Nondestructive detection of defects in friction stir welded lap joints using computed tomography

• Published in: Materials & design 2019-01, Vol.162, p.10-23
• Main Authors: Hamade, R.F., Baydoun, A.M.R.
• Format: Article
• Language: English
• Subjects: Computed tomography; Defect; Friction stir welding; Hounsfield; Mask; Nondestructive testing
• ISSN: 0264-1275; 1873-4197
• DOI: 10.1016/j.matdes.2018.11.034

Based on computed tomography (CT), a method is developed for nondestructive detection of defects in friction stir welded (FSW) joints. Plates of AA6061-T6 and AA1050 are welded to produce short lap welds. Utilized is spindle speed of 1600 RPM and fifteen different tool feeds ranging from 25 to 1000 mm/min (dubbed welded cases #1 to #15). These joints are CT-scanned and DICOM images are produced for digitally sectioned zones within the welded joints. To demarcate metallic zones from defects, a threshold cutoff value for Hounsfield Unit (HU) is needed to create 2D segmented masks determining the periphery between metal pixels from air pixels. Otsu's thresholding selection method from grey-level histograms is utilized for this purpose. These cutoff values are found to vary slightly from one welded joint to another. From the 2D masks, 3D representations of the welded joints revealed the shapes, locations, and volumes of the detected defects. Cross-sectioning of welds is employed as a destructive test to corroborate the locations and areas of CT-detected defects. The results of the defect area analysis from the CT-scans are compared to the results from optical images obtained by computer vision (CV) in a pixel-based analysis. Three projection views and an isometric view of CT scans of an FSW lap welded joint with defects revealed. [Display omitted] •A nondestructive testing (NDT) technique based on X-ray computed tomography (CT) and the Otsu thresholding method is used.•Technique successfully detects and identifies defects in friction stir welded lap joints.•Defects’ sizes are compared to areas determined using computer vision (CV) of images of cross-sections cut from the joint.•Compared with CV, areas of CT-detected defects show discrepancies attributable to the resolution of the CT apparatus.