Enhancement of root crack detection in friction stir welds using ultrasonic non-destructive testing and signal processing technique

The role and scope of non-destructive evaluation (NDE) is continuously changing over time in order to meet the emerging challenges concomitant with new materials and process technologies. Among these are the new inspection methods required by new advanced technologies, such as NDE for the analysis of friction stir welds (FSW) in quality control applications. Recent increases in the industrial application of FSW to join materials that are difficult to fusion weld, has intensified the need for a non-destructive technique to detect and characterize defects that may occur in products produced by FSW. The challenge to meeting this need arises from the tendency of these defects to be very tight and very small, and that they may occur at any orientation. This wide variety of defect orientations is very unusual for conventional weld inspection, and their small, tight geometry adds to the challenge. In this thesis, we compared the sensitivity of several non-destructive methods to determine the optimal approach. Analyses of A-scan signals, experimentation with phased array probes and single element probes were used to test the detection of cracks in FSW. The improvement of detection was verified experimentally on aluminum samples with simulated flaws. Consequently, with some limitations of ultrasonic testing in detecting micro cracks, signal processing technique was applied. Noise suppression of A-scans was done by utilizing wavelet transform. Enhancements of ultrasonic signals were based by the significant increase in signal to noise ratio.