基于深度卷积神经网络的钢材微观组织分类识别  

Classification and Identification of Steel Microstructure Based on Deep Convolution Neural Network

作　　者：段献宝 何惠珍 李平平 张志鹏 魏灏 黄铁 徐云涛 DUAN Xianbao;HE Huizhen;LI Pingping;ZHANG Zhipeng;WEI Hao;HUANG Tie;XU Yuntao(School of Materials Science and Engineering of Wuhan Institute of Technology,Wuhan 430205,China;CRRC Qishuyan Locomotive&Rolling Stock Technology Research Institute Co.,Ltd.,Changzhou 213011,China)

机构地区：[1]武汉工程大学材料科学与工程学院,湖北武汉430205 [2]中车戚墅堰机车车辆工艺研究所有限公司,江苏常州213011

出　　处：《铁道车辆》2022年第1期43-47,共5页Rolling Stock

摘　　要：通过引入多种深度卷积神经网络及分类器构建机器学习模型,对钢材金相图数据集进行学习,研究了一种能够准确、高效识别钢材微观组织的方法。研究结果表明,文章中所涉及的3种深度卷积神经网络在钢材微观组织的分类识别上均表现出优异性能,其中Inception-V3表现尤为突出,其与人工神经网络分类器组合而成的机器学习模型的分类精度可达99.60%。A machine learning model is constructed by introducing a variety of deep convolution neural networks and classifiers to learn the metallographic data set of steel,and a method that can accurately and efficiently identify the microstructure of steel is studied.The results show that the three deep convolutional neural networks involved in this paper have excellent performance in the classification and identification of steel microstructure,especially Inception-V3.The machine learning model combined with the artificial neural network classifier can achieve 99.60%classification accuracy.

关 键 词：金相图 微观组织 分类识别 机器学习 卷积神经网络 

分 类 号：TG115.21[金属学及工艺—物理冶金]