基于卷积神经网络的电熔镁砂块料识别与分类  

Identification and classification of fused magnesia blocks based on convolutional neural network

作　　者：赵鑫 岳静静 游杰刚 张小芳 冯东 侯庆冬 张玲 邓承继[2] Zhao Xin;Yue Jingjing;You Jiegang;Zhang Xiaofang;Feng Dong;Hou Qingdong;Zhang Ling;Deng Chengji(School of Materials and Metallurgy,University of Science and Technology Liaoning,Anshan 114051,Liaoning,China;不详)

机构地区：[1]辽宁科技大学材料与冶金学院,辽宁鞍山114051 [2]武汉科技大学省部共建耐火材料与冶金国家重点实验室,湖北武汉430081

出　　处：《耐火材料》2024年第6期517-520,共4页Refractories

基　　金：国家自然科学基金资助项目(U20A20239,U1908227)。

摘　　要：电熔镁砂的人工分选过程造成电熔镁砂的分类精准度不高,且工作环境恶劣。人工智能的图像识别技术具有高效性和可靠性等优点,采用人工智能的方式对电熔镁砂进行识别和分类将可以很好地解决这一问题。基于卷积神经网络的图像识别技术,通过学习大量电熔镁砂的样本宏观特征后,对不同品种进行了识别和分类,并对高钙电熔镁砂图像进行训练150次迭代。结果表明:运用卷积神经网络对高钙电熔镁砂图像进行训练,在104次迭代时的训练精度最高,高达97.2%。在实际识别中6种不同氧化镁含量的高钙电熔镁砂的预测概率都达到了99.4%以上,不仅可以减少人工,还提高对电熔镁砂的分类效率和识别准确度。The manual sorting process of fused magnesia results in low classification accuracy and has harsh working environment.The image recognition technology of artificial intelligence has the advantages of high efficiency and reliability,so the identification and classification of fused magnesia by artificial intelligence is a good solution to this problem.Based on the image recognition technology of convolutional neural network,after collecting the macroscopic characteristics of a large number of fused magnesia samples,different fused magnesia varieties were identified and classified,and 150 iterations of high calcia fused magnesia images were trained.The results show that using convolutional neural network to train the high calcia fused magnesia image,and the training accuracy is the highest at the 104^(th) iteration,reaching 97.2%.In actual identification,all the prediction probabilities of six fused magnesia samples are more than 99.4%,which can not only reduce labor,but also improve the classification efficiency and identification accuracy for fused magnesia.

关 键 词：电熔镁砂 宏观特征 卷积神经网络 识别和分类 

分 类 号：TQ175[化学工程—硅酸盐工业]