Research On Real-time Weld Deviation Detection And Arc Monitoring For GMAW Robot Based On Wide Dynamic Vision Sensing

Welding automation and intelligence are key technologies of the welding manufacturing.To overcome the shortcomings of the teaching-playback welding robot,detecting weld deviation and monitoring arc using sensor and related algorithm in the welding process,will greatly enhance the level of automation and intelligence of welding robot.In this paper,clear welding images are captured real-timely with only a wide dynamic range（WDR）camera without employing any external light source or filter.On this basis,the GMAW robot system is constructed and the following aspects are studied.1.A weld deviation detection algorithm of CO₂ arc welding based on the weld pool center is proposed.The region of interest（ROI）is automatically extracted using a morphological algorithm based on fast hybrid reconstruction.Then,according to the characteristics of welding images,an improved Canny algorithm is developed to detect groove edges,and both Hough transform and prior knowledge are used to connect these edges.Moreover,in order to fit out the edge of weld pool,a differential-evolution（DE）ellipse detection algorithm based on condition restriction is proposed.Two feature points of weld pool edge are obtained as the constraint conditions of DE iterative optimization.Then,the other three unknown points of the population are initialized and the evaluation function and termination condition are redefined.The mutation factor can be determined adaptively in the DE algorithm.On this basis,the welding deviation is calculated.Finally,the robustness and accuracy of the algorithm are verified by experiments and the real-time performance is described.2.A weld deviation detection algorithm of MAG welding based on the arc top center is proposed.Double threshold method combining basic global threshold and optimal global threshold is developed to segment the precise arc region.According to the connection relationship between the top of the arc and the end of the wire during MAG welding process,a method for determining the wire centerline is proposed.Based on the improved Canny algorithm and combined with the characteristics of MAG welding image,Canny algorithm is further optimized and improved.A groove edge detection algorithm based on Canny theory is proposed.After ROI smoothed by Gaussian filtering,the gradient amplitudes in x,y direction are calculated respectively.The y direction gradient is processed to acquire the binary image,which serves as the mask;then,after the NOT operation of this image,the AND operation of this image with the edges extracted from the x direction is executed to obtain the single-pixel image of the V-groove edges.In addition,self-adjustment method for Gaussian filtering standard deviation is proposed.Further experiment showed that the weld deviation detection algorithm based on the arc top center can meet the requirement of seam tracking of MAG welding.3.The basic principle between the contact tip to work distance（CTWD）and the wire stick-out is expounded.The relationship between the CTWD and the wire stick-out is analyzed by using the CTWD static mathematical model and the wire stick-out model.CTWD step change experiment was designed for testing it.The characteristic of the nozzle edge in the welding image is analyzed and the algorithm for detecting the wire stick-out is proposed.Experiment demonstrated that the algorithm can effectively detect the wire stick-out in the welding image and its accuracy can meet the requirement that maintained CTWD basically constant during GMAW robot welding process.4.Welding images including three different arc shapes-cone shape,bell shape and twisted beam shape-are captured respectively during drop spray transfer mode,streaming transfer mode and rotating streaming transfer mode.The improved Canny algorithm is employed to detect arc edges.To obtain both entire arc shape and closed arc edge,an improved fast marching method（FMM）is proposed.The speed law and stop condition are redefined.Furthermore,a recognition method based on signatures is proposed to identify three arc shapes.Experiment showed that the arc shape extraction algorithm and the arc shape recognition algorithm can effectively extract and identify the arc shape in MAG welding process,which can be applied for monitoring arc real-timely.Finally,the optimal weld deviation is estimated by the Kalman filter algorithm using the predicted value of the previous frame and the detected value of the current frame.The result showed that the optimal estimation was closer to the preset value.Combined with the control diagram of GMAW robot system and the principle diagram of motion execution program,the working principle of GMAW robot system is expounded.Two-dimensional and three-dimensional welding trajectories and arc monitoring experiments were designed for CO₂ welding and MAG welding based on the GMAW robot system.The validity of weld deviation detection,CTWD adjustment and arc monitoring in automatic welding process is verified.