Cold flow analysis involves modeling the airflow and possibly the fuel injection in the transient engine cycle without reactions. The goal is to capture the mixture formation process by accurately accounting for the interaction of moving geometry with the fluid dynamics of the induction process. The changing characteristics of the air flow jet that tumbles into the cylinder with swirl via intake valves and the exhaust jet through the exhaust valves as they open and close can be determined, along with the turbulence production from swirl and tumble due to compression and squish.
This information is very useful to ensure that the conditions in the cylinder at the end of the compression stroke are right for combustion and flame propagation. High turbulence levels facilitate rapid flame propagation and complete combustion during the power stroke. A well mixed and highly turbulent air flow is critical to ensure the right air/fuel ratio throughout the combustion. CFD can also assess the level of charge stratification.
However, since cold flow simulations do not include the significant thermodynamic changes that accompany combustion, the flow characteristics during the power and exhaust strokes do not reflect reality. In terms of validation, experimental PIV (Particle Image Velocimetry) or LDV data in cycling engines is not easy to obtain as with port flow analysis, but transparent pistons and cylinders have been used to gather velocity information for some engine configurations.
Setting up the CFD model for cold flow analysis involves additional work in specifying the necessary information to compute the motion of the valves and piston in addition to the boundary conditions, turbulence models and other parameters. This includes specifying valve and piston geometry, along with the lift curves and engine geometric characteristics in order to calculate their position as a function of crank angle. Since the volume in the cylinder is changing shape throughout the engine cycle, the mesh has to change accordingly. Different approaches to automatically modify the mesh during motion also need to be specified. The CFD calculation is an inherently transient computational problem when involved with moving deforming or dynamic mesh. All the geometric motion is a function of a single parameter, the position of the crankshaft in its rotation, or crank angle.
The preprocessing from geometry to solver setup is typically time consuming and challenging. Here, you have to separate or decompose the geometry into moving and stationary parts. Typically, the intake ports are split off from the cylinder and valves. The region between the valve margin and valve seat, which opens and closes during valve motion may be separated. The combustion chamber and piston region may be also decomposed or separated into smaller parts. Then each part can be meshed accordingly for the solver setup. Any errors at this stage can lead to failures downstream during the solution process.
The run times for solver runs can be fairly long since the motion is typically resolved with small time steps (approximately 0.25 crank angle) to get accurate results and the simulation is run for two or three cycles to remove the initial transients. Finally, the large volume of transient data that results from the CFD solution needs to be postprocessed to obtain useful insight and information. Thus cold flow analysis would also benefit from design automation and process compression.