Abstract: The radial defects of anchor bolt caused by long-term corrosion lead to the reduction of its bearing capacity, which led to a large number of engineering accidents. The defect detection of service anchor bolts was great significance to the removal of safety hazards. However, the current detection methods for radial defects of anchor bolts were still immature. In order to detect radial defects in anchored bolts, guided wave detection tests were carried out on anchored bolts with different defect locations and defect numbers. The propagation decay law of the longitudinal guided wave in the defect bolt was explored, and the applicability of the short-time Fourier transform, the Hilbert-Huang transform and the empirical mode decomposition method to the radial defect identification of the bolt was compared and analyzed. The experimental results show that the detected time-domain signal waveform presents exponential attenuation under the coupling effect of mortar medium, and the more the number of defects, the greater the attenuation degree and attenuation speed of guided waves. The dispersion of the guided wave occurs at the radial defect position of the bolt, so that the peak value of the reflected signal at the defect and the bottom is multi-peak and the range of the wave packet is expanded, which makes it difficult to identify the defect information in the bolt from the time domain signal. The stripes of the time-frequency graph based on short-time Fourier transform are fuzzy, the resolution is low, and the identification of defect information is not clear. The Hilbert- Huang transform has obvious waveform characteristics in the analysis of non-defect anchorage bolt, but it is difficult to judge the defect information in the analysis of multi-defect anchorage bolt. The method based on empirical mode decomposition can identify the defect information of anchor bolt accurately, and the error of fault location identification is 5.3%.