My research activities

Transcription

1 My research activities Gaetano Sardina Division day, 29 th August 218

2 Agenda Challenges with mul1phase flows Bubbles Droplets Par1cles Some new ideas/possible collabora1ons

3 Challenges Mul1-physics: fluid dynamics- chemistry- heat transfer- mass transfer- phase change-fluid/structure interac1ons Mul1-scale: turbulent regime in industrial processesà Many temporal and length scales from the large scales propor1onal to the apparatus length down to Kolmogorov dissipa1ve scales High volume frac1ons: opaque systems where the classic laser-based experimental techniques fail

4 Bubble dynamics

5 Which technique to use? models EE RANS/ STOCHASTIC APPROACH experiments CFD FR DNS EL DNS PIV/CTA/LDS (very dilute systems) X ray MRI EE/EL LES IDEAL SIMULATION > YEAR 27 >SIZE GAP models and comparisons Scale [m]

6 d p Fully resolved DNS VOF Open Source Code- GERRIS Evaluate drag and liq coefficient Detailed dynamics around the single par1cles Limit: DNS low Reynolds number/ small number of par1cles evolved Niklas Ph.d. project

7 d p Fully resolved DNS VOF Open Source Code- GERRIS Evaluate drag and liq coefficient Detailed dynamics around the single par1cles Limit: DNS low Reynolds number/ small number of par1cles evolved Niklas Ph.d. project

8 Eulerian-Lagrangian vs Eulerian-Eulerian d p

9 Eulerian-Lagrangian vs Eulerian-Eulerian EL Filtered EL EE-OpenFoam (developer Klas Jareteg FCC)

10 Droplet condensa1on/evapora1on Clouds are highly turbulent Re 1 6 Climate/Weather predic1on models are essen1ally RANS A small scale varia1on can deeply influence larger scales

11 Droplet condensa1on/evapora1on σ R2 [µm 2 ] A B A B C C D D 1 1 t[s] <s r 2 >[µm 2 ] t 1/ State of the art DNS (124^3+1 9 t[s] lagrangian droplets) Original stochas1c LES model for droplet phase changes (Sardina et al, PRL, 215) Large effects due to turbulence Novel predic1on for the droplet radius growth The new theory has been verified by experimental measures (Chandrakara et al, PNAS, 216 )

12 2 Par1cle transport in turbulent flows 4 A. Nowbahar, G. Sardina, F. Picano and L. Brandt 712 (a) (b) (b) X X2 2 F IGURE 5. (Colour online) Wall-normal slices of an instantaneous configuration of the streamwise velocity field (contours, indicating the non-dimensional instantaneous value) and 3 (black particles have positive vertical velocity and move upward, light grey particles particles 6 with St+ = 25 for the flat (a) and the have negative vertical2 velocity and4move downward) rough (b) cases. by at fasty+streamwise F IGURE 2. (Colour online) Snapshot of an x zchannel plane centre, in thecharacterized buffer layer = 15 forvelocity, and avoid the slow and highly vortical flow regions close to the walls. (a) Newtonian case and (b) polymeric flow at Wi = 5. Black dots represent particles in the The of combined effect of the reduction of the particle concentration close to the slab 1 6 y+ 6 2, and colours indicate the magnitude the streamwise velocity component, wall and its augmentation in the channel bulk region deeply affects the total particle lighter tones being highest and darker lowest. mass flux. In order to quantify this flux, we introduce a new physical quantity named p p p particle bulk Reynolds number Reb = Vb /, where Vb is the particle bulk velocity, PNp p p defined as the averaged particle velocity in the streamwise direction (Vb = i=1 Vx1 /Np, the Newtonian and viscoelastic cases and comparing data unconditioned where Vxp1 the is the pth with particlethe instantaneous streamwise velocity and Np is the total fluid velocity p.d.f.s. number of particles). Figure 6 shows the behaviour of the steady-state particle bulk Reynolds migration number for the six particle populations, relating to the flat (solid line) The turbophoretic drift induces a mean particle towards the wall that + and the rough (dashed line) simulations. For St =, the Lagrangian tracer limit is quantified by the mean particle concentration c/c, defined as the ratio of the is recovered and the particle bulk Reynolds number is equal to the bulk Reynolds p p time-averaged particle number per unit volume normalized by phase the bulk number of the carrier (Reb =concentration. 172 and Reb = 288 for the rough and the flat + respectively). Thus, the mass direction flux for the yflat-wall case is larger than that of Figure 3 shows the concentration profile c/ccase, the wall-normal. against the rough-wall case, because of the roughness function. In the flat-wall case, the For the Newtonian and polymer flows we observe high concentrations of particles larger the particle inertia + (by increasing the Stokes number) the lower the particle at the wall (turbophoresis) as compared withmass theflux, bulk. For St = 1 (figure the accumulating particles (St+ = 25). assuming its minimum value for3a), the most particles are dispersed throughout the channel with only slight accumulation at the wall. Most interestingly, the addition of polymers leads to a decrease in concentration

13 Possible collabora1ons Particle transport in an urban environment PM (1/2.5) preferential accumulation Stable stra1fied boundary layer Code for complex geometry

14 Possible collabora1ons Developing a new multiphase (VOF/diffuse interface) compressible code Bubble implosion/droplet phase changes Compressible schemes (TVD, WENO) to capture shock waves/density-viscosity jumps across the interface

15 Drag reduc1ons for marine applica1ons DNS of turbulent channel flows/boundary layer Polymeric flows (Fene-p model) Microbubble injec1ons Superhydrophobic surfaces