To address the DC-side inter-phase voltage imbalance in star-configured cascaded H-bridge static var generator (SVG) under grid unbalanced faults in new power systems, a ripple suppression method combining a hybrid-frequency topology with triple frequency zero-sequence voltage injection is proposed. First, the working principle and power transfer mechanism of the hybrid frequency topology are analyzed, and its overall closed-loop control strategy is modeled and designed. Second, by establishing a ripple mathematical model, a key characteristic is identified: the amplitude of the mid-frequency ripple component is significantly lower than that of the fundamental component. Based on this, the optimal zero-sequence voltage phase angle is derived, and the ripple suppression effect under different zero-sequence voltage amplitudes is quantitatively analyzed. Finally, simulations under two different triple frequency zero-sequence voltage amplitudes are conducted. The results show that the ratio of peak-to-peak ripple before and after zero sequence voltage injection is approximately 1. 1 and 1. 21 for the two cases, respectively. An experimental platform validates the system performance under unbalanced conditions, demonstrating effective DC-side voltage balancing. All results are basically consistent with theoretical derivations, verifying the effectiveness of the proposed ripple suppression method.
To address the challenge of incomplete boundary features and undetermined center coordinates caused by missing object boundaries in image recognition, this paper proposes a boundary feature topological detection method based on an enhanced Hough algorithm. Firstly, a Canny edge detection algorithm integrated with morphological processing is applied to preprocessed image data to extract edge information, ensuring edge continuity and accuracy. Secondly, the improved Hough algorithm is employed to perform boundary topology and localization on images with incomplete boundary information, effectively inferring the locations and extents of missing boundaries. Experimental results demonstrate that the proposed method achieves a mean localization error of 0. 82 mm, with 97. 92% of errors below 2 mm. The approach significantly enhances the precision and effectiveness of detecting target boundary features, accurately reconstructs missing boundaries, and achieves robust localization of targets, even under conditions of partial occlusion or boundary discontinuities.
Deep learning has shown strong capabilities in modeling nonlinear relationships and complex patterns in temperature prediction. However, it faces limitations in processing multidimensional meteorological data and capturing long-term dependencies. To tackle this issue, this study proposes a WTConv-Informer based temperature prediction method enhanced with frequency-aware channel attention. By integrating wavelet convolution with the Informer model and incorporating a frequency-enhanced channel attention mechanism, the method enables deep extraction of global features and precise identification of dominant low-frequency trends and fine grained high-frequency variations in time series. Experiments conducted on the Max Planck Institute for Biogeochemistry weather station dataset show that the proposed model outperforms traditional approaches in prediction accuracy, achieving a mean absolute error (MAE) of 0. 206 4 and a mean squared error (MSE) of 0. 085 4. The findings indicate that the proposed model delivers high accuracy and robust generalization in temperature prediction and demonstrates notable advantages in handling multivariable meteorological data.
To explore the influence of molecular chain dynamics on high-temperature breakdown performance, nano-SiO 2 particles are added to epoxy composites to modulate chain dynamics. The chain characteristics are analyzed by broadband dielectric spectroscopy, and the breakdown strength is measured. The relationship between molecular chain dynamics and high-temperature breakdown is discussed. The results show that with 1% nano-SiO 2 , stable chemical bonds are generated in the interfacial region between nanoparticles and the matrix, suppressing β-chain relaxation, reducing free volume and bulk conductance, and improving breakdown voltage by 12. 68% . At high temperature, the crosslinked structure reduces the thermal expansion coefficient, further decreasing free volume compared to pristine epoxy, and increasing breakdown voltage by 16. 40% . The results lay theoretical foundation to improve breakdown voltage of polymeric materials at different temperatures by suppressing molecular chain dynamics.
Retinal fundus vessel segmentation technology is of great significance for the diagnosis and prevention of diabetes and retinal diseases. However, pixel-level annotation of blood vessels is very time-consuming and costly. In the case of limited labeled data, blood vessel segmentation remains a challenging task. To address the above problems, this paper proposes a Dual Uncertainty guided Dynamic-competitive Collaborative Mean Teacher framework ( DU-DCMT) for semi-supervised vessel segmentation. The framework utilizes a Dual Uncertainty Region-processing Module (DURM) to expand data distribution, enhance the confidence of pseudo-labels, and guide network attention to correct potential error-prone regions. Additionally, a dynamic competitive teacher model is designed to select high-quality pseudo-label generators, jointly train networks, and preserve network diversity for better segmentation performance. Experimental results show that when the labeling ratio is 0. 2, the sensitivity and accuracy of this method on the DRIVE dataset are 0. 817 3 and 0. 967 3 respectively, and on the STARE dataset are 0. 768 2 and 0. 978 3 respectively. Therefore, DU-DCMT effectively alleviates the problem of data label scarcity, and its segmentation performance outperforms current state-of-the-art methods.
To address the challenge of accurately applying distributed dynamic loads in experiments, a method is proposed whereby concentrated dynamic loads are applied at Gauss points to achieve an equivalent effect. This approach utilizes modal truncation in the modal space, discretizes the distributed dynamic load through a modal force equivalence method, and calculates the coefficients of the concentrated dynamic loads using the Gauss-Legendre quadrature formula. In numerical simulations, dynamic responses of a cantilever beam and a simply supported plate under both distributed and equivalent concentrated dynamic loads are analyzed, with the modal superposition method used to evaluate the dynamics of the beam and plate. Results indicate that the acceleration response error under distributed and equivalent concentrated loads does not exceed 4. 5% , and the displacement response error does not exceed 4. 2% . Additionally, the variation of equivalent results under noise in the concentrated loads is examined. The simulation results validate the accuracy and applicability of this equivalent method for distributed dynamic loads.
In order to advance the application of geometric constants in the study of the geometric structure of Banach spaces,this research introduces a new geometric constant L ′ α,β (X),presents its fundamental properties, and explores the relationships with other classical geometric constants. Furthermore, this constant is utilized to characterize inner product spaces. Lastly, the connection with normal structure is investigated.
In current power systems, the resolution of the Optimal Power Flow (OPF) problem is crucial for ensuring the stable operation and efficiency of the grid. This paper proposes a novel blockchain consensus algorithm for tackling the OPF problem in power systems. The algorithm employs the Alternating Direction Method of Multipliers (ADMM) to decompose the OPF problem into local sub-problems, which are integrated into the blockchain protocol. The effectiveness and superiority of the algorithm are verified through simulation experiments on the IEEE-57 and IEEE-118 benchmark systems. Compared with existing blockchain consensus algorithms, the proposed algorithm exhibits better performance in terms of iteration times and computational load, while maintaining the convergence characteristics of the ADMM method. The proposed method achieves decentralized consensus among grid nodes, can adapt to the structural changes of the grid, and offers good robustness and scalability.
Server energy consumption data is a crucial component of power data assets. Accurate prediction of this data can optimize power usage and enhance the economic benefits derived from power data assets. Existing prediction methods struggle to accurately capture the periodicity, long-term trends, and dependencies in energy consumption data. To address these issues, this paper proposes a server energy consumption prediction method based on xLSTM. This method first decomposes server energy consumption data into trend and seasonality components, effectively extracting long-term trends and repetitive patterns. Then xLSTM is used to capture the temporal correlations between time points and the long-term dependencies across time periods, enhancing the accuracy of energy consumption predictions. The proposed method is compared with existing methods on a self-built dataset and four public datasets, achieving the best results in terms of the evaluation indicators MAE and MSE.
To address the issues of low accuracy in wheat leaf disease identification, such as missed and false detections, a TLeaf YOLOv8 algorithm is proposed for detecting wheat leaf diseases. Firstly, the RCS-OSA module is introduced to replace the C2f module in the backbone network, enhancing feature extraction for leaf diseases through reparameterized convolution. Then, a flexible convolution AKConv is designed in the neck, which can adjust the size and shape of convolution kernels flexibly according to different datasets and targets, ensuring detection accuracy while effectively reducing the model's parameter volume. Finally, a small target detection layer is added to address the frequent missed detections or poor detection effects of small targets, and the LSKA attention mechanism is introduced to prevent the disappearance of small target features in redundant information. The model is validated on the self-made dataset WheatLeafNet, and the results show that the mAP @ 0. 5% of the improved model ( TLeaf-YOLOv8 ) is 1. 9 percentage points higher than that of the original model; the model parameter volume and model size are reduced by 0. 3G and 1. 1MB respectively compared to the original network model. While improving detection accuracy, the goal of lightweighting is achieved.
Using existing detection methods such as profilometer measurement, coordinate measuring machine ( CMM), and optical measurement to measure the geometric tolerance of compressor casings can obtain relatively detailed measurement data. Based on the existing measurement data, how to process the noise in the data and convert it into recognizable information is an urgent problem to be solved. Based on the existing detection methods, this paper proposes a method for extracting fault points based on point cloud data. This method first obtains data points located on the central axis of the casing through the rough extraction of point cloud data. Then, it removes outliers from the point cloud data through the MSAC algorithm to fit a higher-precision central axis of the casing. Finally, the distances from point-cloud data points to the fitted central axis are calculated to judge whether the geometric tolerance of the casing inner surface meets specifications, so as to locate fault points. Simulation verification shows that the proposed point-cloud-based data-processing method can effectively identify fault points on the inner surface of the casing and output their spatial coordinates. At the same time, it can reduce the influence of noise in the data on the results, demonstrating a certain degree of robustness.
An intelligent optimal control strategy was proposed for the chaotic motion of the traction gear transmission system of HXD 2 locomotives under the condition of slight pitting corrosion. Firstly, a three degrees of freedom nonlinear dynamic model of the fixed-axis straight-toothed cylindrical gear pair in the case of slight pitting corrosion was established. Secondly, the dynamic response of the system when the parameter meshing frequency changes was numerically calculated, and bifurcation diagrams, phase diagrams, Lyapunov exponential diagrams, and Poincaré mapping diagrams were obtained, so as to analyze and obtain the parameter analysis criterion of chaotic motion, and the weighted control performance index function was constructed accordingly. Then, the controller was designed based on the Radial Basis Function Neural Network(RBFNN), and the Slime Mould Algorithm(SMA) was used to search and select the optimal parameters of the controller. Finally, the influence of the components of the control performance index function and their weights on the effect of chaotic control was analyzed and discussed. The simulation experimental results show that the proposed intelligent optimal control strategy for chaotic motion is feasible and effective: the system can quickly search and stabilize in the expected periodic orbit during the process of micro-amplitude adjustment of the controllable parameters of the system by using the controller with the optimal parameter outputs of suitable disturbance quantities.
As the proportion of renewable energy connected to the grid continues to increase, inverter systems employing traditional vector current control face reduced stability when output power is excessive, which can compromise the safe operation of the entire power grid. To solve this problem, an impedance tuning method was adopted to improve the vector current control of the inverter. Firstly, a small signal disturbance state space model and impedance model of the inverter were established. Secondly, to improve the output power of the inverter, impedance tuning method was used to counteract the feedback caused by negative resistance characteristics. This method not only preserves fast dynamic response, but also extends dynamic power limitation to near static power limitation, ensuring system stability even during high-power operation. Finally, a simulation comparison was conducted between the two control strategies before and after the improvement, verifying the effectiveness of the impedance tuning strategy.
The complex spatial structure and texture details in breast tissue are key factors in tumor classification. In this paper, a lightweight local and global feature fusion network is proposed to solve the problems of inadequate feature extraction, lack of dependency between features, unbalanced classification and slow classification. First, a lightweight multi-dimensional dynamic convolution (LMDC) module is designed to extract key local features from breast images. This module captures the uniqueness of spatial and channel dimensions across different input images while reducing model parameters. Secondly, a lightweight circular recurrent dilated convolution module ( CIR-C) is designed to increase the receptive field of the network and strengthen the spatial connections of breast features. Finally, the classification weight adjustment factor was introduced into the Focal loss function to solve the classification imbalance problem of breast tumors. Experiments demonstrate that the model proposed in this paper achieves recognition accuracies of 98. 1% and 99. 0% on the BreakHis and BACH datasets respectively. The model's parameter size is merely 9. 04 MB and its computational cost is only 1. 89 GB, which significantly enhances the classification performance of breast tumors.
Intravoxel incoherent motion(IVIM)is the only medical imaging technique that does not require contrast agent.However,existing bi-exponential model cannot reflect the non-Gaussian features of diffusion and perfusion motions,and the fitting method of IVIM parameters is easily disturbed by noise and difficult to generalize.To address this problem,a bi-stretched exponential model was established,and a dual optimized deep neural network(DODNN)parameter estimation method was designed to optimize the neural network weight parameters with the goal of simultaneously minimizing the parameter residuals of the bi-stretched exponential model and the residuals of the diffusion-weighted signals in order to estimate the IVIM model parameters accurately.Comparative experimental results show that the DODNN method reduces the diffusion and perfusion residuals by 14.17%compared to the suboptimal method at low signal-to-noise ratios.In addition,each parameter of the bi-stretched exponential IVIM model has the lowest Coefficient of Variation(CV)and the largest parameter contrast to noise ratio(PCNR),which indicates that the bi-stretched exponential model can provide more information for the accurate diagnosis of early-stage diseases.
Action segmentation is an important task in computer vision,aiming to predict the labels and temporal boundaries of action segments in untrimmed videos.To address the limitations of existing"predict-then-refine"strategies,where the high-level video representations generated by the backbone network are often noisy,and the refinement stage tends to suffer from over-segmentation and blurred boundaries.We propose a Multi-Scale Segment-level Action Segmentation Refinement framework(MSRF-AS)that integrates an attention mechanism.Specifically,MSRF-AS enhances segment representation accuracy through multi-head attention encoding and mask matrices,expands the temporal receptive field using dilated Transformers,and precisely localizes the start and end frames of action segments via segment boundary regression.We evaluate our method on three standard benchmarks:50Salads,GTEA,and Breakfast.Experimental results demonstrate that the proposed method outperforms existing approaches on public datasets.
To address the critical issues of traditional electromagnetic current transformers,such as low detection sensitivity,short transmission distance,and inadequate intelligent interconnection capabilities,a fiber-optic current transformer is proposed,which couples stack-type lead zirconate titanate piezoelectric ceramics(PZT)with a distributed feedback fiber laser(DFB-FL).Based on the principle of electromagnetic induction,a toroidal high-permeability energy-harvesting inductive structure with multi-turn coil windings is established.This structure utilizes the alternating electromagnetic field of the measured conductor to provide a driving voltage signal for the PZT sensor.Under electrical excitation,the PZT actuates vibration of the DFB-FL fiber optic sensor,thereby inducing a shift in the central wavelength of the emitted optical signal.A Michelson interferometer is employed to demodulate the varying optical signal,enabling reconstruction of the measured current signal.Finally,through experimental analysis,the key detection parameters of the fiber optic current transformer are determined.Finally,the key detection parameters of the fiber-optic current transformer are obtained through experimental results analysis.The study demonstrates that the fiber-optic current transformer achieves a detection bandwidth of 0～30 kHz,a current detection range of 0.1～78 A,a maximum nonlinear error of 5.3%,a linear fitting accuracy of 99.96%,and a detection sensitivity of 110.1 mV/A.
According to the requirements of wheel attitude measurement in automobile K&C test,the calibration problem of parallel vehicle wheel attitude measurement mechanism based on cable sensor is studied.Using Matlab software,the kinematics model of the measuring mechanism is first built,and the hinge point errors and cable extension errors are introduced into the positive solution model,and the error model of the measuring mechanism is built.The hinge point compensation is obtained by simplex method,and the pose accuracy is improved by hinge point compensation method.It is of practical significance to optimize the mechanism design of parallel mechanisms.
In addressing the issues of manual dependency,low standardization levels,and algorithmic complexity in quality grading of peaches in small-scale peach orchards,an enhanced YOLOv8n-based model named YOLOv8n-ABTW is proposed.The backbone network incorporates a Coordinate Attention(CA)mechanism to precisely capture target position variations and feature correlations.The Neck network integrates a Simplified Bi-directional Feature Pyramid Network(SBiFPN)and a Context Transformation Module(COT)to achieve cross-scale feature connections,thoroughly exploiting both static and dynamic contextual information such as textures,defects,and positional variations in peach images.The Wise-IoU with a dynamic focusing mechanism is employed to mitigate the penalty imposed by low-quality samples.A dataset comprising 1 226 images across eight categories,representing two varieties("Dongxue Mitao"and"Yufei")and four quality grades,was augmented to 3 650 images.Experimental results on a self-constructed peach quality grading dataset indicate that the proposed model's αmAP@0.5 has improved by 2.1%compared to the baseline model,and it also shows enhancements in αmAP@0.5 when compared to other models,providing a feasible technical approach for the detection of peach quality grading.
Aiming at the problem of large machining deformation easily occurring in the machining process of aerospace structural parts,which seriously affects the machining quality,an orthogonal experiment of radial cutting width,axial cutting depth combination,and feed per tooth is designed.The deformation data are obtained by finite element simulation.The milling parameters of aluminum alloy thin-walled parts are optimized based on the empirical formula and particle swarm optimization algorithm.The results show that the average error of the deformation results of the two sets of simulation and physical experiments is13.1%,which has high accuracy.A set of milling parameters is obtained by using the optimization method:radial cutting depth ae=0.2 mm,axial cutting depth combination is 10.54 mm+12.96 mm,feed per tooth fz=0.131mm/z.Compared with the traditional milling parameter combination in the orthogonal experiment,the maximum deformation of the side wall of the thin-walled part is reduced by30.5%.The results of this study have guiding significance for the selection of milling parameters of aluminum alloy thin-walled parts.
