Abstract:To address the challenge of real-time and precise identification of distributed forces during flexible contact in cooperative target de-tumbling and capture processes, this paper proposes a nested recursive distributed force identification method based on an array pressure sensor. This method achieves real-time characterization of the contact area and distribution characteristics of distributed forces under spatiotemporal coupling conditions. First, a distributed contact force model is established, and its power-exponential function properties are utilized to perform a logarithmic space transformation. A shape function is then introduced to decouple the spatiotemporal characteristics of the distributed force. In the spatial domain of the nested inner layer, an improved weight equation based on the Sigmoid function is proposed, and a weighted least squares method using the S-shaped weight function ( S-WLS) is developed to refine the geometric parameters of the contact area obtained through a neighboring response region search algorithm. In the time domain of the nested outer layer, both a forgetting factor and a weight factor are introduced to construct a recursive formula for dynamically solving the standard equation of distributed force. Accurate identification of the distributed force characteristic parameters is achieved through exponential transformation. Finally, a distributed force acquisition system based on an array pressure sensor is designed, and an equivalent microgravity collision platform is developed to conduct oblique collision experiments. Experimental results demonstrate that the proposed method significantly outperforms the traditional weighted least squares ( WLS ) method in identifying distributed force characteristic parameters, with a relative error range of only ±8. 8% . Furthermore, the Hazen scoring method is applied to perform a normality test on the relative error at a 95% confidence level, confirming the accuracy and effectiveness of the proposed method. This work provides a theoretical foundation and technical solution for the accurate prediction of dynamic distributed forces in space debris detumbling and capture applications.

Abstract:Wind power generation, aerospace, meteorology, and other key industries have an urgent need for high-precision and highreliability detection of wind speed and direction. Compared with the traditional mechanical, ultrasonic convective, and ultrasonic reflective wind speed and direction sensors, the ultrasonic resonant wind speed and direction sensing solution in this article has significant advantages of small size, high accuracy, and no mechanical abrasion. To address the problem of accuracy shift of ultrasonic resonance wind speed and direction sensor under complicated operating conditions, the theoretical model of the error source for the ultrasonic resonance wind speed and direction sensor is formulated. Different temperature environment tests are carried out to analyze the performance of the sensors. The correlation relationship between the working environment temperature and the resonance frequency point of the sensors is constructed, which achieves the temperature compensation of the sensors based on the frequency of the resonance point under different environmental conditions. By establishing a sensor wind tunnel test system, the wind tunnel test of the sensors under different temperature conditions is carried out. Test results show that the wind speed measurement accuracy of the proposed ultrasonic resonance wind speed and direction sensor is significantly improved after the temperature compensation of the resonance state. After the compensation, the accuracy of wind speed can reach ±0. 3 m/ s in the wind speed range of less than 15 m/ s, and ±2. 3% in the wind speed range of 15~ 50 m/ s. The sensor′s measurement accuracy significantly decreases with temperature drift, reducing the measurement error to 2. 30% at an ambient temperature of 17. 1℃ and 2. 09% at an ambient temperature of 29. 2℃ , improving the accuracy by more than 40% compared with the traditional ultrasonic convection / reflection type. In conclusion, it provides support for more effective wind speed and direction measurements for key areas such as improving the power generation efficiency of wind farm turbines and improving the accuracy of elemental measurements in the meteorological field.

Abstract:To address the problem of deviation in calculating the azimuth angle of the drilling tool caused by the measurement error of the magnetic in-place magnetometer, a magnetic inertial beluga whale optimization (MIBWO) based error parameter identification method for the magnetic in-place magnetometer is proposed. Firstly, the output error model of the magnetometer is formulated, the objective function is constructed, and the constraint conditions are established by utilizing the connection between the magnetometer data and the acceleration and gyroscope data. Then, in view of the insufficiency of the beluga whale optimization (BWO) in optimizing the magnetic error parameters in the while-drilling environment, a dimensional adaptive small-hole imaging opposite learning strategy is proposed, three adaptive learning factors are designed to adjust the step size and direction of the individual in searching the magnetic error parameters, and the golden sine search strategy is introduced to improve BWO to obtain MIBWO. Finally, evaluation is carried out through simulation and actual drilling experiments. The results show that the error parameters of the magnetometer identified by the MIBWO algorithm have a significant effect on the error compensation of the magnetometer, and the average value of the absolute error of the calculated azimuth angle is reduced from 5. 1° to 0. 9°. This method can effectively improve the measurement accuracy of the magnetometer.

Abstract:Extrinsic calibration is a key prerequisite for the data fusion of LiDAR and camera. However, the current calibration methods still face several challenges such as dependence on prior conditions, single feature constraints, and low calibration accuracy. To address these issues, this paper proposes a two-stage automatic extrinsic calibration method that integrates the mutual information and multifeature constraints. It combines the advantages of mutual information and multi-feature constraints methods. Firstly, by constructing a mutual information maximization model, the first stage is the coarse extrinsic calibration based on mutual information. This stage obtains the initial extrinsic calibration parameters according to the correlation between LiDAR reflectance and camera grayscale values, not depending on the initial values, set values, or any other prior conditions. Additionally, we design an adaptive gradient algorithm to refine the initial values of the extrinsic parameters. Secondly, the following stage involves the fine calibration of extrinsic parameters with multi-feature constraints, which uses the multiply constraints including the point-to-line, point-to-plane and line-to-plane, to optimize the initial extrinsic parameters obtained from the first stage. Also, the iterative closest point (ICP) algorithm is utilized to minimize the reprojection error between the 3D geometric features of the point cloud and the 2D geometric features of the image. Finally, we conducted the extrinsic calibration experiments in both indoor and outdoor challenging environments using a special-designed hollow circular calibration board, which simultaneously possesses the multi-feature constraints: point, line, and plane. The experimental results proved that the proposed calibration method can automatically and precisely achieve the extrinsic parameters of LiDAR and camera not depending on initial values. Additionally, the method exhibits higher accuracy and stability.

Abstract:In complex observation environments, the number of available GNSS satellites is significantly reduced, and the observed signals are susceptible to multidimensional outliers, resulting in challenges in outlier detection and a marked deterioration in positioning performance. To address this issue, this paper proposes a robust filtering method for quasi-accurate detection based on adaptive classification using Random Forests. Initially, we construct a Random Forest classifier utilizing multidimensional data characteristics derived from dynamic observational data collected in urban settings. This classifier facilitates the adaptive categorization of observational data into two principal classes: quasi-accurate observations and non-quasi-accurate observations. Following this classification, a novel robust positioning method for GNSS quasi-accurate detection is developed. This method leverages the quasi-accurate scores obtained during the classification process to determine the random model for quasi-accurate solutions, substituting the standardized true errors for conventional least-squares residuals in the subsequent IGG3 robustness test. Taking pseudorange differential positioning as an example, we evaluate the effectiveness of the proposed method across two positioning scenarios characterized by different magnitudes of observational errors. Experimental results demonstrate that, in comparison to traditional IGG3 and quasi-accurate detection methods, the proposed approach significantly enhances positioning performance, with accuracy improvements ranging from 16% to 51% across various testing scenarios. The random forest classification model proposed in this paper effectively improves the identification accuracy of quasiaccurate observations, and therefore can effectively improve the positioning performance in complex observation environments.

Abstract:This paper investigates the impact of fiber core diameter on thin-core fiber grating sensing applications, addressing challenges such as low optical imaging resolution, high light energy loss, and poor imaging quality caused by core diameter mismatches in access fibers. Using a dual-grating quadratic dispersion demodulation optical system as a model, a comparative analysis was conducted between conventional single-mode fibers and thin-core fibers with varying core diameters. The study examined the effects on imaging and demodulation accuracy by simulating changes in imaging spot resolution, beam divergence angle, receiving surface irradiance, and the modulation transfer function (MTF) curve at the cutoff frequency. The computational results reveal that reducing the fiber core diameter from 8 μm to 2 μm improves resolution from 0. 8 nm to 0. 1 nm and increases irradiance on the receiving surface. The optical energy utilization on the receiving surface reaches its maximum of 74. 98% when the core diameter is below 4. 8 μm, and the exposure time decreases to 36% of the conventional state. Additionally, the MTF curves at the cutoff frequency show a parabolic correlation with core diameter, with images approaching the diffraction limit when the core diameter is 4. 2 μm. This study establishes and validates the correlation between fiber core diameter and spectral demodulation imaging indices, offering valuable insights and references for fiber grating sensing demodulation technologies with different core diameters.

Abstract:To address the trade-off between bandwidth, sensitivity, and anti-interference performance of the Rogowski coil, a combined PCB differential Rogowski coil with high bandwidth, enhanced anti-interference capability, and improved sensitivity was proposed. The measurement principles and equivalent lumped parameter circuit model of the combined PCB Rogowski coil were analyzed to identify key factors influencing measurement bandwidth. A differential design was implemented to enhance anti-interference performance, which was further analyzed in detail. Finite element simulation software was used to evaluate interference in various electromagnetic environments, confirming the high anti-interference characteristics of the loop-combined PCB differential Rogowski coil. Experimental verification showed that the designed sensor achieved a sensitivity of 3. 46 mV/ A and a linearity of 0. 56% when measuring power frequency currents ranging from 1 A to 100 A. The maximum relative error during 10 A to 100 A measurements was 1. 73% . In anti-interference tests, the sensor exhibited a maximum relative error variation of - 0. 39% , demonstrating strong anti-interference capability. Comparative experiments revealed that the proposed coil effectively resists electric and magnetic field interference while expanding the measurement bandwidth without compromising sensitivity. This advancement broadens the application scenarios of Rogowski coils, enhancing their versatility and flexibility.

Abstract:Precision optical components are extensively used in critical fields such as astronomical observation, space exploration, and projection lithography objectives. These components are characterized by high surface quality, precise profile accuracy, and minimal surface damage. Polishing is the final step in the finishing process of optical elements, requiring an iterative “measurement-processing” cycle to achieve nano-level surface accuracy and sub-nano-level surface roughness. After off-line measurement of the surface shape, the workpiece must be re-installed and aligned on the machine tool twice to ensure precise positioning and posture, which is crucial for subsequent processing accuracy. Traditional contact-based alignment and positioning methods suffer from inefficiencies and risks of surface scratches. To address these issues, this paper proposes a non-contact in-place measurement system for vertical five-axis machine tools, utilizing a spectral confocal sensor. A mathematical model is developed for both the measurement system to obtain workpiece surface coordinates and for the calibration of the spectral confocal sensor on the machine tool. Additionally, a compensation calibration method based on a standard ball is introduced, using a spherical constraint equation. The feasibility and accuracy of this calibration method are validated through practical calibration experiments on a magnetorheological polishing machine tool. Finally, a workpiece position and pose measurement experiment is conducted, with the accuracy of the spectral confocal non-contact in-place measurement system verified against a Renishaw three-coordinate in-place measurement system. Experimental results show that the diameter error of the standard ball is less than 6 μm, and the workpiece positioning error is less than 10 μm, meeting the positioning accuracy requirements for magnetorheological polishing.

Abstract::To address the contact nonlinearity issues in the structure of large-aperture refractive space telescope, finite element methods were employed to perform contact nonlinear analysis and opto-mechanical integrated analysis, ensuring the telescope′s normal imaging performance in a wide temperature range. Firstly, a centering turning process was employed in the structural design of large-aperture refractive space telescope, using tangential contact methods to reduce lens stress levels. Subsequently, contact nonlinear analysis was conducted to evaluate the deformation of large-aperture refractive space telescope under self-weight and thermal loads. The analysis examined the stress levels of the system under static conditions, such as gravity and temperature rise, as well as dynamic conditions, including random vibration and impact. Next, Zernike polynomial fitting was performed using Sigfit software, with the results imported into optical software for opto-mechanical integrated analysis. This step assessed the degradation of optical performance under variations in static mechanical conditions. The analysis results indicated that under both static and dynamic conditions, the maximum contact stress of the optical component is 1. 83×10 7 Pa, with a safety margin better than 0. 82. The deformation of optical components under load led to changes in optical system parameters, resulting in a reduction of approximately 2. 97% in modulation transfer function(MTF), which remains within acceptable optical tolerances. Finally, the complete system underwent machining, assembly, and testing. The test results showed that the wavefront aberration of large-aperture refractive space telescope system is 0. 123λ(λ = 632. 8 nm), and field imaging test produced clear images, meeting the requirements for space-to-ground observation. This provides valuable references and guidance for the further optimization design and development of large-aperture refractive space telescope.

Abstract:The laser tracking testing platform is a spatial large-scale coordinate measuring instrument. The machining and assembly of the shaft system is the base for achieving high-precision measurement of the entire machine. To improve the pointing and positioning accuracy of the laser tracking platform, a laser tracking platform driven by an ultrasonic motor is developed, and a method for evaluating the pointing accuracy of the platform is proposed based on structural error modeling and analysis. Firstly, the platform′s topological structure is established based on the principles of multi-body system theory and coordinate transformation. Secondly, the shafting perpendicularity, coaxiality, feature position deviation, laser module assembly error, and the initial position of each static error are incorporated into the error transfer matrix. Meanwhile, considering the platform′s rotation, dynamic errors such as bearing vibrations and control errors are also incorporated. A well-comprehensive pointing error model of the platform is formulated. Based on the established model, numerical simulation experiments are carried out to quantitatively analyze the influence of each error, as well as the influence of the initial position of the error on the precision of pointing. Building upon the established model, numerical simulation experiments are performed to quantitatively assess the impact of each error on pointing accuracy and the effect of error initial positions. Finally, based on the simulation results, a laser tracking platform driven by ultrasonic motors is manufactured. Axis accuracy and pointing accuracy testing experiments are conducted. The experiments show that the platform has a wobbling of under 2. 2″ for both axes, with an angular measurement error of less than 1. 1″. At a distance of 6. 8 meters, the laser endpoint exhibits a repeat positioning error of less than 0. 20 mm in the Y direction and below 0. 42 mm in the Z direction. The experimental results validate the theoretical analysis process and demonstrate the effectiveness of the proposed method.

Abstract:The geometry and dimensions of atomic force microscopy (AFM) probe tips are critical parameters for precise measurement of surface micro-nanostructures and accurate characterization of local physicochemical properties. While conventional blind tip estimation methods based on mathematical morphology can evaluate tip geometry solely from scanning images, they typically provide upper-bound estimates rather than true tip dimensions and suffer from significant sensitivity to scanning noise, resulting in insufficient measurement accuracy. To overcome these limitations, this study proposes a robust convolutional neural network ( CNN) with an encoder-decoder architecture for stable and accurate AFM tip characterization. During supervised learning, a training dataset was generated by simulating scanning images of nanoparticle structures with varying radii and densities through mathematical morphology dilation operations, representing tips with predefined dimensions. The network parameters were optimized using mean absolute error as the loss function. Experimental results demonstrate that the CNN model achieves accurate tip radius predictions for scanning images when the tip dimensions fall within the training range. However, the model exhibits reduced accuracy for tip sizes outside the training distribution. Notably, the model′ s predictive capability is significantly enhanced through noise-augmented training data, enabling precise tip dimension estimation from noisy scanning images without requiring additional denoising procedures. Validation using actual AFM scanning images confirms the method′s effectiveness in practical applications. Furthermore, simulations and experimental data verify the method′s extensibility for processing tip-effect-distorted images.

Abstract:The relationship between the pointing error and the underlying structural parameters of Alt-Azimuth photoelectric telescopes, which are crucial instruments in astronomical observations and laser communications, has not been fully understood. To establish the connection between the underlying structural parameters and the pointing error, and to provide a theoretical basis for the design and compensation of instrumental errors, this paper first introduces the typical Alt-Azimuth structure and proposes the classification and definition of two-level error sources for pointing errors. A top-down all-link pointing error model, which extends from the underlying structural parameters to the system’s optical axis, is developed using error synthesis theory. The proposed model is then verified using open calibration data from the literature. Subsequently, a sensitivity analysis is conducted on nine processing parameters and nine design parameters within the proposed all-link pointing error model. The results indicate significant variations in the contributions of different underlying structural parameters to pointing errors. It is determined that these structural parameters should be designed differently according to their sensitivities. Furthermore, it is found that the structural parameters significantly influence the nonlinear effects of pointing errors. Finally, based on the proposed model, the pointing error of a 500 mm aperture laser communication Alt-Azimuth photoelectric telescope is compensated. The maximum values of the azimuth and pitch components of the pointing error are reduced from 218. 83″ and - 208. 66″ to 16. 45″ and 6. 89″, respectively, which is more than 12 times better. The all-link pointing error model proposed in this paper reveals the underlying structural factors of pointing errors and their mechanisms. It demonstrates the superiority of the model for error compensation and provides a theoretical basis for the design of instrumental errors and systematic error compensation of Alt-Azimuth photoelectric telescopes, which is of great significance.

Abstract:Electrodes are critical components of lithium-ion batteries, and during manufacturing processes like coating and rolling, their surfaces are susceptible to defects such as scratches and foil exposure. These defects can significantly impact the quality and service life of the batteries. Consequently, defect detection and control procedures for battery electrodes are essential steps in lithium-ion battery production. This article begins by outlining the production process of lithium-ion battery electrodes and analyzing the potential causes and types of surface defects that can occur during manufacturing. Next, it discusses the use of machine vision for surface defect identification, replacing manual labor with automated detection. The article reviews the principles, advantages, and limitations of traditional machine vision defect detection methods. It then delves into the application of deep learning for electrode surface defect detection, focusing on the principles and procedures involved. Particular attention is given to comparing one-stage and two-stage algorithms in target detection for lithium-ion battery electrode defect detection. Finally, the article predicts future developments in machine vision detection methods based on deep learning for surface defect detection of lithium-ion battery electrodes, offering valuable insights for researchers in this field. The advancement of electrode surface defect detection technology depends not only on hardware innovations, such as industrial cameras, but also on continuous optimization and innovation of software algorithms. The synergy between software and hardware can enhance detection accuracy, improve efficiency, reduce costs, and drive the lithium-ion battery production industry toward high-quality development.

Abstract:Interventional surgery is one of the primary methods for treating cardiovascular diseases. Current procedures mainly rely on 2D fluoroscopic images to guide surgeons, which cannot achieve three-dimensional visualization of interventional catheters during surgery, limiting surgical efficiency and safety. This paper addresses the clinical need for precise treatment in cardiovascular interventional surgery by proposing a method for reconstructing the spatial shape of interventional catheters based on the U-Net and epipolar geometry. First, the U-Net network is used to segment the catheter′s contours from biplane fluoroscopic images, and the catheter′s centerline is extracted using a skeletonization algorithm. Then, a stereo vision matching method based on epipolar geometry constraints is developed, where in the intersections of the epipolar lines and the catheter centerline are solved to determine the corresponding points in the biplane projections. By combining the projection model with the catheter centerline to construct spatial rays, the problem of spatial curve reconstruction is converted into a ray intersection problem, enabling accurate reconstruction of the catheter′s three-dimensional spatial shape. Finally, to verify the feasibility of the proposed catheter spatial shape reconstruction algorithm, experiments were conducted using biplane fluoroscopic images to reconstruct the catheter. The results showed that the maximum shape reconstruction error of the catheter was less than 1. 55 mm, the mean square error was less than 0. 89 mm, and the Hausdorff distance was less than 1. 49 mm. This indicates that the proposed method can achieve accurate reconstruction of the three-dimensional shape of interventional catheters, providing a new method and technical foundation for improving the precise navigation and safe manipulation of flexible guidewires in vascular interventional surgery

Abstract:The measurement of crack length is fundamental to the evaluation of crack risk and a prerequisite for crack repair. Aiming at the problems of irregular shapes, small target, sparse data sets and distortions of crack imaging angle, a sparse data driven method was proposed to extract the surface crack length of turbine blades. Firstly, to enhance the Unet model′s precision in handling sparse data, we employ a combination of the GeLu function with the Vgg16 network for feature extraction. The extracted features then serve as inputs for the Unet network′s decoding process. To ensure model compatibility, we incorporate pre-trained weights into the randomly initialized weights and integrate an efficient pyramid compression attention module into the skip connection layer. This approach significantly improves the model′s capability to focus on crack characteristics amidst complex backgrounds. Then, in order to get the unit pixel characteristic curve of the crack, after the fine segmentation, a skeleton structure with eight neighborhood is proposed to preserve the crack backbone characteristic structure. Finally, through an in-depth analysis of camera imaging principles, we discuss the impact of blade chord angles and camera parameters on crack length measurements, establishing a conversion model between pixel size and actual dimensions. Experimental results indicate that when the measuring distance ranges from 100 to 300 mm, the maximum error in crack length is 6. 8% . Compared to X-ray measurements, our method proves to be a viable alternative for measuring the surface crack length of turbine blades. Moreover, the enhanced algorithm demonstrates greater accuracy than the original algorithm in detecting sparse data, with an average cross-over ratio improvement of 7. 14% . The proposed method offers a theoretical foundation and data support for evaluating blade quality and guiding subsequent repairs.

Abstract:Depth completion is a technique for generating high-resolution dense depth maps from sparse depth data for environmental perception. Existing methods struggle with accuracy in low-light or dark conditions, performing poorly under extreme lighting. This article proposes a self-supervised method that fuses thermal images and LiDAR data to complete dense depth maps in low-light or no-light scenarios. The network adopts an encoder-decoder structure, using thermal images and sparse LiDAR depth as inputs. Features are fused at multiple scales in the encoder, and the decoder upsamples them to predict dense depth maps. Multi-modal fusion modules based on self-attention and cross-attention are embedded in the encoder to enhance feature fusion with adaptive weighting, improving prediction accuracy. A self-supervised framework is established with temperature reconstruction and sparse depth losses, removing the need for depth ground truth. Experiments on public datasets show that the method generates dense depth maps stably under various lighting conditions. Mean absolute error decreases by 44. 49% on MS2 and 25. 28% on VIVID compared to benchmarks. By leveraging thermal and LiDAR data′s complementary strengths, this method improves depth prediction accuracy and robustness in low-light environments, offering an effective solution for perception in challenging lighting. Keywords:depth completion; multi-sensor data fusion; thermal imag

Abstract:UWB positioning system was used in an indoor three-line autonomous driving rail transit system, where high-precision positioning of vehicles was a key technology to improve operational reliability and scheduling efficiency. A three-step UWB location algorithm including non-line-of-sight (NLOS) discrimination, ranging error compensation and neural network location error compensation was proposed using a joint convolutional variational auto-encoder and predictor (VAE-CNN), based on the analysis of UWB location accuracy. Firstly, the ranging error and channel impulse response (CIR) data between the tag and the base stations were collected and used to train the VAE-CNN model. The non-line-of-sight ranging values were eliminated according to the confidence threshold of the original and reconstructed CIR. Secondly, the original ranging values were compensated by the prediction errors of the predictor. The coordinates and the direction cosine of the coordinates with respect to the coordinates of each base station were calculated, which were used to train the neural network to fit the relationship between the localization error and the direction cosine. The NLOS discrimination capability of the VAE-CNN model was validated on a publicly available UWB ranging and CIR dataset, which includes both line-of-sight (LOS) and NLOS measurements. The effectiveness of NLOS discrimination and ranging error compensation based on the VAE-CNN model on improving positioning accuracy was also evaluated. The effect of positioning error compensation neural network on improving positioning accuracy was evaluated based on the simulated vehicle trajectories under different ranging variances. An UWB localization system was built to verify the practical performance of the three-step UWB localization algorithm in dynamic localization. The results show that in dynamic localization, in full line-of-sight environment, the algorithm achieved an average localization error of 28. 68 mm, a root-mean-square localization error of 16. 67 mm, and a maximum localization error of 76. 68 mm. In the presence of non-line-of-sight environment, the average localization error is 38. 73 mm, the root mean square localization error is 20. 61 mm, and the maximum localization error is 116. 47 mm. It can be seen that the three-step UWB location algorithm offers high accuracy, low cost, and excellent stability, meeting the positioning requirements of indoor rail transit systems. Keywords:three-line indoor rail transit; UWB location; convolutional variati

Abstract:With the continuous advancement of modern science and engineering technology, the importance of metal temperature measurement has become increasingly prominent for ensuring product quality, optimizing production processes, and ensuring safety. However, traditional metal temperature measurement methods often face challenges such as difficulty in online measurement and low measurement accuracy. To address these issues, this study proposes a non-contact metal temperature measurement method based on magnetic field coupling. First, the study analyzes an approximate linear relationship between the real part of the system′s equivalent impedance and temperature, thereby transforming the unmeasurable temperature changes into measurable system impedance. It provides a theoretical foundation for non-contact metal temperature measurement based on magnetic field coupling. Next, an equivalent model of the non-contact metal temperature measurement system is formulated. Its feasibility is evaluated through a three-dimensional finite element simulation using Ansys Maxwell software. Furthermore, the study proposes and compares two different types of eddy current sensors for temperature measurement, highlighting the multi-coil coupled structure for its compact size, high accuracy, and great stability. Based on this, a non-contact metal temperature measurement system using multi-coil coupling is proposed. The system modeling analysis and physical testing are conducted. The experimental results show that the absolute temperature deviation of the noncontact metal temperature measurement system based on the multi-coil coupling model is less than 2℃ , confirming the reliability and stability of the system. The system can operate stably without a direct visual path, overcoming the reliance on line-of-sight in traditional non-contact temperature measurement techniques. This advancement significantly improves the system′s performance and applicability, enabling precise, real-time measurement of metal temperature even in complex and obstructed environments.

Abstract:Under the action of waves, the ocean buoy device mainly undergoes pitch and heave motions. In this paper, a novel buoy wave energy harvester with multi-source coupling is proposed. The harvester employs a two-degree-of-freedom differential gear train to capture the pitch and heave motions of the buoy, with the output shaft of the differential gear train connected to the rotor of the generator. The power generation device utilizes a pendulum ball rotating around a fixed axis to capture the energy of the buoy′s longitudinal pitch motion, and a gear mechanism with unidirectional bearings is used to convert the reciprocating swing motion of the pendulum ball into unidirectional rotational motion to drive the generator, which serves as one input to the differential gear train. Simultaneously, the heaving motion of the buoy is converted into unidirectional rotational motion to drive the generator rotation through a gear rack mechanism with unidirectional bearings, acting as the other input of the differential gear train. The planetary carrier of the differential gear train serves as the output and is directly connected to the generator rotor, driving the generator to rotate in one direction. Ultimately, the kinetic energy of the buoy′s pitching and heaving motions is transformed into electrical energy. The dynamic equations of the pendulum ball generator, the float body generator, and the coupled pitch-heave motion in the buoy power generation device are established separately. Numerical simulations are conducted to study the dynamic response and power generation of the device under different wave periods and amplitudes. The results show that when the wave amplitude is 1 m, the average power generation of the buoy reaches its maximum at a wave period of 2. 3 s. As the excitation wave amplitude increases, the vibration responses of the buoy′ s pitching and heaving motions become more pronounced, and the power generation gradually increases. When the wave amplitude reaches 1. 1 m, the generated power can reach 525. 7 W. Additionally, when the wave period is 2. 3 s and the amplitude is 1 m, the installation of unidirectional bearings can increase the power generation efficiency by approximately 19. 2% compared to not installing unidirectional bearings.

Abstract:In the position detection of electromagnetic shock absorber (ESA), the commonly used displacement sensors often require large installation space and clean usage environment. To reduce the complexity of the detection method and the limitation of the detection environment, three linear Hall sensors are used in the actuator end of the ESA and are arranged at a spacing of 120° from each other at an electrical angle. They are used for detecting the information of the magnetic density change of the stator′s permanent magnets and converting it into the position information of the ESA. The research of the position detection method is implemented under the consideration of the installation error of Hall sensors, the sensitivity error between multiple sensors and the harmonic component generated by the magnetic field distortion. First, the arrangement position of Hall sensors is determined by analyzing the magnetic density variation, magnetic density peak, and harmonic content in the end region by the finite element method. Then, a mathematical model of the output Hall signal considering the Hall sensor error and harmonic interference is formulated. Based on the characteristics of this model, a dual phase-locked loop (DPLL) combined with a band-pass frequency synchronization extraction filter ( DPLL-BPFSEF) position-solving method is proposed. DPLL is used to suppress the influence of high-frequency harmonics on the signal, and BPFSEF is utilized to suppress the influence of low-frequency harmonics appearing in the signal to compensate for the unstable suppression ability of DPLL on low-frequency signals during the solving process. Finally, the effectiveness of the position detection method under uniform speed and sinusoidal motion conditions is analyzed through experiments. The results show that both DPLL-BPFSEF and DPLL can accurately solve Hall signals and realize the detection of the speed and position of the ESA actuator at the same time. The error rate of velocity is 5. 1% when detecting the sinusoidal motion of the actuator by DPLL-BPFSEF, and the position accuracy of DPLL-BPFSEF is 42. 8% and 37% higher than those of DPLL for both uniform velocity and sinusoidal motion.

Abstract:Wafer alignment is a critical process in ensuring precise positioning of wafers during manufacturing. However, its positioning accuracy is often compromised by residual vibrations caused by sudden changes in acceleration during the high-speed start and stop phases of the chuck. To address this challenge, this study first analyzes the contact characteristics of the chuck surface, develops a mathematical dynamic model for the wafer alignment chuck, and simulates the contact interactions and deformation between the wafer and chuck during motion. Based on this analysis, the relationship between the system’ s vacuum level, maximum rotational speed, and maximum rotational acceleration is explored to determine the wafer’ s amplitude and mode under various acceleration conditions. An acceleration constraint equation is derived, and a symmetric velocity curve for the wafer alignment is constructed using a third-order Bezier curve. Additionally, accounting for motor characteristic variations during acceleration and deceleration phases, the parameters of the symmetric velocity curve are optimized using a Pareto multi-objective genetic algorithm to create an asymmetric velocity curve. This approach helps to further suppress residual vibrations of the chuck, thereby enhancing positioning accuracy and operational efficiency. Simulation experiments confirm the improvement in operational efficiency with the proposed asymmetric velocity curve. A physical wafer alignment device is also built, and physical experiments demonstrate enhanced positioning accuracy and operational efficiency compared to traditional velocity curves. The experimental results show that the asymmetric velocity curve, which accounts for the dynamic characteristics of the chuck, effectively reduces residual vibrations, improving alignment accuracy to 0. 008 mm and reducing operation time by 0. . 14 s. These findings validate the effectiveness of the proposed method in improving wafer alignment performance.

Abstract:There are many problems in the speed measurement and positioning of urban rail transit trains, such as fewer available sensors, more lines with small radius curves and large slopes, frequent changes in operating conditions, and higher real-time and accuracy requirements. In this article, a speed measurement and positioning method based on an innovation-based adaptive Kalman filter is proposed, taking the unmanned metro as the research object. Firstly, based on the prior traction or braking target level constraint, the train is regarded as a one-dimensional rigid uniform mass model and taken into account the dynamic behavior of the train passing through the equivalent grade change point. A train motion model with modified maneuver acceleration is formulated. Then, based on the innovation-based adaptive Kalman filter, the statistical noise affected by the change of operating and line conditions is estimated and modified in real-time. Finally, taking the real train data of 3 typical conditions as an example, the speed measurement and positioning are carried out based on 16 sets of motor axle speed information, comparing its six accuracy evaluation indicators with that of the average axle speed method and conventional Kalman filter algorithm without adaptive noise estimation. The results show that this method can effectively modify the progressive data drift caused by wheel-rail creep and reduce the high-frequency noise in the high-speed area. The root mean square of speed error is 0. 349 0 km·h -1 , and the braking position error is 0. 491 3 m. Under the condition that the axle speed in the high-speed zone has a random loss of 1. 5% , the root mean square of the speed error can be stabilized at about 0. 371 7 km·h -1 , and the braking position error can be stabilized at about 0. 042 0 m, which has strong robustness to the loss of axle speed in the high-speed zone. Under the condition of train sliding, the root mean square of speed error is 0. 360 1 km·h -1 , and the braking position error is 0. 310 5 m, which has strong robustness to train slipping or sliding. The research results can provide a theoretical basis and engineering reference for the accurate speed measurement and positioning of unmanned metros.

Abstract:As China′s space industry advances from being a space power to a space strong nation, the number and density of spacecraft launches have reached new heights. Ensuring the normal operation of spacecraft in orbit has become a crucial task. Spacecraft telemetry data is an important basis for ground control to determine normal operation, and enhancing the anomaly detection capability of telemetry data is key to improving ground control′s support capabilities. Currently, anomaly detection of telemetry data mainly relies on expert experience and fixed thresholds. While these methods are efficient and reliable, they struggle to cope with the complex and dynamic operating environment in orbit, and the detection accuracy still needs improvement. Traditional machine learning methods show limited performance and effectiveness as the volume of telemetry data increases. In recent years, deep learning methods have shown great potential in the field of anomaly detection. However, existing deep learning-based anomaly detection methods for spacecraft telemetry data still face significant challenges. On the one hand, they heavily rely on the accuracy and completeness of anomaly labels, while obtaining a large amount of accurate anomaly-labeled data in practical engineering is difficult. On the other hand, existing methods lack the ability for online anomaly detection, which is essential for meeting the real-time monitoring needs of spacecraft in orbit. To address these issues, this paper proposes an online and unsupervised anomaly detection model, Feen-LSTM. This model extracts global spatiotemporal features from multidimensional telemetry data using a Transformer structure and combines LSTM to model local temporal dependencies, thereby achieving an optimized structure for feature enhancement. Experiments on two spacecraft telemetry data sets published by NASA show that Feen-LSTM can effectively improve the accuracy of anomaly detection, especially in the face of complex data and unknown anomaly patterns, and show better performance than other methods.

Abstract:Under the guidance of the “ dual-carbon” goals, wind energy, as a clean and renewable energy source, has been widely harnessed. Wind turbines (WTs), which are crucial for converting wind energy into electrical energy, have seen a growing cumulative installed capacity. However, WTs operate in harsh environments with highly variable conditions, leading to frequent failures. To ensure their safe and efficient operation, fault diagnosis and intelligent maintenance technologies are urgently needed. Aiming to address the complex and variable operating conditions of WTs and the issues of gradient disappearance and explosion in recurrent neural networks during time-series learning, this paper proposes a condition monitoring method for WTs that integrates operating condition recognition with echo state networks (ESNs). First, the maximum mutual information coefficient is employed to select features from the supervisory control and data acquisition (SCADA) system data, prioritizing those with high relevance to the operational status of WTs. Second, the K-means clustering algorithm is utilized to construct a model for effective classification of different operating conditions. Subsequently, ESN models are optimized under various conditions using the differential evolution algorithm to enhance their adaptability to complex operating conditions, enabling active power prediction of WTs under different conditions. Then, by analyzing the residuals of power prediction, corresponding health thresholds are determined to assess the operating conditions of WTs. Finally, case studies of two actual WTs demonstrate that the proposed method can effectively monitor the operating status of WTs. It can detect abnormal operating conditions earlier than the SCADA system when faults occur, thus realizing early fault warning.

Abstract:A sparse signal decomposition method based on K-SVD dictionary learning is proposed to address the issue of complex vibration signals in shipborne antenna transmission systems. These systems face variable and complex environmental conditions during actual operation, leading to highly nonlinear and non-stationary vibration signals, which increase the difficulty of fault diagnosis and health monitoring. Given the difficulty of traditional parameter dictionaries in matching the diverse characteristics of vibration signals, this paper first introduces a fault richness index based on the frequency-weighted energy operator to quantify fault information in signals. Subsequently, the complementary ensemble empirical mode decomposition ( CEEMD) technique is employed for signal denoising preprocessing, enhancing the signal reconstruction accuracy of the K-SVD algorithm in high-noise environments. We detail the application steps of CEEMD in practical signal processing and verify its denoising effect in high-noise environments through experimental data, further improving the signal reconstruction accuracy of the K-SVD algorithm. Additionally, this paper utilizes a sensitive component selection method based on the fault richness index to ensure that the recovered signal retains as much effective fault information as possible during the denoising process. Furthermore, the K-SVD algorithm is applied for secondary signal decomposition, and a novel dictionary initialization method is used to enhance the fault feature expression ability of dictionary atoms, thereby improving the algorithm′s operational efficiency and fault feature extraction accuracy. Finally, the effectiveness and accuracy of the proposed method are validated through simulations and experiments. The results indicate that the proposed method significantly enhances the accuracy and reliability of fault feature extraction, providing strong support for health monitoring and fault diagnosis of shipborne antenna transmission systems

Abstract:A chiller is a critical component of heating, ventilation, and air conditioning (HVAC) systems. Faults in chillers can lead to energy waste and even safety incidents. Therefore, fault diagnosis for chillers is essential for HVAC systems. Data-driven fault diagnosis methods rely on large amounts of historical data, but labeled fault data is often difficult to collect, resulting in reduced diagnostic accuracy of models. To address this issue, this paper proposes a fault diagnosis method based on a consistency loss generative adversarial network (CLGAN). First, CLGAN is trained with a small number of labeled samples and a large amount of unlabeled data to generate realistic fault samples. Next, a balanced dataset containing multiple fault categories is constructed by combining both generated and historical data. Finally, a fault classifier is trained on this balanced dataset to perform real-time fault diagnosis. By introducing a consistency loss function into the discriminator, CLGAN effectively leverages unlabeled data, increasing data utilization. Meanwhile, the generator is guided at multiple scales to meet the discriminator′s requirements, enabling the model to produce high-quality samples even under various disturbances and thus enhancing diagnostic accuracy and robustness. Experimental results on the ASHRAE and HY-31C datasets demonstrate that, with only five labeled samples per class, CLGAN achieves fault diagnosis accuracies of 92. 8% and 95. 9% , respectively, illustrating its excellent performance. Moreover, in noise and cross-condition experiments, CLGAN shows superior robustness and generalization compared with other methods.

Abstract:In response to the challenges of monitoring and accurately diagnosing the state of main bearings in aircraft engines using a single detection method, a method for rolling bearing condition monitoring is proposed, integrating multi-channel vibration signals with oil debris particle information. This approach initially utilizes a weighted fusion model for multi-channel vibration information to combine data obtained from multiple vibration sensors. Subsequently, the fused signal is decomposed using CEEMDAN, and components with strong impact characteristics are selected based on kurtosis-correlation coefficient filtering criteria, leading to the reconstruction of a vibration signal rich in bearing fault characteristic information. Time-domain features, using the total effective value, and frequencydomain features, employing feature energy, are then extracted as characteristic parameters. Through the selection of membership functions and the definition of fuzzy inference rules based on practical considerations and expert experience, fuzzy inference theory is applied to fuse the total effective value and feature energy into the first-level fused vibration information parameter, denoted as F1. The obtained oil metal debris particle count is utilized as the information parameter F2 for debris, which is further analyzed through a secondlevel fusion using fuzzy inference theory. Finally, the rolling bearing status is monitored, and bearing faults are diagnosed. Experimental tests involving the shedding and expansion of main bearing debris in aircraft engines were conducted. A detection system was installed to simultaneously collect vibration and oil debris particle information throughout the entire bearing shedding process. The proposed method was applied to analyze the collected data. Results indicate that the multi-channel vibration signal and oil debris particle information fusion method for rolling bearing condition monitoring enables comprehensive analysis of fault characteristics and effective discrimination of bearing operational states.

Abstract:Currently, residual current operated protective devices (RCDs) solely rely on a fixed threshold as the tripping criterion. As a result, under conditions such as improper parameter coordination, high harmonic content, and high-frequency arc pulses, there is a risk of failure to trip or unwanted tripping. Moreover, they cannot effectively distinguish true electrocution events. To address this issue, this paper proposes a novel RCD tripping criterion based on wavelet packet decomposition and dynamic feature component selection. This criterion can quickly identify various types of faults, including common ground faults, electrocution, and arcing faults. First, the fault onset moment is captured using kurtosis, a high-order statistical measure sensitive to signal impulses. The energy ratio of the differential residual current signal in each cycle before and after this moment is calculated to identify abnormal conditions in real-time. Second, the differential residual current signals from one cycle before the fault and three cycles after the fault initiation are collected for wavelet packet decomposition. The kurtosis, wavelet packet energy ratio, and sample entropy of each node component are combined to form a dynamic optimization index (DOI). The low-frequency and high-frequency signals are then reconstructed based on the contribution of each component′s DOI, highlighting the fault characteristics of different fault types in current waveforms across various frequency bands. Finally, electrical characteristics from the reconstructed signals are extracted, and fault classification is performed accurately through a two-level chain-rule approach. The proposed method has been validated on an RCD prototype. Experimental results show that it performs excellently in detecting series arcs, ground arcs, electric shock faults, and general grounding faults in low-voltage AC distribution networks. The recognition rate reaches 97. 52% , with an average diagnostic time of 79. 6 ms. This method meets the sensitivity and reliability requirements of RCDs, thereby significantly enhancing their practical application value.

Abstract:The aluminum electrolysis production environment is harsh, influenced by the coupling of multiple physical fields such as electric, magnetic, flow, and temperature fields, leading to frequent failures during the production process. The temperature of the aluminum electrolysis cell is a crucial parameter that affects the lifespan and operational status of the electrolysis tank. However, due to the high temperatures and corrosive nature of the tank, no effective online detection or prediction method for electrolysis temperature has been established so far. To address this issue, this study reveals the close correlation between the electrolysis temperature of aluminum electrolyzers and their process parameters through theoretical analysis and on-site experimental validation. Based on this, a deep learning-based model for predicting the electrolysis temperature is proposed. Considering the complexity, nonlinearity, high dimensionality, and temporal sequence of the process parameters, Convolutional Neural Networks (CNN) are employed to extract highdimensional features from the data, while Long Short-Term Memory (LSTM) networks are used for modeling. Additionally, the Attention mechanism is introduced to capture the relationships between different parts of the input parameters and to weigh the data according to its importance. A PID-based Search Algorithm ( PSA) is applied to optimize the CNN-Attention model for the aluminum electrolysis process, reducing training time and improving model performance. Experimental results demonstrate that the proposed temperature prediction model achieves a correlation index (R 2 ) of 0. 963 7, with a Root Mean Square Error (RMSE) of 5. 417 6 and a Mean Absolute Error (MAE) of 3. 382 5. A comparison with single-model algorithms, other prediction models, and different optimization techniques shows that the proposed model significantly outperforms them. The model successfully predicts the electrolysis temperature of the aluminum electrolyzer, enabling real-time, online detection of the electrolysis temperature during production.

Abstract:Under low light conditions, factors such as low image brightness, weak contrast, poor imaging quality and the constraints of on-board arithmetic greatly affect the detection accuracy from the UAV′s point of view. Therefore, researches based on object detection under low light conditions in UAVs is of great significance. Aiming at this problem, this paper proposes a multiscale differential attention fusion detection method with coupled illumination conditions and contrast. First, an information-aware module is designed to guide the multiscale differential attention module. This module deeply fuses the intra- and inter-modal features of visible and infrared images through calculating the light information and local contrast, thereby enhancing the recognition ability under low light conditions. Second, a rotary-wing UAV multimodal target detection system is constructed based on multimodal pods, edge computing modules and selforganizing network radios. This system has a standardized transmission protocol and a unified task management mechanism for communication interaction and realizes synchronous decoding. Subsequently, comparison and ablation experiments are designed, and the results show that the mAP of this method on the LLVIP is 69. 2% , which is 3. 9% better than before the improvement, and outperforms LRAF-Net. Finally, the proposed algorithm is validated at the airborne end of USVs, demonstrating that it can significantly improve the detection capability of UAVs on targets under low light conditions. The average operation efficiency can reach 21. 2 FPS, which meets the requirements of airborne applications.

Abstract:To improve the safety and flexibility of the feeding process for individuals with limited mobility who are unable to feed themselves, this study proposes a multimodal human-robot interaction framework for assisted feeding and a whole-process intention recognition algorithm. First, based on user characteristics and the requirements for safety and flexibility, a multimodal interaction framework integrating vision, touch, force, position, and language fusion is proposed, and an assisted feeding system is developed. Second, a vision-driven whole-process intention recognition method is introduced to address the entire feeding process, including feeding intention, dish selection intention, dynamic feeding point estimation, delivery pose calculation, and chewing intention. Key facial feature points that effectively capture dynamic changes during feeding are selected, and an algorithm combining the aspect ratio of the mouth and the mandible is designed. The user′s dish selection intention is analyzed through gaze vector estimation, and dynamic feeding points are determined based on real-time facial pose tracking, enabling accurate recognition of dynamic intentions throughout the process. Furthermore, in the virtual mapping system for assisted feeding, a feedback mechanism is established by leveraging a large language model to clarify ambiguous intentions and adapt to temporary changes during the interaction, thereby enhancing safety. Finally, the proposed method is validated through simulations and comprehensive experiments. The results demonstrate that the multimodal interaction framework significantly improves the flexibility of the assisted feeding process, while the integration of the large language model provides effective feedback for ambiguous and changing intentions, ultimately enhancing the safety of the interaction. This approach offers a novel care solution for assisting feeding behaviors in the daily lives of individuals with limited mobility

Abstract:The continuum robotic arm with a central backbone configuration utilizes a single continuous medium rod as its main structure, lacking the joint mechanisms found in traditional robotic arms. This design makes shape feedback a long-standing challenge. Currently, most shape measurement methods based on multiple IMUs rely on the piecewise constant curvature assumption. However, this assumption often fails when the robotic arm is subjected to external loads, leading to reduced accuracy in shape estimation. To address this issue, this paper proposes a multi-IMU shape estimation algorithm for continuum robotic arms based on tangent vector fitting. The algorithm employs Cosserat rod theory to mathematically model the continuum robotic arm, enabling a more accurate description of its deformation behavior. The shape at multiple measurement points is estimated using error-state Kalman filtering, and the tangent vector at each point is calculated. Subsequently, B-spline fitting is applied to the discrete tangent vectors to obtain a continuous tangent vector function with arc length as the independent variable. Finally, by integrating this continuously varying tangent vector function, the shape estimation is completed. Experimental results demonstrate that the algorithm achieves high-precision shape estimation under both dynamic trajectories and static loads, particularly when significant shape changes are induced by external loads. The algorithm exhibits strong robustness and stability. Compared to traditional methods based on the piecewise constant curvature assumption, the proposed algorithm significantly improves the positioning accuracy at the end-effector and the accuracy of shape reconstruction. Under high-load conditions, the shape estimation error is reduced by more than 50% compared to existing methods, proving its superiority in complex application scenarios.