Skip to main content
Energy Systems Division

High-Fidelity Large Eddy Simulations (LES)

Demonstrating that demonstrated that large eddy simulations can qualitatively and quantitatively predict equivalence ratio distributions, ignition delay, soot distribution and more

Large eddy simulations (LES)-based turbulence models are being used more frequently for engine simulations. The main questions that need to be addressed are:

  • Can LES provide further insights on spray and combustion processes?
  • Does LES improve qualitative and/or quantitative predictions?

Argonne engineers, in collaboration with Convergent Science Inc., have demonstrated that LES can indeed qualitatively and quantitatively better predict equivalence ratio distributions, ignition delay, soot distribution and more under the conditions investigated.

Research was conducted using a standard RANS-based approach against a high-fidelity LES-based approach under non-reacting and reacting conditions against both quantitative and qualitative data from Sandia National Laboratories. The standard modeling approach consists of using a coarser minimum grid size (0.5mm) with RANS based models (~0.3 million cells), whereas the high-fidelity approach consists of using a finer mesh (such as 0.125mm) with LES models (~2.0 million cells) . Smaller grid sizes were necessary with the LES model for two reasons: (1) since a zero-equation Smagorinsky model was used, it was desirable that the sub-grid scale modeling be reduced, and (2) the possibility of accurately capturing the large-scale flow structures was higher with a finer grid. The comparisons under non-reacting conditions are presented first.

The instantaneous experimental images obtained using Rayleigh scattering imaging are shown on the left, along with the time ASI and the axial length scale (Figure 1). The field of view is 40 mm x 20 mm in the axial and transverse directions respectively. Note that the experimental contours pertain to a ratio between fuel-air number densities (Nf/Na) whereas; simulations plot the fuel mass-fractions. Fuel vapor penetration and dispersion can be clearly seen from the experimental and simulation plots. Both RANS and LES simulations predict the dispersion and vapor penetration fairly well. However, marked differences in the spray structure are clearly observed between RANS and LES cases.  While RANS predicts smooth, averaged profiles, the LES simulation is able to capture the instantaneous structure well. The initiation of instabilities on the surface seems to be occurring further downstream in the case of LES compared to the experiments. Spray dispersion seems to be marginally underpredicted by the LES model as well.

Quantitative comparisons of predicted ignition delays at different ambient temperatures by RANS and LES turbulence models are presented in Figure 2. The data for NHPT were obtained at Sandia National Laboratories. An NHPT mechanism from the University of Connecticut was used for predicting ignition delays for both turbulence models. It is clear that the LES model predicts lower ignition delay values. Also, under the conditions investigated, it seems that the LES model performs marginally better than the RANS model in predicting ignition delays at different ambient temperatures.

The simulations ran for 8-10 days with the high-fidelity LES approach on 24 cores (3 node) with the enhanced spray models and detailed combustion chemistry for NHPT as surrogate. The RANS simulation with coarser meshes ran for about 80 hours on 8 cores (1 node).

All the simulations were performed at the fusion cluster at Argonne National Laboratory. The authors gratefully acknowledge a grant of computing resources at the Laboratory Computing Resource Center at Argonne National Laboratory.


This work is supported by the U.S. Department of Energy’s Vehicle Technologies Program under Gurpreet Singh. Convergent Science, Inc. was a collaborator.


  • S. Som, P.K. Senecal, E. Pomraning, Comparison of RANS and LES Turbulence Models against Constant Volume Diesel Experiments,” 24th Annual Conference on Liquid atomization and spray systems, San Antonio, TX, May, 2012.
  • S. Som, D.E. Longman, Z. Luo, M. Plomer, T. Lu, P.K. Senecal, E. Pomraning, Simulating flame lift-off characteristics of diesel and biodiesel fuels using detailed chemical-kinetic mechanisms and LES turbulence model,” ICEF2011-60051, ASME Internal Combustion Engine Division Fall Technical Conference, Morgantown, October 2011.