Abstract:To resolve the stator overheat problem of the stator permanent magnet dual-rotor high-speed electrical machine (SPMDR-HSEM) due to its mustered heats source and high-speed operation condition, a 2D-3D hybrid thermal network model (HTNM) is built to analyze the temperature field of the electrical machine, and an integrated structure with magnetic isolation and cooling functions (IS-MIC) is proposed for accurately calculating the machine temperature and improving the machine heat dissipation capacity. Firstly, the topology of the SPMDR-HSEM is introduced, and its heat transfer mechanism is analyzed, as well as the conduction and convection heat coefficients are given. Secondly, for the problem of complex structure of the SPMDR-HSEM and the difficulty in calculating the temperature fast and accurately, the 2D-3D HTNM is proposed to analytically calculate the key components temperature rise of the SPMDR-HSEM under rated and overload conditions, its temperature distribution characteristics are summarized. And a test plat is established and the experimental tests are conducted, which verify the accuracy of the HTNM. Then, the IS-MIC is proposed to solve the heat dissipation problem of the SPMDR-HSEM, and the influence of the structure parameters and water flow velocity of the IS-MIC on the cooling effect is investigated, and the optimal structure of the IS-MIC is designed. And the finite element analysis is carried out to verify the effectiveness and rationality of the IS-MIC. It broadens the application field and provides a new temperature field analysis and cooling structure design method for the stator permanent magnet electrical machines.

Abstract:To address the challenges of electromagnetic characteristics regulation and vibration-noise suppression in memory motors, this paper proposes a novel composite topology structure. This design incorporates auxiliary slots in both the stator and rotor, utilizes NdFeB-AlNiCo hybrid permanent magnets to construct a dynamically reconfigurable magnetic circuit, and employs a segmented Halbach magnetization configuration. First, an equivalent magnetic circuit analytical model and a transient electromagnetic-mechanical coupled finite element model are established for the memory motor, enabling the derivation of analytical expressions for vibration and noise. Second, accounting for diverse performance requirements under multiple magnetization states, a hierarchical optimization strategy based on parametric sensitivity weighting is developed for extreme operating conditions. Using this method, the structural parameters of the proposed topology are optimized. Finally, multiphysics co-simulation integrating electromagnetic, structural, and acoustic domains is performed. Electromagnetic validation results demonstrate that the optimized motor maintains stable torque output characteristics across a wide flux-regulation range. Comparative studies reveal that compared to the baseline motor, the proposed design significantly enhances electromagnetic performance under multiple magnetization states while effectively suppressing peak vibration acceleration in stator teeth, reducing sound pressure levels, and achieving superior resonance frequency avoidance characteristics. This approach comprehensively optimizes the motors vibroacoustic behavior.

Abstract:Intrusion detection systems (IDS) face security challenges of generative model inversion attacks. And Federated GAN Attacks are the particularly characteristic data security threat to federated IDS. To improve data privacy in federated IDS, a universal privacy-enhanced federated intrusion detection (PEFID) method is proposed and is validated in diverse attack-defense simulation scenarios. PEFID jointly enhances data privacy at both the feature level and the model level. From the feature level, an improved adaptive privacy enhancing module is proposed to adaptively adjust the regularization degree of representation learning to balance privacy protection and task learning. Besides, controllable perturbations are injected into the hidden variables to further degrade the traceability of the gradient. From the model level, a label smoothing strategy combined with prediction confidence is proposed to deal with label inversion. Each client can individually adjust the soft label value according to the prediction confidence, assigning victim data a more lenient soft label value to mitigate the consistent attack. Experimental results on the CICIDS2018 and UNSW-NB15 datasets show that PEFID can effectively resist federated GAN attacks in various network scenarios. Compared with other methods, PEFID can better balance privacy and performance with controllable time complexity. It can still maintain superior defensive efficacy even in the case of single point penetration. The proposed method is both universal and lightweight, which can be adapted to existing federated IDS to significantly enhance data privacy with minimal performance cost.

Abstract:To address the challenges that denoising diffusion models struggle to adapt to varying noise levels and conventional residual blocks have limited feature selection capability in image fusion tasks, this paper constructs a multimodal image fusion network integrating dynamic gating diffusion denoising and cross-layer attention. Firstly, four groups of expert convolution kernels are designed and incorporated into the dynamic feature extractor module. The optimal convolution kernels are dynamically assembled based on input content, enabling adaptive processing of input features. Secondly, an improved gated feature selection module is proposed to generate gating signals that suppress irrelevant information, enhance the model’s diffusion denoising capability under different noise levels, and achieve precise feature control. Finally, R-Transformer blocks are adopted for feature adjustment. A global-local spatial attention module is constructed to realize cross-layer feature fusion, thereby generating fused images with rich texture information and high color fidelity. Experimental results on the MSRS, RoadScene, and Harvard datasets demonstrate that compared with 9 representative state-of-the-art methods in the field of image fusion in recent years, the proposed method achieves an average improvement of 5.11% to 15.93% across 7 objective evaluation metrics. The proposed method outperforms other counterparts in texture detail preservation and anatomical structure integrity maintenance, conforms to human visual perception characteristics, and can effectively handle multimodal image fusion tasks in scenarios such as various lighting environments and medical image diagnosis.

Abstract:Affected by the direction of temperature and moisture transfer, the damage development of mortar in the early freeze-thaw (F-T) stage exhibits spatial characteristics. However, existing F-T damage monitoring methods mostly focus on the overall average performance of specimens, ignoring the differences in local damage evolution. To quantitatively evaluate the influence of spatial position on the early F-T damage of mortar and further reveal its evolution mechanism, this paper proposed a real-time and in-situ strain monitoring method based on fiber bragg grating sensors and tested the strain at different spatial positions inside mortar during the early F-T stage. Results show that the strain amplitude in the upper layer of mortar is higher than that in the middle layer during the early F-T stage; with the increase of F-T cycles, the peak strain continues to rise, and residual strain appears; micro-morphology analysis results verify the reliability of using residual strain to judge the spatial difference of F-T damage. Based on this, this paper further established an early F-T damage evolution model considering spatial position. The comparative analysis results of macroscopic performance tests and local strain show that the damage of mortar in the early F-T stage is dominated by surface cracking. Although this phenomenon has little effect on the degradation of macroscopic performance, it can provide new transmission channels for environmental moisture to enter the interior of mortar, further aggravating the evolution of F-T damage.

Abstract:In order to solve the problems of ignoring the correlation between pixels in the deep learning fusion algorithm, which leads to the loss of important global texture in the fusion results, and the difficulty of balancing target highlight and scene enhancement, this paper proposed an infrared and visible image fusion algorithm guided by Taylor expansion and composite attention mechanism. Firstly, a Taylor expansion network was designed to decomposition the input image into a mapping layer and a derivative layer, so as to effectively extract the multi-level feature information of the image. Secondly, a dual-branch feature extraction network was used, in which the parallel convolutional network was responsible for capturing local detail features, and the SwinTransformer module focused on extracting global context information to ensure the efficient retention of local and global features. Then, the composite attention mechanism is introduced to further improve the accuracy of feature fusion. This mechanism fuses spatial dimensional features through axial attention, and uses channel attention to strengthen the feature response between channels, so as to achieve more refined feature selection and fusion. Finally, the fused image was obtained by image reconstruction. Experiments are carried out on the public datasets MSRS and RoadScene. The results show that the proposed method is not only more complete in maintaining texture details and global information, but also achieves significant advantages in objective indicators. The research results can provide new ideas for the field of deep learning image fusion.

Abstract:To defend against node replication attacks in wireless sensor networks and maintain network security and stability, this paper proposes a node replication attack detection strategy integrating node creditworthiness and identity authentication (NRADS-NC＆IA). The approach begins by establishing a fuzzy comprehensive trust evaluation model. In the direct trust module, reputation maintenance function, anomaly attenuation factor and reward and punishment factor are introduced. The direct trust value of the node to be evaluated is calculated by comprehensively considering the influencing factors of communication attributes, data attributes, network attributes and physical attributes. A sliding time window mechanism is adopted to dynamically update node trust value, significantly enhancing the timeliness and accuracy of the evaluation. On this basis, a support function is applied to evaluate the credibility of nodes, effectively filtering out the deceptive behaviors of nodes and obtaining more reliable indirect trust values of nodes. The final comprehensive trust value of nodes is derived from the weighted summation of its direct and indirect trust values. Dynamic adaptive thresholds are adopted to screen suspicious nodes, and node ID comparisons are conducted among the suspicious nodes to obtain the determined replica nodes. Simulation results show that NRADS-NC&IA achieves detection rates of over 97% and 94% in static and mobile wireless sensor networks without relying on the spatial position information of nodes. The strategy exhibits strong environmental adaptability and can effectively deal with the security problems of wireless sensor networks in complex dynamic environments.

Abstract:To address color distortion, detail blurring, and structural incoherence in existing style transfer techniques when processing Dunhuang murals——caused by highly saturated mineral pigments, intricate textures, and complex layered structures——this paper proposes a Multi-scale Dunhuang Style Transfer Network based on an improved Cycle-Consistent Generative Adversarial Network for high-quality artistic style transfer. We introduce an adaptive local dilated convolutional-net that dynamically captures detailed texture edges using deformable convolution and enhances long-range texture dependencies through dilated convolution, thereby restoring deep features to preserve brushstroke details. A dual scope net is designed to mitigate information loss and color-layer weakening during style transfer, employing a global attention branch to model overall tonal harmony and a local grouped convolution branch to reinforce stroke details. Additionally, a pathwise fusion net optimizes logical relationships and proportional coordination between elements using multi-dilation-rate depthwise separable convolutions for parallel processing and a dynamic gated fusion mechanism. Experimental results show that the proposed method achieves reductions of 5.81%, 4.36%, and 5.73% in FID, LPIPS, and L2 loss, respectively, and an improvement of 8.12% in SSIM. User studies confirm its superiority in content fidelity, style consistency, and visual appeal. This approach effectively resolves challenges in preserving color layers, texture details, and spatial layouts in Dunhuang murals transfer, offering a novel approach for Dunhuang murals digitization and innovative dissemination.

Abstract:Near-fault ground motions are critical for assessing seismic damage. However, recorded near-fault ground motion during an earthquake is often limited. To address the challenge of evaluating seismic damage to engineering structures in near-fault regions, particularly for supplementing ground motion at sites without observing stations in near-fault regions, this study proposed a near-fault ground motion reconstruction method based on observed data and an equivalent pulse model. The method initially generated ground motion at sites without observing stations in near-fault regions by interpolating response spectra from observed near-fault ground motion. Subsequently, it employed an empirical equivalent pulse model to recover velocity pulses and permanent displacements of ground motion. By taking the Chi-Chi earthquake in Taiwan, China as an example, examples of ground motion reconstruction at two typical locations in the near-fault region were provided. By comparing them with the observed ground motion data, the effectiveness and rationality of the method were validated. The results demonstrate that the method effectively reconstructs ground motions with near-fault pulse-like characteristics and ground motion at sites without observed data in near-fault regions. This offers a practical approach to meet the need for ground motion in near-fault regions where observed data are lacking.

Abstract:Aiming at the problem that it is difficult to select the weight matrix parameters in the vehicle model predictive control(MPC) trajectory tracking controller, which makes the stability and accuracy of the vehicle trajectory tracking control insufficient, this research propose a latin-pelican algorithm(LPOA) that integrates multiple mechanisms to optimize the weight matrix parameters of the vehicle lateral and longitudinal joint model predictive trajectory tracking controller. Firstly, the vehicle transverse MPC controller, the longitudinal MPC upper controller and the lower controller based on the acceleration-drive inverse dynamics model are designed respectively based on the vehicle single-track model;Secondly, a Latin Pelican Optimization Algorithm is proposed to improve the efficiency of the pelican algorithms searching in the solution space. The hierarchical hunting mechanism of the gray wolf algorithm is introduced to reconfigure the prey localization model of the POA, and the convergence speed of the algorithm is improved by the α-pelican guidance strategy. Thus, a dynamic stochastic search strategy is incorporated to enhance the algorithms ability to escape from local extremes in the late iteration by using its heavy-tailed distribution characteristics. Finally, the parameters of the horizontal and vertical MPC controller weight matrices are optimized using the optimization capability of LPOA; and the proposed horizontal, vertical, and horizontal-longitudinal joint optimization control methods are verified through co-simulation on CarSim and Simulink platforms. Results show that the LPOA-MPC controller proposed in this research can effectively improve the stability and accuracy of vehicle trajectory tracking control in horizontal, longitudinal and transverse-longitudinal joint control.

Abstract:Traditional Iterative Closest Point (ICP) algorithms are susceptible to local optima and sensitive to initial pose when applied to large-scale structured scenes, such as underground tunnels, due to their repetitive geometric features. To address these challenges, this paper proposes an efficient two-stage point cloud registration framework. In the coarse registration stage, an improved Sample Consensus Initial Alignment (SAC-IA) algorithm with an enhanced keypoint sampling strategy is introduced to rapidly obtain a globally optimal initial pose from the downsampled point cloud. Subsequently, the fine registration stage employs a novel multi-dimensional weighted ICP algorithm. This algorithm integrates Euclidean distance, normal vector angles, and curvature information into a comprehensive geometric objective function, effectively reducing the mismatch rate in the planar regions of tunnels.The proposed components were individually validated on the Stanford public dataset. Results show that the improved coarse registration algorithm increases computational efficiency by 44.1% over the conventional SAC-IA, while the geometry-based weighted ICP algorithm improves registration accuracy by 23.81% and reduces computation time by 25.9% compared to the traditional ICP. Comprehensive experiments on real-world tunnel point clouds demonstrate that the proposed framework significantly outperforms traditional methods, improving both overall accuracy and efficiency by over 20%. Furthermore, compared to mainstream deep learning methods, this framework obviates the need for high-performance GPUs. It achieves a several-fold increase in processing efficiency on a standard CPU and delivers more robust registration results in realistic tunnel scenarios. Ultimately, this framework provides a highly efficient solution for automatic point cloud registration in large-scale structured environments, fully satisfying industrial demands for real-time processing and lightweight implementation.

Abstract:To enhance model performance, Deep Neural Networks are frequently trained on untrusted datasets, rendering them vulnerable to data poisoning backdoor attacks. Conventional detection methods rely on identifying feature discrepancies between poisoned and benign samples. However, their effectiveness diminishes when attackers optimize trigger generation to obscure this boundary. To address this issue, this paper proposes a novel detection method named reverse forgeting (RFgt). The method exploits the characteristic of backdoor attacks, where the proportion of poisoned samples is low, and employs a reverse optimization strategy. Instead of forcing a poisoned model to forget backdoor features, RFgt compels it to rapidly forget the features of the majority class (benign samples), while simultaneously retaining and reinforcing the learning of suspicious samples to consolidate their poisoned features. This approach significantly amplifies the feature disparity between the two sample types. Ultimately, the prediction entropy of the samples is used to determine whether they are poisoned or benign. Experimental results demonstrate that RFgt effectively detects poisoned samples under various backdoor attacks on the CIFAR-10 and GTSRB datasets, while maintaining a low false positive rate. Furthermore, this method demonstrates strong generalization capability, as shown by its performance on the Tiny ImageNet dataset. Specifically, against four classic data poisoning attacks, RFgt achieves an average True Positive Rate (TPR) of 99.28% and a False Positive Rate (FPR) of only 0.06%, outperforming existing defense methods in overall performance.

Abstract:In complex electromagnetic environments, radar countermeasure reconnaissance signals often suffer from significant distribution differences between training data and actual combat scenarios due to substantial pulse loss and false pulse interference, which seriously degrades the recognition accuracy of the air-to-air working mode of active phased array radar. To address this issue, this paper proposes a recognition model based on self-attention multi-kernel dilated convolution network (SAMKDCN). Centered on dilated convolution, multi-kernel selection, and residual structures, this model constructs a feature-map extraction module for multi-scale feature learning across the temporal dimension. Moreover, a self-attention mechanism is incorporated to adaptively adjust feature-map weights, thereby highlighting critical features and strengthening feature representation, which ultimately enhances the accuracy of AESA radar working-mode identification. Simulation experiments show that SAMKDCN can effectively learn the core features of the air-to-air working mode of AESA radar. Under ideal conditions, it achieves a peak accuracy of 99.14%. With pulse-loss and false-pulse ratios ranging from 0% to 50%, the average recognition rate attains 95.11%; Even under the extreme scenario of 50% loss rate and 50% false-pulse rate, this model retains a recognition accuracy of 88.23%, demonstrating favorable generalization ability and robustness.

Abstract:The accurate prediction of wind turbine blade icing is essential for the safe and stable operation of wind power systems. To address the challenges of insufficient feature extraction, unclear distribution of multi-dimensional sensors, and class imbalance in few-shot learning scenarios, this paper proposes a prediction method based on a residual graph attention network-bidirectional LSTM-siamese network (ResGAT-BiLSTM-SN). First, the supervisory control and data acquisition (SCADA) data is processed through data cleaning, sliding-window sampling, and feature engineering, resulting in a blade-icing dataset suitable for various few-shot learning tasks. Second, based on 19 key variables, a non-fully connected undirected graph is built using mutual information (MI) and a weight matrix to capture the spatial distribution and correlations among sensor data. By integrating the graph Attention network (GAT) and the bidirectional long short-term memory network (BiLSTM) to extract spatiotemporal features, the ResGAT-BiLSTM-SN model is developed to perform 24-hour-ahead icing prediction on the constructed dataset. Simulation experiments are conducted using the data from turbines No.15 and No.21 provided by the 2017 Industrial Big Data Innovation Competition platform. The experimental results show that the ResGAT-BiLSTM-SN model achieves F1 scores above 0.9 across three few-shot learning scenarios, significantly outperforming other baseline models. Compared to the GAT-BiLSTM-SN model, the proposed model demonstrates clear improvements in predictive performance, validating its effectiveness and superiority.

Abstract:To address the vulnerability of frequency-modulated (FM) radio fuzes to amplitude-modulated sweep-frequency information-based jamming threats in complex electromagnetic environments, this paper proposes a classification anti-jamming method based on frequency-domain entropy features and a parrot optimization algorithm (POA) optimized support vector machine (SVM). First, the output signal of the fuze detector stage is transformed from time domain to the frequency domain using the fast Fourier transform (FFT). Three entropy measures — frequency-domain information entropy, exponential entropy, and R-norm entropy are then calculated to construct a three-dimensional feature matrix. Subsequently, the POA is employed to optimize the parameters of SVM classifier. The optimized SVM utilizes a Gaussian kernel function, with its penalty parameter C and Gaussian kernel parameter σ adaptively adjusted by the POA to enhance the classification merit. Experimental results demonstrate that the entropy features of the target and typical interferences (noise, sine wave, square wave amplitude modulation sweep) exhibit significant separability in their probability density distributions. The POA rapidly converged to the optimal solution within 300 iterations, with fitness values stabilizing below 0.001. Validation in a microwave anechoic chamber confirmed that the POA-SVM achieved 96.8% target recognition accuracy and 97.2% interference recognition accuracy, representing significant improvements over traditional SVM and PSO-SVM methods. Furthermore, Modelsim simulations confirmed the algorithms response performance meets millisecond-level operational requirements of fuzes. The proposed approach effectively enhances both recognition accuracy and real-time capability of FM radio fuzes against informational jamming, offering a novel pathway for fuze anti-jamming recognition in complex electromagnetic environments.

Abstract:To address the pronounced nonlinearity of stress-strain relationships during the compression deformation of structured clay and the limitations of traditional constitutive models in terms of mathematical rigidity and applicability, this study proposes an elasto-viscoplastic constitutive model for structured clays based on a unified rational function formalism. First, based on Maxwell elements, this study decomposed the total strain into the sum of elastic and viscoplastic strains, with a focus on investigating viscoplastic strain associated with structural effects. A rational function expression for viscoplastic strain-effective stress-equivalent time was established through one-dimensional compression behaviors. Second, under the assumption that “the viscoplastic strain rate of soil depends solely on effective stress and viscoplastic strain,” the viscoplastic strain rate expression for structured clay was derived by integrating the concept of equivalent time, and then a one-dimensional elasto-viscoplastic constitutive model for structured clay was derived. Third, the parameter calculation method was introduced, transforming the problem of solving parameters into a multiple linear regression task by using linear programming techniques, with matrix solutions implemented through computational tools. Finally, during model verification, the strain variation of soil under sudden loading conditions as a function of real time was derived. The model was applied to simulating conventional consolidation tests and creep tests, demonstrating its applicability and the feasibility of the parameter calculation method. The research results indicate that the proposed equivalent time model for structured clay effectively describes both the effective stress-viscoplastic strain relationship and the total strain-time relationship under one-dimensional compression.

Abstract:Mechanical parameters of rock mass are one of the important indicators for the comprehensive stability assessment of rock slopes. Existing parameter inversion methods are mainly based on the final deformation values under stable conditions, making it difficult to reflect the nonlinear and time-varying characteristics in the actual slope deformation process. To this end, this paper proposed an inversion method for the mechanical parameters of rock mass based on the dynamic prediction of slope deformation. Firstly, the crow search algorithm (CSA) was introduced to optimize the weight and threshold parameters of the online sequential extreme learning machine (OSELM), and the CSA-OSELM dynamic deformation prediction and parameter inversion models were constructed, respectively. Secondly, piecewise cubic Hermite interpolation and wavelet decomposition methods were adopted to preprocess the measured deformation data to extract the trend term deformation. Thirdly, the dynamic deformation prediction model was used to obtain the final deformation value of the slope, which was substituted into the inversion model to output the mechanical parameters. Finally, verification analysis was carried out by taking the southern slope project of the Jingxi-Barak mining area in Xinjiang as an example. The results show that the CSA-OSELM model outperforms other models in prediction accuracy and stability; by substituting the mechanical parameters obtained from the inversion into the numerical model for forward calculation, the average error between the calculated values and the measured deformation values is 6.21%, which further verifies the practicality and reliability of the method in this paper. The research results can provide a new technical approach for rapidly obtaining mechanical parameters of rock mass in practical engineering.

Abstract:To investigate the structural and thermal performance of the connection between point-supporting facade panels and the main structure, quasi-static tests and heat transfer calculation were conducted. The results indicate that test models achieve the failure mode characterized by “strong column-weak beam” and “strong joint-weak component” behavior, with ultimate drift ratios of approximately 1/30. The connection joints remain undamaged. The connection details of precast components could ensure the overall seismic performance of the precast frame structure. Whether the connection bolts of the precast facade panel are tightened or not had little impact on the elastic lateral stiffness of the main frame. Similarly, the differences between bolted connections and wet connections with U-shaped rebars are minimal in terms of their influence on the seismic performance of the main frame before the elastic-plastic drift ratio limit. Under large lateral drift, the U-shaped steel plate damper can reduce the rigid-body rotation of the precast facade panel by 16% to 30%, while having no significant impact on the failure mode, load-bearing capacity, and deformation capacity of the main frame. Therefore, U-shaped steel plate dampers can be employed to control the seismic displacement response of large-scale external wall panels, achieving a reasonable balance between joint width, sealant application convenience, and displacement capacity. The concrete between the inner and outer wythes is the root cause of thermal bridge in the connection area, resulting in a 62.1% reduction in the thermal performance of the facade panels. The heat insulated pads could markedly reduce the thermal bridge effect in the connection by 86.6%, thereby achieving thermal bridge mitigation near the connection.

Abstract:To investigate the bond-slip behavior between ribbed steel bar and coal gangue concrete, 36 cubic specimens were designed and prepared, in which coal gangue replacement ratio, concrete strength grade and rebar diameter were selected as variables, and the center pull-out tests were conducted. The results show that the bond strength of the specimen gradually decreases with the increase of coal gangue replacement ratio and the failure mode changes from pull-out failure to splitting failure. With the increase of rebar diameter, the bond strength of the specimen shows a decreasing trend, and the degree of decline of coal gangue concrete is more significant than that of ordinary concrete. As the concrete strength grade increases, the bond strength of the specimen increases significantly and the failure mode of the specimen changes from the pre-yield failure of steel bar to the post-yield failure of steel bar when the strength grade reaches C55. Based on the experimental data, a bond strength prediction model and a 3D meso-scale finite element model of reinforced coal gangue concrete members were established. The finite element model, the accuracy of which was verified, was utilized to further reveal the damage evolution and bond failure mechanism of coal gangue concrete and ribbed steel bar under pull-out force.