Schedule for: 18w5139 - New Frontiers in Multiphase CFD for the 21st Century Energy Mix

Quadrature-based moment methods (QBMM) are employed to solve generalized population balance equations (GPBE) and are especially useful for poly-disperse multiphase flows. Starting from a closed GPBE, the unclosed moment equations are formulated and closed using QBMM. The accuracy of the closure is controlled by the order of the moments used in QBMM. For example, poly-disperse gas–particle flows can be described by a GPBE for the particle-phase mass-velocity number distribution function (NDF). In practice, the choice of the moments used in the closure is crucial. For particles with a continuous distribution of masses (e.g., same material with different diameters), the mean velocity and granular temperature of each size can be different. Thus, in addition to size moments, the velocity moments conditioned on size are needed to approximate the NDF. Here, the particle-phase model found from the GPBE with the Boltzmann–Enskog collision operator will be used to explain the methodology. Once the moment equations have been formulated, the numerical algorithms used to solve them must be consistent with the underlying GBPE. For example, the numerical methods employed to solve the spatial advection terms and the source terms must guarantee that the transported moments remain realizable (i.e., they must correspond to a NDF). This can be accomplished with kinetic-based, finite-volume methods. With QBMM, the NDF is represented by a finite set of weighted delta functions, corresponding to discrete velocities and sizes, that agree with the transported moments. Thus, it is often convenient to develop algorithms in terms of the quadrature variables in place of the moments. Employing applications from poly-disperse gas–particle flows, several examples of the numerical issues arising with QBMM will be discussed, along with some open issues related to the numerical algorithms.

Heavy inertial particles in spatially and temporally flows can form clusters if their relaxation time is in the order of the dissipation time scale of the flow. This regime, identified by St = O(1), is investigated in this study using analytical tools. We show that the nonlinear variation of segregation versus St can be explained by considering a one-dimensional canonical setting where particles are subjected to an oscillatory velocity gradient that is constant in space. Our analysis shows that the Lyapunov exponent, as a measure of particle segregation, reaches a minimum at St = O(1) and becomes positive at St >> 1 and approaches zero as St goes to 0 or infinity. These predictions, which are corroborated by the numerical results, are directly linked and compared against measurements of the dispersion and segregation in three-dimensional turbulence. Our analysis reveals a strongly nonlinear behavior of the Lyapunov exponents in the straining regimes of strong oscillations. This work was supported by the United States Department of Energy under the Predictive Science Academic Alliance Program 2 (PSAAP2) at Stanford University.

Triboelectrification or tribocharging is a process by which two materials exchange electric charge upon mechanical contact and it has been observed in many industrial applications such as fluidised beds, pneumatic conveying systems and silo flows. In this study, we seek to shed light on dynamics of tribo-electrically charged particles in gas-solid flows through a combination of computational modelling and experiments. To study how tribocharging affects hydrodynamics of gas-solid flows, we have developed Computational Fluid Dynamics-Discrete Element Method (CFD-DEM) augmented by a finite-volume based Poisson solver for electric field and a charge transfer model. In the charge model, the charging tendency of particles is captured by effective work function difference between the contacting surfaces and the electrical field at contact. In recent studies, we performed vibrated and fluidized bed experiments and measured average charge on polyethylene particles at different humidity conditions. We also performed CFD-DEM simulations of the same flow configurations and showed that the predicted charge values were in a good agreement with experimental data. As CFD-DEM simulations are limited to flow systems with a relatively small number of particles and there is a need to examine the interplay of flow and tribocharging in large-scale systems, we have also formulated a kinetic-theory based Euler-Euler model for monodisperse particles with tribocharging. To this end, we derived the mean charge transport equation from the Boltzmann equation allowing for conduction of mean charge through collisions in the presence of electric field, and boundary condition capturing tribocharging at the wall. These models were implemented in an open-source continuum physics software, OpenFOAM. Model predictions were then assessed through comparisons with hard-sphere Euler-Lagrange simulations and experimental studies.

Novel efficient numerical algorithms and large-scale super computers are enabling interface-resolved simulations of turbulent multiphase flows, giving access to details that improve our fundamental understanding and provide input for modelling efforts. In particular, we will first consider heat and mass transfer in particulate suspensions and perform direct numerical simulations to study the heat transfer within a suspension of neutrally buoyant, finite-size spherical particles in laminar and turbulent pipe flows, using the immersed boundary method (IBM) to account for the solid fluid interactions and a volume of fluid (VoF) method to resolve the temperature equation both inside and outside the particles. We examine particle volume fractions up to 40% for different pipe to particle diameter ratios. We report a considerable heat transfer enhancement (up to 330%) in the laminar regime by adding spherical particles, where larger particles are found to have a greater impact on the heat transfer enhancement than on the wall-drag increase. In the turbulent regime, however, only a transient increase in the heat transfer is observed and the heat flux decreases below the values in single-phase flows as high volume fractions of particles laminarize the core region of the pipe. A heat transfer enhancement, measured with respect to the single phase flow, is only achieved at volume fractions as low as 5% in a turbulent flow. Finally, we will consider emulsions in laminar and shear flows and examine the role of surfactants and short-range interaction forces on the rheology. As concerns the turbulent regime, we study homogeneous shear turbulence and report attenuation in the presence of a dispersed second phase. WE show how droplets break up and coalesce to reach a steady state number, with large size following the Hinze prediction.

Euler-Lagrange simulations of pointwise particles in turbulence have been widely employed for understanding the fundamental physics of dispersed flows. Most of the times, particles are modelled as isotropic and rigid. In this paper, we investigate the dynamics of elongated flexible particles in turbulent channel flow. We consider particles that are longer than the Kolmogorov length scale of the carrier flow, and their velocity relative to the surrounding fluid is non negligible. Such particles are modelled as chains of sub-Kolmogorov rigid rods connected through ball-and-socket joints that enable bending and twisting under the action of the local fluid velocity gradients. We examine the effect of local shear and turbulence anisotropy on the translational and rotational behaviour of the fibers, considering different elongation (parameterized by the aspect ratio) and inertia (parameterized by the Stokes number). Velocity, orientation and concentration statistics, extracted from one-way and two-way coupled direct numerical simulations, will be presented to give insights into the complex fiber-turbulence interactions that arise when non-sphericity and deformability add to inertial bias. The physical problem considered here provides a useful foundation for exploring the capability of the point-particle approach to capture the macroscopic features of multiphase flows of elongated deformable particles.

Turbulent spray atomization involves the formation of small-scale liquid structures such as ligaments and sheets that destabilize and break into droplets. Such microscale flow features can be prohibitively expensive to resolve in direct numerical simulations of atomization, making the capture of their dynamics via sub-grid scale models a highly desirable alternative. This talk presents a recently-developed multiscale volume-of-fluid (VOF) framework that enables the modeling of thin liquid and gas films, which can form in many multiphase flows. In spray formation, such thin films arise whenever bag break-up occurs. Classical VOF techniques can represent films as long as they are resolved by at least two grid cells, but under-resolved films will undergo numerically-induced break-up. The proposed framework hinges on a key improvement to standard VOF techniques: the capability of tracking two interfaces per computational cell. Among other advantages, this new capability allows VOF to explicitly represent a numerically thin film, i.e., a film of thickness much smaller than the grid size. We demonstrate how this strategy makes it possible for physics-based models to be introduced to control film break-up. The application of this idea to a turbulent spray atomization problem is explored.

Dispersed multiphase flows with deformable interfaces are frequently encountered in industrial processes involving large scale synthesis of base chemicals and energy carriers. In these flows complex processes, such as formation, coalescence and break-up of the dispersed elements (bubbles or drops), take place with accompanying physical and/or chemical transformations. These processes significantly influence the specific interfacial area, mixing of chemical species, mass and heat transfer rates as well as the large scale circulation patterns and ultimately the performance of multiphase chemical reactors. Due to the inherent complexity of these multiphase flows a multi-scale modeling approach is adopted in which the interactions between the phases can be properly accounted for. The idea is essentially that detailed models are used to generate closures for the interphase transfer coefficients to feed coarse-grained (such as stochastic Euler-Lagrange) models which can be used to compute the system behavior on a much larger (industrial) scale. In this contribution recent advances in the area of multi-scale simulation of dispersed multi-phase flows with deformable interfaces will be highlighted with emphasis on coupled mass, momentum and heat transfer. In addition, areas which need substantial further attention will be discussed.

Gas-solid flows in engineering are typically heterogeneous, with significant multi-scale structures. Direct numerical simulation (DNS), which fully resolves the fluid flow structures and stress distributions on particle surfaces, may present an ultimate numerical method for understanding the behaviors of these flows and the mechanisms behind. This presentation will summarize the DNS studies in our group, using the lattice Boltzmann method to solve the gas flow and the immersed boundary method for the particle-fluid coupling. The simulations were carried out on heterogeneous supercomputers using CPUs, MICs and GPUs concurrently. Thanks to the capability of the software and hardware, scale-dependent behaviors of gas-solid flows in period boundaries and their statistical properties, such as particle velocity distributions, interphase frictions and turbulent stresses are analyzed. Pronounced locally non-equilibrium characteristics are found and their implications to larger scale models of gas-solid flows, such as the two-fluid models, are discussed.

Over the last two decades, the focus of research in multiphase CFD has been on the development of advanced numerical methods, which leveraged the rapid increase in computing power. Even with the advent of exascale computing in the foreseeable future, detailed simulations of industrial-scale multiphase flows will remain out of reach for decades to come. Within many engineering processes, such as fluidized bed reactors, two-phase flow instabilities often lead to ‘demixing’ resulting in spatially non-uniform suspensions that obstruct chemical/thermal efficiencies, which current turbulence models fail to capture at large scales. As highly-resolved multiphase flow data continues to come online, new techniques are needed to integrate this information across scales. In addition, to aid in decision making, multiphase flow simulations have to be augmented to estimate underlying uncertainties in simulation components. In this talk, we will present a data-driven framework for model closure of the multiphase Reynolds Average Navier—Stokes (RANS) equations. Data generated from high-fidelity simulations are used in combination with state-of-the-art inverse modeling and machine learning techniques to (i) quantify model form uncertainty in existing models and (ii) infer the functional form of new turbulence models across a broad range of two-phase flow regimes.

During the past two decades, the steady increase in the power of parallel super-computers participated heavily in developing 3D unsteady CFD modeling approaches. In these approaches, where the flow fluctuations are time and space resolved on a computational mesh, the cost of a simulation is directly linked to the size ratio between the largest and smallest resolved scales. Turbulent combustion and primary atomization modeling have both strongly benefited from this evolution as it enabled to increase the gas/liquid or burnt/unburnt gas interface resolution, the problem size and to include more physics. However, Direct Numerical Simulation (DNS) is still out-of-reach for most of practical configurations. Adaptive mesh refinement (AMR) is an appealing technique to reach DNS at a lower CPU cost. AMR has been originally designed for Cartesian grids and the major challenge for its use on distributed memory machines is its parallelization. The local mesh refinement indeed creates load imbalance that needs frequent repartitioning and balancing. The presentation will detail recent numerical developments on dynamic adaptation of tetrahedron-based unstructured grids. The use of tetrahedra has two advantages for practical configurations: complex geometries are easily meshed and the mesh is locally more isotropic than Cartesian grids. The proposed methodology relies on frequent sequential calls to a remeshing library (www.mmgtools.org), which adapt the mesh inside each MPI rank without modifying the interface shared with the other ranks. Then, repartitioning and transfer of cell groups is performed to ensure an optimal load balance and to modify the cells at the interface. All the underlying algorithms have been optimized to reach good performances with grids of several billion cells on more than 10'000 cores. This dynamic mesh adaptation strategy has been implemented in the YALES2 code (www.coria-cfd.fr) and applied to the modeling of premixed flames and primary atomization. In these applications, the local adaptation enabled to reduce drastically the CPU cost compared to the fixed grid approach and to reach unprecedented mesh resolutions at the interface.

Many problems in science and engineering demand that numerical methods be developed on adaptive grids in order to capture small scale details while minimizing computational resources. I will present recent (and less recent) results on the development and application of numerical methods on Octree Cartesian grids with an emphasis on free boundary problems.

Computational fluid dynamic modeling has become a valuable tool to simulate multiphase flow in various energy applications, providing detailed flow hydrodynamics, chemical reaction, and heat transfer data for cost-effective reactor design, optimization and trouble-shooting. In this talk, I will focus on modeling gas-solid flows using the computational fluid dynamics (CFD) coupled discrete particle method (DPM). Specifically, I will briefly cover the evolving verification and validation of MFIX-DEM code, an open-source CFD code developed at NETL, then present our recent efforts to improve the computational speed of discrete particle model for large-scale applications. I will show numerical simulations of different fluidization systems ranging from lab-scale experimental facilities to commercial-scale reactors with systematic verification and validation.

In this talk we will present our current work on particles made cohesive through the existence of liquid bridges. We will present some rheological studies and some numerical simulations all aimed at elucidating the role that liquid bridges play on the “flowability” of particles. We also discuss the modelling efforts and how they could help bridge the gap between the particle and the equipment scale. Whilst we recognise that quasi-static and rapid flows rely on well established theories, the intermediate regime between those two limits is still not well understood. The models presented are aimed to elucidate the physics responsible for the bulk properties of powders flowing at the intermediate regime; the bulk properties strongly affect the performance of powders at the equipment scale. As an example, in slightly wet particles, the liquid bridges between particles reduce the inter-particle friction (lubrication) by switching the frictional contacts to fluid shear resistance. We use a shear cell apparatus to characterise powder stresses. The experimental results are compared with discrete element modelling simulation results. Finally, we use these results to propose coarse-grained models for large-scale applications.

We study dilute gas-solid flows of cohesive grains in unbounded fluidization via CFD-DEM simulations, focusing on two types of heterogeneities: particle clusters of hydrodynamic origin and agglomerates of cohesive origin. Clusters refer to the regions of higher solid concentrations than surroundings, while agglomerates are particles held together by cohesive forces. Both heterogeneities have significant impacts on reaction rates, heat and mass transfer in multiphase flows. By tracking enduring contacts, we isolate agglomerates from the overall system heterogeneities quantified by local particle number density fluctuations. We found that with increasing system volume, the degree of clustering increases, while agglomeration decreases, contrary to the behavior of granular (no fluid) systems, where clustering enhances agglomeration due to the reduced relative velocities of particles in clusters. Based on statistical and theoretical analyses of particle velocities, higher levels of clustering in larger systems are shown to result in increased relative velocities between the gas and solid phases. This increased relative velocity between the phases serves as an added source of granular temperature, thereby enhancing deagglomeration. We coin this newly-identified mechanism as “cluster-induced deagglomeration” and demonstrate its robustness in systems with increasing cohesion, where saturating agglomeration level with increasing cohesion is observed, as opposed to the monotonic behavior in granular systems.

Heat transfer occurs in a broad array of gas-particle unit operations. Here we consider relatively dense systems, either stagnant or slow-moving, where particle-wall conduction is an important mechanism. Particular emphasis is placed on indirect conduction, which refers to conduction in the stagnant fluid present in the thin layer between the particle and solid boundary. Experimentally, we test the effect of particle surface roughness and particle size on heat transfer. Theoretically, we critically assess the validity of a one-dimensional, isothermal-particle model for indirect particle-wall conduction by Rong and Horio (1999) via comparison with our experiments. An unexpected discrepancy leads to a theoretical analysis of the isothermal particle assumption, and subsequent experimental testing. The results are captured for practical use via a critical Biot number for indirect conduction.

Numerical modeling of the phenomena in multiphase flows is today a necessary stage when developing or improving materials, technology as well as processes. Model development is based on the understanding of the basic physical phenomena. For this purpose, it is essential to simultaneously carry out numerical and experimental studies on some basic configurations to understand all the underlying phenomena. Direct Numerical Simulations is a powerful tool to understand basic physics and to interpret some experimental observations. Particularly with the advance of supercomputers nowadays, the limits of DNS are constantly extending. This presentation gives some examples of research projects that simultaneously include DNS and experimental study on e.g., gas-fluidization, bubbly flows and wet granulation. Potential of DNS to describe the physical phenomena in gas-liquid-nanoparticle three phase flows is briefly discussed as well.

Using Discrete Element Modelling we present simulation results of recently published experimental data (in PNAS) on the structure and dynamics of a collection of spherical particles placed inside a cylindrical vessel that is placed on a shaker table. The collision models with appropriate rolling and sliding friction parameters captures the structure formation and the dynamics of oscillations in both cylindrical vessels and in petri dishes. While matching the macroscopic observations coming from experiments, the simulations provide detailed information on the state of translation and rotation of each particle. Depending on the initial setup, multiple equilibrium states are predicted in the simulation, which is yet to be tested in the experiments. Also, depending on the friction coefficients used, a new advective transport or mixing process is also observed in the simulations, that remains to be validated in any experiments. Our CFD-DEM model has also been tested against careful experiments in the Rayleigh-Taylor experiments and rotating drums.

In this work, the fluid-solid heat transfer is studied by means of particle-resolved direct numerical simulations performed by using a Lagrangian VOF approach based on fictitious domain and penalty methods. The first non-isothermal test case studied in detail consists in a cold fluid flow across an array of hot particles at a given temperature. Particles are fixed and randomly arranged, and their volume fraction spans from 0.1 to 0.4 in order to cover the typical values encountered in fluid-particle dense flows. Three Reynolds numbers, with unity Prandtl number, are investigated. From the fully resolved fluid velocity and temperature fields, macroscopic quantities are obtained by means of spatial averages over the entire domain or over planes normal to the streamwise direction. These quantities are then used to investigate the gas-solid heat transfer at macroscopic scale with the goal to support two-fluid model development. At such macroscopic scale, the fluid energy equation makes appear a velocity-temperature correlation term which accounts for the microscopic heterogeneous distributions of both the fluid velocity and temperature due to the momentum and heat transfer with the solid particles. In this study, such a contribution is written in terms of a bulk-temperature tensor. In the presence of a unidirectional mean flow, this tensor reduces to the unique component representing the well-known bulk temperature, as classically defined. A connection between this contribution and the ratio of two Nusselt numbers based, respectively, on the fluid temperature and on the bulk temperature, is pointed out. These Nusselt numbers are computed and compared to the correlations available from the literature. Their ratio is also computed and compared to the literature. On the basis of such a ratio, a model for the fluctuating velocity-temperature contribution is proposed. In a second stage, particle-resolved numerical simulations were carried out for several configurations of non-isothermal liquid-solid fluidized beds with a limited number of particles (280, 520 and 1280) at a given temperature with the fluid flow injected at a uniform and constant cold temperature. Then, in this work, simulations results are mainly analyzed in terms of macroscopic variables. In particular, the measured fluid-particle heat transfer coefficient and fluid velocity- temperature spatial correlations are compared with the mathematical modeling approach derived for the fixed bed.

Multiphase flows are central to many manufacturing and processing technologies, with applications spanning energy generation, material science and biomedical settings. Despite advancing computational resources, the complex dynamics of production processes are challenging to simulate, and the use of empirical correlations remains the norm. This talk will summarise recent efforts to develop predictive numerical tools that can be used as a basis for engineering design. In a multi-fidelity approach, we apply highly resolved hybrid front-tracking techniques to clarify fundamental flow features, and utilise low-cost simulation variants to address engineering design questions. We will share simulation results inspired by industrially relevant problems related to liquid–liquid displacements in pipe flows, mixing of miscible and immiscible fluids, two-phase flows with and without surfactants, and two- and three-phase microfluidics, amongst others. The simulation techniques presented can be coupled to novel surrogate models for exploiting advances in optimisation and machine learning, being the new frontier in multiphase fluid dynamics.

I will present a new numerical scheme to model surface tension for an interface represented by a level-set function. In contrast with previous schemes, the method conserves fluid momentum and recovers Laplace’s equilibrium exactly. It is formally consistent and does not require the introduction of an arbitrary interface thickness, as is classically done when approximating surface-to-volume operators using Dirac functions. Variable surface tension is naturally taken into account by the scheme and accurate solutions are obtained for thermocapillary Marangoni flows.

Mezcal is a traditional alcoholic Mexican spirit distilled from fermented agave juices which has been produced for centuries following an artisanal form of preparation and testing. Its alcohol content is traditionally assessed by pouring a stream of the liquid into a small vessel: if the alcohol content is correct, stable bubbles, pearls, form at the surface and remain floating for some time. By conducting controlled experiments, we found that the pearls form as a result of surface splashing instabilities. Once formed, we studied the extended life-time of pearls by experiments and numerical simulations. It was found that both the changes in fluid properties (resulting from mixing ethanol and water) and the presence of surfactants are needed to observe pearls with a long life time. Moreover, we identified the the conditions for which a bubble can have a long life. These results may have important implications for other flow phenomena and engineering applications.

It has been a misconception that CFD of commercial fluidized and circulating fluidized beds is only in its infancy. Indeed, codes and computers are fast enough to now simulate large-scale unit operations in a reasonable amount of time (i.e., weeks). However, CFD has been used for over 20 years in the application with significant impact. Today, CFD is being used for a wide range of industrial applications and targeting complex problems in the industry such as NOx emissions, particle attrition, erosion, maldistribution of feed gases, debottlenecking, etc. Yet, a successful model stems from how well the modeler and stakeholders understand the fundamentals and the constitutive equations used for closing those fundamentals. A lack of this understanding has never changed in the last 20 years, and has resulted in the tarnishing of a valuable tool (CFD, DEM, DNS). In other words, the promise of higher order modeling is confounded by a myriad of mistakes including not understanding the underlining assumptions and the limitations they come with, oversimplifying boundary conditions, ignoring transient versus steady-state behavior, and not realizing the benefits of a poor fit. All of these issues have been the result of failed opportunities that should have been identified at the very least in retrospect. This presentation will focus on the some of the gaps that a modeler often misses or opportunities that were never capitalized.

Between the late 1990s and the early 2000s a plethora of methods for the coupling between population balance models (PBM) and computational fluid dynamics (CFD) in the context of the Euler-Euler framework have been developed. Among them quadrature-based moment methods (QMOM, DQMOM, CQMOM, EQMOM) have been very successful. In the last years, due also to the availability of computational power, these methods have been extensively applied to the simulation of real industrial multiphase flows. This has shown the potentiality of the methods but also their limitations, mainly related to the lack of reliable physical and chemical models to describe the relevant phenomena involved under industrial conditions. Industrial conditions, in this context, mean dense systems (i.e. high concentration of the disperse phase) and complex chemical compositions. In this talk we will show what are the progresses made in the last five years to address these issues with a specific focus on coalescence and breakage kernels and interfacial force models. Examples taken from our recent work with commercial and open-source CFD codes will be discussed and critically analyzed with particular attention to energy applications: gas-liquid bioreactors under the heterogeneous regime, boiling flows, turbulent liquid-liquid emulsions, complex fluids in porous media and precipitation of materials of lithium batteries.

This talk aims at introducing our current understanding of the rheology of suspensions of non-Brownian particles in non-Newtonian fluids. These complex suspensions can be found in natural settings such as landslides, mudslides, and submarine avalanches as well as industrial applications such as in mining operations, chemical mechanical, conversion of biomass into fuel, the petroleum industry, etc. The main scientific challenge is to establish a continuum framework and refine it through microstructure investigations. Suspensions may vary on the particle scale from Stokesian behavior to inertial behavior depending on the flow configuration, the type of suspending fluids, etc. We present a tensorial continuum framework based on our recent computational and experimental works and discuss how this framework can be used to study the dispersion of solids in industrial processes and geophysical flows.

Granular material subjected to agitation is encountered in many practical applications of material processing (e.g., cooling, heating, granulation, and clinkering). Furthermore, many of these applications also involve heat transfer whereby solids face cooling or heating surfaces, and heat is exchanged not only between individual particles, but also between the particles and external surfaces throughout the duration of the particle–particle or particle–surface contact. This work focuses on understanding the heat transfer mechanisms in a granular bed inside a rotary drum, which is one of the most commonly used process equipment. All modes of heat transfer are quantified under varying operating conditions to establish a strong understanding of the heat transport. For this, CFD-DEM (using MFIX-DEM, an open source multi-solver suite) simulation techniques are used to analyze the thermal behavior. The heat transfer coefficient of the particles is derived to validate the conduction heat transfer model implemented in MFIX-DEM, and the effects of various parameters are studied.

This talk will present recent work investigating the use of elasto-plastic continuum models for predicting the flow behavior of particulate materials. Several constitutive relations have been implemented in a commercial finite element method (FEM) software package to predict flow and stress fields in hoppers and tumbling blenders. The stress and velocity fields and mass flow rates match well with other widely accepted models and correlations. Particle mixing and segregation are incorporated by coupling the FEM-predicted continuum flow field with an advection-diffusion-segregation equation, which includes correlations derived from experiments and discrete element method simulations. The blending and segregation results match experimental results well. These models hold the promise of modeling particulate flows at industrial scales.

Technology is playing an increasingly important role in the oil & gas industry as we move towards harsher environments (for example deep-see drilling), stiffer environmental regulations, and tougher business climates. From operations stand-point, safety of our processes is of prime importance followed by commercial efficiency. As our operations primarily involve flowing hydrocarbons, Computational Fluid Dynamics is playing an increasingly important role in our operational excellence - from understanding the integrity/safety risks to our equipment to extracting optimum yields from our processes. A brief overview of CFD activities across BP (Upstream & Downstream) will be provided. Specific Upstream applications will cover prediction of erosion, corrosion and flow-induced-vibration. On the Downstream side, we will discuss crude blending in large tanks, thermal mixing points, fluid catalytic cracking, and bubble-column reactors. In all these examples, emphasis will be laid on challenges faced in accurately predicting the phenomena of interest due to multi-phase/physics dynamics, moving/deforming boundaries or sheer size of the industrial-scale equipment. Also of importance is the turnaround time of these models – from development to runtime and post-processing. The importance of experimental validation, lack of good currently available data, and the need for industrial best practices will also be discussed.

Gasoline direct injection (GDI), because it improves fuel economy, has seen rapid adoption, despite large emissions of harmful nanoparticles. To overcome this shortcoming, Selective Catalytic Reduction Filters (SCRF), a combination of a Selective Catalytic Reduction system (SCR) and a Particulate Filter (PF) meant to reduce both NOx gas and Particulate Matters (PM), have also attracted OEMs’ interest due to lower cost and volume than existing engine exhaust after-treatment systems. Practically speaking, integrating the two technologies consists in depositing a layer of a catalytic washcoat into the PF, but this usually affects negatively the PM capture and back-pressure in the filter. Seeking a possible synergistic effect such that an optimum balance between catalyst effectiveness, PM capture, back-pressure and cost is found is of prime interest. To predict how the washcoat deposition profile can affect the SCRF performance, a four-step numerical model was developed. It consists of: (1) the numerical reconstruction of a representative volume of the porous wall with various washcoat distributions and coat weights based on X-ray computed tomography (CT) data, the computation of both (2) the pressure drop and (3) the NOx catalytic reduction effectiveness through the coated porous wall by solving, using the Lattice Boltzmann Method (LBM), a coupled problem involving the Navier-Stokes equations and an advection-diffusion-reaction equation, and (4) the prediction of the PF filtering performance by means of the solution of a Langevin problem.

In partnership with GE Power and the U.S. DOE we have been developing a predictive LES model for designing a new generation of high-efficiency, coal-fired, 1.2 gigawatt power boilers. We are using moment methods for tracking the particle distribution transformations during the combustion process of the pulverized fuel. The key quantity of interest is the heat flux distribution on the boiler walls from radiation dominated by the particle size distribution. The objective is to predict this heat flux within 5% uncertainty. Quantifying the uncertainty in the multiphase model form is accomplished by using validation data from three scales and using a hybrid method of traditional Bayes and bound-to-bound consistency analysis to predict the uncertainty in the full-scale design.

Many unit operations in the mineral processing, metal production, energy and chemical process industries involve multiphase flows and in many cases the flows are strongly linked to other physical processes. CSIRO has developed CFD models and applied these modelling to a number of industrial processes to improve our understanding of these processes, improve performance and to develop new novel processes. This paper discusses some of these applications and identifies limitations in the models. While these models have proved valuable there are a number of challenges to improving fidelity of the models and extending them to other applications. Application areas discussed include a multiscale and multi-physics approach for modelling gas-liquid-particle flows in aluminium reductions cells, a novel Hybrid Euler-Euler-DEM model applied to modelling particle behaviour in a coal beneficiation fluidized bed, gas-liquid-particle flow in flotation cells and slurry flows in industrial thickeners.

Magnetic field actuation of magnetic nanoparticles is a potent means for process intensification. Its aim is to devise innovative micro/nanofluidics stimulation strategies thanks to a singular alliance between the science of magnetism and transport phenomena. There is so far little awareness of using magnetically-excited MNPs as remotely-controllable nanomixers. We recently succeeded in aligning MNP spins streamwise to laminar flows in microchannels using rotating magnetic fields (RMF). Pulling the nanoparticle’s gyration out of the grips of liquid vorticity by coercing its magnetic moment to gyrate crosswise –there where only diffusion prevails– spectacularly enhances radial diffusion in microchannels. Our recent finding of RMF-triggered eddies in stagnant/laminar fluids falls within the paradoxes’ dominion in fluid mechanics since eddies are an apanage of turbulent flows only. No doubt that liquids can be intimately stirred by spinning MNPs down to ca. 10nm has far-reaching implications as to how mixing/transport can be enhanced in miniature devices. The fact that one can afford nanomixing down to lengths beneath the ~10^3-nm Kolmogorov or Batchelor scales, which usually delineate the private turf of molecular diffusion, is enthralling. Such possibilities open up a range of intensification strategies hitherto unreachable to laminar micro/nanofluidics where diffusion is a severe transport barrier, e.g., fast catalysis, mass/heat transfer/lateral mixing, high-turnover enzymes. In this contribution, we will illustrate with some examples how MNPs can be taken advantage of as nanoscale stirrers to control passive and reactive (micro-)mixing when these nanoparticles are harnessed by different types of magnetic fields to generate MNP-pinned localized agitation in the liquid phase. Ferrohydrodynamic (FHD) theory devolved to analysis of magnetic nanofluids has been under tremendous improvement over the past decades and despite existing debates, FHD still remains a valuable tool to decipher and understand the mechanisms of magnetic field-assisted triggering of colloidal MNPs. Therefore, it is hold in high regard to also include a specific topic on mathematical modeling of colloidal systems with aptitude of absorbing magnetic energy and its conversion into kinetic energy as a tunable external driving force.

The global warming caused by the increasing CO2 level in the atmosphere is a worldwide concern. Microalgal photosynthesis is a potential approach for CO2 reduction due to its high efficiency in capturing CO2. Furthermore, the microalgae can be by themselves a valuable bioproduct particularly for the energy sector as a substitute to fossil fuels. Compared to conventional tubular reactors, the bubble column biophotoreactors are compact and low-cost as well as easy to construct, operate and maintain even at large scales [1]. One of the common issues encountered in biophotoreactors is that the light penetration depth is limited to a few centimeters due to a shadowing effect resulting from the strong light absorption of individual microalgae cells. To overcome it, a good agitation and microalgae circulation providing a uniform residence time distribution of microalgae in the illuminated zone is required. The knowledge of the residence time of the microalgae cells in the illuminated zone is thus of significance importance for the design, optimization and scale-up of the bubble column biophotoreactor. In this work, an Eulerian-Eulerian-Lagrangian model is developed to track the motion of microalgae cells inside the bubble column photobioreactor. A mesoscale force model, which takes the influence of bubble-induced wake on the motion of micoalgae cells into consideration, is proposed. The paper is organized as follows. First, the model is fully described. Next, the accuracy of the simulation is assessed by comparing the simulation results to Radioactive Particle Tracking experimental data performed in our laboratory. Finally, the effect of the operating conditions, such as superficial gas velocity, alga concentration and properties of liquid phase on the residence time of microalgae cells in the illuminated zone is evaluated.

Euler-Lagrange point-particle (EL-PP) technique has been increasingly employed for solving particle, droplet and bubble-laden flows. Since flow around the individual particles is not resolved, the accuracy of the technique depends on the fidelity of the force law used for representing the fluid-particle momentum exchange that occurs at the microscale. The applicability of the standard EL-PP approach is however limited to (i) particles of size much smaller than the grid scale and (ii) dilute flows where inter-particle interaction is weak. In this talk we will discuss recent fundamental developments that begin to ease these limitations. With increasing numerical resolution, as the grid size approaches the particle size, we face the unpleasant prospect of force law becoming less accurate. This is due to the self-induced flow generated at the particle location, which corrupts the estimation of undisturbed flow velocity that is needed in the force evaluation. We will discuss theoretical approached to properly correcting for the self-induced flow. Finally, we will present the pairwise interaction extended point-particle (PIEP) model which rigorously extends the point-particle technique to higher volume fractions. This model systematically accounts for the precise location of all the neighboring particles in computing the hydrodynamic force on each particle. The model assumes pairwise interaction so that the perturbation flow induced by each neighbor can be considered separately and superposed. The generalized Faxén form is then used to quantify the perturbation force due to the presence of the neighbors. The PIEP model predictions are compared against corresponding DNS in a number of test problems.

The Eulerian-Lagrangian point-particle method has become a leading paradigm in the study of dispersed particle-laden flows. This strategy simultaneously represents the non-continuum physics of the dispersed phase coupled to the continuum physics of the carrier fluid. However, while verifiable strategies for one-way coupled particles have been available for decades, it has not been until recently that verifiable strategies have been developed for two-way coupled flows. The purpose of this talk is to present some of the current state of understanding with regard to verifiable simulation of two-way coupled particle-laden flows. One key idea centers around the notion of undisturbed fluid properties evaluated at the location of each particle. In dilute regimes, we will show that removal of the self-disturbance portion of the flow computed in point-particle simulations is necessary for accurate comparison with particle-resolved simulation. In denser regimes, theory and simulations reveal that particle-particle screening cannot be ignored. We may also draw upon the Squires jet as a means for forming reasonable expectations about point- particle simulations. Time permitting, we will also discuss two-way coupled heat transfer in the context of point-particle simulations.

Fluidized-bed reactors (FBR) are found in a large variety of industrial processes ranging from coal gasification to water treatment. In reactive FBR processes, gas-solid mixing and interactions can be enhanced as well as the chemical reaction rate. These properties are particularly interesting to achieve low-temperature combustion with high conversion efficiency and low pollutant emissions such as nitrogen oxides. The objective of this study is to achieve simulations of a FBR at mesoscopic scale using a coupled CFD/DEM (Discrete Element Method) approach in order to gain insight into the physics of such processes and to extract information to be used in the modeling at macroscopic scale. The chosen configuration is a semi-industrial FBR fed with a mixture of natural gas and air containing 200M sand beads. Experimental data show a shift in the combustion regime above a critical temperature of 800°C. Large-Eddy Simulations (LES) are performed using the finite-volume code YALES2, a low-Mach number solver based on unstructured meshes. Its DEM solver has been designed to perform simulations in arbitrarily complex geometries and optimized for massively parallel computing: it features a dynamic collision detection grid for unstructured meshes, packing/unpacking of the halo data for non-blocking MPI exchanges and a dynamic load balancing algorithm.

Developing new and efficient commercial scale multiphase reactors is extremely difficult as one needs to consider the complex interactions over a wide range of both temporal and spatial scales encountered in these systems (from molecules to macroscale). Therefore, it is important to codify and automate the knowledge collectively acquired so far, reserving human resources for the creative solutions that build upon codifying past learnings. Computational science (algorithms, theory and modeling, computer science, etc.) combined with exponential growth in computing hardware has great potential to provide us valuable tools for improving the effectiveness of existing plants but also design new reactors. Industry is in desperate need of fast simulation tools that are comparable to the accuracy of the single-phase reacting flows to advance the adoption of multiphase flow reactors and primarily to reduce the risk of scale-up. On the other hand, academia and national laboratories are focused on improving numerical methods, submodels or certain physics to push the state-of-the-art and predictability of the models. However, there is tremendous need for rapid integration of various capabilities that exploit latest computer hardware improvements, robust model reduction techniques, etc. that would improve the computational tools available to the industry to deliver the next generation of efficient and cost-effective reactors. These developments typically fall into the valley of death where industry and software vendors feel these efforts are far-fetched while the academic world considers it incremental. One way to address this divide is to develop mechanistic upscaling framework to coarse-grain information from more detailed modeling approaches to fast models accessible to the industry with relatively minimal information loss. I will present such an approach using a gas-phase fluidized bed polymerization reactor as an example. First, I will provide an overview of the various models currently used at the different scales in gas-solids reacting flows. The coupling across the scales will be introduced through few different approaches: a) Discrete-Continuum Coupling using Discrete Element Method for particles b) Upscaling data from Discrete Element Modeling results to continuum based Computational Fluid Dynamics (CFD) c) Upscaling information from CFD to simplified reactor network models In conclusion, the presentation will elucidate the approaches and opportunities in bridging the gap between academic advances and industry needs to model and advance our understanding of gas-solids reacting flow reactors.

Using a direct-forcing immersed boundary method (IBM) at fully resolved grid resolutions, we study the interaction between the fluid in the lid-driven cavity and ensuing response of a collection of identical cylindrical particles immersed in the cavity. The particles are fixed in lattice location but are free to rotate, and reach steady-state angular velocities under low to moderate Reynolds numbers. The goal is to find the effect of different lattice configurations on the dynamical response of the cylinders – particularly the angular velocities that develop spontaneously and the forces exerted on each of them by the fluid motion. Two lattices, one triangular and one rectangular, are considered. The two configurations exhibit quite different responses. The rectangular lattice contains two kinds of flow paths, an axis-aligned primary path, and a secondary path in the diagonal direction. This gives it a rich response to the fluid forcing as Reynolds number is increased. The triangular lattice only has one kind of flow path and behaves in a more consistent manner. By increasing the cylinder radius, the secondary path in the rectangular lattice can be eliminated. This qualitatively alters its nature and response to the fluid forcing, and could be considered as a phase transition.

Accurately predicting fluidized-particle behavior using computational fluid dynamics (CFD) is potentially beneficial to industry for several possible reasons, including, improving fluidized bed operational efficiency, troubleshooting existing operations, exploring new equipment design and improving the speed and reliability of operation scale-up. An essential step to establishing a relevant CFD toolboxes for predicting fluidized particle behavior is accurate prediction of the drag force, a major component of the fluid-particle momentum exchange. A wide range of drag models exist and can be separated into two categories, namely homogenous drag models (based on the assumption of homogenous distributions of particles with each fluid cell) and sub-grid drag models (utilize sub-grid formulations to account for solids volume fractions gradients within fluid cells). A critical, comprehensive review of these models is presented with a focus on the parameter space relevant to development and validation. Key findings are presented that include the limitations associated with various model development and the range of experiments used for validating models and the mismatch in drag model validation across different fluidization regimes. Gaps associated with the experimental data available for model calibration and validation and the limitations associated with filling the gap are discussed.