Abstract：An agile long-term hybrid accumulation method based on micro sliding window is proposed for detecting weak targets in terminal guidance radar under clutter background. Long term coherent accumulation is an effective method for detecting weak targets, but this type of method requires a large amount of parameter search and is not suitable for the real-time performance of missile borne terminal guidance radar. Long term coherent accumulation is divided into short-term MTD processing and long-term PRT transformation non coherent accumulation processing to solve the problem of large computational complexity in existing long term coherent accumulation. In response to the data rate requirements of guidance systems, furthermore, it is proposed to use a micro sliding window to achieve edge accumulation and output of data, meeting the demand for output data rate of terminal guidance radar. Simulation and actual data verify the feasibility and superiority of the method.

Abstract：At present, in the simulation of sea surface echo signals in radar systems, target echo models are established only based on parameters such as distance and relative motion speed. However, in reality, the echoes detected by radar on the sea surface not only include ship targets, but also electromagnetic scattering of waves, resulting in complex detection environments. This article uses composite electromagnetic scene simulation calculation technology to calculate the scattering scene of targets and sea surface coupling under low sea conditions, achieving radar cross section estimating and imaging, high-resolution one-dimensional range profile imaging and inverse synthetic aperture radar imaging of certain typical military targets and background environments. Simulation results of the target were analyzed.

Abstract：The causes of sea clutter are complex and diverse, and the target echo is influenced by various factors, which seriously affects the detection, recognition, and tracking of sea targets. In order to solve the problem of insufficient accuracy of traditional research methods, this paper analyzes the statistical characteristics of sea clutter, constructs a three-dimensional feature vector with amplitude entropy, Hurst index and frequency domain peak as the characteristic components, and adopts convolutional neural network method to realize the obvious distinction between sea clutter and targets in feature space.

Abstract：With the development of information technology, the emergence of virtual reality technology has reconfigured the traditional experimental teaching form. Relying on virtual reality technology, the underwater magnetic target detection experimental platform was independently developed by integrating basic physical knowledge with military problems. Through this technology, a large amount of magnetic field numerical simulation data is realized to be displayed intuitively on the virtual platform, an experimental scenario of real-time measurement and display of magnetic anomalies is constructed, and the experimental environment and teaching module are skillfully designed. The experimental platform can help students deeply understand the physical nature of magnetic phenomena and open the window of professional application.

Abstract：In the development of computer architecture, the von Neumann architecture is undoubtedly one of the most traditional and mainstream architectures. In order to pursue higher-speed computing efficiency, people continue to make optimizations on the basis of the von Neumann architecture until it reaches its development bottleneck. Starting from the traditional von Neumann architecture, this paper analyzes the inherent limitations of the architecture, and lists the basic improvement measures for these limitations. Then it specifically introduces a variety of technical means for innovation on the von Neumann architecture, such as the computing in memory, and data flow calculation. Finally, the development process of other non-von Neumann architectures that break through the original electronic computer physical material system, such as brain-inspired computer and quantum computer, is briefly described, and the development prospects and directions of modern computer architectures are prospected.

Abstract：Low dropout linear regulator (LDO) is responsible for providing stable power supply voltage in the circuit system. It is a widely used power management chip. With the continuous improvement of integration and technology, the capacitor-free LDO gradually replaces the LDO with off chip and becomes the focus of research. However, the capacitor-free LDO needs additional compensation circuit to solve the problem of poor stability and transient response characteristics. In order to know the mainstream design ideas in the current industry, a large number of literatures were investigated, the research status of the capacitor-free LDO was analyzed, and the schemes to improve the stability and transient response characteristics were summarized, the technical points were refined, the advantages and disadvantages were evaluated, at last the room for improvement was proposed and the future research direction was prospected.

Abstract：Aiming at the problems of low accuracy of simple text classification model, complex structure of pre-training model and difficult to be directly used in practical environment, this paper proposes a text classification method based on multi-teacher model knowledge distillation. This model uses the training method of "teacher-student network", and the teacher model is the BERT-wwm-ext and XLNet pre-training models. The probability matrix of the output of the two models is fused into soft labels by weight coefficient. The student model is BiGRU-CNN network. The mean square error function is used to calculate the soft label error, and the cross-entropy loss function is used to calculate the hard label error. The student model is trained by hard label and soft label to minimize the value of the loss function. The test results show that the accuracy of the proposed method have great improvement compared with the student model, and it is close to the pre-training model, which can save the running time and improve the efficiency on the premise of ensuring the classification accuracy.

Abstract：To address the issues of insufficient semantic representation and incomplete feature extraction in Chinese event extraction, a method based on RoBERTa and multi-level features is proposed. Firstly, by using the pre-trained RoBERTa model, word embeddings are constructed and extended based on syntactic and semantic information of trigger words. Specifically, part-of-speech tags and trigger word embeddings are integrated into the word embeddings. Secondly, global and local features are extracted using a bi-directional long short-term memory network and convolutional neural network, respectively. The self-attention mechanism is employed to capture the relationships among different features, emphasizing the utilization of important features. Finally, a conditional random field is used to achieve BIO sequence labeling, completing the event extraction process. On the DuEE1.0 dataset, the F1 scores of trigger word extraction and event argument extraction reach 86.9% and 68.0%, respectively, which are superior to existing common event extraction models, validating the effectiveness of this method.

Abstract：With the advent of the post-Moore era, it has become a new direction for the development of the integrated circuit industry to meet the miniaturization and integration of the system through system-level packaging technology. In view of the large demand for low power consumption, miniaturization and integration of controllers in the current market, this paper proposes a system-level packaging technology. Based on a series of design methods such as principle design, package structure design, layout design, signal integrity simulation, power integrity simulation, and thermal stress simulation, combined with the current domestic manufacturing process level, a fully localized controller integrated chip was developed. The integrated chip takes FPGA as the control core, integrates the RF local oscillator signal source and DDS signal source, and realizes the functions of control, output and communication of the RF signal source. Finally, through the test and verification of the sample, the chip has stable function and reliable index, which meets the requirements of technical indicators. Compared with the controller board, the designed integrated chip greatly reduces the power consumption of the product, realizes the miniaturization and integration of the controller, meets the current market application requirements, and provides technical and feasibility reference for the subsequent research of controller integrated chips.

Abstract：The TSV, pipeline, sales and micro-module of the micro-system components are carried out from the parameter settings modeling, DC voltage drop analysis and AC impedance analysis of DDR3 power delivery network, signal analysis methods, criteria. Collaborative design based on chip package system is completed combined with the chip power model and PCB boards. The paper proposes a method for setting component parameters, power integrity DC simulation and AC low impedance design method, and complete cascade connection analysis method of DDR3 signal integrity. The effect of the method is perfect obtained from the DDR3 time domain analysis and timing script, and the research can guide the micro-system PISI analysis and the layout optimization.

Abstract：In order to make use of DMA to move data quickly, no CPU operation is required to reduce the workload of MCU, and each pin of the interface module can be programmed as I2C pin or universal I/O pin according to the demand, this paper introduces an I/O configurable I2C bus interface design with DMA function. The basic principle of I2C bus, the overall structure of I2C module, the design and implementation of DMA function and I/O configurable function are introduced. The function simulation of the interface design shows that the data transmission under the two modes of DMA function and I/O configurable function of I2C interface is realized, which fully meets the design requirements.

Abstract：To solve the hybrid electric power quality problems in low-voltage network such as voltage sag and reactive power imbalance, a voltage sag mitigation technology with reactive power compensation function is proposed. As per the similarity of AC/DC converter and inverter unit, in this paper, a power quality control device based on dynamic voltage restorer is studied. When grid voltage is normal, in order to achieve the reactive power compensation function for the power grid system, the device calculates reactive compensation load needs and emits reverse current by takes real-time sampling of load current. When grid voltage sag occurs, the thyristor valve of device will be cut off, the supercapacitors of device acts as a power source and supplies power to the load through AC/DC converter. The experimental results indicate that this device has excellent performance in solving power quality problems, such as rapid dynamic response and short switching time between voltage sag control function and reactive power compensation function.

Abstract：In view of the engineering features of Yangjiacun Coal Mine's "shallow burial depth, high mining height, and long working face", the current bundled tube monitoring system uses an air pump to pump the gas samples from the monitoring point to the ground through the bundled tube,the sample gas needs to be transported by negative pressure in a long-distance bundle tube from the sampling point to reach the well. Long-distance bundled pipes are difficult to maintain, and once air leakage occurs, the detection results will be unreliable. It takes a long time to take gas from a long distance, a cycle is long, and the analysis frequency of continuous detection was insufficient. The actual situation of negative pressure pumping gas samples was easily contaminated, the analysis and monitoring data were inaccurate, the sampling capacity was small, and the daily maintenance cost was high. The system was upgraded and transformed, combining the micro gas chromatography detection technology, the beam tube positive pressure gas transmission technology, and the data network transmission technology and other high-tech, the micro gas chromatography analysis technology was applied to the field of mine safety monitoring. The monitoring equipment was arranged underground, the positive pressure gas transmission method was used, the gas concentration was analyzed by a multi-component gas analyzer, and the transmission Data was connected to the computer through the network. For different usage scenarios such as computers and mobile phones, special software was designed to facilitate users to operate anytime, anywhere. In this upgrade, the chromatograph was installed in the underground coal mine, which can quickly and accurately analyze the fire data, monitor the trend of the underground disast

Abstract：A two-way relay-assisted device to device (D2D) communication system using non-orthogonal multiple access (NOMA) is proposed to solve the capacity scaling for D2D communication system. Firstly, the signal to interference noise ratio (SINR) of each signal is analyzed, and according to the cumulative distribution function (CDF) and probability density function (PDF) of each signal, the solution form of the ergodic sum capacity of the system is obtained. In addition, the ergodic sum capacity of orthogonal multiple access (OMA) system is analyzed and compared with NOMA system. Simulation results show that compared with OMA system, the achievable capacity of the system can be improved when NOMA is used, and the two-way relay scheme can make the system obtain higher capacity than the traditional one-way relay scheme.

Abstract：Aiming at the problem that some users are affected by shadows, fading and other factors in cooperative spectrum sensing, the sensing performance of weak signals in low signal-to-noise ratio environment is dramatically reduced, and some clustering models do not make full use of energy vectors, a cooperative spectrum sensing algorithm based on multi-head self-attention mechanism is proposed. Each user samples data independently, and the fusion center collects the energy value of each user and forms the energy vector as the input of the self-attention network. A network model based on multi-head self-attention mechanism is designed. The network is used to automatically learn the energy vector features of signals and noises, effectively extract local features, and realize intelligent cooperative spectrum sensing. The simulation results show that the performance of the spectrum sensing algorithm is better than that of the comparison algorithm. When the SNR is -14 dB and the false alarm probability is 0.001, the detection probability of the algorithm is 0.29 to 0.4 higher than that of the comparison algorithm.

Abstract：The detection of surface cracks is a crucial aspect of bridge safety inspection. To address the issue of low efficiency of manual crack detection methods for bridges, and the time-consuming nature of conventional convolutional neural network (CNN)-based crack detection models, a real-time bridge crack detection model based on SegFormer semantic segmentation network is proposed. This model leverages the Transformer architecture, which has demonstrated excellent performance in computer vision. The study demonstrates the feasibility of the proposed model, which exhibits superior accuracy, real-time performance, and robustness in crack detection compared to commonly used lightweight CNN-based deep learning models such as LR-ASPP and BiSeNet V2. The proposed model, when combined with UAV aerial photography, has been applied to actual bridge crack detection, resulting in commendable outcomes.

Abstract：Background management pages are necessary for Internet enterprises, but most of these pages need developers to write from scratch, which is inefficient and has high maintenance costs. This paper proposes a component-based background management page generation system. Through the realization of page modeling, element componentization, and component standardization, it provides users with various basic components, and provides component development interfaces and specifications, and supports customized components. The system provided graphical interface for editing and configuration , which improves the development efficiency, lowers the development threshold and reduces the maintenance cost.

Abstract：As an important part of the integrated navigation system, marine seismic exploration route planning has always been the focus of oil companies. Due to the special operation environment of offshore seismic exploration, in addition to the interference of ocean current, obstacles, underwater environment and fishery, the actual construction efficiency of towline operation is low, which directly leads to the increase of acquisition costs in the work area. The purpose of this paper is to propose an automatic planning method of marine seismic exploration route based on Q-Learning. With the survey line drawn in front of the work area as the research object, the navigation route of the towboat is automatically planned to reduce the interference of human factors and improve the working efficiency of the towboat operation, which is conducive to the maximum benefit and sustainable development of offshore oil and gas field exploration.

Abstract：With the increase of the scale and complexity of industrial production, the traditional SCADA system has been unable to cope with the complex scene of large scale and large data, so the fourth generation SCADA system came into being. Unlike the traditional SCADA system using proprietary communication protocols, the Web SCADA system can quickly obtain equipment data through the Internet of Things access, support multiple communication protocols, and support the expansion of monitoring capability of massive data scenarios based on the flexible expansion capability of the cloud platform. Based on the visualization requirements of industrial production process, this paper designs and develops the Web SCADA system. In the system development process, the design idea of distributed architecture is adopted to improve the system's availability, reliability, scalability, and reduce module coupling.SCADA

Abstract：There were two kinds of relations applied in humidity-functional grid asset radio frequency tag: the influence of moisture-absorption nylon on tag dielectric constant, and the influence of tag dielectric constant on tag radio frequency performance. For tag design, it was used that moisture-absorption nylon on package material and narrow bandwidth design on RFID Antenna. Also, it was improved to get center frequency of tag activation sensitivity for read-write handsets. In the test of application simulation, the average absolute difference of relative humidity is 2.0%RH between humidity-functional grid asset radio frequency tag and moisture meter. In this paper, humidity-functional grid asset radio frequency tag achieved to combine sensor function into electronic tag，and nearly can meet the requirement for appropriate measurement accuracy.

Abstract：In order to meet the requirements of antenna radiation pattern test of ultrashort wave omnidirectional antenna installed in the airplane, an out-field antenna measurement method of simulating multi flight state is proposed. Out-field measurement system uses the ground fixed equipment to test the small airplane pitch attitude and the unmanned aerial vehicle to test the large airplane pitch attitude.The test principle of unmanned aerial vehicle and the antenna radiation pattern test method under the unmanned aerial vehicle flying around or hovering are proposed. According to the method, the installation performance of antennas at different pitch angles was tested to verify the airplane performance requirements of communication links under different maneuver conditions, providing basis data and test basis for subsequence radio frequency link, radio frequency compatibility and radio frequency stealth analysis.

Abstract：Due to the high output power of GaAs power amplifier, it is usually used in radar system and other communication systems with pulse system. According to the characteristics of GaAs power amplifier material, its breakdown field strength is low, and under the drain pulse modulation, when the pulse disappears, the current mutation will form voltage overshoot. When the overshoot voltage is high, it is easy to break down the amplifier. In this paper, the generation mechanism of voltage overshoot is introduced, and several measures to improve voltage overshoot are put through theoretical analysis, and the effectiveness of which is verified by actual measurement in GaAs power amplifier.

Abstract：This article presents the design and verification of an I3C controller IP that can be configured with an APB bus. I3C, which stands for Improved Inter Integrated Circuit, is a new serial communication module that is used as a peripheral IP module in System-on-Chip (SoC) designs. Compared to I2C, I3C supports new communication modes, higher communication speeds (up to 12.5 MHz), dynamic address allocation, and built-in interrupts (IBI). These features make I3C significantly advantageous over I2C and its large-scale popularity and commercial use is a future trend. This article first introduces the basic principles and advantages of the I3C protocol and then describes the design architecture and functional modules of the I3C controller in detail. After completing the design, the I3C controller was functionally verified, and its verification results were demonstrated, proving the effectiveness and reliability of the proposed method.

Abstract：This paper proposes a method for generating GPS L1C/A signals based on software defined radio and CUDA acceleration. The system generates intermediate frequency signals based on user-defined trajectories and satellite ephemeris, and transmits them at the corresponding radio frequency points through USRP, thus realizing real-time broadcasting of satellite navigation signals. To improve the parallel computing speed of the GPU, this paper proposes a CUDA optimization implementation scheme based on asynchronous operation idea from the perspectives of device memory optimization, parallel structure optimization, and CUDA stream acceleration for sampling point data acceleration calculation. The test results indicate that the optimized algorithm is capable of generating high-speed and real-time satellite signals. On an RTX1050TI device, the satellite signal can be generated in real time with a sampling rate of 195 MHz and under the condition of 10 visible satellites.

Abstract：Forward-looking synthetic aperture radar (forward-looking SAR) can acquire image information in the area below the front of the platform. Compared with side-looking SAR, forward-looking SAR has the characteristics of short synthetic aperture, close action distance, wider distance-to-scene width, and azimuthal tuning frequency with distance null, etc. Azimuthal compression using traditional SAR imaging algorithm will cause image sector distortion. In this paper, by analyzing the echo characteristics of the forward-looking array SAR, the data skew caused by the range cell migration correction (RCMC) is restored to the distance information of the target by the inversed range cell migration correction (IRCMC) after the azimuth compression, and the two-dimensional imaging of the ground is completed by coordinate mapping. Finally, the correctness of the analysis and the effectiveness of the method are verified by simulation data.

Abstract：The mini-SAR real-time imaging processing system uses GPU to implement the wave number domain imaging algorithm. It adopts a two-step motion compensation method to compensate for the motion error of the unmanned aerial vehicle platform, and then uses the phase gradient autofocus algorithm to compensate for the residual phase error. Based on the NVIDIA TX2 development board, this paper develops and verifies the real-time imaging processing system algorithm, and combines the GPU characteristics to parallelize the imaging algorithm. The system achieved data processing of 16k×8k points in 18.5 seconds. Experimental results have verified the effectiveness of the real-time imaging processing system.