基于CNN的枪械关重件复杂加工特征识别  

Recognition of complex processing features of firearm key important parts based on convolutional neural network

作　　者：邵海瑞 刘伊华 王希阔 王永娟[1] SHAO Hairui;LIU Yihua;WANG Xikuo;WANG Yongjuan(School of Mechanical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;The 208 th Research Institute of China Ordnance Industry,Beijing 102202,China;The 63856^(th) Troop of PLA,Baicheng 137001,China)

机构地区：[1]南京理工大学机械工程学院,南京210094 [2]中国兵器工业第208研究所,北京102202 [3]中国人民解放军63856部队,吉林白城137001

出　　处：《兵器装备工程学报》2022年第8期304-310,共7页Journal of Ordnance Equipment Engineering

基　　金：基础科研项目(JCKY2018209A001)。

摘　　要：加工特征识别是CAD、CAPP以及CAM之间的智能化接口,对实现CAD、CAPP、CAM集成具有重要意义。原有的加工特征识别技术局限于对零件三维拓扑结构的属性邻接图进行操作,对于一些复杂加工特征识别准确率低、效率低。针对此问题,提出一种基于卷积神经网络(CNN)的复杂加工特征识别方法。对枪械关重件的结构特点和机加工艺过程进行分析并分类,采用复杂加工特征多角度捕捉图像法创建复杂加工特征数据集,通过复杂加工特征卷积神经网络模型训练数据集,并利用复杂加工特征分类器进行分类,输出复杂加工特征结果。结合UG平台以枪械某关重件为例验证了复杂加工特征识别过程。Processing feature recognition is an intelligent interface between CAD,CAPP and CAM,which is of great significance to the integration of CAD,CAPP and CAM.The original processing feature recognition technology is limited to the operation of the attribute adjacency graph of the three-dimensional topological structure of the part,and the recognition efficiency and accuracy of some complex processing features are low.In response to this problem,a complex processing feature recognition method based on Convolutional Neural Networks(CNN)was proposed.The structural characteristics and machining process of the firearm key important parts were analyzed and classified.Using the complex processing features,a multi-angle capture image method was adopted to create the data sets of complex processing features,then a convolutional neural network model of complex processing features was used to train the data sets,and a complex processing feature classifier was used for classification,and finally output the complex processing feature results.Combined with the UG platform,the complex processing feature recognition process was verified by taking firearm key important parts as an example.

关 键 词：枪械关重件 复杂加工特征 卷积神经网络 图像识别 

分 类 号：TJ205[兵器科学与技术—武器系统与运用工程]