A large-eddy simulation study on supersonic jets

• Main Author: Toh, Joel Hang Hwa
• Other Authors: New Tze How Daniel
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/69250
• Institution: Nanyang Technological University

A Large-Eddy Simulation (LES) investigation was performed with the aim of analysing flow behaviour in a supersonic jet nozzle. The experiment also sought to understand how the results from the software used in the project, ANSYS Fluent, fair in its ability to produce accurate and similar results performed by previous experiments, done both numerically and experimentally. The initial phase of the project involved a two-dimensional (2D) mesh independence study, comparing the results of supersonic flow from a finer mesh and a coarser mesh with respect to a control mesh. The simulation was set to solve for a steady state solution with the Reynolds-Averaged Navier-Stokes (RANS) model. There were minor differences between the results of the three cases and results presented by the finer mesh have been proven to be more detailed and precise. In the second phase of the project, a 2D transient state simulation was conducted to obtain an approximate visualisation of how the supersonic flow of the nozzle was be expected to be. The simulation was also conducted using the RANS model. A comparison was made with experimental results from the Nanyang Technological University (NTU) research group to validate the flow. In the last phase of the project, a three-dimensional (3D) mesh of the supersonic nozzle was made to solve for a transient state solution. A 3D wedge of the nozzle was constructed to test if the parameters and resolution of the cells of the mesh was workable. The use of a wedge also saves on computational costs and time as opposed to using a full 3D mesh. After which, the full 3D mesh was made for LES. Though there were some complications met during the simulation, the results agrees with well with previous works and results from the NTU research group. In summary, LES from Fluent was effective in producing shock structures brought about by an under-expanded nozzle. Results also show similarities in flow characteristics from 2D RANS results. However, LES demands a high amount of computation, a very fine mesh, and time steps required a very small. A lot of time has to be accounted for in order to obtain a steady state flow. It seems to be more cost efficient to use data collected from the RANS model since differences in data quantities do not differ to a large extent.