Visual Sensing And State Prediction Of Fusion Hole In TIG Backing Welding With Y-groove

The fusion hole is the specific phenomenon in backing welding,and it forms by arc fully melted the grooves on both sides of the root gap.The fusion hole is a semi-open hole and its existence ensures the full penetration of the backing welding.By observing the shape of the fusion hole,adjusting the welding torch posture,and controlling the shape of the fusion hole online,experienced welders can obtain high-quality welds.By using the vision sensor to shoot and process the images of the welding area,the dynamic information of the fusion hole can be obtained.Studying the relationship between dynamic information and fusion hole behavior can lay the foundation for real-time control of the backing welding.This paper builds a TIG welding test platform centered on the welding system,visual sensing system,and motor control system.The visual sensing system controls the two CMOS cameras to shoot the front and back fusion hole images.The motor control system can adjust process parameters such as welding speed and wire feed speed online.Based on the acquired fusion hole image,the gray distribution characteristics of different regions in the fusion hole image are analyzed.Based on the characteristics of fusion hole edge,root gap edge,and arc area distribution,MATLAB toolbox was used to design an appropriate image processing algorithm,and a complete fusion hole-gap edge was obtained.Through the welding test of 6mm low carbon steel backing with constant process parameters,the behavior and mechanism of the fusion hole were studied.This study found that the formation of the fusion hole is affected by the filler metal and the arc heat-force.When the smaller of the filler metal and the stronger of the arc heat-force,the fusion hole is formed more easily.Before the fusion hole instability,the fusion hole is large and tends to be eccentric to one side.Some parameters are used to characterize the stability of the fusion hole,including the difference value between the width of the fusion hole and root gap and the offset of the fusion hole center.Secondly,the influence of the main process parameters on the size of the fusion hole was studied.The size of the fusion hole is positively correlated with the welding current and root gap but negatively correlated with wire feeding speed.Through the dynamic process parameter tests,the behavior of the fusion hole was studied in dynamic process parameters.Studies have shown that wire feeding speed is more suitable than welding speed as a control variable for online adjustment of the fusion hole.For the two types of images of fusion hole formation and non-formation,the HOG algorithm and image processing algorithm are used to obtain feature information.HOG algorithm obtains grayscale feature information and the image processing algorithm obtains arc feature information in the fusion hole image.The feature information obtained by the HOG algorithm and image processing algorithm was used to establish two SVM fusion hole formation prediction models.Comparing the two prediction models,it is found that the performance of the SVM model based on arc characteristic parameters is slightly better than HOG-SVM.However,the HOG-SVM method can accurately distinguish the boundary of fusion hole formation and is more suitable for the establishment of the fusion hole formation prediction model.It is found that the gray characteristics of the upper wall of the fusion hole change obviously before the instability of the fusion hole.For the fusion hole instability images and stable images,a HOG-SVM fusion hole instability prediction model is established,and the prediction model can provide "early warning"in advance.