Abstract:Aiming at the problem of high consumption of rare earth permanent magnet materials in traditional rare earth permanent magnet motor, this paper proposes a rare earth permanent magnet synchronous motor with alternating pole and local Halbach structure composed of NdFeB and ferrite hybrid permanent magnets, and analyzes the structural characteristics of the hybrid permanent magnet motor. The electromagnetic performance of the motor and the traditional one-shaped Nd-Fe-B rare earth permanent magnet motor is compared and analyzed, and the local demagnetization finite element model of the motor is established according to the demagnetization characteristics. The multi-operating conditions of the motor are simulated by magnetic circuit coupling co-simulation. Aiming at the performance loss of local demagnetization of the motor under high-speed field weakening and variable load operating conditions, the anti-demagnetization optimization design scheme of the motor is given to improve the working point of the permanent magnet and reduce the demagnetization rate. The simulation results verify the rationality of the proposed motor structure, and the anti-demagnetization performance of the motor is improved by 10.3% compared with that before optimization.

Abstract:The completeness of key parts is an important verification requirement for gas meters. Although the traditional image feature matching method is used to realize the automation of part detection, its universality is poor. This paper proposes an improved method for Faster R-CNN to identify and locate key parts of gas meters from multiple perspectives. First, Faster R-CNN utilizes Vision Transformer (ViT) to replace the convolutional neural networks, whose self-attention mechanism can help to learn the correlation between image block features and strengthen the representation ability. And then the ViT structure with 14 Transformer layers and 12 self-attention heads is optimized to achieve optimal accuracy. Experimental results show that the mAP of the optimal model is 86.71%, 2.48% higher than that of ResNet50. It is equivalent to the detection accuracy of ResNet101, whose detection efficiency is increased by 5.8%, and effectively reduces the complexity of the model. It takes 1.13 s to accomplish the single detection of key parts of gas meter. The method balances the accuracy and real-time ability for key parts detection of gas meter.

Abstract:In the face of gradient operation conditions, the development of control strategies that simultaneously take into account intervehicle cooperative control and energy economy is one of the key technologies for improving traffic efficiency and exploiting the energysaving potential of vehicles. A hierarchical optimization control strategy based on improved particle swarm optimization algorithm and Q-learning for fuel cell hybrid electric vehicles queue is proposed with the objective of safe driving and optimizing energy consumption. In this strategy, the upper layer controller uses the improved particle swarm optimization algorithm to obtain the energy-saving speed trajectory under the premise of ensuring that safety constraints such as distance or speed limit from the preceding vehicle are satisfied, and utilizes the model predictive control framework to adjust the vehicle speed in real time to ensure the vehicle follows the energy-saving speed trajectory. The lower layer controller builds the Q-learning controller based on the information such as vehicle speed and demand power solved by the upper layer to realize the optimal energy distribution between the fuel cell hybrid electric vehicles power cell and the fuel cell. Simulation results show that the hierarchical control strategy proposed in this paper exhibits good tracking performance and safety performance under slope conditions, and the optimization results are similar to the dynamic planning strategy, indicating the energy consumption economy and feasibility of the strategy.

Abstract:The detailed evolution process of the electromagnetic pulse signal, especially the change information of its steep front, is helpful to deeply understand the generation and propagation process of the electromagnetic pulse signal. It has extremely important scientific and practical significance for the development of our national defense and natural science. In this work, a compressed sensing technology of analog signal based on three parallel low-speed analog-to-digital converter (TPL) is proposed. By under-sampling the electromagnetic pulse signal (the sampling frequency of the signal is lower than the requirement of Nyquist sampling theorem), the high-speed acquisition waveform of the measured electromagnetic pulse signal can be obtained and recovered. Based on the application of TPL compressed sensing technology, the requirement of the measured electromagnetic pulse signal on the sampling rate of the back-end electronic ADC can be reduced. In this paper, the methods of establishing sparse dictionary, observation matrix and signal reconstruction in the process of TPL implementation are deeply studied, and an atomic number adaptive dictionary construction method based on KSVD is innovatively proposed. On this basis, the compressed sensing effect of TPL system on electromagnetic pulse signal is tested by simulation and experiment, which verifies the feasibility of this method.

Abstract:Ultrasonic guided wave detection has gradually become one of the effective methods for weld defect detection due to its characteristics of high detection efficiency and long detection distance. At present, the signal excitation source of ultrasonic guided wave detection in domestic usually combines signal generator and power amplifier, which is not convenient to operate and the latter is expensive and has limited performance. In order to surmount its shortcomings, an ultrasonic guided wave detection broadband excitation source based on FPGA is designed. The excitation frequency and pulse cycle number of the excitation source can be flexibly adjustable according to different detection objects. The power amplifier of the excitation source can achieve an output amplitude of 100 Vpp in the excitation frequency range of 250 kHz~4.5 MHz on a 50 Ω load, and the piezoelectric transducer can be directly driven. Verilog language is used to design the driver of each hardware and the serial communication protocol with the host computer. The test has verified that the excitation source can achieve defect detection with the receiving and acquisition module on a 500 mm long wield. The module is small in size, reliable in performance, and has certain practical value.

Abstract:Aiming at the problem that it is difficult to accurately measure the high rotational speed of high rotation ammunition, a speed measuring device based on geomagnetic sensor is designed. The measurement device mainly uses HMC1053 three-axis geomagnetic sensor and STM32H753 ARM chip to complete the speed measurement function. According to the characteristics of the output signal of the geomagnetic sensor, the time-domain characteristics of the rotational speed signal are simulated and analyzed by using the zero-crossing detection method, and the time-frequency characteristics of the rotational speed information are also simulated and analyzed by using the short-time Fourier transform and continuous wavelet transform methods. The results show that the continuous wavelet transform method has higher frequency resolution and higher rotational speed precision. Finally, the machine speed test and missile-borne flight test are carried out to verify that the maximum error of the machine speed test is 0.011 3 r/s. The test results show that the speed measurement method is feasible, which provides relevant support for the navigation and guidance technology of high rotation ammunition and has certain engineering application value.

Abstract:The speed control model of ultrasonic motor is the basis of its motion control research. In order to study the ultrasonic motor speed control model with drive frequency as the regulating variable, a second-order linear time-invariant model for ultrasonic motor speed control is established based on the measured data between ultrasonic motor drive frequency and speed. Iterative learning identification is used to recognize the parameters of the motor speed control model. The parameter learning law for iterative learning identification is designed by double-parametric optimality theory for the case of weakened parameter convergence caused by different amounts of data in each set of real measurement data. The results obtained from the iterative learning identification are compared with the Hammerstein model. Simulation and experimental results show that the iterative learning identification can effectively identify the model parameters of the ultrasonic motor. The parameters converge quickly and well, the accuracy of the constructed model is high and the modeling method is effective.

Abstract:It is key issue to accurately simulate the radar echoes of offshore wind turbine and extract micro-Doppler features, and they can solve the reradiation interference of offshore wind farms to neighboring radar stations. Aiming at the problem that the existing algorithms treat directly the sea surface background of wind turbine as a planar good conductor and are too simple. In the meanwhile, the traditional equally spaced scattering point model cannot characterize the complex surface features of the wind turbine blade. Therefore, we introduce a multipath model with a forward complex reflecting coefficient model of rough sea surfaces in radar echoes simulation, also build 3-D scattering point equivalent model of wind turbines. So, a numerical simulation algorithm for radar echoes of offshore wind turbine based on time domain echo electric field is proposed. According to the though of discrete electric field equation of RWG function in the method of MOM, we establish a 3-D scattering point equivalent model of wind turbine. Considering the distortion effect when the electromagnetic wave acts on the sea surface, the time domain return electric field solution equation under a forward complex reflecting coefficient model of rough sea surfaces is derived. We perform vector superposition short-time Fourier variation about radar echoes, and obtain the simulation results of the radar echoes of the offshore wind turbine. By comparing simulation results and experimental results of the scaleddown model, it verifies the correctness of our algorithm. Finally, the micro-Doppler characteristics of the wind turbine echoes under the sea surface background are analyzed in comparison with those under free space,and it prove that the sea surface background cannot be neglected. The influence of sea state about micro-Doppler shift are further analyzed by controlling a single variable. It is，the root mean square of the wave height is negatively correlated with the Doppler shift. These results also provide theoretical reference for the subsequent identification and filtering of offshore wind turbine echoes.

Abstract:Accurate estimation of the state of charge (SOC) of lithium batteries is crucial to improving the dynamic performance and energy utilization of batteries. Aiming at the problems of low accuracy and poor stability of existing neural network SOC estimation methods under complex working conditions, this paper proposes an improved GRU model algorithm to estimate SOC. Firstly, combine 1DCNN and Bi-GRU and add attention mechanism to build 1DCNN-Bi-GRU-ATT model. Secondly, in order to eliminate the phenomenon that the ReLU activation function is prone to dead neurons, it is improved to PReLU activation function. At the same time, in order to solve the problem that MSE-Loss is easily affected by abnormal battery data in complex working conditions and the convergence speed of MAE-Loss is slow, Huber-Loss is used instead as the network loss function. Finally, the Adam algorithm is improved to Nadam algorithm using Nesterov accelerated gradient. The experimental results of lithium battery SOC estimation show that the average values of root mean square error and mean absolute error of the model algorithm under 12 complex operating conditions are 1.181 7% and 0.924 1%, respectively. Compared with the model before improvement and other models, the comprehensive performance of this model in 12 cases is more stable and accurate, and it has higher generalizability.

Abstract:An improved cascaded Taylor compensation algorithm for the time mismatch error in high-speed high-precision TIADC is proposed. Specifically, the linear approximation principle is used to estimate the time mismatch error, then an improved cascaded Taylor compensation structure is used to compensate for error. The error compensation module and error estimation module together form a feedback calibration structure to enable real-time estimation and calibration of the time mismatch error. A 4-channel TIADC system with 16 bit and clock sampling frequency of 500 MHz is established in MATLAB for simulation and verification of the time mismatch error calibration model. The experimental results show that when the input signal frequency is in the whole Nyquist frequency band, and after the 3rd order calibration, the SFDR and SNR of the TIADC system are improved by 56.2 and 55.6 dB on average. Compared to the traditional cascaded Taylor compensation structure, the hardware implementation is further reduced in size.

Abstract:When the transmission tower foundation in landslide area is displaced, the maximum displacement of the tower and the maximum stress of the rod will change. The state prediction model of the tower can be established to obtain the maximum displacement of the tower and the maximum stress of the rod, so as to prevent the occurrence of disaster accidents. Proposes an improved sparrow search algorithm to optimize the prediction model of BP neural network. Firstly, Sin chaotic sequence and the dynamic adjustment strategy of step factor are used to optimize the sparrow search algorithm. Secondly, the optimized model is used to optimize the weights and thresholds of BP neural network to obtain the prediction model. The displacement value of the tower foundation in the direction XYZ is taken as the input of the prediction model, and the maximum displacement value of the tower and the predicted maximum stress value of the tower members are obtained. Compared with the model of BP neural network, the root error RSME value decreased by 63.4%, the average relative error MAPE value decreased by 60.4%, and the absolute mean absolute error MAE value decreased by 62.6%. At the same time, the predicted value of the prediction model in this paper was in line with the changing trend of the real value. In conclusion, the prediction model can accurately predict the operation state of the transmission tower and provide strong guarantee for its safe operation.

Abstract:To accurately identify the elderly′s fall posture and timely carry out the medical intervention, a human fall posture identification method based on multi-modal time series images was proposed. First, the resultant acceleration is decomposed into three sub-sequences by wavelet packet, and then three time series image algorithms are used to transform the resultant acceleration into three three-channel time series images. Then, its high-dimensional features are extracted through ResNet-18, and multimodal feature fusion is used. Finally, the fusion results are combined with the improved random forest algorithm to complete the identification of human fall posture. The accuracy of 98.7% and 99.3% were verified in UMAFall and SisFall. The results show that the method has high accuracy in the identification of human falls, and can provide timely help for elderly people who fall.

Abstract:In the field of engineering vibration measurement, the vibration measurement method based on machine vision is a commonly used technical means in recent years. For small amplitude vibrations, visual measurement methods relying on traditional image processing algorithms are difficult to measure accurately. To solve this problem, a machine vision-based motion magnification method was proposed. It achieves more accurate motion information measurement by amplifying the tiny motion that is difficult to detect by the human eye and combining with the image motion extraction method. First, this paper summarizes the commonly used machine vision-based vibration measurement methods, focusing on the detailed introduction of machine vision-based motion magnification technology. Motion magnification methods include Lagrangian magnification approach and Eulerian magnification approach. The origin of both motion magnification methods is introduced. The two technologies’ feature are analyzed and compared. Then the various improvement methods of Euler magnification techniques are introduced in detail, including the characteristics and limitations. Finally, the development prospects of small amplitude vibrations measurement technology based on Euler motion magnification are discussed.

Abstract:The image quality of locomotive driver′s room video is easily disturbed, especially when the image brightness abnormality caused by external lighting changes, which leads to the decrease of system detection accuracy. To address this problem, this paper proposes an adaptive nonlinear color enhancement algorithm based on side-window filtering for pre-processing, and designs a novel driver behavior recognition system scheme. Using the principal clustering presumption algorithm, an image illumination classification model is established to classify 6A driver′s room video images into three scenes: low illumination, normal illumination and exposure. Then the algorithm proposed in this paper is used to enhance the low-illumination 6A driver′s room video image, which effectively improves the image brightness, contrast and enhances the detail information in dark areas. YOLOv3-based driving behavior detection model is established using a deep learning method. To prove the feasibility of the method, the locomotive 6A video from a railroad bureau′s locomotive depot was selected for experiments on an NVIDIA video analysis server. The results show that the low-light image enhancement algorithm proposed in this paper can better improve the image quality, and the object detection accuracy of the item point reached 97.20%, which was improved by 6.33% compared with before optimization, and meet the actual demand of video intelligent analysis in the locomotive depot of railroad bureau.

Abstract:Aiming at the detection difficulties of small targets in pictures, such as less information and large scale changes, this paper proposes a feature fusion small target dynamic real-time detection model (HCD-YOLOv5s) based on YOLOv5s. In view of the problems that sampling under the model is easy to cause the loss of small target information and insufficient expression of deep network location information, a detection head for detecting small targets is added from the shallow layer; Aiming at the problem of feature confusion caused by feature fusion, this paper designs a feature fusion method CCAT to reduce the loss of location information and semantic information in the detection layer; In view of the inconsistency between the detection task and the activation function adapted to the different data distribution, the DConv module is designed to separate the regression task and the detection task, so as to realize the dynamic detection of the model. In this paper, the Ablation Experiment of the model is carried out on the visdrone data set, and the three modules promote each other. Select pictures with different input sizes to test the speed and accuracy of the model. On the basis of YOLOv5s, the mAP50 of HCD-YOLOv5s is increased by 10.2%, the detection accuracy and parameter quantity are significantly better than YOLOv5m, and the FPS reaches 90. Finally, the experimental verification is carried out on DOTA-v1.0, and the mAP50 and mAP are increased by 1.8% and 2.0% respectively, which proves that the HCD-YOLOv5s proposed in this paper has better performance in small target detection.

Abstract:The PPLCFaster-YOLOv5 model is proposed to address the problems of low accuracy, low recall and poor real-time performance of existing PCB surface defect detection methods. The method uses the modified PPLC-Net as the backbone network and the Focus structure as layer 0 of the network to improve the feature map′s ability to express location information. A channel blending mechanism is introduced within the depth-separable convolutional structure so that the features obtained by each grouped convolution have equal contribution to the global features; a Dropout mechanism is incorporated to limit the imbalance regularisation factor. A low parametric number G4Head feature fusion network structure is proposed to incorporate more shallow information into the feature fusion to improve the model′s ability to locate defects; add residual connections between the backbone network and feature fusion to improve the contribution of backbone network information to feature fusion; and adopt the SIOU loss function to accelerate the convergence of the regression frame. The trained model was deployed using the Flask server framework. Experiments show that the deployed PPLCFaster-YOLOv5 model can detect surface defects on DeepPCB as well as the Peking University PCB surface defect detection dataset in 0.009 s, and the accuracy and recall rates are improved compared with other mainstream models.

Abstract:Digital mural inpainting is an important application of computer vision in the field of image inpainting. Digital mural inpainting model based on improved two-stage generative adversarial network was proposed to solve the problems of ambiguity, structure disorder and detail loss in the process of inpainting. Firstly, a feature optimization fusion strategy is designed in the first=stage generator. The features of different scales in the encoder are optimized and fused in the decoder in proportion to reduce the loss of feature information in the convolution process. Then, in the second-stage generator, the dailated residual module is used instead of the dailated convolution process, and the dailated convolution with small expansion rate is combined with the residual module to increase the receptive filed and reduce the accumulation of holes, which effectively alleviates the repaired grid artifact phenomenon. The experimental results show that the proposed method has obvious advantages in visual effects and tube indexes on the mural dataset compared with other restoration algorithms, in which the peak signal-to-noise ratio is improved by 3~5 dB on average, and the Structural Similarity is improved by 2%~6% on average.

Abstract:Mixer is an indispensable part in electronic communication system, and is the front-end circuit of heterodyne receiver and microwave measuring instrument. Whether it is microwave communication, radar and many microwave measurement systems, the microwave signal must be mixer down to IF for processing. With the development of optical communication, optical mixer in coherent optical communication plays an important role in coherent receiving terminal and optical measurement system. This paper starts with the electric mixer, expounds the principle and development status of the optical mixer, and briefly introduces the work of Xi′an University of Technology in the field of optical mixer, and prospects the development trend and prospects of mixer in the future.

Abstract:The resource scheduling problem under the condition of limited ATDM forward link resources in broadband satellite communication system is studied. Taking the transmission of real-time services preferentially, taking into account the user priority and system throughput as the optimization goal, and the modulation and coding mode, the number of multiframes and the priority as constraints, the resource scheduling objective function is established. An improved ant colony optimization algorithm with the initial solution set construction and enhanced global search as the core is proposed to solve the resource scheduling problem and avoid the slow search speed and local search ability of the traditional ant colony optimization algorithm caused by the lack of pheromone in the initial stage. It is weak and easy to fall into local optimum, which improves the application of the algorithm in the satellite scheduling process with strong real-time and high-efficiency requirements. Simulation results show that the proposed algorithm can accurately obtain the optimal solution, with the accuracy of 99.8%, and its convergence speed is 55.6% higher than that of the traditional algorithm. Compared with the traditional algorithm, the objective function F, comprehensive weight Y, and system throughput of the scheduled service of the proposed algorithm are increased by 8.4%, 6.6%, and 12.1% respectively. It has good accuracy, convergence and optimization performance in resource scheduling, and it is optimized with the same type.

Abstract:With the increasing complexity of FPGA design, physical design requires a large number of optimization iterations to achieve. Cabling congestion affects chip area, delay and other performance indicators, so accurate and rapid prediction and early resolution are required. A FPGA routing congestion prediction model CBAM-CGAN is proposed. The model extracts feature in the layout phase to synthesize learning images, and introduces attention mechanism learning to enhance the importance of each feature channel of the image, so as to improve the routing congestion prediction performance. The experimental results show that the method achieves good results in routing congestion prediction in the layout phase. Compared with the conditional countermeasure generation network model, the average value of structure similarity is increased by 0.89%, the average value of peak signal to noise ratio is increased by 1.37%, the average value of normalized root mean square pixel difference is decreased by 3.8%, the average value of pixel accuracy difference is decreased by 0.06%, and the prediction time of a single image is about 0.1 seconds. Experimental data prove the accuracy and rapidity of the model in FPGA routing congestion.

Abstract:Aiming at the low accuracy of traditional DV-Hop positioning algorithm in wireless sensor network node positioning, this paper proposes an improved DV-Hop algorithm based on ranging correction and Harris hawks optimization algorithm. The algorithm uses multi-communication radius to adjust the minimum hop number of network nodes, optimizes the average hop distance of network nodes by using the minimum mean square error and weight factor, uses the improved Harris hawks algorithm to replace the least square method for position calculation, and introduces Tent chaotic mapping and elite group system. And the sine and cosine optimization strategy is used to avoid the algorithm falling into local optimization prematurely, and the approximate coordinate value of the network node is obtained by solving the optimal solution. The simulation results show that under different conditions, the improved algorithm can have better positioning ability compared with the traditional DV-Hop algorithm and ABCDV-Hop algorithm, the node positioning error is reduced by 20.13% and 7.74% on average, and the positioning accuracy is higher.

Abstract:With the increasing resolution of satellite payload cameras, the amount of image data acquired by them also increases rapidly. How to transfer payload data to the back-end device for high-speed and reliable processing is the current problem to be solved. In this paper, the high-speed SERDES interface chip named TLK2711 and three-way homologous clock working principle are studied and applied, and the transmission error in the high-speed data link of the satellite TLK2711 is analyzed. A high-speed data interface design based on three-way homologous clock is presented, and the specific design of the high-speed data interface is described in detail. Firstly, the disadvantage of the original scheme, that is, the output data signal to TLK2711 by the field programmer without additional reference clock, is analyzed. Based on the original scheme, an improved scheme is proposed. Three-way homologous clocks are added to the original circuit to provide reference clocks for the field programmer and TLK2711. The causes and effects of bit error rate are analyzed in depth, and the optimal phase detection and RS encoding are proposed, and the feasibility of its application in high-speed data transmission interface is verified. The interface design is validated. The experimental results show that the TLK2711 high-speed data transfer interface can achieve up to 2.5 Gbit/s data transmission. Compared with the original scheme, the data clock jitter of TLK2711 high-speed data transfer interface design based on three-way homologous clock is reduced by 59.5%, and the RS encoding error correction capability is strong, which greatly reduces the error rate of CRC, significantly reduces the error rate of hardware implementation, and enhances the working stability of the interface.

Abstract:Aiming at the problem that high frequency signal receiving and storing function in motor bearing monitoring system is easy to lose data and how to realize accurate diagnosis of bearing faults, a motor bearing fault monitoring system is designed and developed by using LabVIEW, Access and MATLAB hybrid programming. Through the producer and consumer structure of LabVIEW, the system realizes the high-speed receiving and real-time saving of vibration signals by TCP/IP communication. Through LabVIEW UDL to achieve Access database add, delete, change, check operation; Aiming at the problem of bearing state pattern recognition, a bearing fault diagnosis method based on Variational Mode Decomposition combined with permutation entropy and Self-Organizing feature Map neural network was proposed. After experimental verification, the receiving speed of high frequency signal of motor bearing fault monitoring system reaches 12.577 KSps, which can realize real-time data access. The average recognition preparation rate of bearing fault diagnosis method based on VMD-PE-SOM neural network proposed in signal analysis function reaches 99.06%. The system integrates the function of vibration signal acquisition and fault diagnosis together, which has the advantages of fast receiving speed, no packet loss, good interaction and high fault recognition rate.

Abstract:A migration learning network model based on the ResNet50 model was studied for accurate and fast classification of the manufactured casings. By constructing a neural network model as well as obtaining sausage casing samples from a cooperative factory and making a total of 2 000 data sets of four grades A, B, C and D according to the actual quality. A new fully connected layer is designed based on the ResNet50 model. And divided into training set and test set in the ratio of 7∶1. Experimentally, it can be seen that the accuracy of migration learning is 99% far better than the accuracy of 94% of the ordinary deep learning model, and the accuracy is significantly improved. Finally, the trained model is made into a user interface using pyqt, a Python graphical tool, for practical application. The migration learning-based intestinal coating quality detection system established in this study can achieve fast and accurate classification of intestinal coating quality, reduce labor cost, and provide a basis for future intestinal coating quality detection.