Andrei V. Smirnov

Large eddy simulations of in-cylinder turbulence for internal combustion engines: A review

Abstract While engineering applications of the large eddy simulation (LES) technique are becoming a common reality in many branches of engineering and science, its application to engine flows has lagged behind due to the relatively more complex nature of both the flow and the geometry relevant to in-cylinder flows. In this paper a review of the limited number of LES applications to engine flows is given, and most significant results from these studies are presented. Also, the LES formulation appropriate for engine applications is briefly described, along with the main characteristics of in-cylinder flows. As expected, engine applications of LES are not of the so-called ‘high-fidelity’ type, but rather they employ formally second-order accurate numerical schemes in conjunction with finite volume formulation. The subgrid scale (SGS) models used are also kept as simple as possible, mostly using a variant of the Smagorinsky model. Nevertheless, this review reveals that even with relatively coarse grids, LES captures much more interesting features of in-cylinder flows, such as the large coherent vortical flow structures developed during the intake stroke. In the opinion of the present authors, a low-resolution LES provides a better solution than RANS (Reynolds averaged Navier-Stokes) with moderate grid resolution because the important features of flow dynamics cannot be reproduced in RANS due to the high level of non-physical diffusion. Of course, some overheads in computational costs must be paid for this benefit obtained from LES.

Prediction of In-Cylinder Turbulence for IC Engines

Abstract This paper presents the preliminary results of some of a few of its kind efforts in large eddy simulation (LES) of engine flows to predict turbulent fluctuations, and the statistics of turbulence quantities inside IC engine cylinders. For this purpose, the well-known engine simulation code, KIVA, is used with special precautions to keep the numerical accuracy at a sufficiently high level, as well as using relatively fine grid resolution. The capabilities of this code are tested against benchmark cases, such as lid-driven cavity flow, and swirling and non-swirling free jet flows. It is then applied to a typical engine geometry under motored conditions. In particular, turbulence generated during the intake stroke, and the instabilities induced by a typical piston-bowl assembly are investigated. The computed velocity fluctuations, correlation coefficients and energy spectra of turbulent fluctuations are compared to experimental results. The predictions seem to extend well into the inertial range of turbulence and depict a good qualitative agreement with measurements. The results also shed light into the mechanisms by which turbulence may be generated by the piston-bowl assembly.

Influence of Radiative Heat Transfer on Variation of Cell Voltage Within a Stack

In this study, a numerical investigation of cell-to-cell voltage variation by considering the impact of flow distribution and heat transfer on a stack of cells has been performed. A SOFC stack model has been previously developed to study the influence of flow distribution on stack performance (Burt, et al., 2003). In the present study the heat transfer model has been expanded to include the influence of radiative heat transfer between the PEN (positive electrode, electrolyte, negative electrode) and the neighboring separator plates. Variations in cell voltage are attributed to asymmetries in stack geometry and nonuniformity in flow rates. Simulations were done in a parallel computing environment with each cell computed in a separate (CPU) process. This natural decomposition of the fuel cell stack reduced the number of communicated variables thereby improving computational performance. The parallelization scheme implemented utilized a message passing interface (MPI) protocol where cell-to-cell communication is achieved via exchange of temperature and thermal fluxes between neighboring cells. Inclusion of radiative heat transfer resulted in more uniform temperature and voltage distribution for cases of uniform flow distribution. Non-uniform flow distribution still resulted in significant cell-to-cell voltage variations.© 2003 ASME

Domain coupling with the DOVE scheme

Publisher Summary This chapter discusses the domain coupling with the DOVE scheme. When domain decomposition is applied to an unstructured mesh, the realization of conformal mapping leads to highly irregular boundaries for each sub-domain, which in turn creates extra errors in the gradient approximations. One possibility to avoid this is to construct all parallel domains independently, and then combine them together using inter-domain interpolation. This study describes the implementation of one such method based on automatic detection of domain overlaps, and setting of communication interfaces (DOVE). The usability of this scheme is demonstrated in the case of two continuum solvers: a Poisson solver and an artificial compressibility flow-solver. This scheme can be easily extended to handle dynamic overlapping regions when the domains may change shape or move with respect to each other. One of the advantages of the proposed scheme is that it can be used to construct large and complex domains without the need to store the entire grid on a single computing node at the same time. The DOVE scheme can be easily extended to handle dynamic overlapping regions when the domains may change shape or move with respect to each other. In this case, the DOVE initialization functions should be called from within the main solver at certain times to update the changing geometry of the overlap region and re-generate the connectivity arrays.

Express Risk Assessment throughWeb Access to Simulation Data

A prototype of an express risk assessment system is proposed and implemented as a combination of a distributed simulator and a Web interface. A case of aerosol dispersion in an urban environment is investigated. Fast response of the system is achieved by replacing complex 3D simulations with information retrieval from aerosol dispersion database. The database is created by extensive simulations of different aerosol dispersion scenarios. Cluster or grid computing environments are most suitable for these type of simulations. The reduction of system response from a few days to several seconds can be achieved compared to realtime 3D simulations

Application of Reactive Molecular Dynamics to Simulate Diffusion and Reaction in a Solid Oxide Fuel Cell Pore

In a typical solid oxide fuel cell (SOFC), the kinetics of the gas phase reactions in the porous anode and electrochemical reactions at the triple phase boundary are generally unknown. Due to the unavailability of non-destructive experimental methods, factors affecting the performance of SOFC systems, especially the loss in performance due to contaminants, are usually deduced from many days of experiments. In this paper a Reactive Molecular Dynamics (ReMoDy) model based on collision theory is introduced and applied to simulate the behavior of species inside a SOFC pore. Using novel simulation methods, algorithms and visualization techniques ReMoDy has the ability to simulate chemical reactions involving tens of millions of molecules and determining the thermo-physical properties of the fluids from intermolecular energies and forces. In the current work two cases of molecular dynamics simulation inside a micro pore were analyzed. In the first case diffusion of hydrogen molecules was studied inside a 0.03125 μm3 cube. The diffusion coefficients obtained from this simulation are compared to the ones obtained using Chapman-Enskog correlations. In the second case gas phase and surface reactions were modeled for Syngas oxidation in a 1 μm3 cube representing a SOFC electrode pore. For this case detailed gas phase and surface reaction mechanisms involving 13 species and 63 reactions is included. Future studies will include the calculation of diffusion coefficients, rates of formation of different species, and comparison with published data. The results can be used for the verification of continuum models.Copyright

Physically based node distributions for mesh generation

Abstract A new physically based approach to mesh generation is proposed, which uses the variants of Monte-Carlo (MC) technique. The mesh nodes are treated as interacting particles and their positions are determined using energy minimization. When the approach is extended with the grand-canonical MC scheme the optimization is performed for both the number of particles and their positions. It is shown how a molecular dynamics technique can be applied to accelerate the convergence of the simulations. Local mesh refinement can be achieved with appropriate node spacing functions. The final mesh is created by connecting the generated node distributions with constrained Delaunay triangulation. Well-shaped triangular or tetrahedral mesh elements are usually obtained. The proposed method is simple, flexible, and works in an identical way in spaces of any number of dimensions.

Embarrassingly Parallel Computations of Bubbly Wakes

Publisher Summary The chapter discusses embarrassingly parallel computations of bubbly wakes. The idea of “embarrassingly” parallel computations presumes a complete absence or a small degree of interprocessor communications during the parallel execution of a multiprocessor problem. In applications to continuum mechanics, it usually means that the domain decomposition is either not used at all or there is a very loose coupling among parts of the decomposed domain. This idea is attractive because of low communication overhead and linear scale-up of the computations. However, this kind of parallelism can only be realized for limited classes of problems. One such class involves statistical modeling, where the results are obtained as an average over a set of independent statistical ensembles. In this case, each ensemble can be simulated on a separate processor and no communication among the processors is necessary. Unfortunately, most problems of fluid dynamics do not fall under this category. Parallel computational fluid dynamics (CFD) computations are notoriously hard to realize because of the inherent strong coupling between the variables and sub-domains.

Multi-Physics Simulations of Fuel Cells Using Multi-Component Modeling

Multi-physics simulations based on multi-component multi-solver modeling approach were performed for high-temperature fuel cells. The developed approach was primarily aimed at the design of complex multi-component engineering systems. It extends the libraries of earlier designed multi-physics system with component classes, which makes it particularly suitable for modeling of fuel cell systems. The C++ based class hierarchy enables simple implementations of different physical models based on general 3D PDE (partial differential equations) solvers, or simplified engineering 1D or 2D models. Simulations of solid-oxide fuel cells were performed using a combined transport solver in multi-species environment. The components included the PEN complex (anode, cathod, electrolyte), air/fuel channels, interconnects, seals and ambient environments. Species concentrations, mass, momentum, energy fluxes were solved for different components. Models for unsteady fluid dynamics of species, heat transport, electrochemistry and electric currents were combined within different components and interfaced for common variables at the inter-component boundaries. The results include distributions of temperature, species concentrations and mass fluxes inside co-flow and cross-flow fuel cells with different number of channels.Copyright

Large Eddy Simulations of a Bubbly Mixing Layer

In this paper we present the results of the application of a Lagrangian particle dynamics (LPD) method and a random flow generation technique (RFG) developed by the authors used in conjunction with large eddy simulations (LES) applied to the case of a two-phase bubbly mixing layer. The objective is to validate the present hybrid Lagrangian-Eulerian approach within the context of LES. The dynamics of the vortices and bubble concentrations reproduced in simulations are in a good agreement with experimental data.Copyright