内蒙古电力技术

2024, 02, 78-83

基于深度学习的火电机组用钢金相组织评级方法研究

1.内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司，呼和浩特 010020 2.内蒙古农业大学机电工程学院，呼和浩特 010018

基金项目(Foundation): 国家自然科学基金项目“基于深度学习的火电厂的耐热钢显微组织金相图像分割方法研究”（52061037）；内蒙古电力（集团）有限责任公司内蒙古电力科学研究院分公司自筹科技项目“基于深度学习的金属监督设备金相组织智能评判研究”（2022-ZC-05）

邮箱(Email):

DOI: 10.19929/j.cnki.nmgdljs.2024.0029

27	0	979
下载次数	被引频次	阅读次数

引用本文下载本文

PDF

引用导出

GB/T 7714-2015 MLA APA Refworks EndNote NoteExpress NoteFirst

摘要全文参考文献出版信息相关文章

摘要：

针对火电机组用钢金相组织评级受人为因素影响，效率低、重复性差等问题，提出建立金相检验图像样本数据集，采用ConvNeXt-T卷积神经网络模型研究基于深度学习的火电机组用钢金相组织评级方法，同时用混淆矩阵对所建模型的性能进行评估，模型对铁素体+珠光体组织球化评级的准确率均值为98.7%、精确度均值为97.3%、灵敏度均值为97.2%、特异度均值为99.1%、F_1-Score均值为97.2%，表明该方法能够对火电机组用钢金相组织进行较为准确的评级，提升评级效率，为火电机组用钢金相组织智能评级提供一种新的方法，同时助力电力行业金相检验向数字化、智能化发展。

关键词： 金相组织; 智能评级; 深度学习; 卷积神经网络;

Abstract：

In response to the issues of susceptibility to human factors, low efficiency, and poor repeatability in metallographic structure rating, this paper uses metallographic images to establish a sample dataset, and uses ConvNeXt-T convolutional neural network model to study the deep learning based metallographic structure rating method for steel used in thermal power units. At the same time, the performance of the constructed model on the validation set is evaluated using a confusion matrix. The average accuracy rate, precision, sensitivity, specificity, and F_1-Score of the model for spheroidization rating of ferrite and pearlite structure are 98.7%, 97.3%, 97.2%, 99.1%, and 97.2%, respectively, which indicates that this method can accurately grade the metallographic structure of steel used in thermal power units, overcome human factors, improve rating efficiency, and form an objective evaluation, and provide a new method for intelligent grading of metallographic structure of steel used in thermal power units, and assist the power industry in moving towards digitalization and intelligence in metallographic examination.

KeyWords： metallographic structure; intelligent rating; deep learning; convolutional neural network;

参考文献

[1] 姚文德.20钢高温珠光体球化行为及对性能的影响[D].大连：大连理工大学，2022.

[2] WANG X, ZHANG W, GONG J, et al. Low cycle fatigue and creep fatigue interaction behavior of 9Cr-0.5 Mo-1.8 WV-Nb heat-resistant steel at high temperature[J]. Journal of Nuclear Materials, 2018, 505: 73-84.

[3] 杨波.火电锅炉末级再热器管开裂原因分析[J].机电工程技术， 2023，52（12）：273-276. YANG Bo. Analysis on the Cracking Causes of the Final-stage Reheater Tube of Thermal Power Boiler[J]. Mechanical & Electrical Engineering Technology, 2023, 52(12): 273-276.

[4] 王兴文，杨鹏，周正华.金相复膜技术在电站闸门缺陷诊断中的解析[J].云南电力技术，2023，51（1）：84-88. WANG Xingwen, YANG Peng, ZHOU Zhenghua. Analysis of Metallographic Replica Technology in Defect Diagnosis of Radial Gate of Power Station[J]. Yunnan Electric Power, 2023, 51(1): 84-88.

[5] 朱婷，张德屯，吕一楠，等.热网加热器管裂纹产生原因分析[J].电网与清洁能源，2021，37（10）：25-29，35. ZHU Ting, ZHANG Detun, LYU Yinan, et al. An Analysis on the Cause of the Heat Network Heater Tube Crack[J]. Advances of Power System & Hydroelectric Engineering, 2021, 37(10): 25-29, 35.

[6] 韩光海.电站锅炉12Cr1MoVG+TP347H异种钢接头管的开裂原因分析[J].机电工程技术，2020，49（3）：221-223. HAN Guanghai. Analysis of Cracking Causes of 12Cr1MoVG+ TP347H Dissimilar Steel Joint in Thermal Power Boiler[J]. Mechanical & Electrical Engineering Technology, 2020, 49(3): 221-223.

[7] 杨进文，余建国.基于深度学习的钢板表面缺陷识别方法及系统开发[J].机电工程技术，2024，53（1）：233-237，303. YANG Jinwen, YU Jianguo. Defects Recognition Method and System Development of Steel Surface Based on Deep Learning [J]. Mechanical & Electrical Engineering Technology, 2024, 53(1): 233-237, 303.

[8] 张艳飞，张永志，公维炜，等.基于GoogLeNet Inception V3模型的火电用钢金相组织智能识别[J/OL].材料导报:1-15[2024-01-25]. http://kns.cnki.net/kcms/detail/50.1078.TB.20231204.1612.025.html.

[9] AUST J, SHANKLAND S, PONS D, et al. Automated defect detection and decision-support in gas turbine blade inspection [J]. Aerospace, 2021, 8(2): 30.

[10] DOMíNGUEZ R G, GONZALEZ S J A, ROSADO C J, et al. Grain boundaries and phases identification of metallographic images by a normalized sobel operation and the edge thinning process for further numerical simulation[J]. Metals and Materials International, 2019, 26(9): 1-12..

[11] JOHANNES W, JESSICA G, DOMINIK B, et al. A new analysis approach based on Haralick texture features for the characterization of microstructure on the example of low-alloy steels[J]. Materials Characterization, 2018, 144: 584-596.

[12] 冯雨歆.基于多尺度神经网络与注意力机制的图像分割研究[D]. 西安：西安电子科技大学，2021.

[13] DA Costa F H, FUKUGAUCHI C S, DOS Santos Pereira M. Metallographic analysis of a TRIP 800 steel using digital image processing[J]. Materials Science Forum, 2015, 805: 236- 241.

[14] GOLA J, WEBEL J, BRITZ D, et al. Objective microstructure classification by support vector machine (SVM) using a combination of morphological parameters and textural features for low carbon steels[J]. Computational Materials Science, 2019, 160: 186-196.

[15] BORDAS A, ZHANG J, NINO J C. Application of Deep Learning Workflow for Autonomous Grain Size Analysis[J]. Molecules, 2022, 27(15): 4826.

[16] 张琦.基于集成学习的金相组织自动评级方法研究[D].镇江：江苏大学，2018.

[17] 包金叶.基于全卷积神经网络的金相图像自动晶界分割与评级方法研究[D].镇江：江苏大学，2020.

[18] 张佳宁.基于深度学习的金相晶粒度自动评级系统[D].北京：冶金自动化研究设计院，2021.

[19] 许帧英，张琦，朱建栋.基于纹理特征的集成学习金相自动评级方法[J].信息技术，2018（9）：71-74，78. XU Zhenying, ZHANG Qi, ZHU Jiandong. Metallographic automatic rating method based on ensemble learning with texture features[J]. Information Technology, 2018(9): 71-74, 78.

[20] 陈先昌.基于卷积神经网络的深度学习算法与应用研究[D].杭州：浙江工商大学，2014.

[21] GOPALAKRISHNAN K, KHAITAN S K, CHOUDHARY A, et al. Deep Convolutional Neural Networks with transfer learning for computer vision-based data-driven pavement distress detection[J]. Construction and Building Materials, 2017, 157(30): 322-330.

[22] 白格滔.基于深度学习复杂金相组织识别与评级方法的研究[D]. 呼和浩特：内蒙古农业大学，2023.

[23] 刘颖，雷研博，范九伦，等.基于小样本学习的图像分类技术综述[J].自动化学报，2021，47（2）：297-315. LIU Ying, LEI Yanbo, FAN Jiulun, et al. Survey on Image Classification Technology Based on Small Sample Learning[J]. Actaautomatica Sinica, 2021, 47(2): 297-315.

[24] KOO K M, CHA E Y. Image recognition performance enhancements using image normalization[J]. Human-centric Computing and Information Sciences, 2017, 7(1): 1-11.

[25] 赵立新，侯发东，吕正超，等.基于迁移学习的棉花叶部病虫害图像识别[J].农业工程学报，2020，36（7）：184-191. ZHAO Lixin, HOU Fadong, LYU Zhengchao, et al. Image recognition of cotton leaf diseases and pests based on transfer learning[J]. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36(7): 184-191.

[26] 刘冬冬，李友荣，徐增丙.选择性集成迁移算法在轴承故障诊断领域的应用[J].机械设计与制造，2020（5）：28-31，36. LIU Dongdong, LI Yourong, XU Zengbing. Application of Selective Ensemble Transfer Algorithms in the Field of Bearing Fault Diagnosis[J]. Machinery Design ＆ Manufacture, 2020, (5): 28-31, 36.

[27] 张永志，辛全忠，王永亮，等.基于迁移学习的钢金相组织分类与识别方法的研究[J].材料导报，2021，35（24）：152-157. ZHANG Yongzhi, XIN Quanzhong, WANG Yongliang, et al. Research on Classification and Recognition Method of Steel Metallographic Structure Based on Transfer Learning[J]. Materials Reports, 2021, 35(24): 152-157.

[28] 张永志，李旭英，辛全忠，等.自组织CNN建模识别耐热钢金相组织研究[J].材料导报，2022，36（12）：160-165. ZHANG Yongzhi, LI Xuying, XIN Quanzhong, et al. Research on self-organized CNN modeling and recognition of metallographic structure of heat-resistant steel[J]. Materials Reports, 2022, 36(12): 160-165.

基本信息:

DOI：10.19929/j.cnki.nmgdljs.2024.0029

中图分类号:

引用信息:

[1]张艳飞, 张永志, 白格滔等.基于深度学习的火电机组用钢金相组织评级方法研究[J].内蒙古电力技术,2024,42(02):78-83.DOI:10.19929/j.cnki.nmgdljs.2024.0029.

基金信息:

国家自然科学基金项目“基于深度学习的火电厂的耐热钢显微组织金相图像分割方法研究”（52061037）；内蒙古电力（集团）有限责任公司内蒙古电力科学研究院分公司自筹科技项目“基于深度学习的金属监督设备金相组织智能评判研究”（2022-ZC-05）

请选择需要下载的pdf数据

内蒙古电力技术

Summary