High fidelity large eddy simulation of reacting supercritical fuel jet-in-cross-flow using GPU acceleration

When liquid fuel is injected into a combustion chamber, the dispersion of the fuel and the ensuing combustion are characterized by the ambient pressure. At elevated pressures, unlike the classical liquid atomization observed at sub-critical pressures, the fuel is dispersed by diffusion-dominated mixing. Numerical simulation of fuel-air mixing and the resulting reactive flow is dependent on accurate modeling of the supercritical flow and is very challenging. In most cases, the combination of the complex thermodynamics and the computational grid resolution requirements render the numerical investigations computationally intractable. However, with the advent of novel heterogeneous computer architectures, it is possible to overcome the computational resource constraints and investigate mixing and combustion of supercritical fuels. In this paper, we study a reacting n-decane Jet-In-Cross-Flow (JICF) of air at high ambient pressure. The flow properties are computed using routines accelerated for Graphics Processing Unit (GPU) computation. The code assisted by GPU computation is nearly 3 times faster due to the reduction in the cost associated with the routines for thermodynamics and subgrid closures. Followed by the dispersion of the supercritical fuel due to the interaction with the turbulent cross flow, the combustion ensues in the leeward side of the jet. The flame is anchored in the aft and spreads downstream following the mixing zone. Due to the heat release, in comparison to the non-reactive case, the downstream turbulence is reduced. The temperature in the mixing zone is correlated to the mixture fraction of the fuel and results in formation of hot spots that convect with the vortices generated by fuel-air mixing.