The geometric design of flow channels in bipolar plates is one of the critical features of proton exchange membrane fuel cells (PEMFCs), as it determines the power output of the fuel cell and has a significant impact on its performance and durability. The function of the bipolar plate is to guide the transfer of reactant gases to the gas diffusion layer and catalytic layer inside the PEMFC, while removing unreacted gases and gas–liquid byproducts. Therefore, the design of the bipolar plate flow channel is directly related to the water and thermal management of the PEMFC. In order to improve the comprehensive performance of PEMFCs and ensure their safe and stable operation, it is necessary to design the flow channels in bipolar plates rationally and effectively. This study addresses the limitations of existing bipolar plate flow channels by proposing a new coupling of serpentine and radial channels. The distribution of oxygen, water concentrations, and temperature inside the channel is simulated using the multi-physics simulation software COMSOL Multiphysics 6.0. The performance of this novel design is compared with conventional flow channels, with a particular focus on the pressure drop and current density to evaluate changes in the output performance of the PEMFC. The results show that the maximum current density of this novel design is increased by 67.36% and 10.43% compared to straight channel and single serpentine channels, respectively. The main contribution of this research is the innovative design of a new coupling of serpentine and radial channels in bipolar plates, which improves the overall performance of the PEMFC. This study provides theoretical support for the design of bipolar plate flow channels in PEMFCs and holds significant importance for the green development of energy.

Traditional fluid network theory often underestimates pressure losses in complex pipe-bundle systems operating under vortex-dominated flow conditions, with deviations exceeding 20% in many cases. To address this limitation, this study proposes a vortex-based correction method. Three-dimensional simulations were performed on a multidirectional parallel pipe bundle to analyze vortex formation and to quantify the effects of fluid properties (viscosity and inlet velocity) and structural parameters (branch diameter, manifold cross-sectional ratio, and manifold arrangement) on pressure loss. To account for vortex-induced energy dissipation that is overlooked by conventional one-dimensional network models, an additional vortex-induced loss coefficient, α, is introduced to modify the pressure-loss formulation. Results indicate that higher viscosity, larger branch diameter, a higher manifold cross-sectional ratio, and a co-flow arrangement improve flow uniformity and prediction accuracy. Conversely, higher inlet velocities and counter-flow arrangements intensify vortex effects and increase prediction deviations. Least-squares fitting indicates that α ranges from 1.15 to 1.37. Implementation of the proposed correction reduces pressure-loss prediction errors to within 5%, demonstrating the method’s effectiveness and extending the applicability of fluid network theory to vortex-dominated flows.

Against the backdrop of the digital transformation in engineering education, this study developed and implemented a Learning Analytics (LA)-driven and Artificial Intelligence (AI)-supported blended learning model to address structural challenges in the “Combustion” course, including highly abstract theories, experimental safety risks, and compressed instructional hours. Moving beyond mere technical stacking, the model establishes a closed-loop data ecosystem that integrates “pre-class adaptive diagnosis, in-class contextualized internalization, and post-class personalized transfer,” while deeply embedding engineering ethics and sustainability issues related to carbon neutrality. A one-semester quasi-experimental study (Experimental N = 60, Control N = 60) was conducted, utilizing a triangulated assessment of final exam scores, platform-based behavioral trajectories, and semi-structured interviews. The results showed that the experimental group achieved significantly higher final assessment scores than the control group (82.4 ± 5.7 vs. 73.2 ± 6.9), with normality tests supporting the use of parametric analysis and Analysis of Covariance (ANCOVA) indicating a significant instructional effect after controlling for Grade Point Average (GPA) and pre-test scores. Furthermore, behavioral analysis confirms that the LA mechanism significantly enhances students’ self-regulated learning and engagement by increasing the visibility of the learning process. This study provides an evidence-based reform paradigm for engineering curricula to achieve the synergistic cultivation of knowledge acquisition, competency development, and value alignment within constrained instructional timeframes.