Abstract

In this work, a discontinuous Galerkin scheme is employed to perform large-eddy simulations of supersonic jet flows. A total of four simulations are performed with different meshes and order of accuracy to identify the resolution requirements to reproduce the physical characteristics from experiments. The number of degrees of freedom from the simulations varies from 50 × 106to 400 × 106. The results indicate that by increasing the resolution of simulation, in general, the results got closer to experimental data. The comparison of velocity distribution in the jet centerline and lipline from the simulation with 400 × 106with experimental shows that important characteristics of the flow are represented. The study investigated a procedure of using lower-order simulations to initialize high-order simulations to reduce the total computational cost of the calculation. This strategy is successful and allows the performance of high-order simulations with only 6% more computational effort than a second-order simulation with the same number of degrees of freedom.