Abstract:With the increasing demand for indoor positioning, ultra wide band (UWB) technology with good communication functions and positioning performance plays an important role in the field of indoor positioning. In view of the problem that UWB communication signals are susceptible to interference in indoor complex environment, resulting in positioning errors, this paper establishes the K-means algorithm to conduct cluster analysis of the collected data, eliminates the wrong ranging values generated when there is signal interference, and improves on the basis of the classic Chan algorithm to create a Chan-IDW model to determine the actual coordinates of the target location, and then measures the accuracy of the positioning model by mean squared error (RMSE). The experimental results calculated that the average error of the two-dimensional coordinates of the target located under signal interference was 5.67 cm, and the average error of the three-dimensional coordinates was 11.34 cm, and the error was in the centimeter level, indicating that the coordinates of the target solved by the model were very close to the real coordinates. Therefore, it is concluded that the Chan-IDW model can effectively solve the problem of accurate UWB positioning under indoor signal interference.

Abstract:Torque ripple is an important factor affecting the torque characteristics of IPM oil submersible motor. The purpose of reducing torque ripple is achieved through the design of the magnetic flux barrier of the motor rotor. Taking the 10 pole 12 slot IPM "一" magnet type oil submersible motor as the basic model. The two-dimensional transient field of oil submersible motor rotor is analyzed by finite element method. Based on the influence of flux barrier effect on the radial flux density of air gap, the distribution of radial electromagnetic characteristics of oil submersible motor rotor under various operating conditions is obtained, and the torque characteristics such as cogging torque and electromagnetic torque are compared and studied respectively. After optimization, the torque ripple of oil submersible motor is significantly reduced. Finally, the results verify that the design of magnetic flux barrier has a good inhibitory effect on the cogging torque of oil submersible motor. Because the stator structure of submersible motor is limited, the optimal design of flux barrier is great significance.

Abstract:For the requirements of contact force in the process of industrial robots such as polishing, a fuzzy PID constant force control method based on particle swarm (PSO) optimization is proposed. Firstly, the force and gas flow model of the flexible force-controlled flange are analyzed to establish the system model of the flexible force-controlled flange device; secondly, the fuzzy PID controller based on PSO algorithm is designed so that the control parameters of fuzzy PID can be adaptively adjusted, and the fuzzy PID control optimized based on PSO algorithm is compared with ordinary PID, fuzzy PID and PSO algorithm through MATLAB simulation; finally, the experimental verification of the constant force control output performance of the flexible force-controlled flange is carried out by building a LabVIEW-based grinding experiment platform. The simulation experiment results show that: Compared with the traditional PID and fuzzy PID control methods, the fuzzy PID control optimized based on PSO algorithm has no overshoot, the system response is faster, and the system reaches stability at 0.43 s; the actual contact force output error of the flexible force control flange is less than 0.85 N when the constant force grinding experiment is conducted; the roughness of each area of the shell surface after grinding is stable between Ra0.1~Ra0.2. The method can effectively suppress the contact pressure fluctuation and has stronger robust performance.

Abstract:In order to solve the problems that the parameters of traditional analog PI controller are not easy to adjust and the parameters can not be adjusted online when the controlled object changes, memristor is introduced to realize the online adjustment of controller parameters. At the same time, aiming at the problem that the resistance value of the memristor is not easy to determine, a fuzzy algorithm is introduced, and a fuzzy memristor PI controller is combined with the fuzzy algorithm to design a fuzzy memristor PI control system, which realizes the adaptive adjustment of the controller parameters. Based up on the oretical derivation，not only the Multisim simulation was carried out，but also the circuit experiment was conducted. The results show that compared with traditional analog PI controller, the parameters of the adaptive memristor PI control system based on fuzzy algorithm can be adjusted online, and has strong adaptive ability. Moreover, the adaptive memristor PI control system based on fuzzy algorithm can reduce the overshoot of the system by 66.27%, and has better tracking performance.

Abstract:Using COMSOL, the mechanical model of ceramic products detection by coin-tap sound is constructed to predict the maximum impact force in the impact dynamic quantity. The ceramic structure is simplified as a cube ceramic test piece, and its stress signal is analyzed in time domain. The difference of the impact signal waveform of the cointap sound between qualified ceramics and defective ceramics, as well as the response difference of the size of the defect to the percussion impact force, are studied. The results show that the peak value of impact force signal of qualified ceramics is higher than that of defective ceramics. The width and length of the defect affect the amplitude of the impact force signal. The wider the defect width or the longer the defect length, the lower the peak value of the impact force signal. When the defect depth is small, the influence on the peak value of impact force signal is not obvious. The research results reveal the difference in the response characteristics of qualified ceramics and defective ceramics to impact force, which can provide an effective theoretical support for the accurate identification of defective ceramics in the online detection process of automatic production line.

Abstract:This paper proposes a broadband rectifier with wide dynamic input power range, which is mainly used in low input power condition. The proposed rectifier uses unidirectional conductivity of Schottky diodes and a parallel-type circuit topology for rectification. In the proposed structure, a broadband impedance matching network consisting of a twostage microstrip lines and a matching inductance are adopted to reduce the mismatch loss and achieve high radio frequency (RF) to DC power conversion efficiency (PCE) over the operating range, where the matching inductance is to offset the capacitance of the diode. The proposed rectifier is designed, optimized, simulated, fabricated and measured. The measured results agree well with the simulations. With an input power range of -1~10 dBm and a bandwidth of 1.8~2.8 GHz (fractional bandwidth of 43.4%), the PCE is greater than 50%. Besides, a peak PCE of 76.4% is obtained at an input power of 10 dBm. Comparing with previous similar work, this proposed rectifier operates at the lowest power level, making it amazing in the development of RF-harvesting.

Abstract:For the radio frequency micro-electro-mechanical system (RF MEMS) switch life test, the cost is high, the test connection is complex, and with the size of the RF MEMS switch shrinking, the traditional test efficiency is low, and the test task can only be carried out in the laboratory. In this paper, a RF micro-electro-mechanical switch test system with functions of signal generation, real-time waveform monitoring, lifetime calculation, data recording and test report printing is set up, and a RF micro-electro-mechanical switch is tested and evaluated preliminarily with this system. The results show that the design has well completed the test of the switch voltage, switch time, cold life, and thermal life of the RF micro-electro-mechanical system switch, and can meet the test requirements of the RF micro-electro-mechanical system switch, which proves the practicability of the test system and the accuracy of the test work.

Abstract:The traditional ant colony algorithm (ACA) is difficult to overcome the problems of suboptimal path and slow convergence in path planning. To solve these problems, a jump point optimization ant colony algorithm (JPOACA) is proposed. By introducing the value function of jump point search (JPS) algorithm, low-cost neighborhood nodes are selected, and then the multi neighborhood of ACA is used to expand the neighborhood of JPS algorithm, expand the vision of JPOACA, increase the number of lowcost neighborhoods, design angle heuristic information function and step size heuristic information function in the low-cost JPS algorithm neighborhood, improve the path optimization ability of the algorithm, and finally supplement pheromones at the jump points, In order to improve the convergence speed of the fusion algorithm, a pheromone supplement method is added to the hops of the optimal path. The simulation results show that the path planned by JPOACA is smooth and better, and the convergence speed and adaptability to complex terrain are significantly improved.

Abstract:Aiming at the problem that the gray wolf optimization algorithm is easy to fall into local optimum and low efficiency in the path planning of mobile robots, a genetic simulated annealing gray wolf optimization algorithm was proposed. An adjustable nonlinear convergence factor is used for the early search and the late search of the balance algorithm. At the same time, the adaptive genetic hybridization strategy was used to hybridize the gray wolf population with a certain probability to produce new individuals, so as to effectively enhance the diversity of the gray wolf population. The candidate wolf is accepted by simulated annealing operation at the later stage of iteration to avoid the algorithm falling into local optimal solution. The path length and path smoothness are taken as the fitness evaluation indexes and the evaluation function is established to evaluate the effect of path planning. Finally, the experimental results of path planning show that the fitness of the improved algorithm in this paper is optimized by 2.10, 3.15 and 3.94 respectively compared with the gray wolf optimization algorithm on three maps of different sizes, and the path planning effect is significantly better than other related algorithms.

Abstract:With the access of a large number of distributed photovoltaics, the operation status of the distribution network has become more complex. In order to accurately evaluate the real-time health status of distribution system, a real-time health status assessment method of distribution network is proposed, which takes into account the importance of network nodes and lines. First, consider the optimal power flow, load importance and load loss of a node in the photovoltaic access network to improve the LeaderRank algorithm, consider the power loss after a node outage and the LR value of the node to improve the load moment algorithm, use the improved electrical LeaderRank algorithm and line load moment algorithm evaluate the importance of nodes and lines in the distribution network. Then, based on the electrical and non-electrical parameters of the operation of the distribution equipment, the health indices of the distribution nodes and lines are calculated respectively. Comprehensively consider the health and importance of nodes and lines in the distribution network to calculate the health index of the distribution network to judge the health status of the distribution network. Finally, taking a 10 kV distribution network as an example, the health index is 2.266 3, which is in a general defect state.The health of the distributed PV system has been improved by appropriately increasing the PV penetration rate in the distribution network, which further verifies the effectiveness and rationality of the proposed method.

Abstract:Aiming at the problems that traditional DWA algorithm is easy to fall into local optimum and dynamic obstacle avoidance is poor in complex dynamic environment, an improved DWA algorithm based on A* was proposed. Firstly, a collision cone is introduced into the DWA algorithm to detect the static and moving obstacles, and the speed with collision threat is eliminated through the speed window to optimize the constraint space. Secondly, the evaluation function is improved according to the static and moving obstacle information to improve the trajectory evaluation ability of DWA algorithm. Finally, the improved DWA algorithm is integrated with A* algorithm to solve the problem that DWA algorithm is easy to fall into local optimum in complex environment. Simulation results show that compared with other similar algorithms, the proposed algorithm can improve the traveling speed and safe distance by more than 50%, which not only can make the robot travel according to the global optimal path, but also effectively improve the robot′s obstacle avoidance ability in complex dynamic environment.

Abstract:In order to improve the wide-range speed regulation accuracy and dynamic response performance of the permanent magnet synchronous motor system, and solve the problems that it is difficult to adjust the weight coefficient of traditional cost function in predictive speed control, this paper proposes a fast predictive speed control strategy for PMSM system without weight coefficient, and proposes a new regulation range widen method combined with voltage vector optimization. Firstly, the discrete model of PMSM is established, and then the reference voltage vector is predicted. Then, the cost function is improved to the voltage dimension to eliminate the weight coefficient, and the alternative vectors are selected according to the sector position of the reference voltage vector, so as to shorten the calculation times. At the same time, to widen the speed regulation range, the current limit and voltage limit are analyzed from the perspective of α-β coordinate combined with voltage comparison method, and the current and voltage limit are realized supplemented by the correction of the voltage vector adjustment time. Finally, the experimental study proves that the proposed control strategy can meet the dynamic and steady-state requirements of wide-range speed regulation without the weight coefficient.

Abstract:This paper proposes a switched Hammerstein model identification method based on special historical data segment mining. Special data segment refers to data in stable state and stable slope response. First, a random sampling consensus algorithm is used to identify the static nonlinear subsystem based on the steadystate data. Secondly, based on the stable slope response data, the density peak clustering algorithm is used to identify the dynamic subsystem structure and the corresponding operation interval. Finally, the least squares algorithm is used to identify the model parameters of the switched dynamic subsystems based on the data sets in the operation interval. The results of numerical simulation and experimental cases show that, compared with the standard Hammerstein identification method, the proposed method can realize the structure identification and operation interval division of multiple linear dynamic submodules with different switching points, reduce the influence of switching dynamic subsystems on parameter identification when the model structure is unknown, and improve the identification accuracy of switching Hammerstein models.

Abstract:As the core equipment of the urban lighting system, the regular operation of streetlights is of great significance to urban lighting. Currently, streetlight fault detection is limited to preliminary fault phenomenon. To detect the streetlight′s specific fault category, this paper takes the streetlight operation data of the streetlight monitoring and data acquisition system as the object. A streetlight fault detection model based on improved VMD and feature selection is proposed combining the traditional method, the datadriven method and the signal processing method. First, principal component analysis filters the main variable parameters of streetlight operation data, and variational mode decomposition is used to decompose the screened parameters. Similarly, the whale optimization algorithm is introduced to improve the adaptability of the variational mode decomposition. Secondly, the Pearson coefficient selects the relevant IMF components, and the sample entropy is used to construct the fault feature vector. Through the experimental verification of the streetlight fault statistical data of the urban lighting monitoring system in Chongzuo City, Guangxi, the results show that the proposed fault diagnosis method can effectively extract the fault feature information of different fault states of streetlights. The correct rate of fault diagnosis is 93.75%, which provides a new way for streetlight fault diagnosis.

Abstract:The commonly used Power Back-Off method can improve the nonlinear problem of the amplifier to some extent, but its application scope is limited due to its low efficiency. Based on the Power Back-Off, this paper proposes a method combining polynomial fitting to improve the 6~10 dB backoff range constraint of the traditional Power BackOff method. In order to verify the effect of the combined method, the photoelectric amplification test platform based on LabVIEW was built in this paper to carry out signal amplification tests on sine wave, square wave and triangle wave signals at different given frequencies. When the amplifying function of the system is tested, the output signal is analyzed by power cycle map, and the harmonic distortion state can be determined. The experimental results show that the improved method can reduce the power value of each harmonic component by 34.6% compared with the Power Back-Off, which proves the effectiveness of the proposed method to further improve the nonlinear problem of the amplifier.

Abstract:To address the existing deep neural network models for insulator defect detection are large in size, high in computational resource consumption, low in detection accuracy and difficult to deploy at the edge end. In this paper, we propose a lightweight insulator defect detection model ACAM-YOLOv5s with asymmetric convolution and attention mechanism based on channel pruning and YOLOv5s method. The ACAM-YOLOv5s model uses the asymmetric convolution module ACBlock to replace the standard convolution in the residual structure of the YOLOv5s backbone network, combined with the attention CBAM of channel and spatial blending for feature fusion to enhance the expressiveness, feature extraction and robustness of the backbone network. PIoU, which is highly sensitive to the size and position of the bounding box, was introduced as a localisation regression loss to address the problem of low defect detection localisation accuracy due to high insulator aspect ratios. The experimental results show that the pruned ACAM-YOLOv5s model has relative advantages over the original YOLOv5s in terms of detection accuracy, computational volume and model size, which can meet the needs of edge device deployment and has potential value in the field of UAV aerial insulator defect detection.

Abstract:An improved U-Net street image semantic segmentation method based on a dual attention mechanism is proposed to improve the segmentation effect of multi-scale targets and enhance the feature extraction ability.After the fifth convolutional block in the U-Net encoding stage, the feature pyramid attention module is added to extract multi-scale features, fuse contextual information, and enhance the target semantic features.Instead of using the feature stitching method of U-Net in the decoding stage, a joint spatial domain-channel domain attention module is designed to receive the low-level feature maps from the jump connection and the high-level feature maps from the previous attention module.Experimental results on the Cityscapes dataset show that the introduced attention module can effectively improve the street view image segmentation accuracy, and the segmentation performance metric mIoU improves by 2.0~9.6 percentage points compared with methods such as PSPNet and FCN.

Abstract:To address the problems of low contrast, edge detail loss and excessive enhancement in existing low-light image enhancement algorithms, a low-light image enhancement method based on the combination of global double gamma correction and improved SSA algorithm is proposed. In addition, to improve the convergence performance of the algorithm, elite backward learning and Lévy flight strategy are introduced to improve the sparrow algorithm, optimize the selection of parameter (α), and realize the detail enhancement of the image by finding the optimal gamma value. The simulation experimental results show that the algorithm enhances the image with larger peak signal-to-noise ratio and structural similarity index, less image color distortion, and sharpens the edges, and the overall enhancement effect is better than other comparison algorithms, which has better processing effect.

Abstract:With the rapid development of computer 3D vision, point cloud data containing spatial geometric information is widely used in robots, autonomous driving and other scenes. However, due to occlusion, angle limitation and other reasons, geometric information is often missing. In order to solve this problem, SRC-Net is proposed. Firstly, a geometric skeleton is reconstructed from the incomplete point cloud using a designed skeleton reconstruction network that fuses a dynamic graph convolutional network encoder and a folding network decoder, and then the auto-encoder structure is used to establish the mapping from the geometric skeleton to the uniform and complete point cloud. Finally, the completion results on the MVP dataset show that SRC-Net can generate high-quality complete point clouds that are more evenly and smoothly distributed than existing completion networks, and can achieve more detailed completion effects. It provides a new idea and method for point cloud deep learning completion, and has certain guiding significance.

Abstract:The classification and detection of fresh cut rose is of great significance to its sales. At present, the classification and detection of fresh cut rose is mainly manual. In order to reduce the loss of fresh cut rose flowers in the process of manual classification, a set of classification and detection system of fresh cut rose flowers was built based on machine vision method and Halcon software. Firstly, the experimental platform was designed and the classification standard of fresh cut rose flowers was established. Then, image enhancement and data enhancement technology are added to improve the image effect, increase the number of samples, and use the median filter method to eliminate the image noise, so as to ensure the accuracy of classification results. Finally, the training samples are added to five models for training, and the training results of each model are compared. Mobilenet_v2 model is selected to join the image classification system to classify the top view of fresh cut flowers, and a one-dimensional measurement system is established to measure the length of flower stems; Establish the evaluation criteria model to complete the classification of fresh cut rose flowers. After testing, the classification accuracy of the top view classification system is 94%, and the flower stem length measured by one dimension is within the error range.

Abstract:Aiming at the problem of low defect detection accuracy caused by complex steel surface background in industrial production, this paper proposes a steel surface defect detection algorithm based on improved YOLOX. First, the Swin Transformer module is introduced to capture the global context information of the surface area of the defective steel and extract more differentiated features. Secondly, the weighted bidirectional feature pyramid network (BiFPN) is used to facilitate cross-scale feature fusion. Finally, we improved the original target localization loss function and established a CIoU loss function that fuses the center position of the bounding box to achieve high-precision localization of the target frame. Experiments showed that the mAP of our algorithm on the NEU-DET dataset is 80.7%, which is 6.2% higher than the original YOLOX-S network, and it is also significantly higher than some other mainstream algorithms, with high accuracy and practicality.

Abstract:In the outdoor inspection task of substation, the robot is vulnerable to the complex environment of outdoor wind, fog and uneven road surface. It is prone to jitter and angle deviation, resulting in blurred photos, instrument tilt and other problems, and it is difficult to ensure the accuracy of the identification reading of the pointer instrument. In order to solve this problem, combined with YOLOX target detection, DeblurGAN-v2 image enhancement, DeepLabV3+ semantic segmentation neural network algorithm, a fuzzy pointer meter reading correction and recognition method is proposed. Firstly, the YOLOX network is improved to extract the instrument panel, pointer area and instrument text information, and obtain the instrument parameters. Secondly, enhance the feature extraction ability of DeblurGAN-v2 network to remove the fuzzy influence in the image, then use deeplabv3+ network to divide the dial and pointer. In the aspect of image correction, perspective change and text rectangle correction are used to achieve high-precision correction of the instrument.Experiments show that this method can more effectively solve the impact of complex environment in the detection task, and the detection accuracy is as high as 97.55%, which meets the requirements of automatic detection in industry.

Abstract:To improve the counting efficiency of steel bars in construction sites, an improved lightweight YOLOv4 algorithm is proposed based on the insufficient computing power of hardware equipment in construction units and the dense occlusion of steel bar image objects. GCA-MobilenetV2 lightweight network is proposed to replace CSPDarknet53 as the main feature network of YOLOv4 algorithm. Aiming at the situation of dense steel bar images and serious occlusion between objects, attention-CSP-PANet structure integrating channel attention mechanism is proposed. Aiming at the large number of SPP structure parameters in deep network, DepthLite-SPP structure is proposed to enhance the receptive field of deep network and improve the detection speed of the algorithm. In view of the imbalance between positive and negative samples of the one-stage regression algorithm, CIOU-Focal loss function is designed. The experimental results show that the detection accuracy of the steel bar data set is 98.78%, which is 3.36% higher than that of the original algorithm. The detection speed FPS is 7.6, and the number of parameters is only 1/3 of the original algorithm.

Abstract:For the issue of low accuracy, poor realtime performance and large network model parameters of the existing insulator defect detection technology, an insulator defect detection model based on improved YOLO v4 is proposed in this study. Firstly, a modified VGG convolutional neural network was applied in the backbone feature extraction. In addition, to reduce the complexity of the model, depthwise separable convolution was introduced in the enhanced feature extraction and prediction networks. Moreover, channel attention mechanism was utilized in the enhanced feature extraction network to enhance the important features. The object recognition ability of the model for insulator defect was further strengthened. Finally, employing Average Precision, Frames Per Second, Parameter Scale, etc. as the evaluation indicators, ablation and comparison experiments were conducted on our constructed dataset based on the public dataset CPLID. The results show that the detection accuracy of the improved YOLO v4 model is 98.35%, which is 6.4% higher than that of the traditional YOLO v4 model. Moreover, the detection speed and parameter scale of the improved model are 1.5 times and 37.5% of those of the traditional YOLO v4 model. Accurate and real-time detection of aerial insulator defect imagery can be realized. Furthermore, the improved model also has higher accuracy, higher speed, and smaller parameter scale compared with other mainstream models YOLO v5-M and Faster R-CNN.

Abstract:For the inherent hysteresis nonlinearity of magnetostrictive materials, this paper presents a composite control method based on CMAC (cerebellar model neural network) feedforward inverse compensation and PID. Firstly, CMAC neural network is used to learn and obtain the hysteresis inverse model of giant magnetostrictive actuator (GMA) for compensation, and then the CMAC model is used to learn and adapt online quickly, and PID controller is used to reduce the error and disturbance during tracking control, so as to realize the precision control of GMA. CMAC feedforward inverse compensation controller and CMAC-PID compound control model are established by MATLAB. Finally, the effectiveness of the proposed method is verified by simulation experiments. The results show that the proposed hysteresis model approximated by CMAC neural network has satisfactory accuracy. Under the action of CMAC-PID composite control scheme, the maximum relative error between the expected displacement and the actual displacement of the system is only 2.39%, and the average relative error is less than 0.5%. It shows that the control strategy can adapt to the nonlinear change of the control object and effectively improve the tracking accuracy of GMA.

Abstract:To further study the mapping relationship between narrowband hydroacoustic signal features and the direction of the arrival (DOA), an improved DOA estimation model combining regional Lipschitzs coefficients and local Lipschitzs coefficients is proposed based on the topological ordering of acoustic signal feature vectors based on three-layer self-organizing neural network mapping. This method is used to check the non-expansive mappings formed by the mapping of signal features to angles of arrival, which is a discussion of the regional Lipschitz coefficients as well as a judgment on the superiority of the mapping, using a self-organizing neural network as trainer, based on the topological ordering of feature layers, and combined with local Lipschitzs coefficients to construct an integrated DOA estimation law based on the 1-neighborhood-rules. The simulation experimental results shows that the method is effective in estimating the angle of direction of arrival，with the average error and variance within 10-2 degree; the estimation results also shows good robustness against other commonly used DOA estimation algorithms, when the signal-to-noise ratio (SNR) decreased from 20 dB to 2 dB.