Zhijun Fu

Bearings serve as core yet vulnerable components in rotating machinery, making accurate fault diagnosis indispensable for safeguarding equipment operational safety and prolonging service life. However, Transformer-based bearing fault diagnosis models are often limited by two key challenges: local feature distortion caused by rigid data segmentation, and insufficient modeling of the intrinsic temporal evolution of signals. To address these issues, this study proposes an augmented Vision Transformer (ViT) framework termed BSFRformer (Boosting Spatiotemporal Feature Representation Transformer). Specifically, 1D convolutions are integrated to enhance the patch embedding of ViT, enabling capture of local dynamic features from input signals prior to tokenization. Moreover, a Bidirectional Long Short-Term Memory (Bi-LSTM) network is embedded into the multi-head attention module during query and key vector generation, improving the encoder's capability to capture temporal dependencies and extract multi-scale features. Additionally, Optuna is adopted to optimize key hyperparameters of BSFRformer. Experimental validation on a self-built dataset and the Case Western Reserve University (CWRU) bearing dataset demonstrates that BSFRformer achieves high diagnostic accuracy and excellent generalization performance, outperforming state-of-the-art counterparts. The proposed framework provides a reliable technical solution for enhancing the operational safety of rotating machinery and extending its service life, thus holding substantial practical significance and engineering application value.

Accurate bearing fault diagnosis in electric drive systems is crucial for ensuring the safety and reliability of new energy vehicles. Aiming at the problem of inaccurate bearing fault diagnosis caused by the failure to fully utilize the time-frequency feature information of bearing faults and the lack of an adaptive selection mechanism for features, an intelligent bearing fault diagnosis method based on wavelet packet decomposition (WPD) time-frequency feature enhancement is proposed in this paper. First, the collected vibration signals are enhanced using WPD to obtain the full-frequency-band time-frequency information, which provides input for the bearing fault diagnosis model. Second, a hybrid neural network CNN-BiLSTM-AM for bearing fault diagnosis is constructed. On the basis of using the convolutional neural network (CNN) improved with cross-convolutional layers to extract multiscale spatial features of the input data and the bidirectional long short-term memory network (BiLSTM) to capture the bidirectional temporal dependence between features, the attention mechanism (AM) is introduced to adaptively weight and enhance key global features. Finally, a fully connected layer is employed to achieve intelligent classification of bearing fault states. Validation on a laboratory test dataset shows that the proposed method achieves an average diagnostic accuracy of 98.67%, outperforming existing benchmark models and exhibiting strong generalization ability. This study provides an effective and practical intelligent fault diagnosis scheme for bearings in electric drive systems.

The fuel consumption of power-split hybrid electric vehicles (PSHEVs) can be effectively reduced via mode transition that includes the engine process. However, factors such as engine torque ripple, system parameter uncertainties, and variations in torsional vibration characteristics can easily induce drivetrain vibration. These factors not only degrade ride comfort but also lead to a fundamental control challenge. The inherent trade-off between rapid response and stability is difficult to reconcile. In addition, the lack of adaptive mechanisms further limits consistent performance under varying conditions. To tackle these problems, a scenario-adaptive composite robust control (SACRC) strategy is proposed. The strategy consists of a UIO (unknown input observer)-based torque observation module, an adaptive VSS-LMS approach, and an H∞ controller with self-tuning parameters. Firstly, a six-degree-of-freedom dynamic model of the PSHEV transmission system is established with excitation sources, considering the characteristics of dual elastic elements. Secondly, a UIO-based torque observer is designed using a simplified dual-elastic-element model. By using engine speed and output shaft speed, the observer can accurately identify the torque transmitted by the torsional damper and drive shaft. Then, an adaptive VSS-LMS and H∞ controller with self-tuning parameters is constructed to ensure a balanced performance between fast torsional vibration suppression and control stability. Finally, simulation and experimental results demonstrate that the proposed strategy provides favorable adaptability to complex scenarios, and unifies the performance goals of rapidity, stability, and robustness.

Bearings serve as core yet vulnerable components in rotating machinery, making accurate faultdiagnosis indispensable for safeguarding equipment operational safety and prolonging service life.However, transformer-based bearing fault diagnosis models are often limited by two key challenges:local feature distortion caused by rigid data segmentation, and insufficient modeling of the intrinsictemporal evolution of signals. To address these issues, this study proposes an augmented VisionTransformer (ViT) framework termed MSFEformer. Specifically, 1D convolutions are integrated toenhance the patch embedding of ViT, enabling the capture of local dynamic features from input signalsprior to tokenization. Moreover, a Bidirectional Long Short-Term Memory (Bi-LSTM) network isembedded into the multi-head attention module during query and key vector generation, therebyimproving the encoder’s capability to capture temporal dependencies and extract multi-scale features.Additionally, the Optuna framework is adopted to optimize the key hyperparameters of MSFEformer.Experimental validations on the Case Western Reserve University (CWRU) bearing dataset and a real-world test platform demonstrate that MSFEformer achieves high diagnostic accuracy and excellentgeneralization performance, outperforming state-of-the-art counterparts. The proposed frameworkprovides a reliable technical solution for enhancing the operational safety of rotating machinery andextending its service life, thus holding substantial practical significance and engineering applicationvalue.

Bearings serve as core yet vulnerable components in rotating machinery, making accurate faultdiagnosis indispensable for safeguarding equipment operational safety and prolonging service life.However, transformer-based bearing fault diagnosis models are often limited by two key challenges:local feature distortion caused by rigid data segmentation, and insufficient modeling of the intrinsictemporal evolution of signals. To address these issues, this study proposes an augmented VisionTransformer (ViT) framework termed MSFEformer. Specifically, 1D convolutions are integrated toenhance the patch embedding of ViT, enabling the capture of local dynamic features from input signalsprior to tokenization. Moreover, a Bidirectional Long Short-Term Memory (Bi-LSTM) network isembedded into the multi-head attention module during query and key vector generation, therebyimproving the encoder’s capability to capture temporal dependencies and extract multi-scale features.Additionally, the Optuna framework is adopted to optimize the key hyperparameters of MSFEformer.Experimental validations on the Case Western Reserve University (CWRU) bearing dataset and a real-world test platform demonstrate that MSFEformer achieves high diagnostic accuracy and excellentgeneralization performance, outperforming state-of-the-art counterparts. The proposed frameworkprovides a reliable technical solution for enhancing the operational safety of rotating machinery andextending its service life, thus holding substantial practical significance and engineering applicationvalue.

The electrification and intelligence of vehicles have led to a shift in driving and riding modes, accompanied by an increasingly severe problem of passenger motion sickness. Accurate behavior monitoring is crucial for proactive comfort regulation. However, in-vehicle cabin environments often face challenges such as sudden illumination changes, physical vibrations, and occlusions, making it difficult for traditional purely visual or full skeletal point models to balance real-time performance with robustness. To address this, this paper proposes a geometry-prior guided dual-stream behavior recognition network, GP-DSNet. The model introduces a head pose temporal analysis module, which utilizes high-precision Euler angle sequences to construct a geometric prior stream, providing physical constraints for visually similar actions and thereby effectively disentangling motion-sickness-related behaviors. Building upon this, we design a hierarchical cross-stream attention fusion mechanism. This mechanism achieves adaptive alignment and deep integration of the visual and geometric streams across spatial and temporal dimensions, thereby significantly enhancing the model's robustness against dynamic noise like vehicle vibrations. Furthermore, windowed attention is adopted to optimize feature contribution weights, which further reduces parameter redundancy and enhances the model's deployment efficiency in resource-constrained in-vehicle scenarios. Experimental results based on a real-world driving scenario dataset demonstrate that the proposed method achieves a significant improvement in classification performance. It attains an overall accuracy of 91.05%±8.9% and a macro-average F1-score of 91.9%±8.0%, surpassing mainstream models such as ViViT and R3D-18. This provides a reliable and efficient perceptual foundation for proactive motion sickness defense systems in intelligent cabins.

This study addresses the limitations of existing collaborative control systems for mixed traffic environments where connected and automated vehicles (CAVs) coexist with human-driven vehicles (HDVs), which overemphasize functional safety and energy efficiency loops while neglecting comfort. We propose a hierarchical model predictive control (HMPC) framework incorporating occupants’ motion sickness. The upper layer generates globally optimal speed sequences through dynamic prediction of signal phases, while the lower layer adopts a variable-weight MPC optimization method with a composite cost function integrating travel time, delay, and motion sickness indicators. To address varying CAV penetration rates in mixed traffic, heterogeneous vehicle dynamics models are developed, where CAVs and HDVs employ Cooperative Adaptive Cruise Control (CACC) and Intelligent Driver Model (IDM), respectively. The simulation evaluation results demonstrates that the proposed method achieves significant performance enhancements across diverse CAV penetration rates and traffic saturation scenarios: traffic efficiency is improved by 6.30% and 13.94%, while motion comfort is improved by 51.91% and 25.07%. Field evaluation at the Dongfeng-Huayuan Road intersection in Zhengzhou further confirms these findings, showing 28.97% and 37.87% reductions in travel time and delay, together with 57.81% and 18.18% declines in MSDV and RMS-Jerk, thereby confirming the control strategy’s robustness in real-world perturbed environments.

With the development of electrification and intelligent technologies in vehicles, ride comfort issues represented by motion sickness have become a key constraint on the performance of autonomous driving. The occurrence of motion sickness is influenced by the comprehensive movement of the vehicle in the longitudinal, lateral, and vertical directions, involving ACC, LKA, active suspension, etc. Existing motion sickness control method focuses on optimizing the longitudinal, lateral, and vertical directions separately, or coordinating the optimization control of the longitudinal and lateral directions, while there is relatively little research on the coupling effect and coupled optimization of the longitudinal and vertical directions. This study proposes a coupled framework of ACC and active suspension control system based on MPC. By adding pitch angle changes caused by longitudinal acceleration to the suspension model, a coupled state equation of half-car vertical dynamics and ACC longitudinal dynamics is constructed to achieve integrated optimization of ACC and suspension for motion suppression. The suspension active forces and vehicle acceleration are regulated coordinately to optimize vehicle vertical, longitudinal, and pitch dynamics simultaneously. Simulation experiments show that compared to decoupled control of ACC and suspension, the integrated control framework can be more effective. The research results confirm that the dynamic coordination between the suspension and ACC system can effectively suppress the motion sickness, providing a new idea for solving the comfort conflict in the human vehicle environment coupling system.