Methodologies for the Numerical Simulation of Fluid Flow in Internal Combustion Engines
Abstract
The increase in the capability of computing in conjunction with the development of new mathematical models and numerical methods, allow to deal with the resolution of complex problems of importance for both science and engineering. Among these, the CFD (Computational Fluid Dynamics) problems in moving domains, such as Fluid-Structure Interaction (FSI) problems, are a topic of particular interest for researchers because of the difficulty that they present and the large number of applications in which these kind of problems are present. One of such problems is the computation of in-cylinder flows in internal combustion (IC) engines.
The modeling of IC engines is a multidisciplinary subject that involves chemical ther- modynamics, fluid mechanics, turbulence, heat transfer, combustion, and numerical meth- ods. In this thesis, the focus is placed on some aspects of the computational resolution of the fluid dynamics problem. In particular, the topics addressed are the mesh dynamics problem, the resolution of flows at low Mach numbers, and the coupling of 1D/multi-D domains for compressible flows. When an Arbitrary Lagrangian Eulerian (ALE) strategy is applied to solve problems with deformable domains, it is necessary to have a Computational Mesh Dynamics (CMD) technique to resolve the dynamics of the mesh. While the movement of the mesh is an artificial field in a FSI problem, its significance is relevant because it affects considerably the efficiency and accuracy of the computation. For in-cylinder flows in IC engines the movement of the boundary domain is known a priori. In these cases the domain has a very high relative deformation and even changes on its topology. This demands great robustness from the CMD strategy to avoid an excessive deterioration of the grid quality and to reduce the number of remeshing needed in the whole simulation.
The flow inside of an IC engine is characterized by a low Mach number, except in the early moments in which the exhaust valve (or port) is opened. The numerical methods for compressible flow based on the density fail when they are applied to flows with low Mach numbers, which is due to the bad conditioning of the system of equations. For this reason, it is necessary to apply a technique that allows the resolution of compressible flows in all the range of Mach numbers, especially in the low Mach limit.
Then, to perform a simulation in an IC engine is necessary to have a CFD code able to compute compressible turbulent flows with low (and also relatively high) Mach numbers in deformable 3D domains [65]. Given the highly complex geometry of the engines and the physical processes that occur within them, it is at present only possible to solve one part of such machines with a 3D model. In this way, and because of its dynamic behavior, another difficulty that appears is related to the boundary conditions to impose to the model. Usually, these problems are addressed by the simulation of the rest of the engine through 0D/1D models, which is achieved in one hand, modeling the entire machine simultaneously (but the level of detail varies depending on the model) and, on the other hand, providing appropriate conditions to the 3D code. Applying the above approximation, the need to couple appropriately the solutions obtained in the computing domains arises, which can be calculated by different codes.
The large spread in length and time scales of in-cylinder flows in IC engines requires a high degree of refinement in the finite element mesh and, then requires very large computational resources. Thus, a parallel code is needed in order to achieve accurate results in that problems. In addition, due to explicit and semi-implicit schemes have demonstrated to be inefficient when they are applied to IC engines [32], a full implicit scheme might be used. [Tesis presentada como parte de los requisitos de la Facultad de Ingeniería y Cs. Hídricas de la Universidad Nacional del Litoral para acceder al grado de Doctor en Ingeniería Mención en Mecánica Computacional. Defendida 2009-05-04]
The modeling of IC engines is a multidisciplinary subject that involves chemical ther- modynamics, fluid mechanics, turbulence, heat transfer, combustion, and numerical meth- ods. In this thesis, the focus is placed on some aspects of the computational resolution of the fluid dynamics problem. In particular, the topics addressed are the mesh dynamics problem, the resolution of flows at low Mach numbers, and the coupling of 1D/multi-D domains for compressible flows. When an Arbitrary Lagrangian Eulerian (ALE) strategy is applied to solve problems with deformable domains, it is necessary to have a Computational Mesh Dynamics (CMD) technique to resolve the dynamics of the mesh. While the movement of the mesh is an artificial field in a FSI problem, its significance is relevant because it affects considerably the efficiency and accuracy of the computation. For in-cylinder flows in IC engines the movement of the boundary domain is known a priori. In these cases the domain has a very high relative deformation and even changes on its topology. This demands great robustness from the CMD strategy to avoid an excessive deterioration of the grid quality and to reduce the number of remeshing needed in the whole simulation.
The flow inside of an IC engine is characterized by a low Mach number, except in the early moments in which the exhaust valve (or port) is opened. The numerical methods for compressible flow based on the density fail when they are applied to flows with low Mach numbers, which is due to the bad conditioning of the system of equations. For this reason, it is necessary to apply a technique that allows the resolution of compressible flows in all the range of Mach numbers, especially in the low Mach limit.
Then, to perform a simulation in an IC engine is necessary to have a CFD code able to compute compressible turbulent flows with low (and also relatively high) Mach numbers in deformable 3D domains [65]. Given the highly complex geometry of the engines and the physical processes that occur within them, it is at present only possible to solve one part of such machines with a 3D model. In this way, and because of its dynamic behavior, another difficulty that appears is related to the boundary conditions to impose to the model. Usually, these problems are addressed by the simulation of the rest of the engine through 0D/1D models, which is achieved in one hand, modeling the entire machine simultaneously (but the level of detail varies depending on the model) and, on the other hand, providing appropriate conditions to the 3D code. Applying the above approximation, the need to couple appropriately the solutions obtained in the computing domains arises, which can be calculated by different codes.
The large spread in length and time scales of in-cylinder flows in IC engines requires a high degree of refinement in the finite element mesh and, then requires very large computational resources. Thus, a parallel code is needed in order to achieve accurate results in that problems. In addition, due to explicit and semi-implicit schemes have demonstrated to be inefficient when they are applied to IC engines [32], a full implicit scheme might be used. [Tesis presentada como parte de los requisitos de la Facultad de Ingeniería y Cs. Hídricas de la Universidad Nacional del Litoral para acceder al grado de Doctor en Ingeniería Mención en Mecánica Computacional. Defendida 2009-05-04]