Abstract:For the traditional trajectory similarity calculation method, the measurement effect is not good, and the similarity measurement is difficult to achieve good results when the time series data is excessively distorted . Based on the accuracy and real-time requirements of many practical applications, this article is based on the dynamic time warping , combined with the idea of trajectory translation and the idea of global variable constraints, and gives an improved dynamic time warping algorithm through algorithm optimization and parameter analysis. Numerical experiment results show that the improved algorithm has a recognition rate of 90% in the measurement of trajectory similarity, which is an increase of 41.25% compared with the classic algorithm, and the measurement accuracy is significantly improved. Furthermore, as a trajectory similarity measurement function combined with spectrum clustering algorithm, it is applied to trajectory data clustering analysis. Experimental results of simulated trajectory data shows that clustering analysis based on the improved algorithm can clearly distinguish trajectory clusters and the clustering effect is ideal.

Abstract:In order to reduce the interference of the on-line calibration of the ultrasonic flowmeter by the on-site spoiler, the large eddy simulation method is used to study the influence of the installation position along the flow direction and circumferential direction on the measurement accuracy when there is a 90° elbow upstream of the ultrasonic flowmeter, so as to obtain the best installation position of the ultrasonic flowmeter. The results show that the elbow causes the downstream flow pattern distortion. The farther the ultrasonic flowmeter is from the elbow, the lower its sensitivity to the circumferential installation angle, and the maximum error of the correction coefficient decreases from 0.11 to 0.03; the correction coefficient fluctuates with the change of the circumferential installation angle, and the higher the flow velocity in the pipe, the smaller the amplitude of the correction coefficient fluctuation, and the maximum fluctuation amplitude is reduced by 20.5%. In addition, when the circumferential installation angle is 0° and 180°, the sensitivity of the flowmeter to the installation position is lower, the correction coefficient is closer to 1.0, and the maximum error is within 0.04. The results are beneficial to improve the on-line calibration accuracy of the ultrasonic flowmeter under complicated flow conditions caused by the on-site spoiler.

Abstract:In view of the low detection efficiency of waste plastic bottles, limited environment and limited color recognition by current waste sorting algorithms, this paper proposes an effective method for identifying and locating waste plastic bottles, which extracts original pictures based on high-pixel images. , Through the shallow enhancement feature of the YOLOv3 algorithm, the target in the picture is subjected to a series of convolutions to obtain different features, and each detection branch is input for detection, and the feature maps of different scales are processed by the k-means clustering algorithm as anchor boxes, and the position is used Predict to achieve the final recognition and location detection results. Through model testing, the YOLOv3 algorithm is superior to other algorithms in terms of recognition speed and complexity of the algorithm. The average recognition accuracy is 90%, the detection time is within 0.4s, and the positioning accuracy is about ±5cm. It proves the effectiveness and practicability of this algorithm for target detection of waste plastic bottles in complex environments.

Abstract:In order to study the application of flexible graphite grounding material in power distribution systems, a finite element simulation model of flexible graphite grounding device is constructed, by which the grounding resistance of a group of 10kV power distribution system flexible graphite grounding devices is calculated. It is proposed that the rectangular frame structure should be preferentially used in urban distribution networks to reduce grounding resistance by combining with the inherent characteristics of flexible graphite grounding material and common resistance reduction methods. The temperature rise of the flexible graphite grounding device under short-circuit fault and lightning fault has been simulated and calculated. The test device can operate normally under short-circuit fault conditions, but the local temperature rise is over standard under the lightning fault that once in 30 years.The research results have important reference value for the design, selection and application of flexible graphite grounding devices in 10kV urban distribution networks.

Abstract:In view of the limitations of traditional nondestructive testing methods for wind turbine blades, this paper proposes an electromagnetic ultrasonic testing technology. The technology has the advantages of non-contact, no coupling agent and strong environmental adaptability. In this method, according to the principle of electromagnetic ultrasonic detection, the research objects are 13mm non-destructive blades and defective blades with bubbles, inclusions and degumming. Firstly, the electromagnetic ultrasonic transducer with square permanent magnet and track coil is built by using COMSOL. Then, the stress nephograms and waveforms of various blade models are obtained through the electromagnetic ultrasonic transducer. Finally, the defect waveform is compared with non-destructive waveform to analyze the defect condition of the blade. The results show that the simulation model can detect three kinds of defects inside the blade, which proves that the relationship between the defect echo and the defect radius is basically linear, and the relationship between the bottom wave and the defect radius is negative, and verifies the feasibility of the electromagnetic ultrasonic finite element simulation analysis of the wind turbine defects.

Abstract:ECG plays an important role in the diagnosis of arrhythmia, but ECG signal is easily disturbed by various noises, which will change the shape of the signal and affect the accurate diagnosis of arrhythmia.This paper proposed an improved empirical wavelet transform denoising algorithm of ecg signals, divide the spectrum according to the frequency domain characteristics of ecg signal, and get rid of the noise by calculating the zero rate and the correlation coefficient of empirical mode functions, use wavelet threshold method to remove the residual noise. The experimental results show that the baseline drift and EMG interference are effectively filtered. SNR of the denoised signal is 17.7382db and RMSE is 0.0203, which is better than other common methods. The improved empirical wavelet transform algorithm has obvious noise reduction effect and can effectively restore the original ECG signal features.

Abstract:In the research and development test process of missiles, rockets and other weapons and equipment, it is necessary to collect, store, and read back and analyze the data generated during the test. For traditional technology, it is impossible to simultaneously perform high-speed data from two data sources with large speed differences. For the problem of data storage, a dual FLASH data logger based on FPGA is designed. Two 8GB NAND FLASH parallel storage solutions are used to store 400Mbp Gigabit Ethernet data and 10Mbps PCM data at the same time. FLASH adopts the Multi-Plane method for data reading and writing. Due to the inherent characteristics of FLASH, the stored data will cause errors. Design Hamming code check code error correction scheme corrects the error code. Test and data analysis results show that the comprehensive data write data rate can reach 410Mbps, the readback data rate can reach 310Mbps, and the data logger's read and write test error rate is 0, which meets the data storage test requirements of weapons and equipment, and can be stored stably Data generated during the test of weapons and equipment in harsh environments such as high temperature and high impact.

Abstract:Piezoelectric actuators have the characteristics of fast response, large mass ratio, high rigidity, etc., and have been widely used in the field of nanotechnology. As a kind of inherent nonlinear characteristic, hysteresis characteristic greatly affects the performance of hysteresis control. This paper proposes a nonlinear model predictive control (NMPC) method to solve the displacement tracking problem of piezoelectric actuators. First, the RBF neural network is used to realize the "external input nonlinear autoregressive moving average" (NARMAX) model of the piezoelectric actuator; secondly, the NMPC principle is used to transform the tracking control problem into an optimization problem, and then the gradient descent algorithm is used to solve it. In order to verify the effectiveness of the proposed modeling and control methods, MATLAB and COMSOL simulation experiments were carried out. The results show that the proposed RBF prediction model has satisfactory accuracy, the maximum absolute error between the expected displacement and the actual displacement obtained by the NMPC method tracking is 0.016um, and the average absolute error is 0.0121um, which has high accuracy.

Abstract:In response to the traditional RRT algorithm in the complex environment, the mobile robotic arm avoidance planning is low, and the current work demand problem cannot be met, an improved RRT avoidance path algorithm is proposed. By introducing an obstacle factor, the number of samples is reduced, by constraining the new node generation mode, the growth direction of the new node is always toward the target point, which speeds up the search speed. Adopting a gentle-step strategy, effectively avoiding oscillating phenomena near the target point. The planned path is smoothed by simplifying the path and combining with local cubic B-spline curve, the length of the path is shortened, and the stability of the robot arm is improved. MATLAB simulation results show that the average path length of the improved RRT algorithm is shortened by 20%, the average number of sampling points is reduced by 62%, and the number of turning points is reduced by 96%. The convergence speed is fast and the time is optimal, which can meet the obstacle avoidance path planning of the mobile manipulator in special environment, and has good feasibility and effectiveness.

Abstract:Aiming at the problem that the traditional K-means clustering algorithm was difficult to adapt to underdeveloped overseas cities because of chaotic planning and unevenly distributed power loads, a method based on improved K-means clustering algorithm for the distribution network planning of underdeveloped overseas cities was proposed. Firstly, the impact of capacity margin on partitions was considered, and the weighting factor was added to improve the Euclidean distance. Secondly, the selection of power supply units were considered by the characteristics of the substations, and the distance between the clustering center point and the substation was calculated. Finally, a distribution network planning model was constructed with the most power supply units between stations and the smallest total Euclidean distance considering power distribution as the objective function. Taking the transformation of the distribution network in the Dhanmondi area of Dhaka, the capital of Bangladesh, as an actual engineering application example, the results shows that the average difference of load partitions obtained by the improved K-means clustering algorithm is reduced by 34.35% compared with the traditional method.

Abstract:In order to improve the accuracy of convolutional neural network for water meter pointer reading recognition and realize the operation of convolutional neural network transplanted into STM32 microcontroller, A data set containing 2913 water meter pointer pictures was used for transfer learning and testing of GoogLeNet and ResNet-18. The accuracy of GoogLeNet test set was 89.37%, and that of ResNet-18 test set was 93.24%. Based on the jumping connection idea of Resnet-18 model, the method of feature fusion of high and low levels is used. On the premise that the size of receptive field remains unchanged, the 7×7 large convolution kernel is replaced by 3 3×3 small convolution kernels in series to reduce the number of network parameters, reduce the depth of the network, and speed up the convergence of the network during training. Then, a convolutional neural network model with higher accuracy and faster convergence for water meter pointer reading is proposed. The test set accuracy of this model is 95.11%. In order to overcome the difficulty of extremely limited storage resources of STM32 microcontroller and further reduce the network size and the number of network parameters on the condition of ensuring high accuracy, the test set accuracy of the designed model is 91.51%. The training process is completed using MATLAB deep learning toolbox on PC, and the generated onnx model is only 948KB in size. The running footprint of RAM is 437.14KB.

Abstract:In response to the problem of low sensitivity and unstable performance of traditional platinum resistance, metal and other temperature sensors, this paper selects a new nanomaterial reduced graphene oxide as the temperature-sensitive material, preparation of forked-finger electrode thermistors using mems process. The reduced graphene oxide morphology was described by scanning electron microscopy (SEM), and a test system was built to test its capabilities. The results show that the resistance of the sensor decreases with increasing temperature in the region of 30°C to 210°C. The sensitivity of the sensor is about 85.97 Ω/°C, the linearity is up to 0.98049, while it has a very small hysteresis and good stability.

Abstract:Aiming at the shortcomings of traditional atmospheric boundary layer sounding system, such as insufficient detection performance, low data accuracy, high operating cost and threatening aviation safety, a new atmospheric boundary layer sounding instrument based on GNSS ( Global Navigation Satellite System ) is designed. The instrument uses satellite navigation wind measurement technology to realize the inversion of wind speed and direction at high altitude, and uses multi-sensor fusion technology to realize the measurement of meteorological elements such as temperature, humidity and air pressure. At the same time, it can monitor PM2.5 and other air pollutants through the external module. The test results show that the average deviations of the radiosonde in temperature, humidity, wind speed and wind direction respectively are 0.19 °C, 0.62 %, 0.25 m / s and 0.62 °, which meet the accuracy requirements of high-altitude meteorological detection. It can be widely used in the study of atmospheric boundary layer characteristics in meteorological detection, urban planning and environmental evaluation.

Abstract:Monocular image depth estimation is a basic problem in the field of computer vision, Convolutional Spatial Propagation Network (CSPN) is one of the most advanced monocular image depth estimation methods. Aiming at the deformation problem of some objects and the boundary mixing problem caused by the blurring of the edges between objects in the dense depth map predicted by the network, we have improved CSPN from the network structure and loss function respectively. The input sparse depth map is downsampled three times with different sizes and added to the corresponding coding process and skip connection part of the U-Net module, so that it can more accurately capture the structure of objects with different scales. The original loss function is replaced by the improved loss function formed by the weighted combination of depth error logarithm, depth information gradient and surface normal. The experimental results on nyu-depth-v2 data set show that compared with CSPN, the root mean square error RMSE and average relative error REL of ICSPN are reduced by 17.23% and 28.07% respectively. The ICSPN in this paper makes full use of the input sparse depth map to reduce the deformation of the object structure in the predicted dense depth map. At the same time, the loss function with gradient loss is used to monitor the training process, which reduces the edge position error of the object and the problem of boundary mixing.

Abstract:In view of the specular reflection of the protection panel and cabinet door, the accuracy of identifying the protection plate state through image is low. A state identification method of substation protection platen based on improved sparse representation is proposed. On the basis of eliminating high light interference, the platen state identification is realized. Aiming at the problem of highlight interference, firstly, the two-dimensional maximum interclass variance method is used to detect the highlight region in the image, and then the sparse representation repair algorithm is improved to eliminate the highlight interference in the image. For the repaired image, the state of the protective pressing plate is identified according to the minimum circumscribed rectangle method. Finally, the effectiveness and accuracy of the proposed method are verified by two groups of comparative simulation analysis with or without highlight interference. Among them, for 200 different types of platen samples without high light interference, the accuracy of the minimum external rectangle method is 98.0%, which is better than 83.5% of the inclination identification method. For 240 image samples with high light interference, the improved sparse representation algorithm is used for state identification, and the accuracy can reach 97.92%.

Abstract:This paper designs a lane line detection algorithm for multi-road scenes based on Raspberry Pi embedded platform in the image preprocessing stage. In the image preprocessing stage, an adaptive binarization extraction algorithm for lane lines is designed. By comparing the pixels to be measured with the vertices of the diamond space where they are located, the binarized lane line information is completely extracted. at the same time, combined with the method of maximum classes error (OTSU), the interference information is effectively filtered out by means of image fusion. In the lane line fitting stage, the slope constraint and distance limitation of the Progressive Probabilistic Hough Transform are improved, and the edge points of the lane line are accurately calculated after further filtering out interference information. Finally, the least squares method is used to fit the lane line. The test results show that the algorithm has a stronger anti-interference ability, and the detection accuracy of multiple road scenes can reach 90.24%. And the running speed on the Raspberry Pi platform is 25fps, which meets the real-time requirements.

Abstract:In order to solve the problems of insufficient correlation between visual features and word features, low training efficiency, errors in generated natural language and low index scores in the process of video description, a video description model based on the attention mechanism of dilated convolution is proposed. In the encoding stage of the model, Inception-v4 is used to encode the video features, and then the encoded visual features and word features are input into the attention mechanism based on dilated convolution. Finally, the video is decoded through the long short-term memory network to generate the natural description statement of the video. A comparative experiment was conducted on the public video description data set MSVD, and the model was verified by evaluation indicators (BLEU, ROUGE_L, CIDEr, METEOR). The experimental results showed that the video description model based on the attention mechanism of dilated convolution has significantly improved in all indicators. Compared with the baseline model SA-LSTM (Inception-V4), the BLEU_4, ROUGE_L, CIDEr and METEOR indicators have increased by 4.23%, 4.73%, 2.11% and 2.45% respectively.

Abstract:At present, the situation of epidemic prevention and control is grim. Real time and rapid mask wearing detection in crowded places can effectively reduce the risk of virus transmission. Aiming at the low efficiency of manual detection, a lightweight mask wearing detection algorithm based on YOLOv3 is proposed. ShuffleNetv2 is used to replace the original backbone feature extraction network to reduce the amount of network parameters and computing power consumption. SKNet attention mechanism is introduced into the feature fusion network to enhance the ability of feature extraction at different scales; CIoU is used as the boundary box regression loss function to further improve the detection accuracy. Experiments on the constructed face mask detection data set show that, compared with the original YOLOv3, the proposed algorithm improves the detection speed by 34FPS while maintaining high detection accuracy, and effectively realizes accurate and fast mask wearing detection. Compared with other mainstream target detection algorithms, the algorithm also has better detection effect.

Abstract:Ground Based Augmentation System (GBAS) is a terminal area approach guidance facility designed to replace the traditional Instrument Landing System (ILS). One set of equipment can provide approach guidance service for multiple runways of the airport. For some large airports, two VDB transmitters are usually installed on the ground to increase system coverage. However, traditional flight inspection cannot identify which station emits the signal from the reception, which causes failure of station coverage evaluation. This paper proposes a multiple slot data analysis method based on EVSF1000 unit, which can analyze and evaluate the VDB signals of multiple ground stations in one flight by decoding and parsing data from 8 VDB time slots, to achieve the flight verification requirements of GBAS system with dual VDB stations. The method is verified in an GBAS inspection flight which completed the inspection of dual VDB stations installed in Tianjin Binhai Airport.

Abstract:To solve the signal synchronization problem caused by the wide range of time delay and frequency shift of the satellite-based automatic identification system (AIS), a new frame header detection algorithm based on differential phase waveform matching is proposed. The theoretical analysis of the phase characteristics of the signal of the training sequence and the start flag in the AIS frame structure, and the use of a specific matching algorithm to detect the frame header. The simulation results show that the algorithm has good properties in frequency shift resistance, and has a high detection rate, low frame error rate, and false alarm rate under low SNR. Compared with other algorithms, the computational complexity proposed in this article is very low, and the hardware resource consumption of FPGA is low.

Abstract:In order to meet the performance requirements of long line high speed signal transmission with zero BER, a long distance data transmission system is designed using LVDS interface technology and optical conversion method. Hardware design uses MLH series serial cable equalizer and driver to work together to enhance the coaxial cable differential signal driving capability with adaptive compensation. Extension of the long-line transmission distance by means of combined photoelectric signal transmission. The logic design solves the high-speed data transmission out-of-lock phenomenon by optimizing the judgment flags for valid and invalid numbers. After several high-capacity data reception tests in various experimental environments, it has been demonstrated that LVDS signals can be transmitted without error codes at 300Mbits/s over 100m twisted pair and 2km fiber optic transmission links, and has been put into engineering applications.

Abstract:In order to optimize the energy efficiency spectrum of the double-hop amplification and forwarding cooperative multi-relay network under the incomplete channel state, a strategy of combining the relay selection scheme based on the network model with the joint optimization scheme of transmission rate and total transmission power is proposed. Specifically, in a two-hop amplifying and forwarding cooperative multi-relay network under an incomplete channel state, a relay selection method based on relay location is first designed to select the optimal relay. Then, when power optimization, combined with the classic 0-1 fractional programming problem solving algorithm-Dinkelbach method, design the rate selection algorithm when transmitting data with the minimum power in the incomplete channel state. The simulation results show that in the double-hop amplification and forwarding cooperative multi-relay network with channel estimation errors, the scheme proposed in this paper enables the network to have an optimal transmission rate. In addition, compared with the method of random relay and equal power allocation, this strategy can improve energy efficiency by nearly 22%.

Abstract:The Automotive Open System Architecture (AUTOSAR) is an open and standardized automotive architecture that enables development partners to integrate, exchange, re-use and transfer functions within a vehicle network and improves their efficiency of development. AUTOSAR CAN communication stack is an important protocol stack of AUTOSAR, which is designed to pave the problems of uneven software quality and poor reusability in the communication layer of the automobile CAN network. The purpose of this paper is to design and implement the AUTOSAR CAN communication stack base module based on the AUTOSAR 4.0.3 standard in the NXP MPC5748G platform, including CAN controller state control, channel communication control, transmit buffer mechanism, transmit cancel mechanism, and software receive filter mechanism. In the experiments comparing the cycle sending delay of traditional CAN communication software and the CAN communication software implemented in this paper, the results show that AUTOSAR CAN communication software can reduce the average delay of high priority message cycle sending by around 90%, proving that the AUTOSAR CAN communication stack implemented in this paper effectively improves the CAN communication software performance.

Abstract:In order to improve the problem that the change of a single physical quantity has too much impact on the final result due to the linear relationship between physical quantities in the process of indirect torque measurement, a nonlinear multi model soft sensing method based on weighted K-means clustering and LSSVM is proposed in this paper. Firstly, multiple easily measured variables are selected as auxiliary parameters, and the data are preprocessed by using subjective and objective comprehensive weighting theory. Secondly, K-means clustering algorithm is used to form clusters of data with similar physical characteristics. Finally, multi model of data cluster is established and measured based on least squares support vector machine algorithm. The results show that under the same experimental conditions, the root mean square error of the proposed model is reduced by 0.484 and 0.263 respectively, and the average absolute percentage error is reduced by 1.003 and 0.292 respectively, which effectively improves the accuracy and stability of the measurement.

Abstract:Aiming at the problem that the rolling bearing fault under strong noise is weak and the characteristic frequency is difficult to extract, which makes it impossible to diagnose the fault accurately, a fault diagnosis method based on trisection EMD fusion Autogram threshold algorithm is proposed. EMD is used to reduce the noise of the signal, and a trisection method based on M index is proposed. EMD reconstructs all IMF into three components (Write M1, M2, M3), and M2 is the required fault component; The Autogram algorithm is used to process the M2 component to determine the resonance frequency band, and the resonance signal is processed by the threshold envelope spectrum to obtain three threshold spectra. The fault type of rolling bearing is diagnosed according to the fault characteristic frequency in the threshold spectrum. In this paper, the simulation signals and the measured data of the inner and outer rings of rolling bearings are used to prove the effectiveness of this method. Experimental results show that the fault diagnosis rate of this method is over 95%

Abstract:The portable embedded-based automatic weather station fault detector has been developed to meet the demand for rapid diagnosis of automatic weather station faults. It integrates a low-power microcontroller, high-precision electronic components and high-speed data communication module in a unified way. It isolates the microcontroller data bus and the weather element processing circuit bus through a four-channel digital isolator to enhance the anti-interference between signals, and realizes the simulation and acquisition of wind direction, wind speed, rainfall, temperature and humidity, and barometric pressure signals to realize the fault detection of automatic weather stations. The test results enlighten that the instrument can quickly locate the fault point, which comes from the automatic weather station sensor, collector, power module, communication module, etc. The error of the output analog weather signal is not more than 0.3, and the average error of signal acquisition is stable around 0.2.

Abstract:High-voltage circuit breaker operating mechanism vibration signal contains important information about the status of the circuit breaker, which is of great significance for the diagnosis and identification of the operating status of the operating mechanism. Aiming at the complex characteristics of random and non-smooth vibration signals, a circuit breaker fault diagnosis method based on bispectrum analysis and a two-stream flow shallow convolutional neural network is proposed. Bispectral analysis and wavelet analysis are performed on the vibration signal. The 2D bispectral matrix and 1D wavelet band energy are extracted as the dual-channel features of the two-stream convolutional neural network, respectively; supervised model training of vibration signals collected from circuit breaker simulation experiments for five operating conditions. The results show that the bispectral analysis can suppress Gaussian noise, retain the main peak morphological features of the operating mechanism under different operating conditions and fuse wavelet band energy features, and the proposed model can achieve a high recognition accuracy of 98.33% in 5 training iterations to achieve fault diagnosis and identification of the circuit breaker operating mechanism.