Flows over porous walls at friction Reynolds number Reτ≈198 are investigated using the lattice Boltzmann method, where the rough and porous walls are treated as spherical arrays. Compared with the smooth wall, the wall roughness significantly enhances the Reynolds shear stress in the near-wall region due to the stronger upwash and downwash motions, and the downwash motion is stronger in the permeable-wall case. As the turbulent fluctuation is closely related to the near-wall coherent structures, the characteristic structures are investigated. In both cases, the near-wall turbulent structures and the relatively large-scale spanwise rollers are identified. Through the clustering method, the spatial distributions of the coherent structures with different wall-normal scales (ly+) are analyzed, and it is found that the near-wall vortical structures with ly+) < d (d is the roughness height) dominate below the logarithmic region (y≲0.1H, where H represents the distance from the crest of the sphere to the upper wall), while the relatively large-scale spanwise rollers with ly+≳d are distributed within the logarithmic region. Furthermore, turbulent fluctuations contributed from different scales of vortical structures are analyzed. It is observed that the fluctuation of the coherent structures is reduced in the permeable-wall case compared with the rough-wall case, and the reduction of the near-wall vortical structures is stronger. Finally, the conditional averaged flow reveals the similarity of the spatial distribution of low/high-speed structures in the rough and permeable wall, but the normal size of the high-speed fluid structure is decreased in the permeable wall, indicating the downward motion.

Computational fluid dynamics (CFD) plays a crucial role in optimizing micro-centrifugal pump geometries; however, conventional workflows often suffer from fragmentation, manual intervention, and poor interoperability among software tools. In this study, an integrated CFD-based simulation–optimization platform was developed to establish a closed-loop workflow covering parametric modeling, automated meshing, steady CFD analysis, and multi-objective optimization. Using a micro-centrifugal pump as a case study, optimal Latin hypercube sampling, a Kriging surrogate model, and a multi-objective genetic algorithm were combined to quantify the relationships between key structural parameters and hydraulic performance and to identify Pareto-optimal designs. Sensitivity analysis showed that blade count and volute throat height were the dominant factors affecting pump head and efficiency. Compared with the baseline design, the optimized schemes achieved average improvements of 40.36% in head and 22.89% in hydraulic efficiency, with Scheme 1 showing the best overall balance. Experimental validation using hydraulic performance testing and particle image velocimetry showed that the deviation between predicted and measured heads was within 10%, and the measured flow-field trends agreed well with the CFD results. The proposed framework provides a reproducible method for the design optimization of micro-centrifugal pumps and other small-scale turbomachinery.

There is a widely accepted conclusion that the wall roughness do not always enhance the heat transport of the turbulent thermal convection. In this paper, the heat transfer efficiency is statistically investigated from the perspective of turbulent structure. The effect of turbulent structure on the heat transfer of Rayleigh–Bénard convection with triangular rough element on the top and bottom plates is numerically simulated by a lattice Boltzmann method. We use a clustering method to identify complex turbulent structures associated with intense events. The reduction of the Nusselt number is obtained for small roughness height H/L, while the enhancement of heat transport appears for large H/L. For the large H/L case, the positive temperature structures $T_{cp}$ occupying the negative heat transfer events $⟨Tv{⟩}_n$ reduce the efficiency of the heat transfer. On the contrary, the negative temperature turbulent structures $T_{cn}$ boost the heat transfer. By analyzing the conditional average field, we found that the enhancement of the heat transfer for large H/L cases is due to that the negative temperature structures play a dominant role. For small H/L cases, the positive temperature structures $T_{cp}$ inhibit the heat transfer. Furthermore, the more positive and negative temperature structures for large H/L cases are generated near the solid wall and the corner of the box. The physical explanation for the Nu enhancement is that the more secondary vortices are generated by the interaction of these turbulent structures and the rough wall, leading to more plumes ejected from the boundary layers to the bulk.