Abstract:Aiming at the difficulty of synchronizing the clock of TDOA in the ultra-wideband indoor robot positioning system and the various interference problems in measurement, a new method for indoor robot TDOA asynchronous non-reference node ranging is proposed. This method uses the time calculated from the actual measurement as a correction factor to construct a ranging algorithm model for the master base station without a reference node under an asynchronous clock. The measurement information of the master and slave base stations can directly calculate the robot's arrival distance difference; propose an association weight The new algorithm of value filtering combines the moving average to obtain the arrival distance difference state matrix, which effectively reduces signal interference. And from the measurement performance, ranging and positioning and other aspects of error analysis. Experimental results show that 98% of the ranging accuracy is maintained within 10cm, and measurement deviation will not occur with long-term operation, and it has good positioning accuracy, which can meet the positioning needs of indoor robots.

Abstract:In order to characterize the defects of carbon fiber reinforced plastics (CFRP) by using the capacitance directly measured by planar capacitance sensor, a planar capacitance sensor is constructed in COMSOL based on the capacitance edge effect, and the influence of electrode parameters on its performance was analyzed and optimized, The sensitivity has been improved by 2% ~ 52%, and the penetration depth can reach 4.13mm. When the sensor area is constant, we can find the optimal length to width ratio of the sensing electrode, and then we can fit the linear relationship between the electrode length and area. The change in the capacitance value of the sensing electrode is detected when the defect size is varied with a single electrode DC excitation. The results show that the capacitance value of the sensing electrode decreases as the defect size increases, and the quantitative evaluation of CFRP defects can be achieved using the fitted curve with a correlation coefficient of 0.996 or more.

Abstract:Active encoding source non-contact stereo measurement method is proposed with one camera and one projector in non-specific posture. A hybrid calibration board is used to calibrate the camera and projector in the measurement system respectively and the mapping relationship between two-dimensional image coordinates and three-dimensional spatial coordinates is established. Feature points is projected from the horizontal and vertical Gray codes and their multi frequency division measurement fringes, and the three-dimensional spatial coordinates of target is obtained through inversion. This method avoids position and orientation constraints in the monocular stereo measurement and the use of external precision measuring equipment. At the same time, feature point matching in binocular stereo vision measurement is solved by this method. Experiments show that in the range of 1.2-1.6m measurement distance, the measurement accuracy can reach 1.5mm, which can realize the active encoding source stereo-measurement with monocular camera.

Abstract:INS/BDS integrated navigation data comes from different subsystems, but the hardware and software factors between different systems will lead to data synchronization between the systems, which will affect the final integrated navigation accuracy. It can be seen that data alignment is very important in the application of integrated navigation system. Aiming at this problem, a INS/BDS integrated navigation data alignment method is realized by using FPGA+DSP small navigation solution platform. The adjustable pulse excited by 1PPS of BDS receiver is used as interrupt, and the inertial sensor and satellite navigation module are captured and framed by FPGA control chip. Then the delay error is calculated and compensated by DSP, so as to achieve the synchronous alignment of INS/BDS real-time data. Experimental results show that this method can improve the position accuracy by about 50%, and can meet the requirements of data alignment, which has certain engineering application value.

Abstract:In order to overcome the defects of narrow bandwidth, low efficiency and high cost of existing microstrip antennas, an X-band broadband and efficient stacked microstrip patch antenna is proposed on a low cost FR4 substrate. The antenna is composed of multilayer FR4 substrate stacks. Through the introduction of air cavity, the effective dielectric constant and dielectric loss of the antenna are reduced, and the efficiency and bandwidth of the antenna are improved. The use of parasitic patches increases the reflection zero in the band, which further enhances the bandwidth of the antenna. At the same time, in order to compensate the parasitic inductance of the feed probe, the input impedance of the antenna is matched by microstrip line. Simulation and experimental results show that the frequency range of input VSWR less than 2.0 is 7.8~11.6GHz, and the relative bandwidth is 40 %. The antenna gain is 7.6dBi and the efficiency is 83%. Compared with the traditional microstrip patch antenna, it has the characteristics of broadband, high efficiency and low cost, and has broad application prospects.

Abstract:In order to improve the intelligent level of blast furnace production, a prediction model of pressure difference in low part of blast furnace with integrated learning algorithm is proposed, which solves the problem of accurately predicting the lower pressure difference based on online data. Through systematic analysis of the internal mechanism of the blast furnace, the raw material parameters, operating parameters, state parameters and index parameters of the blast furnace are comprehensively selected as the input of the model. The actual field data is used to obtain the correlation coefficient between the variables, and the important characteristic variables related to the pressure difference in the lower part of the blast furnace are determined. The extra tree ensemble algorithm is used to establish the pressure difference prediction model, and combined with the prediction accuracy of the model, the forward selection method is used to optimize the input of the model. By selecting the hyperparameters of the model algorithm, the optimal hyperparameter set is obtained. The accuracy R2 of the lower pressure difference prediction model established by the parameter set reaches 0.8264, and the MSE is close to zero. The test results prove that the model has good prediction accuracy and generalization ability, and has important guiding significance for the on-site operators to predict the operating conditions of the blast furnace and adjust the furnace conditions in advance.

Abstract:Due to the dispersion effect and aperture fill effect, essentially, the traditional active phased array antenna is a narrow-band system. To solve this problem, true time delay (TTD) method is massively employed in active phased array antenna design. In this paper, a multi-stage TTD method is proposed to improve the instantaneous bandwidth of antenna array. Firstly, the theory of multi-stage TTD is briefly introduced. Then the antenna radiation pattern performance employing sub-array TTD and multi-stage TTD method are analyzed and compared based on simulation results. Finally, the typical broadband performance of the antenna is tested. When the measured wave position’s frequency is 1.5GHz deviating from the center frequency and azimuth/elevation sweep angle is 60°, the pointing error is reduced from 1.24°/1.63° to 0.47°/0.24° after employing multi-stage TTD method, and the results prove that the multi-stage TTD method significantly expand the instantaneous bandwidth with good engineering realizability.

Abstract:Now the traditional indoor positioning technology mostly needs to install labels or terminal equipment on the located object, which has great limitations in some specific occasions. Therefore, an indoor positioning system based on computer vision is proposed. The system detects a specific object and obtains its image coordinates through the target detection algorithm, judges whether the object is in the area to be located through the target point determination algorithm, and finally obtains the map coordinates corresponding to the object through the projection transformation algorithm. When the object to be located has no perception, the positioning is completed and the positioning error is within 1m. In order to facilitate users' intuitive observation, the actual map and object position are displayed on the terminal device in the form of web page.

Abstract:Based on the measured data of a subway open cut station, the main factors affecting the surface settlement around the foundation pit of the subway station are discussed, analyzed and predicted. Firstly, correlation analysis is used to obtain the correlation degree of various monitoring data and correlation parameters; Secondly, the nonlinear regression model is established by multi-element fitting regression analysis to analyze the law of surface settlement and predict the settlement; Finally, the measured data of open cut deep foundation pit of a subway station are used for experiment and result inspection. After inspection, it is found that the maximum error between the predicted value and the measured value is 0.273mm, and the predicted settlement trend is completely consistent with the measured settlement trend, indicating that the method is feasible.in order to reduce the observation frequency of surface settlement, reduce the workload, more scientifically understand the causes of surface settlement and enrich the prevention strategies.

Abstract:The simulation modeling and control optimization simulation of the pitch control channel in the flight control autopilot of a certain type of civil aircraft were carried out. The model of the pitch control channel of the aircraft autopilot was modeled, and the model system of the channel was improved. The modeling of the control law of the pitch channel was completed in the development environment of Matlab, and the control optimization simulation of the system was completed in the development interface of Simulink, which involves the call of C language functions in Matlab, the construction of the automatic control system and the call of each module in Simulink. The PID controller and the ripple-free minimum beat digital controller were used to improve the system, the former improves the stability, accuracy and physical realizability of the system, and the latter optimizes the time parameters of the system to make it faster. The simulation results show that the adjusting time of the system is controlled within 1.1s, the overshoot is no more than 5%, the steady-state error is less than 1%, and the physical realizability, stability, accuracy and rapidity of the system can meet the basic requirements, the adopted method is direct and efficient, and the calculation precision of the model parameters can basically meet the requirements. Choosing Matlab/Simulink as the development tool is conducive to the construction of the control system model and the subsequent development of the modeling structure, which greatly reduces the difficulty of the development of the controller and the optimal simulation models for the system control, and provides powerful technical support and reference for the further development and application of the deeper self-test function of the more automatic driving instruments and system.

Abstract:In indoor Visible Light Communication (VLC) system, the data carried by Light-emitting diode (LED) light is transmitted through the optical channel through multipath dispersion. Different indoor reflective surfaces are incorporated into the model through MATLAB simulation modeling, and the channel impulse response algorithm is proposed. Direct Current biased Optical Vector Coding (DCO-VC) technology is introduced in the system. The DCO-VC system uses M-ary pulse amplitude modulation (M-PAM) as modulation scheme, to adjust the system data transmission efficiency and accuracy. Intercell interference is introduced into the simulation model. Studies have shown that the higher the modulation parameter is, the higher the data transmission efficiency of the system is. When the modulation parameters are lower, the transmission efficiency of the system is relatively reduced, but the transmission accuracy is higher and the system can still output effectively in the presence of inter-cell interference. In real application, when the system signal-to-noise(SNR) ratio is lower than 35dB, 2PAM or 4PAM modulation should be applied to ensure the transmission accuracy. When the SNR ratio exceeds 40dB, use the 32PAM modulation to increase the transmission rate.

Abstract:Aiming at the DC side voltage control problem of APF of active power filter, the traditional PI control method has the contradiction of low control precision and high speed. Automatic disturbance rejection control (ADRC) greatly improves its anti-interference ability by introducing interference compensation. But at present a single setting ADRC parameter optimization algorithm can not improve the control precision. In order to solve the problem that the BAS algorithm is easy to fall into local optimum, the mutation link is introduced into the algorithm to make it jump out of local search and get the global optimal solution more rapidly. By building MATLAB/Simulink simulation and APF experimental platform to verify that the algorithm optimization can reduce overshoot,achieve accurate current tracking and reduce the harmonic rate that from 3.32% to 1.34%.

Abstract:An improved particle swarm optimization (FPSO) algorithm is proposed, which makes particles focus on different optimization tasks according to the fitness value, and applies it to the UAV 3D path planning problem. The traditional particle swarm optimization (PSO) algorithm sets uniform control parameters for all particles, the optimization process is not flexible enough, it is easy to fall into local extremes and the convergence speed is slow. The improved FPSO algorithm proposes three optimization strategies, that is, the combination of PSO algorithm and genetic algorithm (GA), setting dynamic inertia weight and introducing step factor, so as to give full play to the search advantages of particles with different fitness values and make them dynamically focus on local search or global search. The simulation results show that FPSO algorithm has better search results, fewer iterations, and the average consumption time is 22.0% shorter than PSO algorithm and 39.6% shorter than GA algorithm.

Abstract:Due to the sparse data and large reconstruction area in the actual explosion test, the amount of data is insufficient. The traditional iterative reconstruction algorithm has its limitations in the reconstruction of the shock wave overpressure field. In order to improve the imaging effect under the condition of incomplete projection data with single projection angle, a reconstruction method combining total variation minimization and dictionary learning is proposed in this paper. Combining the advantages of compressed sensing in sparse constraints, the TV regularization method is used to optimize the edge information of the shock wave overpressure field, and the dictionary learning method is used to improve the detail characteristics of the shock wave field, which can reconstruct the shock wave overpressure field with less data. The analysis shows that compared with the SART algorithm, the proposed method can significantly improve the reconstruction quality, its RMSE value is reduced by nearly 40m/s, and the relative error in each grid is reduced by about 2.5%, and a more efficient reconstruction method is realized. It has certain theoretical significance and engineering application value in weapon and ammunition damage assessment and engineering protection.

Abstract:False data injection attack (FDIA) in power information system affects the normal operation of power grid by maliciously tampering with the state data of corresponding physical system. This paper proposes a false data injection attack detection method based on dynamic kernel principal component analysis (DKPCA), in order to solve the time correlation of FDIA events in power information system (dynamic) problem, and the problem that it is difficult to separate nonlinear variables. This method solves the dynamic autocorrelation between variables by constructing a dynamic augmented matrix, uses the kernel matrix to map nonlinear variables into high-dimensional space and convert them into linear variables, introduces principal component analysis to establish DKPCA model, obtains the control limit of statistics, and judges whether there is a fault by detecting data in real time. The experimental simulation is carried out on TEEE-30 node system. Compared with KPCA, PCA, NPE, TNPE and other detection methods, the results show that the detection rate of DKPCA model is as high as 100%, while maintaining a low false positive rate of 0.2%. It is proved that the proposed method can detect the attack data in power information system in real time, effectively avoid fault omission and ensure the data security of power information system.

Abstract:In order to fully exploit the relationship between temporal features in NOx emission data and improve the accuracy of NOx emission forecasting results, this paper proposes a NOx emission forecasting method based on a hybrid neural network model of convolutional neural network (CNN) and long short-term memory network (LSTM). Taking the historical data of a 300MW coal-fired boiler as a sample, the k-means clustering method is used to group NOx emission sample sets. Then the high-dimensional mapping relationship of NOx emission variables is extracted based on the convolutional layer and pooling layer of the CNN network to construct a high-dimensional time series feature vector, which is input the LSTM network. A NOx emission prediction model is established based on CNN-LSTM by training LSTM network parameters. Through the actual data verification of coal-fired boiler, the Mean relative percentage error of the proposed prediction model for training and testing samples are 1.76% and 3.85%, respectively, which are much lower than other models. The results show that the proposed NOx emission prediction model has significant advantages in terms of prediction accuracy and generalization ability.

Abstract:In order to solve the problem that it is difficult to select the optimal combination of linear quadratic controller (LQR) weight parameters through experience, a Lévy flying particle swarm algorithm with greedy principle is proposed to optimize the LQR weight matrix Q. The traditional particle swarm algorithm is easy to locally converge. On this basis, the Lévy flight principle and the greedy selection method are added to expand the parameter optimization range and increase the convergence speed, and the optimal weight matrix Q is obtained. The acceleration response and displacement response of the vibration isolation object under the two-layer precision vibration isolation system model under the control of LQR are analyzed under the conditions of sweep frequency excitation and random excitation. The simulation results show that the Lévy flying particle swarm algorithm with the principle of greed is effective Increasing the uniformity of the population distribution improves the convergence speed to find the optimal solution, and can effectively achieve vibration isolation in both excitation conditions.

Abstract:To address the problem that the traditional UNet is ineffective for segmentation of skin malignant melanoma images with variable size and shape, the improved method is mainly implemented to fully utilize the multi-scale features through two improvements, firstly, in the encoder, the global dense network, the local dense network and the dilated convolution design, and later, in the decoder, the local residual design and the classification regularization. Compared with UNet, the improved method improves 0.82%, 0.03%, 1.99%, and 1.03% in the Dice coefficient, accuracy (ACC), sensitivity (SE), and intersection-to-merge ratio (IOU) metrics, respectively. The experimental results demonstrate that the improved method can improve the image segmentation of skin malignant melanoma and is an effective underlying network structure.

Abstract:Aiming at the difficulty of side crack detection of cylindrical honeycomb ceramics, a detection method based on machine vision is proposed. Through the demand analysis of side crack detection, COMS camera and LED white parallel light source are selected. The collected image is filtered, and the median filter is selected to remove salt and pepper noise. According to the characteristics of the image, the ROI region is selected, the global threshold segmentation operator threshold is used for image segmentation, and the expansion method is used to connect the fracture region. When extracting surface defects, the connection operator is used to segment the image region, and then three features of area, length and width are selected to extract surface defects. The test results show that when there are 50 samples, the time required for qualified, unqualified and mixed samples by this method is 12.50 min, 6.64 min and 10.58 min respectively, which has higher detection speed and better real-time performance; The accuracy rates are 96%, 84% and 90% respectively. The accuracy rate needs to be improved and needs further research.

Abstract:In order to solve the problem of image defogging and enhancement in haze weather, this paper proposes a defogging and enhancement algorithm that combines image layering and dark channels. The algorithm first establishes a dark channel model for the input image, estimates the atmospheric light value and transmittance, and restores the image by defogging. Next, the image is subjected to bilateral filtering and transformation to stretch the gray-level area of the pixels in the low-frequency image information. Or compress, normalize the high-frequency image information, then use the normalized histogram and nonlinear S-curve to perform gray-scale transformation, and finally use the weighted fusion method to effectively merge the low-frequency and high-frequency image information to obtain Output image. Experimental results show that the average gradient and information entropy of the algorithm in the three sets of images are 0.0734 and 7.1733, respectively, which are better than the other three algorithms, and the average contrast and time consumption of the algorithm are 422.6 and 0.76, respectively. It is feasible.

Abstract:In order to solve the problems of low recognition accuracy and poor real-time performance of traffic police gestures in the environment of uneven illumination and complex background. In this paper, based on the yolov5 network model, part of the convolution layers are replaced by self-calibrated convolutions to increase the range of the receptive field. Shuffle attention module is introduced to improve the feature extraction ability of the algorithm. Aiming at the complex and changeable environment of traffic police, the Focal loss function was replaced by Generalized Focal loss function to improve the expression ability of target frame in complex environment. Experimental results show that on the basis of real-time performance, the average accuracy of the improved algorithm for traffic police gesture detection is as high as 98.54%, which is 3.39% higher than that of the unimproved algorithm, and the loss value of the loss function is smaller.

Abstract:Aiming at the problem that the existing shadow detection algorithms are difficult to extract irregular and fragmented shadows in complex farmland scenes, a shadow detection algorithm for remote sensing images of farmland crops by UAV is proposed. Combining the color characteristics of the shadow/non-shadow area of the UAV image, construct a new gray scale transformation method based on dual-channel difference and G-band enhancement, and use the maximum between-class variance method to automatically threshold the grayscale image to obtain shadow detection result. Experiments with the data collected by the team at the National Corn Industry Technology System Experimental Demonstration Base show that the detection results of the proposed method are closer to real shadows, with an average overall accuracy of 0.9868 and an average F1 score of 0.9567.

Abstract:In order to realize nondestructive testing of ferromagnetic objects, a magnetic field scanning system based on FPGA and magnetic sensor array is designed. FPGA is the main controller of this system that includes AD configuration module, AD interface module and data cache module. FPGA controls the transmission device to scan, and the magnetic field is converted to the corresponding voltage signal by the magnetic sensor array and the circuits. The voltage signal is transmitted to FPGA through the analog-to-digital conversion module with the core of AD7768, and high-speed magnetic field data acquisition system with eight parallel channels is realized. Then, through bilinear interpolation and pseudo-color conversion algorithm in FPGA, the collected magnetic field data are converted to the image of magnetic field distribution, which is stored in SD card and displayed in LCD in real time. In the experiment, the magnetic field scanning system is used to carry out magnetic field scanning for the ring magnet, magnetized wire, iron nails embedded in the wood plate, wire stripping pliers and needle-nose pliers. The program, based on magnetic flux and OTSU method, is used to locate and identify the object. The average root mean square error of positioning is 3 mm, and the measured length is close to the length of the actual object. The position and direction of the magnetic object could be identified, and the magnetic object could be identified by the characteristic magnetic field of the object, so as to achieve the purpose of nondestructive detection and identification of ferromagnetic objects. FPGA ZYNQ7010, AD7768, HMC1001 are used to collect magnetic field data in the system. It has the advantages of high acquisition rate, convenient expansion, low power consumption, and can obtain magnetic field image with spatial distribution . It can be applied to the location and recognition of ferromagnetic targets, such as safety detection, magnetic target detection of underground and underwater.

Abstract:In order to achieve guided wave non-contact automatic inspection, based on air-coupled ultrasonic inspection technology, the A0 mode lamb wave is used to orthogonally scan the corrosion defects of the preset 0.5mm and 1mm depth grooves in the aluminum plate to obtain the defect. Location information, using the energy of the lamb wave packet and the correlation coefficient between the lamb wave virtual time reversal reconstruction signal and the original signal as the damage index to characterize the size of the defect. Finally, based on the probabilistic damage imaging method, the full multiplication method is used to calculate the defect damage index. Fusion imaging. The imaging results show that the imaging results of the virtual time reversal algorithm are better than those of the conventional energy algorithm. The results show that the position and shape of the defect are more consistent with the actual situation, and it has a better ability to distinguish the depth of the defect.

Abstract:With the development of the testing verification of the equipment system, there are a huge number of fault samples, long test cycle and high cost in the radar transmitter system. A method based on improving the information entropy and Bayes posterior risk criterion is proposed to optimize the fault sample size. Based on the virtual test data of radar transmitter subsystem, using the method of improving information entropy, obtain the system-level test data by converting the system subsystem, and then determine the total test samples and the maximum number of failures, which determine the test scheme. Taking the fault detection rate of the radar transmitter system, the test fault sample size of the proposed method is 165, reducing the fault sample size by 64.97% and 51.18% compared with the classical sampling and sequential probability ratio test methods, respectively, verifying the effectiveness of the proposed method to some extent.

Abstract:A circuit soft fault diagnosis method based on Improved Whale (IWOA) Optimized Support Vector Machine (SVM) is proposed to address the problem of DC-DC circuit soft fault diagnosis with low accuracy. First, a VMD is performed on the fault signal to extract the feature vectors. Then we improve the global search ability of the traditional whale algorithm to prevent falling into local optimum by referring to the feedback mechanism, and change the linear factor to a nonlinear factor to balance the global search and local exploitation ability to improve the whale algorithm to solve the problem of easily falling into local optimum and low local exploitation ability. Finally, the IWOA-SVM model is used for soft fault diagnosis of circuits, and finally, the problem of low accuracy of soft fault diagnosis of circuits is achieved with high efficiency. Based on the results of fault diagnosis, it is shown that the improved whale algorithm optimized support vector machine has better diagnosis effect compared with other methods compared in this paper. Fault recognition accuracy of 99.1667%.

Abstract:In order to realize the intelligentization of circuit board fault detection, an electroluminescent circuit board fault detection system was designed, its structure was introduced, and a circuit board fault detection method was proposed.In the image preprocessing module, the original image is enhanced by gray scale, and the predetermined bits of the original image are realized by using the method. The code of PCB file saved in the form of ASCII code is deeply analyzed, and the PCB file data is read by MATLAB software, saved in a specific structure, and the template image is drawn by drawing function. Using color image segmentation method to extract the interested part of the circuit board pre-position image, and then the template image drawn by software is registered with the original image of pre-position, and a more ideal result is achieved. Experiments on luminescent circuit board fault detection and location show the effectiveness of the method.The experimental results show that the method has less damage to the circuit board, and is effective in detection, fast operation, short cycle and easy to implement in engineering.