Embedding the intrinsic symmetry of a flow system in training its machine learning algorithms has become a significant trend in the recent surge of their application in fluid mechanics. This paper leverages the geometric symmetry of a four-roll mill (FRM) to enhance its training efficiency. Stabilising and precisely controlling droplet trajectories in an FRM is challenging due to the unstable nature of the extensional flow with a saddle point. Extending the work of Vona & Lauga (Phys. Rev. E, vol. 104(5), 2021, p. 055108), this study applies deep reinforcement learning (DRL) to effectively guide a displaced droplet to the centre of the FRM. Through direct numerical simulations, we explore the applicability of DRL in controlling FRM flow with moderate inertial effects, i.e. Reynolds number $\sim \mathcal{O}(1)$, a nonlinear regime previously unexplored. The FRM’s geometric symmetry allows control policies trained in one of the eight sub-quadrants to be extended to the entire domain, reducing training costs. Our results indicate that the DRL-based control method can successfully guide a displaced droplet to the target centre with robust performance across various starting positions, even from substantially far distances. The work also highlights potential directions for future research, particularly focusing on efficiently addressing the delay effects in flow response caused by inertia. This study presents new advances in controlling droplet trajectories in more nonlinear and complex situations, with potential applications to other nonlinear flows. The geometric symmetry used in this cutting-edge reinforcement learning approach can also be applied to other control methods.

The dynamics of flow over an isolated surface-mounted hemisphere are investigated with tomographic particle image velocimetry (PIV). The 10 mm height hemisphere is completely submerged in the laminar boundary layer, and the height-based Reynolds number is 1530. The evolution of typical coherent structures around the hemisphere are discussed, with emphasis on the hairpin vortex (HV) and side hairpin vortex (SHV) formed periodically in the middle and both sides of the wake, respectively. Proper orthogonal decomposition (POD) analysis is conducted to explore the vortex dynamics. The shedding processes of the HV and SHV are each dominated by two different POD modes with correspondingly different characteristic frequencies, which has not been reported before in the literature. Furthermore, the coexistence of symmetric and asymmetric shedding patterns is explored for the first time in the shedding process of the HV at such a low Reynolds number. The asymmetric behaviour is controlled by the asymmetric shedding POD mode, whose dominant frequency is exactly half of the symmetric mode. In addition, SHVs on both sides of the wake are throughout formed and shed alternately, and the streamwise extensions of a horseshoe vortex also oscillate asymmetrically, which are responsible for the formation of the asymmetric shedding pattern of the HV. These findings help to fill the gaps in the related field and contribute to studies on the vortex dynamics of the flow over a hemisphere.

The skin-friction coefficient is a dimensionless quantity defined by the wall shear stress exerted on an object moving in a fluid, and it decreases as the Reynolds number increases for wall-bounded turbulent flows over a flat plate. In this work, a novel transformation, based on physical and asymptotic analyses, is proposed to map the skin-friction relation of high-speed turbulent boundary layers (TBLs) for air described by the ideal gas law to the incompressible skin-friction relation. Through this proposed approach, it has been confirmed theoretically that the transformed skin-friction coefficient $C_{f,i}$, and the transformed momentum-thickness Reynolds number $Re_{\theta ,i}$ for compressible TBLs with and without heat transfer, follow a general scaling law that aligns precisely with the incompressible skin-friction scaling law, expressed as $ (2/C_{f,i} )^{1/2}\propto \ln Re_{\theta ,i}$. Furthermore, the reliability of the skin-friction scaling law is validated by compressible TBLs with free-stream Mach number ranging from $0.5$ to $14$, friction Reynolds number ranging from $100$ to $2400$, and the wall-to-recovery temperature ratio ranging from $0.15$ to $1.9$. In all of these data, $ (2/C_{f,i} )^{1/2}$ and $\ln Re_{\theta ,i}$ based on the present theory collapse to the incompressible relation, with a squared Pearson correlation coefficient reaching an impressive value $0.99$, significantly exceeding $0.85$ and $0.86$ based on the established van Driest II and the Spalding–Chi transformations, respectively.

Here, we show that the thrust force of oscillating airfoils calculated within the linearised potential flow approach by means of the vortex impulse theory coincides with the one resulting from the integration of the unsteady pressure distribution around the solid obtained by Garrick (1936) when the vertical component of the wake velocity is calculated self-consistently and the analysis retains the contribution of the flux of horizontal momentum induced by the starting vortex. The limitations of the self-consistent linearised potential flow approach for predicting the thrust force of airfoils oscillating periodically with small amplitudes but large values of the reduced frequency are also discussed, as well as the reasons behind the ability of other results in the literature to approximate measurements better than Garrick’s theory. In fact, for those cases in which the airfoil oscillates periodically, the flux of horizontal momentum induced by the starting vortex is negligible and the vortices in the wake are convected parallel to the free-stream velocity, we have deduced an equation for the mean thrust coefficient which differs from previously published results and is in agreement with experimental and numerical results. In addition, for those cases in which the airfoil is suddenly set into motion, we have also deduced an equation that retains the effect of the starting vortex and correctly quantifies the transient thrust force.

The function of aortic heart valves is to prevent regurgitant flow from the aorta into the left ventricle. A higher regurgitant flow is observed in bileaflet mechanical heart valves (BMHVs) compared with bioprosthetic heart valves (BHVs) because of their delayed closure. Here, we investigate this behaviour through fluid–structure interaction simulations of a BMHV compared with a trileaflet mechanical heart valve (TMHV) and a BHV under similar conditions. We find that the TMHV and BHV begin to close during the systolic deceleration, whereas BMHV only begins to close when the flow reverses. We found this to be related to hemodynamics as the TMHV and BHV, when fully opened, generate a central jet-dominant flow, whereas the BMHV generates triple jets with lateral jets being wider than its central jet. The flow deceleration of the central jet during late systole is higher than that of the sinuses, which results in a lower pressure in the central region than the sinuses to drive the leaflets of the TMHV and BHV towards the centre for closure. Conversely, the pressure on the sinus- and central flow-sides of the BMHV leaflets is nearly the same until the end of systole. We, contrary to what classically believed, did not find any evidence of sinus vortices generating high pressure or viscous stresses to initiate valve closure. Overall, the results suggest that the generation of a strong central jet and the direction of the leaflets’ closure towards the centre are the design principles to ensure an early valve closure and minimise regurgitation.