Abstract:The on-line measurement of steam humidity has an important effect on the safe operation of condensing steam turbine. In order to accurately calculate the experimental error of the resonant cavity affected by temperature, the theoretical analysis and experimental study of the resonant cavity temperature characteristics were carried out. Based on the research status at home and abroad, exhaust steam humidity is measured according to Maxwell equation. According to the size change of the cavity in the temperature change and the analysis of the vapor humidity data measured at different temperatures, the influence degree of temperature on the vapor humidity measured by the resonant cavity is obtained. According to the data, the conversion is about 43% steam humidity, and the normal steam turbine exhaust steam humidity is not more than 13%. The measurement error can be greatly reduced by changing the resonant cavity installation mode. The research results are of great significance to improve the on-line measurement accuracy of steam humidity.

Abstract:Due to the influence of non-linear factors on the optical fiber displacement sensor, the measurement error of the sensor is relatively large. To this problem, this paper proposes a compensation measure to reduce the impact of these non-linear factors, using the firefly algorithm (FA) to optimize the back propagation neural network (BPNN) hybrid algorithm to improve the optical power value received by the sensor. The algorithm not only uses the search performance of the Firefly algorithm to find the best position of the particle population, but also utilize the strong local optimal weight threshold search performance of BPNN, and finally achieves the goal of preventing BPNN from falling into the best optimization situation in some samples. During the experiment, the optical power values received by the two parts of the sensors are exploited as data input into the FA-BP algorithm for training and optimization. Finally, compared with BPNN and particle swarm optimization BP neural network (PSO-BP), the FA-BP algorithm with higher convergence accuracy and few iteration steps can effectively improve the accuracy of sensor data and the running speed of the program.

Abstract:In order to meet the requirements of high-precision current feedback in power control and real-time current detection in fault monitoring of grid-connected inverters, a new method of grid-connected inverter current acquisition and rapid overcurrent detection based on Sigma-Delta ADC technology was proposed. In this method, two SinC3 filters with different extraction rates were used to demodulate the filtered data stream respectively, which achieved the flexible usage of the modulation on the same data flow. The data stream was modulated by a second order sigma delta modulator,SinC3 filters were implemented in the digital signal processor (DSP). Test results show that the proposed method realizes the inverter grid current 13 effective resolution bit isolation sampling and 3.4μs rapid over-current detection.

Abstract:In order to study the influence relationship between the output voltage of the infrared detector made by the thermopile principle and the key structure size, to obtain a high output voltage, a microelectromechanical system (MEMS) Thermopile Infrared Detector based on circular structure was proposed. The size of the infrared detector is 1.8mm×1.8mm×0.35mm and the cavity size was 1.2mm as the research basis of modeling and simulation. Modeling the Infrared Detectors, by changing the absorption layer coverage length, thermocouple logarithm, the cold end, presence of etching rules hole size parameters of the key structure, such as the Infrared Detector thermoelectric heat simulation and path analysis, ultimately determine the optimal infrared detector simulation results, the simulation results show that by changing the Thermopile Infrared Detector size parameters of the key structure, The output voltage of the designed Infrared Detector is up to 125.08μV, which greatly improves the output voltage of the external Infrared Detector compared with the traditional four-terminal beam structure.

Abstract:In order to deal with and locate the pipeline leakage in time and solve the problem of misreporting and omission rate and inaccurate positioning of the current pipeline leakage monitoring system, a leak monitoring and positioning method combined with negative pressure wave and flow balance method is proposed, and a combination of LabVIEW and MATLAB is used to develop the real-time pipeline leakage monitoring and location software. Firstly, the software controls the NI demodulator SM130 and the electromagnetic flowmeter, realizes the synchronous acquisition of fiber Bragg grating sensor signals and electromagnetic flowmeter signals, which can monitor the pipe pressure and flow state parameters timely. Leak warning is provided when sensor detects negative pressure wave signal and the flow difference between upstream and downstream exceeds the threshold value. Finally, it calls the positioning method written by MATLAB to locate the leakage and connect with MySQL database for data management. The test proves that the system can accurately identify the real leakage and maintain good monitoring reliability for the function of leak warning. For the leak locating test, the averaging method of multi-group locating based on sensor array is put forward, and the accuracy and stability of locating are verified in repeated tests. Its advantages of good interactive performance and strong scalability are in line with the development process of future pipeline automation management and have important engineering practical value.

Abstract:In view of the problem that the conventional single isolation ship rolling mode in the shipborne TT&C system cannot meet the requirements of TT&C under complex sea conditions, this paper proposes to adopt multi-combination isolation ship rolling technology for TT&C. This paper analyzes the system principle of isolation ship rolling mode of shipboard measurement and control system, and puts forward the method of selecting appropriate combination of isolation ship rolling according to different environment to realize stable tracking target and improve the isolation degree of ship rolling. Mainly introduced the principle of swinging ship various isolation technology and the similarities and differences, using the test data, this paper compares and analyzes the results for convenient combined isolation in different sea environment, the application of the results show that through the combined isolation more accurate selection, can better and more stable tracking target, to show more swinging ship combined isolation technology for system stability tracking and essence of determine the rail has important significance.

Abstract:Fiber coating is very important to the performance of fiber. In the process of measuring the geometric parameters of the coating layer, because the cutting process of the coating layer is easy to produce a large deformation, and often causes the edge fuzzy, so the method of direct edge fitting to obtain the geometric size of the coating layer will cause a large error. In this paper, the wavelet threshold method is used to remove the noise part in the high-frequency signal of the image to protect the edge information. The effective edge is extracted by the Canny calculation of adaptive threshold. The variable radius is used to limit the edge in the case of large deformation, and the geometric parameters are obtained by function fitting. Taking the diameter of a coating layer as an example, the diameter measured by the standard instrument is 185.298µm, and the repeat measurement accuracy is 0.112µm. When the coating layer is not cut properly, the measurement value of the instrument becomes 184.516µm and the measurement accuracy becomes 0.198µm. The measured values before and after the method in this paper are 184.599µm and 184.613µm, and the measurement accuracy is 0.054µm and 0.075µm respectively. Experimental comparison shows that the method in this paper effectively improves the accuracy and stability of the measurement of the geometric parameters of the optical fiber coating.

Abstract:In the field of small target detection, many algorithms improve the accuracy at the cost of increasing the complexity of the model, but it brings a large computational burden and equipment requirements. Aiming at the contradiction between complexity and detection accuracy in the model, an improved resampling strategy algorithm on image pyramid is proposed. The algorithm only needs to calculate a small amount of sample data and introduce a lightweight attention mechanism module with a few parameters. In the experiment, training and testing were carried out on the COCO dataset. The resampling strategy mAP value was 40.6%, and the improved value was 42.1% with the introduction of the attention module, and the weight file size only increased by 2% with the introduction of the attention module. The experimental results show that the improvement of the sample resampling algorithm can improve the detection accuracy while reducing the computational burden, which verifies the effectiveness of the lightweight attention module.

Abstract:In the field of small target detection, many algorithms improve the accuracy at the cost of increasing the complexity of the model, but it brings a large computational burden and equipment requirements. Aiming at the contradiction between complexity and detection accuracy in the model, an improved resampling strategy algorithm on image pyramid is proposed. The algorithm only needs to calculate a small amount of sample data and introduce a lightweight attention mechanism module with a few parameters. In the experiment, training and testing were carried out on the COCO dataset. The resampling strategy mAP value was 40.6%, and the improved value was 42.1% with the introduction of the attention module, and the weight file size only increased by 2% with the introduction of the attention module. The experimental results show that the improvement of the sample resampling algorithm can improve the detection accuracy while reducing the computational burden, which verifies the effectiveness of the lightweight attention module.

Abstract:With the rapid development of wireless communication technology, indoor positioning technology based on wireless sensor networks is widely used in daily life. However, indoor environment is easily disturbed by obstacles, which causes reflection and scattering during signal transmission, resulting in energy loss and lower ranging accuracy. In order to improve the accuracy of received signal strength indication (RSSI), this paper analyzes the relationship between distance and RSSI through experiments, selects the best communication distance as 4m, applies it to the propagation model, and proposes a dynamic path loss parameter ranging method, finds out the best neighborhood where the node to be measured is located, calculates the path loss parameter n by using the regional centroid, and obtains the position coordinates of the node to be measured by using trilateration positioning algorithm. Experiments show that compared with the fixed path loss parameter location method, the positioning accuracy of the dynamic path loss parameter location method is improved by 56%, and the ranging error is greatly reduced.

Abstract:To solve the problem of downlink power control of dense UAV networks, reduce the mutual interference of UAVs and improve the energy efficiency of the system, a downlink transmit power control algorithm for dense UAV networks is proposed. First, convert the power control problem of the UAV network into a mean field game theory model in a high-dimensional system state to reduce the mutual interference between UAVs.Second, convert the mean field game theory model into a Markov decision process, in order to obtain the equilibrium solution under the dense deployment of UAVs.In addition, a mean field game theory algorithm based on deep reinforcement learning is proposed, which obtains the optimal power control strategy of the system by using a deep neural network to maximize the energy efficiency of the system.Finally, the proposed method is compared with the other three algorithms through simulation analysis. The experimental results show that the proposed method can effectively interact between drones and the environment, effectively reduce the mutual interference of drones, and enhance the system network communication performance; at the same time, compared with the other three methods, the proposed method has a faster convergence Faster, more energy efficient, with good convergence and reliability.

Abstract:In view of the characteristics of strong coupling, non-linearity and natural instability of the balancing car and the defects of the conventional PID control, the balance is difficult and the reliability is low. This paper analyses the conventional PID control and single neuron control, the combination of value through balancing the car unit of input error, according to the rules of self learning, adjust each parameter and control links in proportion, integral and differential control, again by self-tuning fuzzy controller on the gain coefficient, thus improve the adaptive ability of the system. Finally, through MATLAB simulation, the controller proposed is compared with the conventional PID controller, which verifies the superiority of the proposed method. And the improved single neuron PID controller has better tracking performance and anti-interference capability.

Abstract:This paper propose an improved yolov4 algorithm for detecting fatigue driving. First, under the transfer learning, the weight of VOC dataset has been trained as pre weight for training. Then in the training, convolution is added before and after the SPP structure of feature pyramid in the frame to improve the extraction of deep features. And introduce the dilated convolution to increase convolution output receptive field and the ability to obtain image location information. Experimental results show that the map value of the improved yolov4 algorithm during the test is 97.29%, 1.98% higher than raw yolo v4 algorithm, the detection of eye parts increased by 6%. Add the frame delay to the detection, for avoiding other behaviors affecting the results and reducing the probability of misjudgment.

Abstract:This paper designs a set of digital identification system for the taxi meter, which is used to replace the manual work of unmanned verification of the meter. Through the SIFT template matching algorithm, the position of the meter in the video screen is accurately located and the target image is corrected into a front view. Finally, locate the position of the jump number according to the label of the template image. Aiming at the number type displayed by the meter, a number recognition algorithm based on the seven-segment code principle is developed. Field tests show that the recognition rate of a single frame image reaches 100% when the digital positioning is accurate. In continuous video, when the digital transition process, the image is captured, which will cause recognition errors. After optimization, this error is judged as a digital jump. After detecting the jump of the meter's mileage, it immediately sends a signal to the verification station and records it. Experiments show that the video processing speed reaches 25 fps, which far exceeds the limit of human eye resolution, and can replace the manual verification of the digital jump of the meter.

Abstract:A line spectrum detection method improved by U-net network is proposed, Which is aiming at the problem of low detection accuracy of the irregular fluctuating line spectrum in passive sonar system. The system framework is based on U-net network, and the residual structure is introduced to increase the depth of the network and strengthen the learning ability of the model feature. Meanwhile, the feature channel attention mechanism is brought in the encoder part which can help the model learn the importance of different features between channels, thus improving the feature expression ability of the model. Finally, the DUpsampling up-sampling method is used in the decoder part, and the redundancy capability in the segmented label space is utilized which can accurately restore the pixel-level prediction. The improved model is compared with the HMM model and CEM model on the line spectrum detection effect. Under the SNR of -24dB to -20dB, the accuracy of line spectrum detection of the improved model is 0.314-0.526, which is better than that of HMM model and CEM model.

Abstract:On the issues of solving several problems such as low accuracy, water waste and low efficiency in traditional Standard Tank verification methods. This paper provides a set of standard tank automatic verification method based on the principle of visual technology. The standard facility of capacity verification is developed by this method. Moreover, we developed the mechanical structure automatically docking with metal tank and designed a verification system, whose software and hardware is based on LabVIEW platform and PLC. In addition, it conducts the uncertainty evaluation on the devices as well. According to results of this principle, the standard facility of this structure can maintain stability of capacity and state parameters during verification. Experimental results indicate the verification efficiency is improved by more than 30%, the uncertainty of verification is 6.4mL(k=2), which can meet the requirements of detection for metal tank.

Abstract:In order to solve the problems of automatic hyperbolic feature extraction, irregular curve elimination and incomplete hyperbola fitting in the ground penetrating radar B-scan (GPR B-Scan) image, a multi-layer fusion processing GPR B-Scan image double Curve extraction model. First, the preprocessing layer uses the OTSU threshold segmentation algorithm and the morphological corrosion method to preprocess the image, and then through the clustering layer connection clustering algorithm to complete the automatic extraction of hyperbolic features to eliminate irregular curves, and finally the hyperbolic fitting layer uses robust orthogonal The distance fitting algorithm simplifies the hyperbolic fitting process. Finally, this paper first built the GprMaxs platform simulation experiment, the average fitting accuracy rate reached 97.64%, and then carried out the actual GPR B-Scan data experimental verification in the two actual working conditions of concrete and wet soil, the average fitting accuracy rate reached 92 %the above. The results show the validity and correctness of the design model in the article, which lays the foundation for subsequent application and promotion.

Abstract:Water level monitoring is of great significance to the comprehensive management of water resources and navigation. With the development of machine vision technology, the image water level method is expected to replace observing the water gauge. Firstly, the history and the situation of water level detection are briefly described. The advantages of image method are illustrated by comparing various water level detection methods Then, Three key steps of image water level detection are analyzed: water gauge segmentation and distortion correction, water level detection and water level conversion. The methods and development of each step for each step are summarized at home and abroad. Finally, the problems of image water level detection at present are summarized and the future direction is prospected.

Abstract:In remote sensing image target detection, the remote sensing image is usually arranged in any direction under the top view angle. This situation makes common detection algorithms have good detection results in natural scenes, but the detection results are often unsatisfactory in remote sensing images. Aiming at the problem of unsatisfactory detection in remote sensing scenes, this paper proposes a remote sensing image target detection algorithm (CSL-YOLOv5) based on the rotating target frame and attention mechanism based on the single-stage detection network YOLOv5. First of all, the original network feature extraction network (CSPDarknet53) was modified to increase the number of output feature maps and optimize the detection effect of the network on small targets. Then, an attention mechanism that combines the channel module and the spatial module is added to the residual block to enhance the expression effect of image features. At the same time, Focal loss is used to optimize the training effect, and the detection accuracy is improved on the basis of ensuring the detection speed. Finally, the long-side representation based on circular smooth labels is used to achieve the rotation of the target frame, and the effect of angle periodicity on training is solved by turning the regression problem into a classification problem. The experimental results show that the CSL-YOLOv5 algorithm proposed in this paper achieves a detection accuracy of 76.24mAP in the DOTA data set, which has a higher accuracy compared with the previous single-stage algorithm, and has an increase of 8.06% compared to the mAP of YOLOv5. The algorithm in this paper has high detection accuracy and good robustness in remote sensing scenarios.

Abstract:The accuracy of power dispatching speech recognition system is related to the effect of language model. In order to improve the accuracy of power dispatching speech recognition, this paper proposes a class label language model based on double dictionaries (general dictionary and power dispatching domain word dictionary). The model improves the n-gram language model and adds class label information, so as to improve the accuracy of power dispatching speech recognition. At the same time, a joint method of word segmentation and part of speech tagging based on double dictionaries is proposed. The system is used for word segmentation and class label labeling of corpus, and then improves the adaptability of class label language model based on double dictionary to power dispatching language. Finally, the comparison experiments between the proposed language model and the common statistical language models are carried out on the collected command set of power dispatching. In addition, the joint system and other word segmentation and part of speech tagging systems are compared by experiments. The simulation results show that the efficiency of word segmentation and part of speech tagging is higher in the joint system. Considering the comprehensive factors of semantic information, dictionaries, word segmentation and part of speech tagging system, the error rate of the proposed model in power dispatching language recognition is only 4.14%.

Abstract:The identification and classification technology of printed circuit board (Printed Circuit Board) surface mount components plays an important role in the production process of modern electronics industry. A target detection method based on YOLO v3. First, an industrial camera is used with an optical lens to construct a data set of mounted components, and secondly, the feature pyramid structure FPN (Feature Pyramid Networks) of YOLO v3 is redesigned, and then the K-means method is used to improve the clustering of the mounted component data set. Get the Mouted anchor and corresponding parameters. Finally, use Mounted anchor and network structure to retrain the improved YOLO v3, and compare experiments with the original network to verify the recognition and classification effect of the mounted components. The experimental results show that the improved YOLO v3 mounted component recognition and classification technology has an average accuracy rate of 9% higher than that of the original network, and a slight increase in the recall rate.

Abstract:In view of some problems in the traditional method of using Hough transform to identify the number of pointer meter, an improved reading method of pointer meter is proposed. Firstly, the algorithm corrects the tilt of the instrument image which can't be acquired due to the angle deviation. Through the comparison of several methods, it is proved that the proposed method is better. And according to the problem that directly using Hough transform to fit the pointer line will appear the influence of the border or other marks, which leads to the poor effect, the depth learning yolov5 model is introduced to realize the direct positioning of the pointer, so that the time required for line detection will be greatly reduced. After preprocessing the pointer image, image denoising, binarization and thinning, Hough linear transformation, angle method is used to realize the indicator recognition of pointer instrument. The experimental results show that the algorithm solves the limitation of the traditional recognition method, and the error between the recognition indication and the actual value of the pointer meter is - 2.73-3.64v, and the relative error is less than 2.47%, which meets the requirements of industrial meter reading.

Abstract:When detecting dim and small targets with low signal-to-noise ratio in infrared image sequences of sky background, the simple traditional algorithm has some problems, such as complex preprocessing process, difficult feature design, difficult to determine control parameters, low detection accuracy and so on. Through the introduction of deep learning technology, a combined algorithm is proposed, which can significantly improve the detection effect of the algorithm. In the infrared image sequence, firstly, the moving target is detected with high accuracy by using the spatio-temporal feature extraction network based on YOLOv3 in the starting frame, and then the traditional method based on local contrast feature is used to detect the target quickly in the subsequent frames according to the speed and luminance characteristics of the target. In the infrared image sequence test data set of the sky background, the combined method achieves higher accuracy and recall than the existing methods, and the computing time also meets the real-time requirements. The results show that the two methods cooperate with each other and achieve a good balance in real-time and accuracy.

Abstract:The development of Real time error compensation controller can detect and identify the temperature field of five axis machine tool with double turntable by using temperature sensor.It can also measure the thermal error of the tool in the directions of X, Y and Z relative to the worktable by the displacement sensor. The communication interface between compensation controller and CNC system of machine tool is established by using the developed communication interface system of machine tool displacement. So the displacement signal of the machine tool can be transmitted to the compensation controller in real time and the temperature sensor for compensation is installed at the established sensitive heat source point.The connection between temperature sensor and compensation controller is established by the temperature acquisition system. The temperature signal is introduced into the compensation controller, which  can measure the temperature of machine tool and thermal error of spindle in three directions manually and automatically. Compared the measured data with the model output, the residual error of prediction model is small. The compensation effect is better.

Abstract:With the development of acquisition technology, the accuracy of acquisition and the amount of data are getting larger and larger, and higher demand is put forward for data storage and transmission. In order to solve the problems of fast acquisition speed and strong real-time performance of embedded system, this paper designs a data transmission system that combines field programmable gate array (FPGA), NAND FLASH and Gigabit Ethernet PHY chip VSC8641. Through in-depth research on the pipelining storage mode of FLASH and Gigabit Ethernet transmission technology, The design of high-speed FLASH controller and the full duplex mode grouping and unpacking of Ethernet frame format based on UDP and IP protocol are realized. And on the hardware platform developed by ourselves, it is verified that the scheme can store the collected data in FLASH, and transmit it stably and high-speed to PC for processing.

Abstract:Aiming at the problems of low precision and poor quality of analog signal acquisition in telemetry system, In this paper, a higher precision analog output card is designed under the trend of increasingly accurate, integrated and complex analog signal. In this design, FPGA centered logic controller and 16 bit high-precision DA converter are used to realize the high-precision DA conversion. The analog signal is switched out through the analog switch adg1606 with 16 bit channel, The output signal self check collects the analog signal in a round-loop switching mode at the switching rate of 175ns per channel through the analog switch, which ensures the high-quality acquisition of the output signal. After the design demonstration and test, the output voltage precision meets the design requirements ± 0.15%, the recovery voltage signal meets the design requirements ± 0.1%, which indicates that the output of analog signal is stable and reliable.

Abstract:Objective: Jogging is a favorite sport modality for urban population. A Monitoring and evaluating system is proposed as it is challenging to monitor the accident in the jogging and evaluate the jogging performance. Methods: The adaptive Right-leg Driven circuit is proposed to overcome the unstable input impedance in hardware, the dual median filter approach is utilized to figure out the ECG motive artifact in software. Subsequently, the accident monitoring and autonomic nervous function evaluation are accomplished based on instantaneous heart rate and heart rate variability (HRV). Results: The system proposed in this paper obtain appropriate ECG. Based on the instantaneous heart rate and HRV analysis of 20 years old and 40 years old volunteers, the autonomic nervous function of 40 years old volunteer is relatively weaker. Conclusions: The experimental results demonstrate our wearable ECG system could monitor and evaluate the jogging effectively.

Abstract:In order to solve the problem that it is impossible to evaluate the results of video flow measurement in the actual hydrological measurement, a specific test device is needed to test the accuracy of various flow measurement methods. This paper designs a non-contact flow velocity measurement result inspection device for open scene by using programmable logic controller (PLC). Through this device, the measurement results of various video flow measurement methods are inspected, thus providing a new calibration method for video measurement. Firstly, this paper designs a non-contact flow velocity measurement device for open scenes; secondly, high frame cameras are used to capture experimental videos with different frame rates; finally, three mainstream video flow measurement algorithms are used to measure experimental video flow velocity, and the measurement results are compared with real data, the results show that the error of the three algorithms is more than 20%. The experimental results show that the mainstream video current measurement methods have great errors in low velocity.

Abstract:Aiming at the frequent faults of electric valves in nuclear power plants and the various types of faults, this paper analyzes the typical failure causes and fault information manifestations of electric valves by analyzing the structure and operation principle of electric valves, and studies the signal detection methods suitable for electric valve fault detection. Combining the data analysis and processing of various types of signals acquired, the electric valve signal processing and characteristic quantity analysis methods are proposed. By extracting the electric valve fault characteristic quantity, the fault state of the valve is further characterized. Combined with the actual valve operating state test, it is shown that the extracted characteristic quantities can reflect the true fault of the electric valve, which lays the technical foundation for the fault diagnosis and failure trend analysis of the electric valve.