Detection of indentation damage in carbon fibre-reinforced composites via EIT based on mechanical-electric coupling simulation and BT-CNN

In damage detection of carbon fibre-reinforced polymer (CFRP), electrical impedance tomography (EIT) is widely used due to its low-cost, zero radiation, and fast response advantages. However, the inverse problem of EIT is highly underdetermined and nonlinear. Researchers usually use regularisation or machine learning methods to solve the inverse problem. In learning-based approaches, most researchers use artificial through-hole damage for their simulation datasets. But this type of damage hardly ever occurs in real components and structures. Therefore, it is necessary to create a realistic damage dataset. In this work, the mechanical-electric coupling simulation was used to build a simulation dataset based on quasi-static indentation, and the conductivity change image was reconstructed by a bagging algorithm combined with transfer learning (BT-CNN). Simulation and experimental results show that BT-CNN is more robust than traditional algorithms and other machine learning algorithms in reconstructing realistic damage images.