Damage Detection of CFRP by Electrical Impedance Tomography Imaging Based on RM-DeeplabV3

Due to the self-conductive characteristics of carbon fiber reinforced polymer (CFRP), electrical impedance tomography (EIT) can be utilized for its noninvasive and visual structural health detection. However, traditional image reconstruction algorithms for EIT suffer from low accuracy due to the nonlinear, ill-posed, and underdetermined nature of the EIT inverse problem. Deep learning algorithms, on the other hand, have gained significant attention in the field of CFRP damage detection as they can achieve higher image accuracy through network training with large datasets. In this article, an EIT imaging algorithm based on RM-DeeplabV3 + deep network is proposed for damage detection of CFRP laminates. This algorithm is proposed to solve the problem of low-quality image reconstruction due to the EIT inverse problem in the case of small CFRP structural damage data. The algorithm is composed of a premapping module, feature extraction module, and segmentation module to realize the image reconstruction of CFRP. The advantages of this algorithm include: 1) effectively addressing the issue of gradient disappearing in deep networks and improving the training effectiveness of the network; 2) introducing a new loss function that combines the dice loss function and focal loss function to improve segmentation performance; and 3) removing artifacts of the reconstructed image through the segmentation network in the feature extraction and segmentation module. In this article, the correlation coefficient (CC) and structural similarity index (SSIM) are used to assess the quality of reconstructed images. Simulation and experimental results demonstrate the image reconstruction algorithm based on RM-DeeplabV3+ can get clearer boundaries, and more accurate damage recognition compared with other imaging algorithms.