Study On Automatic Identification Of Welding Defect Based On Ultrasonic Time Of Flight Diffraction

Ultrasonic Time of Flight Diffraction (TOFD) technology can provide a more accurate characterization for location, length, height and morphology feature of defects in medium and heavy plates, thus this technology has a wide application prospect in domestic nuclear industry, power, pressure equipment, railways, bridges and other projects. While, the identification, localization and quantification of defects in TOFD image remain in the stage of manual operation. Therefore, the study on automatic identification of welding defect based on TOFD is of great significance to NDT automation of welding defects.In this paper, a TOFD ultrasonic acquisition system is built and scanning experiments are completed in austenite stainless steel test blocks. A-scan signals of defects echo are collected and TOFD images are synthesized by post-processing.The calibration of scanned image is directly related to the accuracy of defect positioning in TOFD method. Based on features of TOFD image, this paper develops a complete set of calibration algorithm for direct waves, including smoothing and denoising, morphological feature extraction, threshold segmentation, search of lines and peak detection, by which the peak of direct waves are located precisely. The achievement of image calibration makes the foundation of quantification automation of defects.Aiming at the positioning of subsurface defects using TOFD method, this paper improves the direct wave processing algorithm for business software. Through the improvement, the new algorithm performs better in straightening and removal of direct wave.To solve the problem of low lateral resolution in TOFD B-scan, the synthetic aperture technique is applied to image analysis. In this paper, a parameter model of defect echo is established and the Doppler frequency shift of echoes is studied.On the basis of theoretical calculation of TOFD synthetic aperture, this paper gives a suitable aperture setting, and proposes and demonstrates an improved algorithm based on gain compensation, which significantly increase the image contrast.