A comparative study of time-frequency analysis techniques in the case of signal processing for ultrasonic NDT

Ultrasonic non-destructive testing of components made from granular materials (such as austenitic steels) is complicated by grain noise and randomly distributed intergranular stress corrosion cracks of different shapes. The objective of the present study is to perform numerical simulations of the ultrasonic non-destructive testing procedure and application of various signal-processing techniques in the case of structural noise. Such simulations have been performed in order to adapt the appropriate signal processing technique to the effective improvement of the signal-to-noise ratio (SNR) and to increase the probability of defect detection as well. The advantage of the numerical simulation represents a possibility to simulate a test object with various reflectivity levels of the ultrasonic reflections from its internal structure, which also results in different SNR. Such a numerical approach does not require having different objects with different internal granular structure, like during a real experiment. Comparative results of the SNR improvement have been obtained using the following signal processing techniques: Wigner-Ville distribution (WV), Wavelet transform (WT) and Hilbert-Huang (HHT) transform.