Abstract:With the extensive deployment of industrial sensors and the rapid development of artificial intelligence algorithms, data-driven intelligent fault diagnosis technology has become a key part and a hot topic in Prognostics and Health Management of mechanical equipment. However, such methods rely on substantial labeled data and demand strict consistency in data distribution, thereby causing a significant drop in the accuracy and robustness of related methods in real industrial scenarios. Transfer learning, as an effective approach to tackling the problems of inconsistent data distribution and small-sample fault diagnosis, has drawn widespread attention from both academia and industry. It markedly enhances the generalization performance of the model in the target domain by transferring the knowledge acquired in the source domain to the target domain. To investigate the current state of transfer learning-driven intelligent fault diagnosis methods for mechanical equipment and the crucial technical challenges that urgently need to be resolved, an analysis and summary of the existing literature in this field have been carried out. Firstly, the research progress and current status of intelligent fault diagnosis for mechanical equipment in domestic and international studies have been systematically reviewed. Then, focusing on transfer learning technologies, the advantages and limitations of various transfer learning fault diagnosis methods have been analyzed and compared. From the perspectives of different application scenarios and key technical issues in the industry, the intelligent fault diagnosis technologies for mechanical equipment driven by transfer learning have been summarized and critically evaluated. Finally, current research hotspots have been explored and the technical bottlenecks have been thoroughly analyzed, and potential solutions to existing challenges along with future development trends are identified. Studies show that while transfer learning has garnered widespread attention in the field of intelligent fault diagnosis for mechanical equipment, many technical issues remain to be resolved. With the rapid development of artificial intelligence technologies and the continued efforts of experts and scholars in advancing transfer learning theories and applications, a solid theoretical and technical foundation can be established for the development of intelligent fault diagnosis methods for mechanical equipment.

Abstract:To deal with the threat of space debris to on-orbit spacecraft, modular docking systems are added to the four-corner autonomous maneuvering units of the flexible net capture system. The units can assemble into a combined towing spacecraft, which is convenient for the implementation of subsequent deorbiting tasks, effectively preventing target escape and increasing the capture success rate. Firstly, to address the vibration problem caused by the flexible tether net in the on-orbit capture process and the challenges in coordinated actuation of the four-corner autonomous maneuvering units, a new control method is designed by combining the higher-order sliding mode algorithm with the consensus formation cooperative strategy. Secondly, the single-degree-of-freedom simulation analysis of dynamic model of the tether net capture system based on the mass concentration method is carried out. The control effect and fuel consumption of different sliding mode algorithms and the algorithm in this paper are compared. Finally, the optimal control combination of the super-twisting sliding mode algorithm and the leader-follower multi-agent consensus method is established, and successfully applied to the full-degree-of-freedom on-orbit capture simulation. The simulation results show that the new controller can complete the on-orbit capture of the target within 50 s, which exhibits excellent robustness and effectiveness. At the same time, the four-corner autonomous maneuvering units can keep the attitude angle fluctuation less than 3° during the capture process, which fully meets the objective conditions for docking systems of autonomous maneuvering units. The research effectively connects the transition between the on-orbit capture stage and the deorbit towing stage.

Abstract:To explore the cavity evolution and ballistic characteristics during the oblique water-entry of asymmetric structures, experimental research was conducted on various asymmetric-headed bodies. High-speed imaging technique was utilized to capture the cavity evolution and body position during the water-entry process of asymmetric-headed moving bodies. Utilizing digital image processing, trajectories and attitudes of these bodies were extracted. A comparative analysis was then performed to assess the impact of head shape and entry mode on cavity evolution and ballistic characteristics. The results show that the evolution of the water entry cavity of asymmetric-headed moving bodies has unique cavity flow characteristics such as secondary open cavity, secondary splashing, primary cavity attachment, cavity fusion, and secondary cavity collapse. The shape of the head of the moving body will seriously affect the evolution of the water-entry cavity and the ballistic characteristics of the moving body. As the shape of the head of the moving body changes from a convex shape to a concave shape, the width of the splash water curtain, the size of the cavity curtain, the size of the cavity and the attitude angle of the moving body will gradually increase. On the contrary, the closure time of the secondary cavity will gradually decrease. This characteristic has a profound impact on practical engineering applications.

Abstract:To analyze the aerodynamic characteristics of flying-wing aircrafts after being hit by air defense systems, a combination of wind tunnel test and numerical simulation was used to conduct force measurement analysis on a battle-damaged flying-wing aircraft under the condition of Reynolds number Re=1.47×10⁵, and the LES method was used to study the flow field characteristics of some working conditions, which reveals the reasons of the changes in roll and lateral characteristics of flying-wing aircraft caused by battle damage holes. It is found through the wind tunnel test that the battle damage has less influence on the longitudinal aerodynamic characteristics of the flying-wing aircraft, and more influence on the roll and lateral aerodynamic characteristics. The rolling moment coefficient of the battle-damaged flying-wing aircraft within the angle-of-attack range of 10°~30° within the angle-of-attack range obviously larger than that of the undamaged aircraft, and the absolute values of the rolling moment coefficient and lateral force coefficient of battle-damaged flying-wing aircraft of model2 are the largest; within the angle-of attack range of 10°~24°, the absolute values of the rolling moment coefficient and lateral force coefficient of battle-damaged flying-wing aircraft of model3~model5 decrease as the battle damage holes move towards the tip chord direction. The high-precision simulation of the flow field of the battle-damaged flying-wing aircraft through the LES method reveals that the airflow on the lower surface of the wing will flow to the upper surface through the battle damage holes, which induces the flow separation of the wing by the wind area in advance, thus leading to the asymmetric flow separation on the wing surface of the flying wing layout aircraft, and causing an increase in the absolute values of the rolling moment coefficient and lateral force coefficient of the aircraft. And the closer the battle damage hole is to the root chord, the larger the flow separation area induced out of the wing leeward area is, and the more obvious the asymmetric flow phenomenon in the leeward area of the flying wing layout aircraft is. By analyzing the wake vortices of a battle-damaged flying-wing aircraft, it is found that the battle damage hole induces multiple vortex systems behind it, and each vortex system is close to each other and entangles with each other. The vortices induced by the damage holes move towards the tip chord and merge with the tip vortices as the damage holes move towards the tip chord.

Abstract:To address the issue that support vector machines (SVM) frequently encompass a considerable number of redundant samples during classification, which restricts the computational complexity of SVM when confronted with large-scale datasets, a SVM sample selection algorithm based on local density minimum uncertainty is put forward. This approach efficiently identifies influential boundary data points that significantly affect the decision boundary, subsequently reducing computational costs by isolating potential support vectors from the training set, thereby bolstering SVM’s overall effectiveness. Firstly, the number of K nearest neighbors and Gaussian kernel density estimation of the training samples are computed; Secondly, the sum of the number of K nearest neighbors and Gaussian kernel density estimation is derived for each sample point to acquire the K local density and obtain the density matrix; Subsequently, employing the local density uncertainty balancing optimization method, the density matrix undergoes a triple-mapping process to minimize uncertainty changes, yielding the optimal threshold. This threshold then partitions the density matrix into center data and boundary data. Finally, the boundary data are extracted and utilized as training samples for the SVM, enabling the establishment of an effective classification model. To experimentally evaluate the efficacy of our method, we compared it with six commonly utilized sample selection techniques on UCI datasets, employing accuracy and preservation rate as key performance metrics. The findings indicate that the method introduced in this paper significantly reduces the number of redundant training samples, thereby effectively alleviating the training burden on SVM and enhancing its classification performance.

Abstract:To address the challenges of low efficiency and insufficient success rates in odor source localization (OSL) within complex and dynamic indoor plume environments, particularly where robots struggle to accurately perceive the environment and navigate effectively under turbulent conditions, this paper proposes an auxiliary value and wind-guided proximal policy optimization (AVW-PPO) algorithm based on deep reinforcement learning. First, an auxiliary value network is introduced into the original PPO framework to reduce the estimation bias of a single value network, thereby improving prediction accuracy and stabilizing policy updates. Next, a wind-guided strategy is designed to integrate local wind field information into the state space and reward function of the reinforcement learning framework, enabling the robot to better perceive dynamic changes in the plume environment and optimize its decision-making path, thus significantly improving the efficiency of odor source localization. Finally, a gas diffusion model in a two-dimensional environment is constructed to test the proposed algorithm under three different turbulence conditions. Experimental results demonstrate that, under identical environmental conditions, the AVW-PPO algorithm outperforms other comparable algorithms in terms of average search steps and success rates, achieving a localization success rate of over 99%. Notably, the wind-guided strategy significantly boosts search efficiency, helping to reduce the time required for the robot to complete tasks. This study provides new insights and methodologies for addressing odor source localization problems in complex turbulent indoor environments.

Abstract:To address the issues of sparse keypoint features in weakly textured indoor environments, insufficient utilization of structured features in structured scenes, and keyframe tracking failures during rapid camera movements, a stereo visual-inertial SLAM method based on the fusion of point and line features is proposed. First, the EDlines line segment extraction method, combined with a Gaussian image pyramid, is employed to achieve multi-scale line segment extraction, enhancing the scale invariance of line segment matching. Simultaneously, the uncertainty of line segment endpoints at different scales is modeled, and binary descriptors of line segments are partitioned using tiling technology to accelerate line segment matching, thereby improving the robustness and efficiency of line feature matching. Second, the pre-integration model of the inertial sensor is optimized, and a sliding window nonlinear optimization is performed by fusing the point feature reprojection error from stereo vision, the line feature reprojection error, and the pre-integration constraints of the inertial sensor, thereby improving the system’s pose estimation accuracy. Finally, extensive experiments are conducted on the EuRoC dataset which includes complex environments such as low-texture, structured scenes, and rapid camera movements. The experimental results demonstrate that the proposed method achieves a root mean square error of 0.031 m and an average error of 0.027 m on the EuRoC dataset, exhibiting stronger robustness and higher localization accuracy, especially in low-texture and rapid camera movement scenarios where the accuracy advantage is particularly significant.

Abstract:To address issues such as structural complexity, low transmission rates, and power-signal crosstalk in bidirectional power and signal synchronous transmission systems for inductive wireless power transfer, a novel bidirectional power-signal synchronous transmission method based on amplitude modulation and tuning is proposed. In this paper, the forward signal is amplitude-modulated by changing the phase shift angle of the inverter phase-shifted full-bridge control to achieve forward signal transmission. The reverse signal is tuned-modulated by changing the secondary resonant capacitance to achieve reverse signal transmission. This results in the current at both ends of the coil carrying signal characteristics. A current transformer is then used to feed the current signal with signal characteristics into the demodulation circuit to restore the signal. First, theoretical analysis of the circuit structure is conducted. Then, simulations are performed using Matlab/Simulink to verify the theoretical analysis. Finally, a 120 W experimental platform is constructed based on the simulation results. Experimental results demonstrate that at a power 120 W, the load voltage fluctuation is less than 3%, and the impact of power transmission on signal transmission is minimal. The system achieves half-duplex communication with a forward rate of 4 kbps and a reverse rate of 20 kbps at a bit error rate of 0.1%. Both experimental and simulation results confirm that this method effectively achieves bidirectional power and signal synchronous transmission in inductive wireless power transfer systems, offering high transmission rates and low bit error rates, which provide valuable guidance for wireless power transfer system design.

Abstract:To address the low generalization and poor robustness in feature matching methods that rely on pre-defined parameterized models, based on the observation that the spatial distribution of correct match and mismatch has significant differences, a feature matching algorithm guided by local density difference (RFM-LoDD) is proposed. Firstly, the putative feature matches are converted into spatial sample points that can characterize the nature of the feature matches, and the probabilistic distance is introduced to calculate the local density of the sample points. Secondly, the optimal parameter settings of the algorithm are tested on 40 randomly selected image pairs involving different transformation models,determining the globally optimal density threshold and other parameters. Finally, the local density of the sample points is compared with the density threshold. When the local density of a sample point is greater than the density threshold, the putative feature match represented by the sample point is considered to be a correct match, otherwise, it is a mismatch. Experiments conducted on representative image pairs and public datasets demonstrate that the RFM-LoDD algorithm maintains good robustness in various matching scenarios. Notably, it achieves leading F-scores on the Retina dataset and AdelaideRMF dataset with low inlier rates compared to advanced algorithms. Additionally, the RFM-LoDD algorithm has quasi-linear time complexity, with an average run time of about 40 ms on the four public datasets, significantly reducing time cost by two orders of magnitude compared to the classical random sample consensus (RANSAC) algorithm.

Abstract:To improve the fatigue life of a vehicle bearing shaft under bending-torsion loading a microscopic finite element model containing surface roughness and surface residual stress was established for the bearing shaft material TC11 titanium alloy. A bending-torsion fatigue life prediction model of TC11 was established based on the critical plane method to study the effects of surface roughness and axial/circumferential residual stress on the fatigue life. The fatigue samples were processed by turning and ultrasonic rolling respectively, and the surface integrity of the samples was compared under both processes. Fatigue tests were carried out to verify the accuracy of the prediction model, and the fracture mechanism was studied by analyzing the fracture morphology. The results show that the influence of surface roughness and axial residual compressive stress on life is significant. As the surface roughness S_a decreases from 1.6 μm to 0.4 μm, the life is increased by 135%, and the axial residual compressive stress increases from 100 MPa to 400 MPa, the life is increased by 123%, and the influence of circumferential residual stress is small. The rolling specimens have lower surface roughness, higher work-hardening degree, higher axial/circumferential residual compressive stress, and higher fatigue life. The SWT model can accurately predict the bending and torsional life of TC11 within the 25% error scatter band, which provides a theoretical basis for process optimization. After rolling, the fatigue grain spacing decreases from 0.8-0.9 μm to 0.3-0.4 μm, the crack propagation rate slows down, and the main fracture mode transitions from intergranular fracture to transgranular fracture in the machining-affected layer.

Abstract:To systematically and comprehensively understand the dynamics of the moderate-Stokes-number particle-laden jet (MSPJ) and to verify the applicability of Taylor’s fluid particle theory for smaller particles, an experimental analysis of the velocity evolution of moderate-Stokes-number particle jet is carried out. Firstly, a particle image velocimetry (PIV) experimental bench is built. Then, six sets of experiments including macroscopic large-scale and mesoscale measurements are carried out at different initial velocities. Finally, the evolution characteristics of instantaneous velocity, average velocity and fluctuating velocity of particles at two scales are compared and analyzed, and the MSPJ velocity decay is predicted and verified by combining with Taylor’s particle-laden fluid theory model. The results indicate that the average velocity of particles along the jet centerline decays similarly to the gas phase, exhibiting an initial increase followed by a decrease. In contrast, due to the mixing of low-velocity particles rebounding off the wall and high-velocity particles in the center of the jet, the attenuation of particle fluctuating velocity exhibits a different trend: it first decreases, then increases, and finally decreases again. Moreover, a significant difference is observed between the distributions of the particle pulsating velocity field near the nozzle and the average velocity field; the fluctuating velocity field displays a profile characterized by lower velocities in the center and higher velocities at the edges and in the transition zone, while the average velocity field shows the opposite pattern. The maximum cumulative error of Taylor’s fluid particle theory in predicting MSPJ particle velocity attenuation is 6.16%. Additionally, the velocity self-similarity of heavy particles decays more rapidly due to significant inertial effects. This study provides a reference for further research in the areas of slip velocity, drag force, and engine oil spray.

Abstract:To solve the problems of fault identification in models stemming from the difficulty of acquiring composite fault data in complex industrial environments, a novel zero-shot rolling-bearing composite fault diagnosis model based on contrastive feedback generation is proposed. Initially, continuous wavelet transform is employed to convert vibration signals into time-frequency images, thus preserving the temporal and spectral characteristics of faults more effectively. Subsequently, an attention-guided ConvNeXt feature extraction module is introduced, harnessing channel and spatial attention mechanisms to enhance fault feature representation, mitigate interference from extraneous information, and augment the distinctiveness of fault characteristics. Integrating adversarial training and attribute feedback alignment networks ensures that the generated pseudo-fault features accurately reflect their corresponding fault attribute information, achieving high-fidelity fault feature generation. A contrastive learning module is incorporated to produce fault features that are proximate to positive samples while maintaining distance from other samples, thereby further enhancing the performance of the feature generator and the discriminative power of the features. Finally, calculating the similarity between the pseudo-fault features and the unknown composite fault features, the category label with the highest similarity is assigned as the label for the unknown composite fault, thereby achieving its diagnosis. Experimental results demonstrate that the feature extraction network augmented with the attention mechanism improves diagnostic accuracy by 8.42% compared to other networks; enhances by 14.67% over using only the WGAN-GP generation module; and significantly elevates fault diagnosis accuracy by 28.67% compared to other models, thereby validating the effectiveness and superiority of the proposed model, offering an innovative solution for the intelligent maintenance of mechanical equipment.

Abstract:To identify the flow-induced noise sources in a helical-axial pump and analyze the influence of the unsteady flow on the flow-induced noise, a numerical simulation analysis of the flow-induced noise in the pump was conducted using a combination method of computational fluid dynamics (CFD) and acoustic finite element method (FEM). Firstly, the 3D unsteady flow within the pump was simulated using the shear stress transport (SST) turbulence model in ANSYS CFX to obtain the pressure pulsation characteristics of the flow field. The pressure pulsations on the solid walls were treated as excitation sources, and the flow-induced noise field in the pump is numerically calculated using the FW-H equation with the software of LMS Virtual Lab. In order to identify the noise source location, the focus was put on the frequency domain characteristics of flow-induced noise and the contribution of different components and regions to noise. Finally, based on the temporal signal correlation and coherence theory, the spatial-temporal correlation of the pressure pulsation signal on the impeller blade surface was deeply analyzed to reveal the flow characteristics and noise contribution mechanism of the key noise source region. The results indicate that, compared with the downstream of the diffuser, the flow-induced noise at the upstream of the impeller is greater and exhibits a slower attenuation with increasing frequency, with the highest noise contribution occurring in the frequency range of 0~3 000 Hz. The maximum sound pressure peak appears at the blade passing frequency. The total sound pressure level of the rotating source at the monitoring point is 25.4 dB higher than that of the stationary source, with the suction surface of the blade generating more noise than the pressure surface, particularly in the first 50% chord length region, which contributes significantly to the overall noise. Spatial-temporal correlation analysis of pressure pulsation signals at different positions on the suction surface of impeller blades reveals that the pressure pulsation in this region primarily originates from boundary layer separation.

Abstract:Tool tip frequency response function (FRF) changes with the position of the machine tool spindle, spindle speed, and tool parameters. In order to quickly and accurately obtain the frequency response function of the machine tool tip, a method based on a small number of experimental samples to predict the rotating tool tip FRFs under different tool parameters is proposed by introducing transfer learning. Firstly, an orthogonal programming table for the position and speed of the machine tool spindle is generated, and the prediction model of the rotating tool tip FRFs related to the machining space and the spindle speed is established based on the combination of the self-excitation method and the convolutional neural network （CNN） algorithm; Then, considering the influence of parameters such as tool elongation, diameter and type, transfer learning is adopted to train the prediction model of tool tip FRFs under different tool conditions by a small amount of data. Finally, the prediction model is trained using experimental data from hammer impact experiments and self-excitation experiments at the machining center VMC80IV. The modal parameters of each order output by the model are reconstructed to obtain the tool tip FRFs using the modal superposition method, and the predicted values of the model are compared with the actual measured values. The experimental results show that for the prediction model of the FRFs of the rotating tool tip under different tool working conditions, the prediction error of each order of natural frequency did not exceed 2%, and the prediction error of damping ratio did not exceed 5%, which verifies the effectiveness and accuracy of the prediction model.

Abstract:The integration of mobile energy storage capabilities of electric vehicles (EVs) into building energy system scheduling has gradually become a significant measure for advancing the green and low-carbon development of the construction sector, which can effectively enhance the energy efficiency of buildings while reducing operational costs. For large-scale building integrated photovoltaic (BIPV) public buildings, a two-level optimization configuration and scheduling method for BIPV building energy systems, considering EV integration, is proposed. First, a Gaussian mixture distribution model is used to fit the daily travel behavior of EVs, establishing an EV travel pattern model based on public buildings. Next, to simultaneously determine the configuration and scheduling results of the system, a two-level optimization model is established: the upper level aims to minimize the annual planning cost of the system to obtain configuration results, while the lower level aims to minimize the daily operational cost of the system to obtain power load scheduling results. Finally, a solution method for the two-level optimization model is provided, and a case study is conducted using typical daily weather load data from the heating season of an office building on a university campus. The results indicate that configuring and scheduling the building energy system according to the solution results can effectively improve photovoltaic utilization and extend the service life of energy storage equipment. Furthermore, by comparing different energy storage schemes in the system, it is found that utilizing only EV energy storage is more advantageous in both long-term and short-term economics, and achieve good system operational performance.

Abstract:To study the development status and the key problems that need to be solved urgently of machine learning in motion prediction of marine structures, this paper comprehensively discusses the research on motion prediction of marine structures in the field of marine engineering in the past ten years. With the increasing demand for motion prediction of marine structures, the traditional prediction methods based on fluid mechanics theory cannot meet the practical application requirements in terms of both prediction accuracy and real-time performance. The emergence of machine learning methods makes it possible to accurately predict the future motion response and realize advanced control of structures according to the response. Based on the modeling principles of forecasting methods, they are classified into four categories: statistical regression methods, general neural network methods, intelligent neural network methods and hybrid forecasting methods, and the four categories of methods are comprehensively reviewed, analyzed and synthesized. Finally, the existing shortcomings and problems are analyzed, and the future development directions are given from the aspects of prediction method, framework and data set, which can provide reference for the development of motion prediction of marine structures such as ships and offshore platforms. The research shows that the research of machine learning in the field of marine structure motion prediction is still in the initial stage, and there are still many technical problems to be solved. However, with the vigorous development of AI large model and the deepening of machine learning research by researchers in this field, it can provide a solid foundation for the development of characteristic prediction methods in this field.