Computational Methods
CFD-DEM: Modeling the Small to Understand the Large Ray Cocco Particulate Solid Research, Inc. (PSRI) William D. Fullmer Peiyuan Liu Christine M. Hrenya Univ. of Colorado, Boulder
Coupling the traditional computational fluid dynamics (CFD) approach to model the fluid phase with the discrete element method (DEM) to model the particles creates a powerful numerical method to study multiphase particulate flows.
Cyclones Circulating Fluidized Bed
Riser
Bubbling Fluidized Bed
Standpipe
Mixing Pot
p Figure 1. DEM simulations are capable of modeling subsystems or highconsequence regions. Displayed here are snapshots from DEM simulations of dense gas-solid flow in the bubbling bed of a circulating fluidized bed (CFB) and dilute gas-solid flow in the riser of a CFB.
38
www.aiche.org/cep September 2017 CEP
C
omputational fluid dynamics (CFD) emerged in recent decades and reduced the costs associated with scale-up, optimization, safety analysis, and industrial operations that involve conventional fluids. It provides numerical information that enhances and supplements experimental data and eliminates the need to rely entirely on expensive experiments to gather data that are hard or impossible to measure directly. For example, CFD simulations allow engineers to isolate the effects of a single variable, whereas in experiments, it may be impractical to change only one variable at a time. For the numerical simulation of solids flows, a counterpart to CFD is the discrete element method (DEM), which tracks the motion of every particle by solving Newton’s equations of motion for particles in free flight and employs a collision model for the particles in contact (1). Numerous collision models exist and incorporate the conservation of momentum and kinetic energy loss due to inelasticity and/or friction. It is relatively straightforward to incorporate microscale physics such as cohesion into the DEM framework because each contact is resolved (2). Multiphase flows — which are pervasive throughout the chemical, petrochemical, energy, pharmaceutical, mining, and Copyright © 2017 American Institute of Chemical Engineers (AIChE)
other key industrial areas — present an additional challenge to numerical analysis. Specifically, gas-solid particulate flows pose challenges related to interfacial interactions (heat and mass transfer, chemical reactions, etc.), instabilities (clustering and bubbling), multiscale phenomena (particle-scale physics affecting system-scale operation), and locally defluidized regions (enduring and multiparticle contacts), among others. Researchers have coupled CFD for the gas phase with DEM for the particulate phase to tackle the challenges of gas-solid flows. The use of this CFD-DEM method has ballooned in recent years, especially within the academic community (as evidenced by the rapid increase in publications on the topic). CFD-DEM strikes a better balance between numerical accuracy and computational requirement than direct numerical simulation (DNS) and coarse-grained methods (discussed later). However, the biggest barrier to widespread adoption by industry is the computational overhead of CFD-DEM, which remains an issue for larger multiphase systems. While modeling every particle in a very large industrial unit will remain impractical for the foreseeable future, simulating subsystems and other high-consequence regions is currently possible (Figure 1). Additionally, as the parallel capability of computers continues to advance, the number of particles CFD-DEM is capable of handling also continues to rise, making it a promising tool to aid in the design of many industrially relevant problems. This article provides a general overview of several key methods used for the numerical simulation of gas-solid multiphase flows and explains where CFD-DEM fits into the picture. It discusses the current state of the art of CFD-DEM, with an emphasis on the main challenges and bottlenecks, (a) and outlines areas for improvement aimed to address industrial needs.
Computational methods for gas-solid flow The prediction of multiphase fluid-particle flows is an inherently multiscale problem. Industrial applications involve behavior on the order of meters to tens of meters. Unfortunately for industrial practitioners, bubbling and clustering instabilities at the mesoscale (appreciably larger than the particle scale, yet still significantly smaller than the system scale) strongly influence macroscale behavior. In addition, interfacial phenomena — mass, momentum, and heat transfer — that occur at the microscale (i.e., tens to hundreds of particles) influence mesoscale behavior. Furthermore, truly microscopic properties, such as particle surface roughness and humidity in a gas carrier phase, can also influence behavior at the mesoscale and beyond. Reference 3 provides an interesting case study that illustrates how microscopic properties affect macroscopic behavior. As a consequence of this multiscale nature, no single Copyright © 2017 American Institute of Chemical Engineers (AIChE)
numerical method can directly simulate the wide range of scales needed to completely model the dynamics of particulate flows. Instead, researchers have developed a multitude of different methods to handle different scales. The higherresolution methods are more reliable and desirable, because fewer constitutive relations, which are inherent sources of uncertainty, are needed to close the model. The trade-off is that higher resolution comes at an increased computational cost. Figure 2 compares some of the numerical methods commonly used to study fluid-particle flow.
Continuum methods Since industrial processes often involve many particles, continuum approaches that only attempt to solve for the bulk behavior are the preferred approach for modeling large-scale systems. A two-fluid model (TFM) or Eulerian-Eulerian (E-E) model simulates both the actual (molecular) fluid phase and the solids phase (a “fluid” composed of the particles) with independent Navier-Stokes-like equations (e.g., continuity, momentum, energy) weighted by the fluids’ volumetric concentration and connected through interfacial transfer phenomena (e.g., phase change, drag, etc.). In addition to the interfacial transfer terms, TFMs require the user to specify unknown quantities such as solids-phase pressure and viscosity — opening the door for significant uncertainties. A kinetic theory (KT) approach (4) is com(d) monly used to derive such constitutive expressions, particularly when the fluid phase is a gas (5). Recent efforts have shown that the results obtained with KT-based TFMs compare favorably (b)
Microscale DNS
(c)
Mesoscale CFD-DEM TFM
Macroscale MP-PIC Filtered-TFM EMMS
p Figure 2. Different numerical methods are available to simulate particle systems depending on system size. (a) Direct numerical simulation (DNS) is appropriate for studying the detailed (microscale) gas flow field around particles. Image of flow around a fixed bed of particles courtesy of S. Subramaniam, et al. At the mesoscale, the local gas flow is resolved down to a scale of several particle diameters (flow around an individual particle is not resolved): CFD-DEM simulates a gas-solid homogeneous cooling system in a clustered state (b) and a fully resolved two-fluid model (TFM) simulates unbounded fluidization (c). Macroscale methods include filtered TFM, energy minimization multiscale simulation (EMMS), and multiphase particle in cell (MP-PIC), which is used here to model a fluidized catalytic cracking (FCC) catalyst in a pilot-scale (3-ft-dia.) fluidized bed with an air sparger (d). Contours show gas-velocity magnitude in (a) and (b) and solids concentration in (c) and (d); hot colors (red) indicate high values and cool colors (blue) indicate low values, and in (a) and (b), white indicates the interior of (monodispersed) particles.
CEP September 2017 www.aiche.org/cep
39
Computational Methods
to more-accurate discrete particle data (6, 7). However, a grid resolution of approximately ten times the particle diameter (Δx ~ 10dp) is still required to resolve all scales of motion relevant to cluster dynamics (8). Even finer resolutions are required in dense beds that exhibit bubbling (7), because sharp gradients can exist at the cluster/dilute or the emulsion/bubble interfaces. Since instabilities in gas-particle flows are ubiquitous in industrial devices (9), high-resolution TFM simulations (sometimes referred to as KT-TFM or microscopic TFM) are too computationally intensive for most industrial applications. An additional level of averaging is required to tackle industrial-scale problems. One common strategy is filteredTFMs (10–12). Filtered-TFMs rely on an averaging procedure similar to the filtering of the Navier-Stokes equations for single-phase large-eddy simulation (LES). However, in multiphase flows, many more sub-grid-scale closures are needed. Although filtered-TFM methods have been applied successfully to many cases, robust model development remains an active area of research.
Discrete methods Unlike the continuum approach for solids described previously, discrete methods treat the particles as discrete entities. The coupled approach that combines a discrete method
for particles and a continuum model for the fluid is referred to as a Eulerian-Lagrangian (E-L) method. Two treatments of the fluid phase are common, DNS and CFD-DEM. In DNS, all scales of fluid motion are resolved down to the no-slip boundary condition on the surface of each particle. No closure laws are required; drag is an output rather than a required input. Consequently, the Eulerian grid needs to be smaller than the particles, and resolutions on the order of Δx ~ dp/10 are typically required to resolve such fine details for flows with moderate particle-Reynolds numbers. An approach with such fine resolution is extremely computationally expensive. Thus, DNS is unlikely to make a significant industrial impact for some time. For moving particle suspensions (as opposed to fixed-bed simulations, where the particles are static), current DNS capabilities allow for simulations of a few thousand particles. Reference 13 provides a comprehensive review of particle-laden DNS. A less computationally expensive approach is to use a larger grid for the fluid phase while still using DEM to track the motion of every particle. Additional constitutive models are required, specifically to account for: • the unresolved boundary between the particles and the fluid, i.e., interfacial momentum transfer (drag law, etc.) • the unresolved turbulent motion of the fluid (shear- and particle-induced).
An Alphabet Soup of Numerical Analysis CFD
Computational Fluid Dynamics
Any type of numerical approach to solve fluid flow. This term is commonly used to describe the solution of Navier-Stokes-like continuum equations on a discretized computational grid.
DEM
Discrete Element Method
A method in which particles are modeled as discrete entities, typically with a soft-sphere approach.
CFD-DEM
A coupled approach using DEM for particles and CFD for the gas. It is typically used when the fluid grid size is larger than the particle size (Figure 2).
DNS
Direct Numerical Simulation
An approach in which particles are modeled with DEM and all scales of turbulent motion of the fluid are actively captured. It is typically used when the fluid grid size is significantly smaller than the particle size (Figure 2).
E-E
Eulerian-Eulerian
A category of models in which both phases are modeled as a continuum. E-E encompasses mixture models, two-fluid models (TFMs), and energy minimization multiscale simulation (EMMS) models, among others.
E-L
Eulerian-Lagrangian
A category of modeling in which particles are modeled discretely and fluids are modeled as a continuum. E-L encompasses DNS, CFD-DEM, and multiphase particle in cell (MP-PIC) methods.
EMMS
Energy Minimization Multiscale Simulation
A modified drag law for coarse-grained methods to account for the effect of unresolved clusters.
LES
Large-Eddy Simulation
A turbulence modeling technique in which large eddies are dynamically captured and small eddies are modeled with a sub-grid-scale viscosity.
MP-PIC
Multiphase Particle in Cell
An E-L strategy in which the discrete elements of the solid phase represent parcels or clouds of many particles.
RANS
Reynolds Averaged Navier-Stokes
A common turbulence modeling technique in which flow variables are decomposed into mean and fluctuating terms and the resulting Reynolds stress is modeled, e.g., via a Prandtl mixing length model, k-ε model, etc.
TFM
Two-Fluid Model
An E-E strategy in which solid particles are modeled as a continuous fluid phase, interpenetrating the actual (molecular) fluid phase.
40
www.aiche.org/cep September 2017 CEP
Copyright © 2017 American Institute of Chemical Engineers (AIChE)
This approach is commonly referred to as CFD-DEM. This may be a bit confusing, as all the methods discussed in this article fall under the CFD umbrella, but the terminology used here is standard in the gas-solid simulation community. To avoid errors when filtering the discrete particle data into a concentration field, grid cells are typically larger than the particle diameter, yet resolutions of Δx ~ 2dp are required for accurate solutions (i.e., grid-insensitive results) (2, 14). Although this grid size is 20 times larger than the typical grid size in DNS, CFD-DEM is still quite computationally expensive. Simulations of large systems containing on the order of 106 particles that are run serially (or on only a few, parallel cores) are unlikely to deliver results in a practical time frame. However, with the explosive growth of computational power over the last few decades, it is now possible to use thousands of processors for a single simulation — allowing for larger system sizes to be modeled. Currently, simulations on the order of 107 particles have been achieved (14, 15), but even those are still woefully short of full, industrial-scale systems. Nonetheless, CFD-DEM offers several advantages over continuum approaches for the solid phase (i.e., TFM): • There is no need for complex closures for solids-phase constitutive relations, which are only tractable for very idealized particles. • It is straightforward to incorporate particle properties such as nonsphericity, size distributions, and variable densities. • Implementation of new physics describing interparticle interactions, such as friction, van der Waals forces, electrostatics, capillary bridging, etc., is straightforward (16–18). Therefore, while CFD-DEM may be computationally
The Survey and its Participants
M
ost of the opinions and discussion in the sections on current status and capabilities, examples, and future outlook are based on results of a survey to gauge industrial perspectives on current and future use of CFD-DEM. The survey consisted of 15 questions and was presented to 34 CFD/CFD-DEM champions working in industries relevant to particle flows. The 18 respondents (a return of over 50%) represented a wide range of the industrial spectrum: • chemicals • energy • petrochemicals • research and development • pharmaceuticals • specialty chemicals • engineering and consulting. The respondents work for companies ranging in size from less than 100 to more than 10,000 employees. If you work with CFD-DEM and would like to participate in the ongoing survey, please visit: https://www.psri.org/ demsurvey?acm=1_56.
Copyright © 2017 American Institute of Chemical Engineers (AIChE)
The detailed, particle-level information provided by CFD-DEM is too valuable to relegate to academic studies of idealized systems. expensive, it is being widely pursued because it can provide valuable information. Another discrete method that has found widespread support in industry is multiphase particle in cell (MP-PIC). MP-PIC is similar to CFD-DEM except that a particle is replaced with a “parcel.” Each parcel, or “cloud,” may represent hundreds, thousands, or even hundreds of thousands of particles. This allows much larger systems to be solved than with CFD-DEM. However, the parcel-parcel collision models remain relatively ambiguous and, unlike CFD-DEM, continuum closures such as the solids pressure are also required because individual particle collisions are not resolved.
Current status and capabilities The use of CFD-DEM as a computational tool has grown rapidly in recent years, particularly in the academic community. This recent growth has certainly crossed over into the industrial community, but not yet to the same extent. The pharmaceutical industry, for instance, which can utilize the granular approximation (no fluid phase) for relatively large pills in a spray coating tumbler, has taken a strong interest in DEM (solids only). However, only about 40% of the individuals surveyed (see sidebar) believe that DEM (multiphase or granular) is already a valuable tool (Figure 3). Coupled CFD-DEM (gas and solids) is primarily used in academic research, likely due to its large computational overhead and complexity. Computational constraints. Although CFD-DEM solutions provide a highly desirable level of detail, they are still computationally expensive and generally cannot tackle full-scale industrial devices. Some of the largest CFD-DEM simulations performed to date employed a CFD grid of over 500 million cells and modeled on the order of 107 [denoted O(107), where the notation O(X) means “on the order of X”] (14). However, large industrial-scale circulating fluidized beds (CFBs) may contain 500 trillion particles, O(1014). Subsystems of the CFBs, such as the bed, the riser, and the primary riser cyclone (including diplegs), at high mass loadings may contain O(1013) to O(1014) particles at any given time. Even smaller components like regenerator cyclones and secondary cyclones may contain O(1010) to O(1012) particles — several orders of magnitude beyond even the largest CFDDEM simulations performed on supercomputers. The large gap between current computational capability and that needed for industrial scale is not easily bridged and is likely a primary reason that the majority of the survey respondents do not yet CEP September 2017 www.aiche.org/cep
41
Computational Methods
Now 39%
In 5 yr 44%
p Figure 3. Most respondents do not yet see CFD-DEM as a valuable industrial tool, although the vast majority of them expect it to achieve such value in less than five years, considering the current rate of software and hardware enhancements.
42
www.aiche.org/cep September 2017 CEP
Example 1: Fully developed flow Recently, Morris, et al., (15) used CFD-DEM to study the flow of particles through a solar receiver in a concentrating solar power (CSP) plant. (Editor’s note: For more on CSP, see the special section in the July 2017 issue of CEP.) The proto100 Respondents, %
In 10 yr 11%
In 20 yr 6%
Half of the survey respondents are willing to hire fully trained PhD holders to implement CFD-DEM tools within their organization. The other respondents would be willing to invest in some type of training on using CFD-DEM tools, ranging from participation in a users group to a short course. The clear emphasis on training reflects the difficulty of selecting the appropriate code. To meet the ever-expanding needs of practitioners, many codes are available to perform CFD-DEM simulations. Not surprisingly, commercial codes dominate the market. Five codes, four of which are commercial, are each used by more than 20% of the survey respondents. However, some respondents also used various open-source and inhouse codes. In addition, 44% of the respondents reported using four or more different codes within a single organization. The survey suggests that support and documentation, i.e., having help available for using the code, is more important than cost as the reason for selecting a particular code (Figure 4). However, in a separate question, no respondents said that they would be willing to spend more than $50,000 for a single license. Therefore, while cost may not be the dominant factor, it is still important. Current capabilities. Despite the constraints, the detailed, particle-level information provided by CFD-DEM is too valuable to relegate to academic studies of idealized systems. With a bit of cleverness and ingenuity, engineers can use CFD-DEM to gain valuable insight into real-world, industrially relevant situations that are not as computationally taxing as a complete CFB, such as the sub-component studies illustrated in Figure 1. The following sections present interesting examples that illustrate the use of CFD-DEM.
80 60 40 20
89%
83%
50%
50%
39%
28%
0
W el l-S up po rte d Ve Doc rif um Va ica en lid tio te at n a d io n n d Av a C ila on b su ility lti o ng f Fr Low ee -o Co f-C st ha or N um rg e be ro fU se rs O pe nSo ur ce
see CFD-DEM as a valuable industrial tool (Figure 3). Time constraints. CFD-DEM can also be taxing for other reasons. It may take several weeks to several months to perform a simulation, even if a sufficiently large number of CPU cores are available. Several factors contribute to the relatively long time needed for CFD-DEM simulations: • To accurately capture the mesoscale dynamics, transient simulations in a statistically steady state need to be run for a sufficiently long period of time so that measured properties do not depend strongly on the duration of time averaging. • Simulations need to be run for an initial period to reach a statistically steady state at which data are usable. • Regardless of the number of particles and number of processors, DEM simulations are limited to a relatively small time step to accurately resolve the collisions, particularly for fine particles (19). Interparticle forces, e.g., van der Waals forces (2), require a realistic spring constant (related to the Young’s modulus of the material) to accurately capture the correct contact time during collisions, which makes time-step limitations even more challenging. Other interparticle forces (electrostatics, capillary bridging, etc.) may present similar computational challenges but have yet to be fully investigated. Fortunately, industry is becoming more accepting of simulations that may take several days or longer to produce results. A majority of the survey respondents were willing to wait a week or a month (44% and 28% of the respondents, respectively) for value-added results from CFD-DEM simulations. Other constraints. The user cost associated with CFD-DEM can be high as well. Applying the underlying methods and interpreting simulation results can be quite challenging; often simply using the software correctly can be difficult. The challenges for using CFD-DEM affect personnel decisions and computer code decisions. A recent bachelor-level graduate is unlikely to be fully equipped to implement a complex CFD-DEM code. Some undergraduate research opportunities offer hands-on experience working with graduate students and professors that position some recent graduates to be ideal CFD-DEM users.
p Figure 4. Respondents say that when choosing a code, they consider support and documentation to be the most important factors in their decisions. Copyright © 2017 American Institute of Chemical Engineers (AIChE)
type design consists of a bank of hollow hexagonal tubes and a field of heliostats (mirrors) that direct the sun’s rays toward the tubes to heat them. A heat-transfer fluid comprised of solid particles collects and stores heat when it flows through gaps between the tubes (Figure 5). The researchers determined that the flow became approximately fully developed after only three rows of tubes. Based on this, they simplified the simulation to a much smaller domain that could provide valuable insights into the behavior of the complete system. In addition, although three-dimensional (3D) simulations are necessary to accurately capture the particle packing structure near the walls (which strongly influences the heat-transfer efficiency), the researchers found that a relatively thin domain (in the direction into the page in Figure 5) was sufficient to accurately capture this behavior. CFD-DEM simulations were also used to probe the effects of different design parameters. For instance, reducing the horizontal spacing of the tubes from 5.04 cm down to 4.44 cm creates a significant stagnation zone on the top row, because the particles choke at the point where the streams converge (just above the vertical segments). In addition, simulation results revealed that thermal gradients near hot spots (high-temperature regions with insufficient particle-wall contacting) produce unacceptably large mechanical stresses, which ultimately required a redesign of the prototype solar receiver. Information of this type can be invaluable when designing novel systems with many variables.
Example 2: System-size-independent metrics Many metrics of gas-solid flows are scale-dependent, presenting a significant challenge to scale-up studies. For example, in a bubbling fluidized bed (BFB), the bubble parameters (size, frequency, and rise velocity) all depend on the size of the system due to wall and bed height effects, and running small-scale simulations provides little useful information about a larger system. However, careful investigations have shown that not all metrics depend on system size. For instance, a successful comparison between small-scale CFD-DEM and large-scale experimental data demonstrated that the superficial gas velocity at which defluidization of a BFB occurs is independent of system size (20). Identifying more system-size-independent measurements of industrial interest may broaden the immediate application of CFD-DEM under current computing capabilities. Example 3: Modeling of coarse-grained closures Coarse-grained methods suffer from the challenges associated with not simulating the mesoscale. Although some of the dynamics are actively captured, a significant portion of clustering and/or bubbling dynamics are simply washed out by coarse grids and/or large parcels. CFD-DEM can be used in smaller, idealized domains that represent only a single grid cell (or parcel) in a coarsegrained simulation. Information can be extracted and used to form closure relations for the coarse-grained methods, which Enhanced Physical Models Particle Shape Particle Attrition (11%)
Tp, K
Accuracy of Collison Model (28%)
(5%)
316 314 312 310 308 306 304 302 300
Interparticle Forces Agglomeration and Deagglomeration (56%) Computational Ease of Creating Simulation Domain Meshing (17%)
Visualization and Post-Processing (11%)
Speed (72%)
p Figure 5. Researchers used CFD-DEM to model particle flow through a solar receiver in a concentrating solar power plant (15). Here, 300-μm particles and air at 300 K enter through the top of the reduced domain and contact heat exchanger tubes held constant at 600 K; flow becomes fully developed after the third row of tubes, just before the domain exit.
Copyright © 2017 American Institute of Chemical Engineers (AIChE)
p Figure 6. The majority of respondents view interparticle forces (agglomeration and deagglomeration) as the most important physical enhancement and speed as the most important type of computational improvement needed for value-added DEM simulations.
CEP September 2017 www.aiche.org/cep
43
Computational Methods
can then be applied at a much larger scale. One of the most important closures is the impact of unresolved structures on the mean drag law. Radl and Sundaresan (21) recently used highly resolved CFD-DEM simulations to constitute a drag law for MP-PIC simulations as a function of the parcel resolution, rectifying one of the outstanding concerns about the coarse-grained method (21, 22). A similar approach was also recently used to propose a TFM boundary condition for conductive heat transfer (23). The successes of this approach bode well for the modeling of other physical mechanisms, particularly chemical reactions, which involve inherently microscale phenomena.
Future outlook While some companies are already using DEM, a significant margin for growth remains among industrial users, especially in energy, chemical, and petrochemical engineering. Based on survey results, we believe improvements in the following four key areas will help spur the adoption of CFD-DEM by the industry at large. Physical improvements. For years, most DEM simulations considered only monodisperse, spherical particles. Today, significant progress has been made to model realistic particle shapes using glued spheres, ellipsoids, cylinders, superquadrics, and other methods (24, 25). Some algorithm improvements for polydispersity have been made, but wide size distributions remain a challenge (26). Figure 6 (top) shows that one of the major needs among industry today is the development and implementation of interparticle force models (e.g., van der Waals forces, liquid bridging, electrostatics), which control the agglomeration or deagglomeration of constituent particles into or out of granules. Consideration of interparticle forces is inherently linked to a second significant need area — modeling the actual collision. It becomes especially important to resolve the collisions correctly when considering interparticle forces. More specifically, using artificially soft particles significantly overestimates the collision duration, increasing the dissipation and making the particles appear more cohesive than they should be (2). Reference 24 provides a comprehensive review of the current state-of-the-art capabilities of DEM for nonspherical particles, breakage, attrition, cohesion, and agglomeration. Computational improvements. While highly valuable, CFD-DEM remains computationally expensive. Figure 6 (bottom) shows that a majority of the survey respondents desire faster existing codes. One promising path to increase the speed of existing codes is for code developers, often engineers with a physics background, to work side-by-side with computational scientists who specialize in highperformance computing (27). Codes written from scratch are rarely developed with performance in mind and typically flow in a logical path that fails to leverage vectorized opera44
www.aiche.org/cep September 2017 CEP
tions and other advanced programming strategies. Another failure of the most current generation of codes is a lack of massive scalability. CFD-DEM codes are unable to properly utilize the computing facilities of supercomputers, which evolve more rapidly than the codes do. Porting computations from CPUs to graphical processing units (GPUs) and reprogramming essential routines is another option to improve the speed of existing codes. Several commercial codes already offer GPU capabilities. Hardware improvements. Although large academic CFD-DEM simulations can model O(107) particles and will soon reach O(108), such simulations are being performed on hundreds to thousands of cores, a computational requirement that is too costly for many companies. The current machines need to become cheaper for industry to tackle simulations on thousands of cores. Next-generation coprocessors offer another possibility for speed improvements. Again, a close
RAY COCCO, PhD, is President and CEO of Particulate Solid Research, Inc. (PSRI) (Email: ray.cocco@psrichicago.com), a consortium-based company with 34 member companies worldwide. He was chair of the AIChE Particle Technology Forum (Group 3) before joining PSRI, and is on the Editorial Advisory Board for Powder and Bulk Engineering and Powder Technology. He is a member of AIChE’s Chemical Engineering Technology Operating Council (CTOC) and the advisory boards for several universities. Cocco is also a member of the International Fluidization Conference Advisory Committee and the meeting chair for the next World Congress in Particle Technology (WCPT8). He has over 40 publications, three book chapters, three patents and patent applications, and numerous invited presentations. He received a BS in chemical engineering from the Univ. of Florida and a PhD from Auburn Univ. WILLIAM D. FULLMER, PhD, is a postdoctoral researcher in chemical engineering at the Univ. of Colorado, Boulder (Email: william.fullmer@ colorado.edu). His interests range from fundamental research of multiphase flow instabilities using computational fluid dynamics to nuclear reactor safety analysis. Fullmer earned his BS, MS, and PhD from Purdue Univ. School of Nuclear Engineering. PEIYUAN LIU, PhD, is a postdoctoral researcher in chemical engineering at the Univ. of Colorado, Boulder (Email: peiyuan.liu@colorado.edu). His experience covers discrete element modeling of particulate flows and multiphase flows, including interparticle forces, agglomeration, and breakage in gas-solid flows. Liu received his BS in materials science from the Univ. of Science and Technology Beijing, China and PhD from the Univ. of New South Wales, Australia. CHRISTINE M. HRENYA, PhD, is a professor of chemical engineering at the Univ. of Colorado, Boulder (Email: hrenya@colorado.edu). Her research program in multiphase flows has resulted in over 100 journal publications and 120 invited lectures. Recent recognitions include the 2014 AIChE Lectureship Award in Fluidization and the 2013 Univ. of Colorado Excellence in Teaching Award. Hrenya served as chair of the 2016 AIChE Annual Meeting, and has previously served as chair of the 2006 Gordon Conference on Granular Flow (Oxford Univ.). She also serves as an associate editor of the AIChE Journal. She received her BS from the Ohio State Univ. and her PhD from Carnegie Mellon Univ., both in chemical engineering.
Acknowledgments The authors thank Aaron Morris, Shankar Subramaniam, Bo Sun, and Sudheer Tenneti for contributing to the figures, and Jesse Capecelatro and Casey LaMarche for stimulating discussions. Funding for this work was graciously provided by the U.S. Dept. of Energy Grant No. DE-FE0026298.
Copyright © 2017 American Institute of Chemical Engineers (AIChE)
working relationship between code developers and computational scientists is essential to fully leverage emerging technologies. Collaboration. Partnerships between industry and academia could minimize the need to buy machines capable of 1,000-core (or more) simulations. The industrial sector is most likely to face large computational challenges and the academic sector is most likely to have access to large machines. For example, the eXtreme Science and Engineering Discovery Environment (XSEDE) started an Industry Challenge program funded by the U.S. Dept. of Energy’s Crosscutting Research program with the specific intent to bring academic and industrial researchers together. The director of the XSEDE Industry Challenge, David Hudak, stated (28): “Large-scale problems in open science are often
discovered by trying to solve industrial problems. Attempting to solve these problems requires multidisciplinary research teams using the fastest computers available. … . I am confident these teams will produce results that have both scientific and economic impact.” This current work pairs computational scientists with engineers from academia and industry to bridge the gap between basic and applied research by enhancing the speed and scalability of an open-source CFD-DEM code. Its ultimate goal is to simulate an industrially relevant system containing O(108) particles. Projects like this will hopefully help make CFD-DEM more useful to industrial researchers. If you work in an area of gas-solid particulate flow and CFD-DEM is not already in your toolbox, perhaps it should be. CEP
Literature Cited 1. Pöschel, T., and T. Schwager, “Computational Granular Dynamics: Models and Algorithms,“ Springer-Verlag, Berlin (2005). 2. Liu, P., et al., “Fine-Particle Defluidization: Interaction Between Cohesion, Young’s Modulus, and Static Bed Height,” Chemical Engineering Science, 145, pp. 266–278 (2016). 3. Royer, J. R., et al., “High-Speed Tracking of Rupture and Clustering in Freely Falling Granular Streams,” Nature, 459 (7250), pp. 1110–1113 (June 2009). 4. Goldhirsch, I., “Rapid Granular Flows,” Annual Review of Fluid Mechanics, 35, pp. 267–293 (2003). 5. Brilliantov, N., and T. Pöschel, “Kinetic Theory of Granular Gases,” Oxford Univ. Press, Oxford, U.K. (2004). 6. Mitrano, P. P., et al., “Kinetic-Theory Predictions of Clustering Instabilities in Granular Flows: Beyond the Small-Knudsen-Number Regime,” Journal of Fluid Mechanics, 738, pp. R2–12 (Jan. 2014). 7. Fullmer, W. D., and C. M. Hrenya, “Quantitative Assessment of Fine-Grid Kinetic-Theory-Based Predictions of Mean-Slip in Unbounded Fluidization,” AIChE Journal, 62 (1), pp. 11–17 (Jan. 2016). 8. Agrawal, K., et al., “The Role of Meso-Scale Structures in Rapid Gas-Solid Flows,” Journal of Fluid Mechanics, 445, pp. 151–185 (Oct. 2001). 9. McMillan, J., et al., “Particle Cluster Dynamics During Fluidization,” Chemical Engineering Science, 100, pp. 39–51 (Aug. 2013). 10. Igci, Y., et al., “Filtered Two-Fluid Models for Fluidized Gas-Particle Suspensions,” AIChE Journal, 54 (6), pp. 1431–1448 (June 2008). 11. Parmentier, J. F., et al., “A Functional Subgrid Drift Velocity Model for Filtered Drag Prediction in Dense Fluidized Bed,” AIChE Journal, 58 (4), pp. 1084–1098 (Apr. 2012). 12. Schneiderbauer, S., and S. Pirker, “Filtered and HeterogeneityBased Subgrid Modifications for Gas-Solid Drag and Solid Stresses in Bubbling Fluidized Beds,” AIChE Journal, 60 (3), pp. 839–854 (Mar. 2014). 13. Tenneti, S., and S. Subramaniam, “Particle-Resolved Direct Numerical Simulation for Gas-Solid Flow Model Development,” Annual Review of Fluid Mechanics, 46, pp. 199–230 (2014). 14. Capecelatro, J., et al., “On Fluid-Particle Dynamics in Fully Developed Cluster-Induced Turbulence,” Journal of Fluid Mechanics, 780, pp. 578–635 (Oct. 2015).
Copyright © 2017 American Institute of Chemical Engineers (AIChE)
15. Morris, A. B., et al., “Simulations of Heat Transfer to Solid Particles Flowing Through an Array of Heated Tubes,” Solar Energy, 130, pp. 101–115 (2016). 16. Israelachvili, J. N., “Intermolecular and Surface Forces. 3rd ed.,” Academic Press, Burlington, MA (2011). 17. Visser, J., “Vanderwaals and Other Cohesive Forces Affecting Powder Fluidization,” Powder Technology, 58 (1), pp. 1–10 (May 1989). 18. Waitukaitis, S. R., et al., “Size-Dependent Same-Material Tribocharging in Insulating Grains,” Physical Review Letters, 112 (21), pp. 218001 (May 2014). 19. Liu, P. Y., and C. M. Hrenya, “Challenges of DEM: I. Competing Bottlenecks in Parallelization of Gas-Solid Flows,” Powder Technology, 264, pp. 620–626 (Sept. 2014). 20. LaMarche, C. Q., et al., “A System-Size Independent Validation of CFD-DEM for Noncohesive Particles,” AIChE Journal, 61 (12), pp. 4051–4058 (Dec. 2015). 21. Radl, S., and S. Sundaresan, “A Drag Model for Filtered EulerLagrange Simulations of Clustered Gas-Particle Suspensions,” Chemical Engineering Science, 117, pp. 416–425 (Sept. 2014). 22. Benyahia, S., and J. E. Galvin, “Estimation of Numerical Errors Related to Some Basic Assumptions in Discrete Particle Methods,” Industrial & Engineering Chemistry Research, 49 (21), pp. 10588–10605 (Nov. 2010). 23. Morris, A. B., et al., “A Conductive Heat Transfer Model for Particle Flows Over Immersed Surfaces,” International Journal of Heat and Mass Transfer, 89, pp. 1277–1289 (Oct. 2015). 24. Guo, Y., and J. S. Curtis, “Discrete Element Method Simulations for Complex Granular Flows,” Annual Review of Fluid Mechanics, 47, pp. 21–46 (2015). 25. Potapov, A., “Approaches for Accurate Modeling of Particle Attrition in DEM Simulations,” presented at the 2016 Frontiers in Particle Science and Technology, Houston, TX (Apr. 11–13, 2016). 26. Berger, K. J., and C. M. Hrenya, “Challenges of DEM: II. Wide Particle Size Distributions,” Powder Technology, 264, pp. 627–633 (Sept. 2014). 27. Syamlal, M., et al., “Computational Science: Enabling Technology Development,” Chemical Engineering Progress, 107 (1), pp. 23–29 (Jan. 2011). 28. XSEDE, “XSEDE Industry Challenge Program,” www.xsede.org/ industry-challenge-program (accessed in May 2016).
CEP September 2017 www.aiche.org/cep
45