Volumetric distribution and velocity of inertial particles in a turbulent channel flow

Transcription

1 Volumetric distribution and velocity of inertial particles in a turbulent channel flow Filippo Coletti 1,*, Mostafa Toloui 2, Kee Onn Fong 1, Andras Nemes 1, Lucia Baker 1 1: Department of Aerospace Engineering & Mechanics, University of Minnesota, Minneapolis (MN), USA 2: Department of Mechanical Engineering, University of Minnesota, Minneapolis (MN), USA * Correspondent author: fcoletti@umn.edu Keywords: particle-laden flow, turbulence, turbophoresis, two-way coupling, particle tracking velocimetry, digital in-line holography ABSTRACT The segregation of inertial particles in specific regions of a turbulent fluid flow is a well known phenomenon, but experimental observations of its three-dimensional nature have lacking. Here we are concerned with the transport of small inertial particles in a vertically oriented turbulent channel flow. The working fluid is air laden with sizeselected glass particles. We focus on a regime in which both preferential concentration/turbophoresis as well as twoway coupling are expected to be substantial. We measure statistics of particle and fluid using two-dimensional imaging, and we reconstruct volumetric distributions of the inertial particles using digital in-line holography. The inertial particles are found to accumulate near the wall, but in a much less extreme fashion than what simulations indicate. The particles have a flatter velocity profile and higher velocity fluctuations compared to the fluid. Using Voronoi analysis we find evidence of strong particle clustering especially in the near-wall regions, indicating that the preferential concentration is driven by the near-wall turbulence structures. 1. Introduction Particle imaging techniques are widely used to characterized single-phase velocity fields, and in recent years their capabilities in capturing large ensembles of tracer particles in 2 and 3 dimensions have greatly improved. However the latest advances, especially in volumetric imaging, have not been exploited to reconstruct the motions of inertial particles immersed in the fluid. The latter is a problem of utmost significance which recently has been studied mostly by numerical simulations. Notably, the clustering of inertial particles in specific regions of a turbulent flow is a well known phenomenon with far reaching consequences, but experimental observations of its three-dimensional nature are lacking. Here we are concerned with the transport of small inertial particles in wall-bounded turbulence, in particular in the case in which the working fluid is a gas. In this situation the particle-to-fluid density ratio is high, and even micron-sized particles have significant inertia. It is well known that inertial particles in turbulence experience preferential concentration or

2 clustering, i.e. collection in high-strain, low-vorticity regions (Maxey 1987, Squires and Eaton 1991). This mechanism is most effective when particles have sufficient inertia to depart from fluid streamlines, but not enough to take ballistic trajectories, and is primarily dictated by the Stokes number St = τp/τf, the relative strength of the aerodynamic response time of the particle τp to a fluid timescale τf. The latter is typically taken to be either the Kolmogorov time scale or the viscous time in wall units. In wall-bounded turbulence, clustering typically occurs on scales associated with streaks and coherent structures in the inner wall region (Rouson and Eaton 2001), and the collective interaction with such structures results in so-called turbophoresis, i.e. a significant drift towards the wall (Marchioli and Soldati 2002). This is typically most effective for St = Inertial particles in wall-bounded turbulent air flows are relevant to a wealth of industrial, environmental, and biological settings, from dust ingestion in aircraft engines to ocean see spray and aerosol drug delivery. In all these examples, a crucial feature is that mass, momentum, and energy of the gas phase and particle phase are comparable, so that the understanding and solution of the strongly coupled dynamics is indispensable for an accurate prediction of even first-order statistics. This is because, given sufficiently high concentrations, particles can modify turbulence through dynamic coupling between phases (two-way coupling). At even higher volume fractions (typically above 10, still well in the dilute regime) inter-particle collisions -4 become significant (Elghobashi 1994). This is especially true in the context of preferential concentration and turbophoresis, where both intermittent and mean concentration distributions can be locally very high (Sardina et al. 2012). In turn, inter-particle collisions have a strong effect on the concentration distribution, increasing the wall-normal transport and reducing the near wall-concentration of particles (Vreman 2007). Here we investigate a regime in which both preferential concentratiom (and its macroscopic consequence: turbuophoresis) as well as two-way coupling are simultaneously at play. Very few experimental studies are available that considered two-way coupling. In Tab. 1 we list the main studies that have looked at two-way coupling in what is arguably the most fundamental configuration of wall-bounded gas-particle flows, i.e. the fully developed particleladen turbulent channel flows (with air as working fluid). Of these, the studies of Kulick et al. (1994) and Paris (2000), although designed to have a smooth wall, turned out to have effectively rough walls. Moreover, most of these studies had particles of very large Stokes numbers, for which preferential concentration/turbophoresis is not expected. The study of Li et al. (2012) represents an exception, but the volume fraction was so small that the two-way coupling was probably marginal. Moreover, all previous studies have characterized particle transport uing

3 point-wise (e.g. LDV) or at most 2D (PIV) measurement techniques. In the present study we use a combination of 2D PIV/PTV and 3D PTV based on Digital In-line Holography (DIH-PTV) to quantitatively characterize the instantaneous and mean fields of particle concentration and velocity. Reτ St + d + ΦV Kulick et al. (1994) rough wall Paris (2000) rough wall Kussin & Sommerfeld (2002) Benson et al. (2005) Li et al. (2012) Present study Tab. 1 Summary of previous and present studies focused on the fully developed particle-laden turbulent channel flows, with air as working fluid. 2. Experimental facility and measurement techniques Experiments are conducted in an installation consisting of a 2.5 m long channel flow with a 0.24 m by 0.03 m cross-section (Fig. 1). The channel is vertically oriented, so that gravitational accumulation on the walls does not affect the particle-turbulence interaction. A centrifugal blower drives air downward to a bulk velocity Ubulk = 5 m/s, resulting in a friction Reynolds number of Reτ = 300. The air mass flow rate is continuously monitored through a Venturi meter. Before the air enters the channel, size-selected glass beads (density of 2.5 g/cc, 50±6 µm in diameter, see Fig. 2) are issued into the flow through a precision screw-feeder (VibraScrew). A flow conditioning section (consisting of four screens and three honeycombs) disperses the particles uniformly at the channel inlet. The particle Stokes number in wall units is St = (Stη = 10 based on the Kolmogorov time scale). The main measurement station consists of a 0.3 m long, fully transparent Plexiglas section that follows a 2 m long development section. This ensures that the flow is fully developed and the particles have reached terminal velocity in the measurement

4 18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics LISBON PORTUGAL JULY 4 7, 2016 section. The particles exhausted from the channel are collected in a settling chamber, allowing long run times needed to achieve well-converged statistics, without particles being ingested into the blower. The smooth aluminum walls which make up most of the channel length are provided with static discharge wires and are grounded to structural supports. This prevents the particles from accumulating upon impaction and building up unwanted roughness, an effect which has affected past experiments (Benson et al. 2005). Moreover, the Froude number, defined as the ratio of the bulk flow velocity over the gravitational settling velocity of the particles, is large (Fr = U /τ g = 30), gravitational effects can be considered negligible (Sardina et al. 2012). bulk p Fig. 1: Left: schematic of the considered flow configuration. Right: laboratory installation to study particle-laden turbulent channel flow. Fig. 2: Left: schematic of the considered flow configuration. Right: laboratory installation to study particle-laden turbulent channel flow.

5 18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics LISBON PORTUGAL JULY 4 7, 2016 For 2D measurements, an Nd:Yag (50 mj/pulse) laser is used to illuminate a wall-normal 2D plane (Fig. 3, top), synchronzied to a 4 Mpxl CCD camera. The particle-laden air flow is seeded with DEHS oil atomized by a Laskin nozzle into droplets of 1-2 µm. The small size of the oil tracers ensures they can be readily discerned from the inertial particles (Fig. 3, bottom), which is accomplished automatically via an image processing algorithm based on size and intensityand inspired by Khalitov and Longmire (2002). Therefore, using the same sets of images, Particle Image Velocimetry (PIV) can be applied to the tracer particles to measure the air flow field, while Particle Tracking Velocimetry (PTV) is applied to find position and trajectories of the inertial particles. The latter is performed using the relaxation method proposed aby Baek & Lee (1996). Fig. 3: Top: 2D PIV system measuring on wall-normal plane. Bottom: example of acquired image with tracers and inertial particles (left), which can be easily distinguished based on size and intensity, as apparent from the zoomed in view (right). For 3D reconstruction of the inertial particles position and velocity, we use an advanced DIH-PTV method recently proposed by Toloui and Hong (2015). A typical setup consists of a

6 single camera with an imaging lens, a laser source, and other collimating optics (Fig. 4). The camera records the light interference from the forward scattering of tracer particles and unscattered portion of the laser into holograms. The holograms are then digitally reconstructed into 3D optical fields, from which the 3D tracer positions are extracted. Tracking algorithms are subsequently applied to determine tracer displacements and 3D velocity fields. DIH-PTV suffers from major limitations including poor longitudinal resolution, human intervention (i.e. requirement for manually determined tuning parameters during tracer field reconstruction and extraction), limited tracer concentration, and expensive computations. The method of Toloui and Hong (205) consists of multiple steps involving 3D deconvolution, automatic signal-to-noise ratio enhancement and thresholding, and inverse iterative particle extraction. The entire method is implemented using GPU-based algorithm to increase the computational speed significantly, and it was validated for standard velocimetry (i.e. measureing motions of fluid tracers) against laminar flow in a microchannel as well as synthetic tracer flow fields generated using a DNS turbulent channel flow database. Here the method is applied for the first time to inertial particles in a turbulent flow. Fig. 4: Principle of in-line digital holographic PtV (Figure credit: Jiarong Hong, University of Minnesota).

7 18th International Symposium on the Application of Laser and Imaging Techniques to Fluid Mechanics LISBON PORTUGAL JULY 4 7, 2016 Fig. 5: Left: DIH-PTV optical setup. Right: reconstructed 3D of particle distribution The DIH-PTV optical setup is shown in Fig. 5 (left), and consists of a He-Ne continuous laser (633 nm) and a HX high speed camera mounting a 105 Nikkor lens and recording at 22 khz during 2.9 s. The camera acquires 5 by 5 mm holograms, from which 5 by 5 by 30 mm 2 3 instantaneous volumes are reconstructed (Fig. 5, right). In order to reconstruct a more extended volume in streamwise direction, we exploit the temporal resolution of the particles: we assume that the particles, which have significant inertia, advect in a quasi-frozen manner according to the measured wall-normal velocity profile. By converting time to space (similarly to applying Taylor hypothesis) we reconstruct a streamwise extended volume of 400 by 50 by 30 mm (in 3 streamwise, spanwise, and wall-normal directions, respectively). 3. Results Figure 6 shows the wall-normal profile of particle concentration as measured by 2D imaging. This is calculated from the estimate of the illuminated volume, which is in turn based on an estimate of the laser sheet thickness. As expected, the concentration is much larger in the nearwall region due to turbophoresis. However, the near-wall segregation is much less extreme compared to found by point-particle DNS, both in one-way coupling (Sardina et al. 2012) and in two-way coupling (Zhao et al. 2013). This is likely due to the fact that the locally high concentration causes particle-particle collision (as well as particle-wall collisions), which modulate the tendency of the particles to accumulate near the wall. We notice that the average concentration calculated

8 Fig. 6: Wall-normal profile of particle concentration as measured by 2D imaging. Figure 7 shows wall-normal profiles of mean streamwise velocity (left) and rms of the streamwise fluctuating velocity (right). Profiles relative to both particles and fluid are reported. The particles are faster than the fluid in the near-wall region (y < 10), while they lag the flow + farther away from the wall. This finding, which is consistent with previous studies (e.g. Kulick et al. 1994) is a consequence of the inertial particles crossing the fluid streamlines: when a particle moves in wall-normal direction to turbulent motions, it carries its momentum and it finds itself in a region of larger/smaller fluid velocity. For similar reasons, particles with very different velocities can find themselves in nearby positions, resulting in higher rms velocity fluctuations. Fig. 7: Wall-normal profiles of mean streamwise velocity (left) and rms of the streamwise fluctuating velocity (right).

9 Fig. 8: Ensemble of inertial particles reconstructed by DIH-PTV and projected in spanwise direction. Figure 8 displays the ensemble of the inertial particles position reconstructed by DIH-PTV (with the aid of the adapted Taylor hypothesis) and projected in spanwise direction. The higher near-wall concentration is apparent. In order to characterize the clustering of inertial particles, we use the Voronoi tessellation: each particle is associated to a Voronoi cell, a polyhedron whose faces are defined by the distance from neighboring particles. This method has been successfully used to characterize size and shape of particle clusters obtained by 2D imaging and 3D simulations (Tagawa et al. 2012) of dispersed turbulent flows, and here is applied for the first time to 3D imaging data. Figure 9 (left) illustrates the approach, and for clarity it shows a Voronoi diagram obtained from a 2D slice. The clusters, approximated as the ensemble of cells associated to their particles, are identified analyzing the PDF of the Voronoi cell volumes. The comparison against the Γ distribution expected for randomly distributed particles (Tagawa et al. 2012) defines a threshold volume, below which Voronoi cells are labeled as cluster cells (Fig. 9, right). It is found that the Voronoi cell size associated to particles belonging to a cluster have typical dimensions smaller than ~8η. This is consistent with the notion that preferential concentration in turbulence is maximized over length scales of order 10η (Aliseda et al. 2002).

10 Fig. 9: Left: Voronoi diagram illustrating a cluster of inertial particles. Right: PDF of the volumes of the Voronoi cells, normalized by the average cell volume. The wider distribution compared to a random Poisson process indicate the presence of clusters (small volumes) and areas devoid of particles (large volumes). Because the field is not homogeneous in wall-normal direction, we consider separate bins of the particle ensemble, centered at different distance from the wall (Fig. 10, left). The standard deviation of the Voronoi cell volume PDF, which is a metric of the level of clustering, is shown in Fig. 10 (right). Near-wall particles are substantially more clustered than particles at the center of the channel, which confirms how clustering in wall-bounded turbulence is mainly driven by the near-wall structures, such as the low-speed streaks and hairpins (Marchioli and Soldati 2002). Fig. 10: Left: Wall-normal location of the different considered groups of particles, highlighted in color. Right: standard deviation of the PDF of the Voronoi volumes versus wall-normal distance.

11 3. Conclusions We investigate a vertical turbulent channel flow laden with microscopic inertial particles. Unlike previous studies, we focus on a regime in which both preferential concentration/turbophoresis as well as two-way coupling are expected to be substantial. We measure statistics of particles and fluid using 2D PIV/PTV, and we reconstruct 3D distributions of inertial particles using DIH-PTV. The inertial particles are confirmed to accumulate near the wall, but in a much less extreme fashion than what point-particle DNS (neglecting particle-particle collision) indicate. The particles have a flatter velocity profile compared to the fluid, and a much higher rms velocity fluctuation. Evidence of strong particle clustering is found especially in the near-wall regions, indicating that the preferential concentration is driven by the near-wall turbulence structures. References Aliseda, A., Cartellier, A., Hainaux, F., & Lasheras, J. C. (2002). Effect of preferential concentration on the settling velocity of heavy particles in homogeneous isotropic turbulence. Journal of Fluid Mechanics, 468, Baek, S. J., & Lee, S. J. (1996). A new two-frame particle tracking algorithm using match probability. Experiments in Fluids, 22(1), Benson, M., Tanaka, T., & Eaton, J. K. (2005). Effects of wall roughness on particle velocities in a turbulent channel flow. Journal of fluids engineering,127(2), Elghobashi, S. (1994). On predicting particle-laden turbulent flows. Applied Scientific Research, 52(4), Khalitov, D. A., & Longmire, E. K. (2002). Simultaneous two-phase PIV by two-parameter phase discrimination. Experiments in Fluids, 32(2), Kulick, J. D., Fessler, J. R., & Eaton, J. K. (1994). Particle response and turbulence modification in fully developed channel flow. Journal of Fluid Mechanics, 277, Kussin, J., & Sommerfeld, M. (2002). Experimental studies on particle behaviour and turbulence modification in horizontal channel flow with different wall roughness. Experiments in Fluids, 33(1),

12 Li, J., Wang, H., Liu, Z., Chen, S., & Zheng, C. (2012). An experimental study on turbulence modification in the near-wall boundary layer of a dilute gas-particle channel flow. Experiments in fluids, 53(5), Maxey, M. R. (1987). The gravitational settling of aerosol particles in homogeneous turbulence and random flow fields. Journal of Fluid Mechanics,174, Marchioli, C., & Soldati, A. (2002). Mechanisms for particle transfer and segregation in a turbulent boundary layer. Journal of fluid Mechanics, 468, Monchaux, R., Bourgoin, M., & Cartellier, A. (2010). Preferential concentration of heavy particles: a Voronoï analysis. Physics of Fluids (1994-present),22(10), Paris, A. D. (2001). Turbulence attenuation in a particle-laden channel flow. PhD thesis, Stanford University, Stanford, CA. Rouson, D. W., & Eaton, J. K. (2001). On the preferential concentration of solid particles in turbulent channel flow. Journal of Fluid Mechanics, 428, Sardina, G., Schlatter, P., Brandt, L., Picano, F., & Casciola, C. M. (2012). Wall accumulation and spatial localization in particle-laden wall flows. Journal of Fluid Mechanics, 699, Tagawa, Y., Mercado, J. M., Prakash, V. N., Calzavarini, E., Sun, C., & Lohse, D. (2012). Threedimensional Lagrangian Voronoi analysis for clustering of particles and bubbles in turbulence. Journal of fluid mechanics, 693, Toloui, M., & Hong, J. (2015). High fidelity digital inline holographic method for 3D flow measurements. Optics express, 23(21), Squires, K. D., & Eaton, J. K. (1991). Measurements of particle dispersion obtained from direct numerical simulations of isotropic turbulence. Journal of Fluid Mechanics, 226, Vreman, A. W. (2007). Turbulence characteristics of particle-laden pipe flow.journal of fluid mechanics, 584, Zhao, L., Andersson, H. I., & Gillissen, J. J. (2013). Interphasial energy transfer and particle dissipation in particle-laden wall turbulence. Journal of Fluid Mechanics, 715,