All-position TIG Welding Penetration Monitoring And Visual Information Characterization Of Nuclear Power Thick-walled Pipe

As the development of welding automation and welding intelligence,all-position automatic pipe welding is widely used in nuclear pipeline construction.However,due to welding quality influencing factors such as assembly errors,machining accuracy and thermal deformation,it is difficult to guarantee pipe welding quality.Nuclear power pipe has high welding quality requirements and the formation of back weld bead determines the welding quality,but it is difficult to directly observe the formation of back weld bead due to the constraints of welding technology and on-site construction.Thanks to rapid development of visual sensing and artificial intelligence technology,this article has carried out research on the real-time sensing and information processing of back molten pool and solidification weld bead.Based on the real-time back penetration monitoring of narrow-gap TIG welding for nuclear power pipe,this paper developed a real-time inner monitoring systems,including a pipe inner walking mechanism that rotates synchronously with an external welding gun,a backside molten pool vision sensor and a solidification weld bead scanner,realizing real-time collection and processing of back molten pool and solidified bead information.Firstly,an all-position narrow-gap pulse TIG welding experiment was carried out,capturing back molten pool image and solidification weld bead profile.Based on force state of molten pool,the characteristics of back molten pool at different welding position were analyzed.Secondly,based on the collected profile data of back solidified weld bead,a profile feature extraction algorithm was developed to visually characterize geometric size of back weld bead at different weld position.Before welding,the scanner captures back bevel profile and obtains back gap and misplacement geometry using X-axis difference of profile points.During pipe welding,the geometry size of solidified weld bead can be obtained using bead feature extraction algorithm based on difference between weld bead and base metal.The experiments show that the feature extraction algorithm has high accuracy and robustness,which can characterize the geometric feature of back weld bead.Next,a high-speed CMOS camera was used to obtain back molten pool image in real-time during all-position narrow-gap pulse TIG welding of nuclear pipe.Analyze the dynamic back pool image during different welding position,further develop an image feature extraction algorithm based on back pool feature.The Otsu method,a Gaussian mixture model and a convolutional neural network model were used to segment the pool area,and the back molten pool feature and geometric size processing were completed,then feedback to real-time monitoring interface.Finally,combined with back molten pool image and back weld bead characteristic,the force state of all-position welding pool were discussed,analyzing the evolution of dynamic back weld pool from mechanism level.The result shows that the gravity of molten pool in all-position narrow-gap pulse TIG welding is the key factors that affects the shape of molten pool.The formation of back weld bead depends on the accumulated of fused metal.In all-position welding,we must pay attention to the influence of gravity and improve the welding quality.