Sensing And Control Of Surface Defects In Friction Stir Welding Seam Based On Laser Vision

During the friction stir welding process,flash and grooves are prone to occur on the surface of the weld due to various influencing factors such as welding parameters,which seriously affect the welding quality.In order to effectively detect and control the occurrence of defects in real-time during the welding process,this study conducted research on the sensing and control technology of surface defects of friction stir welding based on laser vision.The surface sensing and control system for welds uses the AM5708 evaluation board as the main processor to complete image acquisition,processing,and welding parameter adjustment.The ARM of the AM5708 evaluation board calls a Gig E industrial camera for laser stripe image acquisition,and important data is displayed using Qt.The DSP is responsible for processing the laser stripe image.During the image processing process,the ROI(Region of Interest)was automatically identified through the image gray distribution map to improve image processing speed.The image was preprocessed with Gaussian filtering,median filtering,Gamma transformation,etc.The OTSU algorithm was used for binarization of the laser stripe image.The connected domain area was eliminated to remove interference factors such as the reflection of the stir tool.The image was subjected to morphological processing to smooth the edges.Then,conventional welding seam feature point detection methods,such as image thinning and line detection,were used for the laser stripe image.When the defect size is small,it cannot be effectively recognized.Therefore,an edge fitting+ feature point detection scheme is proposed.Firstly,the Sobel operator is used to extract the edge point set of the laser stripe,and the Visvalingam-Whyatt algorithm was used to fit the point set with line segments to obtain the fitting polygon of the laser stripe.Then,the SIFT algorithm was used to extract the feature points of the fitted polygon to obtain the basic shape and pixel size of the groove defect.For the flash defect,the horizontal projection method was used for processing.Firstly,the centerline extraction and line detection of the laser stripe is performed to correct the image.Then,the laser stripe was horizontally projected to obtain the pixel height of the flash,and the actual size of the defect was calculated based on the mathematical model of the installation positions of the industrial camera and the line laser.After obtaining the actual size of the defect,the fuzzy PID control algorithm was used to obtain the adjustment amount of the downward pressure based on the defect size,.The communication between the industrial camera,AM5708 evaluation board,and the stir friction welding equipment control system has been achieved using the Gig E Vision and ModbusRTU protocols,enabling data transmission among these three components.Experimental verification was conducted on the system,and the results showed that when flash and groove exist in the laser stripe image acquired by the industrial camera,the welding parameters can be adjusted in time to obtain a well-formed surface of the weld.The calculation error rates of the width and height were below 4.5 % and 4 %,respectively,and the time for each frame of image acquisition and processing was approximately 47 ms,which met the real-time requirements.