• Home
• Program
• Contact

14:00   Non-Newtonian Flows 1 14:00 15 mins #199 DNS-DEM SIMULATION OF TURBULENT NON-NEWTONIAN SUSPENSION FLOW Enzu Zheng, Murray Rudman, Shibo Kuang, Andrew Chryss Abstract: Currently there is a move towards higher concentration tailings suspension due to its environmental, economic and social benefits. With increasing solids concentration, the rheologically active fine particles combine with the carrier fluid to form a non-Newtonian slurry and all other solids will be conveyed as coarse burden. These non-Newtonian suspensions typically exhibit shear-thinning behaviour along with the presence of a yield stress. The interaction between non-Newtonian rheology and coarse particle transport in high concentration suspension is still poorly understood, particularly in the transitional and turbulent flow regimes. This paper presents a DNS-DEM model for investigating the underpinning fundamentals of turbulent non-Newtonian suspension. In the model, DNS is applied to capture the unsteady turbulent flow structure, and the DEM is used for modelling the detailed particle-particle interaction. An open-source coupling library CFDEM, which couples OpenFOAM and LIGGGHTS, is applied for implementing the model. A preliminary simulation has been conducted to couple turbulent flow of a Herschel-Bulkley (H-B) fluid to a coarse solid suspension, verifying that the coupling works. Despite that solid-liquid interaction models are so far simple extensions from the Newtonian models, preliminary results have been found to be qualitatively correct, with particle segregation and solids-concentration pulsing observed as in pipe loop experiments (Figure 1). We will introduce new solid-liquid interaction models in CFDEM and discuss validation against data obtained for experimental suspensions, including pressure drop, flow rate and concentration profile. The methodology developed in this project allows an extended understanding of the complex interaction between non-Newtonian fluids and particle transport to be determined. 14:15 15 mins #242 SOME MECHANISM PROCESSES CONCERNING SHEAR-THINNING T-JUNCTION MIXING WITH DIRECT NUMERICAL SIMULATION Haining LUO, Delache Alexandre, Simoens Serge Abstract: The T-Junction mixing is a very interesting configuration that mix convergent flows inducing two kind of separate and localized mixing processes : a mixing layer where shear stress develops independently from wall and a re-circulation bubble from which feeding and rejected fluid depends mainly on intermittent exchanges between re-circulation bubble and external zone. Surprisingly not so many papers tend to describe phenomenology of its flow structure with Newtonian fluid and a fortiori with Non Newtonian fluid flow properties. Studying the first zone could help to describe more precisely interactions between shear thinning and turbulence phenomena mainly in terms of out of wall generated turbulence or bulk-flow turbulence. Even if from dynamical point of view the flow is well established, the concentration field is submitted to a permanent intermittency providing feeding and evacuation fluid, respectively, to and from re-circulation zone, more or less mixed. In order to describe precisely these mixing processes, we perform a direct numerical simulation of T-junction flow by using a finite volume method with OPENFOAM solver. Total cell count is about 20 million and a mesh spacing constraint is applied to wall boundary to ensure y+ = 1 at first wall cell. We simulate both Newtonian and non-Newtonian fluid by using a Carreau-Yasuda law based on experimental measure of Xanthan gum from [3], known to have strong shear thinning properties compared to weak elastic properties. Two regimes are imposed following Ho [1] and Sakowitz et al. [2] : the "deflecting" and the "impinging" flow regimes (ID and IR respectively). We simulate different Reynolds numbers corresponding to those imposed by the experiment of Nguyen et al. [3] : Re = 4800 and Re = 8000 for water; Re = 2400 and Re = 4000 for Xanthan gum solution. Concentration as a passive scalar was imposed on the inlet of the vertical branch. We examined behavior and evolution of the velocity and concentration fields and quantified the mixing. An example of non-Newtonian simulation is show on Figure 1. We will show results leading to the following conclusions by comparing ID vs IR and Newtonian vs non-Newtonian Fluid : • we analyze precisely the structure of flow in both IR and DR. • we observe turbulence peak shifting for non-Newtonian fluid in IR. • we observe a viscous core in re-laminarised case for non-Newtonian fluid where the viscosity is strong whereas a nonNewtonian self-sustaining turbulence is observed in the higher Reynolds case. • we destabilize the re-laminarised case by introduced pulsative conditions in inlet (Strouhal St = 1; 5). as [2] we demonstrate that the mixing efficiency of the flow is largely enhanced in the IR compare to DR. Moreover in the IR, the reached mixing close to the junction maintain its level till the outlet whereas in the ID the mixing evolves in a longer distance to outlet. 14:30 15 mins #553 TEMPORAL STATISTICS IN TWO-DIMENSIONAL ELASTIC TURBULENCE Himani Garg, Stefano Berti, Enrico Calzavarini Abstract: We numerically investigate the temporal statistical properties of a dilute polymer solution in elastic turbulence (ET) conditions, i.e., in flow conditions of vanishing Reynolds and high Weissenberg numbers. The motivation behind this work is to offer a straight comparison between numerical ET measurements and experimental ones [1, 2, 3]. Indeed, the available experimental techniques provide access to the flow velocity at best in few specific positions and at various times, so that the flow spatial statistics can be retrieved only by means of the Taylor’s hypothesis assumption (see [6] for a detailed discussion). To this end we carry out extensive direct numerical simulations of the two-dimensional Kolmogorov flow of an Oldroyd-B viscoelastic fluid, a flow setting first studied in [4] and more recently in [5]. Static point-like numerical probes are placed at several different locations in the flow, in particular in the positions where the mean flow is maximum, minumum and null. The data from such a probe array allows to precisely assess both the degree of inhomogeneity and local isotropy of the flow. The single point statistics, specifically the probability distribution functions of velocity components and of their increments (Fig. 1) reveal features qualitatively similar to the ones known from experimental measurements [3], which are reminiscent of high-Reynolds number turbulence. However the flow isotropy does not seem to be recovered even for the smallest time lags . Furthermore, the applicability of Taylor hypothesis is questioned. For both components, the spectra of velocity fluctuations display power-law ranges with exponents close to -3.8 independently of the probe position, in close agreement with experimental observations and what previously measured in the spatial domain in simulations. We further characterize the flow intermittency by measuring the high-order statistical moments of the velocity increments. 14:45 15 mins #612 Effects of numerical resolution on elasto-inertial turbulence Vincent Terrapon, Yves Dubief, Fuqian Yin, Jacob Page, Rich Kerswell Abstract: Our talk will review time and space resolution requirements for EIT, as well as discussed advances in the understanding of the mechanism of EIT. 15:00 15 mins #164 EFFECTS OF VISCOELASTICITY ON TURBULENT BUBBLY FLOW Outi Tammisola, Daulet Izbassarov, Zaheer Ahmed, Metin Muradoglu Abstract: Fully three-dimensional direct numerical simulations are performed to examine the effects of viscoelasticity on the structure of turbulent bubbly flow in a pressure-driven channel flow. Even though effects of polymers on two-phase flows have been studied, previous numerical simulations were restricted to laminar cases. Thus, the effect of viscoelasticity in turbulent two-phase flow is still an open issue and a very challenging problem. The viscoelastic two-phase solver is based on a fully parallel front-tracking method [7, 2] and a highly scalable FFT-based pressure solver [4, 1]. The FENE-P model is employed to model the viscoelasticity. The viscoelastic constitutive equations are fully coupled with the incompressible Navier-Stokes equations. Detailed description of the numerical method with various validations for viscoelastic two-phase systems can be found in Refs. [3, 4, 1]. In this work, the method was first validated against the turbulent single phase benchmark cases originally by Kim et al. [5] at Re τ = 127.3 and Re τ = 180, where the Reynolds number is based on friction velocity and half width of the channel. Secondly, the method was validated against Newtonian turbulent bubbly channel flow results in the literature [6]. The bubbly flow was initialized by placing bubbles in a vertical channel with a fully developed turbulent flow. The void fraction considered in this study was fixed to 3.44%. Details of the case can be found in [6]. Sample steady state results are shown in Fig. 1. Present results were found to be in good agreement with Lu et al. [6] both for single-phase and two-phase cases. Turbulent viscoelastic bubbly flow was initialized by the same turbulent flow field, where we injected the same number of bubbles but different amounts of polymers (where the polymer concentration is determined by the Weissenberg number and polymeric viscosity ratio). Simulations are ongoing to investigate the effects of viscoelasticity on the overall flow structure of the bubbly flow. The preliminary results show that the viscoelastic stresses alter the dynamics of the drops/bubbles significantly. The present study is to the authors’ knowledge is the first direct numerical simulation on the effects of viscoelasticity in turbulent two-phase channel flow. 15:15 15 mins #459 Two-dimensional elasto-inertial coherent structures in viscoelastic channel flow Jacob Page, Yves Dubief, Rich Kerswell Abstract: The viscoelastic effects introduced by adding long chain polymers to a Newtonian solvent have dramatic consequences for turbulent fluid motion. Perhaps most significant is the substantial reduction in skin friction at high-Reynolds numbers, though viscoelasticity can also seed entirely new chaotic dynamics which are predominantly two-dimensional \cite{Sid2018} - so-called `elasto-inertial' turbulence (EIT, \cite{Samanta2013}) - in which the polymer becomes stretched in thin sheet-like structures with attached patches of intense spanwise vorticity (see figure 1). The drag reduction brought about by viscoelasticity has been linked to a stabilisation of near-wall streaks that results in a supression of high-drag bursting events \cite{Graham2014}, though it has also been argued that flows with large drag reduction are the weakly-elastic limit of EIT \cite{Dubief2013}. Recent experimental evidence suggests that EIT is a distinct dynamical regime unrelated to modified Newtonian turbulence \cite{Choueiri2018}, though the two behaviours coexist at sufficiently high Reynolds numbers. In this talk, we explore the elastic origins of EIT by computing the first exact coherent structures in a viscoelastic channel flow of a FENE-P fluid. Motivated by the recent discovery that elasto-inertial pipe flow is unstable \cite{Garg2018}, we first revisit the stability of viscoelastic channel flow and find a linear instability distinct from the Newtonian TS branch with up-down symmetry in the streamwise velocity. The instability persists down to a bulk Reynolds-Weissenberg number pair $Re_b, Wi_b \sim 50, 25$ (at solvent viscosity $\beta=0.9$) and exists for moderate elasticities $Wi_b \lesssim Re_b \lesssim Wi_b^3$. Just beyond the point of marginal stability, these instability waves grow and saturate onto a relative periodic orbit associated with a pair of large amplitude sheets of polymer stretch that meet at the centreline. We perform branch continuation under varying Weissenberg number to demonstrate that this bifurcation is subcritical and find a finite-amplitude travelling wave which persists down to $Wi_b\sim 10$ (see figure 1). We will also discuss the results of continuing the exact coherent structures upwards in Reynolds number, with a particular focus on possible overlap with EIT at high Reynolds number. Our results indicate that the complicated dynamics of EIT is predicated on the ramifications of the linear instability which exists at sufficiently large $Wi$ and so EIT is distinct from Newtonian turbulence.