Large-eddy simulations of a jet in crossflow using Julia

Abstract

The jet in crossflow (JICF) is a complex flow that has applications in many fields, from pollutant dispersion into air or water to the injection and mixing of fuel in engines. In this thesis, large-eddy simulations, using a stretched-vortex sub-grid model, of a JICF with a non-reactive scalar are performed using a discrete numerical method that is implemented using code written in the computational language Julia. Velocity profiles, trajectories, entrainment, power spectra, turbulent kinetic energy and dissipation of energy are analysed for simulations run at velocity ratios varying between 0.405 and 3.3, crossflow boundary layer thicknesses between 0.28 and 2.06 and Reynolds numbers between 243 and 20500. Simulations are compared to published experimental and simulation-based results, and a full comparison was performed with a simulation provided by Mattner, run on the same computational grid. It was found that the mathematical model used in this thesis performs better at higher velocities and Reynolds numbers. An investigation into the effect of the ratio of average jet inlet velocity to maximum crossflow velocity was performed. It was found that a jet with a higher velocity ratio showed increased penetration into the crossflow. The amount of turbulent kinetic and scalar energy in the system, as well as the amount of dissipation of energy from the system, also increased with velocity ratio. Finally, a comparison of large-eddy simulation (LES) and direct numerical simulation (DNS) of a JICF was performed on the same computational grid for low and moderate Reynolds numbers. At low Reynolds numbers the di↵erences in results between the LES and DNS are minor, although it is not possible to resolve the flow on the computational grid that is used. At moderate Reynolds numbers, above Re = 1 x 10⁴, the differences between the LES and DNS are more pronounced. Deeper jet penetration is seen in the LES than in the DNS, and the distribution of energy in the system is different, with the sub-grid model used in the LES dissipating more energy from the high wavenumber scales.