Abstract: Aiming at the problem that the glass fiber reinforced plastic (GFRP) anchor rod body is prone to shear failure and difficult to detect,a GFRP anchor rod body defect evaluation method based on time inversion method and deep learning is proposed to achieve accurate identification and quantitative evaluation of defects. Based on COMSOL numerical simulation and laboratory similar test,the time inversion method is used to detect the GFRP anchor anchorage structure with different defects, and the focusing signal is obtained. The results show that the focused signal waveform changes little with the rod defect,and the signal waveform coincidence degree is high. The amplitude of the focused signal decreases with the increase of the defect degree of the rod body. The time-frequency diagram of the focused signal obtained by the test is generated by wavelet transform,which is used as the input of the convolutional neural network (CNN)-support vector machine (SVM) model,and the defect degree of the GFRP anchor bolt body is used as the output to construct the defect evaluation model. The model training results show that the evaluation accuracy of GFRP anchor bolt rod defects reaches 100%. The method proposed in this paper can realize the rapid and accurate evaluation of the defect degree of GFRP anchor,which provides an important theoretical basis and technical support for the defect detection of GFRP anchor. 

Key words: GFRP anchor,rod body defects,time reversal method,wavelet transform,CNN-SVM