Definition of a LES Numerical Methodology for the Simulation of Engine Flows on Fixed Grid

To improve the overall engine performance, it is necessary to clearly understand the main unsteady phenomena that occur inside an IC engine. Since experimental technique can provide only lump parameters, the CFD numerical approach has been identified as a valid alternative tool to perform detailed investigations on the fluid dynamics behaviours. The numerical analysis of engine flows is commonly performed by using RANS approach. Adopting a RANS methodology only the mean flow variable distributions could be obtained because the time average of the generic flow variable fluctuation is zero by definition. To perform an effective analysis about the unsteady characteristic of a generic flow and, in particular, of an engine flow it is necessary to improve the numerical solution level adopting the LES (Large Eddy Simulation) approach. LES solves directly the large scales of motion (responsible for the main energy transport inside the flow) while only the small scales are modelled using a Sub-Grid Scale model. Moreover, the LES approach could also be used as test bench case to properly define and understand how it is possible to improve the solution accuracy of RANS simulation. This paper regards the LES analysis of a steady non-reactive wall-bounded flow over a test bench engine geometry. In particular, two LES models, i.e., the Wall Adaptive Local Eddy-Viscosity (WALE) [25] model and the one-equation Dynamic Model by Kim and Menon [23, 24, 29] have been tested. The numerical set-up has been defined performing a preliminary parametric CFD simulations on a basic flow over a backward facing step case. In particular, a bounded second order central differencing scheme was adopted and a discussion of the kinetic energy conservation attitude of such a scheme is performed. LES results have been compared to available experimental LDA measurements of mean and rms fluctuations of both axial and tangential velocity components and with numerical predictions obtained by an optimized RANS simulation of the same case. This paper shows the advantages and the limits of the LES simulation approach applied to IC engine flows.