Abstract:Due to the advantages of non-contact, high efficiency, and rich information content of the detection signal, pulsed eddy current technology is widely used in the defect detection of industrial products, especially those with cladding, heterogeneous, and multi-layer conductive structures that are extremely difficult to detect. The design and optimization of pulsed eddy current probe structure is the key to improve the detection sensitivity and accuracy, and a lot of research work has been carried out around the design and optimization of the probe structure; different materials, structures, and defects of different types or shapes need to be characterized by the use of appropriate features, and the selection and analysis of the features is also the key to the research of pulsed eddy current technology; In addition, since the pulse eddy current detection signal is greatly affected by the lift-off height, the accurate detection of defects with unknown lift-off height faces great challenges, so the suppression of the lift-off effect is also one of the research focuses. In this paper, the research progress of pulsed eddy current detection technology is reviewed from the aspects of probe design and development, selection and analysis of feature quantity, and suppression of lift-off effect, etc. At the same time, in order to better promote the development of pulsed eddy current detection technology, the paper, on the basis of the existing research, makes an outlook on the design of pulsed eddy current probe, feature quantity analysis, and the suppression of lift-off effect.

Abstract:Aiming at the problems of low computing efficiency, poor real-time performance and low estimation accuracy of the existing SOH and SOC estimation methods for electric vehicle power battery, a recurrent gated neural network model is proposed to accurately estimate the SOC & SOH of electric vehicle power battery. Firstly, the calculation methods of update gate and reset gate in the Gated Recurrent Unit are improved and the candidate hidden state activation function is replaced by the ThLU function to shorten the training time and effectively alleviate the gradient vanishing. Secondly, the sequence data input method is optimized, and the loop GRU calculation mode is introduced to improve the model computing efficiency and estimation accuracy. Lastly, the model is based on the convolutional neural network and the improved gate recurrent unit, the full-cycle SOH and SOC are simultaneously estimated using the voltage, current, and temperature data collected by the sensors, and the SOH estimation is included in the SOC estimation to eliminate the adverse effects of the aging factor on the SOC estimation. Experimental validation using the Oxford University battery dataset shows that compared with the traditional estimation model, the SOC estimation accuracy of the model proposed in this paper is effectively improved, and the prediction error basically stays within 0.5%.

Abstract:Person re-identification can be regarded as a form of fine-grained visual classification task. Existing unsupervised person Re-ID methods typically focus solely on global features of human bodies, failing to capture accurate fine-grained local features, thereby hindering the recognition accuracy of the models. To address this issue, we propose a ViT-based fine-grained feature enhancement network. This network leverages a vision-language model to generate masks for local regions of human bodies in images. Subsequently, based on the distinct interaction strategies between learnable tokens and image patches within the self-attention mechanism, the class token and introduced learnable local tokens are utilized to learn global and local fine-grained feature representations, respectively. Furthermore, to further enhance feature representation capabilities, a spatial information enhancement module is designed. This module augments feature learning by mining spatial contextual relationships among representative image patches within local regions of human bodies. Finally, utilizing the extracted global and local fine-grained features, online and offline camera-aware contrastive losses are computed separately to bolster the model’s robustness to person identities in an unsupervised environment. Experimental results on the Market-1501, MSMT17, and PersonX datasets validate the effectiveness of the proposed method, achieving mAP/Rank-1 accuracies of 90.3%/95.9%, 59.2%/83.5%, and 91.3%/96.1%, respectively.

Abstract:A dense cascade convolution and self attention feature aggregation network was constructed for the segmentation of retinal vascular images, addressing the difficulties in segmenting small blood vessels and the occurrence of fractures during the segmentation process. The network utilizes multi-scale dense convolution combined with self attention mechanism; To better extract complex feature information of retinal small blood vessels, a dense aggregation module is constructed as the backbone network of the U-shaped network; Embedding self attention patches and multi-scale aggregation modules at the bottom layer of the network to enhance receptive fields and obtain high-dimensional semantic feature information; The feature aggregation module is used in the skip connection part of the model to improve the segmentation accuracy of the model. The experimental results show that on the DRIVE public dataset, the F1 score of the network reaches 83.19%, the accuracy ACC score reaches 97.11%, and the AUC value reaches 98.94%; On the CHASE-DB1 and STARE datasets, compared with Unet, DUNet, SA Unet, and FR Unet networks, the AUC index of this network has achieved the best results so far. Using this network for retinal vessel segmentation, the accuracy and robustness of segmentation have been improved to varying degrees, achieving excellent results in small vessel segmentation and its generalization ability.

Abstract:Aiming at the problem of low positioning accuracy caused by non-line-of-sight and environmental interference in indoor positioning process, an indoor fingerprint positioning algorithm based on The Multi-Verse Optimizer- Support Vector Regression algorithm has been proposed. Firstly, the ranging values are calculated through double-side two-way ranging algorithm with ultra-wideband communication technology. Then, the ranging values are utilized as the fingerprint features to construct a fingerprint database, based on fingerprint database SVR algorithm is adopted to establish the mapping relationship between the positioning coordinates and the ranging values. Finally, the MVO algorithm is proposed to optimize the parameters of cost and γ in SVR algorithm to improve the accuracy of the positioning results. Experimental results demonstrate that the Radial Basis Function is used as the kernel function in the SVR model to significantly improve positioning accuracy. The results of MVO-SVR were compared and analyzed with those of Trilateration, Random Forest, eXtreme Gradient Boosting, and SVR algorithms. In the X direction, the average absolute error is reduced by 20.12%, 54.43%, 60.66%, and 16.21%, respectively; in the Y direction, it is reduced by 79.57%, 54.18%, 59.29%, and 38.17%, respectively. The average positioning error Ep is decreased by 60.73%, 54.38%, 60.01%, and 22.84%, respectively. Moreover, the average absolute errors in both the X and Y directions for the MVO-SVR algorithm reach the centimeter level. The results confirm that the indoor fingerprinting positioning algorithm based on MVO-SVR significantly enhances positioning accuracy and demonstrates promising application potential in complex indoor environments.

Abstract:To address the limitations of traditional distribution network fault location methods, which rely on a single fault diagnosis model, a new fault location method for distribution networks based on image fusion and dual-channel convolutional neural networks is proposed. The aim of this study is to improve the accuracy of existing methods under complex conditions such as high-resistance grounding, noise interference, distributed power supply grounding, and unsynchronized sampling times. First, the zero-sequence current signals are converted into two-dimensional images using Gramian angular summation field (GASF) and Gramian angular difference field (GADF) techniques, providing a basis for image processing. Next, image fusion technology is employed to spatially fuse the GASF and GADF images, resulting in a comprehensive feature image that fully leverages the characteristics of different images, thereby enhancing the richness and effectiveness of feature representation. Subsequently, a dual-channel convolutional neural network model is constructed, where a one-dimensional convolutional neural network and a ResNet50 network are used to extract features from zerosequence current signals and Gramian angular field images, respectively. This design takes full advantage of the strengths of different convolutional neural networks in processing one-dimensional signals and two-dimensional images. Finally, the fused features are input into a Sigmoid function to achieve fault line selection. Experimental results show that this method outperforms traditional methods under various complex conditions, with an accuracy rate, Kappa coefficient, Matthews correlation coefficient, and recall rate of 99.97%, 0.999 3, 0.999 3, and 0.999 5, respectively. These results indicate that the proposed method not only has high accuracy but also exhibits good robustness and stability, effectively addressing challenges such as high-resistance grounding, noise interference, distributed power supply grounding, and unsynchronized sampling times in practical applications. The proposed method provides a novel and efficient solution for fault location in distribution networks, with significant practical application value and broad prospects for promotion.

Abstract:Power station boiler heated area ash is one of the important factors leading to the reduction of boiler thermal efficiency and even affecting boiler output, so accurate prediction of boiler heated area ash fouling condition is a prerequisite for optimizing and improving boiler energy efficiency. To address this problem, the paper takes the cleanliness factor of the economizer of a 300MW power station boiler as the research object, and proposes a combined model integrating timevarying filter-based empirical mode decomposition (TVF-EMD) and convolutionally gated neural unit (CNN-GRU) to predict the change of boiler heated surface energy efficiency. Firstly, the non-linear and non-smooth cleanliness factor raw data are preprocessed by the improved wavelet threshold method to remove noise and outliers; then the processed data are decomposed by TVF-EMD to obtain the preset intrinsic modal components, and the components with the threshold value greater than 0.2 are superimposed and reconstructed according to the autocorrelation coefficient, so as to reduce the impact of the low correlation components on the prediction accuracy while retaining the important features of the raw data; finally, the reconstructed model is combined with convolutional gated neural unit (CNN-GRU) to predict the energy efficiency change of the boiler heating surface. Finally, the reconstructed signal is used to establish the nonlinear relationship between input and output by using the powerful feature extraction capability of convolutional neural network (CNN) and the temporal memory capability of gated recurrent unit (GRU) to achieve more accurate time series prediction. The results show that the decomposition using TVF-EMD can improve the prediction accuracy by 9.628 67% compared with the direct prediction, which provides theoretical and technical support for the subsequent optimization and the formulation of more reasonable soot-blowing strategies.

Abstract:The production workshop environment is complex, with numerous equipment and highly autonomous and uncertain personnel activities. Traditional manual observation methods are difficult to achieve efficient real-time control when facing massive monitoring data. To improve the automation monitoring level of workshop personnel behavior and ensure production safety, a behavior recognition algorithm based on improved DETR is proposed. Through on-site research in the smart workshop, various work behavior and abnormal behavior data were collected to construct an infrared behavior dataset for the workshop, and an improved algorithm was designed based on this. In response to the shortcomings of the original algorithm, relative position encoding is introduced and a spatial modulation joint attention mechanism is adopted to improve the network’s localization accuracy of the object to be detected in the global features. In addition, by introducing Gaussian distribution weights of the object to be detected, the network decoder can more efficiently recognize behavioral features. The experimental results show that the improved algorithm has improved recognition accuracy by 6.97% on self built datasets compared to the original algorithm, and also performs well on public datasets. This improvement method not only provides a more efficient solution for monitoring the behavior of workshop personnel, but also provides strong technical support for the automation and intelligent development of smart workshops.

Abstract:Aiming at the problem of limited wireless sensor network resources and intrusion detection system strategy optimization, this paper proposes a wireless sensor network intrusion detection method based on complex network evolutionary game. Combined with the small world model theory, the connection relationship between network nodes is simulated, and the network connectivity is enhanced and the transmission energy consumption is reduced without changing the original relationship of nodes. Then, the attack and defense game model of wireless sensor network about cluster head nodes and malicious nodes is constructed. The node income is calculated by the income matrix, and the reward and punishment mechanism are used to describe the income change of nodes choosing different strategies in the game process. At the same time, the empirical weighted attraction learning algorithm is introduced to improve the strategy update rules of the traditional game and the algorithm is applied to the intrusion detection system, so that the cluster head nodes can dynamically update the strategy selection and obtain the optimal strategy of intrusion detection under different conditions. The experimental results show that compared with the traditional method, the diffusion depth of the cluster head node detection strategy of the proposed algorithm can reach 79%. Under this algorithm, the cluster head nodes choose to detect the attacks in the sensor network as much as possible while ensuring its own detection income, so as to ensure the network detection rate and reduce the consumption of various resources in the network.

Abstract:Addressing the challenges posed by variability and missing samples in multimodal data, we introduce a Multi-modal Information Fusion (MIF) technique that leverages both vibration and infrared image data. This innovative approach facilitates an effective and rapid assessment of power transformer fault states. First, a bidirectional gated recurrent unit (BGRU) is employed to extract features from the textual data of vibrations, the frequency images derived from vibrations, and the infrared images captured from the power transformer. The BGRU then yields feature vectors corresponding to different modalities. Subsequently, a cross-attention mechanism is utilized to establish relationships between these diverse modalities, enabling feature vector fusion. Finally, a combination of convolutional and fully connected layers determines the fault status of the power transformer. The experiment data come from the 10 kV power transformer, which contains the vibration signal and infrared images. Comparative analysis reveals that the MIF method outperforms benchmark techniques across four evaluation metrics, achieving a commendable fault diagnosis accuracy rate of 96%. Furthermore, the MIF methodology demonstrates its robustness by delivering highly reliable diagnostic outcomes under varying voltage and current conditions, offering a promising solution for the detection of faults in transformer multimodal data.

Abstract:Accurately estimation of the state of health (SOH) of satellite lithium-ion batteries using in orbit measurable parameters is crucial for the safe and reliable operation of satellites. For the SOH estimation of satellite lithium-ion batteries, both performance degradation characterization and assessment should be self-adaptive to different operating conditions. Therefore, to address the insufficient validity of characterization parameters and the reliability of assessment results caused by uncertainty in the model and data for satellite lithium-ion batteries under different operating conditions, this paper proposes a probabilistic SOH estimation method based on BNN-PF. Extracting different health indicators (HI) from measurable parameters during satellite lithium-ion batteries charging process to characterize performance degradation, combining Permutation Entropy with principal component analysis to improve feature recognition ability for different tasks. Furthermore, a Bayesian neural network (BNN) is applied to infer the SOH of lithium-ion batteries and quantify uncertainty. The uncertainty obtained by integrating empirical model with particle filter (PF) algorithms further enhances the adaptability of the proposed method to different operating conditions. The experimental results illustrate that the method proposed in this paper demonstrates good adaptability and universality for SOH estimation of satellite lithium-ion batteries under different operating conditions. The cross-validation test results show that the maximum estimation error is less than 0.01, and the estimation interval coverage of most results is higher than 0.95, indicating that the method has good prospects in spatial application scenarios.

Abstract:To address the issue of serious torque ripple in synchronous reluctance motors, a multi-objective intelligent optimization method for the rotor structure of synchronous reluctance motors is proposed based on machine learning. First, the rotor structural parameters to be optimized for the synchronous reluctance motor are obtained through magnetic circuit analysis, and sensitivity analysis is conducted using the finite element method to determine the variables and their ranges for optimization. Second, a deep neural network is introduced to establish a non-parametric rapid calculation model for the synchronous reluctance motor, and a nonlinear mapping relationship between the optimized variables and torque is constructed to accurately model the electromagnetic characteristics of the motor. Based on this, an improved particle swarm optimization algorithm based on reinforcement learning is proposed. This approach adjusts the learning factors of the optimization algorithm online according to the reward function mechanism in reinforcement learning, improving the convergence speed and global optimization accuracy of the particle swarm optimization algorithm. Finally, with the objectives of minimizing torque ripple and increasing average torque, the improved particle swarm optimization algorithm and the deep neural network model are used for global optimization of the motor rotor structural parameters under multiple operating conditions. The simulation and experimental results show that the optimized synchronous reluctance motor using the proposed method not only has lower torque ripple compared to the initial motor model, but also slightly increases the average torque.

Abstract:A fault diagnosis method based on multi-information weighted fusion and two-dimensional convolutional neural network (MIWF-2DCNN) is proposed to effectively monitor the operating status of hub motors for distributed electric vehicles under complex operating conditions and improve the accuracy of bearing fault identification. Firstly, the multi-directional vibration monitoring signals of in-wheel motor bearing were reconstructed by two-dimensional data reconstruction and time-frequency transformation respectively, and then converted into grayscale images one by one. According to the direction order, the time-domain grayscale atlas and time-frequency domain grayscale atlas were stacked as the input of the fault diagnosis model. Secondly, the network structure of efficient channel attention mechanism (ECANet) was improved, and the improved efficient channel attention mechanism (iECANet) was proposed. The core idea of IECANET was to add a global maximum pooling (GMP) branch on the basis of global average pooling (GAP), and update the weight coefficient of each branch based on the contribution of effective information. Then, the fault features in time domain and time-frequency domain were extracted to realize the weighted fusion of multi-information. Thirdly, GMP was used to simplify a fully connected layer of the traditional two-dimensional convolutional neural network (2DCNN) model to achieve network lightweight. Finally, based on the experimental data of in-wheel motor under different working conditions, the corresponding verification under the same working condition, cross validation under different working conditions and ablation experimental verification were carried out. The results show that the proposed MIWF-2DCNN model can effectively extract the fault features of in-wheel motor bearing, and the fault recognition rate remains above 95% in complex environments and variable working conditions, which is better than the traditional LeNet-5 and 1DCNN models.

Abstract:The study of model assisted probability of detection in eddy current nondestructive testing requires a large amount of simulation data, while high-precision physical model calculations demand considerable time and are often impractical. The surrogate model is an efficient mathematical model that can replace time-consuming and complex physical models, and is widely used in design optimization problems. CoKriging, a model that fuses high and low-precision data, utilizes a large amount of low-cost and low-precision data, and a small amount of high-cost and high-precision data, which significantly improves the modeling efficiency over Kriging model. It is a very promising surrogate model. This article applies the CoKriging model to the study of the probability of detection aided by the eddy current nondestructive testing model. In the case of detecting groove defects on the surface of a metal plate using a finite-section coil, the CoKriging model is constructed using physical model calculations for some training points. After verifying the accuracy, the CoKriging model can replace the physical model for MAPoD analysis. By comparing the key parameters of MAPoD calculated by the physical model, the accuracy and efficiency of the CoKriging model are verified. The results show that compared with the Kriging model, the CoKriging model only requires fewer sample points to train the model to meet the defined accuracy requirements and in the best-performing example, its construction time is only 7% of that of the Kriging model, greatly improving the efficiency of the MAPoD.

Abstract:Small object detection in drone images is one of the key and difficult research areas. Compared with large targets, small targets have fewer features and are more susceptible to interference from occlusion and complex backgrounds. To address this issue, a multi model fusion object detection network YOLO-DA based on YOLOv7 tiny is proposed. Firstly, add layers for detecting small and extremely small targets to enhance the network’s ability to learn small target features; Secondly, the spatial adaptive feature fusion ASFF-L detection head is introduced to suppress the inconsistency of features at different scales by learning spatial filtering conflict information, achieving adaptive fusion of multi-scale features; Finally, DCNS deformable convolution was introduced and a modulation mechanism was designed to expand the range of deformable modeling, enhance the modeling ability of the model, and reduce the impact of occlusion overlap on detection. Through experimental verification, the proposed method achieved an average accuracy of 44.7% and a inference speed of 71 fps on the Visdrone2019 dataset. The average accuracy was improved by 9.7% compared to the baseline algorithm, and the model memory was 63.8 M, enabling real-time detection. Through ablation and comparative experiments, it has been shown that YOLO-DA significantly reduces false positives and false negatives in drone aerial image detection, and has higher detection performance. Moreover, the algorithm parameters and computational complexity can meet the real-time detection requirements of edge devices such as drones.

Abstract:Aiming at the problems of low search efficiency, poor path smoothness and security of A* algorithm, and low real-time pathfinding efficiency of DWA integrated with global path planning algorithm, a hierarchical smooth optimization A* guided DWA (HSA*-G-DWA) path planning method for mobile robots is proposed. Firstly, the double dynamic weighting factor is introduced into the cost function of A* algorithm and the collision constraint is developed to avoid the search of unrelated extension nodes, so as to improve the efficiency and security of path search. Secondly, the hierarchical smoothing optimization strategy is designed to eliminate redundant nodes and turning nodes in the path, and reduce the number of path nodes and the path length. After that, the initial global path is generated by segmented interpolation of lines without any obstacle constraints and arcs with obstacle constraints to ensure the safety and smoothness of the path. Then, if the mobile robot encounters unknown obstacles in the process of tracking the global path, it uses the global path to guide DWA to generate the local dynamic correction path for obstacles avoidance and returning to the remaining global path, which reduces the amount of real-time calculation. Finally, the simulation results show that the path search time and path nodes of the proposed HSA*-G-DWA algorithm are reduced by 88.43% and 86%, respectively, and the smoothness and security of the path are better than the A* algorithm in the static environment; and the HSA*-G-DWA algorithm can avoid unknown obstacles in the unknown environment in real time. Compared with the DWA algorithm, Dijkstra algorithm, RRT algorithm and other fusion algorithms, the path length is reduced by 25.78%, 18.65%, 30.48% and 14.59% on average, and the path search time is reduced by 67.39% on average.

Abstract:Aiming at the demand for rapid detection of food residues on the surface of equipment during food processing, a portable fluorescence imaging detection system was developed using spinach juice as an example, and on this basis, the quantitative function of the system for chlorophyll A, which is a substance with high content in spinach juice, was added. The system is mainly composed of a light source, a CCD camera, a display and a control and processing unit, and the light source and filter used to stimulate the substance to produce fluorescence can be adjusted and replaced according to the actual demand, while the control software of the detection system is developed, and the display is used for human-machine interactive operation, mainly realizing the functions of image data acquisition and processing.Qualitative experiments were carried out on spinach juice residues on the surface of three widely used food processing materials, and the detection results showed that the portable fluorescence imaging detection system was able to better identify chlorophyll residues on the food processing surface when the surface itself did not have a fluorescence effect.The quantitative detection results of chlorophyll A showed that when the concentration of chlorophyll A solution was in the range of 0~8.5 μg/mL, the determination coefficient R2 of the concentration prediction model can reach 0.99, and the root mean square error (RMSE) of the prediction results was not higher than 0.24 μg/mL. The results of this study show that the developed portable fluorescence imaging detection system is feasible and effective for the detection of substances with specific fluorescence effects, and the system has the advantages of a large field of view, small size and easy to carry, etc. The system can be used for real-time on-site detection, and in addition to the detection of food residues on the surface of food processing, it can also be extended to the detection of plant pathology and other fields, which has a broader application prospect.

Abstract:Gas-liquid two-phase flow, pervasive in energy and chemical industries, presents significant measurement challenges due to its inherently complex and dynamic nature. Accurate quantification of flow parameters remains elusive, given the variability in phase distribution and interaction dynamics. To address the need for accurate gas-liquid two-phase flow measurement, a novel sensor design is introduced. Leveraging the acoustic emission sensor’s capability to detect flow-induced noise and the pronounced variation in near-infrared absorption across different media, the proposed sensor is specifically tailored for plug flow characterization in two-phase systems. Acoustic emission probes were dually installed in both the venturi pipe and its extension section, complemented by dual near-infrared photodetectors positioned within the venturi pipe. A synchronized acoustic emission and near-infrared acquisition system was developed. Utilizing this setup, 54 datasets of plug flow were meticulously gathered on the high-precision gas-liquid two-phase loop at Hebei University. Through integrated processing, characteristic parameters of the two-phase flow were successfully extracted. Time-domain analysis was employed to extract the standard deviation and skewness from the acoustic emission and near-infrared datasets. In conjunction with parameter fitting techniques, a predictive model for two-phase flow was formulated, followed by comprehensive error analysis. Through verification, the relative deviation of 92.6% of the predicted flow value is within ±20%. The results show that the multi-sensor information fusion scheme based on acoustic emission sensor and near-infrared sensor provides a new way to study the flow characteristics of gas-liquid two-phase flow.

Abstract:A sensorless control method for synchronous reluctance motor based on super-twisting sliding mode adaptive observer was presented for the problems about low accuracy of speed estimation and poor dynamic performance about the sensorless control based on model reference adaptive system. Firstly, the inductance nonlinear model of synchronous reluctance motor was constructed by finite element simulation, and the inductance parameters were updated in real time according to different operating conditions of the motor, so as to improve the accuracy of the model in the observer. On this basis, the super-twisting sliding mode algorithm was used to replace the PI adaptive link, and the super-twisting sliding mode adaptive observer was constructed with the selected linear compensation matrix to reduce the estimation error. Finally, the integrated global fast terminal sliding mode speed controller was introduced to improve the dynamic performance of the system. Simulation and experimental results indicate that, compared to the sensorless system based on model reference adaptive control, the proposed strategy achieves a faster speed response and smaller overshoot during the startup phase, ensures smoother motor operation, exhibits minimal speed fluctuations and faster speed error convergence under sudden load changes, and demonstrates higher speed and rotor position identification accuracy throughout the entire variable-speed operation phase, thereby showcasing excellent dynamic and steady-state performance for high-performance motor operation.

Abstract:According to the current development trend of wireless communication systems, in order to significantly improve the efficiency of communications. a low coupling multiple-input multiple-output antenna based on split ring resonator and complementary split ring resonator is designed for Ka-band. The overall dimensions of the antenna are 40 mm×25 mm×1.2 mm. First, the front side of the antenna consists of an elliptical radiating patch with a hollow circle, trapezoidal microstrip feedline and split ring resonator, to ensure that the antenna operates in the Ka-band(26~40 GHz). Second, on the back is a rectangular ground plate with a complementary split ring resonator, which can effectively achieve the purpose of decoupling. Thus realizing the effect of low coupling. Simulated and measured results show that, operating bandwidth of 26~40 GHz（the relative bandwidth of 50%）, return loss less than -10 dB, coupling less than -26 dB, envelope correlation coefficient is less than 0.001, good radiation directional map, stable gain, high radiation efficiency. As a result, the designed antenna not only has a simple structure, compact size and wide frequency coverage, but also has superior performance. it can be widely used in 5G millimeter-wave related areas. In summary, the application of SRR/CSRR has led to the reduction of the mutual coupling of millimeter-wave MIMO antenna, the feasibility and effectiveness of the SRR/CSRR technique as a novel decoupling technique are verified.

Abstract:When mechanical vibration occurs in a three-phase asynchronous motor, the poor electrical contact points in the main circuit will generate a series of fault arcs under the influence of vibration, which will compromise circuit safety and potentially lead to electrical fires. The vibration condition complicates the fault arc signal, so this paper proposes a highly real-time series fault arc detection method under vibration conditions. First, experimental current data is dynamically preserved by constructing a sliding memory matrix. Secondly, the texture features of the sliding memory matrix are extracted using orthogonality direction local ternary pattern (OD-LTP). Finally, the amplitude of the grayscale distribution histogram of the statistical OD-LTP images is taken as the feature vector. A vibrating series fault arc detection model is established using support vector machine (SVM) optimized by sand cat swarm optimization (SCSO). By comparing different matrix parameters, the proposed method achieves an accuracy of 99.2%. Through a comparative analysis of different feature extraction methods under various working conditions, it is shown that the proposed method is not only suitable for industrial motor inverter systems under different working conditions, but also exhibits higher real-time performance compared to other feature extraction methods.

Abstract:The application of artificial intelligence technology to the field of fault diagnosis can realize the automation and intelligent diagnosis of power equipment and improve diagnosis accuracy and efficiency. Taking the single-input multiple-output flyback switching power supply as an example, for the problem of abnormal circuit performance caused by the failure of its fragile components, a non-intrusive switching power supply fault diagnosis method fusing the input current and output voltage information is proposed by analyzing the signal characteristics and divisibility of different fault modes. A multidimensional feature vector fusing time-frequency domain information consisting of time-domain features and frequency-band wavelet packet singular entropy features is constructed, and the mapping relationship between fault features and fault modes is established. Then, a deep neural network (DNN) fault diagnosis method based on artificial intelligence technology is proposed to monitor the operation status of the flyback switching power supply in real time, identify the fault location in time through data analysis, and provide early warning for potential faults. The experimental results show that the method proposed in this paper has a good diagnostic effect on both single-fault and multi-fault modes, the diagnostic accuracy can reach 97.9%, and the method can show high diagnostic accuracy and strong anti-interference performance under different working conditions.

Abstract:In response to the tracking control challenges faced by freight trains under multi-source disturbances, this paper proposes a fixedtime replacement sliding mode control method based on a cascaded disturbance observer. Initially, a multi-mass dynamic model that takes into account the inter-vehicular forces is constructed. To address the scenario where both matched and unmatched disturbances coexist, a cascaded structure disturbance observer is designed to estimate multi-source disturbances concurrently, thereby relaxing the traditional prerequisite of disturbance observers that dictates disturbances must vary slowly. Utilizing the disturbance observation data, the train dynamics model under unmatched disturbances is transformed into a matched mode. Ultimately, a distributed control strategy based on the alternative sliding mode approach is put forward. Simulation results demonstrate that the proposed cascaded disturbance observer can accurately estimate multi-source disturbances within 0.5 seconds. Compared with traditional research on train tracking control, the proposed control strategy manages to swiftly handle a series of destabilizing issues caused by multisource disturbances. While ensuring the stability of inter-vehicular forces, it achieves robust tracking control of both speed and displacement indices. Relative to conventional control methods, the system convergence time is enhanced by more than 5 seconds, reflecting superior real-time performance and robustness.

Abstract:Aiming at the problem that the internal defect detection signal of gear is easily disturbed by noise and it is difficult to extract the defect feature information, an ultrasonic detection signal of gear defects denoising method based on dynamic local entropy and adaptive decomposition is proposed. The empirical mode decomposition is used to decompose the ultrasonic detection signal of gear defects adaptively, and the preprocessed signal is obtained. The local entropy distribution of preprocessed signal is calculated by dynamic local entropy theory, the defect echo interval is determined with the local entropy threshold, and the defect echo signal is then obtained. The defect echo signal is denoised based on the empirical wavelet transform, and the signal smoothness is improved through the quadratic polynomial least square algorithm. The final denoised result of the ultrasonic detection signal of gear defects is obtained. Simulations and actual measurement experiments are carried out to verify the denoising performance of the proposed method. The simulation experiment results show that for simulation signals under different SNR conditions, the mean SNR of the proposed method after denoising is 21.34 dB, and the mean square error MSE is 0.000 2 V. The denoising effect of the proposed method is significantly better than EMD and Wavelet Transform. The mean SNR and mean square error of the two methods are 10.43 dB, 12.56 dB and 0.001 9 V and 0.001 4 V respectively. In addition, the proposed method has better denoising effect and stronger robustness under different SNR conditions. The experimental results show that the proposed method can remove the complex noise interference in the ultrasonic detection signal of metal gear defects, and effectively improve the quality of the ultrasonic detection signal of metal gear defects.

Abstract:Vortex flowmeter is a flowmeter based on the principle of fluid oscillation to measure the flow rate, but the output signal at low flow rate is very weak, in the actual measurement process will superimpose a variety of on-site noise. At present, the small diameter vortex flowmeter used in industry has measurement limitations under small flow conditions, and the minimum diameter is generally only DN15. Aiming at the existing problems of vortex flowmeter, the structure and parameters of small diameter vortex flowmeter DN10 and below are studied, to meet the new measurement requirements in industry. Based on the simulation calculation platform, the numerical calculation model of the micro-diameter vortex flowmeter with different size parameters is established and the simulation calculation is carried out, and the optimized structure size of the vortex flowmeter is proposed. By studying the intensity and frequency of pressure fluctuation at different pressure monitoring points, the best position of sensor is analyzed. Finally, the electromagnetic flowmeter is used as the standard table to test the developed micro diameter vortex flowmeter in the micro range. The error and accuracy of the flowmeter are tested, and the relationship between Re and St and the instrument coefficient are analyzed. The results show that the measurement characteristics obtained by simulation are in good agreement with those measured by experiment. The relative error of the micro diameter vortex flowmeter is better than 0.5%, repeatability and other indicators can meet the measurement standards in the micro-range.