基于多模态信息融合的分流孔分类方法  

Porthole classification method based on multimodal information fusion

作　　者：周钢亮 秦昊 林卓胜 张昱 ZHOU Gangliang;QIN Hao;LIN Zhuosheng;ZHANG Yu(Faculty of Intelligent Manufacturing,Wuyi University,Jiangmen 529020,China;Guang-dong Institute of Intelligent Manufacturing,Guangzhou 510075,China)

机构地区：[1]五邑大学智能制造学部,广东江门529020 [2]广东省科学院智能制造研究所,广东广州510075

出　　处：《模具工业》2023年第10期20-25,共6页Die & Mould Industry

基　　金：广东省重点领域研发计划项目(2021B0101220006)。

摘　　要：为探究机器学习方法在挤压模设计智能化上的应用,结合型材截面图与数值分析,提出基于多模态信息融合的分流孔分类模型(CMMIF)。CMMIF利用卷积神经网络(CNN)实现型材面图像特征抽取,利用ReliefF算法对数值数据进行特征筛选,将2种不同的图像特征与数值特征进行融合。试验结果表明,与同基线模型对比CMMIF能有效表示分流孔特征,分类模型F1值为91.83%,实现对多种模态信息的准确分类,该研究成果为实现挤压模分流孔的智能设计提供支撑。In order to explore the application of machine learning method in the intelligent design of extrusion die,combined with the profile section and numerical analysis,a porthole classifica⁃tion model based on multimodal information fusion(CMMIF)was proposed.The CMMIF model uti⁃lized convolutional neural networks(CNN)to extract features for profile cross-section image fea⁃ture extraction,used the ReliefF algorithm to carry out feature screening of numerical data,and two different image features were fused with numerical features.The experiment results showed that the CMMIF model could adequately represent the characteristics of the porthole compared with the same baseline model,the F1 value of the classification model was 91.83%,and the accu⁃rate classification of multimodal information was realized,which provided strong support for the intelligent design of extrusion die porthole.

关 键 词：挤压模设计 机器学习 智能设计 多模态融合 

分 类 号：TG375.41[金属学及工艺—金属压力加工]