Abstract:Aiming at the vibration frequency of automotive powertrain systems with problems such as multiple orders and narrow vibration bands of a single order， this paper proposes a multi-frequency line spectrum notch-filter vibration active control method based on the filtered-x least mean square algorithm. First， the method is based on the rotational speed signal to obtain the reference signals of multiple order vibration frequencies， and then the minimum mean square filter is used to calculate the offset signals of these order vibration frequencies. Secondly， based on the orthogonality of the vibration signals at different frequencies， the method further calculates the control signals with linear superposition to obtain the control signals of active suspension and eventually realizes the active control of multi-frequency line spectral vibration. Compared with the traditional filtered-x least-mean-square algorithm， this algorithm improves the acquisition of reference and control signals and thus has the advantages of less computation and faster convergence speed. Simulation and experimental results show that the proposed method reduces both the number of convergence by 81.25% and the steady-state error by 15%， respectively， compared with the traditional filtered-x least-mean-square algorithm. The maximum reduction is 34.01 dB under idling condition and 17.5 dB under WOT condition， respectively， compared with the traditional filtered-x least mean square algorithm.

Abstract:To improve the objectivity， accuracy， and interpretability of device wear evaluation， an improved entropy-weighted TOPSIS method for wear state evaluation is proposed based on multidimensional element data obtained from oil spectroscopy detection， compressively using the DTW， entropy-weighted TOPSIS and system clustering methods. Firstly， the similarity of wear element sequences is measured using the DTW method to extract wear-sensitive features. Secondly， an entropy-weighted TOPSIS wear state evaluation model is established. The contribution weights in wear evaluation are determined based on the information content carried by the wear features and their degree of variation， and the degradation index indicator that quantitatively describes the degree of wear is obtained. Finally， an adaptive hierarchical clustering is performed on the wear degradation indicators to divide the wear state into three stages： normal wear， abnormal wear， and severe wear. A case study on wear evaluation of the power-shift steering transmission is presented， where the DTW distance is used to extract Fe， Cu， and Pb as wear-sensitive features； the wear evaluation weights of Fe， Cu， and Pb are determined to be 0.193， 0.341， and 0.466， respectively， based on entropy calculation； wear degradation index indicator is obtained by the entropy-weighted method to quantitatively describe the wear trend； and the wear state is clustered into three stages using the degradation index， forming a tree-like structure with clear boundaries and strong interpretability. The effectiveness of the improved entropy-weighted TOPSIS method for wear state evaluation is verified， which can provide a scientific basis for health monitoring of comprehensive transmission devices.

Abstract:When applying active suspension control， challenges such as parameter perturbation and the inability to directly acquire state variables in the algorithm may arise. Therefore， developing a robust control algorithm based on state observation is crucial. In this paper， a dynamic model of semi-vehicle roll suspension is established. A nonlinear filtering function coordinates the suspension deflection and the vertical acceleration of the vehicle body. It is then combined with a fuzzy sliding mode algorithm to achieve continuous sliding mode switching by utilizing fuzzy approximation， aiming to improve the chattering problem. On this basis， the stability of the control system under parameter perturbation is proven through the Lyapunov method， and a parameter adaptive law is designed. Additionally， for the state variables that cannot be directly measured in the algorithm， a particle filter state observer is designed to estimate their values in real-time. Finally， simulation analyses are conducted under typical working conditions， such as sinusoidal road excitation and random road excitation. The results demonstrate that the designed observer can provide real-time and accurate state information required by the control algorithm， and the fuzzy sliding mode controller with parameter adaptability exhibits good robustness and greatly improves vehicle posture and ride comfort.

Abstract:To accurately measure the reliable life of the multi-way valve of the backhoe hydraulic excavator， two reliability accelerated life estimation test methods for the normal operation of the multi-way valve were proposed based on the measured load data of the excavator during excavation operation as well as the Archard wear equation. Taking the 20 t backhoe hydraulic excavator as the research object， the load data under the typical operating conditions of the hydraulic excavator are first collected， and 100 groups of typical excavation operation cycle data are extracted from the load data according to the pressure characteristics of the multi-way valve port during the operation of the excavator. To measure the reliable life of the normal operation of the multi-way valve in a short time， two accelerated life test methods of synchronous loading and simulated loading were proposed， where synchronous loading is the A/B port plugging， the rated flow rate of the main relief valve is input， and it is kept constant. In the process of continuous reversing， the rated pressure of the main relief valve is used to load the pressure of the multi-way valve， and the steady-state time of the valve core back to the median position is obtained by AMESim simulation and the pressure holding time of each valve port is determined by factors such as material deformation. Simulated loading means that each valve port is loaded following the action sequence of each valve port when the excavator performs a 90° rotation operation. Finally， the acceleration multipliers of the two acceleration tests are calculated using the Archard wear equation and combined with the actual working conditions of the multi-way valves. It concludes that the acceleration multipliers of synchronous loading are larger， while the simulated loading is closer to the actual working cycle.

Abstract:Due to the inability of the current theory to accurately calculate the changes in the stiffness of the gear during loading and impact， the Hertz nonlinear contact theory is employed to construct an analytical model of length-limited line contact elasticity theory. The normal contact stiffness of the gear under different loads is obtained from the changes in the three-dimensional contact spot size and the principal curvature radius in the meshing process. Later， using the Weber energy method， a time-varying meshing stiffness correction algorithm considering the load condition is proposed to clarify the mapping relationship between load deformation and stiffness hardening. At the same time， the mathematical model of off-line meshing impact is established， and the change law of the meshing stiffness caused by the meshing impact is investigated according to the dynamic contact relationship of the gear pair at the impact time. The research findings demonstrate that the contact stiffness curve is consistent with the change law of the principal curvature radius. When the meshing impact is considered， the meshing stiffness of the gear pair changes abruptly from the large to the small. As the load increases， the contact and comprehensive meshing stiffness increase non-linearly， the single tooth meshing region gradually decreases， and the impact on the stiffness is more markedly.

Abstract:During the charging and discharging of a parallel-configured battery pack， the temperature distribution of each cell in the battery pack is inconsistent. There is a big difference between core and surface temperature for each cell， which directly affects the thermal safety evaluation of the battery pack. To solve the problem that the surface temperature of the battery measured traditionally cannot reflect the core temperature distribution， a battery core temperature estimation algorithm based on the combination of Rauch-Tung-Striebel （RTS） and unscented Kalman filtering （UKF） is proposed. Based on the data processing method of RTS， the future information is combined with the UKF algorithm. The results of the UKF algorithm are corrected using the future information Value to improve the accuracy and stability of the estimation. The hybrid pulse power characteristic （HPPC） experiment at different temperatures is used to identify the parameters of the equivalent circuit model. The current distribution model and the lumped thermal model of the parallel-configured battery pack are established. The model of the parallel-configured battery pack is experimentally verified. Under the dynamic stress test （DST）， the accuracy and stability of battery core temperature estimation by the RTS-UKF algorithm are improved compared with the UKF algorithm， and its estimation standard deviation is 4.2%.

Abstract:To meet the requirements of SOC estimation accuracy and robustness， a temperature-dependent second-order RC equivalent circuit model is established， taking a lithium-ion battery cell as the study object and considering the effects of temperature change on open-circuit voltage， polarization resistance， polarization capacitance， and capacity. Simulation results show that the model has higher accuracy than the second-order RC model. Based on this model， the extended Kalman filter algorithm with multiple suboptimal fading factor is used to estimate SOC. The results show that the root mean square error of SOC estimation of the SMFEKF algorithm based on the variable temperature model and the constant temperature model is reduced by 42.7% and 48.2%， respectively， compared with the EKF algorithm under variable temperature conditions， which can ensure strong robustness of estimation results. The maximum relative error and root mean square error of SOC estimation results based on the variable temperature model are smaller than those of the constant temperature model under the DST condition in the variable temperature environment， which proves that the model has strong temperature adaptability and can have higher estimation accuracy under the variable temperature condition.

Abstract:A object detection algorithm based on improved YOLOv8s is proposed to address the issues of low accuracy and low efficiency in surface defect detection of hot-rolled strip steel. Firstly， an SPPD module based on feature map secondary stitching and incorporating GAM is proposed， which enhances the model’s multi-scale information fusion ability. Secondly， a feature extraction module DCN-block that integrates deformable convolution is proposed to increase the receptive field of the model and extract complete defect information. Finally， the C2f module in the feature fusion network is replaced with a BoT （bottleneck transformer） structure， and the multi-head self-attention mechanism in the Transformer is fused with convolution to enhance the model’s global position information perception ability. The experimental results show that the proposed algorithm achieves mean average precision （mAP） of 80.5% on the NEU-DET dataset， which is five percentage points higher than the original YOLOv8 algorithm. At the same time， the detection speed reaches 83 frames per second， meeting the requirements of real-time detection.

Abstract:To solve the problem of large computation and low accuracy of the current mainstream algorithms for small object detection， this paper replaces the backbone network in YOLOv4 with the lightweight network MobileNetV3， and replaces some ordinary convolutions in the neck network with depthwise separable convolutions. At the same time， a new loss function IF-EIoU Loss is defined for small object detection. Therefore， MDS-YOLO object detection model is constructed. This model has a high detection speed and good detection performance for small object. To verify the effectiveness of the model， experiments are carried out on MS COCO dataset and Visdrone2019 dataset， respectively. Compared with the YOLOv4 algorithm， on MS COCO dataset， the average detection accuracy of the MDS-YOLO algorithm is improved by 1.5 percentage points， the detection accuracy of small object is increased by 3.3 percentage points， and the detection speed is also increased from 31 frames per second to 36 frames per second. On the Visdrone2019 dataset， the MDS-YOLO algorithm increases the average detection accuracy from 14.9% of YOLOv4 to 16.3%. The experimental results show that the MDS-YOLO algorithm proposed can effectively improve the detection accuracy of small object.

Abstract:To address the problem of accurately inferring the content of missing regions in an image when they are closely related to the surrounding textures and structures， we propose a single-stage image inpainting model. The model first compresses， reconstructs， and enhances features through convolutional layers and the FastStage module， while self-attention and multi-layer perceptron are incorporated to capture contextual relationships among features. Furthermore， in order to enhance the attention and importance perception on features， we propose EMMA in the models， which avoids the shaking and oscillation during updating the model parameters， thereby improving the performance of the generator and the quality of the generated results. Lastly， we introduce a discriminator to evaluate the consistency between the inpainted image and the original image. The end-to-end experimental results conducted on CelebA， Places2， and Paris StreetView datasets demonstrate that， compared with classical methods， the inpainting results of this model exhibit better visual semantics， and it is capable of finely inpainting details， textures， and local features of images.

Abstract:The field of vision of deep-sea submersibles is usually limited， and it is difficult to comprehensively observe the surrounding seabed through the video image of a single field of vision， which makes it more difficult for researchers to understand the overall distribution of seabed substrate. To solve the above problems， this paper proposes a fast stitching method of seabed substrate image based on the deep-sea submersible image. Firstly， the red channel of the video frame is corrected based on the image enhancement method of channel compensation， and the brightness enhancement and CLAHE processing are carried out. Secondly， the CUDA accelerated SURF algorithm is used to extract feature points and descriptors， and the KD tree algorithm is used to initially match the feature points of the front and back frames. Then， the KNN classification algorithm is used to eliminate the mismatching， and the interframe motion estimation is carried out for the screened matching points. The base map is generated through the transformation matrix and the front and rear frames are spliced. Finally， the feature point coordinates and interframe motion information are fused， and the above process is repeated to generate a continuous mosaic image. The video images obtained by “Jiaolong” in a certain voyage are used for mosaic processing experiments. The results show that this method has a good mosaic effect， and its feasibility is verified.

Abstract:With the continuous expansion of China’s road network， road disease detection has become an indispensable part of road maintenance and traffic safety， and road disease detection based on deep learning has become a research hotspot in this field. Aiming at the problems of low accuracy and generalization ability of road disease identification in complex scenes with multiple diseases， a road disease detection model called Receptive Ghost Triplet-YOLOv7 （RGT-YOLOv7） in complex scenes is proposed in this paper. A triplet attention mechanism is introduced in the backbone network to improve the correlation of disease features in different channels and spaces， and to solve the problem of low feature extraction efficiency. The original SPP module is replaced by the SPPF module， the Ghost module is added to improve the utilization rate of redundant features， and the original redundant features and the newly extracted features are fused to get more diverse and rich feature information with different scales. In order to improve the model perception field， RFBs module is added in the feature enhancement part， and the feature map is extracted from different directions by using cavity convolution with different sizes to enhance the extraction of horizontal and vertical features. Experimental results show that the average accuracy and balanced F score are improved by 6.9 percentage points and 3.9 percentage points， respectively， compared with YOLOv7， especially the longitudinal fracture identification is improved by 22.3 percentage points， and it also has good performance improvement compared with Faster R-CNN， YOLOv5， and recently proposed algorithm models. It is an effective road disease detection algorithm for proposed RGT-YOLOv7 under complex scenes.

Abstract:Accurate detection of ship violations in bridge area waters is very important for the pre-control of ship-bridge collisions. To ensure the safety of ship navigation， this paper presents a detection model of ship violation facing bridge area waters. AIS data of continuous bridge area in the Wuhan section of the Yangtze River is real-time collected and preprocessed， and the Convolutional Neural Network （CNN） with powerful feature learning ability is used to extract ship behavior information.And combined with the Long Short Term Memory （LSTM），a deep CNN-LSTM is established to learn the spatiotemporal behavior characteristics of ships， and the experimental analysis is carried out based on three kinds of illegal behaviors on ship overspeed， turning around， and overtaking. The results show that the DCNN-LSTM model proposed has a strong advantage over the CNN， LSTM， and Support Vector Machine （SVM） models， and its accuracy rate， precision rate， and F1 are 88.96%， 96.49%， and 92.87%， respectively， realizing the accurate identification of ship violation. The validity and superiority of DCNN-LSTM are further demonstrated by analyzing the violation of ships in typical waters. The research results provide a reliable theoretical basis for ship safety supervision in bridge waters and promote the development of ship intelligence.

Abstract:Malicious attackers can easily deceive neural networks by adding human-imperceptible adversarial noise to natural samples， leading to misclassification. To enhance the model’s robustness against such adversarial perturbations， previous research has predominantly concentrated on the robustness of single-modal tasks， with insufficient exploration of multimodal scenarios. Therefore， this paper aims to improve the robustness of multimodal RGB-skeleton action recognition and introduces a robust action recognition framework based on a Feature Interaction Module （FIM）， which extracts global information from adversarial samples to learn inter-modal joint representations for calibrating multi-modal features. A corresponding loss function tailored to this framework is also developed. Experimental results demonstrate that against CW attack， our method achieves a RI of 25.14% and an average robust accuracy of 48.99% on the NTURGB+D dataset， outperforming the latest SimMin+ExFMem method by 8.55 and 23.79 percentage points， respectively. These findings confirm that our approach surpasses others in enhancing robustness and balancing accuracy rates.

Abstract:As the current algorithms fail to fully extract local features and result in significant degradation of network accuracy when performing geometric transformations such as translation， scaling， and rotation on point cloud data， this paper proposes a dynamic graph attention-based 3D point cloud recognition and segmentation algorithm based on adaptive generated convolutional kernels. Firstly， the positional information of the center point in the receptive field is used to enhance the contextual information perception of neighboring points. The receptive field is reconstructed to enable sufficient interaction of feature information within the receptive field and enhance the contextual information by improving the self-attention mechanism. Then， an adaptive generated convolutional kernel is constructed to capture changing point cloud topology information and adaptively generate convolutional kernel weights to enhance network performance. Finally， a dynamic graph attention convolutional operator is built， and a dynamic network for point cloud recognition and a U-shaped network for segmentation are designed. The experimental results show that the proposed algorithm achieves a recognition accuracy of 94.0% in the ModelNet40 point cloud recognition dataset， and the instance mean intersection over union reaches 86.2% in the ShapeNet Part point cloud semantic segmentation dataset. The algorithm proposed can extract critical feature information from 3D point clouds and is capable of 3D point cloud recognition and segmentation.

Abstract:The traditional attack detection methods struggle to identify advanced persistent threat （APT） attacks launched using zero-day vulnerabilities. To address this issue， this paper proposes an APT attack activity identification for zero-day attack method （APTIZDM）， which consists of three key components. The first component is the cyber situation perception ontology construction （CSPOC） method， which provides a formal description of critical activity attributes and features in IoT systems. The second component is the malicious command & control （C&C） DNS response activity mining （MCCDRM） method， which identifies malicious C&C communication activities in APT attack scenarios while effectively controlling the scope and starting time of the identification process， thereby reducing computational overhead. The third component is the zero-day attack activity recognition method in APT attack （ZDAARA） scenarios， which utilizes Bayesian networks and security risk propagation theory to perform correlation analysis on system call information. It calculates the malicious probability of each system call instance and effectively identifies zero-day attack activities missed by intrusion detection systems. Simulation experiment results demonstrate that MCCDRM and ZDAARA， as the core components of the APTIZDM， achieve high accuracy and low false positive rates， effectively collaborating to identify APT attack activities.

Abstract:Aspect-based sentiment analysis （ABSA） aims to identify users’ opinions expressed about specific text aspects using elements such as aspect words， opinion words， and sentiment polarity. However， the existing research mainly focuses on individual tasks， which neglects feature interactions between different parts and causes error propagation issues. A sentiment triplet extraction method based on a multi-feature weighted graph convolutional network is proposed to jointly model multiple subtasks. Then， a double affine attention module is employed to capture the relational probability distribution among word pairs. Additionally， prior information such as text semantics， syntax， and location is encoded into multi-feature vectors. Finally， graph convolution operations are utilized for achieving multi-feature fusion and realizing the joint extraction of aspect term-opinion term-sentiment polarity . Through the estimation test based on 2 benchmark datasets， the experimental results reveal that the sentiment triplet extraction method based on a multi-feature weighted graph convolutional network can effectively alleviate the error propagation issues in pipeline methods. Moreover， feature interaction among each factor of the triplet set is proposed， and it is proved that the model in the current work performs much better than the previous benchmark model at triplet extraction.

Abstract:Aiming at the problem that the performance of the stage multi-axis synchronous system cannot meet the time limit of the control task due to the degradation of the actuators， and the existing maintenance strategy is difficult to reach the optimization， this paper proposes a reinforcement learning-based predictive maintenance strategy for the stage multi-axis synchronous system. Firstly， reinforcement learning is introduced in a cascaded manner， and constructing a control architecture with capabilities for lifespan prediction and autonomous maintenance， which operates with different sampling rates. Secondly， focusing on the intervening maintenance strategy and the influence of multi-source uncertainty on the actuator degradation process， based on the algorithms of Kalman filtering， Expectation-Maximum， and Rauch-Tung-Striebel smoothing， by the real-time perception and estimation of actuator degradation state， and a daptive update of degradation model， the prediction accuracy of the remaining life of the multi-axis synchronous system is ensured. Combined with the real-time perception， deviation of remaining life prediction， and the actuator degradation state， the objective function of a Q-learning algorithm is constructed. The optimal adjustment of maintenance control is carried out through trials and errors to obtain the maximum life extension reward and realize intelligent optimization maintenance of the stage multi-axis synchronous system. Finally， the effectiveness of the proposed method is verified by simulation experiments of the stage multi-axis synchronous control system， improving the system maintenance efficiency.

Abstract:Medical degradable β-TCP/WE43 composite was fabricated via multi-pass friction stir processing （MP-FSP）. The microstructure of the β-TCP/WE43 composite was characterized by metallography， scanning electron microscopy， and energy dispersive spectrometer， and the mechanical properties， corrosion resistance， and corrosion-wear performance of the composite were tested and analyzed. The results showed that， after three passes of friction stir processing （FSP）， β-TCP particles were well dispersed in the magnesium matrix， and the grains in the stirring zone were refined. MP-FSP can effectively improve the distribution of β-TCP particles in the magnesium matrix and refine the grains， thereby simultaneously enhancing the mechanical properties， corrosion resistance， and corrosion-wear resistance of the composite. These research findings can provide significant guidance for the development of biodegradable magnesium-based implants.

Abstract:Two kinds of CO2-responsive linear polymers， DTS-LP1 and BT-LP2， were reported. Both can change their hydrophilicity under CO2 regulation and thus can uniformly disperse in aqueous solution. At the same time， after the reaction， the polymer photocatalyst can be quickly separated and reused simply by removing CO2 from the system. In addition， the photocurrent and impedance were investigated by electrochemical measurements， where DTS-LP1 demonstrates a greater light current response intensity and small charge transfer impedance. The band position of these two polymers can be obtained by cyclic voltammetry （CV） characterization， indicating that it meets the oxidation condition of photocatalysis sulfides. When DTS-LP1 and BT-LP2 were employed as photocatalysts， the conversion rates of photocatalytic sulfide oxidation within 10 h were 97% and 42%， respectively， indicating that the polymer photocatalyst has excellent catalytic performance.

Abstract:In response to the poor dispersibility of microcrystalline graphite （MG） as a rubber reinforcing agent， this paper selected 25% ethanol solution with similar surface tension to graphene as the ball milling solvent， and prepared MG ultrafine powder with a median particle size of 1.209 μm through ball milling method. By partially replacing carbon black （CB） as a reinforcing agent for natural rubber （NR）， NR/CB/MG composites with better mechanical properties were obtained. The dispersion of MG and CB in the NR matrix was characterized using field emission scanning electron microscopy （SEM）， and the effects of MG particle size and substitution amount on the properties of NR/CB/MG composites were studied. The results indicate that an appropriate amount of MG can promote the dispersion of CB in the NR matrix and shorten the positive vulcanization time of NR/CB/MG composites； The finer the particle size of MG， the better the mechanical properties of NR/CB/MG composites. Compared with NR/CB composites， its tensile strength， tear strength， 100% elongation stress， and 300% elongation stress are increased by 6.3%， 7.5%， 8.4%， and 6.4%， respectively. The fatigue temperature is risen and the permanent deformation rate is decreased by 2.5% and 44.7%， respectively.