Abstract:With the increase of switching frequency, the switching loss in LLC resonant converter will increase. Gallium nitride (GaN) device with high frequency characteristics is applied to LLC resonant converter with planar transformer and synchronous rectifier technology to achieve the double improvement of switching frequency and efficiency. Firstly, the working principle of LLC resonant converter was analyzed. Then, the resonant element parameters were designed to reduce the loss. On this basis, two sets of control experiments were established, and the losses in the power tube, transformer and rectifier tube in the two experiments were calculated. Finally, two experimental prototype platforms were built. Relevant experiments show that the efficiency is increased by 11.25%, which verifies the rationality of the design scheme.

Abstract:Firstly, different kinds of power lithium battery pack equalization topologies are compared and analyzed in this paper. Then the energy flow in five cell balancing models is analyzed quantitatively, they are series-based cell-cell, module-based cell-cell, layer-based cell-cell, series-based cell-pack and module-based cell-pack. Finally, according to the topology of module-based cell-cell, the simulation model of eight lithium batteries in series is built. The results show that after 3.2 s, the equalization within the module is completed, after 6.1 s, the equalization among the modules is completed, after 66.5 s, all the eight batteries’ balancing in the battery pack are finished, and their SOC are 100%. It is verified that the module-based cell-cell equalization topology has fast balancing speed and high efficiency.

Abstract:Aiming at the problem that typhoon disaster can easily cause distribution lines to fail, thus affecting the safe and reliable operation level of distribution network, an identification method of vulnerable lines in distribution network under typhoon disaster is proposed. Firstly, considering the local and global network topology, power flow distribution, equipment parameters, external environment and fault loss, four multi-angle evaluation indexes of line vulnerability are put forward from the structural level and equipment level of distribution network. Then, normalizing the index data. Finally, a comprehensive evaluation index is obtained through the analytic hierarchy process and the entropy method, and a multi-angle identification system for vulnerable lines of the distribution network under typhoon disasters is constructed. The simulation results of IEEE RBTS BUS6 F4 feeder system show that the normalized values of degree, intermediate number, failure rate, loss degree and comprehensive vulnerability of the 10th vulnerable line in the system are 0.4733, 0.0860, 0.7970, 0.4207 and 0.4889 respectively. The index values of the top 10 lines are larger than the above values, and their vulnerability is relatively high; The index values of the lines ranked after 10th are smaller than the above values, and their vulnerability is relatively low. The line with high vulnerability is characterized by being located in the upstream position, trunk line, branch or long length of distribution network.

Abstract:In order to solve the problem of low accuracy of existing on-line SOH estimation methods for lithium batteries, In this paper, we propose an improved particle swarm optimization (IPSO) method to optimize the support vector regression (SVR) model. Firstly, the association analysis of the extracted health indicators is carried out, and the heuristic algorithm is used to optimize the hyper parameters of SVR. At the same time, the weights of the extracted multiple health indicators are optimized, and the SVR model with the optimal parameters trained by the optimal data sequence is obtained. Using three battery data sets B5, B6, B7 of the PCoE Research Center of the National Aeronautics and Space Administration to test and analyze the proposed method, the results show that the average absolute percentage error (MAPE) and root mean square error (RMSE) of the three groups of batteries are reduced by 62.3% and 65.5% respectively after the algorithm is optimized. Finally, by comparing with the existing prediction models, it is proved that the proposed method has higher online estimation accuracy.

Abstract:A new type of bidirectional DC-DC converter with a coupled inductor was designed for the output voltage of the fuel cell power generation system and the input current rod wide. The supercapacitor was used to reduce the impact of current ripple to the fuel cell to increase its power generation efficiency, and improved the output voltage gain by changing the number of turns ratios of the duty ratio and the coupled inductor. A simulation model was created in the Matlab/Simulink software, and a control circuit was built with average current mode. The switching status and working characteristics of the converter were analyzed in the boost and buck mode. The simulation results showed that the input current ripple wave of the converter was about 1%, and the boost voltage gain was up to 16 when the coupling inductance ratio was 1. The results showed that the proposed converter could meet the low frequency current ripple requirements of fuel cell power generation system and achieve high voltage gain at the same time, which verified the feasibility of the proposed topology.

Abstract:Aiming at the difficulty of urban parking space sharing, a shared automatic induction parking lock system based on the combination of intelligent hardware and mobile phone WeChat applet was designed. The system consists of three parts: shared parking lock, mobile phone WeChat applet and server parking information management system. The main work flow and function module design of the system are given. Parking lock system, with STM32F103 as the main control chip, using RC522 technology, GPS positioning, wireless communication, display and other modules, realize the positioning and navigation of shared parking lock, pre-booked parking card unlocking, alarm and other functions. WeChat applet uses MySQL database to store information, and uses WeChat developer tools to complete the basic functions of parking space selection, parking space reservation, parking space entry and parking space fee settlement. The parking lock is tested by reading the IC card of the vehicle, and the test results show that the parking lock has realized the designed function. WeChat APP also completed the design function after testing. The whole system can complete the automatic perception of parking space, warehouse management and cost settlement function, complete the unmanned management of parking space, and then improve the utilization efficiency of parking space.

Abstract:The ability to accurately detect metal particles in the electrolyte solution is of great significance to food and drug safety. In the field of food and medicine, metal detection technology is mainly based on the principle of electromagnetic induction. When the electromagnetic method is used for metal detection, the output of free conductive ions under the metal detection system is similar to that of metal particles. When metal particles are in an electrolyte solution, free conductive ions will have a huge impact on the detection of metal particles. This paper studies the detection of tiny metal particles in the electrolyte solution. Using a balanced double-coil metal detection structure, a metal detection system for the detection of tiny metal particles in the production of food and medicine is developed, and the system is used The standard metal sample detection experiment under the electrolyte solution environment was carried out. Through experiments, it is found that: "In the phase component waveform of the output signal after the electrolyte solution and metal particles pass through the detection system at the same time, the effect of the metal particles on the phase component waveform is obvious and stable, but the strength of the effect decreases with the size of the metal particles. The gradual decrease is related to the position of the metal particles in the electrolyte solution. According to the change of the phase component waveform, there is the possibility of distinguishing the metal particles from the electrolyte solution." Due to the relatively complex waveform shape of the phase component and the weaker influence of small volume metals, real-time detection algorithms such as the curve length integral method and the double-sided cumulative area method are used to detect the influence of metal particles in the phase component, and the above-mentioned metal detection Realized on the system.

Abstract:In order to solve the problem that the traditional detection methods are limited to be unable to achieve accurate positioning function in the specific situation of narrow space, combined with the characteristics of ethylene glycol coolant volatilization and ionization by ultraviolet lamps，a sensor based on photoionization is fabricated, an array sensing method is proposed, by creating a closed space in the prone part of leakage, and placing multiple array sensors in the closed space for detection, the array detection platform for coolant leakage was constructed by combining the data processing module and PC. The test results show that the system has broken through the key technology of real-time online monitoring of coolant leakage in narrow space, and the whole system is sensitive to small leakage. In the simulated micro-leakage test with 0.1ml ethylene glycol coolant, the alarm time is kept within 10s after the alarm concentration is set to 3ppm, and the response is timely and the system is strong scalability, which provides a practical detection method for timely troubleshooting coolant faults in complex cooling environment.

Abstract:In order to reasonably and scientifically select the smart meter with the smallest error to supply electricity to customers, a data mining and analysis method for smart meter error based on Gaussian mixture model is designed. First, the basic ideas of Gaussian mixture model and EM algorithm are analyzed. Secondly, the standard deviation of the smart meter error data is calculated as the modeling data, and the error data model of the smart meter based on the Gaussian mixture algorithm is established. Finally, it is combined with the traditional Kmeans. Class algorithm model for comparison test. The experimental results show that compared with other clustering algorithms, the designed method has a larger contour coefficient value and better performance. It can be used to find the smart meter with the smallest error in a large amount of data, and it can give feedback to the smart meter manufacturer on the product, and it also has functions such as product market division.

Abstract:After ocean space survey ship gradually deploys controllable equipment with independent property rights, in order to solve the problem that satellite communication bandwidth resources are limited during missions at sea, and cannot meet the increasing demand for open source software such as maritime measurement and control missions, scientific research and accompanying experiments, a kind of open source operating system software repository, Python software repository and git distributed management repository are designed based on the cloud computing platform. By building the OpenStack platform, the Debian/Ubuntu operating system software repository is built by using apt-mirror, the Fedora/CentOS software repository is also built by using yum-utils, the python software repository is designed by pip2pi, and the git distributed management service is established using gitlab. The result is that the data volume of each operating system is more than 55GB, the Python data volume is 150GB, the gitlab shared data volume has also reached 50GB, and the offline mode transmission rate is greater than 8MB/s on the sea, during the voyage, the open source software repository has received considerable access, among them, gitlab has reached the highest average daily visit volume of 8 times per day. The results show that the open source software repository built through the modular cloud platform is simple to operate, flexible, and extensible, and can meet the requirements for downloading open source software and collaborative work during the long-term seagoing of the survey ship.

Abstract:In order to deal with the energy risks caused by the aging of the satellite biaxial driving solar array and the collision of small space debris, this paper made a comparative analysis of the historical state of the same satellite biaxial driving solar array and the state of different satellite solar array in the same series through the method of horizontal and vertical comparison, and made a preliminary judgment on the damage of the solar array. After the abnormality is determined, the difference between the actual output and theoretical output of the solar cell array is analyzed through the method of split array joint calculation, and the damage situation is specific positioning and quantitative. Finally, the solar cell array output of the satellite with damaged sailboard meets the requirements of the platform and load, as well as the charge and discharge requirements of the battery bank are analyzed. The damaged state analysis and energy assessment of the solar array can be carried out quickly and in detail, which is an important basis for energy management after the damaged satellite solar array.

Abstract:Aiming at problems such as missing and false alarms that may be caused by the complex drawer structure due to its irregular structure, small area for installing sensor modules, and many monitoring nodes. This paper designs a concentration gradient algorithm that accurately locates the leakage point.Through the experimental test, it is proved that the whole monitoring system can work normally in the environment of - 15 ℃ to 55 ℃.The false alarm rate of the system is less than 3% and the missing alarm rate is less than 1%, which verifies the stability and practicability of the whole system.In other similar complex drawer structures with the risk of gas or liquid leakage, such as chemical industry, agriculture, medicine, military, environmental protection and other specific scenarios, this technology meets the needs of a wide range of universal testing.

Abstract:This article expands from the research of randomness testing of signal generated by superlattice random number generator (SRNG). The thesis tests the generated random signal for this new signal using some common machine learning methods to preprocess one part of random signal and try to train clustering or network model, and then testing other part of random number to judge the quality of randomness. These methods are used on normal distributed random number and SRNG signals, comparison of which shows better performance of SRNG signals.

Abstract:In order to realize the control precision and speed curve of stepper motor in real time, the speed control method was studied.Based on the working principle and mathematical model of two-phase hybrid stepper motor, the model of its open-loop control system was built in Simulink, and a real-time parabolic speed curve algorithm was designed and compared with the typical trapezoidal and exponential speed curve algorithm. The simulation results show that the maximum no-step rotation Angle of the parabolic acceleration and deceleration curve is 990° in unit control period, which is at least 22% higher than the other two algorithms. The minimum control period of driving stepper motor is 42ms, which is 22.2% higher than that of exponential type. It is proved that the acceleration and deceleration algorithm has faster response speed and higher control precision, reduces the occurrence rate of out of step and overshot, avoids mechanical flexible impact and prolongs the system life.

Abstract:The technology of measuring water depth and underwater terrain with solid-state laser radar sensor is a new measurement method developed in recent years. The solid-state LiDAR sensor technology, as well as their application status and development prospect of water depth measurement, underwater and land integrated topographical surveying in offshore, reefs and inland rivers/lakes/reservoirs are reviewed and introduced. This paper first introduces the propagation characteristics of light waves in water and reflection characteristics of light waves on water surface, and the bathymetry principle of dual-frequency LiDAR (light detection and ranging) based on UAV (unmanned aerial vehicle). Then, the present typical airborne LiDAR bathymetry (ALB) systems and UAV-mounted solid-state LiDAR bathymetry systems are reviewed. Finally, the technical difficulties in measuring application of UAV-mounted LiDAR bathymetry system are summarized and analyzed, in addition, the development prospect of solid-state LiDAR bathymetry technology and its application are pointed out.

Abstract:The design of phase frequency detector and charge pump used in the phase-locked loop system has been achieved in this paper. The sampling period is divided into odd and even to avoid the overlap problem caused by the excessive reset delay signal. The phase frequency detector is divided into odd and even period by the data selector. The structure of the source-switch charge pump is adopted, and an operational amplifier is added to the structure to improve the matching characteristics and output voltage range of the charge pump. The circuit designed by TSMC N28 nm CMOS was simulated and verified in the Spectre. The simulation results show that when the output voltage of the charge pump circuit is 0~0.9 V, the charge and discharge current mismatch is less than 0.4%, and when the output voltage is 0.45 V, the charge and discharge current mismatch is only 0.028%; the frequency detector and the dynamic characteristic curve of the charge pump cascade simulation has good linearity, and when the two input signals phase of the frequency detector are same, the average output current of the charge pump is only -4.159 nA, which has good dynamic matching characteristics.

Abstract:Due to the complex underwater environment, underwater images are usually degraded low-quality images. Therefore, a multi attention mechanism guided binocular image super-resolution reconstruction algorithm is proposed to selectively learn image feature information for achieving high-quality image reconstruction. Aiming at the low resolution of underwater image, a network with double attention module is designed to enhance the learning of important details; Then, aiming at the disparity characteristics of binocular images, a parallax attention module is proposed to fully learn the prior information of left and right-hand images, and improve the image quality effectively. The PSNR of the reconstructed image with x2 and x4 on the Middlebury dataset is 33.3dB and 28.39dB respectively. It shows that the algorithm can improve the spatial resolution of the image and better retain the image details. At the same time, the reconstruction effect of this algorithm is better than other algorithms on the underwater dataset in real underwater scenes, indicating that it can achieve higher quality underwater image super-resolution reconstruction.

Abstract:To solve the problem that electroluminescence imaging cannot be performed due to the position deviation of photovoltaic (PV) modules, A machine vision inspection method for PV modules position deviation was studied. The edge features of the cell modules are obtained by image preprocessing, and the sub-pixel edge detection method is used to subdivide the edges to obtain the exact edge point coordinates. The subpixel edge points are then fitted using iterative reweighted least squares (IRLS) to obtain the equation of the edge lines. By comparing the fitting results with the basic least squares method, it is demonstrated that the algorithm is robust to outliers on the edges. The performance of the algorithm was then evaluated through position deviation detection experiments, and the results showed that the detection error of this algorithm is about 1.792%, and the average detection time is only 224.7ms, so it can meet the requirements of practical applications. Finally, a comparison experiment was conducted to verify the higher accuracy of the algorithm compared with the traditional method.

Abstract:SMD components AOI automatic re-judgment can improve the detection of straight-through rate and accuracy, reduce the time and economic costs of manual visual inspection. Chip components as the most commonly used PCB circuit SMD components, its error types are numerous, including solder insufficient, more tin, open solder, horizontal and vertical offset, and the body of the missing parts, wrong parts and other issues. For Chip components, this paper proposes an AI-based AOI re-judgment detection method, including the body and solder positioning, body color or silk screen detection and component soldering detection, each step according to the different needs of the design of the corresponding network model, and the above models were trained and tested. The test results show that the body and solder detection network has a mAP of 0.93 and a detection accuracy of 94.94% under the condition that the IOU threshold is 0.7 and the confidence threshold is 0.6; the average accuracy of the color judgment network is 99.41%; the detection accuracy of the screen-print character detection network is 97.6%; and the accuracy of the welding problem detection network is 92.27%. Finally, the overall process was tested according to the 233 NG data collected and 30,000 OK data, and the overall false alarm rate of the method was 11.2%, with an average run time of 97.5ms per frame, under the premise of ensuring that 233 NG data were not missed. currently, the method has been tried on the production line, and the preliminary results show that the method can effectively reduce the component false alarm rate and reduce the labor cost, which has a good application prospect.

Abstract:Aiming at the problems of poor performance of traditional pedestrian detection and tracking algorithms and false detections in public places, crowded or complex background, this paper proposed an NMS algorithm based on the siamese network, which is to improve the detection and tracking of pedestrians in crowded accurately. This method uses less occluded visible parts to remove redundant detection boxes. In order to obtain the visible part, this paper proposed a double-box model (DBM) to predict the whole body part and the visible part of the pedestrian at the same time, which is to ensure the correspondence between the two boxes in the entire network, so as to achieve a better pedestrian detection task performance. This paper has carried out experimental verification on the CrowdHuman dataset, and the experimental results show that it has good robustness and detection accuracy for pedestrian detection in crowded and is better than other models by about 5%.

Abstract:Massive multiple-input multiple-output (MIMO) technology, as a key technology in 5G mobile communications, greatly improves communication capacity, and is an important development direction for rail transit communications. In this paper, the channel measurement of the 32x32 massive MIMO system is carried out at 3.5 GHz and 5.6 GHz frequency bands in a real subway tunnel environment. According to the measurement data, the characteristics of the massive MIMO channel in subway tunnels are analyzed. It is found that the path loss index of the 5.6 GHz band is slightly lower than that of the 3.5 GHz band. By comparing the polarization crossing rate, it is found that the arc tunnel has a stronger depolarization characteristic for +45° polarization and horizontal polarization. The root mean square delay spread (RMS-DS) of the signal in the tunnel is generally distributed within 10 ns. In addition, it is found that the channel capacity gradually decreases with the increase of the transceiver distance. The 32x32 MIMO channel capacity at 3.5 GHz frequency band is 54.64 bit/s/Hz, which is slightly higher than the 53.35 bit/s/Hz at 5.6 GHz frequency band. And it is found that when the polarization mode of the antenna element matches the channel polarization characteristics or the number of elements increases, the channel capacity will increase. These findings can provide a basis for the application of massive MIMO technology in subway tunnels in the future.

Abstract:The leakage of buried oil pipeline will cause pollution of the surrounding soil and groundwater, which will seriously threaten people's property even life safety. This paper first studies the influence of buried oil pipeline leakage on the porosity and water saturation of the surrounding soil, and then obtains the dielectric constant of contaminated soil. Then the radar echo of pipeline and its surrounding soil under different leakage conditions is simulated by computer, and the influence of pipeline leakage on radar echo is analyzed.It provides support for improving the effectiveness of radar detection data interpretation and detection accuracy of buried pipelines and their conditions.

Abstract:The state data of rocket in flight is an important content to evaluate its flight performance. Aiming at the collection and storage of various sensor data during rocket flight, a collection system is designed. In the aspect of system hardware design, modular design is adopted, and stm32h753 chip of arm Cortex-M series is used as the control core, and AD7606 chip of 16 bit high-speed ADC is used to ensure the efficiency of the system, the synchronous performance of the system is improved by using switch sampling control chip ADG1206. Finally, the NANDFLASH chip MTFC32GAKAECN-4M is used to ensure the system storage. In terms of software design, the scheme of A / D conversion and DMA transmission is adopted to ensure the correct data acquisition, effective transmission and stable storage. Finally, in the system test, the analog voltage signal is input to simulate various sensor data during the rocket flight, and the collection and storage experiments are carried out. The collection accuracy, stability and storage performance of the system are calculated and analyzed, and the reliability and correctness of the system are verified.

Abstract:In order to solve the problem of error codes in the process of reading and writing NAND FLASH memory data, an FPGA-based NAND FLASH error correction coding scheme is proposed. The solution uses Micron’s MT29F64G08ABAAA memory chip and is controlled by the FPGA controller for data storage. Every 64Bytes generates 18bits Hamming code check code, and writes the Hamming code check code in FLASH and the Hamming code check code generated when reading the data. The code comparison can determine the position where the error code occurs, and the error code can be corrected by inverting the data bit. The test results show that if the error correction coding scheme is not used for writing and reading 8GBytes data, there are errors in the test at high and low temperatures and normal temperature. The error correction coding scheme is tested at 20℃, 60℃ and -40℃. Error-free, realizes the function of NAND FLASH error correction, and improves the accuracy and stability of data storage.

Abstract:The network traffic analysis system is an important technology that ensures the security of cyberspace. The traditional analysis system is difficult to deal with the massive amount of data. It is urgent for a specific technology to pre-filter the network data packets in order to reduce the burden of the analysis system. Considering the need, this paper designs a 160 Gbit/s network packets filtering system based on FPGA. The system first parses the input network data packets, extracts the keywords, and then uses a hash table to lookup. A round-robin arbiter is designed for the multi-channel query. Finally, filter and output the packets that meet the requirements. System-level simulation and board-level test results show that the average delay during processing is 1.5us, and the effective filtering ratio is over 99.5%, which meets the requirement of 16*10 Gbit/s line-rate filtering. On the XC7VX485T FPGA, RAM resources occupy less than 80% when supporting 40960 keywords queries. The system is high-throughput, low-latency, and high-reliability. It is practical and has great value in network security fields such as backbone networks and data centers.

Abstract:In view of the difficult situation of ordinary vehicles on uneven and complex roads, A new type of bionic robot of Bull Demon King is designed and developed. The mechanical structure and control system of the robot are researched and designed. Solidworks software is used to simulate and test the robot. The simulation results show that: The design of the Bull Demon robot to achieve stable walking, fast walking speed, The walking speed can reach 50cm / s; The stable and fast turning is realized, The turning speed is about 45cm / s, Flexible operation, strong mobility. The robot with good design, simulation and test is assembled and debugged with high-quality materials, Applied in the National Undergraduate robot competition in 2019, A robot can quickly climb 40 cm high steps, The competition test effect is good, has obtained the outstanding result.

Abstract:The driver controller of the subway train is the core control equipment of the vehicle and is the main command appliance for subway train traction and control. Aiming at the problem that the existing driver controller has a complex structure and low digitalization level and cannot interact with the data of metro train control and diagnostic system, a digital driver controller design scheme based on STM32 is proposed and a prototype is developed. The new design adopts two 9-channel encoders to detect the position of traction handle and 12 opt isolators to detect the switching status of driver controller IO nodes, and the measured handle position information and node status information are mutually corroborated to satisfy the error to be valid; meanwhile, a 2-way industrial Ethernet interface is designed to comply with TRDP communication protocol, which can output the driver controller status in real-time. Finally, in order to verify the feasibility of the system was carried out type test, network consistency test and actual loading debugging, the results show that the designed digital driver controller is compact, highly reliable, functional and basically meets the needs of the actual loading. The design improves the digital level of the subway train driver controller, enables self-diagnosis and recording of many types of faults, and improves system reliability.

Abstract:This paper studies the high-power phase-shifted full-bridge converter, and designs the FPGA-based digital control method for the circuit topology. The system control strategy of switching power supply is given, and many key technologies based on FPGA of the power supply digital control system such as digital PI algorithm, digital control system program architecture, and the method of processing floating-point operation are explained. A digital control power supply prototype with a rated power of 6kW was built, a number of functional performance tests were carried out, and the expected design indicators of the switching power supply were achieved. Experimental results show that the FPGA-based digital control system can not only ensure the stable operation of the switching power supply, but also have great advantages in operation speed and event parallel processing. This technology is not limited to the phase-shifted full-bridge circuit topology, and can be applied to other switching power supply topology circuits by appropriately adjusting the modulation and control strategy.