Advanced turbulence models are tested for internal combustion engine intake simulations. The ubiquitous k ? ? model served as the baseline test case. 3 Non-Linear RANS type turbulence models, 2 algebraic Reynolds Stress Models, the v 2 ? f model and 5 LES models are tested for piston engine intake simulations.
Simple flow tests are conducted with backward facing step and square duct flows. Mean flow as well as turbulence intensities are compared to experimental values. A simple 2 dimensional piston-cylinder assembly with a moving mesh is used to further identify the most suitable RANS type turbulence model. Due to the need of a 3 dimensional geometry for proper LES model testing, simulations with a full 3 dimensional cylinder mesh of a contemporary diesel engine mesh is used for both RANS and LES turbulence models.
The Non-Linear models are presented to be a mathematically convenient way to incorporate the non isotropic nature of turbulence but in practice proves to be cumbersome because of the non-linear terms added to the strain rate tensor which hampers the stability of the model. Algebraic Reynolds Stress models have exactly the same structure as the non-linear models thus exhibit the same behavior. They differ only in a more rigorous derivation procedure of the coefficients of the non-linear terms. The v 2 ? f model proves to be the best RANS type turbulence model with predictions closest to experimental values.
LES models are inherently different from RANS models, where the filtering process is spatial rather than over a period of time or cycle as it is with RANS. LES is less dissipative depending on the model and grid size, showing more detail and spatial variability. Consequently, results of RANS turbulence models and experiments are not directly comparable. Due to the constraints of a single cycle for the 3 dimensional simulation of a diesel engine intake, spatial averaging is utilized in order to compare LES results with RANS and experimental results.