Prediction of Weld Widths for Laser-MIG Hybrid Welding Using Informer Model

In recent years, the field of industrial automation production has witnessed a growing interest in welding quality monitoring. Among the various factors influencing welding quality, weld width plays a crucial role as it serves as an indicator of the overall welding quality. The top weld width provides an intuitive measure of weld quality, whereas the back weld width reflects the degree of penetration. This article proposes a novel weld widths prediction approach for laser-melt inert-gas arc welding (MIG) hybrid welding. Visual information in laser-MIG hybrid welding process is captured by a high-speed imaging system. Despite the strong interference caused by the arc and reflected light, the U-Net model is employed to accurately segment the weld pool region from the welding images. Subsequently, image processing algorithms are used to extract the width of the weld pool. In addition, the 3-D scanning point cloud of the weld undergoes processing via the second-order differential method to acquire the measured value of weld width. Taking the width of weld pool as input, an Informer model is employed to predict both the top weld width and back weld width. Experimental results confirm the validity of the proposed approach in terms of accuracy and efficiency. The approach holds promise for enhancing welding quality monitoring in laser-MIG hybrid welding applications.