Abstract:Aiming at the problem of tough extraction about tiny wear particle in the signal output on the inductive debris detection sensor caused by the interference of noise, a new method of noise reduction and extraction of characteristic information to the oil debris signal based on variance stability is proposed in this paper. Making use of the discrepancy in the pre-processed signal about various components, the variance stability has been measured in the first stage. Then, the adaptive threshold is extracted according to the statistical features of normalized variance stability, and the pre-processed signal is segmented by the threshold on this basis. Finally utilizing the characteristic identification index of target signal to further realize the recognition and counting of all debris induced voltage signals. Experiment show that the proposed algorithm can successfully extract the induced voltage signal generated by the tiny debris with the equivalent diameter of 50 μm. Compared with the traditional noise reduction algorithm based on the decomposition principle, the new method proposed in this paper can effectively eliminate the background noise in the detection signal, and its advantages of protecting the morphological characteristics of debris signal can ulteriorly improve the detection ability of small wear particle in the intense interference environment.

Abstract:A new type of flow sensor integrated with uniform velocity tube and vortex is proposed in this paper, which can simultaneously measure the dual-parameter measurement of the differential pressure signal and frequency signal. According to the different extraction methods of the vortex signal, they are divided into piezoelectric type and differential pressure type. The structure, theory, signal acquisition methods of the two flow sensors are mainly designed and expounded in this paper, and the measurement characteristics of them are compared through gas and water. The experimental results show that the combination of the differential pressure flow sensor and the empirical wavelet transform (EWT) can widen the range ratio from 10:1 to 140:1. The absolute value of the maximum measurement error of single-phase gas is reduced from 1. 46% to 0. 49%, and the repeatability is reduced from 0. 17% to 0. 05%. The absolute value of the maximum measurement error of water is also reduced from 1. 00% to 0. 49%, and the repeatability is lower than 0. 07%. Compared with the piezoelectric flow sensor, the differential pressure uniform velocity tube vortex flow sensor has certain application advantages in measurement accuracy and measurement range, and provides a new idea for wet steam measurement.

Abstract:In order to solve the problems of missing alarms, false alarms, and complex data processing due to inaccurate original data in the existing big data calculation model of abnormal electricity consumption, an online monitoring sensor for distributed installation of line nodes on the primary side was developed. The reliability of data acquisition is ensured through the design of the hardware; the mathematical model of the sensor sampling error is established, the mechanism of the error formation after the sensor core is introduced into the air gap is analyzed, and the simulation optimization is carried out. The fluctuating index and coefficient of variation of the collected current signal are introduced, the centroid of the data sample is selected, and the outliers are screened according to the centroid and outlier algorithm, which greatly reduces the complexity of data calculation and improves the reliability of electricity abnormality discrimination. Functional tests show that the current data collected by the sensor is highly accurate, the relative error is less than 0. 2%, and the data synchronization error is less than 5 μs, providing reliable collection data for the calculation model.

Abstract:To improve the Brillouin optical time domain analysis (BOTDA) system performance limited by signal noise, we proposed a noise reduction method based on complete ensemble empirical mode decomposition with adaptive noise combined wavelet threshold (CEEMDAN-WT). First, the noise reduction principle of CEEMDAN-WT algorithm is analyzed. Besides, the BOTDA signal demodulation model is set up based on the principle of microwave sweep frequency measurement. The simulation results show that CEEMDAN-WT method can remove the random noise in the signal effectively. Then, the BOTDA system is set up for temperature measurement. The CEEMDAN-WT method is adopted to suppress the noise of each measured power curve at different sensing distances and different spatial resolutions. Finally, the experimental results show that the signal-to-noise ratio ( SNR) at the end of the optical fiber is improved by 8. 89 dB utilizing the CEEMDAN-WT noise reduction method, under the sensing distance of 30 km and the spatial resolution of 2 m. Research show that the CEEMDAN-WT method provides an effective scheme for improving the SNR of BOTDA.

Abstract:As a single infrared sensor cannot obtain other information beyond the distance of obstacles, this paper designs an infrared sensor array detection system, which not only detects the distance information of moving targets, but also identifies the moving direction of the target. When the moving target passes through the infrared array, the system detects that the infrared reflected echo signal is converted into the target distance information, uses K-means clustering to separate the moving target signal from the background noise, and then extracts the target motion direction features through least square fitting. Finally, eight different motion directions are effectively perceived and classified through the classifier. The experimental results show that the method can successfully identify eight moving directions of the target. When the classifier parameter K is 5, the average recognition accuracy is more than 0. 83. The experimental results verify the feasibility of the proposed method based on infrared sensor array to accurately detect and identify the moving direction of the target, which lays a foundation for realizing dynamic obstacle avoidance of multimodal fusion robot.

Abstract:Aiming at the problem that the existing stage division strategy does not consider the dynamic and multi-stage characteristics of batch processes at the same time, resulting in poor process detection effect, an improved multi-phase kernel principal component analysis based on combined value function (CVF-MKPCA) algorithm is proposed. Firstly, the three-dimensional data of the batch process are expanded in the corresponding direction, and the dynamic characteristics between the data of the batch process are extracted by constructing the expansion matrix. Secondly, an improved combined value function is constructed to evaluate the structural similarity between different time series information; then, according to the evaluation requirements of dynamic structural similarity, the bottom-up search method is used for stage division, and MKPCA method is used for stage modeling. Finally, a new combine statistic is constructed to detect faults in each stage. In the simulation process of penicillin fermentation, the false alarm rate of the proposed algorithm is 3. 40% when the control limit is 95%, and 7. 98% when the control limit is 99%, compared with the comparison method, the false alarm rate is reduced by 2. 12% and 1. 26% respectively, which proves that the proposed method has better fault detection performance.

Abstract:Aiming at the problem that the feature extraction performance of variational modal decomposition (VMD) is affected by its parameters and the poor real-time performance of fault state tracking, an early warning approach and adaptive VMD method are proposed and applied to mechanical part fault detection. Firstly, the degradation characteristics of the full-life vibration signal of mechanical parts are extracted, and then the state warning line is constructed based on the 2σ criterion. Through the early warning line, the degradation state of mechanical parts can be tracked and the fault early warning points can be detected. Then, the energy entropy and mutual information are introduced to construct the fitness function, and an adaptive VMD model is constructed by grasshopper optimization algorithm (GOA) to detect the fault state of mechanical parts near the early warning point. The results show that the proposed state early warning line can detect the fault early warning points timelier and more effectively, and the adaptive VMD can detect the faults of mechanical parts more accurately, which have good application value.

Abstract:Aiming at the problem of difficulty in improving the classification accuracy of industrial fault detection caused by its limited number of labeled training samples which would consume a significant amount of manpower, which a large number of unlabeled samples containing rich information are not fully utilized, this paper puts forward a semi-supervised classification model of combining active learning (AL) and the optimum-path forest (OPF). Firstly, the high-value samples are selected in sorting batch mode according to the value of samples that are comprehensively measured based on BvSB and cosine similarity criterion, and the value of each sample is obtained to expand the initial labeled sample set. Secondly, semi-supervised label propagation is achieved by constructing the optimumpath forest. Finally, the experimental verification was carried out using laboratory collected pipe condition datasets. The experimental results show that the method can achieve an overall recognition accuracy of 96. 68% when the number of labeled samples is 10%. Compared with active learning methods in one-by-one sampling mode and semi-supervised methods that extract global structural information of training samples based on distance metrics, the proposed method has higher Recall value and F1-score value.

Abstract:Aiming at the problem that the unbalance of transformer fault data weakens the ability of fault classification, a transformer fault diagnosis method based on hybrid sampling and improved honey badger algorithm ( IHBA) and optimized support vector machine (SVM) is proposed. Firstly, K-nearest neighbor denoising, K-means and SMOTE are used for hybrid sampling of data to alleviate the shift of diagnosis results to the majority class. Then, the traditional honey badger algorithm (HBA) is improved by using tent mapping, roulette random search mechanism and optimal individual perturbation strategy, and the SVM parameters are optimized by IHBA to further improve the transformer fault identification ability. Finally, the simulation results of the proposed method show that, compared with the traditional transformer fault identification method, the fault diagnosis model combining K-Nearest Neighbor denoising, K-means, SMOTE hybrid sampling and IHBA-SVM obtains the highest macro F1 and micro F1 values, reaching 0. 877 and 0. 886 respectively, which indicates that the proposed model not only has higher overall classification ability, but also can better identify minority faults.

Abstract:Aiming at the difficulty of fault feature extraction and fault identification of distribution transformers, a fault diagnosis method combining vibration signals, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and graph convolution neural networks (GCN) was proposed. Firstly, the vibration signal from the acceleration sensor is processed by CEEMDAN to obtain a set of intrinsic modal functions. Secondly, its marginal spectrum information is taken as the feature vector. Then, an undirected weighted complete graph is constructed for the eigenvector matrix, and an improved gray wolf optimization algorithm is used to optimize the Gaussian kernel bandwidth. Finally, an improved GCN model with multi-channel and multi-connectivity is built for feature secondary mining and fault classification. At the same time, an index called peak factor is added to the model to realize the identification of unknown faults. In the case analysis, the fault simulation of oil-immersed transformer and dry transformer is carried out respectively, and samples of different states are extracted for testing. The experimental results show that the accuracy of the proposed method for oilimmersed transformer and dry transformer fault identification is 97. 73% and 95. 6%, respectively, which is better than the other two comparison methods. In the face of unknown types of faults and operating conditions change, it also has a high ability to identify.

Abstract:Aiming at the fault problem of wind turbine gearbox overtemperature, a fault early warning model of wind turbine gearbox based on improved parameter optimization machine learning algorithm is proposed. Firstly, the characteristic variables are determined by random out-of-the-pocket estimation, and the input variables are filtered by sliding smoothing filtering. Secondly, the gray wolf algorithm optimization support vector regression model is constructed, and the state identification index is determined according to the residual value of the optimal model output. Finally, the threshold range is set by the time-lapse sliding window, and the alarm is immediately issued when the status identification indicator exceeds the threshold range. Experimental results show that the model can issue a fault warning for the temperature abnormality of the wind turbine gearbox 87 minutes in advance, and the early warning effect is better than that of the distance correlation coefficient-GWO-SVR model, Pearson-GWO-SVR model and OOB-SVR model.

Abstract:In order to detect pipeline spanning and adapt to the fast-moving state of in-pipeline detectors in the field and the low power consumption requirements of stress sensor arrays, this paper proposes a method of measuring the pipe wall stress based on magnetoelastic under weak alternating current. Firstly, stress characteristics of spanning pipeline are studied via finite element simulation, and the stress sensor deployment scheme is determined. Secondly, steel stress measurement model of electrical impedance method is established, and stress sensor is designed and fabricated. Finally, experimental verification is carried out. The results show that the axial stress at the top and bottom of the pipeline can sensitively characterize the spanning pipeline state; when the elastic stress varies from 0~ 50 MPa,the sensor coil impedance and the stress have a negative linear relationship, and the linear fitting goodness of the sensitivity coefficient is greater than 0. 92; the sensitivity increases as the excitation frequency increases, and when the excitation frequency is 30 kHz, the average sensitivity coefficient is -0. 018 Ω/ MPa; the power consumption of a single sensor is less than 0. 1 mW

Abstract:In view of the problem that the amplitude of pulse voltage changes due to the migration of VHF pulse resonance frequency with the change of circuit impedance, and the bipolar characteristics of VHF pulse make the traditional detection method unsuitable, a VHF EDM discharge state detection method based on inter-electrode partial voltage detection was proposed. The conversion of high frequency and high voltage bipolar pulse to low voltage and single polarity DC signal is realized by the combined action of voltage division, rectifier and filter modules. Meanwhile, the current open-circuit voltage and the change of inter-electrode discharge state can be directly characterized by the detection of DC signal. In the simulation and experimental validation testing method on the basis of feasibility, design discharge condition detection system based on the test scheme and corresponding detection algorithms, successfully verified the micro-nano machining the pulse frequency, VHF automatic optimization technology and the effectiveness of the very high frequency discharge condition detection, is conducive to improve VHF discharge during the process of electrical discharge machining precision.

Abstract:Aiming at the problems of particle swarm optimization in mobile robot path planning, such as precocity and local optimum, a path planning method based on region search-adaptive particle swarm optimization algorithm (RS-APSO) is proposed. Firstly, the region search algorithm is used to preprocess the original map to reduce invalid information in the map. Secondly, two variable operators are proposed to adjust the inertia weight factor and improve the acceleration factor adaptively to enhance the search ability of the algorithm in different periods. The new acceleration factor is used to remove the bad region quickly from the particle. Finally, the robot can safely avoid moving obstacles through dynamic obstacle avoidance strategy. Simulation results show that compared with PSO algorithm, the average running time of RS-APSO algorithm is reduced by 30. 3%, the average number of iterations is reduced by 43. 9%, and can also generate safe path in dynamic environment.

Abstract:Time-frequency analysis is a powerful tool to deal with non-stationary signals. As one of the traditional time-frequency analysis methods, s-transform can change the scale of its window function with frequency. However, the scale variation of time-frequency window function is fixed, which cannot be applied to the local characteristics of different signals, resulting in poor energy aggregation. In this paper, an adaptive generalized S-transform algorithm is proposed, and a generalized Gaussian window function controlled by four adjusting parameters is designed. The adjustment parameters are optimized by the adaptive concentration measurement to seek the best time-frequency characterization effect. According to the results of time-frequency analysis, instantaneous frequency recombination and component reconstruction are used to obtain the instantaneous frequency of each component, and at the same time, smooth processing is carried out to achieve the parameter estimation of multi-component signals. Simulation results show that the proposed adaptive generalized S-transform algorithm, combined with instantaneous frequency recombination and component reconstruction signal method, greatly improves the time-frequency resolution of multi-component signals and the accuracy of signal separation.

Abstract:In order to improve the motion safety and interactivity of mobile robots in an indoor human-robot integration environment, a dynamic obstacle avoidance method for indoor mobile robots integrating pedestrian motion information is proposed, considering both task constraints and social rules. First, the YOLO v3 algorithm and the Deep Sort algorithm are used to detect and track pedestrians in indoor environments in real time, respectively, and calculate the historical trajectories of pedestrians in the past. Then, the Social-GAN algorithm is used to build a pedestrian interaction model to achieve trajectory prediction. On this basis, the motion state of pedestrians is integrated into the robot obstacle avoidance algorithm, the evaluation function is designed according to social rules, and the sampling speed samples of the robot are evaluated, so that the robot can bypass pedestrians in a safe and comfortable way and ensure social acceptability of mobile robots in a human-robot integration environment. Through experimental comparison and analysis, compared with the traditional DWA method, the method in this paper not only improves the efficiency of robot navigation and obstacle avoidance, but also increases the navigation obstacle avoidance time from 23. 56 s to 19. 38 s in the same indoor scene, and can effectively reduce the collision with pedestrians risk, and ensure the safety and sociality of robot navigation.

Abstract:The standard UKF algorithm of GNSS / SINS integrated navigation system lacks the ability to adjust the measurement noise variance and system status anomaly adaptively, which affects the filtering accuracy of the integrated navigation system. A robust adaptive UKF algorithm is proposed in order to solve the above problem. Firstly, this algorithm introduces the variational Bayesian estimation principle to estimate the measurement noise variance in real time. Then, an adaptive factor is constructed to reduce the influence of abnormal system state on the navigation solution, based on the predicted residual of the filter. Finally, this algorithm is applied to GNSS / SINS integrated navigation system. The simulation results show that, compared with the standard UKF algorithm and the robust UKF algorithm, the proposed algorithm can improve the position accuracy by 51. 2% and 9. 3% respectively in the whole simulation period when the statistical characteristics of the measurement noise change, and can reduce the influence of abnormal system model disturbance and filter initial value deviation on the navigation solution. The experimental results show that the proposed algorithm has strong adaptability and robustness, and can improve the accuracy of integrated navigation system in complex environment.

Abstract:In order to suppress the strong coupling between extremely close-range multiple-input multiple-output ( MIMO) antenna elements, a decoupling method is proposed to arrange MIMO antenna elements feeding position in an extreme adjacent distance where belongs to their weak electromagnetic field region. The weak electromagnetic field region is mainly generated by changing the feeding structure of the microstrip antenna. Firstly, a two-element MIMO antenna with an edge-to-edge distance of 1 mm (0. 011λ0 ,λ0 is the free-space wavelength at the center frequency of 3. 5 GHz) is proposed. The simulation and measurement results show that the proposed decoupling method can improve the isolation of the MIMO antenna by 42 dB in the operating frequency band. The isolation performance of the proposed antenna is much more obvious than that of the reference antenna. Furthermore, a three-element MIMO antenna is constructed by arranging the improved antenna elements along 120°. The experimental results show that the coupling suppression is greater than 30 dB in the working frequency band of 3. 45 ~ 3. 55 GHz; the maximum isolation is 52 dB at the center frequency of 3. 5 GHz. Moreover, the reflection coefficient of each antenna is not affected by the proposed decoupling structure. It is worth noting that the proposed decoupling method has a good application prospect as it does not need additional decoupling circuit and components compared to the traditional approach with respect to mutual coupling reduction.

Abstract:In order to solve the problem of low detection degree and high missed detection rate of occluded pedestrians in pedestrian detection, an attention mechanism based UAST-RCNN network is proposed, which is improved on the basis of Faster-RCNN network. Firstly, Swin-Transformer is selected as the backbone network to improve the global receptive field by using a window multi-head selfattention mechanism. Then, the feature pyramid is improved for the quality of feature samples through the hierarchical resampling module, and a progressive focus loss function is introduced to balance positive and negative samples. Finally, in the preprocessing stage of the experiment, the improved data preprocessing was used to extend the City Persons dataset for multi-scale training. The experimental results show that the algorithm has a significant improvement in the detection of occluded pedestrians compared with the original model, in which the AP is increased by 6. 3%, and the MR (miss rate) is decreased by 4. 1%. The feasibility of the proposed algorithm in pedestrian detection is verified, and it can meet the detection requirements of occluded pedestrian scene.

Abstract:The lightning monitoring and location are of great significance to reduce the lightning disaster, and the location accuracy is the key index of the lightning location system. Aiming at the problem that the positioning accuracy of traditional positioning methods is greatly affected by gross error of data and station network layout, a positioning method based on time difference and clustering algorithm is proposed. The lightning positioning data is combined based on the principle of time difference of arrival, K-means clustering algorithm is used to classify the combined positioning values and remove outliers to obtain the final accurate positioning value. Taking the lightning detection network in Jiangsu Province as an example, the numerical distribution of location error, the time measurement accuracy and the influence of station network layout on the location error are quantitatively studied. The results show that compared with the traditional arrival time difference method, the regional positioning error calculated by this method is reduced from 5 km to 2 km; The root mean square error of four station positioning is reduced from 0. 89 km to 0. 07 km; The location error is slowly increased due to the sharp increase of the time measurement error, and the impact of website layout on the location error is significantly reduced.

Abstract:Obstacle recognition and path planning are the necessary means for robot to move autonomously. Based on depth camera, this paper proposes an obstacle recognition method based on the fusion of depth continuity and color feature points. The spatial location information of objects is obtained by depth camera and mapped to the existing map to construct the obstacle space. A path planning method of PRM-D∗ is proposed. Firstly, the improved random probability roadmap ( PRM) is used to complete the overall path planning. Then, according to the obstacles identified by the camera, the local map is set up, and the D∗ algorithm based on graph search is used to carry out local dynamic planning to complete the dynamic obstacle avoidance task. Through the experiment, the detection accuracy of the proposed obstacle recognition method is greater than 80% even in dim indoor environment, and the accuracy of conventional environmental detection is higher than 95%, and it has good robustness and real-time performance; The path planning method of PRM-D∗ not only shortens the overall planning time, but also ensures the success rate of path planning. The single dynamic planning time is less than 0. 02 s, and has good dynamic obstacle avoidance performance.

Abstract:In order to improve the defect of UWB-MIMO antenna, such as large size, insufficient port isolation and weak antiinterference ability, a miniaturized UWB-MIMO slot antenna with notch characteristics fed by coplanar waveguide (CPW) is proposed, which cutting the square radiation patch to broaden the bandwidth, by connecting the two antenna unit coplanar ground and adding fence shaped decoupling branches between the antenna units, the isolation of the antenna is improved. The antenna size is only 30 mm× 56 mm×0. 8 mm. In addition, a C-slot open resonator ring ( SSR) structure is etched on the square radiation patch to filter the interference of WLAN band signals (5. 15~ 5. 85 GHz) to the UWB system. The measured results of the antenna show that the working frequency band of the antenna is 2. 73~ 10. 71 GHz, the port isolation is less than-20 dB, and the envelope correlation coefficient ECC is less than 0. 01. It shows that the antenna is suitable for UWB-MIMO communication system.

Abstract:In view of the frequent leakage of oil pipelines in key areas such as power plant oil depots and chemical water workshops, a pipeline leak detection method in key areas of power plants based on improved YOLO v5 is proposed. The improved YOLO v5 detection algorithm first incorporates CBAM module to strengthen the learning of regional features of pipeline oil leakage images. The CBAM makes the model more focused on the extraction of pipeline leakage features, and weakens the influence of complex backgrounds on detection results. Secondly, a bidirectional feature pyramid network is used for multi-scale feature fusion. It also reduces redundant calculation, and improves the detection ability of the algorithm for small targets. Finally, Focal EIoU Loss is used as the loss function to make the regression process more focused on high-quality anchor boxes. It improves the regression accuracy, speeds up the convergence speed, and improve the robustness of the model. The experimental results show that the improved algorithm performs well in real samples, with an average accuracy rate of 79. 6%, which is 38. 4% higher than the original YOLO v5s algorithm. The false positive rate and the false negative rate in the complex background of the power plant are significantly reduced. It shows that the improved YOLO v5 detection algorithm can be effectively applied in the actual production environment.

Abstract:Aiming at the fast and accurate measurement of rail wear based on 3D structured light scanning, this paper proposes a point cloud registration algorithm based on 4PCS (4-points congruent sets) and SICP (sparse iterative closest point), which is used to quickly and accurately register the standard rail point cloud and incomplete worn rail point clouds with noise. Since the wear rail data obtained by one-time scanning of the three-dimensional structured light scanner is usually incomplete and contains noise, 4PCS with good robustness for low overlap point cloud registration is firstly used to coarse registration of the rail point cloud, which provides a good initial transformation matrix for accurate registration. Then, the SICP with good robustness for noisy point cloud registration is used for accurate registration. Finally, the rail head wear is calculated according to the accurate registration results. It quantitatively analyzes the influence of different levels of down-sampling on registration accuracy, time and calculation accuracy of rail head wear, which demonstrates the advantage of 4PCS and SICP in fast and accurate registration of incomplete and noisy rail point clouds. It is concluded that different levels of down-sampling have no influence on the calculation accuracy of rail head wear. Meanwhile, the robustness of SICP in accurate registration of worn rail point cloud with noise is verified by quantitative comparison analysis of the point cloud registrations with different levels of noise.

Abstract:To detect structural damage degree and damage axial direction, reduce detection range and achieve accurate damage location with high efficiency and low cost, a transmission tower structural damage detection method based on MIMU was proposed. Firstly, based on the Angle of transmission tower structure, the MIMU micro-inertial measurement system is used to monitor the structure change of transmission tower, and the translation and rotation entropy matrices are constructed combined with the structural entropy. Then, the damage degree and damage axial direction of transmission tower structure are detected by using the damage index of entropy distance and variation entropy. Finally, simulation and in-service transmission towers under various working conditions are designed to verify the rationality of the algorithm. The results show that the error between the entropy distance index and the theoretical value is less than 3%, which can effectively detect the damage degree of the structure. The variation entropy index of damaged axial direction is 20% ~ 53. 9% higher than that of undamaged axial direction, which can effectively detect the damaged axial direction of structures. This method can provide a practical basis for truss structure damage detection.

Abstract:For the problem of offset detection of dies on matrix test trays, a vision-based die offset detection system is built, which mainly includes loading and off-loading channels, a 3D moving platform, and a vision detection module. This system takes a standard sample as the benchmark and builds an image correction algorithm combining template matching and affine transformation to achieve the generality between different inspections. A rectangular measurement algorithm for a specific area is designed, based on the grey scale gradient and the direction of the gradient; the system calculates the die offset using center coordinates of the line which connecting the top left and bottom right points of the die as the reference point. The experimental results show as follows: The offset detection error range of this system is -2. 145 to 4. 257 μm, the average time of algorithm execution is 72. 56 ms, and the operation speed is 20 mm/ s, and the average detection time for chips on matrix test trays with 5 × 12 is 64. 5 s/ tray, which can meet the demand of the actual processing process; the accuracy of this system is tested using the bias and linearity analysis method in measurement system analysis (MSA), the linearity of bias is significant. To conclude, the bias and linearity of the system meet the requirements.

Abstract:Automatic and accurate location of insulator components in catenary images is the basis of detecting insulator fault. The insulators in the catenary images are incline-angled, so it is difficult to locate the object accurately by using horizontal box. To solve this problem, an insulator accurate location approach was proposed based on improved RetinaNet. To begin with, the efficient Ghost module was adopted to replace the convolution operation in the original feature extraction network to obtain multi-scale feature maps and reduce the computational burden of the model. Next, in order to suppress the influence of secondary features on object detection, the attention mechanism was embedded in the network. Then, the rotating box was introduced as the prediction box of the model to realize the accurate location of insulators and reduce the interference of redundant background noise. Finally, the positive and negative samples were redefined in the training process. By doing so, the problem of learning wrong samples that caused by adding rotating box was resolved. Experimental results demonstrate that the proposed approach featuring good detection performance can locate the insulator accurately and prevent redundant background information. Compared with original algorithm, the detection accuracy increases by 2. 8%, the detection speed reaches to 25. 6 FPS, and the number of network parameters reduces by 42. 8%.

Abstract:In recent years, with the full use of 5G base stations, higher performance of the S-band power amplifier has been required. In order to meet the needs of high gain, high efficiency, wide bandwidth, large power, miniaturization and so on, this paper designs a radio frequency power amplifier worked at 3. 4~ 3. 6 GHz with the parallel open circuit microstrip method based on the step matching circuit. The design method proposed in this paper aims to solve the difficulty of accurate measurement to the dynamic impedance of transistor and the non-negligible difference between theory simulation and practical testing. In this design, the input and output matching network always remain in a dynamic matching condition by utilizing the parallel open circuit microstrip with the adjustable physical length and width. This kind design is beneficial for the later debugging circuit to reach the best matching state. In order to verify the feasibility of the design scheme, this paper designs a radio frequency power amplifier operating at 3. 5 GHz with CREE CGHV40030 and completes circuit implementation and test. The simulation and test results demonstrate that it’ s easy to debug circuit, determine the optimum matching network, realize the test results very closed to the simulation results and obtain good performance by utilizing the proposed design method. The AM-AM distortion is less than 1, the AM-PM distortion is less than 5°/ dB, S11 is less than -5. 1 dB, S12 is larger than 18. 3 dB, the output power is larger than 45 dB, the gain is larger than 12 dBm, and the drain efficiency is larger than 66%.