A Study of the Effect of Extended Hounsfield Unit Range and Voxel Size on Defect Detection in Friction Stir Welds

A study of the effect of extended Hounsfield Unit range and voxel size on defect detection in friction stir welds

Ahmad M. Baydoun, American University of Beirut, Mechanical Engineering Department, P.O. Box 11-0236, Riad El-Solh, Beirut 1107 2020, Lebanon, Email: amb52@mail.aub.edu  

Ramsey F. Hamade, American University of Beirut, Mechanical Engineering Department, P.O. Box 11-0236, Riad El-Solh, Beirut 1107 2020, Lebanon, Email: rh13@aub.edu.lb

Friction stir welding (FSW) is a novel welding method that is garnering attention, in part, due to its ability to join dissimilar materials. One of the challenges in producing dissimilar friction welded joints is achieving proper material mixing in the weld zone and, thus, achieving good joint strength. Another challenge is producing welds that are free of internal defects. One way to ensure proper mixing and defect-free welds is achieved by inspecting the joints via nondestructive testing (NDT) methods that insures that the sample is kept intact. The non-destructive detection of these defects has been an active research area in recent years. Various methods, including the utilization of ultrasonic waves, gamma rays, X-rays, and X-ray CT, have been implemented to this end.This study explores the effects of two critical parameters of X-ray CT scanning namely 1) voxel size and 2) Hounsfield unit (HU) range (12 bit and 16 bit depth) on the detection of internal defects in lap welded dissimilar AA1050 AA6061-T6. Small voxel size (higher resolution) improves the ability of detecting internal defects and improves the effectiveness of the thresholding process. Higher HU range results in a wider separation between detected material peaks, thus enhancing the effectiveness of the thresholding process as well.

Dissimilar AA1050/AA6061-T6 welds are produced in a lap weld configuration. The samples are scanned via X-ray CT at two different voxel sizes (2.457 E-02 and 1.420 E-03 mm³) and two HU ranges 12 bit and 16 bit depth (2^¹² and 2^¹⁶). The generated Digital Imaging and Communications in Medicine (DICOM) images are segmented based on proper HU thresholding found via the Otsu thresholding method. The effect of voxel size on the thresholding method is determined by analyzing scans made at the two resolution levels for the 16 bit HU range. The impact of the HU range on the effectiveness of the thresholding method is investigated by analyzing scans performed at 12 bit and 16 bit depth for the same voxel size. Finally, the combined effect of both parameters is explored.