Abstract:To meet the application requirements of multi-antenna beamforming in 5G mobile communications, this paper designs and implements a multi-channel Software Defined Radio (SDR) communication system, and based on this system, achieves broadband multi-channel multi-frequency point synchronization calibration. To address the issue of multi-channel synchronization drift due to environmental factors such as temperature, this paper innovatively proposes a real-time synchronization scheme based on out-of-band calibration signals. This scheme inserts calibration signals into the redundant bandwidth of OFDM communications to track the drift of multi-channel response errors over time and compensates for the in-band effective signals. At the 1 GHz frequency point, the system effectively compensates for the phase response error drift of 2.8 ℃ and amplitude response error drift of 0.2 dB caused by temperature variations through real-time calibration. After initialization calibration and real-time calibration, the phase error is controlled within 0.4 ℃, and the amplitude response error is controlled within 0.05 dB in the temperature range of 40 to 80 ℃. This scheme not only achieves higher precision but is also completely transparent to the end user, allowing calibration without interrupting communication operations, making it highly significant for the design of 5G multi-channel synchronization.

Abstract:With the rapid development of the integrated circuit industry, higher requirements are put forward for integrated circuit testing, and the Precision Measurement Unit is the core unit for integrated circuit DC parameter testing. A PMU circuit for integrated circuit testing is designed in this paper, which uses a Field-Programmable Gate Array to control the DAC module to apply voltage excitation. The excitation signal after the PI regulator and power amplification is applied to the Device Under Test through the resistor matching network, Then, the ADC module reads back the test response data to realize the parameter testing functions such as applying voltage to measure current and applying current to measure voltage. The designed PMU circuit has the advantages of wide test range and high measurement accuracy, with an applied or measured voltage range of -10 V to +15 V and a maximum current of ±1.838 A. The system performance in different test modes was calibrated and functionally verified with high-precision resistive loads, and the experimental results show that the system calibrated test error is better than 0.05%, which is able to meet the requirements of DC parameter testing of general-purpose integrated circuits.

Abstract:Using the accurate parameters of the heating system has guiding significance for monitoring system status and identifying abnormal conditions. However, a large amount of terminal data may have distortion problems. To address this, this paper proposed a method for detecting and cleaning abnormal data. Signal modal decomposition combined with deep learning was used to construct a detection and cleaning model. The first step involves conducting CEEMDAN mode decomposition of the heating load obtained by DeST. Subsequently, the intrinsic mode functions and residual quantities generated from the decomposition are input into the CNN-LSTM deep learning prediction model to achieve high-precision prediction results. Finally, based on the deviation between predicted values and data to be cleaned, abnormal detection and data cleaning are completed. The CEEEMDAN-CNN-LSTM combined model in this paper achieves superior accuracy and F1 scores of 91.36% and 86.21%, respectively, outperforming the other three models. Moreover, the predicted values can be used to replace abnormal values, ensuring the integrity and accuracy of the final data set.

Abstract:Aiming at the problem that the dynamic performance of the fuel cell air supply system is susceptible to load changes and external environmental factors, a tandem-type control strategy combining super-helical sliding mode control and terminal sliding mode control is designed. A control-oriented fifth-order dynamic model is established, and the control problem of tracking the optimal oxygen excess ratio, maximum net output power and cathode pressure measurement of the air supply system during load change is proposed; the optimal expectation value is extracted and the controller is designed, and the closed-loop stability is verified using the Lyapunov method. Simulation analysis shows that the constructed observer perturbation estimate is within 0.01% error from the theoretical actual value. Compared with PID control, the response time of tracking the optimal expected value of the oxygen excess ratio is improved by 6.9%, the maximum net power output is increased by 0.2%, and the response time of cathode pressure is improved by 60%. From the results, it can be concluded that the strategy of this paper can effectively control the oxygen excess ratio of the gas supply system to track the optimal desired value and output the maximum net power when the load changes, can accurately and rapidly estimate the cathode pressure, can better estimate the perturbation, and has a strong anti-disturbance ability.

Abstract:With the continuous advancement of artificial intelligence technology, the scale of data in neural networks is gradually expanding, leading to a rapid increase in computational complexity. In order to reduce the computational load of SpatioTemporal Graph Convolutional Neural Networks (ST-GCN), decrease hardware resource consumption, and improve processing speed in practical applications of human skeleton recognition systems, a hardware accelerator based on ST-GCN was designed and developed using Field Programmable Gate Arrays (FPGA). By optimizing the structure of the original network model and quantifying the data, the computational load of FPGA implementation is reduced by about 75%. Based on the sparsity of adjacency matrix, an optimization method for multiplicative and additive operation of sparsity matrix is proposed, which reduces the multiplier resource consumption by about 60%. Experimental validation on human skeleton recognition demonstrated that compared to CPUs, FPGA-accelerated ST-GCN units achieved a speedup of 30.53 at a clock frequency of 100 MHz. The FPGA acceleration for human skeleton recognition achieved a speedup of 6.86.

Abstract:Aiming at the problems of low sensitivity and many local extreme points of the clarity evaluation function during the focusing process of the photoelectric imaging system. This paper proposes an image clarity evaluation algorithm based on regional weighting. First, the algorithm adopts a threshold based on the traditional Laplacian evaluation function to improve the anti-noise and the ratio of clarity. Then, the algorithm also uses the image gradient map to calculate the regional clarity weighting factor which can optimize the variance of flat part of focusing curve. Experimental results show that compared with the most traditional clarity evaluation functions, the clarity ratio of this algorithm is increased by about 2.7 times、the sensitivity increases by about 1.9 times and the variance of flat part of focusing curve can be reduced to 1/6 of the traditional Laplacian evaluation function.In general, this algorithm has the advantages of high clarity ratio and sensitivity、low variance of flat part of focusing curve and has better evaluation performance when the image content is complex.

Abstract:The modulation recognition problem of subcarriers in the universal filter multi-carrier (UFMC) system for non-cooperative communication needs to be addressed. Therefore, a modulation recognition algorithm based on feature fusion is proposed for the UFMC system. Firstly, the receiver signal of the UFMC system is obtained and input features such as in-phase and quadrature sequence and amplitude phase sequence are extracted. Subsequently, a neural network module is constructed by connecting a convolutional neural network with a long short-term memory network in series, while also incorporating a gated recurrent unit in parallel. Finally, modulation recognition of UFMC system subcarriers is performed. The experimental results demonstrate that the constructed neural network effectively identifies five signals (BPSK, 4QAM, 8QAM, 16QAM, 64QAM) with a recognition accuracy reaching 100% when signal to noise ratio greater than or equal to 6 dB.

Abstract:Aiming at the problem that visual SLAM is easily affected by moving objects such as vehicles and pedestrians in real environment, resulting in low pose estimation accuracy, a dynamic visual SLAM algorithm combined with lightweight YOLOv5s is proposed. The improved lightweight YOLOv5s is used as an object detection algorithm to judge moving objects. Combined with the proposed method of dynamic feature points elimination, dynamic feature points are eliminated, and only static feature points are used for pose estimation and map tracking. Experiments on TUM data set show that compared with ORB-SLAM3 algorithm, the pose estimation accuracy of the improved algorithm on high dynamic sequence is improved by 89.29%, 65.34% and 94.42% respectively. The results show that the improved algorithm can effectively eliminate dynamic feature points. The pose estimation and positioning accuracy of visual SLAM algorithm in dynamic environment are improved.

Abstract:An improved pedestrian dead reckoning algorithm is proposed to solve the problem of inaccurate pedestrian location in indoor space. In the step detection stage, a three-threshold peak detection method based on motion segmentation is proposed to realize the accurate step detection of pedestrians in different motion states. Step size estimation is achieved by using an improved Weinberg model. A heading Angle correction algorithm based on the principal direction assumption is proposed to realize the pedestrian′s heading Angle correction. Finally, the information of step number, step length and course Angle is integrated to realize the dead reckoning of indoor pedestrians. Experimental results show that the improved Pedestrian Dead Reckoning algorithm has good indoor stability, and the average indoor location error <5%, which is 9.53% lower than that of traditional PDR algorithm, and improves the indoor location accuracy of pedestrians.

Abstract:The speed control system of permanent magnet synchronous motor with the traditional proportional integral dual closedloop control structure has drawbacks on antiinterference performance and response speed. Therefore, under the research of PMSM mathematical model and analysis of traditional PI parameter design methods, this article proposes a PMSM speed control optimization strategy based on the fractional order proportional integral with improved sparrow search algorithm. By introducing good point set initialization strategy and golden sine update strategy into conventional SSA, the problem of SSA easily falling into local optima is solved, which improves the convergence speed and accuracy, leading to better response performance of the PMSM speed control system. Finally, the MATLAB simulation results show that the ISSA-FOPI speed regulation improvement strategy has the shortest adjustment time of 1.2, 0.9, 9.9 and 12.1 ms respectively under the four working conditions, and the minimum overshooting is 0.16%, 0.16%, 0.05% and 0.6% respectively. This fully embodies the effectiveness and superiority of the strategy.

Abstract:Aiming at the contradiction between the large position tracking error in the traditional backstepping sliding mode position servo system and the contradiction between the dynamic response and the chattering suppression when the fixed sliding mode surface is adopted, a composite control strategy based on disturbance observer and backstepping time-varying sliding mode is designed. Firstly, a state variable is introduced into the traditional exponential reaching law and an adaptive law is designed. A new variable exponential reaching law is proposed to improve the response performance of the system. In order to further improve the convergence speed of the system in each stage, a time-varying sliding mode surface with online optimization is designed by using the idea of ergodic optimization in predictive control, and the stability is proved. Finally, a nonlinear load disturbance observer is designed to estimate and compensate the disturbance of the system. The experimental data show that compared with the traditional backstepping sliding mode control, the steady-state error of the composite control strategy proposed in this paper only accounts for 9/17 of the traditional method, the dynamic response speed is improved by 30 ms, and the antiinterference performance is obviously superior.

Abstract:In order to improve the measurement accuracy of dynamic weighing and realize realtime monitoring and fine management of intelligent pasture, a dynamic weighing algorithm based on chaotic sparrow search algorithm to optimize LSTM neural network is proposed. The data is collected by the dynamic weighing platform, and the Kalman filter algorithm is used to process the interference data. The CSSA-LSTM neural network model is established by using the Tent mapping strategy and the sparrow search algorithm after Gaussian mutation to optimize the parameters of the LSTM neural network. The results show that the average absolute percentage error of CSSA-LSTM neural network is within 1.5%, the average absolute error is reduced by 0.874, and the root mean square error is reduced by 1.115 3. The comparative experiments show that the hybrid algorithm has the smallest prediction error and effectively improves the measurement accuracy of dynamic weighing.

Abstract:In order to enhance the convergence speed and accuracy of UAV path planning in complex environments, and to avoid falling into local optima, a novel three-dimensional UAV path planning method based on the improved spider wasp optimizer algorithm is proposed. This paper introduces an adaptive t-distribution disturbance mutation and cubic mapping strategy for updating the search stage positions within the traditional SWO algorithm, which helps to prevent premature convergence to local optima. Furthermore, a periodically random amplitude dynamic adjustment for the pursuit and escape phases is incorporated to assist the algorithm in escaping local optima. The spiral update mechanism and Levy flight strategy are combined to enhance the global optimization capability of the algorithm, thereby improving its convergence precision. Finally, the performance of the ISWO algorithm is validated through experiments on eight test functions, and simulation results indicate that the execution time of the ISWO algorithm in complex terrain environments is reduced by 26.86% compared to the traditional SWO algorithm, and by 13.80% to 28.27% compared to other algorithms such as CPO, COA, GOOSE, PSO, and GWO. Additionally, the minimum fitness value achieved by the ISWO algorithm is 49.76% lower than that of the traditional SWO algorithm, and at least 27.73% lower than that of other algorithms. Consequently, it is concluded that the proposed improved algorithm can efficiently obtain a shorter and safer path in complex terrain environments.

Abstract:To enhance lithium-ion battery remaining useful life (RUL) prediction accuracy, we propose an integrated model using hybrid scale health factors. We address challenges of noisy data, limited quantity, and incomplete capture of nonlinear characteristics. Firstly, we use singular value decomposition (SVD) to process capacitance signals, optimizing variational mode decomposition (VMD) for denoising and reconstructing the direct health factor, SR. We introduce an amplitude-phase perturbation (APP) data augmentation method to generate artificially labeled data, ESR, based on changes in SR data distribution. Combined with three indirect health factors, selected using GRA algorithm, we establish a comprehensive mixed-scale life characteristic information. Additionally, we improve Transformer model′s decoder structure with LSTM and optimize key hyperparameters using Optuna framework. Experimental results on NASA data show RMSE within 2.39% and MAE within 1.59%, with improved stability and narrower 95% confidence intervals compared to RNN, LSTM, Transformer, and existing models.

Abstract:To address issues of high memory usage, computational complexity, and inadequate detection speed in defect detection algorithms for complex scenarios, this paper proposes a lightweight forged defect detection algorithm based on YOLOv8. First, magnetic particle inspection images from the production line of heavy truck steering knuckles were collected to construct a forged surface crack dataset. Then, a lightweight convolution module, GSConvns, was introduced to enhance feature interaction and reduce computational load. The Shape-IOU loss function was employed to optimize training performance. Finally, the LAMP pruning strategy was used to remove unnecessary weight parameters, reducing model size and increasing detection speed. Experimental results show that the model achieves a mAP of 83.8%, with parameter and computational reductions of 85.05% and 80.25%, respectively. Detection speed improved from 38.7 FPS to 65.6 FPS, significantly outperforming other mainstream algorithms, making it more suitable for real-time detection. The algorithm′s generalization capability was further verified on a public dataset, with the unpruned improved algorithm′s mAP value increasing by 2.0% compared to the baseline. In summary, this algorithm significantly enhances detection speed and resource efficiency without substantially compromising detection accuracy.

Abstract:Aiming at the problem of small target size and huge weight file difficult to deploy in PCB defect detection, an improved YOLOv8 small target defect detection method is proposed. The method incorporates the SE attention mechanism into C2f, which enables the network to assign different weights to different locations in the image based on the information in the channel domain to obtain more important feature information; introduces Basic RFB in SPPF to enhance the network sensing field and improve the feature extraction capability of the network; adds a new small target detection scale to improve the model′s ability to detect tiny defects; discards the large target detection scale to reduce the computational load and shrink the weight file. The experimental results show that the improved YOLOv8 improves the average accuracy by 2.6%, shrinks the weight file by 27.3%, and achieves an FPS of 34.4 ms/frame over the original algorithm in the publicly available PCB defective dataset.

Abstract:When measuring the diameter of circular parts under the illumination methods of ring light source and strip light source, the problems are usually produced, such as the chamfer features of parts are prone to wide edges in the images, the shadows on the boundaries of circles are appeared due to the influence of thickness, the efficiency of image processing is affected by the surface textures and scratches, and the high-resolution panoramic image of large-sized circular parts is acquired multiple times due to insufficient camera field of view. In this paper, a measurement method for large-size circular parts size based on improved SURF image stitching is proposed, and the high-precision measurement of parts size is realized by illumination optimization, image stitching and sub-pixel edge detection. Firstly, the illumination methods of the strip light source arranged at 45°, the combination of ring light source arranged vertically downward and strip light source arranged at 45° are proposed respectively, thus, the wide edges, shadows, textures and scratches in the images of parts are eliminated, which influence the measurement of parts size. Secondly, the SURF feature matching method is improved to locate overlapping regions of stitching images for coarse matching of feature points. Thirdly, the RANSAC algorithm is proposed for accurate matching of feature points, the image registration method is improved to expand the stitching images of parts to the same size and complement the missing background areas. Fourthly, the weighted average fusion algorithm is proposed to smooth the stitched image, and the high-resolution parts panoramic image is obtained. Finally, the least squares fitting of circles is improved to fit the edge circle in the image, and the actual diameter measurement value of circular parts is obtained through pixel size conversion. The experimental results show that the proposed method is more accurate than the traditional visual measurement method, and the relative error with the CMM measurement reference value is less than 0.044 4%.

Abstract:As one of the most common diseases of roads, the timely and accurate identification and localization of cracks is of great significance to the maintenance and continuous healthy operation of roads. However, the detection of pavement cracks is easily affected by complex factors such as road illumination, road shadows, and road environment, which leads to low segmentation accuracy of pavement cracks and prone to fracture and other problems. In order to realize fast and accurate semantic segmentation of pavement crack images, this paper proposes a pavement crack segmentation model based on pixel intensity order transform (PIOT) and UNetFormer. Firstly, the PIOT algorithm is used to preprocess the crack images, and according to the intensity order between each pixel and its neighboring pixels, the image is converted into a four-channel image with higher contrast along the four directions of the diagonal, which retains the intrinsic features of the crack curve structure and effectively enhances the contrast between the cracks and the background pixels. Then, based on the structural characteristics of UNet and Transformer networks, the high-precision segmentation of pavement cracks is accomplished by constructing the UNetFormer segmentation model, in which the global-local attention mechanism is designed and invoked to fully capture the pavement crack feature information. Finally, three open-source datasets, CFD, Crack200 and Crack500, are used for example validation, and the experimental results show that the F1-score of the crack segmentation model proposed in this paper reaches 83.4%, 82.6%, and 81.9%, respectively, and the model parameter is only 37.7% of that of the UNet network model, which provides higher segmentation accuracy compared to the existing crack segmentation model and stronger generalization ability than the existing crack segmentation models.

Abstract:In order to solve the problems of foreground occlusion, background noise interference, and low contrast in industrial CT images, an image enhancement method based on the improved whale optimization algorithm is proposed. First, tent mapping is used to initialize the population of the whale optimization algorithm, and the global search capability is improved by making the whale population more widely distributed. L-vy flight is then used to update the position of individual whales to further improve the local search capability of the algorithm. Finally, the improved whale optimization method is applied to find the best parameters of CLAHE and achieve the adaptive optimal enhancement of the image. Experimental results show that compared with several popular image enhancement algorithms, the method proposed in this article can better enhance industrial CT images, significantly improve image quality, and retain more detailed information.

Abstract:In order to meet the needs of remote time users for standard time UTC (NTSC) nanosecond distribution, a standard time remote reproduction system has been established. The single-reference terminal service model currently used in the system has certain risks. If the single receiver fails, it will affect the accuracy of the resulting deviation between local time and satellite clock (i.e., the time difference between satellite time and local time).To solve the above problems, this paper proposes a method based on robust estimation for GNSS common-view multi-reference stations to fuse the time difference between satellite time and local time. In this method, the median absolute deviation (MAD) is used to detect the outliers of the time difference between satellite time and local time data, and the robust estimation method of IGG III. equivalent weight function is used to fuse the above data of multiple stations, to output a set of stable reference station reference values. Comparing the outlier detection performance of the MAD method and the 3.Sigma method through the Matlab simulation, the MAD method is more applicable for the time difference between satellite time and local time data of the reference stations. At the same time, the fusion method based on robust estimation is compared with the equal-weight fusion method, the standard deviation of the fusion data of the former is 0.11~6.65 ns lower than that of the latter, which improves the stability of the system.

Abstract:This study focuses on the development of an airborne wireless flexible distributed testing system to address the challenges of large space occupation, high cable usage, complex wiring, and numerous testing channels in flight experiments. The distributed system consists of nodes distributed at different locations, and the internal clocks of these nodes may become unsynchronized due to various factors, which can affect the accuracy of data acquisition. By employing a flexible time slot allocation mechanism, the AUTBUS bus can be dynamically adjusted to ensure efficient and stable communication within the system. The synchronization of all network node devices is achieved through the utilization of 1PPS, TOD, and AUTBUS bus-based timing algorithms, ensuring synchronized operations among all nodes based on the same time reference. The synchronous data acquisition of the testing data is accomplished through an externally triggered analog-to-digital conversion module, enabling the synchronized data collection of all nodes during coordinated events across the network. The test results demonstrate that the system achieves a synchronization data acquisition accuracy of 822.5 ns, meeting the data acquisition requirements for airborne testing in flight experiments.

Abstract:Meters can accurately reflect the operation status of substation electrical equipment, to overcome the manual inspection caused by meter leakage and misreading and so on, put forward a PAT-Unet neural network based on the number of intelligent reading algorithms for the leakage current meter display intelligent reading. Firstly, the PAT-Unet neural network is designed to segment the pointer and dense scale of the meter. The feature aggregation module and residual feature dispersion module are constructed in the coding layer of the network to enhance the feature extraction ability. Design the transformer feature concentration module for deep semantic information fusion to enhance the segmentation accuracy of acceptable targets; introduce the pyramid slicing attention mechanism to strengthen the information interaction between the network coding and decoding layers. The information interaction between the coding layer and the decoding layer of the network is enhanced. Ability. Combine the contour detection algorithm and the minimum outer rectangle algorithm to calculate the key point of the scale segmentation results, and use perspective transformation to complete the correction of tilted meters; then use the K-means clustering algorithm to locate the center of the leakage current meter; finally, according to the center of the metering circle, use the polar coordinate transformation to expand the sector dial into a rectangle, and get the number of the leakage current meter by calculating the distance between the zero scale, the pointer and the maximal scale respectively. The leakage current meter is obtained by calculating the distance relationship between the zero scale and pointer and the maximum scale respectively. Experiments have demonstrated that the proposed algorithm can correct the tilted dashboard and provide intelligent readings of the leakage current meter representation while ensuring the accuracy of the readings.