Na, HUANG Tao, SUN Peng-lin, et al. Building extraction of high-resolution remote sensing imagery on optimized Mask-RCNN[J]. Remote Sensing Information, 2022, 37(3): 1.) [16] LIN Y, ZHAO Q. Mask-RCNN with spatial attention for pedestrian segmentation in cyber-physical systems[J]. Computer Communications, 2021, 180(25): 109. [17] 刘子全, 付慧, 李玉杰, 等. 基于Mask-RCNN迁移学习的红外图像电力设备检测[J]. 数据采集与处理, 2021, 36(1): 176.(LIU Zi-quan, FU hui, LI Yu-jie, et al. Electrical equipment detection in infrared images based on transfer learning of Mask-RCNN[J]. Journal of Data Acquisition and Processing, 2021, 36(1): 176.) [18] 叶阳, 卢奇, 程时伟. 基于质心法的车联网目标跟踪方法与应用[J]. 计算机科学, 2021, 48(s2): 340.(YE Yang, LU Qi, CHENG Shi-wei. Centroid method based target tracking and application for internet of vechicles[J]. Computer Science, 2021, 48(s2): 340.) [19] 徐军, 张先锋, 杨帆, 等. 基于机器视觉的风力灭火机性能检测系统[J]. 控制工程, 2021, 28(2): 360.(XU Jun, ZHANG Xian-feng, YANG Fan, et al. Performance detection system of wind extinguisher based on machine vision[J]. Control Engineering of China, 2021, 28(2): 360.) [20] 蔡朕, 王景良, 吕梦樊, 等. 基于改进布谷鸟搜索算法的圆度误差评定[J]. 组合机床与自动化加工技术, 2020, 61(7): 40.(CAI Zhen, WANG Jing-liang, LÜ Meng-fan, et al. Roundness error assessment based on improved cuckoo search algorithm[J]. Modular Machine Tool and Automatic Manufacturing Technique, 2020, 61(7): 40.) [21] 雷贤卿, 张亚东, 马文锁, 等. 轴承滚子凸度轮廓的最小二乘拟合与误差评定[J]. 光学精密工程, 2018, 26(8): 2039.(LEI Xian-qing, ZHANG Ya-dong, MA Wen-suo,et al. Least square fitting and error evaluation of the convex contour of bearing roller[J]. Optics and Precision Engineering, 2018, 26(8): 2039.) [22] 王春光, 李俊杰, 王伟, 等. 基于机器视觉的机油冷却管圆度检测与分拣研究[J]. 机床与液压, 2022, 50(15): 31.(WANG Chun-guang, LI Jun-jie, WANG Wei, et al.Research on roundness detection and sorting of oil cooling pipe based on machine vision[J].Machine Tool and Hydraulics, 2022, 50(15): 31.)