点焊接头抗剪强度的SVM回归预测  

SVM regression prediction of shearing strength of spot weld nugget

作　　者：张宏杰[1] 周超[1] 侯妍妍[2] 

机构地区：[1]天津工业大学机械电子学院,天津300160 [2]河北省行政学院建筑工程系,石家庄050031

出　　处：《天津工业大学学报》2008年第4期65-69,共5页Journal of Tiangong University

摘　　要：通过对同步采集的点焊焊接过程动态电阻、电极位移信号的特征分析,提取出若干特征参量,依据所提取特征参量与焊点接头抗剪强度之间的相关性分析结果,选取相关系数较大的6个特征参量建立数据集作为SVM的输入,分别利用线性核函数、多项式核函数、RBF核函数实现监测参量与焊点强度之间的映射.采用交叉有效性检验对不同核函数下预测结果进行比较,结果表明,RBF核函数支持向量机各项预测误差最小,有较好的应用前景.Through the feature analysis of synchronously-gathered dynamic resistance and electrode displacement signals of welding process, several characteristic parameters are extracted. Based on the analysis result of correlation between extracted characteristic parameter and welded spot strength, six characteristic parameters which have big correlation coefficient are selected to set up a set of data as the input of the SVM. The sophisticated mapping of monitor parameters and welded spot strength is realized by means of linear kernel function, polynomial kernel function and RBF kernel function respectively. The comparison of predictive results under different kernel functions by means of cross-validation estimation shows that SVM of RBF kernel function has the smallest predictive error and has a better prospect of application in the future.

关 键 词：电阻点焊 相关性分析 支持向量机 回归分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] U463.82[自动化与计算机技术—计算机科学与技术]