Research On Visual Sensing And Tracking Detection System For Narrow Gap Swing GTAW Vertical Welding Process

Narrow Gap Swing Tungsten Inert Gas Welding(NGS-GTAW)is widely used for large structural components with high requirements for comprehensive mechanical properties.However,its application in large-scale structural welding with a significant amount of pipeline systems and thicker workpiece thicknesses poses challenges due to limited technological means for intelligent sensing and control of the swing welding process,lack of self-adaptive capability for groove processing errors,and inability to achieve adaptive control under conditions such as welding gun swing,variable groove forms,and thermal accumulation deformation.This limits the automation and intelligence level of the NGS-GTAW welding process,leading to difficulties in ensuring stability and quality of welding results,thereby hindering its widespread application in engineering.This article addresses the challenges of adapting welding structures in the single-layer and multi-pass NGS-GTAW welding process under complex working conditions.A sensor detection system based on integrated passive visual information technology is developed to establish an accurate,versatile,and highly robust welding seam tracking detection system.The system’s ability to detect and control in actual working conditions has been verified.In order to steadily obtain the characteristic information of welding torch position in narrow gap bevel and swing state,the passive vision sensor is used to obtain the welding images when the torch stays on the left and right sides in one swing cycle in real time,and the most suitable image pre-processing parameters are calculated by image grayscale information features,and the welding images with different brightness caused by random changes of arc light are processed adaptively so as to eliminate noise interference and accurately enhance the characteristic information of welding images from different sources,and then the enhanced characteristic edges and contours are detected by linear detection and contour analysis methods to realize the parameterization of position characteristic information such as bevel edge,tungsten tip edge and melt pool contour.In order to improve the accuracy,real-time and robustness of the feature information detection method,according to the characteristics of the weld image feature distribution,the feature detection interest area applicable to different bevel sizes and different weld layers is set,and in order to adapt to changes in the position of the weld features in the field of view and ensure that the detection interest area can be accurately arranged in the weld area feature position in real time,the deep learning target detection technology is used to accurately detect the weld area in the field of view in real time,forming an adaptive correction technology for feature detection interest area based on deep learning target localization to accurately and stably obtain the feature information in the NGS-GTAW standing welding process.The bevel information,tungsten edge information,tungsten tip positioning information and melt pool contour information are obtained by adaptive image feature detection method,and the gun position is characterized by the tungsten edge information based characterization method,and the tungsten tip position is characterized by the tungsten tip positioning information based characterization method,so as to calculate the relative position relationship between the gun and the weld seam and the tungsten tip and the melt pool,and the gun alignment and arc length deviation information are characterized by two-dimensional images.The gun alignment and arc length deviation are characterized by two-dimensional images.A feature information evaluation system with certain judgment criteria is also designed to filter out effective information that can be accurately controlled to avoid erroneous control signals from the detection system,thus improving the self-adaptive capability of the oscillating GTAW welding process for narrow gap bevels of medium-thick plates.An integrated narrow gap submerged-arc welding process sensing-detection-control software and hardware system was developed based on the Lab View platform and Mod BusTCP protocol.A vertical up NGS-GTAW welding platform was constructed to simulate actual working conditions,and the effectiveness of the NGS-GTAW adaptive tracking and detection software system was validated.The experimental results showed that the adaptive tracking and detection software system had a high detection frequency,obtaining two welding gun position deviation information within a swing period of 1.13 seconds,while maintaining a detection accuracy of over 80%.The system could stably obtain information on centering and arc length deviation under a two-dimensional field of view.The detection accuracy of arc length deviation information needs improvement,but the adaptive control of centering position can maintain within ±0.3 mm range and has good adaptability to tilted weld seams in the field of view.Compared with manually adjusted welding processes,the adaptive adjustment of centering is more accurate and precise,avoiding over or under adjustment,making the welding process more stable.