关键词: 焊缝, 偏差识别, 小波滤波, Laplace特征映射, 外延算法, 支持向量回归机

Abstract:

In order to improve the identification precision of welding seam offset,first,the welding current signals based on the rotational arc sensor are filtered by wavelet,followed by the reconstruction of a sample data set via the pretreatment.Next,an extension algorithm of Laplace feature mapping is proposed based on the support vector regression（SVR） machine,which is applied to the dimensionality reduction of the sample data set and the new sample.Then,the sample data set after the dimensionality reduction is used to train the SVR machine and identify the offset for the new sample. Finally, the proposed identification method is compared with the traditional method without dimensionality reduction. Experimental results indicate that the dimensionality reduction based on Laplace feature mapping may result in an average increase of identification precision by 25 %.

Key words: welding seam, offset identification, wavelet filtering, Laplace feature mapping, extension algorithm, support vector regression machine