The interior permanent magnet synchronous motor (IPMSM) is known for its high output torque, strong overload capacity, and high power density, making it a popular choice in the electric vehicle industry. This paper proposes an improved multi-objective artificial hummingbird algorithm that combines chaotic mapping, adaptive weights, and dynamic crowding entropy. An optimization strategy that combines the Taguchi method with the Improved Multi-Objective Artificial Hummingbird Algorithm (IMOAHA), is proposed to minimize torque ripple and back electromotive force in the interior permanent magnet synchronous motor while simultaneously increasing the average torque of the motor. Taking the 8-pole 48-slot interior permanent magnet synchronous motor as an example, the optimization objectives include back electromotive force, average torque, and torque ripple. The rotor-related structural parameters are used as optimization variables. First, the Taguchi method is employed to identify parameters that significantly influence the optimization objectives. Subsequently, response surface fitting is used to establish the relationship between the optimization objectives and parameters. Finally, the multi-objective artificial hummingbird algorithm is utilized for optimization. By comparing the finite element analysis of the motor models before and after optimization, it is evident that the improved multi-objective artificial hummingbird algorithm can effectively enhance the performance of the interior permanent magnet synchronous motor.

Due to low contrast, chromatic aberration, and generally small objects in underwater environments, a new underwater fish detection model, BGLE-YOLO, is proposed to investigate automated methods dedicated to accurately detecting underwater objects in images. The model has small parameters and low computational effort and is suitable for edge devices. First, an efficient multi-scale convolutional EMC module is introduced to enhance the backbone network and capture the dynamic changes in targets in the underwater environment. Secondly, a global and local feature fusion module for small targets (BIG) is integrated into the neck network to preserve more feature information, reduce error information in higher-level features, and increase the model’s effectiveness in detecting small targets. Finally, to prevent the detection accuracy impact due to excessive lightweighting, the lightweight shared head (LSH) is constructed. The reparameterization technique further improves detection accuracy without additional parameters and computational cost. Experimental results of BGLE-YOLO on the underwater datasets DUO (Detection Underwater Objects) and RUOD (Real-World Underwater Object Detection) show that the model achieves the same accuracy as the benchmark model with an ultra-low computational cost of 6.2 GFLOPs and an ultra-low model parameter of 1.6 MB.

In this paper, a comprehensive method is proposed to derive the boost DC–DC converter from a given gain formula. The given gain formula is obtained by analyzing, generalizing, and summarizing previous boost structures in the literature. The analysis is based on the volt-second balance theory of inductors. Thus, the gain formula is only related to two elements, the connection between components and the duty cycle of the switches. Therefore, the topology corresponding to the gain can be derived and synthesized according to the connection of the inductors and the capacitor components during the commutation of the switch to meet the demand for different boost converters in different applications. Meanwhile, all the three-order gain formulas generated by this method are analyzed and three of them are selected for topology simplification and focus analysis, and the correctness of the selected topologies is verified by the simulation results and experimental results.

The rapid expansion of vehicular networks has intensified congestion and privacy risks. Although Federated Learning (FL) addresses both challenges through decentralized model training while preserving data locality, existing FL-based client selection strategies often fall short in highly heterogeneous vehicular environments. Specifically, the inherent heterogeneity in vehicular networks—characterized by diverse data distributions and system resources—complicates current methods. The imbalanced nature of local data further exacerbates model divergence, resulting in degraded overall system performance. To solve the above problems, in the process of vehicle selection, this paper comprehensively considers three kinds of heterogeneity and proposes an FL model with adaptive proximal term. The weight coefficient of this model is dynamically adjusted based on the difference between the local and global model parameters. Based on this, a contribution score-based vehicle selection strategy (CSVS), considering the dynamics of vehicles, is proposed to alleviate the problem of traditional model weight divergence and minimize system cost. Experimental results on two classic datasets demonstrate that the proposed strategy significantly outperforms baseline methods in reducing system cost and improving model training performance, particularly in highly heterogeneous vehicular environments.

Wheelset bearing fault diagnosis becomes challenging under time-varying train speeds, where traditional diagnostic methods often fail to remain effective. To address this issue, a multi-source physical feature graph attention network based on vibration–rotational speed data fusion is proposed to identify wheelset bearing failures under time-varying speed conditions. The method fuses three types of information to enhance rotational speed adaptability. First, a two-channel convolutional neural network is constructed to extract features from vibration signals and key-phase signals. Next, a self-attention mechanism combined with global average pooling is introduced to fuse the two feature sources. To further obtain effective information, a recursive graph auto-encoder exploits the recursiveness of vibration to generate inter-sample relationship graphs, enhancing feature interpretability. Finally, a graph attention network is used to fuse the extracted features with the recurrence graph. Experiments demonstrate that the proposed method achieves higher accuracy and rotational speed adaptability under time-varying speed conditions.