摘要
中走丝机床放电加工过程中,放电状态的检测是数控系统中的一个极其重要环节,而传统的检测是基于放电电压的平均值对其辨识,但是它对过渡电弧放电和电弧放电等状态检测不敏感。针对上述问题,提出一种基于深度网络学习的机器视觉方法,首先将放电状态的电压信号转换为灰度图像;其次对不同周期的放电图像压缩到统一的图像尺寸中(256×256像素);最后基于Tensorflow并行计算框架,采用K邻近方法、Logistic Regression方法和深度网络学习三种方法对放电状态进行辨识。结果表明深度网络学习方法最好,准确率最优为95.63%,这对中走丝机床放电状态传感器的设计具有重要的指导意义。
During the process of MS-WEDM,the detection of discharge state is very important in the numerical control system. The traditional detection is based on the average value of discharge voltages. However,it is not sensitive to the transition arc or arc. To solve this problem,this paper proposes a machine vision method based on deep network learning.First,the discharge voltage signal is converted to gray level images,and then the discharge images of different periods are compressed to a unified image size(256*256 pixels). Finally,based on the Tensorflow parallel computing framework,the Knearest neighbor,regression logistic and deep network learning methods are used to identify the discharge status. The results show that the deep network learning method is optimal and the accuracy is 95.63%,which is of important guiding significance for the design of the discharge state sensor of MS-WEDM.
作者
沈娣丽
刘冬敏
都金光
明五一
SHEN Di-li;LIU Dong-min;DU Jin-guang;MING Wu-yi(School of Mechanical-electronic and Automobile Engineering, Zhengzhou Institute of Technology, He’nan Zhengzhou 450052, China;Mechanical and Electrical Engineering Institute, Zhengzhou University of Light Industry, He’nan Zhengzhou 450002, China)
出处
《机械设计与制造》
北大核心
2019年第8期146-149,共4页
Machinery Design & Manufacture
基金
河南省科技厅基础前沿基金资助项目(16230041 0039)
郑州轻工业学院博士基金资助项目(2014BSJJ 024)
