A 3D-CFD methodology to investigate boundary layers and assess the applicability of wall functions in actual industrial problems: A focus on in-cylinder simulations

In the industrial practice, 3D-CFD in-cylinder simulations still largely rely on RANS turbulence models and high-Reynolds wall treatments, i.e. based on wall functions. However, the use of the latter represents a potential source of error, leading to poor estimations of shear stress and heat flux at the wall. In fact, universal laws of the wall can be claimed only under very restricted conditions, which are hardly (to say never) met in industrial applications. As a result, typical dimensionless profiles of velocity and temperature on the combustion chamber walls are far from standard wall functions. In the present paper, a methodology to investigate the presence of dimensionless profiles comparable to universal wall laws in boundary layers of actual industrial problems is presented. In particular, attention is focused on 3D-CFD in-cylinder simulations. While the existing literature deals with DNS or hybrid URANS/LES approaches applied to simplified geometries and low revving speed conditions (for computational cost reasons), in the present paper a RANS k-ε turbulence model with a low-Reynolds wall treatment is adopted. In addition, an alternative strategy to extract velocity and temperature dimensionless profiles from the computed fields is proposed. The methodology is preliminary tested on a 2D plane channel (where the existence of wall functions is a priori acknowledged), at both quasi-isothermal and highly non-isothermal conditions. Afterwards, it is applied to the well-known “GM Pancake” engine test case, showing that both u+ and T+ calculated on the combustion chamber walls remarkably differ from analytical standard wall functions. Finally, in order to demonstrate the importance of dimensionless profiles to properly predict heat transfer, two different high-Reynolds simulations of the “GM Pancake” engine are proposed, one with standard wall functions and one with u+ and T+ profiles provided by the low-Reynolds analysis. While the former underestimates heat fluxes, the latter provides results in good agreement with the experiments