核主成分分析法的MIG焊电弧声信号特征选择  

Feature Selection of Arc Sound in MIG Welding Based on Core Principal Component Analysis

作　　者：毕淑娟[1,2] 韩玉杰[1] 兰虎[3] 刘立君[4] 

机构地区：[1]东北林业大学机电工程学院,黑龙江哈尔滨150040 [2]哈尔滨学院理学院,黑龙江哈尔滨150086 [3]哈尔滨理工大学荣成学院,山东荣成2643004 [4]哈尔滨理工大学材料科学与工程学院,黑龙江哈尔滨150040

出　　处：《哈尔滨理工大学学报》2011年第5期30-33,共4页Journal of Harbin University of Science and Technology

基　　金：黑龙江省自然科学基金(E2007-01)

摘　　要：以不同焊接状态电弧声信号识别为目的,从MIG焊平板对接射流过渡电弧声信号入手,探寻电弧声产生机理和传播特性.提出电弧声是气体放电的伴生物,是等离子体集体振荡以波形方式传播的结果,也是声源和声道共同作用的产物,其频谱特性主要取决于声道的频率响应.在此基础上,采用线性预测分析技术,提取了表征焊缝熔透状态的10维电弧声信号特征向量,并采用核主成分分析技术成功进行了特征级参数的融合,实现了原始特征空间压缩,最终以2维合成向量取代高维特征向量,为后续利用电弧声信号实现MIG焊典型熔透状态的分类提供了技术支持.Aiming at the recognition of the arc sound signal in different welding status,the mechanism of arc sound generation and propagation are investigated from MIG butt welding with spray transfer.It is proposed that arc sound is an associated thing of gas discharge,the result of transmitting by the wave when the plasma collectively vibrates,and the outcome of their combined action,including the sound source and the sound track.Among them,the spectrum characteristic chiefly depends on the frequency respond of sound track.Based on the above mentioned,the 10-demensional engenvector were extracted with linear prediction analysis,which can characterize weld penetration status.Subsequently,the feature-level parameters were successfully fused utilizing the concept of core primary principal component analysis.Ultimately,the high-dimensional engenvector was replaced by the synthesis of 2-demensional vector,which achieves compression for feature space and provides technical supports for pattern classification of typical penetration status with the help of arc sound signal in MIG welding in the future.

关 键 词：核主成分分析 电弧声 线性预测分析 特征提取 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]