| 电弧焊熔透ICA-BP神经网络识别模型 |
| |
| 引用本文： | 高向东,林俊,萧振林,陈晓辉.电弧焊熔透ICA-BP神经网络识别模型[J].焊接学报,2016,37(5):33-36. |
| |
| 作者姓名： | 高向东 林俊 萧振林 陈晓辉 |
| |
| 作者单位： | 1.广东工业大学机电工程学院广东省计算机集成制造重点实验室, 广州 510006 |
| |
| 基金项目： | 国家自然科学基金资助项目(51175095);广东省协同创新与平台环境建设专项资助项目(2015B090901013);广东省重大科技专项资助项目(2014B090921008);广州市科学研究专项资助项目(1563000554);佛山市科技创新专项资助项目(2014AG10015) |
| |
| 摘 要： | 以氩弧焊熔透状态识别为研究对象,研究一种基于ICA (Imperialist Competitive Algorithm) 的BP(Back Propagation)神经网络识别模型方法. 首先利用ICA全局搜索不易陷入局部极值及搜索速度快的特点对神经网络权值和阈值初始化,再用BP算法对神经网络进行训练. 通过摄取焊接过程中的熔池图像,提取熔池面积、熔宽以及熔池质心位置作为神经网络预测模型的输入量,分析熔池图像三个特征与焊缝熔透状态的映射关系,最终建立熔透状态预测模型. 结果表明,采用ICA-BP神经网络能够有效地预测焊缝的熔透状态.
|
| 关 键 词： | 熔透 识别 ICA BP神经网络 |
| 收稿时间： | 2014/10/31 0:00:00 |
Recognition model of arc welding penetration using ICA-BP neural network |
| |
| GAO Xiangdong,LIN Jun,XIAO Zhenlin and CHEN Xiaohui.Recognition model of arc welding penetration using ICA-BP neural network[J].Transactions of The China Welding Institution,2016,37(5):33-36. |
| |
| Authors: | GAO Xiangdong LIN Jun XIAO Zhenlin and CHEN Xiaohui |
| |
| Affiliation: | 1.Guangdong Provincial Key Laboratory of Computer Integrated Manufacturing, School of Electromechanical Engineering, Guangdong University of Technology, Guangzhou 510006, China2.Guangzhou Panyu Gofront Dyeing & Finishing Machinery Manufacturer Ltd., Guangzhou 511400, China |
| |
| Abstract: | A BP neural network model based on ICA (Imperialist Competitive Algorithm) is proposed to recognize the arc welding penetration status. The weights and thresholds of the neural network are initialized using ICA which has the features of uneasy accessibility to local extremum and fast search speed. Then the BP algorithm is used to train the neural network. By capturing the images of the molten pool in welding process, three features of a molten pool image are processed. The features includes the weld pool area, weld pool width and the distance between the weld pool centroid and the bottom. These features are as the inputs of neural network to create the mapping relationship between the three features of molten pool and the weld penetration status, and eventually a predicted model of penetration status is established. Welding experimental results show that the welding penetration status can be accurately recognized using the ICA-BP neural network. |
| |
| Keywords: | penetration recognition ICA BP neural network |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《焊接学报》浏览原始摘要信息 |
|
点击此处可从《焊接学报》下载全文 |
|
