基于支持向量机(SVM)的点焊质量在线监测  

On-line quality inspection of spot welding based on support vector machine

作　　者：张宏杰[1] 李雅峰[1] 侯妍妍[2] 

机构地区：[1]天津工业大学机电学院,300160 [2]河北省行政学院,石家庄市050031

出　　处：《焊接》2007年第10期33-37,共5页Welding & Joining

摘　　要：电阻点焊过程动态信号蕴含大量直接或间接反映焊点质量的动态信息,通过对焊接过程电极位移、动态电阻信号的同步采集和分析,从两种信号中提取了12个特征参量,建立表征点焊过程的数据集,以焊点抗剪强度作为焊点质量评价指标,利用支持向量机(SVM)统计学习方法,将焊点试样动态监测参量与焊点抗剪强度之间低维的复杂非线性映射关系,映射到一个高维的特征空间(Hilbert空间),原试样数据空间的非线性关系相应变化为高维特征空间的线性关系,在不增加计算复杂度的情况下,实现对未知焊点试样抗剪强度的分类及预测。SVM测试结果表明,支持向量机在小样本情况下具有较好的泛化能力,分类、预测速度快,准确率高,能较为满意地完成焊点强度的分类、预测任务,可以作为进一步研究的方法。Lots of dynamic information,which can directly or indirectly reflect the quality of welding spot,are included within the dynamic signals of resistance spot welding. The electrode displacement and dynamic resistance signals are synchronously gathered and analyzed. 12 characteristic parameters are picked up from the two signals to set up a set of data which characterizes the process of spot welding. By means of SVM( support wector machine) statistics theory,the sophisticated lower dimension mapping equation for describing the relationship of welding process inspection parameters to welding spot strength is changed into a higher dimension one called Hilbert space,when taking the shear strength of the welding spot as the estimation index of welding spot quality. The original nonlinear relation can be replaced by linear one in high space. Under not increasing the calculational complexity circumstance,the classification and prediction rule of the welding spot can be realized. The support vector machine test showes that the SVM provides better data generalization ability on the case of small samples,and can satisfactorily accomplish the classification and prediction of the welding spot quickly and efficiently. So it can be regarded as the forward research method.

关 键 词：电阻点焊 统计学习理论 支持向量机 

分 类 号：TG453.9[金属学及工艺—焊接]