The combustion process in a diesel engine was simulated using KIVA-II, a multi-dimensional computer code. The original combustion model in KIVA-II is based on chemical kinetics, and thus fails to capture the effects of turbulence on combustion. A mixing-controlled, eddy break-up combustion model was implemented into the code. Realistic diesel fuel data were also compiled. Subsequently, the sensitivity of the code to a number of parameters related to fuel injection, mixing, and combustion was studied. Spray injection parameters were found to have a strong influence on the model's predictions. Higher injection velocity and shorter injection duration result in a higher combustion rate and peak pressure and temperature. The droplet size specified at injection significantly affects the rate of spray penetration and evaporation, and thus the combustion rate. Contrary to expectation, the level of turbulence at the beginning of the calculation did not affect fuel burning rate.
THE CLEAN AND EFFICIENT conversion of the chemical energy of the fuel to useful work in internal combustion (IC) engines is influenced by the in-cylinder processes and the intake and exhaust flows. Design engineers need to understand these phenomena in detail using available experimental and analytical tools. A new generation of multi-dimensional reacting flow computer models has evolved in the last fifteen years (e.g. the KIVA-II [1, 2, 3 and 4], RPM [5] and PHOENICS [6] codes). These simulations promise to be an invaluable companion to laboratory experiments.
A widely used multi-dimensional program is KIVA-II [1, 2, 3 and 4], developed at Los Alamos National Laboratory. KIVA-II is a two or three-dimensional code that describes the fluid, chemical and spray processes that occur in IC engines. The model solves for the conservation of momentum, energy, and mass for an arbitrary number of species in the gaseous phase. The spray dynamics are solved using a discrete particle technique. The effects of evaporation, droplet breakup, collisions (coalescence or deflection) and viscous and turbulent interactions with the ambient gas are accounted for. The effects of turbulence are represented by a k-e model modified to include the effects of spray interaction. Chemical reactions can be specified either as equilibrium (for fast reactions) or kinetic reactions. The kinetic reaction rates are expressed in Arrhenius form.
KIVA-II models the physical phenomena which occur inside the combustion chamber from first principles. Despite the sophistication of the submodels, certain assumptions must inevitably be made due to lack of knowledge about the details of the mechanisms which characterize the various processes. These simplifications often result in errors in the predictions. Invariably, some empirical calibration constants are introduced due to the simplifications. Furthermore, some of the physically meaningful parameters need to be adjusted to compensate for modeling errors. For example, the chemical reaction constants must be adjusted to account for intermediate reaction steps not included in the mechanism. Finally, some of the physical variables that must be provided as inputs to the program, as for example the size of the droplets injected, cannot be measured directly [7]. Therefore, the “proper” values of the modeling constants and physical parameters must be established for the code to produce realistic predictions. To achieve this goal, experimental validation and computational sensitivity studies must be performed to assess the importance of these variables on code predictions.
Despite the widespread usage of KIVA-II and its predecessors CONCHAS, CONCHAS-SPRAY and KIVA, only a limited number of validation studies appear in the literature. These studies [8, 9, 10 and 11] have compared some of the model's predictions on combustion, spray characteristics, heat transfer, and emissions with experimental data; they also tested the code's sensitivity to a small number of input parameters. The objective of the present work is to conduct more comprehensive sensitivity studies to examine the effect of various parameters related to the spray, mixing, and combustion processes on the prediction of combustion in a direct injection diesel engine. Before we proceed with the presentation of our results, previous studies are summarized and the modifications we implemented in the KIVA-II code are presented.