Combustion simulation involves simulation of the power stroke during the engine cycle, starting from closing of valves to the end of the compression stroke. Since the valves are closed or in the process of closing, the combustion chamber is the chief flow domain, and the piston the sole moving part. These simulations are also known as "in-cylinder combustion" and though multi-dimensional, are less complicated geometrically than a port flow simulation. In addition, if the geometry is rotationally symmetric and has a single feature like a very high pressure spray that dominates the flow in the calculation, the entire domain can be modeled as a sector to speed up the calculation.
Typically, the initial flow field at this stage is obtained from
a cold flow simulation if the full geometry is used
patching-in based on a cold flow analysis
running the piston without combustion to obtain charge compression
As with cold flow, a moving deforming mesh model is used for the piston motion. Geometric decomposition is not required here, since only the piston motion is included in the simulation. Hence, in-cylinder combustion simulations typically do not include the modeling of the fluid dynamics in the valve port region and their effect on combustion.
Models are used to account for the fuel spray, combustion and pollutant formation. For direct injection engines, the fuel spray from the tip of the nozzle injector is introduced at the specific crank angle and duration using a spray model. For port fueled engines, it is assumed that the combustion charge is well mixed. A chemical mechanism describing the reaction of vapor fuel with air is used to describe combustion, and models for turbulence-chemistry interaction are specified. Sub-models for NOx and soot formation are used to calculate pollutant formation, which can be coupled with the combustion calculation or calculated as a postprocessing step.
With in-cylinder combustion, the main challenge lies in the physics for spray modeling and combustion. The spray is composed of a column of liquid entering the domain at high speed which subsequently breaks into droplets due to aerodynamic forces. These droplets can coalesce into larger droplets or break into even smaller droplets, all while exchanging mass with the surrounding gases. Sub-models for coalescence and breakup, as well as heat and mass transfer, are used to capture spray dynamics. The CFD mesh has to be sufficiently resolved to capture the coupling between the liquid droplets and the gases in the cylinder accurately. If liquid spray impinges upon the cylinder walls, it is possible to form a thin liquid film which undergoes its own processes of movement and vaporization and requires a separate treatment.
To calculate combustion, detailed chemical mechanisms for pure fuels that constitute the components of diesel fuel and gasoline involve hundreds of species and thousands of reactions. These reactions are coupled with the fluid dynamics due to the similar time scales of fluid mechanical motions and chemical reactions. The energy release from combustion increases pressures and temperatures for the fluid flow, which affects the fluid motions inside the cylinder. A direct computation of this coupled interaction without submodels while including detailed chemistry is staggeringly expensive in terms of computation time and is impractical for complex geometries.
Reduced order mechanisms capture most of the essential chemistry in a narrower range of temperature and equivalence ratio, and are used along with a submodel for turbulence-chemistry interaction. One such model is the Probability Density Function (PDF) approach which allows an efficient computation of turbulence-chemistry interaction. Flame propagation is modeled using a progress variable based approach such as the Zimont model, which calculates the transient flame front speed and location. These approaches allow computation of the combustion process on large meshes in complex geometries with a reasonable computational power.
Simplified mechanisms are used to compute the NOx formation due to
high temperatures (thermal NOx)
nitrogen in the fuel (fuel NOx)
fuel reactions in the flame front (prompt NOx)
sulfur oxides in the fuel (SOx)
soot formation
Since these pollutants are generally a very small percentage of the total mass in the cylinder, these calculations can be decoupled from the calculation of the main energy release. In some cases, this is done as a postprocessing operation at the end of the simulation, but it is more accurate to include the pollutant formation in the simulation, especially for the pollutants arising from incomplete combustion that oxidize later in the cycle. In terms of automation and process compression, the problem setup at the solver stage can benefit from automation.