A. Nassiri-Toosi, Ph.D.
Automotive Engineering Department,
Iran University of Science and Technology
Engine Modeling and KIVA
The role of real-world mobile source emissions on air quality is well established. Due to increasing environmental concerns, there has been a growing requirement to develop cleaner engine. This is particularly true in the vehicle industry where there are strong competitive pressures to bring new vehicles to the market in shorter times to respond quickly to costumer needs.
For this purpose, new engine technologies are becoming more rapidly implemented such as gasoline direct injection, turbo-charging, exhaust gas re-circulation and treatment. These recent advances would have been impossible without the progress in CFD modeling.
This technique is used to predict fluid properties within the engine. This method has extremely large computing overheads, but is invaluable as a tool for greater understanding of the airflow patterns within the engine. Various CFD codes such as KIVA, STAR CD, FIRE, VECTIS and FLUENT have been used for model development and application. Mesh preparation has been a bottleneck for 3-D simulations of in-cylinder phenomena of internal combustion engines.
KIVA is a large FORTRAN program developed at Los Alamos National Laboratory for internal combustion engine simulation. It was designed to numerically simulate two- and three-dimensional chemically reactive turbulent fluid flows with sprays, as in the simulation of internal combustion engines. With the introduction of KIVA-3V in March 1997, an effective approach to modeling moving valves became available.
KIVA code has been integrated into most of the top CFD modeling programs and is used by the automotive and heavy-duty diesel engine industries to develop energy-efficient, emission compliant internal combustion engines. Also it is widely used for model development in academia due to the availability of the source code. However, its capability for resolving complex geometries is limited.
Along with KIVA_3V, there is a pre-processor called K3PREP. K3PREP is a basic grid generator and is not designed to generate very complex geometries that KIVA-3V is capable of running. Hence it only can be applied for simple or simplified engine geometry. Time consumption is another concern of K3PREP. Hence one can say mesh preparation is the KIVA bottleneck for 3-D simulations of engines with complex geometry.
Nowadays there are commercial softwares for mesh generation which are user friendly and the support is available. Since a block-structured mesh with connectivity defined through indirect addressing is adopted in KIVA-3V, most of commercial softwares could not generate mesh to meet the requirements of KIVA-3V. The only commercial software which supports KIVA_3V requirements is ICEM CFD. It is a product of ANSYS ICEM CFD Engineering With a built in KIVA-3V translator. It can be used as a stand-alone mesh generator which assigns vertex flags and cell boundary conditions to directly produce the KIVA-3V grid file (itape17).
Also, since engine simulation involves moving piston and valves, it is necessary to re-mesh the computational domain for different piston and valve positions. In KIVA-3V code the dynamic mesh management includes snapping and rezoning processes which based on the characteristics of the mesh nodes. The rezoning subroutine in KIVA-3V is not universal or generalized, hence for a particular engine geometry, one need to develop a proper rezoning algorithm.
The engine used in this study belongs to Pride. This car is designed by KIA motors and nowadays is manufactured in Iran by SAIPA Company. As can be seen in Figures 1-5, the engine has got complex geometry and also to generate the intake runner by using K3PREP, it is time consuming and difficult.
A CAD version of the geometry was generated using Solidwork package. It was imported into ICEM CFD package. By applying the proper boundary conditions, the itape17 file was generated using special interface program for KIVA, utilized in ICEM CFD package.
The generated mesh file (itape17) was imported into KIVA-3V code and during the full cycle calculation only 36 non-convex cells was detected. The total number of cells was 43447.
Figure 6 show a non-hydro run of the engine and Figures 7 & 8 show the valves movements. In Figures 9 & 10 the dynamic meshes are shown. As can be seen the mesh is adjusted according to the piston and valves position.
Typical results of the engine simulation are shown in Figures 11 & 12. In figure 11 the fuel particle trace, injected via an injector is shown. In Figure 12 velocity components in the same plane are shown.
Recently further study is carrying out to develop a general dynamic mesh management algorithm in order to be used for wide range of different engine geometries without any need for a modification.