Abstract:Aiming at the uncertainty of the offset angle of the receiving coil relative to the transmitting coil in the magnetically coupled resonant wireless power transfer (MCR-WPT) system, which leads to large fluctuations in the transmission efficiency of the system, a method for calculating the offset angle of the receiving coil in MCR-WPT system based on the reflection impedance is proposed. According to Kirchhoff’s voltage law, the relation between the reflection impedance of the transmitting coil and the mutual inductance of two coils is established, then the relation equation of the mutual inductance and the coil offset angle is derived according to the mutual inductance formula of two coils. Therefore, the expression for the reflection impedance and the offset angle is derived. The coils model and peripheral circuits are built using Maxwell and Simplorer simulation software for co-simulation. Meanwhile, a MCR-WPT experimental system is built, and multiple sets of experiments are conducted under different coil distances. The simulation results show that the mean absolute error between the set angle of two coils and the calculated angle is 2. 57°. The experimental results show that the aforementioned mean absolute error is 2. 94°. The offset angle of the receiving coil of the short-distance wireless power transmission system can be calculated accurately through the above method.

Abstract:The temperature drift in Hall current sensors can affect the accuracy of the sensor, especially in extreme high and low temperatures, which limits the application of Hall sensors. This article designs and implements a wide temperature range temperature drift compensation circuit suitable for Hall current sensors to address this issue. The temperature drift compensation circuit compensates for the temperature drift error of the Hall element by combining gain compensation with bandgap compensation, and using a load resistor that is consistent with the shape and material of the Hall element in the Hall voltage amplification circuit. At the same time, using the high and low temperature compensation current generated by the bandgap reference circuit to achieve temperature compensation for the tail current of the amplifier circuit, the Hall current sensor can maintain sensitivity stability over a wider temperature range. Adopting GF0. 18 μm BiCMOS process, simulation verification shows that under a 5 V power supply voltage, the circuit has a wide temperature range of -40 ℃ ~ 140 ℃ , a sensitivity temperature drift error of less than 0. 3%, and a temperature drift coefficient of 35 ppm/ ℃ . Compared to other temperature compensation designs, this design achieves compensation for high-order temperature errors of Hall sensors, resulting in a wider operating temperature range and smaller temperature drift errors. It does not require additional digital processing circuits and has high engineering application value.

Abstract:With the unique advantage of passive and self-stability levitation, high-temperature superconducting maglev is regarded as a promising element for a wide range of applications such as high-speed bearings, flywheel energy storage system and rail transportation. In order to investigate the levitation behavior of high-temperature superconducting maglev system, a high-temperature superconducting maglev force three-dimensional measurement system is developed, which is suitable for measuring the levitation behavior of maglev bearings. Firstly, based on the triaxial linear module with servo control and triaxial force sensor measurement unit, the hardware of measurement system is designed. Also, the data acquisition unit is developed in LabVIEW so as to realize the continuous sampling and imaging of data. Secondly, the measurement precision of high-temperature superconducting maglev force three-dimensional measurement system is validated under the constant load and permanent magnet levitation system. Finally, a finite element model of the superconductor-permanent magnet levitation system is established. Then, the levitation force behaviors of high-temperature superconductor were simulated and measured under field cooling and zero field cooling conditions. The comparison of the simulation data with the measured data shows that a well agreement between the two can be achieved. The results show that this measurement system has the characteristics of high measurement precision ( 0. 5%), high location accuracy in three-dimensional space ( ± 0. 02 mm) and synchronous measurement of three-dimensional force ( maximum levitation force 500 N, guidance force 200 N) at movement of the measured specimens.

Abstract:With the future wireless communication demand growth, communication systems need to apply more frequency bands and standards. Aiming at the shortcomings of the narrow operating bandwidth of reconfigurable power amplifiers in each mode, this paper proposes a design method to expand the operating bandwidth of reconfigurable power amplifiers based on the simplified real-frequency technique and reconfigurable theory. By incorporating the broadband design method of simplified real-frequency method in the reconfigurable theory, a new error function is added in the design process to discriminate the reconfigurable circuit structure in variable modes, then realize the reconfigurable broadband power amplifier design. In order to verify the effectiveness of the method and meet the actual design specifications, a frequency-switchable wideband reconfigurable power amplifier for GSM network and LTE network is designed and fabricated by using LDMOS transistors independently developed by the Institute of Microelectronics, Chinese Academy of Sciences. The test results show that the reconfigurable power amplifier can operate in the frequency bands of 0. 6 ~ 1. 1 GHz and 1. 1 ~ 1. 6 GHz in different modes, respectively, with a saturated output power of more than 40 dBm, and a drain efficiency (DE) of between 50% and 60%. Therefore, the design method proposed in this paper can reduce the design difficulty of reconfigurable broadband power amplifiers, better utilize the transistor performance, and reduce the cost, which is of great significance for application in the design of RF circuits of practical base stations.

Abstract:In order to reduce multipath loss and anti-polarization mismatch, and meet the multi-band needs of wireless devices, a dualband circularly polarized antenna loaded with polarization torsional artificial magnetic conductors is designed. The polarization torsion artificial magnetic conductor adopts a double-layer structure to increase the phase response, uses a rectangular ring to extend the current path, and adds rounded corners and truncated rectangular patches to the rectangular ring which can cause surface impedance imbalance, and achieve efficient polarization conversion in the 2. 45 GHz and 5. 8 GHz bands. A polarization torsion artificial magnetic conductor is applied under the dual-band monopole antenna. By using the 90° polarization rotation effect achieve left-handed circular polarization in low-frequency bands and right-handed circular polarization in high-frequency frequencies. The simulation and experimental results show that the working bandwidth of the designed antenna is 2. 2 ~ 2. 58 GHz and 3. 5 ~ 6 GHz, the 3 dB axial ratio bandwidth is 2. 3 ~ 2. 56 GHz and 5. 6~ 6. 2 GHz, respectively, and the peak gain is 16. 8 and 7. 5 dBic, respectively. Experimental results confirm that the use of polarization torsion artificial magnetic conductors can reduce the complexity of dual-frequency circularly polarized antenna.

Abstract:This article explores the microwave non-destructive testing technology for slit defects in medium to long distance pressure pipes. The characteristics of microwave transmission in bent pressure pipes and methods for slits detection were theoretically analyzed. The feasibility of detecting slit defects in straight and bent pipes was verified by simulating the transmission of microwave in medium to long distance pipes and the changes in microwave mode after passing through bend. Straight and bent pipes with a length of more than 5 m and a diameter of 8 cm were constructed and microwaves were introduced into the pipes using a vector network analyzer and a spiral antenna to detect axial, circumferential, and oblique slits, and the amplitude of the S11 parameter was varied in the case of slitless, slit, and even multiple slits to determine the location of the slits. For axial, circumferential, and oblique slits detection in straight pipes, the errors are 0. 047%, 0. 018%, and 0. 298% of the total pipes length, respectively. For axial, circumferential, and oblique slits detection in bent pipes, the errors are 0. 074%, 0. 065%, and 0. 203% of the total pipes length, respectively. This indicates that an effective remote detection method for medium to long distance pipes slits has been established.

Abstract:In order to meet the needs of modern wireless communication systems to the direction of broadband, integration, low cost and high performance. Based on the theory of collaborative design, a wideband filter power amplifier with integrated filtering characteristics is designed by using the self-developed RF-LDMOS device. A “ T-type” prematching network is designed to enlarge the transistors impedance and a wide-band filtering network is formed by integrating the load coupling microstrip lines. The continuous wave test results show that in the frequency band of 1. 2~ 2. 6 GHz, the saturation output power is more than 40 dBm, the drain efficiency is greater than 45%, the power gain is about 11. 5 dB, and the second harmonic inhibition ability is - 62 dBc. When the average output power is 32. 5 dBm, ACPR is better than - 37 dBc in the range of frequency bands. In this design, the power amplifier has the function of amplification and filtering, and the relative bandwidth is extended to 74% while the circuit integration is improved, which is in line with the needs of today’s wireless communication system.

Abstract:Currently, there are lots of technical difficulties in the measurement of transient electromagnetic pulses. In order to solve the problem of narrow dynamic range of transient electric field measurement, an optical fiber electric field sensor that can adaptively adjust the measurement range was developed. Firstly, the monopole PCB antenna was modeled on the CST platform, and its electromagnetic characteristics and size effects were simulated. Then, the automatic gain control (AGC) circuit of the transmitter was designed, and the threshold of AGC circuit was adjustable. Quadratic interpolation was used to compensate for transmitter ’ s non-linearity. The transimpedance amplifier of the receiver were optimized to make its noise in the microvolt level. The sensor was calibrated and tested using the standard field method, and the uncertainty was quantitatively calculated. The test results show that the input dynamic range of the sensor reaches 54 dB, the average response time is less than 3 ns, the linear correlation is 0. 98, the sensitivity is 0. 025 V/ (kV· m -1 ), and the extended uncertainty is 2. 67. The sensor can meet the electromagnetic environment measurement of lightning pulses and partial discharge positioning.

Abstract:A planar near field measurement method based on unmanned aerial vehicles (UAVs) is proposed to address the difficulty of precise measurement of ultra large ground antennas with apertures exceeding 10 meters due to their extremely high cost. Firstly, simulation analysis was conducted on the main factors that affect the antenna field measurement results, such as drone scattering, positioning accuracy, and measurement field selection. The measurement area was determined to use multi rotation drones and highprecision flight control technology. Under the condition of high-performance drones with the best 20 mm flight control accuracy and 5 mm measurement accuracy, the RF link design of the antenna field measurement system was carried out, and two near to far field data transformation methods were provided, they can effectively solve the problem of unreliable measurement results caused by poor positioning accuracy of drones in higher frequency bands. The effectiveness of the method was verified through case analysis, providing a good solution for low-cost and accurate measurement of ultra large ground antennas.

Abstract:The standard orifice plate flowmeter faces limitations in accuracy and applicability in industrial applications, and to improve its performance, multi-hole orifice flowmeter is increasingly used. In order to study the influence of structural parameters of multi-hole orifice plate on the performance of orifice plate flowmeter, the multi-hole orifice plate with inner diameter of 40 mm, multi-hole orifice diameter ratio of 0. 4 and thickness of 3 mm is used as the research object, and the multi-hole orifice plate model with different number of holes and clearance rate is designed. Numerical methods are used to analyze the flow characteristics of the multi-hole orifice flowmeter. The results show that the pressure loss coefficient of the multi-hole orifice flowmeter decreases with the increase of the number of holes, while the discharge coefficient increases with the increase of the number of holes. The clearance rate has little effect on the pressure loss coefficient and discharge coefficient of multi-hole orifice flowmeter, but increasing the clearance rate in a certain range can effectively reduce the pressure recovery length.

Abstract:To address the issue of reconstruction networks in unsupervised defect detection failing to preserve detailed information of normal regions while simultaneously suppressing abnormal reconstructions, a defect detection method that combines self-attention and defect highlighting is proposed. First, the discrete wavelet transform ( DWT ) is introduced in the reconstruction network for downsampling, and the inverse discrete wavelet transform ( IDWT) is used for upsampling. Compared to traditional reconstruction networks, this method reduces the loss of detail information and performs frequency decomposition on features. Then self-attention modules are added into the skip connections to re-encode the features, enabling the features to focus on the details of the normal region. Additionally, a defect region highlighting module is designed, which utilizes features from normal samples to construct a feature library. By comparing the features extracted from the test image with the features in the library, an abnormal map is obtained. Finally, the abnormal map is combined with the reconstruction residual map to improve the results of defect localization. The proposed method is tested on the industrial defect detection dataset MVTec AD and achieved 99. 3% area under the receiver operating characteristic curve (AUROC) at the image level and 98. 3% at the pixel level, demonstrating high detection accuracy and robustness in unsupervised defect detection.

Abstract:In response to the micro-newton thrust measurement requirements for space gravitational wave detection missions, a singlewire torsion micro-newton thrust measurement system based on closed-loop control has been developed. Based on the principle of torsional pendulum micro-newton thrust measurement, the detailed design scheme of the micro-newton thrust measurement device is given, and the twisted wire structure, angular displacement measurement, electromagnetic calibration force, and closed-loop control system are mainly analyzed. The micro-newton thrust measuring device is calibrated through experimental measurement, and finally the error analysis is carried out. The test results show that the torsion wire stiffness is about 0. 003 25 N·m/ rad in the open-loop state, and the error with the theoretical value is about 4. 0%; the thrust measurement range in the closed-loop state is 0. 1 ~ 246. 0 μN, and the minimum resolution is better than 0. 1 μN, which is relatively uncertainty error is 1. 174%, which meets the measurement requirements for the thrust measurement range, resolution and accuracy of micro thrusters.

Abstract:True random number generators have broad application prospects in the fields of hardware and information security. In order to improve the throughput and reduce the hardware overhead of true random number generator, an autonomous Boolean network is constructed with coupled basic logic units as the entropy source. A first-order high-frequency oscillation loop is used to enhance the network refresh frequency and multi-level nonlinear amplification, thereby obtaining a high entropy chaotic signal. Combining a postprocessing circuit composed of DFF and XOR, a true random number generator is designed and implemented on an FPGA platform. The sampled output data is extracted using the ChipScope online tool, then NIST SP800-22 and SP800-90B randomness tests are performed on the data, and their performance such as offset, autocorrelation, and maximum Lyapunov exponent are evaluated. The results show that the proposed true random number generator can generate a random number sequence with an entropy value of 0. 994 847 bit / sample at a throughput rate of 600 Mbit / s, and is of low offset and no autocorrelation, and low hardware overhead.

Abstract:Aiming at the short transmission path and low measurement accuracy of the small ultrasonic wind measurement array, four ultrasonic transducers integrating transceivers are designed, arranged in a square, with the probes placed vertically downwards, and the ultrasonic signals are reflected multiple times between the wind measurement areas. The wind measurement is completed, and the volume of the one-time reflection ultrasonic wind measurement array is reduced by 20%. Based on the built model, an adaptive wind measurement algorithm is proposed to reduce the influence of shadow effect on wind measurement accuracy. Using the SpaceClaim Design Modeler software, the wind array and wind tunnel model were established, the grid was divided by Fluent Meshing, and the wind speed and direction in the simulated wind tunnel were changed using Fluent software, and the designed small ultrasonic wind array and adaptive weight were verified. The superiority of the wind measurement algorithm in weakening the shadow effect and improving the measurement accuracy. The experimental results show that compared with the traditional wind measurement algorithm, the speed measurement accuracy of the adaptive weight wind measurement algorithm is increased by 12. 44%, and the angle measurement accuracy is increased by 10. 40%.

Abstract:Aiming at the problem that the detection of the magnetic anomaly signal is usually interfered by the geomagnetic noise, a modified OBF magnetic target detection method based on the Karhuen-Loeve expansion is proposed, after the research of the magnetic interference compensation. Firstly, the ellipsoid fitting model of the magnetic interference is built, and the compensation of the interference is realized by solving the coefficients of the fitting model. The compensation ability is about 30 nT. Then, the OBF detection algorithm is modified by the Karhuen-Loeve expansion to improve the detectivity under the colored background noise. Finally, the experimental research on the detection of magnetic targets by an unmanned aerial vehicle is carried out. The experimental results show that the modified OBF detection algorithm presents better detectivity, especially for the detection of low SNR targets, and the SNR gain is improved by about 2 dB.

Abstract:Environment-aware millimeter wave radar (EMWR) is one of the most important sensors in autonomous driving system of car. With the rapid increase of the number of cars installed with millimeter-wave radar, the problem of mutual interference between radar is increasingly prominent, which seriously affects the road safety of cars. The interference types of millimeter wave radar can be divided into two types: Same frequency interference (SFI) and different frequency interference (DFI). Aiming at the influence of SFI and DFI of millimeter-wave radar on radar false alarm and leakage alarm, this paper proposes the anti-interference algorithm of environmental awareness millimeter-wave radar. The random initial frequency and random starting time are used to reduce the probability of signal collision to inhibit the interference. The Hilbert transformation is used to extract the envelope and gets the location of the interference signal. Furthermore, Lagrange interpolation is used to improve the accuracy of the location in DFI to reduce the effective of different frequency interference on EMWRs. The simulation results show that the above method can achieve an ideal effect.

Abstract:To address the issue of low weighing accuracy caused by the tilt angle of the scale body during the use of electronic hand pallet truck forklift scales, a multi-directional variable-angle tilt simulation and measurement experiment platform was constructed to simulate the different tilt states of the forklift scale. By analyzing the weight and tilt angle change data measured by the experiment platform, a compensation algorithm model for the weighing error of the electronic forklift scale under the tilt state was established, and the model parameters were identified using both the least squares method and the multilayer perceptron non-linear regression. The compensated forklift scale with a range of 1 500 kg and sensitivity of 0. 02 kg was tested under tilt experiment, and the results showed that within a ground tilt angle not exceeding 1. 5°, the instrument achieves an accuracy of 3 000 divisions with a maximum permissible error (MPE) of less than 0. 75 kg in any tilt direction.

Abstract:By introducing the time slotted channel hopping mode and IPv6 protocol, the industrial Internet of Things 6TiSCH exhibits significant advantages in supporting massive node connectivity, deterministic and reliable transmission, and low-power operation. While the defined 6TiSCH operation sublayer provides a standardized scheduling execution process for user-defined scheduling strategies, it still faces the challenge of personalized design for upper-layer scheduling functions ( i. e. , decision entities ) in different application scenarios. In this paper, the control theory is introduced into the scheduling application of the deterministic industrial wireless network, the communication resource control between distributed nodes is generalized as a closed-loop control problem, the principles of resource scheduling in 6TiSCH network are elucidated by combing the classic PID control concept. A distributed PID scheduling algorithm for 6TiSCH network, which is specifically tailored for burst traffic scenarios, is designed, and an upper-layer scheduling function that is friendly to burst traffic is constructed. The experiments conducted on an OpenMote-B based 6TiSCH network platform demonstrate that the proposed method enables adaptive scheduling of communication resources between nodes with dynamically changing traffic demands, and presents rapid responses especially for various types of burst traffic.

Abstract:The firing zone images of rotary kilns contain rich flame information and accurate combustion state recognition are the premise of optimal control for the rotary kilns. The working conditions can be quickly recognized, and the automation level of the rotary kilns can be improved through the convolutional neural network-based methods, but there are problems with large network size and high computational resources required. Therefore, a working condition recognition method based on lightweight network and knowledge distillation was proposed in this paper. The teacher model and the student model were improved by introducing the collaborator differential layer after the convolutional layer of the network. The improved lightweight network MobilenetV2 was used as the backbone network of the student model while the improved Resnet50 was used as the backbone network of the teacher model. The rich classification label information contained in the teacher model was transferred to the student model by constructing the mixed distillation loss function and the student model obtained by distillation training was used as the working condition recognition model to improve the recognition accuracy of the high similar rotary kiln flame images. The experimental results show that the overall recognition accuracy of the improved student model is increased by 3. 33% compared with the original model, and the recognition accuracy of the three working conditions in the test set reaches 93%, 99%, 90% respectively. The accuracy and network size are better than those of other mainstream networks, and the requirements of real time and low cost in actual production process are met.

Abstract:In industrial production, intelligent detection of product defects is crucial. Field-programmable gate arrays ( FPGAs) are embedded devices with features such as high arithmetic power and low power consumption that enable small convolutional neural networks to be deployed in them. In this paper, a set of improved YOLOv2 target detection algorithm is designed based on Xilinx Zynq series FPGAs, and a reordering layer is added to the model framework to complete the detection of surface defects on aluminum sheets by parallel computing processing of the slice map before reorganisation. The algorithm is designed at a high level ( HLS), then RTL converted and IP cores are packaged and imported into the project to complete the SoC design. Generate bitstream files through comprehensive layout and wiring, import them into PYNQ images, and complete industrial defect detection on the surface of aluminum sheets. The experimental results show that this system can accurately detect defects and reduce power consumption to 2. 494 W.

Abstract:In order to enhance the efficiency of the federating for disjoint segments method in multi-robot wireless sensor networks (WSNs), this paper presents a solution for optimizing the federating for disjoint segments in WSNs based on partition energy balance, along with an approximate algorithm to address this optimization problem. Firstly, this research incorporates relevant model assumptions and symbol definitions to propose a sophisticated connection mechanism. Drawing inspiration from divide-and-conquer and local priority approaches, the proposed mechanism ensures energy balance within the network. Furthermore, a round-robin iterative process is introduced to iteratively refine the method, which is then precisely formulated using mathematical representations. Secondly, a relay deployment algorithm is designed using heuristic algorithms, taking into account the energy balance of the robots. Finally, through comparative experiments conducted with existing methods, the results demonstrate the effectiveness of the proposed approach. Notably, the approach successfully conserves connection costs while maintaining energy balance and significantly improves the efficiency of the federating for disjoint segments method in WSNs and demonstrates its capability to prolong the overall lifespan of the network. Overall, this research contributes a comprehensive solution for optimizing the federating for disjoint segments method in multi-robot WSNs.

Abstract:In order to solve the problem that the label is not accurate and the background interference makes it difficult to obtain the predefined local area in the weak supervision vehicle re-identification method. A vehicle re-identification network based on weak supervised attention and knowledge sharing is proposed. In the weak-supervised attention module (WAM), the weak-supervised method is used to generate the vehicle component mask, and the component channel alignment step enables the module to perform adaptive feature alignment under complex background. Aiming at the problem that the mask of WAM module is unstable due to the low accuracy of labels in weak supervision method, a knowledge sharing module is constructed in local branches. The module uses migration learning to extract vehicle component features from WAM module, and performs multi-scale component feature extraction to prevent unstable vehicle component mask generation. Through experiments, mAP, CMC@ 1 and CMC@ 5 reached 82. 12%, 98. 50% and 99. 12%, respectively, which are better than the existing methods and verify the effectiveness of this method.

Abstract:Currently, the supervised models trained on synthetic paired datasets have weak generalization ability and perform poorly in diverse real underwater environments. Although unsupervised models are not dependent on paired datasets, the lack of feature information may result in the generated images with poor visual quality. Therefore, with the architecture of cyclic generation adversarial networks, the underwater image enhancement method of multi-feature selection and bidirectional residual fusion is proposed. On one hand, a multi-feature selection module based on mixed attention is designed to select multiple features of underwater images. Furthermore, the bidirectional residual fusion is used to optimize traditional U-shaped skip connection, which realizes high-efficiency expression of image features and effectively restores the texture and color of underwater images. In addition, mixed attention is introduced and content-aware loss and style-aware loss are proposed in the discriminator to ensure that the enhanced image is consistent with the clear image in terms of global content, local texture, and style features. The PSNR of the proposed model is improved by 6% and 2%, respectively, compared with the existing unsupervised and supervised models. Additionally, SSIM is improved by 4% and 3%, respectively. With a significant enhancement effect on underwater images, the proposed method demonstrates superiority over other existing methods in terms of color fidelity and saturation.

Abstract:In view of the harsh environment with uncertain external disturbances such as irregular airflow while in high overload and high dynamics motion state during the in-flight alignment of guidance projectile, an adaptive reduced dimension fifth-order cubature Kalman filter based on sequential quadratic programming ( SRD5-CKF) is proposed in this paper, which uses the projection statistics ( PS) principle to detect the innovation vector in the sliding window. If the innovation vector has outlier, it will be reweighted to adjust the measurement noise covariance matrix in real-time. At the same time, the sequential quadratic programming method with fast solving speed and good convergence is used to solve the adaptive factor matrix, so as to realize the optimal estimation of the measurement noise covariance matrix under complex interference. Simulation experiments have shown that the SRD5-CKF proposed in this paper has a faster convergence speed and higher convergence accuracy under complex disturbances such as short-term strong interference and instantaneous impact. The alignment accuracy of the upward misalignment angle reaches 0. 25°, and the alignment accuracy of the eastward and northward misalignment angles reaches 0. 1°, which can meet the requirements of rapid in-flight alignment of guidance projectile.

Abstract:To solve the problem of underdetermined blind source separation (UBSS), an UBSS method based on wavelet packet hybrid optimization is presented. The method, adopting wavelet packet transform, decomposes the observed signal, expands the dimension of the observed signal, and removes redundant signal components with the cross-correlation value, transforming the problem of UBSS. Then, with the singular value decomposition method under the Bayesian information criterion, the number of source signals is estimated, and the signal dimension is reduced through the whitening process. At last, the spiral bubble net hunting behavior and levy flight strategy in the whale optimization algorithm (WOA) are introduced to improve the gray wolf optimization algorithm, and the improved hybrid gray wolf optimization algorithm is integrated with the independent component analysis algorithm to separate the reconstructed positive definite whitening signals, and rewarding separation performance is achieved. The performance of the algorithm is tested by simulation experiments, and the results show the feasibility and effectiveness of the proposed method.

Abstract:The quality of electrode is extremely important for the display effect of liquid crystal panel. To solve the problem of difficult detection due to the variety of electrode defects, small scale and complex background, an electrode defect detection method for liquid crystal panels is proposed based on improved YOLOv7 algorithm. Firstly, the CBAM attention module was embedded into the YOLOv7 backbone network to suppress background information interference and strengthen defect features. Secondly, the feature information of the shallow networks and deep networks was fused by cross-level connection operation. Then, the C2f module was integrated into the feature pyramid network to lightweight the model and improve the training speed. Finally, the WIoU was used to replace the loss function of the YOLOv7 model to reduce the harmful gradient caused by low quality labeling and improve the defect location performance. By a customized electrode defect dataset, the results showed that the proposed algorithm was able to achieve an average detection accuracy of 67. 8% for large scratch, scratch, shell and dirt on electrodes with a per-sheet detection time of 5. 6 ms.

Abstract:To reduce the impact of thermal errors on CNC machine tools, improve the machining accuracy of workpieces, and solve the problem of poor thermal error prediction accuracy under different working conditions. The thermal error measurement experiment of the CNC machine tool feed system is conducted under working conditions of a feed speed of 10 m/ min and an ambient temperature of 20°. The Pelican optimization algorithm is used to optimize the neural network, determine the optimal weight and threshold of the BP neural network, and the thermal error of the feed system prediction model of POA-BP is established. The experiment is compared and analyzed with traditional BP neural network, GA-BP neural network and the SCN random configuration network. The results show that the average relative error of traditional BP neural network prediction is 12. 23%, the average relative error of GA-BP neural network is 11. 5%, the average relative error of SCN prediction model is 12. 71%, and the average relative error of POA-BP prediction model is 9. 93%, which improves the accuracy. Conclusion: The neural network improved by the proposed Pelican optimization algorithm has strong effectiveness and accuracy in thermal error prediction, which can improve the accuracy of feed motion and provide theoretical guidance for the realization of thermal error compensation.