The Feature Vector Of Welding Electrical Parameters In Seam Tracking Of Robot-based Weaving Welding

Under the condition of arc welding robot-based seam tracking, the weaving frequency and amplitude, arc voltage, root gap as well as with or without seam tracking, are all the factors affecting the weld penetration. Thus, to find the relationship between the process conditions and influencing factors of weld penetration, it is of great theoretical and practical significance in oscillation seam tracking welding.In this paper, 48 kinds of experiments, under the condition of arc welding robot-based seam tracking, were conducted based on Vee groove of thick mild steel plates which is widely used in engineering construction equipment. The test samples are made, and the corresponding macrophotographies of weld at cross section were obtained. The weld penetration, deposit area and the cross section area of the welds at different process conditions are measured. According to the experiment data, the suitability of the pre-selected welding process parameters and the groove shape are testified, and the correlation between weld penetration depth and root gap & welding process parameters is studied quantitatively.In the welding process, the electrical signals (welding current and arc voltage) are measured in real-time and processed off-line. The electrical signals are processed by two kinds of statistical methods, first the transient electrical signals are processed into a 10-dimensional feature vector M10. On the other hand, the electrical signals are firstly processed into probability density distribution values (PDD's), and then the PDD values are transformed to another 10-dimensional feature vector S10. Both feature vectors M10 and S10 are set up for describing the welding process, respectively the marked differences among the feature vectors (M10 and S10) under different welding conditions were found. It indicates the existence of correlation between welding process conditions andthe feature vectors.Five kinds of welding experiments with different levels of weaving amplitude and frequency were carried out under the condition of arc welding robot-based seam tracking on V groove of 10 mm-thick mild steel plates. The transient electrical signals were statistically processed into the 10-D feature vector S10. Employing the fuzzy pattern recognition technique, the relationship between the feature vector and the weaving amplitude & frequency is developed.