In the inverse problem of tomography field, the solution of image reconstruction is often ill-posed and the prior information about imaging features is limited. We expect to learn imaging autonomously by learning algorithms and representative samples. So in this paper, two deep learning image reconstruction algorithms SSAE+RBF and optimized fully connected (FC) are proposed to learn imaging in electromagnetic tomography (EMT). It is a preliminary attempt of sample training algorithm in EMT. Furthermore, a loss function is proposed and 30000 image samples for training, verification, and test are designed. Simulation experiments show the following. First, for the 26000 training samples, both of two algorithms have the ability to basically reproduce the actual distribution of object field. Second, for the random 2000 test samples, which has similar type with training sample but doesn't learned, both of the two algorithms are superior to the traditional algorithms in image reconstruction. In addition, the mean value of image correlation coefficient (ICC) and relative image error are 0.817 and 0.530 for optimized FC network without noise. Third, when 0%-7% noise levels are added to the test set, the standard deviation of ICC in two algorithms are 0.007 and 0.040. To a certain extent, it proves the robustness of these networks. Fourth, in addition, our deep learning algorithm has an advantage in computing speed with graphic processing unit.

Electrical tomography (ET) has emerged as a safe and real-time imaging technology for decades due to its fast, non-invasive, and radiation-free characteristics. By reconstructing the spatial distribution or dynamic changes in the electrical properties of target regions, ET finds broad applications in non-destructive testing, industrial process monitoring, and biomedical research. Recent advancements in deep learning have significantly enhanced the accuracy, efficiency, and robustness of ET image reconstruction, particularly in electrical impedance tomography (EIT), electrical capacitance tomography (ECT), and electromagnetic tomography (EMT). Despite the growing interest, challenges persist in handling the ill-posed nature of ET imaging, especially for EMT, which remains underexplored compared to EIT and ECT. This review delves into the fundamental principles of ET and surveys recent progress in deep learning-driven approaches. We focus on two key paradigms: direct reconstruction networks (data-driven) and hybrid frameworks that integrate deep learning with classical algorithms (image-driven). The discussion highlights emerging trends, including multimodal ET integration, dataset optimization, and the design of advanced neural architectures tailored for ET-specific constraints. This study underscores the potential of deep learning to redefine ET systems, paving the way for more intelligent, versatile, and accurate imaging solutions.

AC bridge techniques commonly used for precision impedance measurements have been adapted to develop an eddy current sensor for rail defect detection. By using two detection coils instead of just one as in a conventional sensor, we can balance out the large baseline signals corresponding to a normal rail. We have significantly enhanced the detection sensitivity of the eddy current method by detecting and demodulating the differential signal of the two coils induced by rail defects, using a digital lock-in amplifier algorithm. We have also explored compensating for the lift-off effect of the eddy current sensor due to vibrations by using the summing signal of the detection coils to measure the lift-off distance. The dominant component of the summing signal is a constant resulting from direct coupling from the excitation coil, which can be experimentally determined. The remainder of the summing signal, which decreases as the lift-off distance increases, is induced by the secondary eddy current. This dependence on the lift-off distance is used to calibrate the differential signal, allowing for a more accurate characterization of the defects. Simulated experiments on a sample rail have been performed using a computer controlled X-Y moving table with the X-axis mimicking the train's motion and the Y-axis mimicking the train's vibrational bumping. Experimental results demonstrate the effectiveness of the new detection method.

A non-contact Laser-EMAT integrated system (LEIS) that uses a pulse laser as a generator and an Electromagnetic acoustic transducer (EMAT) as a detector is proposed. To explain the LEIS could simultaneously measure the width and the depth of metal plate, the finite element model (FEM) of the LEIS, which can describe the entire process from laser generates ultrasound to EMAT receives ultrasound, is established. In detail, the intensity ratios of Rayleigh wave, Shear wave and Longitudinal wave excited by laser under the consideration of three components, the particle displacement, the particle vibration velocity in out-plane direction and particle vibration velocity in in-plane direction, are focused. Combining the characteristics of the ratio and the EMAT as a velocity sensor, the Rayleigh wave and Longitudinal wave in the case of the particle vibration velocity in out-plane direction are chosen as the using waveforms, and the changes of directivity of various waveforms with laser parameters are studied. Furthermore, after the coil structure parameters and magnetic field direction of EMAT were reanalyzed and reconsidered, it is decided to use the coil with the wire spacing of 0.1mm and a number of turns of 8 to combine with the transverse magnetic field to simultaneously detect the Rayleigh wave and Longitudinal wave generated by the laser. Finally, the signal-to-noise ratio of the two waveforms reach a balanced value, and can be clearly distinguished and extracted to achieve the purpose of measuring the depth and width of the metal plate at the same time.

This paper presents electromagnetic inspection of in-service train wheel online for high-speed railway using a fast demodulation algorithm. The electromagnetic simulation demonstrates the defect can be identified by the amplitude variation of differential voltage induced in two pick-up coils. To demodulate the induced voltage, the Kalman-model-based demodulation algorithm is proposed, which can fast demodulate resultant signal of any arbitrary frequency and have higher anti-noise ability at high speed. Furthermore, the experimental platform for online electromagnetic inspection of in-service train wheel is designed. Both defect identification and speed estimation experiments are carried out. Finally, the experimental results demonstrate that the design of online electromagnetic inspection of in-service train wheel is effective and two defects can be identified by the differential induction voltage at speed over 300km/h. Furthermore, the Kalman-based demodulation have better performance in noise suppression at higher speed compared with other traditional demodulations.