Online monitoring of surface forming quality in laser-arc hybrid butt welding with uneven joint gaps based on arc signal diagnosis

Abstract

High-power laser-arc hybrid welding is commonly employed in the welding of medium-thickness steel. In full-penetration butt welding, butt joint gaps are particularly prevalent, especially as plate thickness increases, which in turn tend to exacerbate underfill and surface collapse defects. Consequently, it is essential to monitor the surface collapse defects during the welding process for medium-thickness steel. In this paper, a novel method is proposed to monitor surface collapse defects for laser-arc hybrid welding of medium-thickness steel with uneven joint gaps based on voltage signal diagnosis and machine learning. Firstly, the interrelation mechanism between the butt joint gap, surface collapse, and the voltage signal is analyzed with high-speed images, which is adopted to guide the time/frequency feature extraction in the proposed method. Then, it is found that the changes in peak and base voltages, along with the instantaneous frequency of arc pulse captured through variational mode decomposition (VMD), reflect the surface collapse defect, significantly affected by the butt gap. On this basis, a back propagation neural network (BPNN) is constructed to identify the extracted features, realizing the monitoring of the surface collapse defects with a precision of 94.24%.