Overhead transmission lines are continuously exposed to wind-induced aeolian vibration, which may cause conductor fatigue and threaten grid reliability. To overcome the limitations of battery-dependent monitoring approaches, this work proposes a self-powered intelligent vibration monitoring system based on an electromagnetic-triboelectric hybrid nanogenerator (EMO-TENG). Within the system architecture, the electromagnetic generator (EMG) provides harvested energy for intermittent operation, while the triboelectric nanogenerator (TENG) serves as the vibration-sensing unit, forming an integrated energy-sensing structure. The EMG achieves a peak output power of 4.2 mW, enabling low-duty-cycle sensing and wireless transmission. To bridge physical sensing and vibration parameter estimation, a physics-informed signal feature extraction strategy is first developed, where dual-peak amplitude and interpeak interval characteristics are constructed based on the TENG contact-separation mechanism. These features capture intrinsic relationships between electrical output and mechanical vibration states. On this basis, an intelligent inversion framework using an improved group optimization algorithm (IGOA)-assisted CNN-GRU-attention neural network is introduced. The optimization algorithm performs adaptive hyperparameter search, enhancing nonlinear feature mapping and multiparameter decoupling capability. Experimental results demonstrate accurate quantification of vibration frequency, acceleration, and amplitude, with $R<^>{2}$ values exceeding 0.996. By integrating energy harvesting, physics-informed feature construction, and optimized intelligent inversion within a unified architecture, the proposed system provides a scalable and low-maintenance solution for autonomous transmission-line vibration monitoring and contributes a system-level methodology for intelligent electric Internet of Things (eIoTs) sensing.

Abstract Electromagnetic vibration pulse based condition monitoring has shown considerable potential for power electronic devices. Existing studies mainly rely on single turn-on or turn-off electromagnetic vibration pulse signals extracted from double-pulse tests, which are not fully representative of practical converter operation under pulse width modulation. In this paper, continuous electromagnetic vibration pulse signals excited by multi-cycle switching in TO-247 packaged IGBT devices are investigated through theoretical analysis and experiments. The time- and frequency-domain characteristics under different switching periods and duty cycles are analyzed, revealing spectral and modal features specific to multi-cycle excitation. A regularization-enhanced cycle-stacking inversion method is proposed to reconstruct single-cycle signals from multi-cycle responses, and two indicators, Rqs and Rm, are introduced to evaluate inversion feasibility. Simulation and experimental results show that reliable inversion is achieved only when inter-cycle interaction is sufficiently weak. This study represents a first essential step toward EMVP-based degradation assessment under practical PWM operation by establishing the feasibility and limitations of extracting single-cycle features from continuous signals. The findings provide a methodological basis for subsequent fault diagnosis and degradation evaluation of IGBT devices using EMVP features reconstructed from continuous operating signals.

Eddy current testing (ECT) is an important nondestructive evaluation technique. However, in high-speed inspection, the excitation frequency in conventional ECT must be adjusted to match the specimen velocity, which aggravates the skin effect and limits subsurface sensitivity. Motion-induced eddy current testing (MIECT), in which eddy currents are generated by the relative motion between a static magnetic field and a conductive specimen, has recently attracted attention as an alternative approach for high-speed inspection. Existing MIECT studies have mainly focused on defect detection, whereas simultaneous estimation of thickness and conductivity remains challenging. In this paper, an equivalent-frequency-based inversion method is proposed for MIECT. A theoretical model is established to relate probe velocity to an equivalent excitation frequency, showing that varying the specimen velocity can emulate the role of frequency sweeping in conventional multifrequency ECT. Theoretical and numerical validation shows that the proposed equivalent-frequency formulation can estimate the MIECT skin depth over 2–50 m/s with errors below 5.5%. Based on this principle, an improved Dream Optimization Algorithm (IDOA) is employed to achieve simultaneous inversion of specimen thickness and conductivity using multi-velocity MIECT signals. Experimental validation, limited by the maximum speed of the rotating-disk platform, was conducted over 2–10 m/s. Repeated experimental results show that the thickness inversion error is below 6%, while the conductivity inversion error is below 5.2%. The proposed method extends MIECT from defect detection to specimen parameter inversion and provides a feasible framework for simultaneous estimation of thickness and conductivity.

Copper and aluminum foils, serving as current collectors in lithium-ion batteries, are manufactured by continuously rolling thick plates to the desired thin-foil thickness. Once defects occur during the rolling of thick plates, they inevitably propagate into the foils, thereby compromising production quality and further deteriorating the electrochemical performance and safety of batteries. Conventional detection methods based on a single permanent magnet or DC coil often generate weak responses to small defects, making them unsuitable for highspeed manufacturing. To overcome this limitation, this study proposes a magnetic-field-reinforced excitation structure and establishes a theoretical model to analyze the effects of magnet spacing and lift-off distance on magnetic field characteristics. Finite element simulations are employed to investigate their influence on motion-induced eddy current (MIEC) signals. Experimental results show that the proposed structure enhances defect signal amplitudes by 59.9-92.7 % compared with single magnet configurations, with particularly significant improvements for small defect sizes and short lift-off distances. This work provides an effective method for highspeed defect detection and quality control of copper and aluminum foils in battery manufacturing.

With the rapidly increasing demand for metal products, production lines operate at ever higher speeds, making motion-induced eddy current testing (MIECT) an effective technique for the nondestructive evaluation of conductive materials under high-speed motion. However, in conventional eddy current responses, electrical conductivity and specimen thickness are intrinsically coupled, which makes their simultaneous determination challenging. This article proposes a model-driven inversion method based on a compact one-magnet–two-sensor (1M2S) configuration for the synchronous estimation of conductivity and thickness of a moving conductive plate. By exploiting the motion-induced upstream–downstream asymmetry of the magnetic field, the proposed 1M2S configuration generates two physically nonequivalent observations under the same excitation. These dual-location magnetic field components exhibit complementary sensitivities to conductivity and thickness and thus provide an independent information source for parameter decoupling. A Jacobian-based identifiability analysis further shows that neither motion alone nor dual-location sampling alone is sufficient; rather, reliable simultaneous estimation is enabled only by their combination, which renders the inverse problem locally identifiable. Numerical simulations verify the identifiability, robustness, and convergence behavior of the proposed method over a prescribed conductivity–thickness range, while experiments on 2011 and 1100 aluminum alloy specimens with a nominal thickness of 10 mm demonstrate its practical feasibility and repeatability under the tested conditions.

Bond-wire lift-off is a common failure mode in wire-bonded IGBT modules and is difficult to detect and localize using conventional electrical indicators. This paper proposes an electromagnetic voiceprint (EMVP)-based method for bond-wire lift-off fault detection and localization in multi-chip IGBT modules. A finite-element model is established to analyze how bond-wire cut-off changes the current distribution and electromagnetic excitation at different chip locations, and the peak-to-peak EMVP amplitude is adopted as the fault indicator. Simulation results show that this indicator increases monotonically with fault severity and exhibits clear fitting relationships at different sensing positions, with the highest R2 reaching 0.9972. Experimental results on three modules confirm the same trend and show good module-to-module repeatability. At the two sensing positions, the normalized peak-to-peak indicator follows approximately linear and quadratic relationships, respectively, with R2 values of 0.963 and 0.999 for the averaged responses. Compared with the on-state voltage and Miller plateau duration, the proposed indicator exhibits larger relative variation, especially at low lift-off fault severities. For fault localization, a Singular Spectrum Analysis (SSA)-based modal deviation metric is further introduced. The deviation at the lift-off location remains consistently higher than that at the intact location, providing at least an eightfold separation in the present experiments. These results demonstrate the feasibility and sensitivity of the proposed EMVP-based method for bond-wire lift-off fault detection and localization under controlled test conditions.

Motion-induced eddy current (MIEC) testing is a nondestructive testing method that can effectively detect defects at high speeds. However, the design and optimization of array probe for defect imaging remain underexplored. First, a three-dimensional analytical model is proposed by solving the Maxwell's equations, allowing calculation of the MIEC density and magnetic field. Then, the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) algorithm is applied to the model to design the arrayed magnets, which results in a uniform eddy current distribution and reduces the sensitivity differences across sensing positions. Finite element simulations and experiments are performed to verify the optimization results. Finally, high-speed defect detection experiments are conducted using the optimized MIEC probe. Experimental results show that array probe can realize defect imaging and identify crack orientations.

With the increasing demand for stainless-steel strips, accurately detecting defects during high-speed production has become essential for ensuring product quality. Motion-induced eddy current (MIEC) testing is a nondestructive testing (NDT) method particularly suitable for detecting defects in high-speed moving materials. Existing research typically utilizes a single permanent magnet as the excitation source to detect surface defects, without optimizing the excitation structure for the detection of buried defects. In this study, an opposing-pole structure is proposed. The magnetic field characteristics of this structure are theoretically analyzed, and its design parameters are optimized through numerical simulations. Based on the simulation results, an experimental platform is established to validate the effectiveness of the proposed structure. Experimental results show that, when detecting five types of groove-shaped defects buried 0.5 mm below the surface of 316 stainless steel, the MIEC signal amplitude obtained using the opposing-pole structure is more than 125% of that generated by the single-magnet excitation structure. Furthermore, the signal enhancement decreases with defect size, confirming the superior performance of the opposing-pole structure for detecting deeply buried defects.

During the cracking process of concrete, electromagnetic radiation (EMR) signals are emitted as a result of transient stress variations occurring on crack surfaces. By analyzing the characteristics of these EMR signals, researchers can monitor the health status of concrete structures. This study investigates the mechanical and electromagnetic properties of concrete during the cracking process while proposing mechanisms for piezoelectric and frictional charge generation on crack surfaces, based on the theory of electric dipole oscillations. Considering the effects of crack propagation velocity and surface stress, a model for the emission of EMR signals during the propagation of a single crack in concrete has been developed. The model correlates crack distance, applied stress, and the amplitude and frequency of EMR signals. Experimental observations from point load and hydraulic fracturing tests were conducted to investigate the variation patterns of EMR signals within the frequency range of 100 Hz-100 kHz during the evolution of single and multiple cracks in concrete. The experimental results indicate that the amplitude of the EMR signals generated during concrete cracking is directly proportional to the applied stress while inversely proportional to the square of the observation distance. The proposed signal generation and emission mechanism is in strong agreement with the observed experimental phenomena. This theoretical framework provides a deeper understanding of the intrinsic causes of EMR signal generation during concrete cracking, enhancing the interpretability and applicability of EMR-based monitoring techniques for concrete cracking.

The presence of defects significantly affects the performance of flexible graphite sheets. To strengthen quality control, we propose a non-destructive testing method for detecting internal defects in the production process of flexible graphite using the motion induced eddy current detection principle. Experimental results demonstrate that the designed apparatus can successfully detect through-holes resembling defects exceeding 500 mu m on two pieces of flexible graphite with different thicknesses and achieve precise localization. This method proves effective in detecting both visible and hidden defects, constituting a valuable asset for future production processes of flexible graphite. Its application is crucial for continuously monitoring the quality of flexible graphite and lays a solid foundation for future high-quality applications.

Monitoring and evaluating the aging state of power semiconductor devices is crucial for the reliability of power electronic converters. Existing methods exhibit low sensitivity and struggle to quantify the early degradation of the solder layer in power devices. During switching transients, power semiconductor devices emit electromagnetic voiceprint (EMVP) signals containing aging information, which can be captured by acoustic emission (AE) sensors. However, the impact of power electronic device degradation on the measured EMVP signals has not been thoroughly investigated. This study investigates insulated gate bipolar transistor (IGBT) devices and employs a finite element model to simulate EMVP signals under different solder layer void ratios. The results reveal that the energy accumulation curve can serve as an effective indicator of device aging. To validate this finding experimentally, an AE sensor was used to measure EMVP signals during the turn-off transient of IGBT devices. The measurements indicate that as the number of thermal cycling aging cycles increases, the maximum accumulated EMVP energy exhibits a linear growth trend. Compared to conventional aging characterization methods, such as conduction voltage and thermal resistance, the EMVP method demonstrates significantly higher sensitivity to early-stage minor degradation. These findings highlight the potential of the EMVP method for accurate aging state monitoring and characterization of power electronic devices.

Concrete generates weak electromagnetic radiation (EMR) signals during the cracking process. The EMR monitoring method for concrete cracking is a noncontact and nondestructive monitoring approach. However, substantial interference from environmental EMR noise degrades the consistency of monitoring signals, resulting in low recognition accuracy and a limited monitoring range. To address this, this study introduces a resonance noise reduction method based on the structure of EMR monitoring sensors. The proposed approach enhances the clarity and interpretability of monitored EMR signals through frequency-domain amplitude transformation and discrete wavelet decomposition. The time-frequency characteristics of both EMR signals and noise are comprehensively analyzed by utilizing continuous wavelet transform. A parallel network architecture integrating a convolutional neural network (CNN) and long short-term memory (LSTM) is employed to extract features from EMR signals associated with concrete cracking. Additionally, multiscale convolutional kernels and attention mechanisms (AMs) are integrated according to the time-frequency characteristics of the signals. Experimental results reveal that the proposed method effectively distinguishes EMR signals from four common types of EMR noise, achieving an average signal recognition accuracy of over 98% on the validation and test datasets, with a crack identification accuracy rate of 94.6%. This approach significantly enhances the applicability and potential use of EMR monitoring in concrete cracking scenarios.

Health status monitoring of power electronic devices is essential for ensuring their safe and reliable operation. In recent years, monitoring switching transient electromagnetic voiceprint (EMVP) signals has emerged as a promising new indicator for assessing device health. However, the identification of aging states based on EMVP still relies on manual interpretation, leading to low accuracy. This article presents a health status monitoring method for insulated gate bipolar transistor (IGBT) devices using EMVP signals, applicable across various operating conditions. Specifically, we propose a spatiotemporal feature fusion cross attention neural network for aging state identification. Experimental results demonstrate that the network achieves an accuracy of over 95% in detecting the aging state of IGBT devices. Furthermore, a transfer learning approach is introduced to improve the model's effectiveness and generalization ability when working with small sample datasets. The proposed monitoring method facilitates accurate and rapid evaluation of IGBT device aging states under diverse operational conditions.

State monitoring technology is essential for the fault prognostics and health management of power electronic devices. Electromagnetic voiceprint (EMVP) signals, produced during the switching transients of these devices, carry substantial information about their health status. In recent years, EMVP signals have gained attention as a promising state monitoring indicator for power electronic devices. However, the mechanism by which the health status of a device affects its EMVP signals remains unclear, limiting their use in health monitoring. This article investigates the impact of insulated gate bipolar transistor (IGBT) health status on EMVP signals and develops a mechanistic model to illustrate the influence of bond wire lift-off and solder aging on EMVP signals. Simulation models were developed to analyze the impact of both scenarios on EMVP signals. Typical features are proposed to characterize the relationship between IGBT health status and EMVP signal. Experimental validation supports the theoretical framework and simulation results. The research provides a theoretical foundation for the development of reliability evaluation methods based on EMVP.

Defects in flexible graphite sheets can significantly reduce their performance. However, existing detection methods, such as X-ray and ultrasonic testing, have limitations in real-time, non-contact, and comprehensive detection. To address these issues, this paper proposes a defect detection technique for flexible graphite sheets based on Lorentz force eddy current (LET) testing. The design and structural optimisation of the detection sensor were systematically studied through finite element analysis and an improved Multi-Objective Manta Ray Foraging Optimisation (IMOMRFO) algorithm. A corresponding detection system was subsequently developed. Experimental results demonstrate that the designed sensor and detection device can effectively identify defects larger than 500 mu m in size at a speed of 7.5 m/s. Furthermore, the detection sensitivity is found to be directly proportional to the relative motion speed. This method offers a non-contact, high-speed defect screening capability, making it suitable for industrial production lines and filling an important gap in the quality control of flexible graphite materials.

Lithium-ion batteries are recognized as a critical technology for modern energy storage, with copper foil frequently used as the current collector for the anode. However, the presence of micro-defects in copper foil can significantly degrade the electrochemical performance of lithium-ion batteries. Therefore, a method capable of accurately detecting micro-defects during high-speed production and subsequently quantifying their sizes is required. Motion-induced eddy current (MIEC) technology, characterized by relative motion, has been employed to detect defects in conductive materials during high-speed movement. However, limited research has been conducted on the quantification of defect sizes based on MIEC signals. In this study, a novel method for quantifying the size of micro-defects in copper foil based on MIEC signal features is proposed. Wavelet soft-thresholding and morphological filtering are combined to effectively remove high-frequency noise and signal fluctuations caused by variations in lift-off distance. Eight distinct features of the micro-defect-induced eddy current signals are defined, and a feature dataset for various sizes of copper foil micro-defects at different motor speeds is established. The IVY algorithm, known for its excellent optimization speed, is used to optimize the hyperparameters of a conventional convolutional neural network (CNN)-bidirectional long short-term memory (BiLSTM)-Attention model, resulting in the development of the IVY-CNN-BiLSTM-Attention model. The experimental results demonstrate that superior performance is achieved, with accurate micro-defect size quantification in copper foil.

In the production and manufacturing of non-ferromagnetic materials, precise control of motion speed and accurate detection of defects are critical to ensuring the quality and reliability of the final product. However, existing sensors face limitations in practical applications, as they are unable to simultaneously measure the motion velocity and detect defects in non-ferromagnetic thin plate materials. To address this issue, this study presents a Lorentz force sensor based on the principle of the Lorentz Force Particle Analyzer (LFPA), which is capable of simultaneously measuring the motion velocity and micro-defects in non-ferromagnetic thin plate materials. To enhance the sensor's detection sensitivity while mitigating the effects of the permanent magnet's weight, a detailed analysis of the influence of support beam parameters on stiffness was conducted, and the folded beam parameters were optimized through numerical simulations. Static performance tests confirmed that the sensor can accurately detect force variations on the order of micro-Newtons. Experimental testing demonstrated that the sensor is capable of detecting planar defects as small as 500 mu m in 0.1 mm thick copper foil. Compared to similar sensors, this sensor can detect defects approximately 3.5% of the size of those detectable by other sensors, thereby exhibiting higher sensitivity. The design and implementation of this sensor significantly improve the accuracy and operational ease of on-site measurements in the production and manufacturing of non-ferromagnetic thin plate materials, providing technical support for optimizing production processes and enhancing product quality.

As production speeds continue to increase, accurately measuring defects during the high-speed motion of samples becomes of significant practical importance. Motion-induced eddy current (MIEC) is a nondestructive testing method well-suited for detecting defects in high-speed moving samples. Existing studies typically employ permanent magnets or coils as excitation sources, without addressing the impact of the central magnetic field and magnetic field gradient of the excitation source on the generated MIEC signals. This study analyzes, both theoretically and through simulations, the influence of the central magnetic field and magnetic field gradient on MIEC signals. Based on these findings, a new excitation source structure is designed, and an improved multiobjective RIME optimization algorithm is proposed to optimize the shape of the magnetic yoke surrounding the permanent magnet. Numerical simulations demonstrate that the optimized excitation source structure significantly enhances the amplitude of defect signals across various sample motion speeds. Finally, an experimental setup is constructed to validate the results. The experimental results indicate that the optimized excitation source structure effectively enhances the MIEC signal amplitude for five different slot-shaped defects ranging from 0.5 to 2.5 mm in length on the aluminum disk, with a minimum amplitude increase of 52% and a maximum of 74.1%.

Lithium-ion batteries are a key technology in the energy storage field, and aluminum foil, a common cathode current collector, can develop micro-defects that significantly affect battery performance, especially during high-speed manufacturing. Motion-induced eddy current (MIEC) sensing is effective for detecting defects in conductive materials. However, lift-off fluctuations degrade the signal-to-noise ratio (SNR), making it challenging to detect small or shallow defects. Traditional methods, such as differential sensor outputs, help reduce these fluctuations, but baseline signal deviations caused by varying relative positions between the sensor and the permanent magnet limit their effectiveness. To address this, this study proposes a spatial compensation method for differential tunnel magnetoresistance (TMR) sensors, which reduces baseline deviations caused by position differences, mitigating lift-off fluctuations and improving the SNR for micro-defect detection.

Copper and aluminum foils serve as predominant materials in fluid collectors, and defects within them can significantly impact the electrochemical performance of cells. However, existing methods for detecting defects within non-ferromagnetic thin metals, such as copper and aluminum foils, have several limitations. This study aims to address the need for detecting micrometer-scale defects on 0.1 mm copper foils, aligning with industrial field requirements. We devised an inspection device based on the induced magnetic field detection principle and explored the impact of copper foil undulations on micrometer-scale defect detection using COMSOL modeling. Subsequently, we introduced a coherent cumulative-differential algorithm to effectively mitigate the influences of circuit noise and sampling heave noise on defect signals. Consequently, the signal-to-noise ratios of 100- and 200-micron defect signals were significantly improved by 157% and 234%, respectively. This approach shows promise for detecting micrometer-scale defects in non-ferromagnetic thin metals and lays a robust foundation for future defect identification and inversion endeavors.