This PDF file contains the front matter associated with SPIE Proceedings Volume 12254 including the Title Page, Copyright information, and Table of Contents.

Performance of SLD light source in fiber optic current sensor is easily affected by temperature, vibration, and device aging failure, and the failure mechanism is not clear. This paper proposes a fault diagnosis method for fiber optic current sensors based on light source monitoring data. In this paper, according to the working principle of fiber optic current sensor and SLD light source, we analyze the influence of light source parameters on fiber optic current sensor and establish light source data monitoring. The feature vector is dimensionalized by feature extraction. Using the light source monitoring data, a support vector machine-based fault diagnosis model with different kernel functions is designed to establish a fault diagnosis model for fiber optic current sensors. Experiment results indicate that the SVM fault diagnosis model takes less time and has better diagnostic performance, which basically meets the field diagnosis requirements.

In order to improve the operation and maintenance level of the distribution room in the park, according to the actual demand, this paper proposes a comprehensive on-line monitoring system for the distribution room in the park, which monitors the real-time state of the distribution equipment and operation environment from multiple dimensions, realizes the visual browsing of the data, finds the equipment defects and accident risks in real time, and indirectly improves the power supply reliability.

Based on 45nm process, a 14 bit 20 MS/s pipelined analog-to-digital converter (ADC) was designed. In this paper, the ring amplifier is used to replace the traditional operational amplifier (OTA). The ring amplifier has the advantage s of large bandwidth and good linearity, and the minimum power supply voltage of the ring amplifier is low, so the core power supply of the analog module and digital module of ADC is 1.2V, which reduces the overall power consumption of ADC. The ADC consisted of 7 stages, in which 2.5-bit/stage was applied from 1st to 6th stage and the last stage was a 2-bit Flash ADC. The ADC is tested with a sampling frequency of 250MS/s. The simulation results indicate that when the input signal frequency is 10.5 MHz, the SFDR of the ADC is 78.7dB and the ENOB is 11.36 bit; when the input signal frequency is 89.5 MHz, the SFDR can still reach 73dB and the ENOB can reach 11.2 bit.

For the problem of reliability in the long-term use of integrated circuit chips, this article aims to work reliably for eight years, that is, after eight years of work, all devices can work normally, and a specific assessment plan for the life test is proposed. In this paper, a physical unclonable function chip is taken as an example, and related life-span experiments are carried out for the reliability target, and a method of life evaluation of a physical unclonable function chip is proposed. This method of evaluating life can also be used for other integrated circuit chips or components.

Converter open circuit fault prediction plays an important role in improving the intelligent operation and maintenance of energy storage systems. Consider the existing methods for non-intrusive identification of the IGBT open circuit faults in power conversion system (PCS), signal feature extraction occurs difficulties, explosion of data dimensions, and instability of the threshold judgment interval. An open-circuit fault prediction method for energy storage converters based on mel frequency cepstral coefficient (MFCC) features is proposed to support the normal operation and maintenance of PCS. First of all, to communicate the three-phase current on the side is the input signal, and the MFCC fault characteristic data set is constructed by analyzing the energy distribution and envelope characteristics of the signal spectrum in different frequency intervals, and then combined with the nuclear principal component analysis to realize the nonlinear fault under charging and discharging conditions Feature dimensionality reduction screening; secondly, taking the low-dimensional fault feature set as input, build a fault state prediction model based on Bayesian optimization algorithm and one-dimensional convolutional neural network (CNN-1D); Finally, taking the simulation data of different fault states under different working conditions as examples, it is verified that the proposed method has good robustness and accuracy even in a complex noise environment, and has better advantages compared with the existing methods.

Buck converter is widely used in solar power generation and electric vehicle batteries. In order to control the output voltage, this paper builds Buck converter simulation model using PID and fuzzy-PID controllers and examines improvements to the Buck converter performance. The PID and fuzzy-PID controllers regulate the duty cycle of Buck converter so that the output voltage follows the reference voltage. The simulation results present that the settling time of fuzzy-PID is faster than PID and the voltage deviation of fuzzy-PID is smaller than the PID. The performances of system with fuzzy-PID controller are better than system with PID.

A new hybrid switching logical control circuit and an efficient capacitance DAC array is presented by combining split capacitor algorithm and common-mode switching algorithm in order to quickly establish the comparative logic of Successive Approximation Register Analog-to-Digital Converter(SAR ADC) during successive approximation. This circuit is a fully dynamic logic circuit that achieves high-speed conversion and is controlled by asynchronous timing in a different way. The split switching algorithm is adopted at the most significant bit (MSB) in the capacitive DAC array(CDAC), and the rest bits use the common-mode switching algorithm. This logic circuit can maintain the common mode level and reduces the switching power consumption as well. A 10bit SAR ADC is designed at the end. According to Spectre simulation, at a 1.8V supply and 25Ms/s, the ADC achieves an SNDR of 56.66 dB and the effective number of bit (ENOB) of 9.12bit, and the capacitance is reduced by 95.3% compared with the conventional method.

With the continuous development of communication technology, the nonlinear requirements of RF power amplifier are higher and higher. The nonlinear measurement of broadband power amplifier involves signal synchronization, nonlinear measurement and memory effect characterization. Aiming at the synchronization problem of test signals, a crosscorrelation interpolation synchronization algorithm suitable for unknown modulation signals is designed, and the corresponding amplifier test platform scheme is proposed. Compared with other papers, the algorithm in this paper has stronger applicability and anti noise performance. Comparing the test results of instrument manufacturers, the relative error of synchronization time is much less than 0.01%, the relative error of AM/AM is less than 0.1%, and the RMS phase error of AM/PM is less than 0.5°.

In order to improve the accuracy of the estimated value of battery state of charge, this paper builds an electric vehicle battery management system platform, and proposes an extended EKF algorithm based on THEVENIN battery model, and a detailed analysis and introduction of the BMS software and hardware structure, at the same time in order to facilitate debugging, designs the input and output system and the host computer data receiving system, carries out the simulated working condition test of the lithium iron phosphate battery pack through MATLAB/SINULINK simulation, and gets the result that error of the estimated battery state of charge (SOC) is within 5%, confirms the reliability of BMS test platform and software system, accuracy of battery model and estimation algorithm.

With the rapid development of sensors and machinery manufacturing, handling trolley has also gradually become a hot topic. This paper introduces the design of a single-handed control handling trolley based on real-time operating system RT-Thread, which has high application value, including using for daily life assistance of disabled people and non-contact small goods transmission and other fields. The trolley consists of a gesture acquisition terminal, a server transmission network and a trolley control terminal. The gesture acquisition terminal uses wearable devices to realize the gesture recognition. The uploaded data are sent to the trolley control terminal through the server transmission network. The trolley control terminal collects the video stream and transmits it to the server network auxiliary operator to implement the remote control. The system shows good control performance on the experimental platform.

Previous research methods have focused on reuse of application software about embedded software development among smart distribution terminals. However, few investigations have been able to study the reconfigurable method of this software prospectively and in well-layered settings. Moreover, no studies have determined whether the embedded hardware development in the era of power Internet of things that influences the performance improvement of the software platform. The purpose of this paper was to explore the reconfigurable method of architecture level for embedded software development that can be realized only through four steps: start-data operation-control operation-end, and describes the design steps and the application of available high-performance techniques which greatly improves the efficiency of software development. By constructing the three-level reconfigurable software design method of ‘function element-component-system architecture’, this paper realizes the reconfigurable method of smart distribution terminal software development. We further confirmed the feasibility of this method by large-scale application of intelligent distribution terminal to low-voltage distribution station area.

In order to improve the accuracy of active sonar underwater target tracking, an improved algorithm is proposed based on the existing forward-backward prediction algorithm of unscented Kalman filter. The algorithm can adaptively perform forward-backward prediction operation, and replace the measured value with the state estimate value of the past moment to avoid nonlinear problem in reverse prediction. The simulation results show that when the total calculation amount is almost the same, the algorithm can improve the accuracy of state estimation better than the standard algorithm.

When a space manipulator captures an object in orbit, the excess contact force during collision causes the uncontrollable vibration of the manipulator and object. To devise a strategy for reducing this vibration, this paper leverages the multiple configurations of a redundant manipulator and seeks to identify the optimal configuration and contact direction based on the effective mass method. In this paper, a collision model is first established between the end effector and the object based on the continuous contact force method. Then, the peak force, maximum compression, and contact duration between the two bodies during the collision are obtained. For a certain task, the magnitude of the contact force only depends on the effective mass of the space manipulator. Hence, the mapping relationship between the effective mass and the configuration and contact direction of the space manipulator is established, and an ellipsoid is used to describe it. Accordingly, a contact force minimization strategy is designed. The optimal manipulator configuration and collision direction corresponding to the minimum contact force are obtained, and the correctness and effectiveness of the calculations are verified using MATLAB and Adams simulations.

With the aging of population and the loss of hand motor function in stroke patients, finger rehabilitation training robot has attracted more and more attention. This paper introduces the research status of finger rehabilitation robots in recent ten years and summarizes the different characteristics of finger rehabilitation robots with different actuators. The structure and control mode of exoskeleton robot are classified and discussed. The key technologies and future research emphases of finger rehabilitation robot are analysed.

Intelligent temperature and humidity control system is widely adopted in many applications. However, in the greenhouse and storage facilities with large area, there is a conflict between cost and accuracy of temperature and humidity control. Therefore, this paper proposes a temperature and humidity detection module (THDM) suitable for storage applications. The THDM includes STM32 which is one popular single chip microcomputer, DH11 which is one kind of temperature and humidity sensor, ATGM332D module which supports the function of GPS and Beidou dual-mode positioning, and HC05 which is low-power Bluetooth communication module. The THDM not only transmits the temperature and humidity information, but also uploads the geographic location information. Therefore, the placement work is convenient, and the location of THDM can be freely configured according to the special needs and occasions. The maintenance work is also small. With the help of THDM, the intelligent temperature and humidity control system will be able to accurately obtain the temperature and humidity and other environmental information of the specific location, and make correct control decisions.

DC motor has been widely used in industry control area. This paper proposes a speed control system of DC motor. Classical proportional-integral-derivative (PID) control is not flexible for the parameter changes of the system. In the system, the auto-tuning PID controller based on the relay feedback approach are adopted to achieve precise speed, which are respond to disturbances and adjust suitable parameters rapidly. In order to verify the model, the proposed controller is evaluated experimentally with data acquisition card by LabVIEW programming software. The results showed that the system had the characteristics of fast response of speed adjusting and low error.

In Bayesian multi-target filtering, knowledge of detection probability is of critical importance. Significant mismatches in detection probability parameter result in biased estimates. In this paper, we propose a time-varying probability hypothesis density (PHD) filter to deal with unknown detection probability. The detection probability for each target is assured when the predicted states associate with the measurements, and closed-form solutions to the PHD filter are derived by some modifications of the measurement updated formula. Unlike other methods for this challenging class of the PHD filters, the novel method can deal with the unknown and time-varying detection probability effectively, without the complicated process of parameter estimation. Simulations show that the proposed algorithm is suitable to solve the unknown detection probability problems and have better performance, implying good application prospects.

In order to solve the problems of weak battery life and few charging places of electric vehicles, dynamic wireless charging system (DWCS) came into being. In this paper, the significant mutual inductance fluctuation and the internal coupling of the traditional DD coil in the process of dynamic charging are deeply studied. The size and parameters of the coil are optimized by theoretical design, and a large number of software simulation experiments and experimental verification are carried out on the design scheme. Firstly, the size of transmitting coil is optimized to solve the problem of high mutual inductance fluctuation in dynamic charging process. On this basis, the magnetic field distribution of DD coil was analyzed, the influence of magnetic coupling structure change on coil parameters was compared, the optimization direction of coil was established, a new DD coil was designed, and the mutual inductance function of new DD coil was deduced. Then the electromagnetic simulation software is used to simulate and the experimental platform is built to verify the rationality and feasibility of the design.

Load forecasting is the basis of economic operation of power system. In the time series forecasting method based on deep neural network, single load forecasting method can not meet the requirements of load forecasting in the new period. Therefore, a short-term power load forecasting method based on CNN-LSTM neural network is adopted in this paper. This paper first introduces the data set used and the pre-processing operation of the data sample. Then, a sequential hybrid model consisting of single-layer CNN and two-layer LSTM is designed. Finally, the experimental results obtained by using the hybrid model are compared with those obtained by using CNN alone and LSTM alone. The experimental results show that the power load prediction method proposed in this paper gives full play to the advantages of multiple models, effectively deals with the linear and nonlinear characteristics of data samples, and further improves the accuracy of load prediction.

In order to solve problem of the complex structure of electromagnetic AC&DC current comparator, in this paper a principle of AC/DC current transformer based on self-excited oscillation frequency voltage conversion is proposed, and the basic principles of magnetic flux detection of AC/DC current transformer based on self-excited oscillation is introduced. The gain and sampling circuit of the current transformer is proposed to make the output voltage more accurate. According to the principle, the circuit schematic is given, after preliminary tests, the DC current measurement error of the transformer can reach 3%, and the measurement error of AC current is not more than 7% within the range of 1000Hz.

At present, the new instruction set architecture RISC-V has attracted the attention of many chip researchers. Research based on RSIC-V processor has also developed rapidly. However, many RSIC-V-based processor designs focus on short-stage pipelines such as two-stage pipelines or three-stage pipelines, and there are few classic five-stage pipeline designs. Based on this, a five-stage pipeline processor is designed using system verilog. In order to avoid the influence of the branch jump instruction on the execution efficiency of the processor, a prediction circuit based on the dynamic branch pre-method is added. And the data related controller is added to solve the data-related problem. Finally, Modelsim software is used to build a verification environment, and the basic instruction types of RISC-V such as logic operation instructions, arithmetic operation instructions and shift operation instructions are simulated and verified.

With the development of society and the continuous improvement of people's living standards, the amount of domestic waste is constantly increasing. At present, there are almost no mechanical products on the market for waste sorting and most waste transfer stations use manual sorting. In this article we will introduce a waste bin that can sort waste intelligently. This product intelligently identifies the waste put in, matches it with the information in the database, determines the type of waste and puts it into the appropriate bin. The device also has a conveyor belt and a laser distance measuring device. When several pieces of rubbish are put on the conveyor belt together, the laser distance measuring device will intelligently determine the quantity and position of the rubbish and the mechanical claws will sort the rubbish out, identifying and sorting it in turn. This ensures that the waste enters the waste portal one by one and enables precise sorting of individual waste. This has the effect of intelligently sorting the waste. The device also has an ultrasonic full detection module, which alerts people to change their rubbish bags when the bin is full, so that residents can simply put the appropriate rubbish bag into the corresponding bin at the rubbish drop-off point, saving time and reducing the sorting pressure at the city transfer station (Qu 2011).

To improving the reliability and flexibility of the power sensors on the transmission lines and towers, and overcome the shortcomings of the traditional power supply with wind-solar hybrid generators and chemical batteries, a piezoelectric wind energy harvester (WEH) with the low onset speed utilizing the enhancement effect of interaction between vortex induced vibration (VIV) and galloping was developed. The VIV-galloping interactive WEH with a square cross-sectional bluff body was designed by letting the ratio of the theoretical onset speed of galloping to that of VIV as low as 1.27, much lower than the critical ratio for full VIV-galloping interaction. The volume of the harvester is about 8.13cm3. To demonstrate the application of the interactive WEHs, a self-powered wireless sensor node for temperature and humidity sensing was designed. The harvester prototype and the wireless sensor node were tested in a small wind tunnel. Experimental results verified that it is a VIV-galloping interactive device and the onset speed is as low as 2.4-2.7m/s. The working range of the prototype is very wide and the electrical output increases with the wind speed from 2.4m/s to 9.9m/s (about 3.67-4.13 times the onset speed). When the wind speed is 6.5m/s, it took about 202 seconds to charge the capacitor of 100μF from 0 to 4.0V and the self-powered wireless sensor node measured and transmitted the ambient temperature and humidity every 63 seconds, which can meet the need of Transmission Lines Condition Monitoring.

The contactor failure of important systems and key loops will lead to the speed reduction or even shutdown of the rail transit, affecting the operation. It is necessary to study on the deoxidation technology of contactor in rail transit, while most of the failures are found because of the increases of the resistance of contact in contactor due to the oxidation. The study shows good results that electric-deoxidation by making the contact pass through a larger current by load resistive and inductive reduces the failure of the contactor in rail transit applications.

Pepper is a kind of seasoning food. In the face of the current situation of high water content, perishable and not resistant to storage of fresh pepper, a molecular sieve drying device based on PLC is developed by using PLC + molecular sieve to solve the drying problem of pepper dehydration. The practice shows that the device can well meet the process requirements of pepper drying, and has the characteristics of low power consumption, high production efficiency and easy scale and popularization.

The use of electric energy water heater, because water temperature is controlled, often appear to deviate from control temperature is big, temperature achieves steady state time long problem, cannot get popular all the time. To solve this problem, a fuzzy PID adaptive water temperature control system is designed. The system mainly collects temperature data through DS18B20 and filters temperature data by weighted average filtering method. According to the feedback data, the fuzzy PID controller controls the power of the heater by changing the pulse width modulation signal to keep the temperature stable. The results show that compared with the conventional controller, the fuzzy PID controller has the advantages of short adjusting time, small overshoot and fast response speed, and can achieve better control effect when applied to the water temperature control system.

In order to improve the short-term power load prediction results, this paper uses long short-term memory network to predict short-term power load. With the improvement of accuracy requirements, the traditional load forecasting does not consider problems such as time series and eigenvalues. This paper proposes a long short-term memory network model to predict load. First, the data set is processed, cleaned and normalized, and the data set is divided into training set and sample set. Then, a prediction model is built, and appropriate parameters and eigenvalues are selected for the model to study the impact on the short-term power load under the LSTM model. This paper uses the Short-term electricity load forecasting (Panama) dataset on the kaggle platform to verify the model, and uses multivariate and multistep to forecast short-term electricity load.

The current power grid is a hierarchical and partitioned power grid, and the power supply adequacy of the partitioned 220kV power grid needs to be evaluated in real time. The article analyzes the sensitivity between the branch power flow and the node load, the sensitivity between the branch power flow and the generator active power, and deduces the optimization model of the power supply adequacy of the district power grid. This paper uses a linear optimization method that takes into account the power flow section constraints and the branch power flow sensitivity to solve the model. This paper uses the IEEE30-node system as a calculation example to verify the above method. The results show that the real-time assessment of the power supply adequacy of the power grid can calculate the maximum power supply capacity of the regional power grid and reflect the current operating conditions of the power grid.

The ground test equipment for the aircraft safety control system can provide energy channels, receive positioning data, and issue commands and control signals for the safety control system. It is one of the core equipment of the safety control system. In this paper, by sorting out the indicators of the ground test equipment, it is divided into different modules according to functions, and the embedded computer-STM32-FPGA architecture is used to design a flexible and configurable ground test equipment. When performing tasks of different models, different function sub-boards can be replaced according to specific requirements, which can meet the needs of ground testing of most aircraft security control systems. At the same time, the research and development cost are reduced, the development cycle is shortened, and the reliability of the product is improved.

Monitoring the condition of oil-filled casing and other equipment with little oil is of great importance for the safe operation of power grid. In this article, through analysis of the existing hydrogen sensor selection, analysis of transformer bushing structure model, design a high-pressure oil casing in hydrogen gas on-line monitoring terminal which include LoRa, solidstate palladium alloy membrane hydrogen sensor and temperature sensor, the terminal can be real-time monitoring casing hydrogen content in the oil and the oil temperature, and real-time uploading data to the host. After application in substation, the effect is better. This paper describes the research status of hydrogen sensor and the application of wireless communication technology LoRa in electric power online monitoring and acquisition terminal, and introduces the software and hardware design of monitoring device in detail.

In order to solve the problem of DC capacitor voltage fluctuation during the operation of five level cascaded H bridge, a new DC capacitor voltage balance control method is proposed based on the analysis of the current control method. The superposition of active current to reference current of each phase can control the balance of three-phase voltage, and the superposition of the voltage whose phase is same as the active current but sign is contrary to the modulation wave of two H bridge unit respectively. Then the capacitor voltage on the DC side is almost maintained near the given value. The simulation results verify the effectiveness of the proposed method.

In recent years, with the popularization of the concept of green and healthy life, the sales volume of potted plants has generally shown an increasing trend, but relevant data show that most people have little time to care for pot-ted plants and poor care. Most of the flowerpots on the market are containers for plants, with no other functions, and the functions of some multi-functional flowerpots are only for light supplementation, water supplementation, etc., which are relatively simple. A fully functional pot is an important factor in successfully growing plants. Therefore, this paper will propose a multi-functional smart flowerpot, which endows traditional flowerpots with new nursing functions such as fertilization, loosening soil, etc., and also has a special monitoring device to make up for the lack of manual care time and poor care. While reducing labour costs, it can also promote the healthy growth of plants, bringing users a better pot planting experience.

Data applications become an important part of the metaverse, data center is the core hardware. State Grid aims to build a smart and reliable power grid, and has a proposal to integrate the construction of substation data centers. This paper displays the appearance of a data center on-site controlled integrated substation, discusses the fusion scope of data center and substation systems from a practical point of view, lists the cost difference between traditional data center construction mode and substation integrated mode, introduces related fusion evaluation indexes and propose a method for economic evaluation via cost-splitting simulation. In this way, State Grid can obtain theoretical and technical support when investing in substation integrated data centers.

For a long time, the application of power electronic devices and the access of new energy have affected the power quality of power lines. With the high proportion of photovoltaic and other new energy access recently, the power quality level of power grid will be greatly affected. Electric energy measurement may be affected by various factors in this new environment, which is an important part to ensure the fairness and justice of electric energy transaction. Firstly, the possible power quality problems caused by the high proportion of new energy access are analysed; Secondly, the theoretical effects of harmonic, voltage and frequency deviation, voltage fluctuation and flicker, three-phase imbalance on electricity meters are also studied; Finally, the influence tests of each index are developed to verify the influence of power quality on electricity meters.

According to the requirements of the mobile communication system, radio frequency identification system, ultra wideband system and the mobile digital TV system to the antenna's performance, the present paper originally combines split growth electromagnetic bionic structure, photonic crystal antenna and rectangular array antenna, designed a electromagnetic bionic photonic crystal array ultra wide band antenna, fabricated the antenna sample and tested the radiation characteristics of the antenna. The result of test indicate that this antenna completely covered all working frequency bands of the second generation to the fifth generation mobile communication, the frequency bands of the radio frequency identification system, the frequency band of ultra wideband system, and the frequency band of the mobile digital TV system. The smart terminal using this antenna will have many functions, such as mobile communication, radio frequency card read and write, ultra wideband transmission, mobile digital TV and so on; this terminal can achieve multiple network integration of microwave frequency band.

According to the security problems caused by the software definition of distribution terminals in large-scale applications, it is of great significance to study the evaluation technology of terminal embedded operating system. This paper defines the architecture characteristics of micro application of distribution terminal by studying the software, establishes a progressive systematic detection system of micro application of distribution terminal, and forms the evaluation index and automatic test scheme of micro application of distribution terminal through the research of consistency verification modelling and test packaging method, which is helpful for the real realization of informatization, digitization and Modernization has certain scientific significance.

With the development of the times, people are increasingly pursuing the comfort and convenience of life. Water is an indispensable part of people's family life. At this stage, the faucets used in most families adjust the water temperature through manual and mechanical rotation. On the one hand, the adjustment speed is slow. On the other hand, it is often difficult to accurately achieve people's really comfortable water temperature. The water outlet is either hot or cold(Tan 2013). The temperature control faucet designed in this paper is different from traditional products. It takes the STM32 single chip microcomputer as the control core, uses the high-precision temperature sensor DS18B20 to detect the ambient temperature, sends the digital temperature signal output by the AD conversion module to STM32 for processing. It displays the current ambient temperature value and each gear temperature value preset with the key on the LCD display screen. In addition, using the existing voice signal module, we can convert the voice signal into the input signal of single chip microcomputer, and then use PWM to change the speed of stepping motor, so as to control the water flow and temperature more conveniently and quickly.

The article studies the key technical parameters such as "horizontal position deviation", "distance-counting error", "velocity measurement error", and "time-counting error" of intelligent service terminal of cruising taxi, and determines the appropriate standards for the calibration of the parameters The novel calibration procedure presented in this article has been performed on several typical types of intelligent service terminals of cruising taxis widely used in China. The experimental and the uncertainty analysis results show that the calibration method presented in this article can be adopted for the periodic calibration of the intelligent service terminal of cruising taxi, in order to establish the metrological traceability system of the instrument.

To solve the problem that the existing license plate recognition algorithm process is complex and the real-time recognition detection is ineffective, a license plate recognition method based on the principle of color difference components and clustering is proposed. Firstly, the license plate image is denoised by median filter, then the chromatic aberration image is obtained by mathematical operation using the RGB channel characteristics of the image, then the optimal threshold value is searched by maximum entropy traversal method, the chromatic aberration image is segmented by threshold value, and the minimum bounding rectangle of the segmented area is marked, and the clustering principle is introduced to merge the disconnected areas of some special characters. Distance thresholds are derived from the supervised learning of the two classifications and are ultimately marked for subsequent feature extraction. The results of sensitivity and accuracy index calculation are 97.4% and 96.1%, respectively. The recognition accuracy is 96.6%, which basically meets the practical application.

The ever-increasing amount of data has brought severe pressure to the data center. Facing the large-scale data processing requirements, the data center not only needs to increase the data bandwidth, but also needs to ensure the timeliness of data processing. It is increasingly unable to meet the processing requirements of high throughput and low latency. This topic adopts the stream processing architecture of heterogeneous collaborative computing based on FPGA and CPU, designs a dual-channel separated stream processing system based on FPGA network offload processing, improves the management and interaction capabilities of FPGA, reduces processing delay, and provides a set of manageable The general data real-time processing platform.

This paper adopts a short-chain circulation mode suitable for agricultural products from the background of agricultural product circulation, domestic and foreign research status and achievements. Assuming that all agricultural products in rural areas have been assembled in the rural distribution center, only the short-chain distribution and pickup process from the agricultural product distribution center to each supermarket user is considered, and the improved genetic algorithm is used to plan static customers. Dynamic customer insertion into the original delivery route avoids vehicle overuse to a certain extent.

This paper proposes a track association algorithm based on convolutional neural network. The algorithm uses multi-time history correlation local track estimation, and adjusts the position coordinate estimation in the track data to classify the track association. It is used to solve the technical difficulties of track splitting and redundancy caused by track association of multiple radar stations. After simulation analysis, the results show that the intelligent track association algorithm proposed in this paper has higher correct association probability and better universality.

Intrusion detection technology can effectively evaluate the security state of the system. However, the accuracy of traditional intrusion detection methods still needs to be improved. Therefore, this paper proposed a SVM model based on improved QPSO. QPSO is used to optimize the hyperparameters of SVM to improve its accuracy. In addition, the shrinkage expansion coefficient 𝛼 in QPSO is improved to make QPSO adaptive. Experiment shows that the proposed algorithm has good performance.

The aboveground biomass (AGB) is a key index for predicting wheat yield. In the case of high biomass, the AGB estimation of single spectral feature or image texture is poor. Therefore, this study evaluated the ability of fusion of spectral reflectance and texture to predict wheat AGB. Among them the reflectance spectrum of the wheat canopy was collected by near-earth spectrometer, and the texture features of three bands of RGB were extracted by gray co-occurrence matrix. Partial least squares regression (PLS) model was used to evaluate the relationship between fusion features and AGB. The experimental results based on the validated data set show that the AGB estimation effect of feature fusion is better than that of single feature (R2 = 0.70; RMSE = 0.06). This shows that the combination of spectral reflectance and texture can improve the accuracy of AGB estimation in the later stage.

Traditional phased array radar (PAR) beamforming only has angular dimension degrees of freedom, and most of the research models are linear arrays. Frequency diversity multiple input multiple output (FD-MIMO) radar expands the freedom of beam range dimension and obtains the freedom of transmitter. In order to study the beamforming performance of FD-MIMO radar under planar array model, a signal model of planar array frequency division MIMO radar is established. Based on this array model comparison and analysis three work modes: Two-dimensional frequency diversity MIMO (TFD-MIMO) radar, pitch frequency diversity MIMO (PFD-MIMO) radar, azimuth frequency diversity MIMO (AFD-MIMO) radar beam characteristics and Capon beamformer performance. FD-MIMO radar beamforming introduces range-dimensional degrees of freedom, which has significant advantages in countering main lobe deception jamming. The research of PFD-MIMO radar and AFD-MIMO radar provides theoretical support for the balance between high performance and high data dimension of the radar system. The proposed FD-MIMO radar model has flexible working mode and enriches the research of multi-mode radar system.

Rubbings of China is a great treasure, of which the recorded information is later generations research was one of the important ways in politics, economy and culture. But the rubbings are vulnerable to damage. However, the partition of rubbings can make this information better preserved, which is of great significance to the protection of cultural relics, historical research and cultural inheritance. This paper combined region growing segmentation and connected domain labeling to segment the historical document image. Firstly, the OTSU algorithm has been employed for document image binarization. Second, the binarized image has been corroded to remove isolated noise points, and then the effective connected area is extracted by the connectivity domain marking algorithm to find the initial value seed point. Finally, the seed point segmented the area growth to obtain a clear binarized document image. Experiments showed that the automatic seed region growth leads to performance improvements over the K-means cluster and OSTU method. We tested our method using the DIBCO 2010 dataset, which showed segmentation accuracy of up to 98%.

U-shaped networks are widely used in the field of image denoising with their multiscale and jump connection structures in recent years. The feature extraction structures mainly used in previous works are convolutional neural networks (CNNs), but their ability to extract information at a distance is poor. In this paper, we propose a U-shaped network structure combined with large kernel attention structure for image denoising, in which the CNNs structure can effectively extract local information while the large kernel attention structure has better extraction of global information and lower computational cost compared with Transformer, through which local-global features are learned on the feature maps of the input noisy images at different scales, which can effectively enhance the performance of the network and improve the image denoising task. The performance of the network can be enhanced to improve the performance of the image denoising task.

The rapid 3D reconstruction of field scene has attracted more and more attention in civil and military fields. This paper proposes a fast 3D reconstruction system of field scene. This system mainly includes: UAV aerial photography module, sparse point cloud construction module and 3D scene construction module. The system is decoupled through modular processing to reduce the complexity of the system. Aerial photography is carried out in the field real scene by four-rotor UAV. The field scene is reconstructed according to the aerial image to verify the practical application value of the system.

Electroencephalogram (EEG) signals are easy to be interfered by the artifacts of both electrooculogram (EOG) and electromyogram (EMG) during data acquisition. The two artifacts are unwanted in EEG signals processing but difficult to remove. In this paper, a novel blind source separation method for removal of both EOG and EMG based on the fast independent component analysis (FICA) and the independent vector analysis (IVA) is presented. The system uses both high-order statistics and second-order statistics to analyze the non-Gaussian and weak correlation of the artifact signals. Firstly, the FICA algorithm is used to decompose EEG into independent components. Next, the EOG artifacts will be identified and removed according to the window width of the separated EEG signals. Finally, an IVA Gauss-Laplace distribution (IVA-GLD) algorithm with an adaptive step size is exploited to remove the EMG artifacts. Experiments were carried out by the data set collected by a self-developed equipment. The root mean square error (RMSE) of the proposed method is 19.431, which is lower than the similar method without the peak window (23.118), and the correlation coefficient (CC) is 0.895, which is higher than the similar method without the peak window (0.799).

With the development of information technology, drone cluster have been widely used in electronic warfare (EW). In this paper, the distributed cooperative jamming research based on drone cluster is carried out to meet the requirement of jamming for shipborne air defense and antimissile radar. Firstly, the distributed jamming model of shipborne radar is established, and then the cooperative jamming equation is established, finally, according to the cooperative jamming equation, the numerical simulation of shipborne radar is carried out.

Frequency diverse array imposes a different small frequency increment on each element to provide many promising applications. As a new radar system, FDA radar covariance matrix estimation plays an important role in target parameter estimation, interference suppression and adaptive processing. Firstly, we formulate the structure of the FDA radar covariance matrix with Hermitian and Toeplitz. Then, we present an asymptotic maximum likelihood covariance matrix estimation for the FDA radar with computationally efficient. Numerical results show that compared with sample covariance matrix AML enhances the estimation performances of the FDA radar covariance matrix.

Beidou satellite communication system has the function of short message transmission. The traditional short message "broadcast" function is to use Beidou command aircraft to bind with subordinate end users for "broadcast", which has poor timeliness. This paper designs a short message "broadcast" system using Beidou multi card machine, and carries out data processing at both the sending end and the receiving end, so that it can adapt to various external systems. Through practical verification, when the number of user terminals is not too many, the success rate and timeliness are greatly improved compared with the traditional system.

Aiming at the problem that Distributed Optical Fiber Acoustic Sensing (DAS) system will misjudge external intrusion signals, an intrusion signal discrimination method based on MFCC-Energy entropy feature and FTO-SVM is proposed. Firstly, the former 13-dimensional Mel-Frequency Cepstral Coefficients (MFCC) is extracted from the collected sound signal, and Principal Component Analysis (PCA) is used to reduce the dimension of MFCC; Secondly, the energy entropy of the collected sound signal is calculated, which fused with the reduced MFCC as the feature parameter of the collected sound signal; Finally, the extracted feature parameters are discriminated by using Support Vector Machine (SVM) with the hyperparameters optimized by the Fibonacci Tree Optimization (FTO) algorithm. The results show that the proposed method can effectively improve the system discrimination accuracy for intrusion signals, and is of great significance to perimeter security and fault diagnosis and other related fields.

With the rapid development of modern information technology, the overall level of Chinese design has been continuously improved, and the application scope of new technologies, means and concepts in graphic image and visual communication design has been continuously expanded. In the development of information age, the design technology of computer graphics and images has been adopted in various industries. The development of social economy puts forward higher requirements for graphic and image design. Under the background of social media and popularization, the processing of computer graphic and image should go beyond the traditional simple presentation, and use virtual reality technology to provide a more three-dimensional and subjective design effect. The rapid development of virtual reality technology has opened up a new field for computer graphics and image design. Based on the characteristics of computer graphics and image design and visual communication design in virtual reality environment, this paper analyzes the integration and specific application of computer graphics and image design and visual communication design, hoping to help the development of virtual reality technology in China.

The feature points extracted by the FAST algorithm in the ORB algorithm are densely distributed, and the repeated appearance of the feature points leads to a long time-consuming process in the matching process. Aiming at this problem, this paper proposes an improved method. First, the image is partitioned, then the feature points obtained by the FAST detection algorithm in the original algorithm are sorted, and the feature points that are ranked first and meet the conditions are selected for binary description, and finally the homography matrix is added after hamming matching to further improve the matching. Accuracy. The experimental results show that the algorithm improves the time-consuming of the original algorithm from 1.4s to 0.7s.

Pulse compression signal is widely used in radar detection because of its outstanding anti-interference ability. The parameter measurement of pulse signal by signal analyzer is the basis of accurate detection, and the extraction of single pulse can reduce the complexity of pulse measurement. The commonly used neural network extraction methods have high algorithm complexity and are difficult to implement in instruments. In this paper, a pulse extraction method based on statistical histogram is proposed to achieve pulse power level classification. The paper estimates the pulse top power, base power and pulse amplitude, and propose the algorithm flow of pulse endpoint position extraction. Using MATLAB simulation, it is proved that the relative error of the parameter estimation results of the algorithm in this paper is less than 0.2%. The algorithm complexity is reduced by 20% compared with the traditional method.

In order to improve the performance of low precision INS/GPS integrated navigation system, an effective algorithm of INS-assisted GPS tightly coupled navigation based on EKF was used,which uses the acceleration and angular velocity information measured by INS to estimate the pseudorange and pseudorange rate error measured by GPS, this error is treated as the measurement information of EKF, and uses the corrected INS information to assist the tracking loop of GPS receiver. This method can effectively improve the acquisition and tracking ability of GPS and improve the performance of integrated navigation system in complex environment.

Frequency hopping signal has attracted widespread attention in the fields of civil and military communication since its appearance because of its unique anti-interference performance. At present, the time parameter estimation algorithm of frequency hopping signal is mainly based on the extraction of spectral peak. This kind of method has large amount of calculation, low accuracy and is easy to be disturbed by noise. In this paper, a parameter estimation method based on standard state frequency and tolerance value is proposed for the time parameter estimation of frequency hopping signal. The time parameters of frequency hopping signal are estimated by using the mutation characteristics of frequency hopping signal at the hopping moment. Simulation results show that this method can accurately estimate the time parameters of frequency hopping signal. Among them, the error between the estimated value of the frequency hopping start time, the frequency hopping end time, and the frequency hopping period is less than 0.1% compared with the actual value. Compared with traditional methods, the accuracy can be improved by more than 10%.

In order to obtain clearer CT images at low doses, this paper proposes a super-resolution reconstruction method of lung CT images based on Laplacian pyramid residual network. The SRResNet network is connected in parallel to solve the problem that the traditional network model adopts a single scale. At the same time, the BN layer in the SRResNet residual module is deleted, and the feature information between the residual blocks is deeply fused through the dense series connection between the residual blocks. Enhance the network's perception of image features. The experimental results show that the lung CT image reconstructed by the algorithm proposed in this paper has richer details and clearer edges.

Image has been widely used in various fields, but due to camera motion or scene change, the image will form different degrees of non-uniform motion blur. For image motion blur, the existing motion blur method based on conditional generative adversarial network still has a lot of room for improvement in quality and efficiency. As a result, In this paper, we propose a self-aware Attention Motion Deblurring Using Conditional Adversarial Networks, which is based on Conditional generative Adversarial Networks. (SAD-GAN), adding Self-aware Attention (SA)and Cascaded Parallel Dilated Convolution (CPDC) to the generator. The discriminator uses the ideas of global discriminator, local discriminator and relative discriminator, and the whole network works closely together to capture the context fuzzy information more accurately, so as to achieve high quality deblurring. Experimental results on existing public data sets show that the proposed method performs better than other advanced methods in image de-motion-fuzzy structure similarity (PSNR) and peak to noise ratio (SSIM) quantitative indexes.

Aiming at the problems of large system, complex structure, heavy server load and difficult business function expansion in the enterprise measurement instrument verification system with traditional integrated architecture, a lightweight distributed measurement instrument verification system architecture based on P2P network is proposed. The verification business processing service, traceability approval service, basic information service of measuring instruments and business registration service are distributed to P2P network nodes, and the message communication is used to synchronize the node information and coordinate the tasks of nodes, so as to realize the distributed processing of measurement instrument verification and traceability approval business. The centralized control of basic information of measuring instruments and system operating parameters ensures the data sharing and consistency of the system. In addition, the polling method is improved to realize the load balance of verification service nodes. Experiments show that the distributed architecture has good practicability, smooth business processing, reduces the difficulty of system development, improves the development efficiency, and the system runs well.

The rapid and uncontrollable spread of COVID-19 has seriously threatened global public health. Rapid and accurate diagnosis of COVID-19 is the main key to control and manage the epidemic. COVID-19 segmentation can provide great insights to accelerate clinical decisions. COVID-19 segmentation based on deep learning method has attracted extensive research in the field of medical image analysis. However, most existing networks are heavyweight networks, which causes structural redundancy and expensive computational cost. Moreover, the segmentation problem of low-contrast COVID-19 image and the obscure boundary between the infected area and normal tissues affects accuracy of segmentation. To solve the problems, we propose a lightweight contextual information fusion network, LCFNet, for COVID-19 segmentation. We introduce a contextual information fusion strategy combining multiple global pyramid guidance (GPG) with scale-aware pyramid fusion (SAPF) module and deep supervision (DS) module, which can capture more fine-grained image features. We conduct experiments on two COVID-19 datasets. We perform the ablation studies, demonstrating the effectiveness of key components of the proposed method. Compared with traditional segmentation methods, LCFNet model match more consistently with the ground-truth boundary, which shows the superiority of the proposed method. Moreove, LCFNet has 1.68M parameters, which demonstrates the robustness of the proposed method. Our proposed method can further induce the number of model parameters. Compared with the state-of-the-art methods, the proposed approach has achieved significant improvements and is also superior to other segmentation methods.

As a method of geological disaster management, grouting has been widely used. At present, there is no good scheme on how to evaluate the grouting effect efficiently. In this paper, a dynamic monitoring system of grouting process based on ZigBee wireless network is proposed by monitoring the apparent resistivity change in the grouting area, which provides a basis for the evaluation of grouting effect. The monitoring system consists of a user control terminal, a ZigBee wireless network and a data acquisition terminal. The cascade of multiple data acquisition terminals realizes the expansion of area width, and the user control terminal can monitor multiple monitoring profiles through the ZigBee wireless network, so that the electrodes required for multiple monitoring profiles can be laid in the area to be measured at one time to dynamically monitor the apparent resistivity of the grouting area. After testing, the system can collect high-precision data to monitor the change of apparent resistivity in the grouting area, and can meet the rapid evaluation of filling effect in the grouting area.

According to the application requirements of a certain type of radar turntable thin-walled bearing, the bearing parameter optimization design was carried out, and the optimal load-carrying and friction torque performance was obtained. In order to achieve deformation coordination, a large interference design method was adopted, and the interference amount and bearing were determined by simulation. The initial clearance and assembly clearance of the large thin-walled bearing are then realized by the hot and cold interference assembly technology of the large thin-walled bearing. The question provides an important reference.

With the continuous expansion of the scale of my country's railway electrical system and the increasingly complex equipment functions, the requirements for ensuring the safe, stable and long-term operation of the electrical system are becoming more and more urgent. The operation and maintenance system is mainly through manual maintenance, which is inefficient. With the continuous development and application of technologies such as information technology and intelligent terminals in my country, intelligent monitoring technology has been gradually applied to the maintenance of signal systems. In order to facilitate the electric service personnel to grasp the operating equipment status in real time, optimize the equipment maintenance process and improve the maintenance efficiency. This paper comprehensively considers the daily maintenance, operation, emergency response operation process and related precautions of the signal system, and combines key technologies such as data access, data fusion, data visualization, fault prediction and health management to realize the analysis and design of intelligent operation and maintenance application functions. Provide reference for the improvement of railway signal operation and maintenance level and the transformation of electric service operation and maintenance mode.

Aiming at the problems of traditional fractal slot antenna, this paper optimizes and improves the fractal iterative process, and puts forward a new embedded slot fractal iterative design scheme, which greatly improves the broadband working ability of fractal slot antenna. According to the performance requirements of wireless transmission antenna for unmanned aerial vehicle system working in multi frequency band, a dual frequency fractal slot monopole antenna is designed by combining embedded slot fractal structure, monopole antenna and fractal induced radiation patch. The actual test results show that the antenna can fully cover the Bluetooth communication frequency band, the fourth generation mobile communication working frequency band and ultra wide band communication frequency band. The antenna has excellent omnidirectional radiation performance, small size and high radiation intensity. It is an excellent unmanned aerial vehicle system antenna with broad application prospects.

Aiming at the 17dB loss problem in the 10Gbps rate transmission channel in SerDes, this article implements an equalizer of SerDes receiver with TSMC 28nm CMOS technology. At first, a structure of continuous time linear equalizer (CTLE) is adopted to compensate for the loss in the channel. However, due to the inter-symbol interference (ISI) of signals passing through channel, the phenomenon of serious tailing occurs. Therefore, a 4-tap finite impulse response (FIR) decision feedback equalizer (DFE) is added to eliminate ISI post-cursor. Based on the circuit schematic, the simulation results show that the proposed equalizer structure can effectively compensate for the 17dB channel loss and eliminate the ISI post-cursor. The equalized eye width reaches 0.95UI at 0.9V supply.

In the process of online monitoring based on multi-sensor machining, in view of the problem that the accuracy rate of the monitoring model decreases due to the sudden fault of a certain sensor, this paper proposes a method to rapidly train a new monitoring model, and takes tool condition monitoring as an example to verify the effectiveness of the method. In this paper, the stacked autoencoder (SAE) model is used to monitor the tool conditions during the machining process, and the accuracy rate is 99.9%; when a sensor fault occurs suddenly during the machining process, the monitoring model accuracy rate drops to 73.6%. In order to solve the problem of the monitoring model accuracy rate declining and prevent the monitoring process from being interrupted, the method proposed in this paper to quickly train a new monitoring model only takes 5s to train a new stacked autoencoder model. The average accuracy rate of the new monitoring model is 98.6%, and it can still accurately monitor tool conditions.

In order to evaluate the performance status of the speed sensor for the key components of EMU (Electric Multiple Units), according to the main environmental factors experienced by the service life-time, the type test scheme and corresponding detection requirements of the speed sensor are formulated. Combined with the test results after each type test, the performance status of the speed sensor is checked, the main faults occurring are summarized, and the failure analysis is carried out. The speed sensor is not a full-life product (less than 30 years), so it is recommended to repair it in fifth class, and the main failure location and failure stress of the product are found out, which provides important basis for the maintenance of the repair cycle and repair system.

Aiming at the problems of backward test method, long test period and low test efficiency of communication-based train operation control system, the formalized generation method of CBTC (Communication Based Train Control) system engineering acceptance test sequence was studied. The first is to divide the test line into several test sections; the second is to analyze the constraints of the test cases and the characteristics of the test lines; the third is to design the matching algorithm and the connection algorithm, and to convert the constraints of the test cases and the characteristics of the lines into a formal language; It is to design test sequence generation software to generate several executable test sequences. Through the result analysis, the test sequence generated by this method meets the test requirements of the CBTC system, can cover all test cases, and greatly improves the test efficiency.

The current engineering construction quality control using the installation of sensors and other ways of quality control, not only the efficiency is affected, and can't trace the quality problems of each construction link, resulting in a delay in the construction period, the quality problem rectification effect is poor. Aiming at the defect of appeal, the real-time quality control method of assembly building engineering based on BIM technology is studied. According to characteristics of prefabricated building construction process, analysis of factors influencing the modular construction quality. The appropriate influencing factors were selected to establish the quality control index system, and the entropy weight method was used to determine the index weight. According to the index weight, the application of BIM technology design in different construction link quality control plan, realize the real-time control of the quality of engineering. In the actual case application, the effective rectification rate of project quality problems is as high as 98.5% on average after using BIM technology, and the construction period is effectively shortened.

The DC FSM equipment used for monitoring the inner corrosion of oil and gas pipeline has problems such as low safety, vulnerable to interference and low signal-to-noise ratio. The AC FSM is developed to avoid the above problems. This paper introduce the design of data acquisition and analysis software for the AC FSM equipment, which including data preprocessing module, pipeline corrosion calculation module and data analysis and display module. The experiment showed that the software can display the internal corrosion morphology of the pipeline in the monitoring area in 3D, reflecting the internal corrosion location. The defect quantification of the software is consistent with the actual situation, indicating the effectiveness of the software.

With the rapid development of mobile Internet and the continuous expansion of network scale, the network security situation has become increasingly severe, and the endless network security threats have put forward higher requirements for network security performance. In this paper, we investigate this issue from two perspectives of situation assessment and situation prediction, and the main research work includes, for the traditional BP neural network situation assessment model, the network attack assessment only considers the attack effect, and the model convergence speed is slow due to the inaccurate parameter initialization settings, and the assessment efficiency is poor. An LSTM-based method for detecting malicious network traffic in an encrypted environment is proposed, while the network posture assessment based on the optimized BP neural network for the attack impact index system is carried out to achieve real-time monitoring and analysis of the network security status.

In view of the large scale of Ancient Chinese books and the laborious problem of digitalization by manual input method, the ancient books digitalization model based on CRNN is proposed to realize the efficient text input of ancient books. In this paper, we use Handwritten Characters data set of Chinese Ancient Handwritten Characters Database (CASIA-AHCDB) to generate vertical text image data set of variable-length Handwritten Characters. Finally, after iterative training, the accuracy based on character (similarity measured by text editing distance) reached 97.54% in the training set and 96.58% in the test set. The method can be used to identify and proofread ancient books.

This paper analyzes 48923 examinee's achievement data in a high school academic level examination in a city, and comprehensively evaluates the overall distribution of the examination from the aspects of average score and distribution of examination results. Pearson correlation coefficient is used to study the correlation between courses, cluster analysis is carried out for the courses with high correlation coefficient, and the courses with high correlation with the total score are also analyzed. Improving the score of the course can more effectively improve the total score. Finally, the LSTM method is used to predict and analyze students' performance, and compared with the results of SVM and logistic regression to verify the relationship between courses with high correlation, which provides a scientific reference for teaching management, teachers' teaching according to their aptitude and students' learning.

Air pollution has aroused widespread concern. Predicting air pollution can help people prepare for preventive responses. However, the existing air quality prediction methods are still in their infancy. In this paper, the LSTM (Long Short-Term Memory) sequence-to-sequence model based on the attention mechanism is applied to the prediction of air pollution. We used the meteorological and air pollutant observation data and pollutant discharge point monitoring data of Jinchang City, Gansu Province, and fused the meteorological and air pollution data of the neighboring areas to analyze the six basic air pollutants (PM2.5, PM10 , SO2, NO2, O3, CO), followed by model training and prediction evaluation. The results show that the prediction performance of the model for each air pollutant is different. For the prediction of short-term NO2, O3 and CO concentrations in the future, the A-LSTM-Seq2Seq model designed in this paper performs better; for PM2.5, PM10 and SO2, the prediction error of the model is larger than the previous three, because The observed values of these three are widely distributed and the data is discrete. In the future, it is necessary to improve the prediction performance from the perspective of improving the model. The model pays attention to all six basic air pollutants, and has certain practical value and application potential in the prediction of air pollution on the hourly time scale, and also has reference significance for the prediction of AQI.

In this paper, an airfoil aerodynamic prediction model based on neural network is proposed. In the traditional prediction method based on surrogate model, there are some shortcomings, such as over dependence on airfoil parameters, limited dimension of design variables, and complexity of strong nonlinear engineering design. As an alternative method, convolutional neural network (CNN) has been proved to have good ability of complex image classification, and is widely used in aerodynamic element modeling task. So in this paper, based on enough data sets, a convolutional neural network framework is constructed to replace the traditional CFD solver to predict large-scale nonlinear aerodynamic parameters. Using CNN, the aerodynamic parameters of airfoil at different angles of attack can be predicted（α)，The results of CFD solver are compared. The results show that the proposed method can maintain a high level of accuracy, and its efficiency is far better than CFD method.

Aspect-based sentiment analysis mainly processes natural language and generates its aspect term and corresponding sentiment. Previous studies focused on individual subtasks, did not make full use of large-scale corpus and dig out the semantic information of sentences, and did not consider the different contributions of different words in aspect-based sentiment analysis. In this paper, a unified sequence labeling BIO model is used, and the two subtasks of aspect word extraction and sentiment analysis are fused. By sharing a unified BERT model, and using the auxiliary training of domain-related documents, the attention is focused on aspect words and emotional words. Through experiments, a better effect was finally achieved.

With the development of computer technology, the overlapping computation with communication has been continuously developed and widely applied in the process of computer development from single machine to supercomputer. The degree of parallelization is getting higher and higher, and the data volume of communication between nodes is getting larger and larger. Communication overhead becomes an important factor affecting parallel computing. The overlapping computation with communication is an important way to solve the excessive communication overhead in parallel computing. This paper first introduces the development history of overlapping computation with communication, including hardware optimization compiler upgrade and task optimization scheduling. For the development of numerical weather prediction models, the wide application of overlapping with computation and communication can effectively reduce the communication overhead and improve the computational efficiency of the models. Due to the increasing parallelism of numerical weather prediction models and the increasing communication cost, more and more application scenarios of computation and communication overlap, which can be effectively realized by grouping computation and communication. There is more communication in the three-dimensional data redistribution of spectral mode, which is the direction of performance optimization by overlapping computation with communication in the future.

Aiming at the problem that the uniformly sampled point cloud is used for training in the three-dimensional object detection of RGB-D data, the accuracy of the point cloud in the actual scene is reduced, and the PointConv network is proposed to detect the loss of some key point cloud features. Firstly, by generating a 2D area proposal from the image and positioning it in the 3D point cloud data. Secondly, the 2D-driven 3D target detection is realized by the segmentation network based on PointConv and the 3D bounding box evaluation network. At last, evaluated on the KITTI and SUN RGB-D datasets, this method has real-time capabilities and surpasses the current detection levels of DoBEM, MV3D, Mono3D, Frustum PointNets ,and other methods. Experimental results show that compared with the previous state-of-the- art technology, our method improves mAP by 0.5% to 12.4%, and is 1-3 orders of magnitude faster.

Take-out business has become an indispensable part of the daily life of urban workers. According to statistics, the number of take-out riders in 2020 has exceeded 4.38 million. Facing the huge number of take-out riders in the market, budgeting its income level is a problem worth thinking about and urgently needs to be solved. In full consideration of riding safety and quality services and other factors, a reasonable design scheme of rider delivery time was developed, and a rider delivery model based on overtime penalty cost was established through order segmentation and objective function construction. Both static and dynamic order allocation commission pricing are considered as well as reward and punishment strategies that incorporate multiple factors to maximize riders' overall satisfaction. By establishing the constraint function model based on genetic algorithm, the fitness function of 27 generations is basically convergent, and the overall satisfaction of riders is the highest. Considering global extreme climate conditions, the objective function model based on extreme climate conditions is reconstructed. Through search, a feasible win-win solution is obtained. On the basis of reducing the delivery time of riders, the delivery volume is increased, the time of food distribution platform is shortened, and the cost of merchants is reduced. The VRP model is established to solve this problem, which is transformed into a problem of finding the shortest distance by replacing the unit price commission with the carrying quantity, and then the shortest distance is solved by Freud algorithm.

The remote tower system is mainly used to complete the monitoring and control tasks of the airport and temporary take-off and landing points in the unattended state. It has the advantages of low cost and convenience. The remote tower can reasonably assign flight command tasks to the controllers to reduce the risk of human error caused by boredom or sleepiness due to low work intensity. However, when controllers deal with flights from multiple airports at the same time, there will be the challenge of continuous multitasking, which is very easy to cause fatigue. Eye tracking is an important method to collect evidence of controller fatigue. This paper introduces a non-invasive detection system for evaluating the fatigue of controllers by tracking eye movement. An eye movement instrument is installed in front of the controller to monitor his eye state, collect eye movement data, and understand the fatigue state of controllers by analyzing and processing the data. In the experiment, 20 healthy subjects simulated the control for 2 hours on the control simulator, and the measurement characteristics of fixation time and pupil area were recorded by eye tracker. In order to obtain good data and fast calculation time, the effects of gaze time and pupil area of different controllers are evaluated and analyzed by using fuzzy k-nearest neighbor method. The results show that there are significant differences in the domain value distribution of pupil region between normal control and fatigue control. The results also show that the recognition accuracy of the knife cutting verification method is about 89%, which means that the method has great potential in real-time applicability and can accurately detect controller fatigue.

YOLO (You Only Look Once) of target detection algorithms have complex model structure and large computation. However, in practical application scenarios, they not only need to meet the task requirements of low latency and low power consumption, but also face problems such as difficult deployment and long development cycle of YOLO algorithms. Based on the flexibility of embedded CPU software design and the advantages of Field Programmable Gate Array (FPGA) parallel computing, combined with the structural characteristics of YOLO algorithm, a universal hardware accelerator software/hardware co-design method for rapid deployment of YOLO algorithm is proposed to solve the above problems. In the hardware acceleration design, multi-channel parallel internal and external storage interaction, model parameter reordering, fixed-point, multi-dimensional parallelism tiling and other acceleration optimization techniques are adopted. In the software driver design, multiple versions of YOLO algorithm are compatible to achieve rapid deployment. Zynq7000 is used as the hardware platform to implement a YOLO algorithm universal hardware accelerator system with low delay and low power consumption. The results show that the energy efficiency ratio of this system can be improved by 132x compared with PC CPU and 120x compared with embedded CPU while implementing YOLOv4-tiny algorithm.

With the rapid development of UAVs in recent years, UAVs are more widely used in all walks of life, and there is less discussion on the application of large UAVs. At present, the flight of large UAVs mainly relies on the joint operation of ground station operators and air traffic control systems, and is mainly based on experimental flights in non-converged airspace. Without reducing the acceptable level of safety, large UAVs enter the fusion airspace. Operation will be the inevitable development trend of airspace in the future. In this paper, the Reich model is improved. Based on the operation and error characteristics of large UAVs, a collision model between UAVs and manned aircraft is established. The existing Beijing-Shanghai route is selected for simulation, and the UAVs added to the existing airspace are calculated and compared. Before and after the collision risk, discuss and look forward to the feasibility of large UAVs entering the fusion airspace.

In order to effectively realize the display, control and data storage of multi-channel test data acquisition system, a display and control software of multi-channel test data acquisition system based on UDP protocol is designed and developed. This paper expounds the design and implementation of display and control software, and focuses on the key technology of multi-channel test data transmission and storage. Specifically, the software adopts multi-threading and message response mechanism to ensure the reliability of the software system and the efficiency of data transmission. At the same time, the software adopts the data-packet check confirmation mechanism to overcome the data loss and disorder that may occur in the data transmission and ensure the reliability of the test data. The display and control software has been successfully applied to a multi-channel test data acquisition equipment, which provides data support for the positioning and analysis of targets.

At present, many text classification model methods integrate Convolution Neural Network (CNN) and Long Short-Term Memory (LSTM), although they have achieved good performance, they have not fully Taking advantage of the characteristics of these two models, this paper proposes a text classification model CR-IAN model based on interaction-level attention fusion of multi-channel CNN and Bi-LSTM. First: The local feature vectors of different latitudes of text are extracted through multi-channel CNN; At the same time: Pass the text matrix through the Bi-LSTM layer to obtain the result vector of the Bidirectional semantic learning of the text. Then: calculate the interactive attention according to the bidirectional output matrix and the local feature vector matrix. This can fully consider the characteristics of the text with semantics, the model can learn more text information. Finally: Pass the interactive attention through the fully connected layer and use the softmax activation function to get the classification result.

To realize the real-time visualized monitoring on the manufacturing in digital workshop and to make breakthrough in the key technologies of digital twin workshop setup and operation management, this article brings forward an integrated management and control system framework and technical route which is based on a real-time data driving and used for 3D digitalized shop. In this article, a 3D visual twin environment is established by means of Virtual Reality (VR). Through Virtual Reality (VR) engine, the digital workshop is modularized and generally utilized, and set of integrated management and control system for a digital twin workshop is customized. By taking the digital workshop of an aircraft manufacturing enterprise as an example, processes of equipment operation status, material transfer and scheduling, and remote control in the virtual shop is virtually simulated by such integrated management and control system.

Motion blur and ambient noise are the main reasons that affect quick response (QR) code recognition. In this paper, we propose a novel deep learning approach to deblur the QR codes and realize the effective recognition of deblurring QR codes by using generative adversarial networks (GANs). We estimate the blur kernel and ambient noise of the blur QR code in the dataset using GANs, so as to realize the transformation from the blur QR code image to the sharp image. We also propose an expansion method of QR codes dataset, and achieve better generalization performance of the model. The experimental results show that our approach can effectively estimate the blur kernel and ambient noise that can realize the deblurring of QR code.

The intelligent monitoring technology of grain depots has been constantly developing and improving. In order to reduce the labor intensity of staff and improve the intelligence and efficiency of grain depot monitoring, this paper proposes a grain depot gate monitoring system based on the Internet of Things cloud. Install an access control alarm in the warehouse for data collection. The access control triggers an alarm when any situation such as demolition, door and window movement, or RFID device card abnormality is triggered, and sends alarm instructions and alarm videos to the server; the server is based on the MQTT protocol. After analysis and processing, it will be displayed to supervisors via WEB; supervisors will check and process one by one according to the displayed alarm information.

Aiming at the nonlinear fitting problem of the connection surface during the bolt tightening process, a model of deformation coordination relationship of the bolt connection is established based on the elastic assumption. The model is used to analyze the variable stiffness characteristics of the bolt connection, reveal the internal mechanism of the bolt connection stiffness during the tightening process. The definition of the boundary point—the fitting point of the nonlinear segment and the linear segment of the bolt torsion-tension relationship curve is given, and the fitting point identification algorithm is proposed. The effectiveness of fitting point identification is verified through case analysis, and a new solution is provided for bolted connection process control.

According to the requirements of simulation support software of electronic sand table hardware, this paper proposes a design method of electronic sand table support software. The software is composed of three-dimensional battlefield environment simulation module, military standard, arrow and other plotting modules, a large number of dynamic target loading modules, situation simulation module and real-time communication module. After practical verification, the software can be used for three-dimensional situation generation, rendering and plotting of electronic sand table, supporting scenario generation, three-dimensional simulation, countermeasure deduction and other activities, and the application effect is good.

To improve the safety and automation during the spanning construction of transmission lines with voltage levels of 110kv and below, an intelligent spanning frame monitoring system based on PLC and configuration software is developed. The system performs real-time monitoring and early warning of various parameters such as height, inclination, and wind speed at the construction site of the spanning frame during the construction process, and controls a series of actions such as the synchronous lifting and spanning of the crossing pole, and the insulating mesh distribution and net collection. The system adopts the distributed structure of a monitoring center and four substations to realize distributed hybrid networking by combining wireless communication, serial communication, and 4G communication.

The ultraviolet (UV) spectrometer installed on the balloon platform is a cost-effective instrument to observe the UV spectrum of near space, which can meet the needs of biological evolution and planetary material exchange research. A control system of an enhanced UV spectrometer is designed in this paper. Firstly, the framework of the control system is introduced, and the functions of each module are described. Then, the solar angle sensor and automatic air pressure balance device in the enhanced detector module are introduced, making the detector operate safely and reliably in solar radiation and low-pressure environments. Next, the temperature controller is introduced, which can automatically adjust the heating power through the PID algorithm in the embedded chip according to the data collected from the light body and the atmosphere. Finally, the automatic detection mode is described. Among them, the controller can automatically adjust the system gain and integration time according to the spatial ultraviolet radiation intensity information, thus maximizing the efficiency of data. The flight test results of Honghu HH-20-7, HH-20-9 and HH-19-7 show that the spectrometer works normally in the flight experiment. Compared with the ultraviolet radiation in the MODTRAN database, the ultraviolet spectrum data of adjacent space obtained by the spectrometer has the same consistency, but provides a lot of detailed information.

Complete and available data is of great significance for improving the theoretical line loss calculation in the low-voltage transformer area. However, with the upgrading of equipment, the electrical data presents the characteristics of small granularity and high complexity, resulting in insufficient line loss calculation accuracy. This paper proposes a missing data reconstruction method based on Kmeans and GBDT combined model. Since it is difficult to unify the reconstruction models of different types of data, the data set is clustered and divided by Kmeans, and GBDT is used for training respectively. During the test, the corresponding GBDT model is used for reconstruction according to the sample category. The simulation results show that the proposed method is suitable for actual data and has higher reconstruction accuracy compared with the traditional mean filling, KNN and decision tree methods. The proposed method can reconstruct the missing of multiple datasets and has good generalization.

Rotating machineries account for 80 percent of the total large mechanical equipment. To optimize their work efficiency, research of the simulation on improving the accuracy of life prediction which is about key components in rotating machineries was conducted. Firstly, Confidential Value (CV) can quantify working condition. And then make the training set of the original CV data be in Wavelet Decomposition (WD). Next, the data transformation and accumulative integral can improve GM (1,1) by the optimization of the smooth ratio and background value; improved GM (1,1) is in combination with Adaptive Neuro-Fuzzy Inference System (ANFIS) for new prediction model to predict the CV. Ultimately, improved WD-GM-ANFIS is compared with GM(1,1), ANFIS and LSTM network in three performance indicators, meanwhile, contrast improved WD-GM-ANFIS with Full Convolutional Layer Neural Networks (FCLNN) algorithm and Unscented Particle Filter (UPF) algorithm in failure time prediction. Three indicators' averages of the new model are all better than the other three single models; failure time prediction of the new model is more precise than the two improved algorithms.

In recent years, global optimization methods based on gradient method have been studied. The applications of genetic algorithm and simulated annealing algorithm based on search technology in global optimization problems have attracted more and more attention According to the parameters and requirements of problem d in 2021 Chinese college students' mathematical modeling competition, this paper applies simulated annealing algorithm to solve the optimization problem The hybrid algorithm of simulated annealing algorithm and BFGS algorithm is used to solve the nonlinear optimization problem On the basis of solving the nonlinear optimization problem, K-means algorithm is applied to solve the combinatorial optimization problem of linear programming problem; In the process of billet cutting, aiming at the three cases given in question D, the goal of meeting customer requirements and minimizing cutting loss is achieved.

This paper presents a design method of system simulation electronic sand table software. The software consists of display effect and touch action control module, interface communication module, entity state information processing module, data filtering and secondary development interface module. Based on the software, the electronic sand table can realize the relevant functions of system simulation, accept external simulation information and support users to generate typical scenarios.

In the process of equipment simulation, the environment has an important impact on the equipment Strike ability. According to this demand, this paper proposes a simulation method and software for the interaction between the environment and the equipment Strike ability, including the simulation model of the interaction between the battlefield environment and the tracking target, the simulation model of the interaction between the battlefield environment and the fire control system, the simulation module of hitting the armored target The simulation model of line of sight shooting entry ray, the simulation model of projectile hitting the target and the simulation model of the influence of meteorological environment / electromagnetic environment on shooting accuracy. This method can study the interaction between battlefield environment and equipment strike action, and achieved good results.

With the development of artificial intelligence, intelligent robot has gradually become one of the current research hotspots. Among them, path planning is the key technology of intelligent mobile robot. The planning to enable the mobile robot with a safe and effective path efficiently has always been a challenging research field. This paper designs a mobile robot path planning system based on machine vision, establishes the overall framework of the path planning system, and focuses on the global path search algorithm of ant colony algorithm to realize the optimal path in mobile robot space. According to the basic principle of ant colony algorithm, the workspace environment of mobile robot is established, and the relationship curves between the core parameters and ant colony algorithm performance are obtained by software simulation. On this basis and under the condition of the optimal combination of parameters, the efficient path planning based on ant colony algorithm is realized. The path planning algorithm can provide a corresponding solution for the path searching problem of mobile robot.

The line spectrum detection method based on coherent power spectrum estimation is one of the important research directions of weak line spectrum detection. Aiming at the problem that the phase compensation in the line spectrum detection is not accurate due to the limited frequency search accuracy of the existing coherent power spectrum estimation methods, a line spectrum coherent detection method based on cross-spectral frequency measurement (CDCSFM) is proposed on the basis of the CAPSE method. The cross-spectrum method can more accurately estimate the line spectrum frequency, and then perform more accurate phase compensation and further improve the coherent processing gain. The simulation results and sea test results show that the coherent processing gain of the CDCSFM method has a certain improvement compared with the CAPSE method.

In the process of industrial process gradually moving towards automatic production, wood grinding always re-quires a lot of manpower due to the particularity of its process flow, and the dust and noise generated by wood grinding have caused certain damage to the physical and mental health of workers. In this paper, a wood collaborative grinding bench with high efficiency, safety and friendliness is proposed, in which the double collaborative arms are used as the main grinding body, and the manual drag is used as the direction of grinding path planning. The safety of human-machine cooperation is improved by the force feedback sensor of the collaborative arm itself, and the dust absorption device fixed at the end of the collaborative arm is used to reduce the environmental pollution.

In response to the notice on accelerating the comprehensive reform of agricultural water price put forward by the state, we will strengthen the comprehensive reform of agricultural motor well irrigation water price, and guide agricultural production to save water and irrigate scientifically. This paper takes the calculation of agricultural well irrigation water price, and puts forward the calculation method of well irrigation water price based on BP neural network. The calculation of irrigation water price of agricultural motor well applies to the intelligent power consumption device of agricultural irrigation. Guide farmers to save water and reasonable irrigation, and provide technical support for Rural Revitalization. Use the stage water price charging mode to reasonably control the water consumption according to the land area of each household.

Aiming at the problems of large search range, long calculation time and low efficiency of traditional A* algorithm in highrise fire rescue path exploration, an improved bidirectional A* algorithm suitable for modern fire rescue path exploration is proposed. Firstly, the bidirectional search mechanism is combined to effectively avoid the influence of excessive fluctuation of estimated value on search speed and accuracy. Secondly, the multi-path pruning strategy is used to narrow the search range of each step and reduce the traversal of useless nodes. Finally, the Floyd algorithm is used to process the generated path to make it smooth. Considering the subjective initiative of people, it is more in line with the actual situation. By simulating the traditional A* algorithm and the improved A* algorithm on the same map, the effectiveness of the improved A* algorithm and the practicability of exploring species in modern fire rescue path are verified.

With the increasing competition in the hot pot market, Haidilao must understand the dining needs and consumer psychology of customers, in order to stand firm in the market competition. This paper takes Chengdu Qunguang Plaza Haidilao branch as the research object, and obtains the post-meal evaluation data of the branch on the platform of the public comment platform published online. In this paper, after basic pre-processing operations such as removing duplicate and cleaning of the data, the distribution of customer scores is analyzed to obtain customer satisfaction with the restaurant. Secondly, the text data of customer comments are analyzed by text word segmentation, word frequency statistics, word cloud data display and other methods. Then, the LDA theme analysis model is used to conduct theme analysis on the review data. The experimental results show that the Haidilao branch has advantages in dishes, pot bottom and service. There are deficiencies in food hygiene, environment and queuing. According to the model results, this paper puts forward the corresponding five suggestions. Finally, in view of the shortcomings of this paper, the improvement plan and future prospects are put forward.

Driven by the "civil aviation power" strategy, the pace of development of air transport industry is accelerating, the growth rate of air traffic flow is accelerating year by year, and the demand for pilots is increasing. The pressure of civil aviation pilot training task is increasing year by year. In the case of limited available resources in the airspace, the increasing flight training will inevitably lead to more and more serious flight conflicts, greatly increase the load of controller deployment conflicts, and affect the safety of control operation. In the control operation, the basic unit of civil aviation controller's work is a control sector, and its workload is mainly reflected in air command and inter sector handover. Among them, air command deployment flight conflict is a very key work. Using the real-time ADSB data of the training airport, this paper applies the artificial intelligence technology to the problem of control conflict allocation. The controller on duty obtains the aircraft flight status information in the sector by observing the air traffic control radar screen, analyzes the current overall operation situation in the sector and predicts the possibility of conflict in the future, After determining that there may be a conflict, issue deployment instructions to the conflicting aircraft according to the control operation rules and control experience.

Landslide is one of the world's most destructive natural disasters, in recent years, due to extreme weather and human engineering activities, such as shift caused by the landslide disaster events brought serious harm to the society, therefore, scientific prediction of landslide disaster in order to realize, disaster prevention and reduction is of great scientific significance, based on the guizhou pan city as research area, with the support of vast amounts of big data, The traffic road and residential building information were obtained by high resolution remote sensing image, and the landslide risk and disaster bearing body information were integrated, and regional landslide disaster risk analysis was realized by risk matrix. The results show that the landslide risk areas are mainly concentrated in urban areas, township streets and traffic areas with high economic value, covering an area of 214.2km2, accounting for 5.3% of the total area of the study area. The prediction results are consistent with the actual situation in the study area, providing technical support for disaster prevention and relief.

In recent years, with the rapid development of the Internet, the emergence of deep learning has attracted widespread attention in academia. Based on the scientific nature and strong operability of deep learning, it has changed the random sampling data in the past, with higher accuracy and reliability. Using deep learning to conduct risk management research on military sports training, with the help of the large capacity and accuracy of deep learning, scientific research on the risks existing in military sports training, assessing risks, and coping with risks, thereby reducing training injuries and improving training levels. It is characterized by a large number, variety, high speed, and low value density. Using deep learning for risk management in military sports training can optimize information collection procedures, identify risk sources with massive information, solve information processing speed problems, scientifically assess risks and propose risk response measures, which has important practical and innovative value.

A research model of plasma flow field and electromagnetic (EM) wave around the hypersonic vehicle for solving the blackout problem is developed in this study. Based on the computational fluid dynamics simulation, the plasma flow fields are derived and serve as the background conditions in EM wave transmission. The complicated behavior of blackout is clarified at several altitudes. Moreover, the comparations of VHF (Very-High-Frequency) blackout data between the flight tests and numerical simulations verify the validity of numerical method. In addition, the external magnetic field is applied to this model. The effects of RHCP (right hand circularly polarized), LHCP (left hand circularly polarized) waves and magnetic induction are studied, and the results show that external magnetic field can effectively reduce the transmission attenuation, and the blackout problems can be solved when the proper strength of the magnetic field is applied.

The application of artificial intelligence technology has improved the production efficiency of agriculture and forestry, and the huge demand for fruit picking robots has driven the booming development of related technologies. To address this situation, this paper designs a fruit target detection and segmentation method based on deep learning technology. This paper firstly introduces the necessity of fruit target localization technology, then introduces the technical foundation including neural network, neuron model and activation function, convolutional neural network, etc. Finally, the research and design of fruit contour recognition and segmentation method is carried out and summarized. The method can provide pre-processing methods and data basis for subsequent fruit target recognition, and lay a certain foundation for promoting the development and design of fruit picking robots.

In the era of big data and Internet, teaching informatization and grade management informatization are more and more widely used, How to quickly and efficiently import various grades generated in the teaching platform into the grade management platform has become a concern of many schools and teachers. The FileReader object in front-end development can quickly read various file contents into web pages, and after that, you can complete various operations in the web page according to the read file data. Based on the function of FileReader object, a score import function is designed to import Excel data into web pages, which can be displayed, sorted and modified in tabular form. The design and development of score import function effectively improves the efficiency of score entry, ensures the correctness of input data, and provides convenience for teachers.

Enhanced Loran (eLoran) system plays an important role in the national positioning, navigation, and timing (PNT) fields. The information demodulation capability of its user receiving end also directly affects the application coverage of the system. Using the existing arctangent and matched filter demodulation algorithms, we propose a novel algorithm based on log likelihood ratio (LLR) information, and then analyzing the implementation process of these three algorithms in detail and simulating the system performance difference in an additive white Gaussian noise (AWGN) environment. The results showed that the performance difference of the three algorithms was between 0.5 dB and 1 dB, and the proposed algorithm was 0.5 dB better than the arctangent demodulation which is more widely used in engineering practice. Simultaneously, the proposed algorithm can be used to provide an algorithm basis for the study of the eLoran system to switch to modern channel coding methods with LLR information as the decoding input.

The accurate detection of targets on the road is of great significance (for the big data of road system). However, the traditional detection suffers from missed detection and false detection. In this work, an improved algorithm was proposed based on Yolo-v3 algorithm. The replacement of traditional K-means clustering algorithm with K-means++ algorithm endowed the system a more suitable size of detection frame. Meanwhile, the detection sensitivity of small or incomplete targets was obviously improved with the utilization of cross entropy loss function. The detection experiments on vehicles and pedestrians were performed in two actual situations, normal situation and congested situation. The average detection accuracy was increased by 2.46% with the improvement of clustering algorithm, and the accuracy was increased by 2.08% with the improvement of loss function. The average accuracy was increased by 3.56% with the combination of K-means++ algorithm and cross-entropy loss. The experiments results showed that the combined algorithm possessed a significant improvement in average accuracy, recall and detection speed. This research provides new ideas for the establishment of high-precision algorithms for road dynamic target detection.

With the rapid development of computer technology, algorithm technology is also constantly innovating. The swarm intelligence optimization algorithm has gradually become the most important class of algorithms because of its simple and easy-to-implement characteristics. It has been widely used in optimization problems and provides new ideas for solving complex engineering problems. The latest swarm intelligence algorithms proposed at home and abroad are selected for analysis, including Arithmetic Optimization Algorithm (AOA), Wild Horse Optimization (WHO), Dingo Optimization Algorithm (DOA), and Artificial Hummingbird Algorithm (AHA). The paper uses 12 general benchmark functions and part of CEC2017 test functions for experimental simulation, and further compares the experimental overall performance of these algorithms in phrases of convergence speed, accuracy, and stability. Finally, the future development of the swarm intelligence algorithm is discussed.

Quickly predicting the remaining useful life of bipolar transistors is an effective means of assessing their health and reliability. Aiming at the problems of high modeling difficulty and poor fitting accuracy in the traditional physical model method and data fitting method, this paper proposes a reliability life prediction method for bipolar transistors based on deep learning. Firstly, the experiment is carried out with the accelerated storage experimental data of transistors in the Microelectronics Reliability Laboratory of Beijing University of Technology, and the reverse leakage current I_CBO is selected as the failure sensitive parameter, and a reasonable failure judgment is set. The lifetime of the sample is calculated; Secondly, the average lifetime under three common distributions is given, and combined with the Peak temperature and humidity model, the storage lifetime of the transistor under natural storage conditions is extrapolated; finally, the mean absolute error (MAE) and mean square error (MSE) are selected as evaluation functions, and the life prediction results of the deep learning model and the data fitting method are compared. The results show that compared with the data fitting method, the MAE of the life prediction results of the deep learning model decreases by 82.00%, and the MSE decreases by 98.16%, which verifies the effectiveness of the method.

In order to improve the picking efficiency of warehouses, a multi- objective warehouse storage allocation optimization method based on NSGA-II is proposed for the movable shelf storage system. First, according to the characteristics of the picking operation, a integer number chromosome coding method combining the storage index number and the goods category number is designed; Secondly, a multi-objective optimization model is established with the picking efficiency, shelf stability, goods correlation and the picking channel busyness. Then designed by the number of iterations to determine the adaptive crossover and mutation probability operators; Finally, the simulation experiments show that the improved adaptive NSGA-II algorithm in this paper has significant improvements in running time, DELTA, channel busyness, which can effectively solve the problem of warehouse storage allocation optimization.

With the development of Industrial 4.0, the popularity of information technology (IT) systems in industrial control systems (ICSs) has brought great cyber-attack risk. The proactive defense technology, an important defense approach, plays a vital role in ICSs. This paper introduces the definition, taxonomy and application of proactive defense in ICSs. We divide the existing ICS proactive defense approaches into three categories and summarize their technical characteristics. Finally, the application prospects and open issues of proactive defense in ICS are summarized.

Limited by the shielding effect from the pipeline structure and materials, pipeline robots usually rely on sensors such as odometers and inertial measurement units (IMUs), resulting in significant error accumulation effect and difficulty in ameliorating positioning accuracy. In this paper, we present an internal positioning method in the pipeline based on data fusion and point cloud registration to improve this problem. First of all, the Extended Kalman Filter (EKF) fuses the odometer and IMU data, acquiring the estimated pose's initial value and calculating the derived position. Secondly, the conditional filtering of the known pipeline point cloud is carried out using the derived position to solve the inaccurate problem caused by the significant difference in the number of points between the two groups in point cloud matching. Eventually, the improved iterative closest point (ICP) algorithm is applied to the filtered point cloud and the data of the two line-laser sensors, obtaining accurate pose information of the robot. The proposed method enables the robot to locate only by its sensor data, which is especially suitable for use in pipelines where the environment is harsh and wireless signals are often not received. In our experiments, we first demonstrate the effectiveness and necessity of point cloud filtering, and then compare our algorithm with the traditional intralocalization algorithms under the Gazebo/ROS simulation platform. The results show that the positioning accuracy of our method is better than the traditional internal positioning method.

Travelling Salesman Problem (TSP) is a famous Problem in operational research. many problems in real life can be reduced to this kind of Problem. In practical problems, sometimes exact data cannot be obtained due to many conditions, yielding only a general range. In this case, this paper presents an uncertain TSP problem with grey numbers, and constructs a dynamic programming model for the uncertain TSP problem with grey numbers. Then, combined with the dynamic programming method, an algorithm for solving the uncertain TSP problem with grey number is given by using the grey number possibility degree and the ranking method. Finally, use case calculations to verify the correctness of the algorithm.

Problems such as hunchback and myopia caused by improper sitting posture are becoming more and more serious. It is important to establish posture detection and give proper guidance and supervision. In this paper, on the basis of multifrequency lighting intelligent lamp, the introduction of artificial intelligence technology, from the perspective of machine vision proposed a set of posture correction system design based on face recognition technology. The camera adopts a rotatable structure, and its rotation is realized by STM32F103 driving steering gear. The sitting position information obtained by the camera is transmitted to Jeston Nano through CSI interface. Jeston Nano is the core of image processing. OpenCV + Dlib facial feature technology is used to collect the user's facial feature data. Study the user's work, study time, sitting posture information, to achieve regular rest and correct sitting posture reminder. The analysis results show that the intelligent desk lamp to lean forward, lean back, left, right and other non-standard posture recognition accuracy of more than 92%.

Cardiovascular diseases are the leading cause of death worldwide. Numerical simulation is an important way for studying cardiovascular pathology and testing cardiovascular devices. The arterial physiological characteristics of the human body vary greatly at different exercise levels. It is difficult to accurately simulate the physiological characteristics of arteries in different exercise states. This paper presents an accurate simulation method of motion states using the Proximal Policy Optimization algorithm based on a pneumatic ventricular numerical model. The results show that the proposed method can reproduce arterial pressure and flow at different motion levels with high accuracy. The simulation accuracy is more than 98%. The proposed accuracy simulation method for exercise states can be used for studying cardiovascular physiopathology and testing of cardiovascular devices.

Deep Reinforcement Learning (DRL) has been successful applied to a number of fields. In recent years, many scholars have used the DRL algorithms to solve a classic combinatorial optimization problem, i.e. Vehicle Routing Problem (VRP). The scale of the problems that are solved in the literatures is small, thus it is difficult to apply the algorithm into practice where there are many large-scale instances. To solve large-scale VRPs by using DRL, this paper proposes a pre-training mechanism for online shared networks. The graph pointer network under the multi-head attention mechanism is trained in the dual-network reinforcement learning mode. The trained model can be applied to large-scale VRP with 100/300/500 customers within a certain time. The experiments reveal that our algorithm can obtain good solutions in terms of solution quality and offline solution efficiency.

This article mainly focuses on the cooperative transportation between unmanned aerial vehicles (UAVs) and trucks, i.e., the optimal strategy problem under emergency conditions of multiple truck-UAVs. Under the traditional UAV-truck distribution mode, there are few transportation strategies considering the transportation of emergency supplies, and there is a lack of research on the scheduling of multiple truck-UAV combinations. In order to fill this gap, we propose the FSTSP model with a multipath genetic algorithm. It obtains the optimal solution of visiting customer locations under the condition of the limited number of truck-UAVs. Meanwhile, the minimum number of truck-UAVs for delivery tasks under timelimited conditions is calculated. Such a scheme helps to deliver goods to customers at a lower time cost and improves economic efficiency. The experiments validate the effectiveness of the proposed model.

In order to fully apply the data, mine the relevant information in the existing data, judge and analyse the equipment faults that have occurred, and provide intelligent solutions, this paper studies the equipment similarity recommendation algorithm. Referring to the similarity of spatial coordinate points, this paper expands the equipment attribute to spatial coordinate points, and gives different weights to the attributes according to the different influence factors, the attribute domain space is further expanded, and the parameter calculation is carried out by means of matrix analysis and linear fitting. The algorithm is applied to the actual production management system and achieves good results.

Safe Deep Reinforcement Learning (SDRL) has been studied in many domains. In previous work, based on Augmented Lagrangian Method, we proposed DDPG-ALM algorithm in an industrial oil pump case study to learn a safe controller. In this paper, we apply our algorithm to a more general platform, Safety Gym, to test the performance of our algorithm. To solve the issue of sparse cost in Safety Gym, we propose the notion of State Shaping which aggregates safe states to pay more attention to the unsafe states. Furthermore, to pay more attention to safety-related experience, we combine the advantages of Prioritized Experience Replay and Double Replay Buffer and propose a method called Safe Prioritized Experience Replay (SPER). Our experiments show that, State Shaping and SPER can deal with the sparse cost well and provide safer policies. Ablation studies show that State Shaping reduces average episode cost of all evaluated trajectories from 2.67 to 0.45 and SPER reduces from 0.45 to 0.20 while PER only reduces to 0.30. Moreover, our DDPG-ALM based algorithm not only automatically finds a balance between reward and cost, but also converges faster than other existing benchmark algorithms.

The application of RGV system not only ensures the safety, but also improves the operation efficiency. Aiming at the dynamic scheduling problem of intelligent RGV and using the first call first processing principle and the proximity principle, this paper establishes the RGV dynamic scheduling model based on tabu search algorithm, and obtains the scheduling strategy and solution algorithm in three cases: the material processing operation in one process and two processes, and the failure in material processing, Then the model is tested and the operation efficiency is calculated. It is found that this model has good practicability.

With the development of the Industrial Internet, technological innovation and iterative evolution in the industrial direction are getting faster and faster, and the amount of data generated by industrial equipment is also increasing explosively. The Industrial Internet itself is based on data and supported by intelligence. However, with explosive data, there is an irreconcilable contradiction between the simplicity of data storage and data collection and the scale of data. The system cannot fundamentally solve this contradiction, but through the components of the system itself, data collection can be made more portable under the premise of a large amount of data and ensuring data reliability. This system relies on the data format of industrial machinery and equipment, that is, structured time series data, which is stored in a time series database. In order to ensure data security, the system also uses distributed ceph block storage as the carrier of database storage to ensure the reliability of data storage.

With the rapid advancement of the scientific and technological revolution and the rapid development of information and communication technology, the fields of wireless communication and artificial intelligence have entered a new research stage, which has brought a revolutionary impact on the global industrial industry and gave birth to the birth of the industrial Internet. With the high attention paid to relevant topics at the national level such as German industry 4.0, American industrial Internet and the national action program made in China 2025, the research on industrial Internet and related fields has attracted extensive attention all over the world. Made in China 2025 points out that we should actively promote the transformation and upgrading of industrial enterprises, rely on the industrial Internet platform and realize the databased development of industrial manufacturing industry with the help of big data analysis. Based on this, this paper will discuss the corresponding policy background and current application status of industrial Internet platform and big data technology, and explore new ideas and new paths for the construction of data analysis system based on industrial Internet platform through data visualization and data analysis system.