As the operating speed of high-speed trains continues to rise, aerodynamic noise has become a critical factor influencing train performance and passenger comfort, especially in confined environments like tunnels. In this study, a comprehensive simulation of the aerodynamic noise generated during the entire ‘open air-tunnel-open air’ transition at 450 km/h was performed using the improved delayed detached eddy simulation (IDDES) turbulence model and overset grid method, with results validated against field measurements. The findings reveal that, compared to open-air conditions, the tunnel environment induces higher-flow velocities around the train, stronger pressure fluctuations on the train surface, and more pronounced vortex shedding in the wake. Sound pressure levels (SPLs) at surface measurement points are typically about 10 dB(A) higher in the tunnel, with the primary differences observed in the 100–300 Hz frequency range, while both scenarios exhibit spectral peaks near 530 Hz. In the bogie region, tunnel noise is dominated by pressure-wave coupling and wall reflections, which amplify low- and mid-frequency components, whereas under open-air conditions, local vortex shedding and wake instabilities prevail, resulting in more high-frequency noise. The web structure contributes to the reduction of aerodynamic noise from the bogie area, with a particularly significant decrease observed in tunnel conditions. Along the tunnel wall, the SPL initially increases and then decreases, reaching a peak of 125.7 dB(A). After exiting the tunnel, the measurement points at Section L in the open air show a reduction of approximately 10 dB(A) compared to the corresponding tunnel-wall points. This study highlights the substantial amplification of low-frequency noise caused by wall confinement and demonstrates the noise-reduction potential of the web structure. The results provide valuable theoretical insights and numerical evidence for aerodynamic noise control in higher-speed train tunnels, offering significant engineering implications.

Flapping wing micro aerial vehicles (FWMAVs) with flexible wings offer unique advantages across multiple scenarios due to their high energy efficiency and precision capabilities. This study explores the fluid-structure interaction (FSI) mechanisms and aerodynamic performance differences among three flexible wing motion patterns – passive twisting (PT), chordwise active twisting (CAT), and ‘Figure-8’ active twisting (FAT) – using a bidirectional FSI numerical simulation platform. A bionic wing model (aspect ratio AR = 3.86) and a Multiphysics-coupled framework were developed to evaluate the effects of dynamic wing torsion on lift and energy recovery. Results show that PT enhances aerodynamic performance by delaying flow separation and stabilising leading-edge vortices (LEVs). At 4 m/s, the average lift of flexible wings (FW) with PT is 6.31 times higher than that of rigid wings (RW). Active twisting strategies further improve efficiency: CAT increases average lift by 143% compared to PT at θmax = 40°, while FAT achieves 14.9% energy recovery rate through wake capture and elastic potential energy release, with an elastic energy release rate 2.79 times higher than CAT. Vortex dynamics analysis reveals that active twisting optimises lift by enhancing LEV circulation and proximity to the wing surface. CAT strengthens LEV attachment near the wing root, while FAT stabilises vortices at the wingtip. This research provides insights into energy efficiency optimisation and active control strategies for FWMAVs, highlighting the benefits of flexible deformation and intelligent motion regulation in improving aerodynamic performance and energy management.

Continuous manifold systems (CMS) are widely used for energy and mass transfer. The flow distribution of the branch ports directly affects the flow pattern of the chamber. However, the current understanding of the flow distribution in CMS under unsteady flow conditions and the optimization of their geometry is still limited. In this study, three-dimensional numerical simulations were performed to investigate the flow distribution of branch ports under unsteady flow conditions. Moreover, this study analyzed the effects of the geometric parameters of CMS on the non-uniformity coefficient (α) and total pressure drop (ΔP) under the maximum inlet flow condition (Q = 23.40 L/s). The results show that (i) from 0 to 12 s, the flow ratio (β) at the branch ports decreases in the direction of the fluid flow and then increases in the same direction from 12 to 90 s; (ii) as flow growth coefficient (η) increases, α for the maximum inlet Reynolds number (Re) increases slightly, and the difference in α corresponding to the same Re at different time points decreases; (iii) the increase in area ratio (AR) causes the flow rate of the first branch port to rise and the flow of the terminal branch port to decrease. The value of α increases linearly with AR: α = 0.23AR−0.11; (iv) The most economical approach involves adopting a linearly tapered branch ports width design with b 1/b 10 = 1.50, which decreases α by 65.04% while increasing the ΔP by 9.93% compared to the original design. This study can provide a reference for engineers in designing CMS with unsteady flow.

This study presents a comprehensive numerical investigation into steady-state mixed convection heat transfer within a square ventilated cavity containing a centrally positioned isothermal cold cylinder. The objective is to assess the combined effects of nanofluids and magnetic fields on thermal performance. The working fluids considered include pure water and water-based nanofluids enhanced with copper (Cu) and aluminium oxide (Al2O3) nanoparticles. Simulations were conducted across a range of Richardson numbers (0.1 < Ri < 100), Hartmann numbers (0 < Ha < 100), and nanoparticle volume fractions (0% < φ < 8%), using the finite volume method and the SIMPLER algorithm. Distinct from prior studies, this work bridges two gaps: (i) quantifying how high magnetic fields (Ha > 50) diminish nanoparticle-enhanced heat transfer and (ii) integrating artificial intelligence not only for prediction but also optimisation. Specifically, three machine learning models Decision Tree (DT), K-Nearest Neighbors (KNN), and a Deep Neural Network optimised via Genetic Algorithm (DNN-GA) were trained on 160 high-fidelity simulation datasets to estimate the average Nusselt number. Results demonstrated the DNN-GA’s superior accuracy (R² = 0.999, RMSE = 0.021) over DT (R² = 0.978) and KNN (R² = 0.921). Furthermore, five metaheuristic algorithms Queuing Search Algorithm (QSA), Barnacles Mating Optimiser (BMO), Search and Rescue (SAR), Gradient-Based Optimiser (GBO), and Manta Ray Foraging Optimisation (MRFO) were applied to maximise heat transfer. Optimisation identified Cu nanoparticles at Ri = 109.7, Ha = 9.0, and φ = 6% as optimal (Nu = 34.95), validated experimentally with 0.89% error. The findings confirm that increasing Ri and Ha enhances heat transfer efficiency (by 12–18%), while nanoparticle contribution declines (to 3–5%) at higher Ha. This work offers a dual contribution: advancing understanding of MHD nanofluid interactions in ventilated cavities and demonstrating a robust AI-driven framework for thermal system design. Highlights Analysis of mixed convection in a ventilated cavity using Cu-water and Al2O3-water nanofluids under varying Richardson and Hartmann numbers. Examination of magnetic field impacts on heat transfer and nanofluid flow. Comparative study of Al2O3 and Cu nanoparticles on heat transfer enhancement. Provides valuable insights into the combined effects of nanoparticles, magnetic fields, and convection parameters. Machine learning models are very useful for predicting the Nusselt number. Metaheuristics algorithms are highly effective in optimising heat transfer processes.

To look inside a pulse tube for streaming effects beyond 5–6 Hz is not possible experimentally. Therefore, CFD is a reliable tool for observing the streaming effects occurring in a cryocooler. In the present work, a numerical investigation on the inertance type pulse tube cryocooler (IPTCC) is carried out using computational fluid dynamics (CFD) at frequencies greater than 80 Hz. A numerical model is developed, and simulations are validated against the numerical results published in the open literature. As opposed to the local thermal non-equilibrium model, porous media are modelled using a local thermal equilibrium approach. The streaming effects are analysed in the pulse tube for various sizes and frequencies with temperature-dependent material properties. It was observed that Rayleigh streaming is most likely to occur when the viscous penetration depth is very small compared to the pulse tube radius. Further, it was also examined that streaming is purely acoustical, not due to the transition to turbulent flow in the pulse tube. While varying the whole length of IPTCC, Rayleigh streaming occurred at a particular combination of size and frequency. Furthermore, the streaming at different time periods was analysed for two cases of different IPTCC lengths. It was found that a combination of various factors and flow conditions resulted in the appearance and disappearance of vortices in the pulse tube in various instances. The present study also suggests an interplay between the acoustic power density, size, and frequency of the IPTCC. It is concluded that CFD modelling is necessary to observe the multidimensional effects and complex transport phenomena in IPTCC.

Accurate and efficient prediction of spatiotemporal variations in the distribution of substances in fluids (SIFs) is crucial for various aspects of fluid mechanics related research and applications, involving for instance, material transport quantification, water quality assessment, and engineering condition analysis. This study proposes a framework for resolving the spatiotemporal distribution of SIFs such as salt and suspended sediment based on water levels and flow velocities. The framework incorporates a deep learning model based on a classic neural operator (DeepONet) architecture, which consists of a feature network and a position network to encode the characteristics of input variables and the problem domain. Numerical simulations were performed to generate the needed datasets. The framework was well-validated by predicting salinity and suspended sediment concentration (SSC) distributions in two idealized cases and a real-word case, demonstrating its efficacy and robustness. Time-series validation further demonstrated the prediction accuracy of the framework. The deep learning model is also capable of enhanced-resolution predictions, enabling the generation of high-resolution spatial distributions of SIFs from low-resolution hydrodynamic data. Both bottom and surface layers of the water column were analyzed, revealing that the mapping relationships between hydrodynamics and SIF distributions can be accurately captured throughout the water column, despite variations in correlation coefficients. Due to these capabilities and advantages, additional data sources can be integrated into the framework in the future, highlighting its considerable potential for broader applications in aquatic environments.

To meet the underwater drag reduction needs in engineering applications, one can draw on the drag reduction strategies of marine animals such as sharks, utilizing their surface groove structures and the secretion of viscoelastic drag-reducing mucus to achieve drag reduction effects. However, in the field of engineering research, efficient and accurate simulation methods for viscoelastic fluid flow are relatively scarce. There is an urgent need to develop efficient numerical simulation methods for viscoelastic fluids, and to analyse the drag reduction characteristics of the coupling between surface grooves and drag-reducing agents. Through large eddy simulation and re-development of the ANSYS Fluent platform, efficient and accurate simulation of viscoelastic fluid flow as well as wall infiltration method has been successfully realized. Quantitative analysis of drag reduction characteristics of surface structures and viscoelastic fluid additives within the Reynolds number range of 4×104 to 4×105 has been conducted. Moreover, the synergistic effect of micro-grooves and drag reducer infiltration on drag reduction has been deeply explored. It is found that the maximum drag reduction by the groove surface solely is below 20%, whilst the maximum drag reduction by viscoelastic drag-reducing agent infiltration on the smooth flat plate is up to 28.9%. Notably, when combining the groove structure with drag-reducing agent infiltration, the maximum drag reduction reaches up to approximately 70%. The mechanism for the drag reducing effect is analysed from the aspects of vortex structure evolution on the wall, velocity profile and drag-reducing agent diffusion. This establishes an efficient and accurate simulation method for viscoelastic fluid flow suitable for engineering research, providing an important reference and foundation for underwater drag reduction studies.

Tropical cyclones (TCs) are destructive meteorological phenomena, necessitating accurate predictions of TC track and intensity to reduce risks to human life. This study evaluates three Transformer-based models – vanilla Transformer (Transformer), inverted Transformer (iTransformer), and temporal-variate Transformer (TVFormer) – which are trained, validated, and tested on best track data from 1980 to 2021 from the China Meteorological Administration for TC prediction, integrating temporal, variate, and hybrid token embeddings to analyze temporal and variate correlations. Comparative analysis with four recurrent neural network (RNN) models demonstrates the superiority of the refined Transformer models over RNNs: iTransformer reduces mean absolute error (MAE) and root mean square error (RMSE) by 29.55% and 25.80% (latitude), 50.31% and 46.18% (longitude), 8.71% and 9.98% (pressure), and 8.68% and 9.45% (wind speed), while TVFormer achieves MAE and RMSE reductions of 13.98% and 13.84% (latitude), 39.11% and 38.02% (longitude), 13.69% and 14.02% (pressure), and 12.84% and 12.94% (wind speed) on average. Among Transformer variants, iTransformer excels in track prediction, outperforming Transformer with 21.74% lower MAE and 18.26% lower RMSE for latitude, and 32.73% lower MAE and 24.01% lower RMSE for longitude. TVFormer dominates intensity prediction, reducing pressure errors by 4.42% (MAE) and 3.92% (RMSE) and wind speed errors by 19.21% (MAE) and 14.79% (RMSE) compared to Transformer, while outperforming iTransformer with 4.59% lower MAE and 3.68% lower RMSE for pressure and 3.83% lower MAE and 3.18% lower RMSE for wind speed. Notably, TVFormer also enhances track prediction, with 7.10% reduction in MAE and 7.02% reduction in RMSE for latitude, and 22.84% reduction in MAE and 17.09% reduction in RMSE for longitude compared to Transformer. These results highlight the superiority of iTransformer in track prediction and the efficacy of TVFormer in intensity prediction, thanks to their ability to exploit temporal and variate dependencies, offering potential for TC disaster preparedness systems.

The exploration of deep learning methodologies has recently generated significant interest in the use of Physics-Informed Neural Networks (PINNs) to address complex physical problems governed by partial differential equations (PDEs). The PINN is trained using information from physical laws, including governing PDEs, boundary conditions, and initial conditions. However, achieving a well-trained PINN typically necessitates an appropriate balance between the weights of each loss function, which can considerably increase manual effort. This paper introduces a novel training approach that integrates non-dimensionalization with a multi-task learning technique, termed Automatic Weighted Loss (AWL), to autonomously achieve an optimal balance for each loss function. In the baseline PINN training for solving time-dependent PDEs, multiple weights (usually more than six) must be manually tuned for the model, whereas this method can reduce the number of scaling weights to only one. The proposed approach, referred to as the Non-dimensionalization Automatic Weighted Loss (NDAWL), is evaluated through six free-surface flow problems modelled by the Shallow Water Equations (SWEs). Furthermore, a comparative analysis is conducted between the solutions obtained using NDAWL-PINN and those from the original PINN, which relies on manual fine-tuning of loss functions. The numerical results indicate that the NDAWL-PINN method achieves comparable or superior accuracy to the original PINN, demonstrating its effectiveness in automating the balancing of loss functions.

Energy harvesting from ambient sources has gained attention due to increasing energy demands. Despite VIV-based harvesters showing significant potential, their lock-in region, where significant power is generated, is narrow. Given the continuously varying ambient conditions of fluid currents, harvesters can easily fall into de-synchronization, yielding low energy output. Existing solutions like tunable masses or multiple degrees of freedom systems increase complexity and weight, limiting practical applications. This work introduces a novel variable-diameter cylinder mechanism–a practical technique that actively tunes the cylinder's geometry in real time to enhance energy harvesting efficiency from VIV. The mechanism employs an expanding pulley system that deforms the elastic circular cylinder radially, dynamically altering its diameter to adjust key non-dimensional parameters governing VIV. This real-time adaptability counteracts ambient fluctuations, significantly widening the lock-in range. The cylinder incorporates a piezoelectric patch, and the fluid-structure-piezoelectric interaction problem was analyzed numerically to determine the cylinder motion, voltage, and power output. A partitioned Lagrange-Eulerian approach was employed, coupling a fluid solver with a custom-coded structural piezoelectric solver via the preCICE library, which enables data mapping and exchange between solvers at run time. The analysis was conducted for different diameter profiles, and the results showed an enhancement in the harvester synchronization width and maximum amplitude by 70% and 117%, respectively, compared to the constant diameter case. Additionally, a 113% increase in peak voltage was achieved, and more than 8 times greater power was generated. The analysis also showed the effect of load resistance on harvesting performance, and the varying diameter cases demonstrated resilience against the shunt-damping effect. The proposed control technique is versatile, as it can be used for VIV suppression as well as energy harvesting applications.