Interface-resolved Direct Numerical Simulations Research Articles

Interface-resolved direct numerical simulations of the interactions between neutrally buoyant spheroidal particles and turbulent channel flows

A parallel direct-forcing fictitious domain method is employed to perform interface-resolved numerical simulations of the interactions between neutrally buoyant finite-size spheroidal particles and turbulent channel flows. The effects of the aspect ratio of the spheroidal particles on the turbulence modulation and the rotation of the particles are investigated at the friction Reynolds number of 180, with the ratio of the particle equivalent diameter to the channel width being 0.1, the particle volume fraction ranging from 0.79% to 14.16%, and the particle aspect ratio ranging from 1/3 to 8. Our results show that the flow friction decreases as the prolate particles become more slender or the oblate particles become flatter and is smaller than that of the single-phase flow for the aspect ratio being 1/3 and 8 at the particle volume fraction of 2.36%. Both effects of the low particle volume fraction in the near-wall region and the relatively small Reynolds stress are important to the occurrence of the drag-reduction by the non-spherical particles, and a lower flow drag for the oblate particles compared to the prolate particles at comparable aspect ratios (e.g., 1/3 vs 4) is mainly caused by a lower Reynolds stress contribution. The prolate particles preferentially align their symmetry axes with the streamwise direction, and the oblate particles preferentially align their symmetry axes with the wall-normal direction. However, the most probable orientation of the major axes of both prolate and oblate particles in the near-wall region is not exactly the streamwise direction but has a positive inclination angle with the streamwise direction. Generally the prolate particles have higher spinning velocities and lower tumbling velocities in the entire channel region, compared to the oblate particles.

Clustering and increased settling speed of oblate particles at finite Reynolds number

We study the settling of rigid oblates in a quiescent fluid using interface-resolved direct numerical simulations. In particular, an immersed boundary method is used to account for the dispersed solid phase together with lubrication correction and collision models to account for short-range particle–particle interactions. We consider semi-dilute suspensions of oblate particles with aspect ratio $AR=1/3$ and solid volume fractions $\unicode[STIX]{x1D719}=0.5{-}10\,\%$. The solid-to-fluid density ratio $R=1.02$ and the Galileo number (i.e. the ratio between buoyancy and viscous forces) based on the diameter of a sphere with equivalent volume $Ga=60$. With this choice of parameters, an isolated oblate falls vertically with a steady wake with its broad side perpendicular to the gravity direction. At this $Ga$, the mean settling speed of spheres is a decreasing function of the volume $\unicode[STIX]{x1D719}$ and is always smaller than the terminal velocity of the isolated particle, $V_{t}$. On the contrary, in dilute suspensions of oblate particles (with $\unicode[STIX]{x1D719}\leqslant 1\,\%$), the mean settling speed is approximately 33 % larger than $V_{t}$. At higher concentrations, the mean settling speed decreases becoming smaller than the terminal velocity $V_{t}$ between $\unicode[STIX]{x1D719}=5\,\%$ and 10 %. The increase of the mean settling speed is due to the formation of particle clusters that for $\unicode[STIX]{x1D719}=0.5{-}1\,\%$ appear as columnar-like structures. From the pair distribution function we observe that it is most probable to find particle pairs almost vertically aligned. However, the pair distribution function is non-negligible all around the reference particle indicating that there is a substantial amount of clustering at radial distances between 2 and $6c$ (with $c$ the polar radius of the oblate). Above $\unicode[STIX]{x1D719}=5\,\%$, the hindrance becomes the dominant effect, and the mean settling speed decreases below $V_{t}$. As the particle concentration increases, the mean particle orientation changes and the mean pitch angle (the angle between the particle axis of symmetry and gravity) increases from $23^{\circ }$ to $47^{\circ }$. Finally, we increase $Ga$ from 60 to 140 for the case with $\unicode[STIX]{x1D719}=0.5\,\%$ and find that the mean settling speed (normalized by $V_{t}$) decreases by less than 1 % with respect to $Ga=60$. However, the fluctuations of the settling speed around the mean are reduced and the probability of finding vertically aligned particle pairs increases.

A computational framework for interface-resolved DNS of simultaneous atomization, evaporation and combustion

A computational framework for interface-resolved direct numerical simulations (DNS) of simultaneous atomization, evaporation and combustion process is proposed. The present work utilizes a level set method to implicitly capture the gas–liquid interface and the ghost fluid method (GFM) to accurately address jump conditions across the interface. Specific care has been devoted to the discretization of the convective term and diffusive term for the species and energy equations. The level set method with a sub-cell resolution of the interface is employed and a semi-Lagrangian scheme for the discretization of the level set equation with an evaporation source term is proposed. The one-step global reaction model of n-heptane is used for the vapor combustion. To the best of our knowledge, this is the first computational framework that has the capability to simulate spray combustion with an interface-resolved method. The present framework has been validated in several cases, including the one-dimensional evaporation and the static droplet evaporation for low ambient temperature, the Stefan problem, the Sucking problem, the evaporation of static and moving droplet for high ambient temperature, and the combustion of static and moving droplet. Finally, the integrated simulation of droplet collision and combustion is performed to evaluate the accuracy and robustness of the present method.

Effects of finite-size neutrally buoyant particles on the turbulent flows in a square duct

Interface-resolved direct numerical simulations of the particle-laden turbulent flows in a square duct are performed with a direct-forcing fictitious domain method. The effects of the finite-size particles on the mean and root-mean-square (RMS) velocities are investigated at the friction Reynolds number of 150 (based on the friction velocity and half duct width) and the particle volume fractions ranging from 0.78% to 7.07%. Our results show that the mean secondary flow is enhanced and its circulation center shifts closer to the center of the duct cross section when the particles are added. The reason for the particle effect on the mean secondary flow is analyzed by examining the terms in the mean streamwise vorticity equation. It is observed that the particles enhance the gradients of the secondary Reynolds normal stress difference and shear stress in the near-wall region near the corners, which we think is mainly responsible for the enhancement in the mean secondary flow. Under a prescribed driving pressure gradient, the presence of particles attenuates the bulk velocity and the turbulent intensity. All particle-induced effects are intensified with increasing particle volume fraction and decreasing particle size, if other parameters are fixed. In addition, the particles accumulate preferentially in the near-corner region. The effects of the type of the collision model (i.e., if friction and damping are included or not) on the results are found significant, but not so significant to bring about qualitatively different results.

Turbulent channel flow of a dense binary mixture of rigid particles

We study turbulent channel flow of a binary mixture of finite-sized neutrally buoyant rigid particles by means of interface-resolved direct numerical simulations. We fix the bulk Reynolds number and total solid volume fraction, $Re_{b}=5600$ and $\unicode[STIX]{x1D6F7}=20\,\%$, and vary the relative fraction of small and large particles. The binary mixture consists of particles of two different sizes, $2h/d_{l}=20$ and $2h/d_{s}=30$ where $h$ is the half-channel height and $d_{l}$ and $d_{s}$ the diameters of the large and small particles. While the particulate flow statistics exhibit a significant alteration of the mean velocity profile and turbulent fluctuations with respect to the unladen flow, the differences between the mono-disperse and bi-disperse cases are small. However, we observe a clear segregation of small particles at the wall in binary mixtures, which affects the dynamics of the near-wall region and thus the overall drag. This results in a higher drag in suspensions with a larger number of large particles. As regards bi-disperse effects on the particle dynamics, a non-monotonic variation of the particle dispersion in the spanwise (homogeneous) direction is observed when increasing the percentage of small/large particles. Finally, we note that particles of the same size tend to cluster more at contact whereas the dynamics of the large particles gives the highest collision kernels due to a higher approaching speed.

Clustering and preferential concentration of finite-size particles in forced homogeneous-isotropic turbulence

We have performed interface-resolved direct numerical simulations of forced homogeneous-isotropic turbulence in a dilute suspension of spherical particles in the Reynolds number range $Re_{\unicode[STIX]{x1D706}}=115{-}140$. The solid–fluid density ratio was set to $1.5$, gravity was set to zero and two particle diameters were investigated corresponding to approximately $5$ and $11$ Kolmogorov lengths. Note that these particle sizes are clearly outside the range of validity of the point-particle approximation, as has been shown by Homann & Bec (J. Fluid Mech., vol. 651, 2010, pp. 81–91). At the present parameter points the global effect of the particles upon the fluid flow is weak. We observe that the dispersed phase exhibits clustering with moderate intensity. The tendency to cluster, which was quantified in terms of the standard deviation of Voronoï cell volumes, decreases with the particle diameter. We have analysed the relation between particle locations and the location of intense vortical flow structures. The results do not reveal any significant statistical correlation. Contrarily, we have detected a small but statistically significant preferential location of particles with respect to the ‘sticky points’ proposed by Goto & Vassilicos (Phys. Rev. Lett., vol. 100 (5), 2008, 054503), i.e. points where the fluid acceleration field is acting such as to increase the local particle concentration in one-way coupled point-particle models under Stokes drag. The presently found statistical correlation between the ‘sticky points’ and the particle locations further increases when focusing on regions with high local concentration. Our results suggest that small finite-size particles can be brought together along the expansive directions of the fluid acceleration field, as previously observed only for the simplest model for sub-Kolmogorov particles. We further discuss the effect of density ratio and collective particle motion upon the basic Eulerian and Lagrangian statistics.

Interface-resolved direct numerical simulation of the erosion of a sediment bed sheared by laminar channel flow

A numerical method based upon the immersed boundary technique for the fluid–solid coupling and on a soft-sphere approach for solid–solid contact is used to perform direct numerical simulation of the flow-induced motion of a thick bed of spherical particles in a horizontal plane channel. The collision model features a normal force component with a spring and a damper, as well as a damping tangential component, limited by a Coulomb friction law. The standard test case of a single particle colliding perpendicularly with a horizontal wall in a viscous fluid is simulated over a broad range of Stokes numbers, yielding values of the effective restitution coefficient in close agreement with experimental data. The case of bedload particle transport by laminar channel flow is simulated for 24 different parameter values covering a broad range of the Shields number. Comparison of the present results with reference data from the experiment of Aussillous et al. (2013) yields excellent agreement. It is confirmed that the particle flow rate varies with the third power of the Shields number once the known threshold value is exceeded. The present data suggests that the thickness of the mobile particle layer (normalized with the height of the clear fluid region) increases with the square of the normalized fluid flow rate.

Interface-resolved direct numerical simulation of vertical particulate channel flow in the turbulent regime

We have conducted a direct numerical simulation study of dilute turbulent particulate flow in a vertical plane channel by considering thousands of finite-size rigid particles with resolved phase interfaces. The particle diameter corresponds to approximately 11 wall units and their terminal Reynolds number is set to 136. The fluid flow with a bulk Reynolds number of 2700 is directed upward, which maintains the particles suspended on average. Two density ratios were simulated, differing by a factor of 4.5. The corresponding Stokes numbers of the two flow cases were O(10) in the near-wall region and O(1) in the outer flow. We have observed the formation of large-scale elongated streaklike structures with streamwise dimensions of the order of eight channel half-widths and cross-stream dimensions of the order of one half-width. At the same time, we have found no evidence of significant formation of particle clusters, which suggests that the large structures are due to an intrinsic instability of the flow, which is triggered by the presence of the particles. It was found that the mean fluid velocity profile tends toward a concave shape, and the turbulence intensity and the normal stress anisotropy are strongly increased. The effect of varying the Stokes number while maintaining the buoyancy, particle size, and volume fraction constant was relatively weak.