基于电流多特征融合的窄间隙P-GMAW摆动电弧传感焊缝跟踪方法OA北大核心CSTPCD

Arc sensing is one of the important methods to track the seam in narrow gap welding.In response to the problems of poor stabil-ity and low reliability,fusing multiple statistical features of current information in the swing cycle is proposed to overcome the problem that a single data feature is easily affected by the stability of the arc.Firstly,multiple time-domain features of the current signal are extracted,and the feature matrix is calculated to correlate with the deviation vector.Then,the features with high correlation rate are fused by using the method of principal component analysis,and the first two principal components are adopted as the observation observed data.Finally,a support vector machine model based on multiple classifications is used for the classification test.The test results show that the maximum error is 0.2 mm,and the error within 0.1 mm accounts for 93.75%of the total erro.The method has improved accuracy compared with the traditional method,and the training samples used are less and the training process is simpler compared with the neural network method.