Recognition Model For Micro Gap Weld Seam Using Magneto-optical Imaging

Laser welding is an important technique of joining techniques among nonmetal materials and different metals, which has advantages of high laser power, small laser beam spot, good welding quality, small laser affected zones and large depth and width ratio, and it can realize the connection between different materials. During the laser welding process, laser beam must be controlled to focus on the weld seam center for good welding quality weldment. Weld seam detection is necessary and significant for welding automation. Since the laser beam spot diameter is small(less than 200?m), which is sensitive to the weld seam size. The traditional method of structured light visual sensing could not identify the weld seam while the weld seam gap less than 0.1 mm. There are a lot of smoke, splash and plasma influence during actual industrial environment, traditional camera can not capture clear and accurate images of molten pool and weld seam image, meanwhile, intense heat conversion effect exists during laser welding process, which requires high fixed precision of workpiece assembly, little changes could cause severe welding defect or even scrapped. Therefore, it's the premise to control the laser beam focus on the weld seam for good welding quality.On the basis of current weld seam detection and tracking technology, the advantages and disadvantages of the existing methods of weld seam detection are compared. In the case where the width of a butt joint is less than 0.20 mm, no groove and it is too narrow to measure by naked eyes directly, according to the principle of Faraday magneto-optical effect, micro gap weld seam magneto-optical images were captured by a magneto-optical sensor. Study on the micro gap weld seam magneto-optical imaging characteristics and mechanism during laser welding. Micro gap weld seam magneto-optical imaging experiment platform of laser welding of stainless steel is established(Chapter 2). Firstly, weldment is placed on the servo platform, a magnetic field generator is placed below the weld seam, the induction current and induced magnetic field strength are changed by adjusting the excitation voltage of the magnetic field generator. According to Faraday electromagnetic induction effect and Faraday magneto-optical effect, the flows are affected when there is a gap during the vortex flowing in the flow path, eddy current distortion occurred at the position of weld seam gap and caused the vertical distribution of the magnetic field change. Weld seam magneto-optical imaging by a magneto-optical sensor which change the eddy current magnetic field intensity to light intensity. Research on the relationship between the micro gap weld seam magneto-optical imaging features and the weld seam position. The test results show that the excitation voltage, the distance between magneto-optical sensor and weldment, welding speed, micro gap weld seam width are sensitive to the micro gap weld seam magneto-optical imaging.Secondly, study on the micro gap weld seam magneto-optical image features(include gray level feature, gray gradient feature, color space feature and texture feature),explore the rule between these characteristics and micro gap weld seam position. Analysis of micro gap weld seam magneto-optical image gray and gradient distribution characteristic are shown, through the global threshold and edge operator, weld seam position could be obtained by the center position of the weld seam transition zone, but the threshold is not a universal. In RGB and HSV color space, weld seam position could be computed by the micro gap weld seam magneto-optical image color spatial feature, the measuring accuracy of weld seam detection in color space is higher than grayscale space.Finally, the characteristic of the magneto-optical image sequence is analyzed, weld seam position and the movement of each pixel are determined by the magneto-optical image sequence features in time domain and correlation. Research on the application of optical flow method and the gradient vector flow model in the micro gap weld seam detection. In addition, the neural network, particle filtering and Kalman filtering algorithm were applied to established the recognition model for micro gap weld seam using magneto-optical image and ultimately realize welding seam recognition, tracking and prediction. Through this series of research work, the main achievements are obtained as following:(1) Exploration and analysis of the relationship between the micro gap seam magneto-optical imaging and the actual situation of the welding seam as well as the other factors.Research on the main factors that affect the micro gap seam magneto-optical imaging. For a specific magneto-optical sensor, the main factor that affecting the magneto-optical imaging is Faraday magneto-optical effect, whose related factors include:excitation voltage, the distance between the magneto-optical sensor and weldment, weld shape and welding speed. Experimental results show that for the same micro gap weld seam, the magnetic induction intensity around the weld seam would change when the excitation voltage changes. The magnetic induction strength of the weld seam position placed at the center of symmetry of the two sides of magnetic field(N pole and S pole). The weld transition zone moves up and down with the change of excitation voltage, but the weld seam position offset is constant for two different magnetic field intensities, which can by neglected under a certain magnetic field strength environment. For the same micro gap weld seam width, the weld seam position is basically constant when the welding speed changes. Welding speed only affects the welding penetration and image sampling frame. Under the same excitation voltage, the larger the micro weld seam width, the smaller the transition zone width, the more clear the weld seam magneto-optical image. For the same weld seam width and excitation voltage, the closer between the magneto-optical sensor and weldment the smaller the weld transition zone and more clear the magneto-optical weld seam image.(2) Analysis and extraction of micro gap weld seam magneto-optical image features.The characteristics of the magneto-optical image of the micro gap weld seam include: gray level features, gray level gradient feature, texture feature and image sequence feature. A micro gap weld seam magneto-optical image is scanned through the progressive or column to analyze the gray distribution characteristics. It can be found that on both sides of the weld intensity distribution have obvious difference, the difference of gray distribution in the weld position can accurately detect the weld center position. The gray scale image of the micro gap weld are analyzed, a scanning region is selected according to the experience, the corresponding row of the gray gradient maximum value is used as the upper and lower edge coordinates, finally the center weld position is calculated. The texture features of each image are extracted from the image of the weld region and the base region. The texture features include:average luminance, standard deviation, smoothness, three order moment, consistency, entropy 1, energy, correlation, entropy 2. Texture features can be used to separate the weld seam center using the differences between the characteristics of the weld zone and material.(3) Micro gap weld seam position detection using magneto-optical imaging.Optical flow method is used for determining the position of the respective pixels by using the time domain and the correlation of the pixel data in the weld seam magneto-optical image sequence. The micro gap weld seam position is defined as the position where u component peak of optical flow field in weld seam magneto-optical image. The weld seam position computed by H-S optical flow method is consistent with the actual values. Meanwhile, gradient vector flow field is applied for micro gap weld seam position, which considers the region of interest as a non discontinuity curve, in the control point of the active contour, the energy can be reduced. Under the control point of the internal force, active contour expand to the target area and ultimately achieve accurate identification of the target contour.(4) Recognition model establishment of micro gap weld seam.Based on BP neural network and Elman neural network, the prediction of weld seam position model is established. A feed-forward neural network (forward) neural network is designed to estimate the position of the weld seam through computering the position of the welding seam at the previous moment. The accuracy of BP neural network and Elman network location prediction is compared. Experimental results show that the prediction ability of BP neural network is better than that of Elman neural network, and it can effectively predict the location of weld seam, and the measurement accuracy of BP neural network is better than that of Elman network. Kalman filtering algorithm is used for weld seam tracking and prediction, the optimal estimation of the system state is obtained under the condition of the known position of the weld. The noise interference can be greatly suppressed after filtering, Kalman filtering can effectively improve the tracking precision of weld seam.