Abstract:Metal oxide semiconductor (MOS) gas sensors require their sensing materials to operate at temperatures ranging from 200℃ to 500℃ to achieve sufficient and controllable chemical reactions with target gases. Micro-electro mechanical systems (MEMS) technology enables the monolithic integration of gas-sensitive films, heaters, and signal processing circuits on a single chip, significantly reducing power consumption and device size. Among these components, the micro-heater, which provides a stable operating temperature for the gas sensor, plays a crucial role in determining the overall sensor performance. The electrode morphology, dimensions, and materials of the micro-heater directly influence key characteristics such as temperature distribution, power consumption, and mechanical stress. Furthermore, the temperature uniformity, operating range, thermal response time, power efficiency, and mechanical stability of the micro-heater collectively affect the sensitivity, selectivity, lifetime, and reliability of the sensor. This review focuses on recent advances in micro-heater design over the past five years and examines how optimized designs impact sensor performance. Firstly, the gas-sensing mechanism of semiconductor materials and the influence of operating temperature on sensor performance are introduced. Based on this, the theoretical foundations of heat conduction, convection, and radiation in micro-heaters are presented, along with various modeling and optimization approaches. Secondly, recent research progress on the morphological and structural design of micro-heaters is elaborated, covering geometric configurations, thermal isolation structures, suspension beam optimization, and micro-hotplate arrays. The effects of these structural improvements on gas-sensing performance are also discussed. Subsequently, different materials used in micro-heater fabrication are reviewed, with an evaluation of their mechanical stability and electrothermal properties. Finally, the current research status and key performance parameters are summarized, and future research directions are outlined, providing insights for enhancing semiconductor gas sensor performance through micro-heater optimization.

Abstract:In single-beam spin-exchange relaxation-free (SERF) atomic magnetometers, strong optical absorption induces pronounced polarization nonuniformity in the vapor cell, degrading the magnetometer performance, while the compact configuration hinders further improvement of the polarization uniformity. To address this issue, a pulsed optical pumping scheme for single-beam SERF atomic magnetometers is proposed, in which short-duration, high-power pump pulses are employed to replace continuous-wave pumping. This approach suppresses absorption-induced polarization gradients while preserving the single-beam SERF architecture. Based on a bias-field-assisted mode, we established dynamic and response signal models for the magnetometer under pulsed pumping. Analytical solutions of the models demonstrate that increasing pump power is beneficial for enhancing polarization uniformity. By considering both the magnetometer response and the photodetector shot noise, the optimal duty cycle maximizing the signal-to-noise ratio was theoretically determined to be 37%, which is close to that at 50%. Accommodating the compactness of the single-beam configuration, a miniaturized prototype based on a 1×2 fiber-optic switch was implemented, and a 50% duty cycle was adopted to enable time-division multiplexing of two magnetometers with nearly unchanged signal-to-noise ratio. Experimental results showed that under 50% duty-cycle pulsed pumping, the optimal pump power for maximum magnetometer response in a 3 mm 87Rb vapor cell increased from 0.9 mW (steady-state pumping) to 1.7 mW. With the polarization uniformity defined as the ratio of the average polarization over the cell volume to that at the beam entrance, the uniformity was improved by 46%, leading to an enhancement in magnetic field sensitivity from 14 fT/Hz1/2 to 12 fT/Hz1/2. The proposed method enables sensitivity enhancement in arrayed and integrated ultra-weak magnetic field sensors, with potential applications in high-performance magnetocardiography and magnetoencephalography.

Abstract:To address the critical need for precise detection of early-stage micro-cracks originating from stress concentration around fastener holes in aerospace structures, and to overcome the challenges associated with the poor structural compatibility of traditional piezoelectric ultrasonic sensors and the limited sensitivity of conventional linear guided wave methods, this study proposes a nonlinear guided wave detection method based on a direct-write piezoelectric ultrasonic sensor array, aiming to achieve high-sensitivity identification of micron-scale cracks. The method involves the design and fabrication of a locally enhanced sensor array by integrating the nonlinear effects of guided waves. This array consists of an outer annular excitation element for the fundamental frequency (0.5 MHz) and an inner arc-shaped receiving element optimized for the second harmonic frequency (1 MHz), thereby optimizing the excitation and reception of nonlinear signals. Utilizing the fabrication process of direct-write piezoelectric ultrasonic sensors, an ultrathin and flexible poly vinylidene fluoride-trifluoroethylene (P(VDF-TrFE)) copolymer piezoelectric ultrasonic sensor array is achieved. This enables in-situ and consistent integration with the surface of the structure under test, effectively eliminating the signal instability issues associated with traditional coupling methods. Building on this foundation, the study introduces a multi-path pulse transmission-reception detection strategy. Frequency-domain analysis and feature extraction of the received ultrasonic signals are performed to calculate the relative nonlinear coefficient, a key parameter characterizing crack-induced nonlinearity. This facilitates crack location identification and establishes a quantitative relationship with crack lengths (2.5,4,6 mm). Experimental results demonstrate that the proposed method effectively captures the evolution trend of nonlinear guided wave features, enabling reliable detection of both the size and propagation direction of early-stage micro-cracks. This work provides a promising new technical pathway for the in-situ, high-precision inspection of critical regions in aerospace structures.

Abstract:In low-cost strapdown inertial navigation systems (SINS), the determination of horizontal attitude relies on the specific force measurements of accelerometers to track the gravity vector. However, the specific force measured by accelerometers couples motion acceleration and gravitational acceleration under the maneuvering conditions. Both components are difficult to separate effectively, leading to the increased errors of gravityreferenced horizontal attitude estimation and severely restricting the practicability and reliability of low-cost SINS in dynamic scenarios. To address this issue, this paper proposes a low-cost horizontal attitude measurement method aided by the external-velocity, which avoids the complex dynamic modeling and specific scenario assumptions. The core of this method is to use the external velocity information to directly estimate and compensate the horizontal attitude errors caused by motion accelerations. First, the SINS update algorithm and error model are simplified according to the accuracy characteristics of microelectro-mechanical system (MEMS) inertial sensors. Second, an analytical relationship between the east/north velocities and the horizontal misalignment angle is established. Accordingly a measurement equation is constructed using the external velocity information to directly estimate the horizontal misalignment angle induced by the motion acceleration, thereby enabling the realtime correction of horizontal attitude. To verify the effectiveness of proposed method, both the turntable tests simulating a swaying environment and vehicle tests under the real maneuvering conditions were conducted. Experimental results indicate that the horizontal attitude measurement accuracy of the proposed method is better than 0.02° (RMS), which is comparable to that of the gravity-referenced method in the swaying environment. Under the maneuvering conditions, the root mean square errors (RMSE) of pitch and roll are reduced by 44.8% and 47.3% compared with the conventional integrated navigation algorithms. These results demonstrate the effectiveness and robustness of the proposed method in dynamic environments.

Abstract:To address the problems that traditional electromechanical and piezoelectric/piezoresistive sensors are susceptible to interference in extreme electromagnetic environments such as explosive shock waves, exhibit response lag, and have difficulty capturing the rising edge, this paper proposes and fabricates an all-silica Fizeau-cavity optical fiber miniature dynamic pressure sensor based on the Fabry-Pérot (F-P) interference principle. The sensor integrates the characteristics of high sensitivity, high frequency response, and high spatial resolution. Its sensing unit is formed by fusion splicing a 125 μm single-mode fiber, a silica capillary tube, and a coreless fiber. By length-limited grinding and 40% HF chemical etching, the thickness of the coreless-fiber pressure-sensitive diaphragm is precisely controlled to 2~3 μm, achieving an approximately ideal two-beam sinusoidal interference output. To enable high-precision inversion of transient pressure, a three-wavelength light source excitation scheme and a passive homodyne demodulation technique with arbitrary deterministic phase spacing are adopted to extract, in real time, the interferometric phase shift induced by cavity length variations under shock loading. In combination with three-channel voltage normalization and a phase-jump compensation algorithm, the influence of insertion-loss differences and quantization noise on measurement accuracy is effectively suppressed. Static calibration experiments show that the sensor operates within the elastic regime over the entire measurement range; within 0~60 MPa, the full-scale nonlinearity reaches 2.23%, the repeatability is better than 2%, and the hysteresis error is less than 0.1%. Shock-tube dynamic test results indicate that when the resonance frequency of the high-speed photodetector is set to 20 MHz and the response time is 8 ns, the dynamic response time of the sensor system is less than 50 ns, which allows for the accurate reconstruction of the shock-wave pressure-time waveform. With the merits of microscale sensing and high-frequency response, the sensor is suitable for transient pressure measurements under extreme conditions such as explosive impacts and intense laser-induced near-field plasma shocks, and it can maintain stable output under strong electromagnetic and high-shock conditions, thus exhibiting considerable engineering application value.

Abstract:To address the strong coupling between the structural parameters of micro-electromechanical systems (MEMS) gyroscope and the electrical parameters of capacitive readout circuit, which limits the improvement of the output signal-to-noise ratio (SNR), this paper proposes a mechatronic integrated modeling and force-electrical co-optimization approach for a fully decoupled tuning-fork resonant MEMS gyroscope. An electromechanically-coupled dynamic model of the fully decoupled microstructure with dual-mass anti-phase drive and differential sensing, and a noise model of the capacitive readout interface circuit, are established. Analytical expressions for the equivalent stiffness of the decoupling beams and the structural modal frequencies are derived. The influence of structural dimensional parameters and circuit impedance parameters on signal noise is systematically analyzed. The critical beam dimensions of the microstructure and the capacitance/resistance combinations of the capacitive readout circuit serve as optimization variables, with the SNR as the optimization objective. The method of moving asymptotes (MMA) is employed to perform co-optimization of the electromechanical parameters of the MEMS gyroscope system, achieving an SNR improvement from 17.70 to 37.45 dB. Based on the optimized results, a printed circuit board (PCB) is designed for performance characterization and on-vehicle road test of the fabricated MEMS gyroscope. Experimental results show that the drive-mode and sense-mode resonant frequencies are 8 750.47 and 8 828.63 Hz, respectively, with errors of approximately 2.9% relative to theoretical predictions. Using the half-power bandwidth method, the quality factors (Q) of the drive and sense modes are measured to be 1 008.1 and 1 027.8, respectively. Single-axis rate-table tests yield a sensitivity of 0.486 9 mV/(°/s) and an SNR of 36.31 dB, with an error of about 3% relative to theoretical values. The zero-bias instability calculated by Allan deviation is 28.26(°)/h. Under typical dynamic driving conditions, including vehicle turning and roundabout maneuvering, the angular rate output exhibits good consistency with a high-precision reference gyroscope, thereby validating the effectiveness of the proposed electromechanical co-optimization design method from a system-level application perspective.

Abstract:To address high deployment complexity and cost associated with antenna-array-based radio frequency identification (RFID) positioning systems, this paper proposes a 3D positioning algorithm based on tag arrays. First, the RFID phase backscatter model is analyzed, and the coupling effect is experimentally investigated to reveal their potential impact on angle estimation accuracy. Second, to mitigate the mutual coupling within the tag array, we propose a virtual tag construction method considering the tag arrangement. By optimizing the tag array with sparse configurations, an angle of arrival (AOA) estimation method for sparse tag arrays based on virtual tag construction is developed. Furthermore, to address the interference caused by potential target pose variations in practical applications, an angle-search-based attitude estimation method is incorporated. This method employs a spatial coordinate transformation model to achieve precise estimation of the tag array′s true orientation relative to a known antenna, effectively reducing the negative impact of attitude angles on angular measurement precision. Subsequently, a 3D positioning model is established using one-dimensional angle measurements, where azimuth and elevation angles after attitude compensation are obtained from sparse tag subarrays in two orthogonal directions to achieve 3D positioning through triangulation. Finally, a systematic evaluation of the proposed sparse tag array AOA estimation and localization algorithms was performed on a test platform built with the commercial Impinj R420 reader and tags. Experimental results demonstrate that the proposed system achieves positioning in three-dimensional space, with a median positioning error of 23.5 cm and a standard deviation of 12.5 cm, validating the effectiveness of the proposed algorithm.

Abstract:Pressure gradient vector hydrophones are composed of multiple pairs of independent scalar hydrophones. However, the amplitude and phase errors between the scalar hydrophones can cause the distorted pressure difference directivity and degrade the direction-finding accuracy of underwater targets. To address this issue, an active rotational correction method using the inherent dipole directivity is proposed to correct the intrinsic amplitude and phase errors of pressuregradient vector hydrophone hardware system, which does not require the accurate position of auxiliary sound source. The vector hydrophone is rotated 360° on a precision turntable. Meanwhile a scalar hydrophone located off the rotation axis is selected as the reference, and the amplitude scaling and time-domain shifting optimization are applied to the signals of other hydrophones until the pressure difference directivity curves are corrected to their ideal shapes, thereby compensating for the amplitude and phase errors. Amplitude/phase error correction and sound source localization experiments are conducted in an anechoic tank using a laboratory-developed three-dimensional pressure-gradient vector hydrophone. The results show that the corrected pressure difference directivity curves of vector hydrophone closely approximate the ideal shapes. Compared to the results before correction, the number of test points with azimuth errors below 5° increases by 10.6%, and the number of test points with elevation errors below 10° increases by 92.9% in rotating hydrophone tests; meanwhile the number of test points with azimuth errors below 5° increases by 46.3% and the number of test points with elevation errors below 10° increases by 42.6% in moving sound source tests. Experimental results verify that the proposed method effectively corrects amplitude and phase errors of the pressure-gradient vector hydrophone, which significantly improves the accuracy of underwater target localization. This method reduces the difficulty and cost of amplitude and phase error correction, making it valuable for practical applications.

Abstract:The performance of an analog-to-digital converter (ADC) determines the quality of the entire acquisition system. As system sampling rates and bandwidths increase, nonlinear errors become significantly more detrimental than linear errors. To address frequency-dependent nonlinear errors in ADCs, this paper proposes a novel digital post-calibration method based on an artificial neural network (ANN). This method first conducts spectral analysis on the single-tone signal sampled by the ADC to remove nonlinear harmonics, and takes this result as the reference true value to train the artificial neural network from the perspective of the frequency domain. The proposed method takes the amplitude information of the signal spectrum as the training object of the artificial neural network, retains the phase information, and compensates the output amplitude result after calibration. The amplitude result is reconstructed with the retained phase information to form a complex frequency spectrum, and an inverse Fourier transform is performed to restore the time-domain signal. For experimental validation, a time-interleaved (TI) ADC system is utilized as the application scenario, where various inter-channel mismatch errors and the non-linear error of the ADC are jointly calibrated. Using multiple sets of single-tone signals with different frequencies, the data is divided and the neural network training is conducted by combining the strategy of frequency-layered sampling and time-segmented sampling. Furthermore, the generalization performance of the multi-tone signals is also verified. The universality of the proposed method in ADC nonlinear calibration applications is demonstrated. The network is also verified on a 4-channel TIADC hardware platform with a sampling rate of 20 GSPS, resulting in an overall improvement in the spurious-free dynamic range of the system by about 36 dB while ensuring that the multi-tone signals could maintain the correct phase relationship.

Abstract:With the increasing precision requirements for integrated circuit testing, the long-term operational stability of the test equipment has become a key factor affecting test quality and cost. Traditional monitoring methods based on fixed confidence-level thresholds, due to their static nature, struggle to adapt to the performance drift caused by aging and environmental fluctuations during long-term operation. As a result, they often suffer from false alarms or missed detections, which hinders the implementation of predictive maintenance. To address this issue, a dynamic threshold optimization monitoring method integrating Gaussian mixture model (GMM) and Kullback-Leibler divergence (KL divergence) is proposed. This method first uses a GMM to accurately model the multimodal distribution of test data, effectively characterizing the equipment′s operational state under complex multi-condition scenarios. It then introduces KL Divergence to quantify in real-time the distribution difference between the monitoring data and the healthy baseline model. Building on this, it innovatively updates the anomaly detection threshold based on a rolling historical KL Divergence sequence, allowing the threshold to adaptively adjust according to the natural drift in data distribution. This mechanism fundamentally overcomes the mismatch between static thresholds and dynamic processes, enhancing the monitoring system′s sensitivity to both gradual performance degradation and sudden anomalies. Experimental results show that compared to the traditional K-means clustering method with a fixed threshold, the proposed method achieves significant improvements in both anomaly detection accuracy and F1-score, enabling more sensitive and reliable identification of performance fluctuations and early-stage faults in test equipment. This method not only provides an effective technical solution for stability monitoring and predictive maintenance of integrated circuit test equipment, but also holds potential for extension to other industrial equipment health management fields due to its core framework of sensing distribution changes through probabilistic modeling and achieving threshold self-adaptation. This framework demonstrates strong generalizability, offering a new approach with continuous adaptation capabilities for equipment condition monitoring in dynamic operational environments.

Abstract:Pulse repetition interval (PRI) modulation type recognition of radar signals is a critical component for signal sorting and threat assessment in electronic warfare systems, as its accuracy directly affects battlefield electromagnetic situational awareness and the effectiveness of countermeasure strategies. To address the limitations of traditional methods—namely, strong reliance on expert knowledge and insufficient feature extraction capacity of single models—this paper proposes a deep neural network architecture based on multi-dimensional feature space-time fusion. The approach constructs a cascade of convolutional neural networks (CNN) and long short-term memory (LSTM) networks to capture the intrinsic correlations between the spatiotemporal characteristics of radar pulses and the temporal evolution of modulation patterns. A sequential attention mechanism is incorporated to weight and fuse the extracted temporal features, emphasizing key moments of PRI variation and enabling accurate identification of PRI modulation types. Experiments are conducted on a dataset comprising five typical modulation types: Fixed, jitter, group change, stagger, and slip. Results show that, under ideal interference-free conditions, the model achieves an overall recognition accuracy of 99.40%. Even under 70% pulse loss interference, recognition accuracy remains 70.93%, and under equal-strength false pulse interference, it reaches 96.13%. The proposed model significantly outperforms mainstream comparison models, including CNN, CNN-LSTM, GRU-Attention, and SE-NET, demonstrating enhanced accuracy and robustness in complex electromagnetic environments. Furthermore, the model requires only 1.54 ms for single-sample inference and has 653 814 parameters, offering strong real-time capability and potential for lightweight deployment. The proposed method provides an effective technical solution for real-time, accurate recognition of radar PRI modulation types in complex electromagnetic environments, with both theoretical significance and practical engineering applicability.

Abstract:Efficient path planning and decision-making are essential for the unmanned platforms to achieve the autonomous exploration in unknown environments. However, due to the limitations in perception accuracy and computational resources, the current exploration methods often suffer from the low efficiency and incomplete coverage in the structurally complex scenarios. In particular, the redundant paths and sensing blind zones frequently occur in cornered or occluded regions. This is mainly because most planners focus only on the local information gain or the shortest path, which fail to fully exploit environmental structural features and results in the high-cost backtracking and repeated exploration. To address these challenges, this study proposes a hierarchical active path-planning method based on the edge-and-corner regions for unmanned platforms. The proposed framework adopts a two-layer structure consisting of front-end path generation and back-end trajectory optimization to achieve the efficient and continuous global exploration. In the front-end stage, a fast environmental information preprocessing mechanism is introduced by combining with an adaptive viewpoint push-away and small disturbance optimization strategy to ensure the balanced coverage and uniform viewpoint distribution in corner regions. In the back-end stage, a multi-factor cost model with corner constraints integrated with path sequence optimization, B-spline smoothing, and terminal correction is designed to generate the continuous, safe, and executable trajectories. Experimental results demonstrate that, the proposed method reduces the average exploration time by 14.7%~18.2% and shortens the path length by 17.4%~39.7% compared with the two advanced algorithms in the typical corner-rich environments. Meanwhile, it maintains an average coverage rate over 96.6% by effectively balancing the exploration efficiency and path optimization performance, thereby significantly improving the overall exploration quality and planning effectiveness. The comparison with the learning-based method further verifies its stability and adaptability in the structurally complex scenarios. Furthermore, the real-world indoor experiments validate the feasibility and adaptability of proposed algorithm, confirming its potential for the practical deployment in the complex exploration tasks.

Abstract:Master-slave control is one of the key methods for realizing the control of robot-assisted puncture operations. However, due to the complexity and diversity of puncture-assisted robots as well as their sensing and feedback methods, there exist significant challenges in the rapid construction of master-slave control for surgical robot systems. This paper proposes a rapid construction method for master-slave control of puncture surgical robots, which enables a rapid construction of the master-slave motion mapping and a rapid implementation of the master-slave control. Firstly, a coordinate system for the object motion is established, and the relative position transformation chain for the coordinate transformations of object motion mapping is further defined and constructed to realize explicit expression of the coordinate transformations of the object motions, and thus achieving the rapid construction of the motion mapping. Secondly, a rapid construction strategy for the master-slave motion mapping based on an intuitive mapping model is proposed. For the slave objects with various sensing and feedback modes, the relative position transformation chain for the master-slave motion mapping is quickly constructed to enable the calculation of the multiple coordinate transformations for the master-slave control and to further realize the rapid realization of the customized and diversified master-slave control. Experimental studies are then conducted using the proposed method, and the rapid constructions and implementations of 4 types of the master-slave control for two master objects and two slave objects are implemented, respectively. The experimental results show that: The proposed rapid construction strategy for the master-slave control mapping can realize a rapid construction of the master-slave motion mapping with various sensing feedbacks, and 4 relative position transformation chains for the master-slave motion mapping are obtained, where each transformation chain represents an independent type of the master-slave control to accommodate the requirements of different application scenarios; through the control calculation of the multiple coordinate transformations based on the relative position transformation chains, the rapid implementation of the master-slave control is achieved, which verifies the feasibility and practicality of the proposed rapid construction method for the master-slave control.

Abstract:The traditional hand rehabilitation robots typically employ multiple actuators to independently control finger motions with the high system complexity and cost, which limits the lightweight design and practical deployment. Although the underactuated mechanisms can reduce the number of actuators, their cable length adjustment capability is structurally constrained, which limits the adaptability to different hand sizes and initial postures and also degrades the motion response speed. To address these issues, a slack-adjustable underactuated cable-driven actuator is proposed for hand rehabilitation robots, where the actuator employs a dual-pulley differential mechanism to separate the driving cable from the finger cables. Here a knob-based adjustment mechanism is introduced to independently regulate the initial slack of each finger cable without relying on the motor operation, which effectively reduces the initial dead zones and improves the motion response performance. Meanwhile, the finger cables are arranged in a differential mechanism within the dual-pulley system, enabling the automatic redistribution of cable length and tension according to different finger blockage conditions during grasping, thereby realizing the adaptive grasping. The system is driven by a single motor and integrates gear transmission with the dual-pulley differential mechanism to realize the coordinated control of thumb, index finger, and middle finger. Experimental results show that the proposed slack adjustment mechanism reduces the average system response time by about 91.5% with an initial slack of approximately 10 mm, which significantly enhances the motion responsiveness. Additionally the stable adaptive grasping performance is maintained under the various finger-blocking scenarios. The measured fingertip forces of thumb, index finger, and middle finger reach 8.85, 8.29, and 7.84 N, respectively, which meet the force assistance requirements for the hand rehabilitation training.

Abstract:With the continuous expansion of power grid scale in recent years, a vast amount of unstructured equipment defect records has been collected and accumulated. This data contains entity information crucial for equipment condition assessment and operational decision-making. However, the prevalent nested entity structures in such data lead to increased entity boundary ambiguity and contextual semantic complexity, posing significant challenges to traditional named entity recognition methods. To achieve accurate identification of nested entities in defect records, this paper takes power transformers as a typical example and proposes a named entity recognition method for power transformer defect records that combines pre-training bidirectional encoder representations from transformers with permuted language model(PERT), bidirectional gated recurrent unit(BiGRU), and efficient global pointer(EGP). This method first employs the PERT model as a vector embedding layer for dynamic semantic encoding, leveraging its permuted pre-training characteristics to deeply capture contextual dependencies. Subsequently, a BiGRU network is introduced as the text encoding layer to comprehensively extract semantic features through its bidirectional gating mechanism. Finally, EGP is adopted as the decoding output layer to focus on entity spans and boundaries, enabling precise extraction of nested entities while avoiding the label conflict issues associated with traditional CRF decoding. Experimental results show that this entity recognition method effectively addresses the challenges of nested entities, achieving a comprehensive F1 score of 96.01%, which is 0.55% and 0.70% higher than those of the bidirectional encoder representations from transformers-biaffine attention(BERT-Biaffine) and bidirectional encoder representations from transformers-machine reading comprehension(BERT-MRC), respectively. It attained the highest F1 scores across all five entity label categories. Specifically, for defect equipment and defect component recognition, where nesting is most prominent, the F1 scores of the proposed method reached 100% and 94.74%, representing improvements of 0.57% and 0.13% over the best baseline models.

Abstract:To address the common issues in traditional robotic arm path planning methods, which include significant discrepancies between simulation and reality, low search efficiency, and limitations in path reliability and executability, a digital twin-based robotic arm path planning method is proposed. Firstly, a digital twin platform is constructed according to the actual operating environment of an industrial robotic arm, enabling bidirectional virtual-real mapping and real-time data interaction between the physical robotic arm and its digital twin model, thus providing a real-time and accurate digital twin simulation platform for the simulation verification and practical execution of robotic arm path planning algorithms. Secondly, at the path planning algorithm level, an adaptive gradient-based bidirectional rapidly-exploring random tree* (AG-BI-RRT*) algorithm is proposed. The algorithm incorporates an adaptive conical sampling method based on historical gradient feedback, three expansion strategies (target bias expansion, improved artificial potential field expansion, and random direction expansion), and a multi-factor parent node reselection strategy. These improvements comprehensively optimize the algorithm in terms of search efficiency, obstacle avoidance capability, and path quality, effectively improving the efficiency and quality of path planning. Finally, a path optimization method is introduced to generate a smooth collision-free path through greedy pruning and B-spline smoothing optimization. Comprehensive simulation experiments and physical robotic arm experiments verify the feasibility and effectiveness of the proposed method. Compared with existing methods, AG-BI-RRT* achieves superior performance in path length, iteration time, number of search nodes, and path steering angle. The joint angle deviation between the digital twin model and the physical robotic arm does not exceed ±0.01°, and the average response time between the physical robotic arm and the twin model is 176.721 ms, satisfying the real-time and consistency requirements of digital twin systems. This work provides an effective solution for robotic arm path planning in a digital twin environment.

Abstract:The spaceborne hyperspectral imager′s spectral and imaging performance may be altered by the on-orbit vacuum and low-temperature environment. To achieve high-precision quantitative applications, system calibration must be conducted on the ground by simulating the on-orbit environment. This paper takes the advanced hyperspectral imager (AHSI) onboard the GF-5(02) satellite as the research subject and innovatively proposes a synergistic testing method combining thermal focusing adjustment and vacuum calibration, along with constructing a corresponding experimental system. By adopting a thermal focusing strategy of overall temperature variation combined with local fine-tuning, precise calibration of the imager′s focal plane in a vacuum environment was achieved. Subsequently, vacuum spectral calibration was performed, revealing the thermal-optical coupling mechanism responsible for spectral shifts in the shortwave infrared band. The results indicate that the spectral resolution of AHSI is better than 4.83 nm in the visible-near-infrared band and 8.97 nm in the shortwave infrared band. Compared with calibration under normal temperature and pressure, the center wavelength of the shortwave infrared channel exhibited an average shift of 1.83 nm, which is attributed to thermal deformation of the shortwave infrared detection cooling assembly under vacuum and low-temperature conditions. This synergistic calibration system effectively integrates thermal focusing adjustment and vacuum spectral calibration, enabling accurate characterization of spectral and imaging performance under on-orbit working conditions. The system has been successfully applied in the on-orbit detection of methane point source emissions in the Permian Basin in the United States and alteration information of the Liba gold deposit in the West Qinling region of China, demonstrating its high-precision detection capability in ecological environment monitoring and mineral resource exploration. This study provides systematic technical methods, engineering practice references, and a replicable experimental paradigm for pre-launch vacuum calibration of spaceborne hyperspectral instruments, with notable innovations in the thermal focusing strategy for highly integrated optical systems and the analysis of thermal-spectral shift mechanisms.

Abstract:Atomic clocks are core components of satellite navigation and precision timekeeping systems. However, their signal quality is often compromised by anomalies. To address the limited adaptability of the traditional ordinary least squares (OLS) method to complex anomaly patterns in integrity monitoring, this paper proposes an interference-resistant modeling and anomaly repair method based on the random sample consensus (RANSAC) algorithm. This method utilizes RANSAC to construct highly robust phase or frequency prediction models from noisy data. By combining an inlier optimization strategy with a dynamic threshold mechanism based on median absolute deviation (MAD), it achieves precise detection and repair of anomalies. Validation experiments were conducted using real data from hydrogen masers and cesium atomic clocks, employing datasets containing outliers, phase jumps, and compound anomalies. The proposed method was compared with traditional methods, the robust Kalman filter (RKF), and M-estimation methods. Results demonstrate that the proposed method exhibits superior performance across various anomaly scenarios. In the comparison of robust algorithms, the RANSAC method achieved an F1-score of 0.953 8 in hydrogen clock tests, outperforming M-estimation (0.924 7) and the RKF with optimal parameters (0.817 7). Although its F1-score was slightly lower than that of the RKF with optimal parameters in cesium clock tests, the performance of the RKF degraded significantly under non-ideal conditions with parameter mismatch. Convergence analysis indicates that with appropriate minimum subset sizes and iteration counts, the fitting results achieve significant convergence, with the standard deviation of the fitting slope approaching zero. Furthermore, the processing latency for a single sliding window is in the millisecond range. Under a 1 Hz sampling rate, the computational load is less than 1%, meeting the requirements for real-time integrity monitoring. Experimental results validate the adaptability and robustness of the RANSAC algorithm in the absence of precise prior noise information, providing reliable technical support for autonomous integrity monitoring in precision time-frequency systems.

Abstract:To meet the precision requirements for full-stroke, multi-parameter composite measurement of rolling linear guides in ultra-precision machining, this paper investigates a geometric error compensation method for guide rail contours based on fused line-structured light sensing technology. To address the distortion in measurement data caused by the coupling of feed-axis straightness errors and sensor installation deviations, a unified mathematical model incorporating three-dimensional position error, attitude error, and sensor three-axis installation deviation is established based on multi-body system theory. The feed axis is treated as the "moving body" and the sensor as the "end effector," and the nonlinear coupling mechanism of various errors is clarified. An error identification method using a quadrant detector and a linear laser sensor is designed. Feed-axis position and attitude errors are separated through quadratic polynomial fitting, and sensor installation deviation angles are extracted based on geometric projection relationships. A joint compensation matrix is then constructed to correct and compensate the measurement coordinate system. An experimental platform for the guide rail geometric contour instrument is established. Straightness error identification is verified over an 800 mm stroke, and sensor installation deviation is validated within a 0.5°~2.0° deflection range. Identification errors for both methods are below 3.5%, confirming the accuracy of the proposed method for guide rail contour measurement. Comparative experiments under no compensation, single-error compensation, and joint compensation scenarios demonstrate that, after joint compensation, the average deviation of guide rail profile measurement decreases from 6.229 μm to 2.301 μm, with the mean deviation reduced by 36.9%, the standard deviation by 22.8%, and the maximum deviation by 20.8%. Full-stroke measurement accuracy is significantly improved, verifying the effectiveness of the multi-body system error model and the joint compensation matrix. The proposed method provides a theoretical and technical foundation for high-precision inspection of ultra-precision machining equipment.

Abstract:To ensure the long-term reliable operation and performance optimization of wind turbine blades, real-time deformation monitoring is essential. Therefore, this study develops a LiDAR-based non-contact real-time deformation monitoring method for wind turbine blades using a cascaded structural fusion framework, enabling high-precision tracking of dynamic blade deformation. First a three-dimensional LiDAR is used to collect the point cloud of blade motion, and noise is eliminated through downsampling and statistical denoising preprocessing. Subsequently, a point cloud cascade processing framework of “prediction-correlation-optimization-registration” is established, which utilizes the KF dynamic model to predict the prior positions of feature points, introduces dynamic constraints and observation fusion mechanism, and effectively suppresses the error propagation of point cloud noise; Based on the prediction covariance constraint, KD tree adaptively matches the search domain to improve the accuracy of feature correspondence and enhance the robustness of feature matching under vibration conditions; Integrating point cloud observations to optimize feature coordinates, and inputting the optimized feature set into the ICP algorithm to solve high-precision deformation matrices, forming a closed-loop system that dynamically tracks to global registration, thus achieving a balance between accuracy and efficiency in a cascaded architecture. Experimental results show that the proposed method achieves a deformation measurement accuracy of 0.209 6 mm; 95% of data points have an absolute error of ≤ 0.1 mm, with RMSE = 0.073 2 mm, MAE = 0.057 8 mm,and a correlation coefficient of 0.954 4 (p < 0.01); In terms of system matching performance, the matching success rate reaches 95.4%, significantly better than the single KF algorithm (82.6%) and ICP algorithm (73.1%), verifying the effectiveness of this method in real-time deformation monitoring engineering. Under complex working conditions, it still has excellent measurement accuracy, system stability, and environmental adaptability, which can provide reliable technical support for wind turbine blade health management, early fault diagnosis, and intelligent operation and maintenance decision-making.

Abstract:To address the stringent autofocus requirements of high speed and accuracy for the high-content microscopy imaging systems, this study proposes a peak search strategy based on the active sampling and iterative weighted curve fitting, which aims to achieve the stable and precise focus positioning with the fewest sampling steps. The method begins by establishing a quadratic curve model using four initial sampling points to predict the focus region, thereby constructing a closed-loop mechanism of "prediction-verification-optimization." In each iteration, the system actively selects the sampling position with the highest information gain based on the current fitted model, while dynamically reducing the impact of outlier noise on the fitting results through the weighted least square criterion. To further enhance the reliability of search, an uncertainty assessment mechanism is introduced as the criterion for iteration termination, which quantifies the credibility of model predictions by analyzing the distribution of fitting residuals and the clustering of sampling points. Simultaneously, a management strategy of fixed-scale intelligent point set is employed by consistently maintaining the four most representative sampling points for modeling, which ensures that the model remains focused on the vicinity of optimal region. This approach improves the computational efficiency while enhancing the local characterization capabilities. Experimental results show that the method achieves the stable convergence under different initial position conditions, requiring an average of only 6 to 8 sampling steps to achieve the focusing accuracy within ±4 μm. Compared to the fastest sampling time of 5.75 seconds for traditional methods, the proposed method completes focusing as fast as 2.75 seconds and provides the efficiency improvement of over 50%. Moreover, it maintains the excellent robustness and adaptability in the complex imaging environments with high noises and nonlinear interference. This study provides a reliable technical solution to achieve the rapid and precise autofocus in the high-throughput microscopy imaging.

Abstract:Surface electromyography (sEMG)based lower-limb motion intention recognition has been shown to hold broad application potential in the field of human-machine interface (HMI). However, owing to the inherent inter-subject variability of sEMG signals, significant domain shifts exist in feature distributions across subjects, which severely limits the generalization capability of sEMG recognition systems in cross-subject scenarios. To address this issue, a novel method named cross spatial attention-based dual-stream time-frequency convolutional neural network with gate-controlled feature decoupling (CSACNN-GFD) is proposed in this study. The proposed method adopts a dual-branch time-frequency input structure and employs a multi-scale convolution module integrated with spatial attention to capture the spatial correlation and time-frequency dynamic features of multi-channel sEMG signals, thereby enhancing the capability of motion intention information extraction. Furthermore, a gate-controlled feature decoupling module with a complementary mechanism is designed, together with a decoupling loss function to constrain both feature extraction and gate learning processes. This design enables adaptive feature partitioning in the deep representation space, realizing the disentanglement of motion-related features from subject-related features, and further performing pattern recognition using the cross-subject invariant motion features. In the experiments, sEMG data of five common continuous lower-limb movements were collected from ten subjects, and comparative experiments on motion pattern recognition were conducted against existing generalization strategies under the leave-one-subject-out (LOSO) cross-validation setting, with the results showing that the proposed CSACNN-GFD achieves an average accuracy of 84.29% on unseen subjects. Further validation on a public dataset with eight types of movements yields an average accuracy of 73.83%, improving the average performance by 4.32% and 6.55% respectively compared with the baseline models, and outperforming mainstream strategies including MIXUP, DANN, CORAL, and DIFEX. Meanwhile, the inference time of CSACNN-GFD is only 9.57 ms, demonstrating favorable real-time performance. The proposed method effectively enhances the generalization capability of cross-subject sEMG recognition systems, thus contributing to the universalization development of human-machine interaction technologies.

Abstract:The vibration of the spindle of the worm wheel gear grinding machine has a decisive influence on the quality of gear processing. However, the periodic dressing of the grinding wheel and the continuous tool shifting during grinding will change the amplitude and frequency of the spindle vibration, making spindle vibration prediction difficult. This paper introduces the diameter parameter of the grinding wheel, converts the input grinding wheel linear speed into the grinding wheel rotational speed, and utilizes the dynamic characteristics of the grinding wheel rotational speed to characterize the influence of periodic grinding dressing; at the same time, it introduces the spindle position parameter to establish a compensation function for the grinding wheel spindle position to eliminate the influence of continuous tool shifting during grinding. Based on this, a prediction method for the spindle vibration is proposed, which predicts the spindle vibration of the worm wheel gear grinding machine through grinding process parameters. Firstly, the liquid neural network (LNN) gating mechanism is utilized to dynamically screen the process parameter features, simulate the physical conduction logic between the process parameters and the root mean square (RMS) value of vibration, discretize the process parameters through a continuoustime dynamic system, and use the activation function to capture the hidden dynamic characteristics between them. Secondly, a position compensation function is established based on LNN to capture the hidden characteristics between position information and RMS. Taking the RMS value corresponding to the standard Y-axis position as the benchmark, the RMS values corresponding to other positions are mapped and compensated. Finally, the global dependencies of the features is modeled through multiple stacked Transformer encoder blocks, and the output features of LNN are optimized using residual connections, etc. Finally, the sequence dimension is removed and combined with the compensation value to obtain the vibration prediction value. In the comparative experiments, the R2 of this prediction model reaches 99.19%, the RMSE is 0.074 1, the MAE is 0.051 1, and the MAPE is 0.05%. Compared with the traditional model, the prediction accuracy is higher. Finally, based on this prediction model, a spindle vibration suppression model for the worm wheel gear grinding machine is established. The grinding process parameters are optimized using the quantum slime mold algorithm to suppress the spindle vibration, and the suppression effect is 39.99%.

Abstract:Missing data often occur in multi-channel vibration signals of rotating machinery due to sensor faults, communication interruptions, or environmental disturbances, which can degrade the performance of fault diagnosis models. To address this issue, this paper proposes an improved diffusion-model-based method for missing data imputation. A masked multi-scale conditional diffusion model is developed based on denoising diffusion probabilistic models, where observed data are incorporated as conditional information to guide the stepwise generation of missing values. The joint distribution of multi-channel vibration signals is modeled to effectively capture inter-channel correlations. Regarding network architecture, a U-Net backbone is employed, with multi-scale convolutional residual blocks and linear attention modules stacked in both the encoder and decoder. This design enhances the extraction of temporal dependencies and multi-scale features of vibration signals, improving the accuracy and stability of missing data imputation. Comparative experiments are conducted on bearing and gearbox multi-channel vibration datasets under random point missingness and random block missingness scenarios. The results demonstrate that, across different missing rates, the proposed method outperforms traditional imputation methods and existing deep learning models in terms of root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE). Moreover, the imputed signals preserve temporal and frequency domain characteristics that are closer to the original signals. When applied to fault diagnosis tasks, the classification accuracies reach 95.85% and 94.85% on the bearing and gearbox datasets, respectively, validating the effectiveness of the proposed method in improving multi-channel vibration signal quality and ensuring reliable fault diagnosis performance.

Abstract:Monitoring and evaluating the in-service condition of equipment in real time is essential for maintaining the stable operation of large-scale complex electromechanical systems. Inductive oil debris sensors detect wear particles in lubricating oil via electromagnetic induction, providing a reliable basis for assessing the wear condition of critical mechanical components, and have been widely applied in the maintenance of large-scale mechanical equipment. However, the induced voltage signals generated by debris are usually weak and are difficult to identify accurately by manual feature extraction under the influence of various interferences, which limits the identification accuracy and generalization capability of inductive oil debris sensors. To address this issue, this paper proposes an intelligent recognition method for oil debris signals, referred to as PCatten. Firstly, the morphological stability of debris-induced signals under multiscale filtering is exploited to construct multiscale filtering features, which characterize the key geometric profiles and energy distribution of debris events and provide physically meaningful input representations for subsequent deep learning. Subsequently, a parallel convolution module is designed to perform branch-wise deep convolutional modeling of features at different scales, and an improved fusion attention module is introduced to adaptively recalibrate feature weights along the channel and temporal dimensions, thereby highlighting debris-sensitive components and suppressing complex background interference. Finally, the reconstructed multiscale feature sequence is fed into a Vision Transformer, which captures long-range dependencies and cross-scale correlations through the self-attention mechanism, enabling accurate discrimination of debris-induced voltage signals under strong interference and low signal-to-noise ratios. Experimental results demonstrate that the proposed model achieves excellent performance on both three-coil sensor dataset and high-gradient static magnetic field sensor dataset, the interference elimination rate, debris identification rate, and debris identification accuracy are 99.72%, 98.94%, and 99.44%, respectively, and the proposed method still exhibits superior debris-detection performance compared with traditional algorithms under low signal-to-noise ratios ranging from -5 to 0 dB.

Abstract:To address the performance bottlenecks in real-time capability, accuracy, and noise robustness associated with multi-spot centroid extraction in complex industrial visual measurement scenarios, a fast and high-precision multi-spot centroid extraction method suitable for implementation on field-programmable gate array（FPGA） is proposed. The method integrates template matching, run-length-encoding–based connected-component identification, and a distance-weighted grayscale centroid technique to construct a multi-level collaborative optimization framework. Local grayscale statistics are exploited to achieve adaptive threshold segmentation and edge-noise suppression, thereby enhancing localization stability and computational parallelism. First, coarse spot localization is performed through local grayscale statistics and Gaussian-template cross-correlation, followed by dynamic generation of adaptive thresholds to improve the reliability of spot-region segmentation. Subsequently, a run-length-encoding connected-component structure is designed, which completes region labeling and coarse centroid estimation using only a single-line buffer, effectively reducing on-chip memory consumption. Finally, a distance-weighted grayscale centroid model is developed to improve localization accuracy and robustness under boundary blur and low signal-to-noise ratio conditions. Experimental results demonstrate that, under various spot distortions, noise distributions, and observation distances, the proposed method significantly outperforms traditional grayscale centroid and Gaussian-fitting approaches in terms of localization accuracy and error stability, reducing localization error by approximately 70% and improving robustness metrics by more than 50%. Within a measurement range of 10~30 m, the repeatability of centroid localization is better than 0.02 pixels, and the end-to-end system processing latency is reduced by approximately 89%. With its high accuracy, strong robustness, and low latency, the proposed method is well suited for real-time multi-spot detection in long-range industrial visual measurement applications and provides an effective technical solution for high-performance industrial measurement systems.

Abstract:X-ray digital radiography (DR) has been widely used in the industrial nondestructive testing. However, there will be a large number of workpieces with irregular structures and large thickness variations in the practical applications, which cause the DR detection prone to the underexposure in thick parts and overexposure in thin parts. On one hand, for the detector pixel array of 4 K×4 K, most algorithms can hardly handle these large DR scan images with the customer-level devices. On the other hand it is difficult to obtain a large number of paired labels for the industrial inspection. To address the problem of large-size DR inference and label scarcity, a lightweight unsupervised enhancement framework is proposed by coupling the contrastive language-image pretraining(CLIP) vision-language model with contrast-limited adaptive histogram equalization (CLAHE) priors. The first stage learns prompt vectors to guide a frozen CLIP image encoder with the CLIP enhancement loss, structural consistency loss, and CLAHE feature-map perception loss. The second stage refines the prompts iteratively through the ranking loss and alternately updates the enhancement network until the visual convergence. Experimental results show that peak signal-to-noise ratio(PSNR), learned perceptual image patch similarity(LPIPS), and structural similarity(SSIM) are improved by 1.0 dB, 1.6%, and 2.0%, respectively, outperforming other unsupervised algorithms on multiple metrics. Additionally the inference needs only 0.279 M parameters and processes a 5 732×2 333 image in 1.5 s. Furthermore the model trained with merely 380 casting images generalizes directly to unseen carbon-fiber circuit boards and other materials, demonstrating the strong potential for industrial deployments.

Abstract:Manual inspection of high-voltage cable insulation is inefficient, and traditional algorithms falter at weak boundaries. To address this, we propose a portable measurement system and a two-stage “coarse localization-fine reconstruction” algorithm. First, a novel scale-aware attention U-Net (SA-UNet) performs precise segmentation. It incorporates a multi-scale attention fusion encoder (MSAF), a refined multi-kernel pooling (RMP) module, and a Skip attention feature fusion (SAFF) module to enhance perception of complex details and weak boundaries, effectively mitigating issues from class imbalance. Subsequently, a Pix2Pix generative adversarial network (GAN) reconstructs low-contrast regions, sharpening interlayer boundaries by learning a mapping from blurred to clear images. Finally, insulation thickness is automatically computed using the ray intersection method. Validated on a dataset of 3 300 cross-sectional images, the proposed SA-UNet achieves a 99.36% intersection over union(IoU), outperforming models like U-Net and DeepLabv3+. The GAN reconstruction achieves high structural fidelity (SSIM > 0.98), enabling a final thickness measurement with a mean absolute error (MAE) of only 0.01 mm. This work presents a robust, automated solution for high-precision measurement of high-voltage cable insulation parameters.

Abstract:Conventional visual-inertial SLAM algorithms often suffer from poor feature quality in dim and low-texture environments, leading to unstable tracking, frequent tracking loss, and low localization accuracy. To address this, this paper proposes a stereo visual-inertial SLAM algorithm for UAVs that integrates texture awareness and adaptive feature tracking. First, a noise suppression factor is introduced to prevent noise amplification during image enhancement, and a texture-aware control weight is designed to adaptively adjust the enhancement strength based on local texture richness, applying stronger enhancement in texture-rich regions while keeping the enhancement low in texture-sparse areas, thereby improving overall image contrast while effectively suppressing noise and preserving useful details. Second, the XFeat network is used to replace hand-crafted features, and a density-feedback threshold adjustment mechanism is introduced to address uneven feature distribution by adaptively adjusting the detection threshold according to local feature density, enabling more features to be detected in sparse regions while maintaining a higher threshold in dense regions to retain high-quality features, thereby improving the stability and uniformity of feature tracking. Finally, these improvements are integrated with the VINS back-end optimization to build a complete SLAM system. Evaluations on the EuRoc dataset show our method achieves 29% and 13% higher accuracy than VINS-Fusion and SuperVINS, respectively, in challenging dim and low-texture scenarios. Real-world experiments demonstrate a closure error of 0.486 meters, confirming its superior performance.

Abstract:To address the issues in the mechanical condition monitoring of air circuit breakers using acoustic signals—specifically, the dependence on manual parameter setting and poor interpretability of modal decomposition methods, as well as the limited applicability of short-time analysis techniques—this paper proposes a sound event detection model combining improved symplectic geometric mode decomposition (ISGMD) and a time-frequency attention (TFA) mechanism. The method involves synchronously collecting acoustic signals, main shaft angular displacement, and contact voltage signals during circuit breaker operation to perform time-frequency correlation analysis on closing/opening events. ISGMD is utilized to adaptively decompose the acoustic signals, overcoming interference from invalid components and the limitation of unclear physical meaning. Subsequently, S-transform is applied to construct time-frequency spectrograms, highlighting the time-frequency distribution patterns of the signals and thereby building the dataset required for subsequent model training. Finally, a deep learning network is constructed by embedding the time-frequency attention mechanism into the feature extraction module. This enables the network to dynamically focus on frequency intervals associated with the closing/opening events. Combined with the bidirectional long short-term memory (Bi-LSTM) network to deeply explore long-term dependencies in the sequences before and after sound events, the model achieves accurate localization of the boundaries of closing/opening events, effectively reducing the probabilities of false alarms and missed detections. The results indicate that the proposed method achieves an accuracy, recall, and F1-score of approximately 93%. For data from different microphone positions and distances, the root mean square error (RMSE) is less than 0.44 ms; for different devices, the RMSE is below 0.57 ms, demonstrating good generalization capability and stability. ISGMD provides interpretable signal decomposition from the perspective of physical mechanisms, while deep learning drives the automatic learning of complex event features from the data level. The synergistic approach formed by these two approaches achieves millisecond-level localization of sound events, providing reliable support for the intelligent diagnosis of the mechanical condition of circuit breakers.

Abstract:Complex high-rise steel structures can be partitioned into layered substructures, and sensors installed at each interstory level can be used to monitor damage in the corresponding substructure layer. To address the problem of damage detection in complex high-rise steel structures, a damage identification method based on the impact information transformation index derived from a vector error correction model (VECM) is proposed. First, the fundamental theory of the VECM is introduced, including the basic procedures for model order determination, cointegration testing, and parameter estimation. Subsequently, the damage-related characteristic parameters of the VECM are analyzed. By examining the impact matrix, the damage characteristic impact vector is extracted. Using the characteristic impact parameters and statistical parameters under both the undamaged baseline state and the damaged state of the structure, an impact information distance is constructed. Furthermore, the degree-of-freedom information at the measurement points is transformed into interstory stiffness information to establish the impact information transformation index. On this basis, a complete damage detection methodology and identification procedure for high-rise steel structures is developed, including data acquisition, VECM modeling, extraction of the damage characteristic impact vector, and damage identification based on the impact information transformation index. Finally, experimental studies are conducted on a classical three-story frame structure and a high-rise relay tower model. Non-stationary time-domain excitations are applied to the structures using an excitation device to obtain time-history response data. Damage detection is then performed by comparing structural responses before and after damage. The results show that the traditional cepstral distance–based detection index exhibits certain effectiveness for simple frame structures but performs inadequately for complex high-rise steel structures. In contrast, the proposed VECM-based impact information transformation index demonstrates superior detection performance and can more effectively identify damage locations in high-rise steel structures.

Abstract:Existing electricity theft detection methods primarily focus on high-loss feeders, where anomalies in the daily average line-loss rate are pronounced. However, potential short-term and intermittent electricity theft on non-high-loss feeders is easily masked by load fluctuations, metering errors, and changes in operating conditions, making such theft difficult to detect effectively. To address this issue, this paper proposes a method for electricity theft detection on non-high-loss feeders based on a mutual-information-difference dynamic graph sequence and adjacency features. First, at an hourly resolution, a sliding window is employed to calculate the mutual information between users′ electricity consumption and line-loss electricity, thereby constructing a mutual-information-difference dynamic graph sequence. A window-wise adaptive threshold is introduced to capture the nonlinear dependency between users′ electricity usage and line loss as it evolves over time, avoiding graph-structure instability caused by a fixed threshold. Second, a feature extraction method combining a graph autoencoder and a dynamic weight balancing mechanism is developed. Based on a cross-window adjacency reconstruction task, it captures the temporal evolution of connection patterns within the population network and extracts adjacency features that reflect the robustness of users′ dependency relationships with the group. Meanwhile, a dynamically weighted loss function is applied to mitigate training bias caused by the imbalance between positive and negative samples and to prevent degeneration in graph reconstruction learning. Finally, principal component analysis is used to reduce and select the embedded features, and K-Means++ clustering is applied for unsupervised anomaly identification. Simulation studies and experiments on real-world data demonstrate that the proposed method can effectively detect electricity-theft users in non-high-loss scenarios, verifying its feasibility and effectiveness. From the perspective of group relationship evolution, this method characterizes the dynamic dependencies between users′ electricity consumption and line loss, transforms electricity theft detection into a graph anomaly detection problem, and provides an effective and practically applicable technical solution for electricity theft detection on medium-voltage feeders.

Abstract:Although the ultrasonic testing technology provides the high-precision advantages in pipeline integrity assessment, its dependence on the liquid coupling severely limits the application in the natural gas pipeline environments. To address the challenge of ultrasonic energy transmission difficulties caused by the severe acoustic impedance mismatch at the gas-solid interface of high-pressure natural gas pipelines, this study proposes a high-pressure gas-coupled ultrasonic testing method based on the acoustic resonance principle. By integrating the Redlich-Kwong equation of state with ultrasonic wave propagation theory in multilayer media, a theoretical model for the ultrasonic resonant reflection under the high-pressure conditions was established. The model incorporates a complex wavenumber to characterize the acoustic attenuation and systematically reveal the influence mechanism of pressure on gas acoustic properties as well as resonant response, and the analytical expression for the resonant frequency shift caused by medium damping is also derived. Then the coupling effects of pressure, propagation distance, wall thickness, and resonant order on the resonant frequency were quantified with the simulation analysis. Furthermore a high-pressure experimental platform was established, and the experimental validation was performed using nitrogen as an equivalent medium. Simulation and experimental results demonstrate that the gas acoustic impedance significantly improves when the pressure increases to 1 MPa. Specifically the amplitude of the received resonant signal is enhanced by approximately 69 dB while consistently exciting multi-order resonant modes when the pressure reaches 3 MPa. Under such conditions, the thickness measurement error of steel plates ranging from 6.2 to 8.2 mm remains below 0.15 mm, whose resonant signal fluctuation amplitudes are smaller than 1 V within the 40~100 mm propagation distance range, demonstrating the good adaptability to variations in lift-off distance. The study verifies the accuracy and reliability of acoustic resonance for the wall-thickness measurement of high-pressure natural gas pipelines, thereby overcoming the precision limitations of conventional pulse-echo method in the high-pressure gas environments and providing a new theoretical foundation as well as experimental support for the nondestructive testing of transmission pipelines.