In actual applications, owing to long-term high loads on their surfaces, rail tracks inevitably develop surface fatigue cracks that propagate in depth, which can cause abrupt rail breakages. Therefore, periodic in-situ defect non-destructive testing must be performed on service rails without affecting the normal operation of the railway. Laser ultrasonic technology can achieve non-contact detection and exhibits outstanding on-site testing capabilities, showing promising application prospects in the field of non-destructive testing. Previously, researchers primarily focused on simulation methods for crack detection using laser ultrasonic technology, whereas experimental studies pertaining to crack detection are scarce. Additionally, a few researchers successfully performed crack localization and quantitative detection via overall crack imaging. This study proposes a dual-side line-scanning imaging method based on laser ultrasonics for rapidly detecting surface cracks on rail tracks via line scanning. The defect images obtained can effectively reflect the positions, lengths, and widths of the cracks.

Laser ultrasonic dual-side scanning is adopted in this study to detect surface cracks on steel rails. The laser ultrasonic experimental system includes excitation devices, reception devices, mobile scanning devices, and signal acquisition and processing devices, as shown in Fig.4. By fixing the excitation and detection devices and actuating the sample, a line scan is conducted on each side of the crack to achieve dual-side line scanning for crack detection. Wavelet packet transformation is employed to denoise ultrasonic waveform data obtained from bilateral line scanning. Subsequently, the denoised ultrasonic data are subjected to B-scan imaging processing, which converts the resulting B-scan image into left and right images of the crack and merges the images from both sides to achieve an overall image of the crack. To mitigate errors in crack-length detection using the line-scanning imaging method, a scan position versus maximum amplitude difference graph is plotted to determine the start and end positions of the crack, from which the crack length can be determined, thus facilitating improvement to the accuracy of crack-length detection.

The wavelet packet transform effectively enhances the signal-to-noise ratio of ultrasonic signals (Fig.5). The laser ultrasound single-side B-scan image, as shown in Fig.7(b), allows for the indirect extraction of the position and length of cracks but does not provide the width information. Similarly, the laser ultrasound single-side line-scan image shown in Fig.8 can directly yield the position and length of cracks but not the width. Meanwhile, the imaging method shown in Fig.9 directly provides information regarding the position, length, and width of cracks, with precise detections of crack position and width but with relatively large errors in terms of length detection. Figure 11 shows the scan position versus maximum amplitude difference graphs obtained by line scan on both sides of crack, from which the start and end positions of cracks can be determined directly to obtain the crack length. The detection results indicate that compared with applying the laser ultrasound dual-side line-scan imaging method, using the scan position versus maximum amplitude difference graph significantly improves the accuracy of crack-length detection.

In this study, the principles of laser ultrasonics are utilized to conduct a dual-side line scanning detection of surface cracks on U71Mn steel plates using a custom-developed laser ultrasonic scanning detection system. By employing wavelet packet denoising, applying the dual-side line-scanning imaging method, and plotting the scan position versus maximum amplitude difference graph, surface cracks on rails can be imaged rapidly and cracks can be detected quantitatively with high accuracy. Wavelet packet denoising significantly improves the signal-to-noise ratio of ultrasonic signals, thus facilitating the extraction of defect information after imaging. The dual-side line-scanning method enables an overall imaging of cracks, thus enabling the direct acquisition of the crack position, length, and width from the images. Compared with the conventional point-by-point comprehensive scanning detection method, the proposed method only requires B-scanning on both sides of the rail, thus eliminating the necessity for scanning detection at the middle of the rail, which improves the scanning speed and efficiency. By extracting the maximum amplitude difference and plotting the scan position versus maximum amplitude difference graph to determine the crack length, the relative errors obtained are 1% and 4%, respectively, as compared with relative errors of 7%, 10%, and 8% yielded by imaging methods. Thus, errors are reduced significantly and the accuracy of crack-length detection is enhanced successfully.