In this study, acoustic emission (AE) tests were conducted to detect and locate wire fracture in strands that are widely used in cable-stayed and suspension bridges. To effectively separate fracture signals from unwanted noises, distinct features of fracture, fracture-induced echo, and artificial tapping signals as well as their dependence on loading levels are characterized with short-time Fourier transform. To associate fracture scenarios with their acoustic features, two 20-foot-long (~6.1 m) 270 ksi (~1,862 MPa) steel strands of seven wires were tested with one wire notched off at center and support, respectively, up to 90% of its cross section area by 10% increment. Up to 80% reduction in cross section area of the notched wire, each strand was loaded to 20 kips (~89 kN) corresponding to 35% of the minimum breaking strength and the acquired AE parameters such as hits, energy, and counts were found to change little. With a reduction of 90% of the section area of one wire, both strands were found to be fractured under approximately 16.5 kips (~73.4 kN). The hits, energy, and counts of AE signals were all demonstrated to suddenly change with the fracture of the notched wire. However, only the counts of AE signals distributed over the length of the strands allow the localization of fracture point. The frequency band of fracture signals is significantly broader than that of either fracture-induced echo or artificial tapping noise. The time duration of artificial tapping noises is substantially longer than that of either fracture or fracture-induced echo. These distinct characteristics can be used to effectively separate fracture signals from noises for wire fracture detection and localization in practice.
Nondestructive evaluation and sensing technology have been increasingly applied to characterize material properties and detect local damage in structures. More often than not, they generate images or data strings that are difficult to see any physical features without novel data extraction techniques. In the literature, popular data analysis techniques include Short-time Fourier Transform, Wavelet Transform, and Hilbert Transform for time efficiency and adaptive recognition. In this study, a new data analysis technique is proposed and developed by introducing an adaptive central frequency of the continuous Morlet wavelet transform so that both high frequency and time resolution can be maintained in a time-frequency window of interest. The new analysis technique is referred to as Adaptive Wavelet Analysis (AWA). This paper will be organized in several sections. In the first section, finite time-frequency resolution limitations in the traditional wavelet transform are introduced. Such limitations would greatly distort the transformed signals with a significant frequency variation with time. In the second section, Short Time Wavelet Transform (STWT), similar to Short Time Fourier Transform (STFT), is defined and developed to overcome such shortcoming of the traditional wavelet transform. In the third section, by utilizing the STWT and a time-variant central frequency of the Morlet wavelet, AWA can adapt the time-frequency resolution requirement to the signal variation over time. Finally, the advantage of the proposed AWA is demonstrated in Section 4 with a ground penetrating radar (GPR) image from a bridge deck, an analytical chirp signal with a large range sinusoidal frequency change over time, the train-induced acceleration responses of the Tsing-Ma Suspension Bridge in Hong Kong, China. The performance of the proposed AWA will be compared with the STFT and traditional wavelet transform.