Comparison of Linear, Non-Linear and Generalized RNG-Based k-epsilon Models for Turbulent Diesel Engine Flows 2017-01-0561
In this work, linear, non-linear and a generalized renormalization group (RNG) two-equation RANS turbulence models of the k-epsilon form were compared for the prediction of turbulent compressible flows in diesel engines. The object-oriented, multidimensional parallel code FRESCO, developed at the University of Wisconsin, was used to test the alternative models versus the standard k-epsilon model.
Test cases featured the academic backward facing step and the impinging gas jet in a quiescent chamber. Diesel engine flows featured high-pressure spray injection in a constant volume vessel from the Engine Combustion Network (ECN), as well as intake flows in a high-swirl diesel engine. For the engine intake flows, a model of the Sandia National Laboratories 1.9L light-duty single cylinder optical engine was used. An extensive experimental campaign provided validation data in terms of ensemble averages of planar PIV measurements at different vertical locations in the combustion chamber, for different swirl ratio configurations during both the intake and the compression strokes.
The generalized RNG k-epsilon model provided the best accuracy trade-off for both swirl and shear flows, thanks to the polynomial expansion of coefficients C1 and C2 in the RNG k-epsilon model with an effective ‘dimensionality’ of the strain rate field. Similar performance was seen across linear and non-linear RNG models, which achieves good prediction of in-cylinder swirl flows; however, they noticeably underpredict jet penetration in the case of high-pressure sprays, suggesting the additional computational cost and lower stability of the non-linear model do not justify greater suitability for engine calculations.
Citation: Perini, F., Zha, K., Busch, S., and Reitz, R., "Comparison of Linear, Non-Linear and Generalized RNG-Based k-epsilon Models for Turbulent Diesel Engine Flows," SAE Technical Paper 2017-01-0561, 2017, https://doi.org/10.4271/2017-01-0561. Download Citation
Federico Perini, Kan Zha, Stephen Busch, Rolf Reitz
University of Wisconsin-Madison, Sandia National Laboratories