Abstract:The traditional sliding mode control has the problems of slow approach speed, large buffeting, slow response speed and low precision when applied to a synchronous Buck converter. A novel nonsingular terminal sliding mode control method is proposed. Firstly, the traditional sliding mode surface function is modified. By introducing a nonlinear function and attaching the integral term, the system can be forced to converge quickly, thus reducing the response time. Secondly, based on the traditional approach law, a function of the form sigmoid is introduced, and a saturation function is used to replace the sign function, which avoids the high-frequency switching of the sign function near zero, and thus reduces the chattering effect caused by the traditional sliding mode control. Finally, the corresponding model is built on the MATLAB/Simulink simulation platform, and the results show that the adjustment time is 90 μs and the recovery time after sudden loading is 20 μs. The experimental results show that when the speed instruction changes, the proposed method can adjust the time faster, and the anti-interference ability is stronger after sudden loading.

Abstract:This article addresses the issue of low-frequency mechanical energy in the human body and the relatively low energy harvesting efficiency of traditional contact-separation nanogenerator. A novel assisted-structure nanogenerator designed for self-powered communication systems is proposed to tackle this problem. The assisted-structure can effectively transmit and enhance forces, which can reduce the excitation required by the triboelectric nanogenerator for energy collection by 8.6 times with similar output performance. With the aim of self-powered communication, after encoding information, the human body′s mechanical energy acting on the assisted-structure nanogenerator is utilized to modulate and transmit the signals. In the voltage signal, a half-wave rectification circuit is designed to reduce the influence of positive and negative half-axis waveform asymmetry of the voltage waveform of the TENG. The analog voltage signal corresponding to the power assisted structure nanogenerator is collected and after corresponding demodulation and decoding, the original information can be recovered. The experiment successfully transmitted the string "NJUPT".

Abstract:Aiming at the complexity, switching frequency and control performance of the three-vector model predictive current control algorithm for doubly-fed induction generators, this paper proposes an improved threevector model predictive current control algorithm. The algorithm aims to reduce the computational complexity while fixing the switching frequency and improving the control performance. Firstly, the d-axis and q-axis deadbeat control principle of rotor current is used to fast vector selection is performed on the first and second optimal voltage vectors, so as to improve the efficiency of vector selection. Secondly, based on the d-axis and q-axis deadbeat control principle of rotor voltage, the action time of voltage vectors is allocated to reduce the computational complexity of vector action time. Finally, the action sequence of voltage vectors is optimized according to the principle of fixed switching frequency in each control period to fix the switching frequency. The simulation and experimental results show that compared with the three-vector model predictive current control algorithm, the algorithm shortens the running time, effectively reduces the rotor current, electromagnetic torque and output power ripple while fixed switching frequency and has good control performance.

Abstract:In order to solve the problem of low synchronization efficiency in chaotic multi-channel communication, a new four-dimensional super-chaotic system based on the classical Lorenz chaotic system was proposed, and its potential application in secure communication was explored by analyzing its chaotic characteristics in detail. First, the dynamical behavior was analyzed by phase diagram, Lyapunov exponent spectrum, bifurcation diagram, Poincare section diagram, and complexity method. Then, the numerical analysis results were verified by Multisim circuit simulation. Finally, a multi-channel synchronization control scheme was designed using the displacement projection method. The research results show that the new four-dimensional super-chaotic system has complex dynamic behavior and can generate hidden attractors under different initial conditions. The circuit simulation results are consistent with the numerical analysis, and the multi-channel error system has good chaotic characteristics. This indicates that the system has high reliability and flexibility, laying a good foundation for studying chaotic systems in the field of information security.

Abstract:Aiming at the problems of strong coupling, multivariate, nonlinearity, and time delay in the simulated moving bed chromatographic separation system, a fuzzy decoupling control method for simulated moving bed component purity based on the self-tuning PID parameters of BP neural network with an improved sand cat swarm optimization algorithm is proposed. First, fuzzy decoupling is used to eliminate the coupling between the purity control loops of components A and B. Then, combined with the improved sand cat swarm optimization algorithm and BP neural network, the adaptive adjustment of PID parameters is realized, thereby effectively controlling the purity of components A and B. In the improved sand cat swarm optimization algorithm, the Cubic chaotic map is introduced to initialize the sand cat population to improve the uniformity of population distribution; the variable spiral search strategy is added in the prey search phase to enable the sand cat swarm to have more search paths to adjust its position; simultaneously, the alert mechanism of the sparrow search algorithm is integrated to accelerate the convergence speed of the algorithm. The effectiveness of the improved sand cat swarm optimization algorithm is verified by 12 CEC2022 test functions. The simulation results show that the proposed method can not only effectively eliminate the coupling effect between the purity control loops of components A and B, but also demonstrates excellent performance in various real-world application scenarios. Compared to traditional PID control methods, the proposed approach reduced the settling time by 75.40% and 77.57%, and decreased the overshoot by 91.84% and 81.96% under flow rate fluctuation conditions. This method possesses strong anti-interference ability and good robustness, and improves the control performance of the entire system.

Abstract:DC microgrid, as an important form of comprehensive utilization of new energy, plays an important role in the effective utilization of distributed power generation. A cascaded active disturbance rejection control strategy based on deep deterministic strategy is designed to address the problem of deteriorating power quality of interface converters in DC microgrids under real-time disturbances. Introducing a new observer into the original state observer and cascading it to form a cascaded self disturbance rejection, improving the initial observer′s resistance to disturbances, and establishing a linear feedback law to counteract the observed disturbances, enhancing the system′s ability to suppress real-time disturbances. To suppress the adverse effects caused by the uncertainty characteristics of the system controller parameters, a deep deterministic gradient algorithm is introduced to achieve adaptive tuning of the controller parameters, ensuring the real-time optimality of the system parameters and further enhancing the system′s ability to suppress real-time disturbances. Finally, theoretical analysis was conducted on the tracking performance and disturbance rejection performance of the observer, and the performance differences of different control strategies under various operating conditions were compared through simulation, verifying the correctness and superiority of the proposed control strategy.

Abstract:The state of charge of lithium batteries is the most important state parameter for the safe operation of lithium batteries. In order to improve the estimation accuracy of lithium battery SOC, this paper proposes a dual adaptive volumetric Kalman filter algorithm. The second-order DP equivalent circuit model of lithium batteries is used for offline identification of state parameters, and a high-precision volumetric Kalman filter algorithm is used to estimate a single SOC. An adaptive factor is introduced to estimate real-time noise. On the basis of obtaining SOC, the internal resistance of lithium batteries is estimated in real time, and the dual adaptive volumetric Kalman filter algorithm is used to estimate SOC. In order to fully verify that the conclusions of this paper meet the requirements of actual working conditions, this paper conducted dynamic stress testing, federal city driving, urban driving cycle and suburban driving cycle simulation experiments. The SOC errors of the first three working conditions obtained by the algorithm are within 0.5%, and the error of the suburban driving cycle working condition is within 1%, and it has strong robustness, proving that the algorithm is valid..

Abstract:Aiming at the problem that the dung beetle optimization algorithm (DBO) has insufficient global exploration ability and easily falls into the local optimum, resulting in poor UAV 3D trajectory planning, an improved dung beetle optimization algorithm (EDBO) is designed. Firstly, a leader dung beetle group incorporating the idea of a grey wolf leader group is divided from the stealing dung beetle population to enhance the diversity and robustness of the algorithm; secondly, a population switching strategy is used, where the behaviour of dung beetle individuals with the same number is no longer fixed to improve the algorithm′s ability of global exploration and local mining; and lastly, a Beta-distributed dynamic inverse learning strategy is used to help dung beetles to evolve better. The proposed algorithm is compared with 4 optimization algorithms for UAV 3D trajectory planning, and the simulation results show that EDBO can generate trajectories with smaller cost function values more stably in the 3 scenarios, and the average cost function values are reduced by 9.8%, 10.4%, and 16.5% compared to the original DBO algorithm, respectively.

Abstract:Aiming at the hidden security risks caused by some mobile workers illegally smoking in no-smoking areas of the park, a deep learning method for smoking behavior detection combined with human bone key point detection and improved YOLOv7 cigarette detection was proposed. The method first extracts the coordinate information of the key points of the human body through OpenPose, calculates the ratio of the distance between the hand, nose and neck, and the angle between the hand, elbow and shoulder, and determines whether the smoking posture is met. Then, the improved YOLOv7 algorithm is combined to detect whether there is a cigarette in the image to finally determine whether there is smoking behavior. The improved YOLOv7 algorithm introduces a global attention mechanism module, strengthens the semantic and position information, uses transposed convolution to improve the upsampling method, reduces information loss, and adopts the MPDIOU loss function to enhance the accuracy of the regression results and improve the detection accuracy of small cigarette targets. Through experimental tests, the accuracy of this method reaches 95.45%, which can effectively detect smoking behavior.

Abstract:In response to the challenges of slow convergence and suboptimal path optimization in complex obstacle environments, this paper presents an adaptive path planning algorithm based on IRRT*-Connect that is tailored to the environmental complexity. This algorithm combines Informed-RRT* and RRT-Connect, namely IRRT*-Connect algorithm for initial path planning to improve the efficiency of initial path planning; additionally, it introduces a sampling constraint probability p to confine sampled areas and augment the purposefulness of sampling. Furthermore, a step length calculation method based on environmental obstacle coefficients is devised to dynamically adjust extension step lengths, thereby bolstering the algorithm′s adaptability in traversing complex environments. Through the comparison of multiple sets of experiments, we show that the number of algorithm nodes is reduced by 9.75%, and the path length is reduced by 20.82%, and the planning time is shortened by 3.08%, which proves that the improved IRRT*-Connect adaptive step path planning algorithm has a strong ability to adapt to the environment, high node utilization, and the planning effect is better.

Abstract:Aiming at the problem that the operation state of three-phase PWM rectifier in AC microgrid is easily affected by load disturbance and filter capacitor parameter perturbation, a sliding mode control method of three-phase PWM rectifier based on nonlinear disturbance observer is proposed in this paper. Firstly, the load disturbance and filter capacitor parameter perturbation are regarded as lumped disturbances, and the nonlinear model of voltage loop is constructed. Secondly, a nonlinear disturbance observer is designed to realize the online estimation of lumped disturbances. In order to achieve active anti-interference control and sliding mode chattering suppression, the estimated lumped disturbance is introduced into the sliding mode surface, and a new sliding mode surface with disturbance estimation is obtained. Finally, the voltage outer loop sliding mode controller is designed based on the sliding mode control theory, and the stability is proved by Lyapunov stability theory. The dynamic performance test results show that the nonlinear disturbance observer can estimate the lumped disturbance quickly and accurately. Compared with the double-loop PI control, the transition process time of the proposed control is reduced by 60%, and the voltage drop is reduced by 54%. When the filter capacitor parameters change, the dynamic performance of the proposed control remains good. The steady-state performance test results show that compared with the traditional sliding mode control and the double-loop PI control, the total harmonic distortion of the current of the proposed control is reduced by 70% and 44% respectively, and the sliding mode chattering is effectively suppressed. The test results show that the proposed control method has strong anti-load disturbance ability and good robustness of filter capacitor parameters.

Abstract:In view of the fact that the network selection algorithm in the heterogeneous wireless network has few quality of service indicators, and the frequent switching of users is becoming more and more serious, In this paper, a vertical handover method for heterogeneous wireless networks based on subjective and objective weighting combined with improved deep reinforcement learning is proposed. Firstly, a software-defined network architecture supporting heterogeneous wireless networks was proposed; secondly, an attribute weighting algorithm combining subjective and objective weighting was proposed; finally, the network selection problem is solved by using Dueling-DQN. The simulation results show that the proposed algorithm reduces the number of switching times by 11.25%, 13.34%, 18.76% and 13.75% respectively under different user types of networks, and increases the throughput by 16.64%. Therefore, the algorithm proposed in this paper effectively avoids ping-pong switching, reduces the number of switching times and improves the throughput.

Abstract:Aiming at the problem of map offset caused by deceleration zone or pothole zone encountered by vehicles in automatic parking scenarios, and the problem of poor real-time performance and low accuracy of the system caused by changing dynamic environment, a variable sliding window look-around SLAM algorithm for automatic parking is proposed. Firstly, IMU is used to calibrate the real-time attitude of the collected panoramic image to improve the accuracy of mapping. Secondly, combined with the advantages of multi-sensor fusion, the IMU and odometer data are fused to estimate the vehicle pose. Finally, the variable sliding window algorithm is used to accelerate the back-end optimization and improve the real-time performance and accuracy of the system. The simulation test results show that the method solves the problem of map offset in the deceleration zone or the pit zone, and the efficiency and real-time performance are improved by 31.35% and 25.06% respectively in the sparse feature environment. The real vehicle test results show that the method can achieve the positioning accuracy with an average error of 0.039 m, which provides safety guarantee for parking.

Abstract:In order to solve the problems of low accuracy and large number of parameters in traffic sign detection, this paper proposes an improved SC-YOLOv8 traffic sign detection algorithm based on YOLOv8s. This algorithm uses the downsampling Adown module to replace the ordinary downsampling Conv, improving the model′s perception ability of the target; replace the Bottleneck in C2f with the SCConv module and design a brand new C2f-SC module, significantly reducing model parameters; adding a 160×160 scale detection head and removing a 20×20 scale detection head, effectively improving detection accuracy; finally, the idea of using WIoU loss function is used to improve MPDIoU, replacing the original CIoU loss function with Wise-MPDIoU, alleviating the problem of imbalanced positive and negative samples. The algorithm was validated on the TT100K traffic sign dataset, and compared with the original model YOLOv8s, the accuracy P increased by 4.8%, the recall R increased by 6.7%, the mAP50 increased by 6.6%, and the parameter count Params decreased by 61.5%. Proved the effectiveness of the improvements made.

Abstract:Aiming at the problems of limited accuracy of UAV detection of foreign objects on power transmission lines, high model computational complexity and limited computational speed, a power transmission line foreign object detection method SC-YOLO which improves YOLOv8 is proposed. This method introduces StarNet to construct C2f_Star module to realize the lightweight of Neck network, effectively reducing the number of model parameters and calculation amount, and at the same time improves the feature extraction ability of Neck by increasing the dimension of feature space; adds convolution attention fusion module after the backbone network outputs feature map to improve the backbone network′s preliminary feature extraction ability of input feature map, and enhance the overall detection effect of the model; replaces the original detection head with dynamic detection head to improve the model′s dynamic adjustment ability to different inputs and the degree of attention to key information; uses WIoU as the bounding box loss function and EMA-Slide Loss as the classification loss function to improve the model′s generalization ability and detection performance. Experimental results show that the proposed SC-YOLO has 8.02% fewer computational amount than the original model, and mAP is increased by 1.4 percentage points, reaching a detection accuracy of 95.2%. RC-YOLO reduces the computational complexity while achieving a high detection accuracy, and is highly feasible and practical.

Abstract:Aiming at the existing distracted driving detection algorithms, this paper constructs a YOLOv8-EFM distracted driving detection and recognition model based on improved YOLOv8-pose. Firstly, by replacing the backbone network of YOLOv8-pose with EfficientViT, combined with the complementarity between CNN and VIT, the detection accuracy is improved; secondly, replacing the Bottleneck module in C2f with FasterBlock module, increasing the detection rate and reducing the model parameters; finally, the lightweight MLCA attention module is added after SPPF, achieving a good balance between model size and accuracy. The experimental results show that the YOLOv8-EFM model constructed in this paper can detect mAP 50 with 98.9%, and the model size is only 9.7 M. This method can not only detect the specific distraction behavior, but also detect the human skeleton of the upper body, which can be effectively applied in the detection scene of distracted driving.

Abstract:To reduce the impact of foggy days on transmission line inspection images, a foggy day transmission line inspection image de-fogging method, Diff-EaT, is proposed for the current mainstream de-fogging algorithms that have high computational costs, poor detection performance after image de-fogging, and difficult to deploy. The method adopts a fusion Transformer′s diffusion model structure, and to reduce the computational complexity of the multi-head self-attention in the feature extraction module in the ViT, multi-head external attention is used instead of multi-head self-attention to reduce the computational load and enhance feature learning. Meanwhile, a mixed-scale gated feed-forward network is designed to integrate a pick-and-pass mechanism after the depth-separable convolution of input features to improve local information capture. Tested on synthetic and real datasets, quantitative and quantitative metrics prove their effectiveness with clearer details of recovered images. In the defogging detection system, the real inspection images are defogged and then detected using YOLOv7, mAP@0.5, recall rate, and checking accuracy rate are improved by 6.92%, 9.58%, and 4.11%, respectively, and this paper′s method effectively improves the detection confidence after defogging. Defogging detection systems can be applied in real-world scenarios. Also in ablation experiments to demonstrate the effectiveness of its improvements.

Abstract:In response to the issues of visual performance blurring and underestimation of high echo values in the modeling results of traditional short-term rainfall prediction models for historical radar data, we propose a short-term rainfall radar echo extrapolation model that integrates 3D convolution and dual-end attention mechanism 3DuA-Net. Using ST-LSTM space-time long short-term memory networks as the recurrent units, replacing ordinary convolution with 3D convolution enhances the model′s capability to capture short-term motion features from a global perspective. Additionally, an efficient dual self-attention module DuAtt is proposed to improve the model′s ability to preserve and integrate important local and global features in long-term radar image sequences. Experimentation conducted using publicly available Doppler radar datasets from the Shenzhen Meteorological Bureau shows that, at 10、20、 40 dBz thresholds, the model exhibits an average improvement of 7.74% in the CSI metric compared to the Conv-LSTM model, an average improvement of 5.54% in the HSS metric, a decrease of 3.8% in the MAE metric, and an improvement of 8.86% in the SSIM metric.

Abstract:The failure of rolling bearing in the running process may cause serious consequences, so it is of great significance to carry out on-line detection of bearings. A portable bearing fault online detection system based on STM32 and FPGA is designed to solve the problem of on-line detection of rolling bearings. In terms of hardware, the FPGA chip is used as the data processing unit to realize the A/D conversion and on-line acquisition of bearing vibration signal, and the signal noise reduction, envelope spectrum analysis and fault frequency extraction are carried out. The bearing vibration signal time domain waveform and fault spectrum are displayed in real time by LCD screen. The STM32 single chip microcomputer is used to design the system UI control interface, control the sampling rate, waveform display and display the diagnosis result, and realize human-computer friendly interaction. In the aspect of algorithms, adaptive line enhancement technology is implemented by FPGA to reduce the noise of the collected signals, and the fault spectrum is obtained by envelope spectrum analysis and the fault characteristic frequency is extracted. Finally, the system is tested by the self-built mechanical integrated fault simulation test bench. The experimental results show that the system can effectively extract the bearing fault frequency, and the speed is improved by about 30 times compared with the software detection scheme, which can meet the requirements of online detection.

Abstract:To address the issues of nonlinearity and high dimensionality in industrial process data, a fault detection method combining Canonical Variate Dissimilarity Analysis and Locally Linear Embedding is proposed. The dissimilarity matrix constructed by the CVDA algorithm can effectively monitor faults, but it relies on linear projections and is only sensitive to changes in linear features of the data structure. The LLE algorithm is used to map high-dimensional data to a low-dimensional space by preserving local relationships between samples, further extracting features and uncovering nonlinear characteristics and local neighborhood information. Finally, an isolation forest model is established in the low-dimensional manifold space to obtain anomaly scores of sample points as the fault detection evaluation criterion. Through a set of nonlinear numerical examples and the Tennessee Eastman chemical process data, the proposed method is compared and analyzed with traditional KPCA、PPA and CVDA to verify its effectiveness and superiority.

Abstract:In the field of oil exploration and development, accurate prediction of mechanical drilling rate is crucial for improving drilling efficiency and reducing engineering risks. Accurate mechanical drilling rate prediction provides an important basis for formulating drilling plans and assessing drilling risks. However, traditional drilling rate equations and machine learning methods cannot fully consider the factors affecting the mechanical drilling rate in complex nonlinear drilling systems. This paper presents a mechanical drilling rate prediction model based on a genetic algorithm optimized backpropagation neural network (GA-BPNN), using historical drilling data from an oil field in the South China Sea. The data preprocessing includes SG smoothing, normalization, and comprehensive feature parameter selection through Pearson, Spearman, and Kendall correlation coefficients. The model is compared and verified with BP, RBF, MEA-BP neural network models, and traditional machine learning methods such as ELM, RF, SVM, and KNN. The experimental results show that the GA-BP has an R2 of 0.967, and the predicted values are in good agreement with the measured values, with an accuracy increase of 17.64% compared to the standard BP neural network prediction R2, and more accurate predictions than other models. This hybrid intelligent prediction model can accurately predict and prevent drilling accidents, provide effective data for guiding oil field drilling construction parameters, thereby improving the economic benefits of drilling construction.

Abstract:The traditional harmonic detection algorithm is affected by noise, resulting in low detection accuracy and easy distortion at the boundary. In this paper, based on successive variational mode decomposition, a harmonic detection method combining wavelet denoising and characteristic waveform matching extension is proposed. Firstly, the adaptive wavelet threshold function is used to smooth the signal noise and eliminate the interference of bad data to the decomposition results. Secondly, the characteristic waveform matching extension method is used to extend the edge of the signal and then cut it to curb the distortion at the end of the waveform caused by the boundary effect. Finally, the harmonic signal is detected by successive variational mode decomposition, the amplitude-frequency information of steady-state harmonics is extracted, and the start-stop time of transient harmonics is located. Simulation results show that the proposed method can effectively reduce the noise interference and reduce the waveform distortion caused by the boundary effect. In electric arc furnace example signal simulation，the average amplitude error and the average frequency error are 0.545% and 0.146% respectively.