基于卷积神经网络的CO_2焊接熔池图像状态识别方法  

Recognition of molten pool morphology in CO_2 welding based on convolution neural network

作　　者：覃科[1] 刘晓刚[1] 丁立新[2] 

机构地区：[1]桂林航天工业学院广西高校机器人与焊接技术重点实验室培育基地,广西桂林541004 [2]武汉大学软件工程国家重点实验室,武汉430072

出　　处：《焊接》2017年第6期21-26,共6页Welding & Joining

基　　金：广西自然科学基金资助项目(2014GXNSFAA1183105;2016GXNSFAA380226)

摘　　要：为了通过熔池图像对焊接状态进行判断,将卷积神经网络引入到CO_2焊接熔池图像状态识别中,提出了一种CO_2焊接熔池状态识别卷积神经网络CNN-M。该网络使用简单预处理的熔池图像作为输入向量,避免了人工提取图像特征的主观性对识别率的不良影响。同时,CNN-M采用了ReLU激活函数、随机Dropout及SVM分类器来降低样本集稀少可能导致的网络过拟合现象。试验结果表明,和人工提取熔池特征状态作为输入向量的BP神经网络相比,CNN-M在识别率及识别速度方面均体现出了更好的性能,其良好的泛化能力能够满足在线熔池状态监控的要求。A kind of convolution neutral network CNN-M was proposed to recognize the morphology of molten pool in CO_2 welding.A simple pretreatment of the molten pool image was adopted as the input vector in CNN-M,which avoids the adverse effects brought by the subjectivity of artificial extraction of molten pool image features.In order to decrease the possibility of network over-fitting caused by the sparse training data set,several methods including ReL U activation function,random Dropout and SVM classifier were used in CNN-M.The experimental results show that CNN-M has better performance in recognition rate and recognition speed than the BP neural network does,which adopt the characteristic values of the molten pool as the input vector.The performance of CNN-M is able to meet the requirement of the on-line molten pool monitoring.

关 键 词：焊接熔池 卷积神经网络 状态识别 

分 类 号：TG409[金属学及工艺—焊接]