Comparison and Standardization of Numerical Approaches for the Prediction of Non-reacting and Reacting Diesel Sprays 2012-01-1263

The primary objective of the research discussed here was to compare the commercial computational fluid dynamics (CFD) software, CONVERGE, and a prevalent open-source code, OpenFOAM, with regard to their ability to predict spray and combustion characteristics. The high-fidelity data were obtained from the Engine Combustion Network (ECN) at Sandia National Laboratory in a constant-volume combustion vessel under well-defined, controlled conditions. The experiments and simulations were performed by using two diesel surrogate fuels (i.e., n-heptane and n-dodecane) under both non-reacting and reacting conditions. Specifically, ECN data on spray penetration, liquid length, vapor penetration, mixture fraction, ignition delay, and flame lift-off length (LOL) were used to validate both codes. Results indicate that both codes can predict the above experimental characteristics very well. The simulations incorporated similar approaches for spray and combustion modeling, with some differences in the choice of sub-models and their associated constants. For example, for both codes, the smallest preferred grid size chosen was 0.25 mm and adaptive mesh resolution was used to capture regions characterized by high densities, temperatures, species gradients. In addition, the chemical-kinetic models for n-heptane oxidation were also the same for both the codes. After extensive validation of the modeling tools, parametric studies were performed to characterize the effect of the grid size, time-step size, and turbulence models on the above-mentioned parameters. For the minimum grid sizes chosen, both codes were grid dependent. CONVERGE was time-step independent, but it was more dependent than OpenFOAM was on mesh size for vapor penetration if the larger grid was used. The turbulence model had a greater impact on vapor penetration than on liquid penetration. Under reacting conditions, some differences in ignition characteristics, flame LOL, equivalence ratio, temperature, and OH mass fraction distributions were observed between the codes. In general, there was good agreement between predictions by both CONVERGE and OpenFOAM against experimental data, although flame LOL was better predicted by CONVERGE. Finally, throughout this study the two approaches used different definitions for several terms (liquid and vapor penetration, ignition delay, lift-off), which provides a strong motivation for standardizing the parameters further by enveloping the experimental definitions for spray and combustion parameters.