Abstract:Weak magnetic detection technology is a nondestructive technique based on the magnetic-force coupling effect. By measuring and analyzing the magnetic signal on the surface of the material, the stress concentration, early damage, and degree of damage in the ferromagnetic material can be readily detected and evaluated. The main advantage of the weak magnetic detection technology include no manual magnetization or attached sensor, no surface treatment of component, simple operation and so on. Based on the weak magnetic detection technology in the past 5 years, a theoretical model of the technology was established and the progress made in the description and calculation methods of magnetic signals at defects, the damage assessment criteria in judging and locating defects, the progress of quantitative inversion of defects and so on. Finally, the engineering applications status of this technology was discussed. Based on this review, the future research directions of weak magnetic detection technology were proposed.

Abstract:The electromagnetic flowmeter plays an important role in the industrial production. However, easily influenced by the bubbles in the fluid, fluctuations often appear in the measurement results, leading to a lower measurement accuracy. Quantitative analysis of the influence of bubbles in the electromagnetic flowmeter measurement process has important engineering significance for improving the accuracy of electromagnetic flow measurement and realizing gas-liquid two-phase flow measurement. In view of this, starting with the weight function, the analytic method is used to establish the theoretical model of bubble size, bubble eccentricity, and the number of bubbles. Then, by simulation and experiments, the influence of different states of bubbles on the electromagnetic flow system with a diameter of 100 mm and circular point electrodes with a 5 mm radius is studied. The output voltage sensitivity is used to analyze the influence of different states of bubbles on electromagnetic flow measurement. According to the research results, with the increasing of the bubble size, its impact on the electromagnetic flow measurement is within the range of 0. 3% ~ 5%, the maximum of 5%. The influence of bubbles with a diameter of 10 mm ranges within 0. 25% ~ 0. 6% along the eccentricity direction of the electrode and increases with the increasing of the eccentricity distance. The influence of the increasing eccentric distance along the direction perpendicular to the electrode has a tiny impact on the measurement. It fluctuates within - 0. 2% ~ 0. 2%. Meanwhile, with the bubble flow, a shorter distance between the bubbles and the electrode cross-section leads to a larger impact while the impact gets its maximum at the electrode cross-section. The maximum values for a single bubble are 0. 18%, 0. 22%, -0. 20%. The influence is enhanced with the increasing in the number of bubbles. The maximum values in the case of 3 bubbles and 6 bubbles are 0. 36% and 2. 3%. This paper improves the theories related to the weight function of electromagnetic measurement of gas and fluid double-phase flow, obtains the influences of different states of bubbles on the electromagnetic flow measurement system, and provides technical support for the improvement of electromagnetic flow measurement accuracy and realization of gas and fluid double-phase flow.

Abstract:The magnetic Barkhausen noise effect can reflect the statistical significance of the dynamic rotation and deformation of the magnetic domains during the alternating excitation of ferromagnetic materials, which can be used as a non-destructive detection technology for ferromagnetic materials' stress state, material deterioration and early micro-damage detection and evaluation. At present, the magnetic Barkhausen noise detection for magnetocrystalline isotropic materials has obtained a large number of laws and established engineering applicable methods, but most of the laws and methods often produce erroneous detection results when used for the detection of magnetocrystalline anisotropic materials or produce larger errors. In order to find out the reason for this inapplicability, a circumferential magnetic Barkhausen noise measurement system was built, taking X60 steel as an example to test the distribution of magnetic anisotropy, the distribution of magnetocrystalline anisotropy on the surface of the same bulk material is revealed from three aspects: the direction of the easy magnetization axis, the amplitude and shape of the circumferential magnetic anisotropy map, and the characterization of different characteristic parameters. The study found that the circumferential magnetic Barkhausen noise distribution at different positions on the magnetocrystalline anisotropic material is not uniform, so that the reference calibration curve generated based on the test block of the same material or the same batch of material is no longer available for the actual test piece inspection Validity, which is the key cause of detection biases and errors, and previous studies mostly believed that the magnetocrystalline anisotropy distribution on the same piece of material was consistent, or the influence of its distribution characteristics was ignored. The discovery of this phenomenon poses new problems and challenges for the magnetic Barkhausen noise detection of magnetocrystalline anisotropic materials.

Abstract:In this paper, an active fault-tolerant control method based on interval adaptive threshold and control law reconstruction strategy is developed for electric scooter system with uncertain parameters. Firstly, the uncertain diagnostic bond graph model of the electric scooter system is established, and the interval adaptive threshold based on interval valued analytical redundancy relations is developed to improve the fault detection performance in the presence of parameter uncertainties. Secondly, the sliding mode control law is designed under normal conditions to realize the speed tracking of electric scooter. After that, the sliding mode control law under faulty condition is established where the extreme learning machine is adopted to estimate the term compensating the unknown parameter fault in the control law in a real time manner. The switching of control law can be implemented online using the fault detection result. Finally, through the analysis of experimental results, the effectiveness of the robust fault detection method based on the interval adaptive threshold is proved, and the feasibility of the active fault-tolerant control method in the presence of parameter fault is verified.

Abstract:In order to solve the problem of the acoustic impedance mismatch between wood and air medium and the large attenuation of ultrasonic. These are the reasons that the signal-to-noise ratio of the detection signal is low and the measurement accuracy is not high. The phase coded pulse compression technology was applied to the air coupled ultrasonic detection of wood in this paper. Buck code sequence was selected as phase coding sequence, and the principle of generating and receiving signal demodulation of impulse compression excitation signal was detailed. The defect wood was used as the experimental object, and the difference of the detection signal between the defect and the non-defect was compared by single point scanning testing. Then the accuracy of wood defect detection was quantitatively analyzed by C-scan experiments. The results show that the signal-to-noise ratio of the detection signal is improved by 9. 12 dB when 13-bit buck code is used as the coding excitation signal, compared with sinusoidal excitation signal. The quantitative recognition accuracy of the defect for the node is 90% by C-scan results. The crack with width of 1 mm in the node can be detected effectively. These show that the application of pulse compression technology to the air coupled detection of wood can effectively improve the wood defect identification ability.

Abstract:With the goal to realize the real-time accurate measurement of chemical oxygen demand ( COD) in wastewater treatment process, a soft-measurement method based on TentFWA-GD RBF neural network (NN) was proposed. To solve the problems of network parameters settings and local optima existing in RBF NN based soft sensor modeling for complex industrial processes, as well as improve the model’s prediction precision and generalization ability, tent chaotic mapping was introduced in fireworks algorithm (FWA) to keep the population diversity and avoid the premature convergence by making use of the global ergodicity of chaos movement. Then a novel training method for RBF NN was proposed by combining the improved TentFWA with gradient descent (GD) method to enhance the learning ability. The TentFWA-GD RBF NN was applied to construct the fitting models of four Benchmark functions and the COD soft sensor model of rural domestic sewage treatment process. Simulation and application results showed that the model had lower function approximate error and higher COD prediction precision as compared with other neural network models. In COD soft sensor modeling, the mean square error and mean absolute error of the training results were 0. 18 and 0. 25, which of the test results were 0. 23 and 0. 36, respectively.

Abstract:In the industrial production process, ultrasonic flowmeter plays an important role with the advantages of non-contact measurement and suitable for various fluid media. Aiming at the problems of poor anti-interference ability and low detection accuracy of ultrasonic flow detection, a four-channel non-full tube ultrasonic flowmeter combined with pattern recognition is proposed. The system uses a high-performance chip with floating point operation for FFT calculation and calculates the integrated velocity of four channels, and identifies the liquid level of the non-full pipe with the edge calculation chip, and then improves the identification stability through the liquid level correction model. In the liquid level recognition model, the feature extraction module and the spatial attention mechanism module are used to extract the effective features, and the random forest classification module is used to classify the liquid level. The experimental results show that the DC-SAM algorithm can converge faster than other models, and the accuracy can reach 96. 6%. In the flow experiment, the system can achieve 96. 5% accuracy and good linearity compared with the calibrated flowmeter. The system can accurately identify liquid level and non-full pipe flow, and meet the stability requirements of detection while maintaining high measurement accuracy, which proves the feasibility of operation and deployment of edge calculation in ultrasonic flow detection.

Abstract:Aiming at the problem that the amplitude of electric shock current is relatively small and superimposed on the inherent residual current of cables and lamps, which makes it difficult to detect accurately, a multi parameter fusion detection method for street lamp electric shock accident is proposed. By collecting the instantaneous values of the total residual current, lamp current and phase voltage signals of the street lamp system, the normalized characteristic quantities of the residual current component of the suspected personal electric shock accident, the random capacitive residual current component and the change rate of the effective value of the load current are calculated respectively, then the fuzzy logic device is used to fuse these characteristic quantities to obtain the comprehensive characteristics of the personal electric shock accident. The recognition coefficient is compared with the experience threshold to judge whether there is an electric shock accident. The results of experiment and engineering application show that the method can effectively improve the recognition accuracy of electric shock accidents in street lighting system.

Abstract:In order to improve that the detection range of a single sensor is small, the detection features are few and the detection accuracy is low, a target detection method for visual and 2D laser radar is proposed. In terms of visual detection, an improved GoogLeNet algorithm is proposed to realize the visual recognition of target objects. Compared with GoogLeNet algorithm, this method has improved the recognition accuracy of 6 target objects by 0. 7%. In 2D laser radar detection using European clustering algorithm for 2D laser radar point cloud data clustering, and then use RANSAC algorithm for clustering data point in the cluster to filter, finally using Kalman filter algorithm to estimate the location of the target object, and realize the 2D laser radar in a particular plane 360° to detect and locate the target object tracking. Experimental results show that this method can enlarge the detection range, increase the detection features and improve the recognition accuracy of mobile robot.

Abstract:In order to improve the grinding efficiency of screw rotor, a screw synchronous grinding device is developed. Different ways are used to grind the concave and convex sides of screw respectively, which will produce repeated grinding areas at the joints. In order to ensure that the surface quality of the repeated grinding area meets the requirements of the screw, the surface roughness of the area is predicted. Firstly, based on Rayleigh distribution, the abrasive particle distribution model on the surface of sand belt is established. On this basis, the surface morphology of the ground screw is predicted according to the grinding removal mechanism and screw profile. Secondly, according to the definition of surface roughness profile, the predicted values of surface roughness under different process parameters are obtained. Finally, the accuracy of the proposed algorithm is verified by screw grinding and measurement experiments. The experimental results show that the average error of the proposed algorithm is 6. 27% and the minimum error is 0. 16%. Therefore, the proposed algorithm can provide a theoretical basis for the prediction of surface roughness in screw rotor abrasive belt grinding.

Abstract:Focusing on the problem that magnetic interference of carrier in magnetic anomaly detection will greatly affect the measurement results, a carrier interference magnetic field compensation method for magnetic anomaly detection based on RLS algorithm is proposed. First, classify the interference in the magnetic anomaly detection and establish a measurement model. Then, convert the model to a standard multiple linear equation. Reduce collinearity through orthogonal and linear constraints. Solve the model parameters through the RLS algorithm and the correctness of the algorithm is verified by model simulation. Finally, carry out finite element simulation of the carrier with and without anomalies. The RLS algorithm is used to calculate the compensation parameters from the simulation data without abnormality. The calculated compensation parameters are used to compensate the abnormal simulation data. The results show that the compensation error is significantly reduced. The signal improvement ratio after compensation for experiment without magnetic anomalies is 10. 3. This method can be used for carrier interference magnetic field compensation in magnetic anomaly detection. It effectively improves the anti-interference ability of magnetic anomaly detection, and is of great significance to buried object detection, mineral exploration, anti-submarine and other works.

Abstract:Target perception in complex scenes is one of the most important research fields of deep learning in computer vision, and vehicle detection in complex traffic scenes is the object of research by many scholars today. In the process of video target detection, due to the insufficient utilization of the time dimension feature information of moving objects, time features between long sequences are extremely easy to be ignored. This paper proposes a spatio-temporal consistent video vehicle detection and tracking algorithm. The algorithm is composed of a two-branch network structure: one of branch is composed of transformer network modules based on spatial correlation. The branch network is mainly used to determine the correlation between the previous and subsequent frames, perceive the consistency between adjacent frames, and predict the temporal and spatial consistency of the target vehicle relevance; another network branch is composed of network modules based on cross-feature pyramid fusion. This module mainly extracts the local information of the detected object combined with shallow spatial edge information and high-level semantic feature information. This branch extracts the spatial position of the object characteristic information. The network structure combines the Transformer mechanism and the cross-feature pyramid module, and uses the advantages of Transformer’s sensitivity to the time correlation between long sequences and the feature pyramid network module’s sensitivity to edge information to detect and track video frame objects to ensure neighboring the long-range correlation of the frame is deeply integrated with the feature information of the edge and the deep layer. The experimental results show that the dual-branch network structure designed in this paper achieves better accuracy and faster convergence speed in video target tracking and detection. At the same time, experiments in saliency video target detection show the effectiveness and generalization of the algorithm.

Abstract:Aiming at the problem of low underwater object detection accuracy caused by low-quality underwater imaging, different shapes or sizes of underwater objects, and overlapping or occlusion of underwater objects, an underwater object detection algorithm combining data enhancement and improved YOLOv4 is proposed. By adding CBAM ( convolutional block attention module) to the backbone of YOLOv4—CSPDarknet53, the feature extraction ability of network model is improved. In order to enhance the multi-scale feature fusion ability, the same-layer skip connections and cross-layer skip connections are added to PANet (path aggregation network). To enhance the robustness of the network model, the data enhancement method PredMix (prediction mix) is used to simulate the incomplete display of underwater organisms such as overlap or occlusion. The experimental results show that the detection accuracy of the underwater object detection algorithm combining data enhancement and improved YOLOv4 on URPC2018 dataset is improved to 78. 39%, 7. 03% higher than YOLOv4, which fully proves the effectiveness of the proposed algorithm.

Abstract:In order to reduce the threat of radar detection to stealth equipment such as missiles, aircraft and ships, an ultra-wideband (UWB) high gain Vivaldi antenna with low radar cross section (RCS) was designed. By analyzing the surface current distributions of the antenna at different frequencies, a simple and effective method is adopted to change the shape of the antenna without affecting the radiation performance of the antenna, and a rectangular and semicircular shaped structure is loaded on both sides of the radiation arm. The proposed structure can reduce the scattering on the surface of the radiator, leading to the monostatic RCS reduction. The measurement results show that the proposed antenna has a working bandwidth of 4. 7~ 11 GHz and a maximum gain of 11 dBi. Compared with the original antenna, the proposed antenna maintains good radiation performance. At the same time, the maximum reduction of RCS is 18. 5 dB in the working frequency range. The proposed antenna can be applied to stealth systems with low RCS.

Abstract:Low voltage operational amplifiers and their digitally assisted calibration algorithms are critical in the design of ultra-high speed, high resolution analog-to-digital converters (ADCs). A 500 MS / s, 12-bit pipeline ADC based on a 40 nm CMOS process and operating voltage of 1. 1 V has been proposed. This ADC adopts a sample-and-hold (SHA) less front-end structure and low-voltage interstage operational amplifiers (opamp) to reduce power consumption. A foreground calibration algorithm using digital detection is designed for gain error and capacitance mismatch calibration, effectively improving the overall performance of the ADC using smaller area and power consumption. This digital calibration scheme improves the differential nonlinearity (DNL) and integral nonlinearity (INL) of the ADC from 2. 4 LSB and 5. 9 LSB to 1. 7 LSB and 0. 8 LSB. for a 74. 83 MHz sinusoidal signal, the calibration technique achieves a signal-to-distortion noise ratio (SNDR) of 63. 14 dB and a spurious-free dynamic range ( SFDR) of 75. 14 dB, respectively, with a power consumption of 123 mW, which meets the design targets and demonstrates the effectiveness of a low-voltage pipeline ADC design with digital correction.

Abstract:The accumulation of intermittent faults will cause the deterioration of the health of the electronic system. Correctly identifying the severity of intermittent faults can ensure the safe operation and reduce maintenance costs of the electrical systems. However, it is difficult to extract intermittent fault features accurately, which leads to the failure of traditional identification methods. This paper proposes a method for identifying the severity of intermittent faults based on LSTM network. First, the intermittent faults are injected into the electronic system to obtain sufficient training data of different severity. Then use these data to train the classifier which is constructed by LSTM network and the softmax fully connected layer network. Finally, by injecting faults into typical circuits and using the trained LSTM network to identify the severity of intermittent faults, the experimental results prove the effectiveness and feasibility of the method.

Abstract:In order to solve the problem of denoising power signals with a wide spectrum range, an adaptive wavelet packet de-noising algorithm based on the decider of power spectrum characteristics was proposed to dynamically adjust the number and bandwidth of subband, reasonably preserve the effective high-frequency components of the signal. The main feature of this method is to achieve nonuniform decomposition by using power spectrum amplitude segmentation. The coefficient selection of wavelet packet de-noising is optimized, which improves the adaptability of the denoising algorithm. This method was used to denoise the power signals of the multielectric aircraft ATRU test platform, and the utility of adaptive non-uniform subband decomposition was achieved. Experimental results show that the algorithm can effectively retain 89. 21% high-frequency details of the signal and obtain a better denoising effect.

Abstract:Aiming at the problems of inaccurate attitude estimation and noise in the estimation of a standing ball robot during slope motion due to the system oscillation orientation, an improved complementary filtering algorithm based on Huber robust estimation is designed. Firstly, the dynamics and kinematic performance of the standing ball robot on the slope are analyzed. Secondly, Huber robust estimation is used to suppress the interference noise caused by the fuselage oscillation during the climbing process of the standing-ball robot, and complementary filtering is improved to solve the attitude. Finally, in order to reduce the interference of magnetometer output instability to attitude calculation, a data smooth switching method is designed. The experimental results show that compared with the complementary filtering algorithm, the proposed algorithm improves the estimation accuracy of pitch Angle, roll Angle and yaw Angle by 24. 82%, 11. 73% and 35. 65%, respectively. The algorithm can effectively suppress the noise caused by the body oscillation during the sloping motion of the standing ball robot, and can ensure the accuracy and real-time performance of the attitude solution.

Abstract:Aiming at the long runtime of windowed Fourier filtering ( WFF) in the filtering of fringes of electronic speckle pattern interferometry, a faster Fourier filtering method based on a two-dimensional Kaiser-Hamming window (KH-WFF) is proposed. First, the Kaiser window and Hamming window are combined to design a two-dimensional Kaiser-Hamming window, and then the Gaussian window in WFF is replaced with the two-dimensional Kaiser-Hamming window. The time complexity is reduced due to the removal of redundant non-orthogonal basis functions. KH-WFF and WFF are applied to the simulated fringe and the experimental fringe, and the results show that the smoothness, detail protection, and fidelity of the filtered fringe are not much different. Moreover, both methods have obtained the unwrapping phase with good continuity, while the filtering speed of KH-WFF for a fringe in practical application is 12 times that of WFF. Finally, the time complexity analysis of KH-WFF and WFF under different image sizes, window sizes, and bandwidths proves that KH-WFF is faster. Experiments prove that the proposed method provides almost the same result while reducing the execution time.

Abstract:Aiming at the shortcomings that the important feature information of the original vibration signal of the rolling bearing is submerged by strong background noise, and the extracted time domain features have high redundancy and strong relevance, this paper proposes a new rolling bearing fault diagnosis research method based on maximum relevance-minimum redundancy ( mRMR) feature selection and random forest. First, the original signal is subjected to complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to obtain a series of intrinsic modal functions ( IMFs), analyze IMF and remove high frequency noise and part of false component, then reconstruct the signal and extract its time domain characteristics, mRMR is used to remove redundant and highly correlated feature vectors, so that the selected feature subset has the greatest dependence on the label, and finally the feature subset is input to the random forest classifier for classification. Experiments show that mRMR has an excellent feature search strategy, the important features are selected first. Only three features are needed to achieve a higher classification accuracy, and the efficiency is higher than other feature selection algorithms.

Abstract:The unbalanced traction current and the frequency shift signal of the ZPW2000 series track circuit have a common transmission channel. The traction returns and higher harmonic components will affect the frequency shift signal. When the train relies on the CTCS- 2 train operation control system to provide operation permission, increasing the risk of safe operation. In order to effectively remove highorder harmonic interference and obtain jointless track circuit frequency shift information, a frequency shift signal processing method based on the combination of VMD and Hilbert transform is designed. First, use VMD to decompose the harmonic interference signal into several IMFs of different frequency bands, solve the center frequency of all IMFs, determine the interference frequencies of each harmonic, and determine the eigenmode function corresponding to the current jointless track circuit according to the prediction algorithm and correlation verification; finally, determine the current jointless track circuit mid-shift by performing Hilbert transform analysis on the IMF Frequency information. Through the analysis of simulation and laboratory measured signals, it is found that the method not only effectively suppresses the mode aliasing phenomenon to make the harmonic interference components accurately separate from the mixed signals, but also accurately solves the frequency shift information, which provides a reference for the accurate demodulation of driving permit signals under interference conditions.

Abstract:In order to reduce the impact of satellite signal loss and cumulative error of inertial navigation during vehicle driving, a vehicle heading angle estimation method based on line detection and digital map matching is proposed by combining scene feature extraction and expression with digital map information. Firstly, according to the coordinate points of the map matching, the corresponding points azimuth of the lane line map is calculated, and the angle difference between the vehicle heading angle and the lane line point azimuth is calculated. Secondly, the angle of lane line in image is recognized and calculated by the improved FLD line detection method. The angle of bilateral lane lines is taken as the input of BP neural network, and the predicted angle difference is taken as the output of the network. Finally, the vehicle heading angle is obtained by combining the angle difference and lane line azimuth. The results of the experiments show that the proposed heading angle estimation algorithm has certain advantages over the existing methods and ordinary measurement sensors.

Abstract:To address the problem of poor heat dissipation in the axial ventilation holes of the rotor of the current AC traction motor, a new axial ventilation slot structure is proposed for the teeth of the rotor. According to this distribution diagram, a new rotor teeth axial venting slot structure is proposed for the problem of high temperature of the rotor end rings and teeth; the influence of the new venting slot structure parameters on the motor rotor heat dissipation effect and the motor related performance indexes is analyzed, and a new rotor teeth axial venting slot structure is proposed. The results show that the proposed new rotor teeth axial venting slot structure significantly reduces the temperature of the rotor teeth and end rings compared to the traditional axial venting slot structure, while keeping the motor performance indexes basically unaffected, thus significantly reducing the temperature of the rotor teeth and end rings. It improves the ventilation and heat dissipation effect, which is important for reducing the total body temperature rise of the motor and thus extending its service life.

Abstract:To improve the accuracy of the algorithm for vehicle detection and solve the problem that the original algorithm is not effective in the complex traffic scene, a vehicle detection method based on attention mechanism and improved densely connection network structure was proposed. Firstly, SoftPool was used in the transition layer to consolidate the characteristic information between the dense blocks. Secondly, the expression of effective channel features was enhanced by the lightweight channel attention mechanism, it was used as the deep feature extraction layer of Darknet-53. The CIOU loss was used as the prediction loss term of the bounding box position of the model, and reduce the model volume using deep separable convolution. Compared with the original algorithm, the mAP value is increased by 2. 6%, and the model volume is reduced to 42%. Experimental results show that the algorithm has good detection performance in complex traffic scene.

Abstract:In the large-scale automatic production process, in order to test different types of TV motherboards, it is very necessary that video test sources can output different formats, multi-interfaces and multi-resolution in parallel. Through the exploration of multimedia video transmission principle and video signal interface formats, the video test signal source is designed and realized in an advanced Artix-7 series large-scale programmable XC7A100T chip, which can output both ultra-high-definition and high-definition video HDMI interface signal, and analog standard definition video signal in VGA, CVBS and YPRPB format. First of all, parallel digital baseband video signals with different resolutions and field frequencies are generated according to the principle of video display. Then, the parallel ultra-high-definition and high-definition digital video signals are converted into serial TMDS signals in order to be sent to the HDMI interface for display; and the parallel standard resolution signal is sent to the modulation module and digital-to-analog converter to become analog video signal in VGA, CVBS and YPRPB format. In particular, the parallel ultra-high-definition and high-definition digital video signals are converted into serial TMDS differential signal inside the XC7A100T chip, which saves the video parallel-serial processing chips and reduces the cost of equipment. The whole system improves the integration level of video test signal, reduces the volume of test equipment, and the manufacturers of video equipment can be more convenient to test.

Abstract:Brain-computer interface (BCI) based on motor imagery (MI) has been applied to the plasticity rehabilitation of limb motor function in recent years. Visual assistant stimulus can improve the classification performance of MI-BCI. However, for users with impaired visual system, visual assistant stimulus cannot be used. Therefore, this paper designs ASMI-BCI based on auditory assistant stimulus. It has been found that dynamic acoustic assistant stimulus could improve the excitability of motor related cortex, and enhanced the separability features of related frequency bands. The average classification results of the three experimental paradigms (C-SW, CDA, C-DV ) for 10 college students (5 males and 5 females, with an average age of 22. 6 years old) showed that the classification accuracy of C-SW paradigm was the lowest, followed by C-DA, and the accuracy of C-DV paradigm was the highest. The optimal classification accuracy of the auditory assistant stimulus paradigm was 76. 03% and the average classification accuracy was significantly improved by 8. 83% compared with the traditional MI-BCI paradigm. For 60% of the subjects, the classification accuracy of this paradigm can reach higher than 70%. The dynamic auditory assistant stimulus paradigm can provide a new pattern and method of feature modulation and BCI performance enhancement for patients with visual impairment.

Abstract:In view of the current emotional electroencephalogram (EM-EEG) identification research on time scales is difficult to grasp the time domain information and the spatial domain information is easy to ignore the recognition rate is stagnant, and collect the EM-EEG with too many channels in the excessive information redundancy and increasing cost of information processing problems, it puts forward the space-time convolution based on CNN optimization study on EM-EEG identification fusion network. The fusion network is composed of a parallel long convolution (L-Conv) CNN that extracts EM-EEG time domain information and a CNN that extracts EM-EEG spatial information. Particle swarm optimization (PSO) is used in the time-space optimization of the CNN model. The L-Conv scale in CNN has been optimized, and use the short time power spectrum ( STPS) correlation analysis method of the spatial CNN channel number optimization model, temporal and spatial domain features in EEG are extracted deeply and effectively. The results show that the proposed optimization of space-time convolution integration CNN on SEED IV data set for peace, sadness, fear, happy four final accuracy can reach 90. 13%, compared with the traditional single CNN recognition accuracy rate increased by 4. 76%, and channel number from 62 to 33 road, shrank by 46. 77%, confirmed the feasibility of this method.

Abstract:The network request of smart phone leads to the decline of its endurance. The combined forwarding technology can effectively reduce energy consumption. But setting the optimal combined forwarding time is still the key to technological development. And it is time-consuming and labor-intensive to solve this problem through a large number of manual experiments. Therefore, this paper based on statistical model checking and used the tool UPPAAL-SMC to conduct simulation modeling of user requests and WiFi modules in Android devices with probabilistic time automata. Quantifying energy consumption, delay, user satisfaction and other attributes. And then used statistical model checking to conduct Monte Carlo simulation of scenarios with different request frequencies to obtain the effect of acquisition delay on energy consumption and user satisfaction. Finally, multi-objective optimization was carried out to obtain the general optimal combined forwarding delay time of 22 s, which reduced energy consumption by 20% on average on the premise of satisfying user experience. This method can simulate the general optimal merge forwarding delay time in different application scenarios, which can provide reference for developers.

Abstract:ZY(J)7 electrohydraulic turnout switch equipment has been widely used in high-speed railway, and accurate fault diagnosis is helpful to the daily maintenance of high-speed railway turnout. Taking the fault text data of ZY( J) 7 turnout as the research object, a fault diagnosis model for high-speed railway turnout was proposed, which combined LDA topic model with association rules classification technology. Firstly, this model adopted LDA topic model to extract the feature of ZY(J)7 turnout fault text data. Secondly, due to the unbalanced data of each fault type of turnout, the original association rule classification algorithm was introduced into the concept of class support to deal with unbalanced data, and finally the fault diagnosis of ZY(J)7 switch was realized. Through the experimental analysis of ZY(J)7 turnout fault text data of a railway bureau from 2017 to 2019, the experimental results indicate that the classification precision and recall rate of the proposed fault diagnosis method are 95. 08% and 90. 24% respectively, which not only guarantees the accuracy of the whole classification, but also gets better classification performance of minority class.

Abstract:During the operation of the bulb tubular hydropower unit, due to hydraulic factors, machinery, working conditions and other factors, it is easy to cause the runner blades and runner chamber to malfunction, which seriously affects the safe operation of the hydropower unit. Based on the analysis of the fault signal characteristics of the runner blades and runner chamber of the bulb tubular hydropower unit, an online fault detection method for hydropower units based on K-Means and Wright's criterion is proposed. This method uses principal component analysis (PCA) to reduce the dimensionality of the vibration and noise signal characteristics of the hydropower unit, and integrates the Wright criterion to improve the traditional K-means algorithm to realize the adaptive selection of the K value, and perform online clustering of the features, which can quickly and accurately identify the variable load state of the turbine and the failure of the metal sweeping chamber. The method proposed in this paper is applied to the fault detection of the bulb tubular unit of Wuling Electric Power′s Jinweizhou Hydropower Station. The experimental results show that the accuracy of the online fault detection using this method is 100% and the accuracy of the variable load online detection is 96. 7%, there has been no fault false positives and false negatives in the past 10 months of operation, indicating the effectiveness of the method.