Abstract:Scintillators are widely used in many fields such as nuclear medicine imaging, industrial non-destructive testing, and high-energy physical radioactivity measurement, which greatly promotes scientific and technological progress and innovation in the fields of basic science, medical science, and industrial technology. With the continuous improvement of application requirements, the performance requirements of scintillators are getting higher and higher, especially higher light output yield. Combining scintillators with micro-nano photonics technology research, by preparing micro-nano structures on the surface of scintillators and using their regulation of electromagnetic waves to change the critical angle of photon emission, the technical problem of low light output yield of scintillators due to total internal reflection effect can be effectively solved. In order to achieve the same radiation effect at a smaller dose. This paper describes the research progress of scintillator micro-nano structure preparation technology in recent years, comprehensively reviews the mechanism of micro-nano structure regulating scintillator light output enhancement, summarizes the existing micro-nano structure preparation technology methods, analyzes the influence of different types of micro-nano structures on scintillator light output yield, and summarizes the preparation technology according to the size of scintillator micro-nano structure, and discusses the research and application prospects of various types of scintillator micro-nano structure preparation technology.

Abstract:Aiming at the problems of small targets, scattered distribution, and susceptibility to background and noise interference in the process of transmission line insulator defect detection, an improved YOLOv8based defect detection method is proposed. Firstly, LSKNet is introduced to replace the original path aggregation network, enabling the model to adaptively select and adjust convolution kernel sizes based on the characteristics of different targets. This allows for more precise matching of target features and background information at various scales, significantly enhancing the robustness of defect recognition in complex scenarios. Furthermore, the SPPF-LSKA module is integrated into the network. By fusing global context information, this module greatly improves the aggregation efficiency and discriminative capability of multi-scale features, providing more refined feature representation for defect detection. Additionally, the proposed method incorporates a spatial attention mechanism into the neck network of YOLOv8, enhancing the model’s global feature comprehension and strengthening its focus on key information, particularly for small targets. To address the practical requirements of model efficiency and deployment, part of the conventional convolution layers in the neck network are replaced with GhostConv, effectively reducing the model’s parameter count and computational overhead. This achieves a balance between detection performance and resource efficiency. Experimental results demonstrate that the proposed method achieves a mAP of 93.1%, representing a 4.4% improvement compared to the original model, effectively enabling accurate detection of small targets.

Abstract:Conformal coatings are essential for ensuring the reliable operation of printed circuit boards (PCBs) in complex environments, effectively preventing external corrosion and component aging. However, current conformal coating detection techniques remain inadequate, making it difficult to achieve comprehensive coverage assessment, which affects the coating’s protective capability and long-term stability. This paper proposes a PCB conformal coating coverage detection method based on electromagnetic scanning and fluctuation sequence difference measurement. First, the electromagnetic shielding mechanism of the conformal coating is analyzed, providing a theoretical foundation for system detection. Next, an improved adaptive fluctuation sequence difference measurement method is proposed. This method acquires electromagnetic radiation data of the PCB under normal and weakened coating conditions through near-field scanning, applies wavelet decomposition for noise reduction, and uses an improved difference measurement algorithm to calculate the distance between two fluctuation sequences to identify coating defect areas. Finally, the effectiveness of this method is verified through simulations and experimental measurements. The results demonstrate that this method significantly enhances the accuracy and reliability of coating detection, overcoming limitations in traditional coating assessment methods and offering new perspectives for the theoretical research and engineering applications of electromagnetic shielding technology.

Abstract:Addressing the low accuracy problem in detecting multi-scale, multi-type steel surface defects within complex backgrounds, this paper designs an improved YOLOv5 algorithm that integrates HGnetv2 with an attention mechanism. First, the HGnetv2 network incorporates an attention mechanism as a backbone layer to enhance feature extraction capabilities for small target defects. Second, in the feature fusion layer, attention mechanisms and involution operations are combined to achieve effective aggregation of edge features in shallow layers and semantic information in deep layers. Besides, CBME_C2f replaces the C3_Bottleneck module to improve gradient flow. Additionally, a new bounding box loss function, VCIoU, is used to calculate positional features between the vertices and center points of the prediction and target boxes, enhancing bounding box regression precision. Finally, MetaAconC is introduced to adaptively adjust the non-linearity of activation for each feature map channel, improving the ability to extract feature information from complex backgrounds. Experimental results on the NEU-DET dataset show that the proposed method achieves an mAP50 of 81.4% and an mAP@50:95 of 44.1%, which is 5.4% and 2.8% better than YOLOv5s respectively. For the small defects such as crazing in this dataset, the detection accuracy reaches 55.4%, representing an 18.1% improvement over YOLOv5s, while maintaining a detection speed of 80.6 fps. Compared to other mainstream defect detection algorithms, this algorithm improves accuracy while meeting the real-time demands of steel surface defect detection.

Abstract:In order to solve the problems that the size of different types of defects in magnetic tile images affects the segmentation accuracy, the detection model parameters are large and difficult to deploy in practical applications, and the image pixel distribution is uneven, a magnetic tile surface defect segmentation algorithm based on improved Deeplabv3+ was proposed. Firstly, in terms of structure, the lightweight MobilNetv3 network was used as the backbone network to replace the Xception network of the original model, so that the parameters and computational cost of the model were kept small to improve the detection speed. Secondly, the ECANet attention mechanism was introduced to improve the feature expression ability and generalization ability of the model. Finally, the loss function combining dice loss and focal loss were used to effectively alleviate the influence of sample pixel distribution imbalance on model training. The experimental results show that the average intersection union ratio of the proposed algorithm on the magnetic tile surface defects dataset was 68.25%, the average pixel accuracy was 82.80%, and the accuracy was 79.80%, compared with the original Deeplabv3+ algorithm, the average intersection union ratio was increased by 8.62%, and the average pixel accuracy was increased by 9.96%, the accuracy of the algorithm was increased by 11.52%, which verifies the effectiveness and feasibility of the proposed algorithm, and has certain application value in industrial applications.

Abstract:Timely detection of insulator self-explosion defects is of great significance to the safe and reliable operation of transmission lines. Aiming at the problems such as insufficient detection ability of insulator self-explosion defect with small target characteristics and complex model structure of deep learning model, proposes a lightweight improved YOLOv8n insulator self-explosion detection method for transmission lines. Based on the YOLOv8n network model, a small target detection module is added to capture the details of the insulator self-exploding small target and improve its detection capability. Furthermore, SIoU loss function is introduced to solve the problem that the original CIoU loss function does not consider the direction between the real box and the predicted box, and the target positioning accuracy is enhanced. Finally, channel pruning method is used to prune the improved model, remove the redundant parameters of the model, reduce the floating point operations, and reduce the calculation cost and complexity of the model. The experimental results on the constructed insulator self-explosion data set show that the average accuracy of the lightweight improved method reaches 97.1%, and its floating point operations and volume are 4.9 G FLOPS and 1.82 MB respectively, which is only 60.5% and 29.7% of the original model, which reasonably balances the accuracy of insulator self-detonation detection and the complexity of the model. In another transmission line inspection data set, the proposed method also has good detection accuracy for other types of small targets, and has a good prospect of popularization and application.

Abstract:Time-frequency-spatial features are widely used in motor imagery EEG classification, but effectively utilizing these features to improve classification accuracy remains challenging. Traditional methods often eliminate redundant information through feature selection but tend to overlook the intergroup dependency of time-frequency-spatial features. To address this issue, we propose an EEG classification model based on a feature recalibration network with weight fusion (FRNWF). First, we extract the time-frequency-spatial features to reveal their grouping structure, treating each group of these features as a whole and considering it as a feature map. Two branches are then established to obtain the channel weights of these feature maps: one branch derives the channel weights of global information through global average pooling, while the other derives the channel weights of local information through global maximum pooling. Next, we design a weight fusion operation to merge the two sets of channel weights and rescale the feature maps, thereby achieving intergroup dependency modeling of the time-frequency-spatial features. Finally, two fully connected layers are used for classification. Experimental validation on four publicly available motor imagery EEG datasets shows that the proposed method achieves an average classification accuracy of 80.72%. This outperforms 18 feature selection methods, existing feature recalibration network methods, and most of the recent literature. The experimental results indicate that the proposed method demonstrates significant potential in practical applications and is likely to be widely adopted in future brain computer interface research and motor rehabilitation training.

Abstract:Online measurement of blade tip clearance provides important data support for the performance evaluation and fault diagnosis of aero-engines. Most of the current clearance measurement technologies for aero-engines focus on measuring the clearance of individual blades, and no airborne application cases have been reported. This paper addresses the challenges associated with the capacitive blade tip clearance measurement method, such as high computational complexity, data redundancy, high sampling requirements, and difficulties in solving for the average clearance. A novel method for measuring average blade tip clearance based on the root mean square (RMS) of capacitive sensing signals is proposed. A blade tip clearance signal model based on RMS was developed, and the relationship between the RMS value of the clearance signal and the average blade tip clearance was derived. Simulations were performed to verify the impact of noise and time parameters on the RMS value of the clearance signal, and signal processing parameters for blade tip clearance based on the RMS method were introduced. A scaled blade model was constructed based on the constant duty cycle principle of the clearance signal, enabling dynamic calibration suitable for the RMS processing method. In-flight tests were conducted on the compressor of a specific aero-engine to validate the feasibility and effectiveness of the proposed method. Experimental results show that this method achieves average blade tip clearance measurement at a sampling rate of 10 kHz, with measurement errors less than 29 μm compared to traditional methods.

Abstract:To address the problem of inefficient resource allocation during task offloading in mobile edge computing (MEC) environments for terminal devices, a multi-task partial offloading scheme is proposed that leverages device-to-device (D2D) communication technology to assist the MEC system. The scheme is based on the block coordinate descent (BCD) method to optimize the task offloading and resource allocation strategies jointly. Additionally, a dynamic pricing strategy is adopted to incentivize the service-oriented smart devices (SSDs) to share the remaining available computational resources, aiming to minimize the system-wide execution cost. Firstly, the reconfiguration linearization technique (RLT) and convex optimization theory are utilized to optimize the allocation of computational resources and the offloading ratios, determining the tasks to be allocated to local computing, D2D offloading and edge offloading. Secondly, the appropriate SSDs are selected to perform D2D offloading tasks based on the optimal offloading policy. Simulation results show that, compared with the traditional partial offloading scheme, relay-assisted offloading scheme, and cooperative computing offloading scheme, the proposed offloading scheme reduces the total cost of system execution by 27.62%, 25.58% and 19.98%, respectively, under different numbers of devices, and reduces the total cost of system execution by about 43.35%, 38.19% and 36.79%, respectively, under different conditions of the maximum tolerable delay. The average reduction in total system execution cost under different task data sizes is about 36.47%, 30.60%, and 29.15%, respectively. Further experiments indicate that, compared with the greedy offloading scheme, the proposed offloading scheme optimizes the system execution cost by an average of 7.59%, 0.39% and 3.10% for different numbers of devices, maximum tolerable delays and task data sizes, respectively. Therefore, this scheme effectively enhances the resource utilization while reducing the execution cost.

Abstract:In order to solve the problem that the imaging performance of the instrument was continuously improved but the measurement method was lagging behind in the development process of X-ray three-dimensional size measuring machine, the uncertainty was studied, and an evaluation method based on voxel model was proposed. Firstly, starting from the principle of X-ray detection, the influencing factors of digital space such as original projection, three-dimensional reconstruction, surface measurement and point cloud fitting in the size calculation process of X-ray three-dimensional size measuring machine were analyzed, and the uncertainty model of voxel was established. Secondly, based on the traceability and measurement results of the center distance of the selfdeveloped ball bar, the uncertainty of each parameter of the voxel model was calculated by GUM method and Monte Carlo method, and the uncertainty of the voxel model was analyzed. Thirdly, combined with the traditional uncertainty parameters, the uncertainty of the measurement results of the spherical center distance was synthesized and expanded. Finally, the reliability and consistency of the measurement and evaluation results are verified by the application of the results of the extended uncertainty. The length measurement results of the short and long ball bar standard were (32.947 7 ± 0.002 0) mm and (52.406 5 ± 0.002 2) mm, respectively. The results showed that the evaluation method obtained by the exploration refines the error source of the measurement results of the X-ray three-dimensional size measuring machine, expanded the research direction of the size measurement method of the X-ray three-dimensional size measuring machine, and provided a measurement basis for the improvement of the data accuracy of the X-ray three-dimensional size measuring machine, which had certain scientific value.

Abstract:To address the common issue in traditional multi-load MCR-WPT systems, where misalignment or improper placement of charging equipment often leads to a decrease in charging efficiency, a parameter optimization method for MCR-WPT systems based on an adaptive genetic algorithm is proposed. Firstly, the impact of load radial offset, load quantity, transmission distance, and load resistance on system transmission efficiency in multi-load MCR-WPT systems is analyzed, and the results are verified using a combined simulation with Maxwell and Simplorer. Subsequently, an adaptive genetic algorithm is employed to find the optimal solution for key parameters affecting system efficiency, such as load radial offset, load resistance, and transmission distance, thereby achieving the best transmission efficiency for the system. The simulation results indicate that the overall transmission efficiency of the optimized system reaches 83.2%, a 7.7% improvement compared to the efficiency before optimization. Finally, an experimental platform for the multi-load MCR-WPT system is established for validation, and the experimental results show that the overall transmission efficiency of the optimized system reaches 81.6%, an 8.2% enhancement over the pre-optimization figures. Both simulation and experimental results confirm the effectiveness of the proposed adaptive genetic algorithm-based parameter optimization method for multi-load MCR-WPT systems.

Abstract:In autonomous driving technology, multi-task environment perception algorithm is one of the key technologies to ensure the safe operation of driverless vehicles in complex traffic environments. In view of the poor robustness of the existing environment perception algorithms in dealing with complex driving scenarios caused by weather, illumination, occlusion and other factors, and the problems of missed detection, and blurred segmentation boundary, an improved high-performance hybrid network IPHNet based on HybridNets network is proposed to more accurately complete real-time perception tasks. This network uses a decoder-encoder structure and adopts an improved EfficientNetV2-S as the backbone network to enhance feature extraction capability and processing speed. By reconstructing BiFPN to increase the feature fusion of intermediate information of different levels, the lightweight up-sampling module DySample is introduced to reduce the complexity of the model and retain more information. The innovative design of the segmentation module (SPN) greatly ensures the integrity and accuracy of the underlying information extraction. Experiments on BDD100K dataset show that compared with the baseline network HybridNets, IPHNet achieves 81.4% mAP on vehicle detection tasks, which is improved by 4.1%. The accuracy of the lane line segmentation task reaches 86.84%, which is increased by 1.44%, and the IoU reaches 33.32%, which is increases by 1.72%. The mIoU of the drivable area division task is increased by 1.8%. The FPS reaches 34, which verified that IPHNet has a certain real-time processing capability.

Abstract:Double fed wind turbines (DFIG) have become mainstream models due to their excellent performance and economic benefits. However, the series capacitor compensation technology adopted to improve transmission efficiency can lead to sub-synchronous oscillations (SSO) due to resonance with grid parameters, threatening the stability of the power system and the safe operation of wind turbine units. In order to solve this problem, an in-depth analysis of the SSO characteristics of doubly fed wind farms is conducted, and the typical wind farm at China Hebei Guyuan is equivalent to two far and near wind farms, an equivalent impedance model considering the DFIG converter control link and PLL accuracy is established; Then, the impedance analysis method in the frequency domain is used to analyze the effects of wind speed, line series compensation, wind farm capacity, and PI control parameters of DFIG converter on system stability, it is found that the SSO stability of the wind farm is inversely proportional to the RSC current inner loop PI ratio coefficient and series compensation, and directly proportional to wind speed and the ratio of far and near wind farm capacity; Finally, the time-domain simulation analysis method is used on MATLAB/Simulink software to further verify the correctness and accuracy of the impedance analysis results. The research conclusion of this paper is of great significance for understanding the dynamic characteristics of wind power generation systems and designing effective SSO suppression strategies. Based on the analysis of SSO characteristics, the proposed linear active disturbance rejection control (LADRC) for SSO suppression strategy can effectively suppress SSO.

Abstract:In response to the issues such as the overly large scale of input parameters in the traditional LSTM data-driven model, which leads to an excessive computational burden, improper selection of hyperparameters, high frequency of turbine system failures, and high operation and maintenance costs, a turbine data-driven modeling approach based on PCA-POA-LSTM is proposed, and the turbine fault early warning is achieved by combining with the sliding window method. Firstly, the PCA dimensionality reduction technique is applied to reduce the dimension of the input data. Secondly, the POA parameter optimization method is adopted to select the optimal combination of hyperparameters. Then, the LSTM algorithm is utilized to predict the output parameters of the turbine. Finally, based on the prediction results of the PCA-POA-LSTM turbine data-driven model, the turbine faults are warned by combining with the sliding window method, and the alarm threshold is defined by the standard deviation within the window, thus conquering the difficulty of turbine fault early warning. The results indicate that the turbine data-driven modeling based on PCA-POA-LSTM achieves a relatively high accuracy, with the average absolute percentage error all below 0.396, the average absolute error all below 0.809, and the average root mean square error all below 1.387. Moreover, the fault early warning method can issue a fault early warning signal at least 173 monitoring points in advance, achieving the purpose of turbine fault early warning and providing theoretical basis and technical support for the future development of turbine health management.

Abstract:The accurate and rapid calculation of aircraft pose using three-dimensional point clouds scanned by LiDAR is the key to achieving automatic parking guidance for aircraft. Therefore, a fast target pose estimation algorithm based on precise registration of skeleton point clouds is proposed. In the point cloud on the aircraft surface, the main body structures such as wings, engines, and nose are selected from the perspective of LiDAR to construct a simplified aircraft skeleton point cloud, avoiding erroneous registration of other complex structures and effectively reducing computational complexity. When parking the aircraft, establish a point cloud bounding box based on the aircraft axis to obtain the initial pose of the aircraft and use it as a constraint for registration. Then, a random sampling consistent coarse registration algorithm based on fast point feature description is used to correct the aircraft pose, and a point surface fine registration algorithm based on bidirectional KD-Tree is designed to improve the accuracy of aircraft pose estimation. Finally, the performance of the algorithm was validated through simulation experiments on aircraft pose estimation throughout the entire parking process. Compared with typical algorithms such as Super-4PCS, MSKM-NDT, and AA-ICP, this paper’s algorithm reduces registration error by 32.5% and improves processing speed by 34%. The maximum angle error for pose estimation is 2.0 degrees, the maximum distance error is 0.125 meters, and the single frame processing speed is 0.37 seconds. The actual aircraft pose estimation experiment also verified the effectiveness of the algorithm.

Abstract:Aiming to address the problems of limited sample size, uneven sample distribution, and cross-operating condition fault diagnosis for gas turbine rotor systems, a fault diagnosis method based on the linkage of numerical model and physical model is proposed. A classical central mass method is used to establish the dynamic model of the rotor system, and fault dynamics differential equations are established in the model by introducing misalignment fault and unbalance fault. Finally, the differential equations of rotor system fault dynamics are solved by the Runge-Kutta method (ode-4/5), and the simulated signal of fault displacement is obtained, which is prepared for subsequent data augmentation and model linkage methods. A conditional deep convolutional generative adversarial network (GAN) model is established by combining the orthogonal gradient penalty algorithm, and the model is used to generate signals by inputting the simulated signals obtained from the physical model as the generator input, and inputting the real experimental signals as the discriminator input to obtain a generated signal that integrates the intrinsic characteristics of the physical model and the actual mechanical characteristics. Secondly, a cross-operating condition domain adaptation fault diagnosis model is established based on the theory of transfer learning, and the data is converted into a two-dimensional temporal-frequency image sample using a combined short-time fractional Fourier transform and inverse convolution algorithm, which provides more obvious resolution and features. The data that integrates the intrinsic characteristics of the physical model and the mechanical characteristics is used as the source domain and the other target domain data to be measured at other operating conditions is used to train the fault diagnosis model. The experimental verification shows that the diagnostic accuracy of the five different fault categories at different speeds and fault levels in different operating conditions is all above 91%. The results are explained and analyzed through confusion matrix, and the method can effectively improve the model’s generalization ability and realize cross-operating condition fault diagnosis of rotor systems, which is superior to other mainstream diagnosis methods under the same conditions.

Abstract:To address the problem that the mutual coupling effect between tags has different effects on tags at different positions, which makes the overall performance of the linear array distributed RFID system show nonlinear changes with the change of tag spacing and number. The method involves that based on the working principle of RFID and the theory of electromagnetic wave propagation, the expressions of the mutual impedance and power transmission coefficient of tags in the multitag RFID system are derived. Using power transmission coefficient, the influence of mutual coupling effect the tags at different positions and the overall performance of the system in the case of linear equidistant distribution of tags is analyzed. A mathematical model of the system's minimum power transmission coefficient changing with the tag spacing and the number of tags is established. In an open indoor environment, the response signal power is tested when the tag spacing and number change. Theoretical analysis and experimental results show that when the tags are linearly equidistantly distributed, the power transmission coefficient of the tags is nonlinearly related to the spacing. The influence of the mutual coupling effect on the tags no longer decreases monotonically with the increase of distance, but gradually decreases with the wavelength as the period, and reaches the maximum value when the spacing is a multiple of the wavelength. The location of the system's minimum power transmission coefficient fluctuates between the center and the outside with a wavelength as the spacing increases. The test results of system performance are consistent with the change pattern of the established model.

Abstract:To improve the measurement accuracy of the downhole accelerometer, a method for online compensation of accelerometer errors based on a magnetic-inertial coati optimization algorithm is designed. Firstly, an error compensation model is established based on the sources of error; the constraint conditions of the gravity angle and the magnetic-gravity angle are established using a gyroscope and a magnetometer; the difference between the true value of the acceleration and the modulus of the theoretical value is set as the objective function. Secondly, based on the coati optimization algorithm, the initial search boundary for error parameters is determined according to the recursive gravity acceleration, and the boundary is narrowed based on the relative distance among the current error parameters, the optimal error parameters, and the boundary values; a boundary point selection is designed to screen the initial error parameters, enabling the algorithm to initially search in the direction of high-quality solutions while retaining some inferior solutions to increase the diversity of error parameters; in the global exploration stage of the algorithm, parameters are designed to adjust the search range of accelerometer error parameters based on the error between the current error parameters and the average error parameters. Finally, the ratio of the modulus of gravity is set as the threshold for deep development, and a Gaussian mutation vector is constructed to enable the accelerometer error parameters to break out of local optima. Experimental results show that after MICOA compensation, the accelerometer error decreases, and the range of inclination angle decreases by approximately 62.5%; at different drilling angles, the root mean square error and standard deviation of the inclination angle can be maintained below 1°.

Abstract:Addressing the difficulty in accurately determining the arrival time of echo signals in time-difference ultrasonic gas flowmeters due to inevitable external influences such as circuit noise, acoustic noise, and environmental noise, which distort the initial segment of echo signals and result in a low signal-to-noise ratio, a precise positioning method for echo signal arrival time based on K-SVD dictionary learning for noise reduction is proposed to enhance the detection accuracy of ultrasonic flowmeters. The method involves using the OMP algorithm to perform sparse representation on the matrix constructed from multiple sets of extracted echo signal data, followed by iterative updating of the dictionary using SVD. The two parameters of dictionary size and sparsity are optimized through the control variable method to train an optimal dictionary learning model capable of adaptively extracting echo signal features, forming a complete dictionary, and reconstructing the echo signal to eliminate waveform distortion caused by noise interference in the initial signal segment. Calibration experiments conducted on a critical flow Venturi nozzle-based gas flow standard device show that directly using a threshold method to locate the arrival time of unprocessed echo signals results in significant errors in the final flow calculation. However, applying K-SVD dictionary learning-based noise reduction to process echo signals can effectively improve the accuracy of locating the arrival time, leading to more accurate flow information. Specifically, the indication error in the high flow range is reduced from 1.32% to 1.02%, and the repeatability is improved from 0.154% to 0.054%. In the low flow range, the indication error is decreased from 3.69% to 1.46%, and the repeatability is enhanced from 1.152% to 0.126%. The optimized measurement results meet the national accuracy standard of Grade 1.0.

Abstract:The series arc fault is one of the main causes of the electrical fire. Aiming at the problem that the series arc fault is difficult to detect accurately under unknown working conditions, a series arc fault detection method based on real-time training and updating kernel based extreme learning machine prediction model was proposed. First, the series arc fault experiments under different power supply harmonics, the carrier frequency and operating frequency of the frequency converter and current level conditions were carried out by using the three-phase motor with frequency converter load circuit. Second, the current signal was denoised by using singular value decomposition filtering and improved single exponential smoothing filtering in turn. Third, the kernel based extreme learning machine prediction model was trained and updated by using the first two cycle current signals, and the predicted residual of the next cycle current signal was calculated. Then, a matrix was constructed by using the absolute value of the predicted residual, the residual matrix was reduced to one-dimensional vector by combining the non-negative matrix factorization, and the maximum value of the one-dimensional vector was used as the fault feature. The series arc fault was detected by using a fixed threshold. Finally, the series arc fault detection and anti-noise performance of the proposed method were tested under unknown working conditions. The results indicated that the proposed method can effectively detect the series arc fault under four kinds of unknown working conditions, which are unknown power supply harmonics, carrier frequency and operating frequency of the frequency converter, and current level, respectively. It showed that method has a strong anti-noise ability.

Abstract:To address the problem of how liquid level height variations within containers impact the performance of passive ultra-high frequency (UHF) RFID tags, a link budget model for passive UHF RFID systems was derived based on the electromagnetic wave propagation theory of RFID. Using the power transmission coefficient, the study analyzes how liquid-induced impedance mismatch within containers affects system performance. To validate the theoretical model, a method combining simulations and indoor experiments was employed. Through simulation and analysis, segmented models were established to describe the variation in tag response signal strength (RSSI) with changes in liquid level height for both vertical and horizontal tag orientations. In an open indoor environment, the RSSI of two tags, Alien9662 and Alien9640, was measured as the liquid level varied, covering a range from 0 mm to 140 mm to observe signal strength changes under different liquid levels. Theoretical analysis and experimental results indicate that when the liquid level rises along the bent arm of the antenna, RSSI gradually decreases with increasing liquid level height, whereas when the liquid level rises along the electric small loop, RSSI exhibits a nonlinear trend, initially increasing and then decreasing. The RSSI change patterns for both tags align with the segmented model, verifying its accuracy. These findings provide a theoretical basis for understanding the effects of liquid environments on RFID system performance and offer practical insights for tag deployment and design in real-world applications.

Abstract:In cement production, clinker is the key component, and its quality directly affects the overall performance of cement. The content of free calcium oxide (fCaO) in cement clinker is one of the important parameters to evaluate the quality of cement clinker. In order to make up for the lack of timeliness of traditional laboratory methods, this paper constructs an efficient and accurate soft sensor model for cement clink f-CaO content, which combines channel attention mechanism and long short-term memory network. The feature extraction method combined with attention mechanism was used to extract the features of the data set. Then, the features were input into the LSTM network for learning, so that the model could focus on the core feature channels in an adaptive manner. Due to the poor prediction effect of LSTM on data with large volatility, the above soft sensor model is improved. Based on the original model, the classification module and weighting module are introduced to modify the prediction results of the LSTM network, so that the model can be more flexible to adapt to the differences between different categories. The accuracy of the model prediction is improved. The experimental results show that the coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE) of the improved SE-LSTM prediction model for cement clink f-CaO are significantly improved compared with the ordinary LSTM and SE-LSTM prediction models. Therefore, the improved model proposed in the prediction of cement clinker f-CaO content improves the prediction accuracy and achieves the expected prediction effect.

Abstract:Detecting composite insulator hydrophobicity class is critical in power system inspections. This study proposes an intelligent recognition method for the hydrophobicity class of composite insulators based on MSG-SSD to address the challenges of low detection efficiency, poor real-time performance, and complex model structures in existing methods. Firstly, the detection model is based on the SSD algorithm, employing the lightweight MobileNetV2 as the backbone network to simplify the network and significantly enhance detection speed. Secondly, to improve the extraction capability of watermark features, a high-resolution feature fusion module, Sim-HRFPN, is constructed, which retains high-resolution features during the fusion process to compensate for the accuracy loss caused by the lightweight design. Finally, to further enhance the computational efficiency of the model, traditional convolution is replaced with GhostConv in the additional prediction feature layers, thereby significantly reducing the computational burden while maintaining the high performance of the model. The results indicate that, compared to SSD, MSG-SSD achieves a 48.17% improvement in detection speed and a 4.89% improvement in accuracy, while reducing computational cost and parameter count by 97.63% and 82.99%, respectively. From this, it can be concluded that the improved model accurately identifies and rapidly locates the hydrophobicity class of composite insulators and meets the lightweight deployment requirements of edge inspection devices. This provides an effective method for the intelligent detection of the operational status of composite insulators in power systems.

Abstract:To reduce the temperature measurement error of distributed single-mode fiber temperature sensing systems, the paper proposes a temperature measurement method based on Golay-Simplex hybrid coding. First, four G codes are transformed into S codes to achieve 12-channel G-S hybrid coding modulation. Then, the output signals of the 12 channels are processed through S code decoding and G code decoding sequentially, employing cumulative averaging and wavelet transformation for temperature curve denoising. This verifies that the coding gain of the G-S hybrid coding is the product of the coding gains of the G and S codes. Comparative experimental results show that under conditions of 30 km fiber length, 50 ns pulse width, and 64 bit coding length, the amplitude fluctuation range of the anti-Stokes signal curve in the G-S hybrid coding temperature measurement system is smaller and has a higher signal-to-noise ratio across the fiber length compared to the Golay code-based temperature measurement system and the single-pulse temperature measurement system. The steady-state temperature measurement error of the G-S hybrid coding can be optimized from ±7.3℃ in the single-pulse system to ±2.5 ℃, outperforming the measurement error of ±3.9 ℃ in the distributed Raman fiber temperature measurement system based on Golay codes. Additionally, the spatial resolution can be maintained at 5 m, demonstrating the effectiveness of G-S hybrid coding for long-distance single-mode fiber temperature measurement, potentially providing effective technical solutions for the integrated perception of infrastructure conditions such as temperature changes due to leakage in hydraulic dams.

Abstract:Embedded devices in substations frequently encounter challenges related to real-time performance and detection accuracy, especially in scenarios involving small targets and densely arranged pointer instruments. This paper proposes an enhanced substation pointer instrument recognition model based on YOLOv5s-BCGS, which improves detection accuracy and efficiency. The model employs YOLOv5s as its backbone network, incorporating a coordinate attention mechanism at the neck to enhance spatial feature extraction. Additionally, the original path aggregation network is replaced with a weighted bidirectional feature pyramid network to better integrate positional and detailed information from the feature maps. This modification increases the model’s sensitivity to target location and size, particularly in complex scenarios. To accelerate inference speed and reduce model size, we substitute traditional convolutions with lightweight Ghost Convolutions. Furthermore, the conventional Complete Intersection over Union loss function is replaced by the SCYLLA-Intersection over Union loss function, which improves both the training speed and the inference accuracy for small targets at greater distances. Experimental results show that the proposed model outperforms YOLOv5s on a custombuilt substation pointer instrument dataset, with mAP0.5 increasing by 2.2%, mAP0.75 improving by 3.8%, and mAP0.5~0.95 rising by 6.7%. Additionally, the model size is reduced by 34.07%. When compared to other widely used models such as Faster R-CNN, YOLOv4-tiny, YOLOv7-tiny, and YOLOv8n, our model shows significant improvements in both accuracy and speed. The model, with a size of only 18.0 MB, demonstrates strong generalization and robustness, making it well-suited for lightweight deployment. Inference speeds on a PC and the Jetson Xavier NX development board reach 154.7 FPS and 18.7 FPS, respectively, meeting the performance requirements for embedded devices used in substation pointer instrument inspections.