电阻点焊压痕深度的SVM回归预测  

Predication of Regression Indentation Depth in RSW Based on SVM

作　　者：张鹏贤[1,2] 冯毅 刘兴刚[1,2] 

机构地区：[1]兰州理工大学 有色金属合金及加工教育部重点实验室,甘肃兰州730050 [2]兰州理工大学 甘肃省有色金属新材料省部共建国家重点实验室,甘肃兰州730050

出　　处：《热加工工艺》2012年第17期130-133,共4页Hot Working Technology

基　　金：国家自然科学基金资助项目(50275028);兰州理工大学科研基金资助项目(hz0901005)

摘　　要：针对电阻点焊压痕深度的检测存在离线、滞后等问题,提出了一种基于电极位移信号特征提取的人工智能在线预测压痕深度的实现方法。首先,采用搭建的计算机激光测量系统探索了焊点压痕深度测量方法,确定了以多次测量的平均值hT作为压痕深度的实际评定值。其次,通过对熔核形成过程、电极位移信号与焊点压痕深度的相关性研究,确定了焊接电流I、电极压力F、以及从电极位移信号中提取的特征参量h作为压痕深度的表征参量。最后,采用压痕深度的表征参量作为输入向量,以测定的焊点实际压痕深度hT作为目标向量,建立了SVM(supportvectormachine)回归预测模型。实际测试表明,模型输出的压痕深度预测值和实际测定值间的线性相关度达到了91.18%,通过实时监测熔核形成过程,可以实现焊点压痕深度的预测。Off-line and delay are the main questions in measuring the indentation depth of welding spot in RSW. To overcome the questions, an artificial intelligence method for prediction indentation depth is put forward based on the electrode displacement signal characteristic. Firstly, a measure method is explored for indentation depth by a computer laser measure system. In the basis, the average value of the measure h~ is selected as the actual judgment value. Secondly, through a correlational study between electrode displacement signal and welding spot indentation depth, the measure parameters of indentation depth h, welding current I and electrode force F are determined as character parameters. Lastly, a model of support vector machine （SVM） for predicting the the regression indentation depth was established between the input vector I, F, h and the target vector hr. The results show, the correlation coefficient between the model output and actual measuring values are 91.18%. The model can realize the estimate of the indentation depth by monitoring the process of nugget formation.

关 键 词：电极位移曲线 压痕深度 支持向量机 回归预测 

分 类 号：TG453.9[金属学及工艺—焊接]