Image Analysing And Processing Method For TOFD Testing On Welded Joints Of The Rail

Time of Flight Diffraction(TOFD) method was a novel technique for the inspectionin steel plates, pipelines and vessels, with its advantages such as ease of operation,accurate location and sizing of defects. However, the image processing and interpretationfor TOFD testing on non-planar structure with variable curvature relied heavily on theskilled operator. This paper presented the image processing and reconstruction of TOFDimages on welded joints of the rail which was a typical non-planar structure, to make theTOFD data more comprehensive.The effective information between the lateral wave and backwall signals wasextracted for the image reconstruction. To enhance the image of SNR and obtain goodresults of segmentation, a FIR lowpass filter was designed for noise suppression of theTOFD image, both vertical denoising and horizontal denoising methods were conducted,the result showed that the horizontal denoising performs better. Based on the highcontrast of the lateral waves and backwall signals vs. background, the segmentationalgorithm of TOFD images derived from different rail positions was developed. Thereliability of this algorithm was evaluated by segmenting the effective information ofTOFD images with different parameters, it showed that the success rate of the side of railhead was75%, and the success rate of other positions was higher than80%.The problem that the vertical axis of the image represented time instead of the depthof the defects had to be solved prior to the image reconstruction. Therefore the verticalaxis was rescaled to represent depth by re-sampling the TOFD signals under thethickness intervals0.1mm. Then the TOFD images were reconstructed according to theinspection methods of different rail positions, the location of the defects from thereconstructed image proved to be more easily understood than the original.In order to solve the information overlap problem, the defect signal was extractedfor the image reconstruction. The threshold method for slope of signal peaks wasproposed to extract the defect signal, and the influence of different testing parameters onthe threshold was studied, it was found that images of the same parameter revealed thesame threshold for signal extraction. The reconstruction result of the extracted defectsignals was shown to be more comprehensive, with the depth of the defects quantitative. The study above showed that the defect information was clearer and more intuitiveby the image reconstruction of rail TOFD testing, and this image reconstructionprocedure could be used as reference in interpretation for TOFD testing on non-planarstructure with variable curvature.