Research On Ultrasonic Non-Destructive Testing And Three-Dimensional Visualization Technologies For Carbon Fiber Composites

Ultrasonic testing has a long history of development and has become one of the most valued non-destructive testing methods in the industry.Traditional ultrasonic nondestructive testing technology can only perform two-dimensional scanning of the internal defects of the workpiece,and cannot combine the defect with the three-dimensional morphology of the workpiece to be measured,which is not conducive to the statistics of defects.In this paper,the combination of ultrasonic testing technology and binocular vision technology can display the three-dimensional spatial information and location distribution of the defects of the detection object more intuitively,so that the overall detection is more accurate.The work done is as follows:In order to solve the current problems of low visualization degree of ultrasonic testing and difficulty in locating defects,an ultrasonic testing system based on binocular vision was built by studying the principle of ultrasonic testing and binocular vision technology,and obtaining ultrasonic defect information and spatial position information of carbon fiber composite workpieces.In order to solve the problem of low defect recognition accuracy and poor efficiency in the non-destructive testing of carbon fiber composites,the ultrasonic nondestructive testing results of preset defects of carbon fiber composites were studied by using YOLOv5 algorithm.The YOLOv5 algorithm is improved by adding attention mechanism,prior box optimization and loss function optimization;and the generalization ability of the model is improved through data enhancement;and the average recognition accuracy exceeds 98%;the position calculation error is less than 2 mm;and the area calculation error is less than 5%.The binocular camera was used to capture the cooperative target to obtain the spatial coordinates of the workpiece;the filtering algorithm was performed on the surface type point cloud to remove redundant noise and discrete noise,and the surface deformation caused by the soft ultrasonic roller was excluded by the point cloud smoothing algorithm based on least squares movement.The synchronized data are simplified by voxel filter down sampling;and the optimized ultrasonic defect point cloud is obtained;the ratio of ultrasonic point cloud defect area to actual area is more than 93%;and the roundness is above 0.91.The corresponding feature points of adjacent point clouds are obtained by 3D scale invariant feature transformation algorithm and fast point feature histogram;and the improved random sampling consensus algorithm is used to remove false matching;and the corrected corresponding point set is used for rough matching;and on this basis,the iterative nearest point algorithm is used for phased fine registration.The 48 point clouds were registered in stages,and the final error was 1.17 mm,and the maximum error between the defect coordinates of the point cloud and the actual defect coordinates did not exceed 3mm,so as to accurately realize the three-dimensional visualization of large workpieces.Based on the ultrasonic phased array system and binocular camera on the defect identification and 3D visualization technology of carbon fiber composite workpieces,in this paper,a method for the overall measurement of ultrasonic nondestructive testing of large workpieces is provided,so that the non-destructive testing personnel can analyze the material properties more accurately.