GRU-CNN Neural Network for Electrical Impedance Tomography

Carbon fiber reinforced polymer (CFRP) is widely used because of its high specific strength and stiffness characteristics. However, the impact resistance of CFRP is inevitably subjected to impact during work. Electrical impedance tomography (EIT) has great potential in structural health monitoring (SHM) due to its non-destructive, non-intrusive and low cost. In the inverse problem of EIT, numerical algorithms are used to handle large data sets. However, traditional algorithms are computationally expensive and can be complex to implement. This paper aims to solve the inverse problem of EIT by deep learning. To achieve this goal, GRU-CNN model is adopted to the inverse problem of EIT. The RMSE (root mean squared error) and CC (correlation coefficient) are set as image quality criteria. Both simulation and experimental results prove the performance of this method.