A two-phase data-driven image reconstruction method of carbon fiber composite cable defects in electromagnetic tomography

To address the challenge of clearly imaging and evaluating defects in low- conductivity carbon fiber composite cables, a two-phase data-driven electromagnetic tomography imaging method is proposed to realize the image reconstruction of the defects. Firstly, the electromagnetic model of 8-coil excitation-8-coil detection based on carbon fiber composite cable is constructed by combining its actual electrical properties and fiber orientation. The induced voltage response of carbon fiber composite cable with various defects is simulated using the finite element method, establishing a dataset of real defect images and corresponding induced voltage. Secondly, the objective function of the sensitivity matrix optimization model is constructed by the difference between the real defect image and the reconstructed defect image. Carbon fiber composite cable reconstructed images with defect shape are reconstructed. Using defect reconstruction images for fine-tuning, SAM2_Ran achieves precise defect region segmentation in phase II, with mean Intersection over Union of 0.9142. Compared with LBP and Tikhonov regularization algorithms, the proposed two-phase imaging method presents superior defect imaging performance in both simulation and experiments. The average defect location identification accuracy reaches 94% under a 10% noise level. This study provides a novel solution for accurate and rapid defect detection in low-conductivity carbon fiber composite cable.