Abstract:The demodulation of echo signal information primarily relies on the FFT. This paper proposes an FPGA-based signal processing unit scheme utilizing the split-radix FFT for efficient information demodulation of echo signal. In the original scheme, range dimension demodulation was performed by a DSP, while velocity dimension demodulation was conducted on a PC. However, the single-core DSP could not handle range dimension demodulation and data uploading in parallel, resulting in suboptimal real-time signal processing. Leveraging the large-scale parallel processing and flexible programmability of FPGA, this paper integrates both demodulations into the FPGA, achieving a parallel design for data processing and uploading, thereby enhancing realtime data processing. The split-radix FFT is crucial for achieving this functionality. By designing adaptive computation units and data flow control units, this paper improves the processing speed of the split-radix FFT. Compared to traditional structures, the proposed scheme reduces computation cycles by over 5.4%; compared to the improved split-radix FFT, it reduces computation cycles by over 2.31%. Closed-loop test results demonstrate that the integrated signal processing unit designed in this paper achieves a target signal-to-noise ratio (SNR) of 70 dB, effectively meeting the requirements for signal processing.

Abstract:To improve the sampling rate of time-interleaved analog-to-digital converter (TIADC) systems and address the limitation imposed by the mainstream method of reference channel in time misalignment error correction, a method for time misalignment error correction based on internal channel reference is proposed. In this method, the internal channels of the system are utilized as reference channels for estimating time mismatch errors. The time mismatch error parameters are estimated using the internal channel reference method, and the derivative of the target signal is obtained using the central differencing method. The time mismatch error is then reconstructed and compensated for using the estimated error parameters and the derivative of the target signal. Simulation results demonstrate that the proposed method can achieve an improvement of around 30 dB in spurious-free dynamic range and an increase of about 5 bits in effective resolution without additional restrictions, in a background, adaptive manner. The error suppression capability is comparable to that of the reference channel method.

Abstract:An analog front-end circuit for multi-meteorological sensors which mainly includes LDO, programmable gain amplifier, SAR ADC, and humidity measurement circuit was designed. The programmable gain amplifier uses all-difference rail to orbit as the main structure to suppress the noise, and the continuous-time Auto-Zero calibration technology is adopted to reduce its input imbalance voltage. For the 14-bit SAR ADC, in order to reduce the average power consumption and area of CDAC, a segmented differential DAC capacitor array based on the VCM-based switching strategy was designed. Finally, based on the principle of the relationship between the capacitance value of the humidity sensor and the frequency of the rectangular wave, a humidity measurement circuit was designed. The frequency error of the humidity measurement circuit is 0.03%. The analog front-end circuit is based on Hua Hong′s 0.18 μm CMOS process, and the circuit design, layout drawing and simulation verification are carried out through Cadence Spectre software. The post-simulation results show that the circuit as a whole can realize the function of amplifying the input analog signal and finally outputting the digital code, its effective number of bit (ENOB) is 11.40 bit, SINAD is 70.37 dB, SNR is 71.05 dB, SFDR is 83.85 dBc, and THD is -78.55 dB.

Abstract:The electric-drive-reconstructed onboard charger (EDROC) can effectively reduce costs and occupy the volume of charging equipment by reusing the drive system components into charging system and executing the electric drive and charging functions in a time-sharing manner. However, there is a risk of irreversible demagnetization of permanent magnet synchronous motors in EDROC systems at high temperatures. In view of this, this paper analyzes the degree of demagnetization of the motor under different conditions, and proposes the safe and stable operating conditions of the motor under charging conditions. Firstly, the winding currents of the EDROC system under DC charging mode and in-motion charging mode are analyzed. And then, the irreversible demagnetization model of the motor is established. Moreover, for different charging modes, detailed analysis is conducted on the demagnetization degree of the motor under different temperatures and charging currents. Based on this, the indispensable conditions for safe and stable operation of motor under charging conditions is established. Finally, experiments were carried out based on a 2 kW laboratory prototype to verify the accuracy of the irreversible demagnetization model and the necessity of the conditions for safe and stable operation of the EDROC system.

Abstract:Low-dropout regulator (LDO) without external capacitors can provide low noise and low ripple power supply voltage for highly integrated on-chip system (SoC). To address slow transient response speed and poor stability of the capacitorfree LDO, active feedforward compensation technology was adopted to accelerate the charging and discharging rate of power transistors to promote the transient response speed of LDO; and introduce the left half-plane zeros to improve the stability of the system. Meanwhile, the feedforward compensation circuit can adaptively switch between the two-stage cascaded amplifier and the three-stage cascaded amplifier structures when the LDO changed between light and heavy loads, ensuring that the LDO remained stable throughout the full load range. The circuit was designed using the TSMC 0.18 μm CMOS technology. The simulation results show that when the on-chip load capacitance is 20 pF and the current load varies from 0 to 200 mA, the designed capacitor-free LDO can operate at an input power supply voltage of 1.7 to 2.8 V, stably output a power supply voltage of 1.5 V, and the system has a phase margin of more than 45° across the full load range. The overshoot and undershoot voltages are less than 100 mV when the load current jumps between 0 mA and 200 mA within 1 μs, and the recovery time is less than 0.6 μs. The designed LDO has achieved good performance in high stability, fast transient response, and wide load range.

Abstract:In complex industrial systems due to the dramatic increase in the number of sensors, high-dimensional noise and random disturbances are generated, which seriously affect the data continuity and control accuracy of multivariate time series. However, the existing pairs of multivariate time series have the problems of temporal inconsistency over time, deviation of space vectors, and redundancy of spatio-temporal graphical models. In this paper, a new multivariate time series anomaly detection method STGAD is proposed. First, a gating mechanism is introduced to improve the multiscale convolutional network from a high-resolution granularity level to control the process of information interaction between features. Then, two graph structures are designed to eliminate redundant spatio-temporal dependencies, enabling GAT to effectively learn spatio-temporal correlations. In addition, an attention mechanism-based GRU module is proposed to capture the importance of variables over different time windows. Finally, modules for joint optimization prediction and reconstruction. Extensive experiments on three publicly available datasets show that the average F1-score of the proposed model is higher than 0.94, which significantly outperforms other benchmark models on high-dimensional datasets.

Abstract:In order to improve the throttle response speed and the following characteristics, a particle swarm optimization (PSO) based method was proposed to optimize the non-singular fast terminal sliding mode control (NFTSMC) in plug-in automotive electronic throttle system. Firstly, considering the parameter uncertainty of throttle components and the intake disturbance, the system disturbance is estimated and fed into the controller. Then the parameters of NFTSMC are adjusted by PSO algorithm, and the stability is proved according to Lyapunov′s second theorem. Compared with PSO-SMC, the response time of PSO-NFTSMC is reduced by 82 ms, the maximum value following error is reduced from 0.32° to 0.005°, and the maximum value of error is reduced from 0.5° to 0.004°. It is proved that the proposed plug-in automotive electronic throttle system under the control of PSO-NFTSM has faster response speed and more consistent with the preset error. The proposal and application of this method provide a reference for improving the combustion and emission efficiency of plug-in vehicle engines and improving the dynamic performance of vehicles.

Abstract:A human activity recognition model based on an improved Transformer-BiLSTM network is proposed to address the problem of decreased accuracy in activity recognition methods due to the high dimensionality and large noise of time series collected by wearable sensors. The model leverages the advantages of Transformer encoder in handling long-range dependencies and parallelized computations to enhance the efficiency of sequence feature extraction. Subsequently, the features are passed to a bidirectional long short-term memory network with skip residual connections, where two residual connections replace numerous convolutional layers while retaining essential information. Additionally, an attention layer integrated with time information encoding is proposed to enhance the model′s expressive power and understanding of temporal data. Experimental results show that the model achieves an accuracy of 98.38% on public datasets, effectively improving the accuracy of human activity recognition.

Abstract:To improve the efficiency of internal inspections in power plants, this paper proposes an inspection scheme based on intelligent robots. Given the complexity of power plant environments, achieving efficient and accurate autonomous mapping by robots in unknown settings is crucial. We designed an active SLAM method using composite exploration points, incorporating plane segmentation and vector synthesis to guide exploration trajectories, thereby reducing map uncertainty from random exploration. The boundary point evaluation function is enhanced by considering boundary length gain to improve exploration efficiency. The method involves using plane segmentation to search for boundary points around the target, with an evaluation function based on movement distance and boundary length to determine the optimal boundary point with the largest exploration range. Composite exploration points are created through vector synthesis of the optimal boundary and target points, guiding the robot for simultaneous mapping and tracking. Real-time positioning and mapping technology is used to construct the current environmental grid map, achieving target point tracking and autonomous mapping through sequential exploration points. By setting new target points, tracking and expanding the mapping range are achieved. The proposed algorithm exhibits a tendency towards exploration, performing depth-first search on the grid map while considering the traction effect of target points, thereby avoiding multiple trajectory overlaps and loops. Experimental results demonstrate that this method achieves target tracking and high-precision mapping in unknown environments with fewer exploration steps and shorter paths.

Abstract:Aiming at the issue of high reflux power in the three-level hybrid full-bridge bidirectional DC-DC converter, this paper proposes a reflux power optimization control strategy based on double phase-shift control with the objective of minimizing the reflux power. Firstly, the mathematical model of the three-level hybrid fullbridge DC-DC converter is established, and the characteristic relationship between the converter′s reflux power and transmission power is analyzed; then, taking the reflux power as the objective function, the equation and inequality constraints of different modes are converted into the constraint function by the KKT conditional method, and the Lagrange polynomial functions are constructed to find the optimal combination of the shift ratio; finally, in order to improve the rapidity of the system dynamic response, the direct power control method in the loop is adopted to realize the dynamic response of the converter in the face of unexpected situations. The simulation results demonstrate that the improved control strategy proposed in this paper can effectively reduce the reflux power and improve the dynamic response performance of the system when the input voltage and load change.

Abstract:Multi-source domain adaptation is an important branch of transfer learning. Category shift, a prominent challenge in this field, stems from the mismatch between category distributions in the source and target domains. To address this problem, a category-aware and reweighting-based multi-source domain adaptation algorithm is proposed. The algorithm enhances positive transfer between similar categories through a category-aware strategy and introduces a reweighting moment matching strategy to reduce distribution differences at various levels. Additionally, adaptive weights are constructed using pseudo-labels to effectively mitigate the impact of category shift. Experimental results on the Digits-Five and Office-Caltech10 datasets show that the proposed algorithm achieves classification accuracies of 94.11% and 97.18%, respectively. These results indicate that the proposed algorithm significantly improves accuracy in scenarios with category shift compared to current typical multi-source domain adaptation algorithms.

Abstract:Microwave laboratory measurement of phased array antenna plays an important role in the development of phased array antenna. In the microwave laboratory measurement, in order to solve the problems of phased array antenna measurement content, low efficiency, low precision, poor system universality and high cost, this paper designed a general wave control extension, through hardware design, reserve a variety of communication interfaces, fixed interface protocols, suitable for multi-class phased array antennas. Through software programming, the pulse and beam switching time are accurately measured, and the microwave laboratory measuring system is matched to the plane near-field darkroom and tight field darkroom. The test process and sequence diagram are given. The experimental results show that the scanning accuracy of multi-wave position and single-wave position measurement pattern is basically the same, the test efficiency is significantly improved, and the measurement cost is reduced. Rapid and accurate measurement of different types of multi-wave phased array antennas in the microwave laboratory is realized by the application of general wave controller. In the microwave laboratory measurement, in order to solve the problems of phased array antenna measurement content, low efficiency, low precision, poor system universality and high cost, this paper designed a general wave control extension, through hardware design, reserve a variety of communication interfaces, fixed interface protocols, suitable for multi-class phased array antennas. Through software programming, the pulse and beam switching time are accurately measured, and the microwave laboratory measuring system is matched to the plane near-field darkroom and tight field darkroom. The test process and sequence diagram are given. The experimental results show that the scanning accuracy of multi-wave position and single-wave position measurement pattern is basically the same, the test efficiency is significantly improved, and the measurement cost is reduced. With the application of general wave controller, different types of multi-wave phased array antennas can be measured quickly and accurately in the microwave laboratory.

Abstract:To improve the safety of electrical equipment and ensure the stable operation of the power system, it was necessary to effectively detect and identify partial discharges in high-voltage cables. This study develops a high-voltage cable partial discharge detection and recognition system based on the heterogeneous sensor data fusion. In this system, the distribution of cable electric fields were detected by using the electric field sensors, and identified insulation hazards in cable joints, the occurrence and degree of partial discharge in hazard areas were detected by using the pressure wave sensors, the improved adaptive threshold discrete wavelet transform was employed for signal denoising, the data classification features was enhanced by using the improved Gram angle field feature transform to, and the partial discharge pattern recognition was realized by using the residual convolutional neural network with improvement of the efficient channel attention. The sharp discharge, internal discharge, and surface discharge were selected as the objective to conduct experimental test, the results shows that the system can accurately detect the partial discharge characteristics of cables and effectively identify the discharge mode of high-voltage cable defects. The partial discharge detection rate in the laboratory reached 100%, and the discharge mode recognition rate reached 96.0%. It also performed well in engineering application environments, which is of great significance for the safety of cable use and the stability of power grid operation.

Abstract:The proposed approach presents a design and implementation method for a software-defined broadband signal acquisition and playback system based on SoC, aiming to enhance the simplicity and flexibility of the system. By employing direct RF sampling technology, high-speed signal acquisition and playback are achieved, while digital signal processing technology enables more flexible programmable down conversion and signal processing in the digital domain. This system offers secondary development capabilities for software radio, addressing the issues of complex design and limited flexibility associated with traditional wideband signal acquisition systems that utilize superheterodyne IF sampling. Experimental results demonstrate that this signal acquisition and playback system supports an input/output frequency range of 1 MHz to 4 GHz, with a minimum signal bandwidth of 80 MHz. The sampling clock frequency is not less than 200 MHz. Moreover, the output power range for playback signals spans from -90 dBm to 0 dBm, with adjustable power levels in increments of 1 dB. Overall, this system exhibits exceptional flexibility.

Abstract:The diamond nitrogen-vacancy center has become a very promising magnetic field sensor with its advantages of high sensitivity, sensitivity to external magnetic field, and detectability at room temperature. In this paper,we introduce the principle of magnetic measurement by nitrogen-vacancy center, and a magnetic measurement system based on nitrogen-vacancy center is designed and constructed, in which the noise is reduced to about 1/8 by the laser-fluorescence differential optical path, and the microwave mixing modulation method can enhance the slope of the ODMR spectral line by about 2.5 times, and the sensitivity of magnetic field measurement of 0.934 nT/Hz1/2 is finally realized based on the diamond samples with the concentration of nitrogen-vacancy center <300 ppb. Meanwhile, the NV magnetometry system based on dual microwave modulation and demodulation was designed, and a measurement uncertainty of 1.76×10-4 was achieved at a sampling rate of 10 kHz, and a measurement uncertainty of 10 ppm was achieved at a statistical time of 0.2 s based on the Allen variance method.

Abstract:There are many types of apparent diseases in concrete bridges, and each type of disease has different characteristics such as shape, size, color, etc. A machine vision detection method for surface defects of concrete bridges based on laser scanning is proposed. Obtain surface data of concrete bridges through laser scanning equipment and convert it into two-dimensional images through rendering technology. Sharpen the edge areas of the image to make the image quality more natural and clear. Locate the region of interest in the apparent image of concrete bridges, and fuse the gradient direction histogram features extracted in the region of interest with local binary pattern features to form a set of concrete bridge apparent samples. SVM is used as the classifier, and the fused features are used for offline training of the classifier. Based on the training results, concrete bridge apparent disease detection is carried out. The experimental results show that the proposed method has a maximum fuzzy coefficient of 0.98 and a maximum quality index of 0.99, which is close to the optimal value of 1. This indicates that it can effectively improve the quality of concrete bridge surface images and obtain more accurate results of concrete bridge surface disease detection.

Abstract:To address the issue of existing circular meter reading algorithms being susceptible to adverse factors such as shooting angles and complex environments, this paper proposes a perspective rectification-based automatic circular meter reading system successfully deployed on mobile devices. Initially, a lightweight meter detection model and key target detection model based on YOLOv8n were redesigned. After simplifying the network structure, a mobile-optimized lightweight model was developed by integrating PConv into the LiteFFM module and employing LAMP pruning techniques. The improved model significantly reduces computational requirements, with a parameter reduction of 97.75% and GFLOPs as low as 0.4. Furthermore, the paper introduces an effective circular meter rectification method that constructs a rectification matrix from edge contour point sets of the meter to eliminate distortions caused by shooting tilt, integrating markers and scale characters for precise rotational rectification. Lastly, an enhanced angle method calculates accurate readings, and a rectification guidance mechanism within the app optimizes the user shooting experience. Experiments show that under severely tilted conditions, the correction algorithm can reduce the average relative reading error by 60.68%. The system runs at 9 FPS on mobile phones, with an average relative reading error of only 1.76% in complex environments, outperforming existing advanced methods and demonstrating high robustness.

Abstract:Aiming at the problems of the traditional sensor′s late detection of fireworks and its inability to give details of fireworks, as well as the imbalance between detection efficiency and accuracy of the current mainstream fireworks detection algorithms, an improved YOLOv5s light detection algorithm for fireworks was proposed. The second convolutional module in Backbone is replaced with Stem module, which can improve the model′s detection performance of small target space information and effectively control the total floating point operand. C3Ghost module and Ghost convolution module are introduced in Backbone and Neck to reduce the number of network parameters and improve the performance of fireworks detection. In order to distinguish the importance of different features in the process of feature fusion, a structure of adding learnable weight parameters to PAN is proposed, which significantly improves the average accuracy of fireworks detection. The experimental results show that compared with the original model, the weight of the model is reduced from 14.4 M to 10.2 M, GFLOPs is reduced from 15.8 to 3.7, and the average accuracy is increased by 1.1%. The improved model has improved the performance of pyrotechnic detection while being lightweight.

Abstract:In response to the issues of large model size, complex computations, and high resource demands on computational platforms in safety helmet detection models, a lightweight safety helmet detection algorithm called YOLOv8-MBS, based on an improvement of YOLOv8, is proposed. A new lightweight backbone module was first formed by combining MobileNetv3 with SPPF, reducing the algorithm′s parameter and computational load. Moreover, the algorithm′s feature extraction and representation capabilities are enhanced using a weighted bidirectional feature pyramid network, which also reduces the false detection rate. Finally, the SimAM module is incorporated to improve the network′s correlation between positional information and safety helmet features without increasing the computational burden. Experimental results show that compared to the original YOLOv8n network, the improved YOLOv8-MBS maintains high detection accuracy while reducing computation by 35.96%, the number of parameters by 25.63%, and model size by 23.22%, and increasing the frame rate by 12.52 fps. The lightweight nature of the model reduces deployment costs and provides theoretical support for embedded deployment and large-scale applications.

Abstract:The vehicle safety and stability during the driving process is directly influenced by the tire wear, and the vehicle safety can be improved by detecting the degree of tire wear to find and process the abnormal state of the tire timely. The tire wear degree can be detected with the sensor installed in the tire or the tire image directly. However, the sensor installed method has high cost and cumbersome installation process, and the image-based detection method requires more samples and the detection accuracy is not high. Therefore, a tire wear detection method based on fused texture features is proposed in this paper. The training set was constructed with 25 tire images of 5 different wear degrees, and each image was uniformly cropped into 12 sub-images, and the gray level co-occurrence matrix and local binary patterns features were extracted by median filtering, and the fusion features were obtained by principal component analysis and stitching fusion method. Then, the classifier was trained by sparrow search algorithm and the random forest method with the fusion features. Finally, the algorithm was tested with 225 acquired images of different degrees of tire wear. The results show that the average detection accuracy reaches 97.33%, which is significantly higher than that of a single feature and other classification methods, so, the proposed method can be used to detect the tire wear quickly and accurately.

Abstract:In the process of coal transportation, there are often foreign bodies scratching or tearing the transportation belt, resulting in safety accidents such as coal outlet blockage. Therefore, it is necessary to identify and classify the foreign bodies on the coal conveyor belt in time, so as to carry out early warning, sorting and control to reduce the probability of accidents. Aiming at the problems of large amount of calculation parameters and low classification accuracy in most classification networks, a classification network of coal belt foreign bodies based on residual network is proposed. The network uses multiple small convolution layers instead of the 7×7 convolution of the first layer to enhance the capture ability of local features and adds BN layer and ReLU activation function to make the network converge faster and enhance the nonlinear ability of the network. In the residual block, the depthwise separable convolution is used instead of the ordinary convolution, which greatly reduces the parameter quantity and calculation amount of the network and speeds up the model inference. After adding the CBAM attention mechanism to the convolution layer in the residual block, the network′s ability to learn channel features and spatial features is enhanced, the influence of useless background information on the model is weakened, and the attention is focused on the coal belt area. The deep features are fused with some shallow features to improve the recognition rate of small target foreign bodies such as anchors. The accuracy of the network on the self-built mine data set reached 91.4%, which was 4.7% higher than that of the improved network. The recall rate reached 91.2%, which was 5.8% higher than that of the pre-improved network. The calculation amount was reduced by 20%, and the number of parameters was reduced by 31%. The results show that the constructed network has higher accuracy, lighter weight, faster training speed and stronger real-time performance.

Abstract:To address the challenges of accurately locating target regions and identifying fine-grained features in fine-grained image classification, we propose a fine-grained image classification method based on an improved multi-scale deformable convolution (MMAL). Firstly, by leveraging the variable receptive field principle of deformable convolution, our method dynamically adapts to different scales and shapes of target regions in sample images, enhancing the network′s ability to perceive the position of these regions. Subsequently, we utilize the Grad-CAM gradient backpropagation technique to generate network attention heatmaps, which reduces the interference from background noise and achieves precise localization of the image target regions. Finally, we introduce a positionaware spatial attention module that integrates coordinate positions and dual-scale spatial information, significantly improving the network′s capability to extract fine-grained features of the target regions. Experimental results demonstrate that, compared to baseline methods, our approach achieves improvements of 1.4%, 1.5%, and 1.9% in classification accuracy on the CUB-200-2011, Stanford Car, and FGVC-Aircraft datasets, respectively, validating the effectiveness of the proposed method.

Abstract:Aiming at the problems of extracting the fault features of reciprocating compressors under the strong noise interference environment and the over-reliance on the a priori knowledge in the parameter setting of the VMD algorithm, a reciprocating compressor fault diagnosis method that integrates the VMD algorithm of trend evaluation and MSSE2D is proposed. In this paper, the valve fault data of reciprocating compressor is selected as the research object, firstly, the signal is analyzed and processed in depth by applying the VMD technique optimized by spectral trend evaluation, then the MSSE2D is used for analysis and calculation, and finally the support vector machine is used for validation test, and the experimental results show that this method can extract the fault characteristics of the reciprocating compressor valve efficiently compared with other methods, and it can quickly and accurately diagnose the various states of the valve, the results show that this method can effectively extract the fault characteristics of reciprocating compressor valves compared with other methods, and can quickly and accurately diagnose and distinguish between various states of valves.

Abstract:In order to improve the speed tracking performance and disturbance rejection capability of a permanent magnet synchronous motor (PMSM) under load variations, parameter perturbations, and other uncertainties, a control method based on active disturbance rejection control (ADRC) combined with precision feedback linearization control(FL) is proposed. Firstly, the precision feedback linearization method is employed to decouple the PMSM model into independent current and speed subsystems through coordinate transformation and state feedback. An ADRC is designed for the speed loop to estimate and compensate for the real-time uncertainties and disturbances during the motor operation, thereby improving the system′s speed tracking performance and disturbance rejection capability. Secondly, to address the issue of nonsmoothness at the inflection point of the nonlinear function in ADRC, the function is improved to reduce high-frequency buffeting in the system. Finally, the simulation results of two control methods based on FL and precise feedback FL-ADRC are compared in MATLAB/Simulink simulation platform. The results show that the speed tracking adjustment time of FL-ADRC is shortened by 58.33%, and the recovery time after sudden load is increased by 95.45%, which shows that this method can quickly track the changes of speed and load torque, and has good anti-jamming ability, which verifies the superiority of the proposed method.