Comparing Two Opacity Models in Monte Carlo Radiative Heat Transfer: Computational Efficiency and Parallel Load Balancing

Abstract

Thermal heating from radiative heat transfer can have a significant effect on combustion systems. A variety of models have been developed to represent the strongly varying opacities found in combustion gases (Goutiere et al., 2000). This work evaluates the computational efficiency and load balance issues associated with two opacity models implemented in a 3D parallel Monte Carlo solver: the spectral-line-based weighted sum of gray gases (SLW) (Denison and Webb, 1993) and the spectral line-by-line (LBL) (Wang and Modest, 2007) opacity models. The parallel performance of the opacity models is evaluated using the Su and Olson (1999) frequency-dependent semi-analytic benchmark problem. Weak scaling, strong scaling, and history scaling studies were performed and comparisons were made for each opacity model. Comparisons of load balance sensitivities to these types of scaling were also evaluated. It was found that the SLW model has some attributes that might be valuable in a select set of parallel problems.