高级检索
高理文, 薛家祥, 陈辉, 王瑞超, 林放. 基于相关与时耗复合维归约的弧焊电源动特性自适应在线监测[J]. 焊接学报, 2012, (4): 17-20.
引用本文: 高理文, 薛家祥, 陈辉, 王瑞超, 林放. 基于相关与时耗复合维归约的弧焊电源动特性自适应在线监测[J]. 焊接学报, 2012, (4): 17-20.
GAO Liwen, XUE Jiaxiang, CHEN Hui, WANG Ruichao, LIN Fang. Adaptive online detection on dynamic characteristics of arc welding power supply based on complicated dimensionality reduction of correlation and time consumption[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION, 2012, (4): 17-20.
Citation: GAO Liwen, XUE Jiaxiang, CHEN Hui, WANG Ruichao, LIN Fang. Adaptive online detection on dynamic characteristics of arc welding power supply based on complicated dimensionality reduction of correlation and time consumption[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION, 2012, (4): 17-20.

基于相关与时耗复合维归约的弧焊电源动特性自适应在线监测

Adaptive online detection on dynamic characteristics of arc welding power supply based on complicated dimensionality reduction of correlation and time consumption

  • 摘要: 提出了相关与时耗复合维归约方法,可实现多种弧焊电源动特性自适应在线监测.其核心思想是从一个大的特征库中选择出最贴近监测对象的若干特征.该方法充分考虑特征间的相关性,以及在线监测的时效性.搭建了较为完善的焊接试验数据采集平台,共采集189次焊接过程的电压电流数据作为样本,并以人工评定结果作为文中维归约方法教师信号,即类的标签.随机选择150个样本组成训练集,剩余39个组成测试集.运用文中维归约方法的寻找最优的特征子集.结果表明,找到的最优特征子集的自动化评定准确率达97.4359%,接近应用要求.

     

    Abstract: A complicated dimensionality reduction of correlation and the time consumption was put forward,which realized the adaptive online detection on different dynamic characteristics of arc welding power supply.The main idea of this method was to select some certain features from the complete feature set which were the closest to the detection targets.The correlation among features and the efficiency of the online detection were fully taken into account in the use of this method.A perfect welding data collection platform was set up,the samples of the voltage and current data were collected in the 189 welding processes,and the artificial evaluation results were taken as the teacher's signals,namely the cluster labels,in the dimension reduction.150 samples were randomly selected as training set,while the remaining 39 samples were used as the test suite.The results of the experiments showed that the automatic evaluation accuracy of the chosen one reached 97.435 9% and satisfied the application requirements when the optimal feature subset was chosen based on the dimension reduction method.

     

/

返回文章
返回