Abstract:

In order to improve the precision of the seam-tracking system based on rotating arc sensor,an identification method of welding seam integrating the principal component analysis（PCA） and the relevance vector machine（RVM） is proposed,marked as PCA_RVM.In this method,first,the welding current signals are processed by using a wavelet filter,followed by the cycle partition and data normalization.Then,the data set of acquired welding seam offset is analyzed via PCA and is projected in low-dimension PCA space,and the low-dimension data set is used as the training data set of RVM.The proposed method is tested by some experiments.The results show that（1） the maximum error and mean error of PCA_RVM are respectively 0.54mm and 0.43mm;（2） the precision of PCA_RVM,which is better than those of the methods based on interval integral,neural network and support vector machine,is as high as RVM;and（3） the runtime of PCA_RVM is more than that of the method based on interval integral but is less than those of the methods based on neural network,support vector machine and RVM.It is thus concluded that PCA_RVM is more suitable for the seam-tracking system based on the rotating arc sensor.

Key words: rotating arc sensor, welding, principal component analysis, relevance vector machine