Abstract:For the problem that the balise uplink (BU) signal transmission was interfered in the complex electromagnetic environment of high-speed railway, a denoising method based on the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) combined with wavelet packet adaptive threshold was proposed. Firstly, the CEEMDAN algorithm was used to decompose the simulated BU signal into 12 modal components, and the components were determined to be correlated or irrelevant based on the correlation coefficients; Then, the relevant components were reconstructed into the denoised BU signal after wavelet packet denoising processing; Finally, signal noise ratio (SNR) and root mean square error (RMSE) were selected as evaluation metrics to compare this method with six widely used denoising methods. The SNR increased by 0. 486 1 ~ 6. 144 dB and the RMSE decreased by 0. 054 9 ~ 11. 091. To verify the practical application effect of this method, this joint denoising method was adopted to denoise the measured BU signal. The results of simulation and experimental verification showed that the BU signal denoised by the joint denoising method not only effectively removed the noise component, but also preserved the signal features well, proving that this method can be applied to solve the problem of actual BU signal interference.

Abstract:Aiming at the difficulty of rolling bearing fault feature extraction in the encoder instantaneous angular speed (IAS) signal, a parameter-adaptive SMHD rolling bearing IAS signal feature extraction method is proposed by combining the advantages of the sparse maximum harmonics-to-noise-ratio deconvolution ( SMHD) algorithm, which can extract the periodic impulse fault component in the signal without a priori period. Firstly, the IAS signal is estimated using the forward difference method. Then, the fault characteristics (FC) are utilized as an adaptive criterion for selecting the optimal length of the SMHD filter, achieving adaptive determination of the filter length. Subsequently, the optimized filter length is applied to enhance the IAS signal using the SMHD algorithm. Finally, the fault characteristics of the rolling bearing are revealed through envelope analysis. The effectiveness of the proposed method is validated through analysis of both simulated and measured data.

Abstract:Under the long-term influence of external incentives such as ocean current scour, human activities and geological movements, the buried depth of submarine cables is prone to continuous changes, which leads to a series of serious problems such as a sharp increase in the failure rate of submarine cables and a sharp decline in load capacity. This paper takes 35 kV photoelectric composite three-core submarine cable as the main research object, uses the finite element method to simulate the temperature field of submarine cable under different ambient temperatures, different burial depths and different carrying rates, establishes the relationship curve between cable core temperature and buried depth under different ambient temperatures, and the relationship curve between cable core temperature and buried depth under different ambient temperatures, and fit the corresponding relationship formula. The general formula of the relationship equation for real-time prediction of submarine cable buried depth was integrated. Finally, by using the multi-source data of the actual project monitoring, the buried depth and change trend of submarine cable are predicted in real time through the general formula. At the same time, combined with the measured value of submarine cable disturbance energy and submarine cable buried depth, it is verified that the prediction method is not only feasible and correct, but also has high prediction accuracy.

Abstract:In order to meet the requirements of terahertz high-resolution detection and real-time processing, the photoconductive antenna is used to generate and detect terahertz time-domain spectral signals. Based on the field programmable gate array, the functions of terahertz time-domain spectrum acquisition, Wiener filtering deconvolution processing, transmission and host computer display are realized. The collected terahertz time-domain spectral data is subjected to Wiener filtering deconvolution processing to achieve the effect of restoring terahertz signals, improve time resolution and noise reduction. The data is transmitted to the host computer by Ethernet transmission for real-time display. In view of the wide pulse width after deconvolution of terahertz signals in actual detection, it is proposed to introduce a frequency-related function into the Wiener filtering deconvolution algorithm to optimize the algorithm, so that the pulse width of the signal becomes narrower and the detection accuracy is improved. Compared with the original algorithm, the signal-tonoise ratio of the optimized Wiener filter deconvolution algorithm is increased by 7 dB, the pulse width is reduced by 0. 2 ps, and the higher detection resolution is realized. The algorithm is implemented in FPGA, the accuracy error is less than 0. 7%, the processing efficiency is improved by 14. 29 times, and the post-processing time of the host computer is reduced.

Abstract:In order to improve the accuracy of sliding electric contact failure judgement of the pantograph-catenary, a multi-strategy fusion of improved northern goshawk optimisation algorithm (INGO) and least squares support vector machine (LSSVM) sliding electric contact failure diagnosis model is proposed. Firstly, the self-made sliding electric contact testing machine is used to carry out friction experiments, analyse the change rule of the current-carrying stability coefficient under different working conditions, and determine the criteria for the pantograph-catenary contact failure; secondly, the tent chaotic mapping, uniformly distributed dynamic adaptive weights, and the golden sinusoidal algorithm and the nonlinear convergence factor are used to improve the deficiencies in the NGO, and the simulation is carried out through the test function. Test, the results prove that the improved northern wing algorithm ( INGO) convergence speed and stability is better; finally, using the improved northern eagle optimisation algorithm on the model’ s parameter optimisation, to establish the sliding electrical contact failure diagnostic model. Comparing the proposed model with other diagnostic models, the diagnostic accuracy is improved by 16. 67%, 12. 5% and 8. 33% respectively, which further proves that the diagnostic model has high accuracy and generalisation ability.

Abstract:In order to solve the problem of continuous and accurate localization in complex environments such as slope, feature degradation and GNSS signal loss, a multi-sensor fusion scheme based on ground constraints is proposed in this paper to improve the overall performance of SLAM algorithm. Firstly, the key frame selection strategy under different system states is proposed. By increasing the number of key frames in the starting position, the positioning jump caused by factor map optimization is avoided, and continuous and accurate pose output is obtained. At the same time, in order to prevent the loopback detection failure caused by error accumulation, this keyframe strategy is used to effectively increase the subkeyframe set of the current frame, and improve the robustness of the loopback detection algorithm. Secondly, to solve the problem that IMU drifts too much in the height direction after long-term operation, this paper constructs the ground constraint according to the extracted ground points and introduces it into the factor graph for optimization. Finally, the mobile robot experiment platform is used to complete the data collection of different scenes on campus, and the effectiveness of the proposed algorithm is verified. The comparison test between KITTI data set and LIO-SAM algorithm is carried out, and the error analysis shows that the proposed algorithm has better positioning accuracy.

Abstract:In response to the problem of small differences in blister defect features and background pixels in tire defect images, as well as difficulty in detection, Skip-GANomaly is adopted as the basic framework to propose the fusion attention mechanism generative adversarial network (FAMGAN). Firstly, the skip layer between the encoder and decoder in the generator consists of an attention feature fusion (AFF) module and a convolutional block attention module (CBAM) module, which improves the focus on target features and reduces image feature loss. Then, a joint pyramid upsampling (JPU) module was added to the discriminator to improve the speed of the model in detecting image defects. Finally, the FAMGAN network proposed in this article will be trained, tested, and evaluated on a selfmade tire defect dataset with classic generative adversarial networks in recent years. The experimental results show that the proposed network achieves an accuracy of 0. 837 for tire blister defect detection, which is nearly 30 percentage points higher than the Skip GANomaly network.

Abstract:In the process of load identification of ball mill under varying working conditions, a domain adaptation method based on domain antagonism and classification difference is proposed to solve the problem that domain adaptation method does not consider the target domain sample in the feature transfer between source domain and target domain. The method uses domain adversarial training to align the features between source domain and target domain. At the same time, two classifiers are introduced to detect samples far away from the target domain, and the inconsistency between the maximization and minimization of the classifiers is utilized to realize the adaptive matching of the features of the target domain and the source domain to achieve a better domain adaptation effect. In order to verify that the method of training the classifier error can consider the in-class boundary to improve the load recognition accuracy on the target domain, a migration experiment is designed to analyze the impact of its difference loss function on the model migration performance. The experiment shows: When the classifier loss value is greater than 0. 02, the accuracy of the prediction model will decrease by 0. 8% ~ 1. 2%, and the load accuracy is higher than that of the model without the classifier differential loss, which can reach 95. 78%. Compared with two classical transfer methods, the advantages of this method in the application of mill load identification under varying working conditions are verified.

Abstract:Aiming at the difficulty of obtaining the posterior distribution of measurement model in Bayesian uncertainty evaluation, a method based on accept-reject sampling is proposed to realize Bayesian measurement uncertainty evaluation. For linear/ nonlinear measurement model, the prior information being measured is obtained by using Bayesian hypothesis or Monte Carlo method, the accepted sampling points being measured are obtained based on accept-reject sampling. Then the posterior distribution is formed based on these accepted sampling points, and the measurement uncertainty evaluation results are obtained by statistical inference. Through the two evaluation examples which come from the specification and practical measurement application, it is verified that the Bayesian uncertainty evaluation method using the accept-reject algorithm can obtain reliable evaluation results compared with traditional GUM and MCM methods, the process of obtaining the Bayesian posterior distribution is simple, and it is feasible and practical in the application of measurement uncertainty evaluation under the condition of without / with historical information.

Abstract:In the electronic warfare environment, the performance of radar and jammer will be affected by the lack of understanding of the battlefield game environment and the inaccurate estimation of spectrum information. This paper studied the energy allocation problem of radar transmission waveform and jamming waveform, and radar and jammer can obtain better performance by optimizing their own transmission waveform in the process of confrontation. In reality, because the radar cannot obtain accurate target spectrum and jamming spectrum, and the jammer cannot obtain accurate target spectrum and radar transmitted signal spectrum, the radar and jammer are taken as leaders respectively, a hierarchical game model is established, and a double robust waveform design method is proposed. The optimization model of dual robust transmit waveform is established under the energy constraint, and the dual robust radar transmit waveform and dual robust jamming waveform are solved by Lagrange multiplier method. Simulation results show that the proposed method can improve the radar parameter estimation performance and jammer performance in the worst case.

Abstract:To address the challenges of low transduction efficiency, poor signal-to-noise ratio, and the influence of internal ultrasonic waves within magnets on signal detection in electromagnetic acoustic transducers (EMAT), conventional methods have employed copper foil backings or magnetic flux concentrators. However, while copper foil backings alleviate magnet resonance, they also reduce signal amplitude. On the other hand, existing flux concentrators suffer from complex structures, manufacturing difficulties, and high costs. This paper presents a novel EMAT design that incorporates a trapezoidal silicon steel concentrator positioned between the permanent magnet and the coil. The trapezoidal concentrator, with its simple structure fabricated using adhesive bonding techniques, effectively eliminates ultrasonic waves within the magnet by increasing the distance between the magnet and the coil. Moreover, its unique trapezoidal shape concentrates the magnetic flux lines into the coil’s active region, thereby enhancing the amplitude of the echo signal. The optimal dimensions of the trapezoidal silicon steel concentrator, which significantly impact the echo signal’ s amplitude, were determined through orthogonal experimental design on the COMSOL simulation platform. The simulation and experimental results demonstrate that the signal amplitude of the EMAT with the optimized trapezoidal concentrator is approximately 60% higher compared to the EMAT without a backplate. Furthermore, the signal amplitude of the EMAT with the optimized trapezoidal concentrator is approximately six times greater than that of the EMAT with a 0. 1 mm copper foil backplate.

Abstract:A structure sparse channel prior ( SSCP) blind image deblurring approach is presented to address the issues of inaccurate major structures and unclear edges in the blind image deblurring process. A prior method of SSCP shows that blurred images have less structured sparse channels than sharp images. Using the performance features of SSCP, it is introduced as a new regularization term into the standard deblurring model, and a novel blind deblurring model is created to achieve accurate estimation of the blur kernel. Through the coordinate descent approach alternately optimizes the latent image and blurry kernel variables. Finally, deconvolution is used to obtain deblurred clear restored images, subjective and objective comparison experiments on benchmark datasets and natural state blurred images, and application expansion experiments on human faces and low-brightness real blurred images. The experimental results show that the proposed method outperforms the classical deblurring method in terms of blur removal, structure restoration, edge retention, and visual effect by an average of 1. 72%, and the computing device designed by this method can achieve a high-precision clarity to process blurred images in the field of machine vision.

Abstract:Accurate control of gas odorization concentration plays an important role in ensuring the safe transportation and use of gas. However, due to the lack of monitoring of the actual odorization effect at the end-users, the existing linear control algorithm represented by the gas flow rate proportional open-loop control is difficult to achieve a good control effect. For this reason, based on the least-squares identification of the open-loop step response and taking the odorant concentration at the gas end-users as the control object, a CARIMA model of the gas odorization process is established, and an improved generalized predictive control algorithm is proposed on the basis of the traditional generalized predictive control, and the feasibility of this control algorithm is verified by the simulation tests and field experiments of the gas odorization control. The research results show that compared with the conventional gas flow rate proportional openloop control, the improved generalized predictive control algorithm can accurately control the odorant concentration at the end-users in real-time online, with the steady-state error of less than 2 mg / m 3 , the overshoot of less than 10% when the working condition changes, and the average absolute error of less than 1. 5 mg / m 3 , which can meet the need of accurate control of gas odorization process. The research results improve the efficiency and quality of gas odorization while playing a role in saving costs and improving gas safety, which has good application value.

Abstract:To reduce the impact of dynamic environment on the localization accuracy and stability of visual / inertial navigation system, a dynamic feature removal visual / inertial navigation method is proposed in this paper. Based on the VINS framework, this method uses structural similarity as the cost volume to generate an end-to-end network for dynamic regions detection. Symmetric optical flow screening is then performed on the identified dynamic regions to remove non-consistent outliers and further eliminate dynamic features that affect localization. The cost function is constructed by fusing visual and inertial measurements, and the nonlinear optimization method is used to estimate the unmanned system states effectively. The experimental results show that the visual / inertial navigation method with dynamic feature removal has good localization accuracy and stability, the position root mean square error is 0. 081 and 1. 982 m on EuRoC publicly available datasets and real scenario data respectively, which is only 35. 5% and 24. 9% of VINS. This method can provide accurate position information in complex application environment, and has good practical value in the navigation of low-cost unmanned systems.

Abstract:At present, algorithms for monitoring the driver’s line of sight area usually use deep learning models to directly classify image features. This method relies on the driver’ s line of sight area data collected from a fixed cockpit perspective. However, due to differences in driver appearance, sitting habits, and camera installation positions, it is difficult to obtain a large amount of comprehensive data, resulting in a decrease in classification accuracy. How to improve the accuracy of line of sight recognition using only small sample datasets has become a challenge. This article will design an adaptive line of sight region calibration method based on semi supervised learning theory. Firstly, the L2CS model is used to regress the two-dimensional vector of driver’ s line of sight angle in small sample data. Then, statistical analysis is used to mine the generalization prior knowledge of driver’ s line of sight angle and line of sight area mapping. This knowledge is used for line of sight area calibration, removing invalid line of sight landing points in non-inspection areas, and completing fine classification of driver’s personal line of sight area in a sliding window manner. Through experiments, it has been proven that this method solves the problem of low cross domain capability of end model data, improving accuracy and recall by 22. 4% and 10. 3% respectively, and the calibration results have adaptive adjustment ability.

Abstract:To assist maintenance personnel in detecting the fastening status of railway locomotive bogie bolts through visual image analysis, a railway locomotive bogie bolt fastening status detection method based on image recognition is proposed. Firstly, the YOLOv7 algorithm is used to quickly locate bolts in the image, and the strong robustness and generalization ability of deep learning algorithm are utilized to accurately obtain bolt target detection results images including bolts and their positioning paint in various scenarios of maintenance. Secondly, the bolt target detection result image is converted into YCbCr space, combined with the color characteristics of the bolt positioning paint, the Cr component image is extracted, and an adaptive segmentation algorithm is applied to effectively filter out background pixels to obtain a binary image containing only the bolt positioning paint. Finally, based on the differences in shape, position, and angle of bolt positioning paint, Hu moment features were extracted as quantitative representations of bolt positioning paint status information, and a classification model was established using SVM to obtain the final bolt tightening status detection results. The experimental results show that this method fully utilizes the characteristics of railway locomotive bogie bolts. While ensuring the accuracy of bolt target detection and bolt positioning paint segmentation, the accuracy of bolt tightening status in railway locomotive bogie bolts in all scenarios is 92. 42%, the recall rate is 94. 55%, and the average accuracy rate is 93. 28%.

Abstract:The permanent magnet synchronous motor (PMSM) speed control system is easy to be affected by external load disturbance, this paper presents a fixed-time model-free sliding mode control strategy for PMSM based on improved sparrow search algorithm. Based on the input-output ultra-local model of the speed loop, a fixed-time sliding mode speed controller is constructed, and a fixed-time disturbance observer is designed to estimate the unknown disturbances of the ultra-local model, and through the way of feed-forward compensation to weaken the adverse effect of unknown disturbance on the speed control system, and improve the anti-interference ability of the speed control system, the fixed-time convergence of the speed controller and the observer is proved by Lyapunov function. In order to obtain the parameters accurately, an improved sparrow search algorithm is proposed to optimize the parameters of the controller. The experimental results show that the control scheme proposed in this paper effectively solves the overshoot problem compared with the PI control, accelerating the stability time by 22. 8% and reducing the loading speed change by 75%, indicating the effectiveness of the proposed control scheme.

Abstract:To improve the efficiency of antenna modeling and change the problem of slow speed and low efficiency of traditional modeling methods, an antenna modeling method using improved butterfly algorithm ( BOA) to optimize multilayer feedforward neural network (back propagation neural network ( BPNN)) is proposed. Firstly, the BP neural network optimized by the butterfly algorithm is established with the multilayer feedforward neural network as the base network to solve the problem of low prediction accuracy of the BP neural network. Secondly, the beetle antennae search (BAS) algorithm is integrated into BOA, replacing the local optimization process of the butterfly algorithm with the beetle antennae search algorithm to reduce the spatial complexity of the BOA, solve the problem that the BOA is prone to fall into local minima, and create an improved BOA-BP neural network for accurate antenna modeling. The design example shows that the prediction accuracy of the network reaches 99. 60%, and the prediction error is reduced by 47% and 40. 9% compared with the traditional BPNN and the BPNN optimized by the unimproved butterfly algorithm, respectively. In addition, the running time of the improved BOA algorithm is reduced by 80. 86% and 82. 79% compared with the particle swarm algorithm and the genetic algorithm, which greatly reduces the running time cost of the network. In summary, the modeling accuracy and speed of the improved BOA-optimized BPNN are improved, which verifies the feasibility and effectiveness of the improved butterfly algorithm as a novel neural network optimization strategy.

Abstract:Aiming at the cordless direct drive operation of linear motor vertical lifting system (LMVHS), there is an urgent need for a new type of linear motor with high fault tolerance and high thrust density, a five-phase U-shaped permanent magnet salient pole linear motor (FU-PMSPLM) is proposed. The primary set five-phase winding to improve the fault-tolerant performance, and the secondary permanent magnets (PMs) with U-shaped structure can make full use of the PMs. Firstly, from the perspective of reducing magnetic flux leakage, the evolution mechanism of U-shaped magnetic pole structure is analyzed, and the main structural parameters of the motor are given. Secondly, the finite element method (FEM) is used to compare and analyze the electromagnetic characteristics of the U-shaped and Halbach structure of the motor secondary permanent magnet. Based on the principle of constant magnetomotive force, by taking the phase A fault as an example, the fault tolerance performance of the three-phase U-shaped permanent magnet salient pole linear motor (TU-PMSPLM). Finally, the experimental prototype is made and the experimental verification is carried out, the experimental results are basically consistent with the simulation results, which verifies the rationality and feasibility of the proposed motor. The results show that when the amount of permanent magnet is equal, the U-shaped structure of the secondary permanent magnet is compared with the Halbach structure, the amplitude of the air gap flux density increases by 13. 89%, the average thrust increases by 7. 54% and the thrust fluctuation decreases by 25. 07%. FU-PMSPLM has better phase-deficient operation ability than TU-PMSPLM.

Abstract:Under the action of a single wind direction, more contamination accumulates on the leeward side of the insulator in service, which reduces its insulation performance. The movement of charged pollution particle was simulated by establishing a three-dimensional coupling model, revealing that the backflow and eddy flow on the leeward side of the insulator are the main reasons for pollution accretion. The manual coating method was used to simulate non-uniform pollution on the windward / leeward side, and AC flashover tests were conducted to investigate the effects of average salt density (SDD), salt density ratio (J) between the windward and leeward sides, and leeward area ratio (R) on the average flashover voltage (Uf ). The results show that Uf decreases with the increase of SDD, and increases with increasing J and R. Moreover, J has a stronger impact on Uf under the small values of SDD and R. While R imposes a greater effect on Uf when SDD and J are low. By observing the flashover process of polluted insulator, it was found that the arc develops randomly when uniformly polluted, while the arc always propagates along the leeward side under non-uniform pollution, and the amplitude of the leakage current is larger during the voltage increment.

Abstract:In the context of measurement while drilling (MWD), there is a significant issue related to the accuracy of magnetic azimuth calculations due to substantial errors in the measurement of geomagnetic data using micro-electro-mechanical systems ( MEMS ) magnetometers,improved skill optimization (ISO) error parameter estimation method for magnetometer was proposed. Firstly, a multiparameter error model is established according to the output characteristics of the magnetometer, and according to the modulus relationship between the local geomagnetic vector and the real output vector of the magnetometer, the nonlinear error objective function is constructed by the principle of minimization, and the dot product value between the gravity vector and the magnetic field vector is a fixed value as the constraint function, and the SO algorithm is used for optimization. Due to the difficulty of estimating the magnetic error sources, the ISO algorithm is proposed on the basis of SO, and the Tent chaotic inverse learning initialization strategy is used to improve the randomness of the initial population, retain the optimal solution and increase the diversity of the solution space of the magnetic error parameters. The adaptive skill intensity factor is introduced, and the skill cross-avoidance local optimal between members is increased, the skill step size of the ISO magnetic error parameter optimization is improved, the algorithm running time is reduced, the global search ability is optimized, and the error compensation accuracy of the magnetometer is improved. Finally, the compensation performance of the ISO method is verified by turntable experiments and simulated drilling experiments, and the experimental results show that the proposed algorithm has a significant optimization effect on the error parameters of the magnetometer, and the error range of the calculated geomagnetic modulus is reduced to ±0. 2 μT, and the mean absolute error of azimuth is reduced to 2. 1°. It is shown that the output error of the magnetometer is significantly reduced after parameter optimization, and the method can effectively improve the accuracy of MEMS magnetometer measurement, obtain reliable azimuth angles, and verify the effectiveness of ISO.

Abstract:Screw rotors are mainly used in compressors, screw pumps and other equipment, and their surface quality plays a key role in service performance and service life. Process parameters are one of the main factors affecting the surface roughness of screw rotors. In order to explore the influence of process parameters on the surface quality of helical surface milling, a rotor milling experiment was designed to obtain prediction and experimental comparison samples. The improved northern goshawk search algorithm (INGO) is used to optimize the initial weights and thresholds of the BP neural network, so as to improve the prediction accuracy of the surface roughness of the milled multi-head screw rotor. Experimental results verify the prediction accuracy of the proposed algorithm. The results show that the proposed prediction model outperforms GRU neural network and CNN-GRU neural network models in terms of average training accuracy and prediction accuracy. The average training accuracy and prediction accuracy are about 94. 502% and 95. 523% respectively. Therefore, the proposed algorithm has high prediction accuracy and can provide a theoretical basis for reasonable selection of processing parameters of screw rotor milling.

Abstract:The conventional pulse injection method for low-speed operation of switched reluctance motors utilizes only the information of the non-conducting phase currents, which has the problems of limited conduction intervals and continuation currents affecting the estimation accuracy. In this paper, the low-speed control strategy and the principle of position-sensorless detection are investigated, and a three-phase current slope difference position-sensorless detection scheme with dual current chopper-limited PWM hysteresis loop control is proposed. The proposed scheme adopts dual chopper-limited PWM hysteresis loop control in the conduction zone, which increases the number of current chopping in the conduction zone, improves the performance during low-speed operation, and increases the calculation accuracy of the current slope difference. Compared with the traditional non-conducting phase current comparison method that utilizes only the non-conducting phase information, this scheme adds the on-phase current calculation. After calculating the on-phase current slope difference, the three-phase current slope difference is formed with the two non-conducting phases to estimate the real-time position information of the motor. The related simulation and experimental verification are carried out with a three-phase 12 / 8 structure motor, and the experimental results show that the scheme can effectively solve the problems of fixed on-phase intervals and real-time angle calculation affected by the continuation current that exist in the traditional method.

Abstract:In order to solve the problem of low prediction accuracy caused by insufficient data mining in the prediction task of remaining useful life (RUL) of rolling bearings in industrial production, a multi-channel fusion method for predicting the remaining life of rolling bearings was proposed. In this method, the original vibration signal is denoised and feature enhanced by complementary ensemble empirical mode decomposition (CEEMD) is taken as input. A three-channel network model was constructed, and three different neural networks were introduced: Temporal convolutional networks (TCN), convolutional long short-term memory network (ConvLSTM), and bidirectional gated recurrent unit neural network (Bi-GRU), which differentially extracts features from multiple dimensions such as time series, space, and receptive field. The multi-head attention mechanism (MA) is added on the basis of the structure to readjust the output weight of the network and accelerate the convergence speed of the model. Finally, a feature fusion output module was designed to predict the remaining life of rolling bearings. Experimental verification was carried out on two datasets and compared with the advanced models in other literatures. The results show that the proposed model can capture the bearing life degradation curve more accurately, and is better than the comparison model in a variety of evaluation indicators.

Abstract:A hierarchical equalization circuit is proposed to solve the problem of partial over charge and over discharge of lithium-ion battery, due to the inconsistency of single battery in series group. The hierarchical equalization circuit takes the state of charge (SOC) value of single battery as equalization variable, divides battery pack into three groups, and carries out equalization based on Buck converter within the group; Buck-Boost circuit was built between each group to realize bidirectional flow of electricity between groups and balance between groups. The combination of intra-group balancing and inter-group balancing improves the balancing efficiency. The simulation results built on MATLAB/ Simulink platform show that compared with the equalization circuit based on Buck converter and Buck-Boost converter, the hierarchical equalization circuit can operate under static, charging and discharging conditions, compared with Buck-based equalization circuits, the equalization time is reduced by 21%, 18% and 30%, respectively, and compared with Buck-Boost equalization circuits, the equalization time is reduced by 17%, 29% and 15%, respectively. The equalization experiment of the battery pack shows that the circuit can improve the consistency of these single batteries, control the SOC value range of the battery pack within 0. 1%, solve the problem of overcharge and over discharge of partial batteries, and improve the equalization efficiency.

Abstract:With the continuous expansion of the power grid scale, the power grid monitoring data presents a trend of diversification, highspeed and massive quantities. Consequently, the monitoring work of the power grid has become increasingly complex and challenging. In order to solve the problems of low efficiency and poor synchronization of high-dimensional data processing in power system by traditional methods, a monitoring method based on random matrix theory is being investigated to extract monitoring data feature values and detect power grid anomalies. Firstly, disturbance situations within the power grid are designed, and a matrix window is constructed to select monitoring signals within a time series, resulting in a high-dimensional matrix. Secondly, Marchenko-Pastur law and single-ring law are applied for matrix transformation, and feature values are extracted based on the distribution of these values to judge the system's state. Then, linear statistics based on feature values are used to construct various evaluation indicators, including maximum eigenvalue of sample covariance matrix, energy with minimum eigenvalue, maximum-minimum eigenvalue, and mean spectral radius. Finally, the performance of each statistical indicator during short-circuit faults, open-circuit faults, and fault clearance in the power grid is compared to achieve power grid state recognition, anomaly event detection, and power grid stability assessment. The results show that these indicators can accurately determine whether the system has anomalies, detect the start and end time of anomalies, and evaluate the stability of the power grid. The method proposed by this paper can detect disturbances like open circuits and short circuits, and achieve synchronized processing of global monitoring data with low computational complexity and high efficiency, making it suitable for detecting anomalies in large-scale power grids.