The Compressible Flow CFD (Computational Fluid Dynamics) group is part of the larger Centre for Hypersonics at the University of Queensland. The focus of the group has traditionally been on the simulation of hypersonic flows in support of the experimental activities carried out at the Centre in their shock tunnels and expansion tubes. Numerical simulation is used during the design of the experimental facilities, for full characterisation of the test-flow in the facilities and in the analysis of experimentally obtained data. In addition to hypersonic flows, the group has diversified its interests in the past few years to include other areas of compressible flow, such as blast wave modelling and interior ballistics computations.
The inclusion of strongly interacting physical effects and large variations in length scales leads to the need of the group to run their simulation codes on large, fast computers. The group presently has access to the University of Queensland's HPC facility, the APAC National Facility and, for long-running jobs, their own Beowulf cluster based on a collection of Intel workstations and a very large SUN cluster.
Simulation of Blast Waves
Joseph Tang is working on developing a reasonable fast 3D CFD code for predicting the effects on complex geometries of blasts generated by an explosion. This is done, for example, by monitoring overpressures or impulses. This work can aid in determining critical damage zones and help to improve structural design.
In order to do this Joseph has created code which relies on an Eulerian formulation of the unsteady Euler equations, with detonation products modelled using the JWL equations of state. A number of commercial and research codes with this capability already exist, however the code used by Joseph employs a variant of cartesian cell methodology known as Virtual Cell Embedding (VCE). In this formulation, an explosion is initiated by specifying a region of high pressure and temperature gas as the initial condition. More advanced modelling is incorporated by adding finite energy release capability, where initiation points can be specified for the detonation waves to consume the whole solid explosive. In addition, the detonation products can be modelled using the more realistic JWL or JWLB equations of state.
Using QCIF supercomputers at UQ and shared-memory machines at the APAC National Facility at ANU, Joseph is hoping to validate his code by carrying out a number of different simulations, including modelling explosions in simple 3D environments, blast wave diffraction over protective barriers, and explosions in more complex 3D environments such as around town buildings. The more intensive simulations are expected to require up to 1200 CPU hours and 12 Gb of memory each.
Figures 1 and 2 show some of the numerical results Joseph has been able to produce so far. Figure 1 shows the blast waves generated around protective barriers. Figure 2 shows the results of a simulation in which a blast is detonated just outside the Mansergh Shaw Building at the University of Queensland.
|
|
|
| Figure 1: 3-dimensional blast wave with barriers. | Figure 2:
3-dimensional simulation of blast detonated within UQ (the centre of the explosion is just outside the Mansergh Shaw Building (Bldg #45)). |
Simulation of the X2 Expansion Tube
Current experiments in the X2 expansion tube at UQ are simulating the atmospheric entry of space vehicles into Titan, the largest moon of Saturn. The numerical simulations of the X2 facility for the Titan studies is complicated by the Titan atmosphere, which consists of approximately 95% nitrogen, 5% methane and trace amounts of argon. These species chemically react to give rise to 20 species, all of which need accounting for in a finite-rate chemistry simulation.
During normal operation, the expansion tube is used to accelerate an impulsive test flow which is applied to a subscale vehicle model. The difficulty with extrapolating from the subscale experiments to a true flight vehicle is that the relative contribution of radiative energy in the flow is not conserved on the subscale vehicle. A second part of the experiments in the X2 expansion tube aim to address this problem of imperfect scaling by operating the expansion tube as a nonreflected shock tube. In this configuration, a travelling shock will directly simulate the detached blunt body shock that exists in front of a true space vehicle. Figure 3 shows a comparison of the two types of experiments undertaken. The experiments will also be able to measure the spectral radiation component and some of the species composition directly behind the travelling shock.
 |
|
Figure 3: Comparison of experiments performed in the X2 expansion tube. |
Computational modelling by Rowan Gollan is aiding the final design of the experiment, and will be an important part of the post-experiment analysis. The computationally intensive work inherent in the problem comes from two sources:
- the finite-rate thermochemical effects in the gas behind the travelling shock, and
- the coupling of fluid dynamics and flowfield radiation in the region near the shock.
Using shared memory machines at the APAC National Facility, Rowan is hoping to complete a grid refinement study (on medium and fine grids) for the complete X2 expansion tube.
Simulation of Compressible Turbulent Flows
Presently, viscous drag represents a significant proportion of the total drag on a hypersonic body. Hence, any reduction in the drag, even if small in comparison to the total drag, will result in large gains in performance of high speed vehicles. One proposed way of achieving a reduction in viscous drag or skin friction is through combustion in the boundary layer. Previous experimental work performed at UQ indicated that drag reductions of 50% were possible. These results
 |
|
Figure 4: Contours of density
gradient showing turbulent structure within a supersonic boundary
layer. |
Andrew Denman is using a turbulence modelling technique called Large-Eddy Simulation (LES) to obtain much more accurate results, which has the added benefit of revealing more flow information. This technique allows the most significant energy carrying component of the turbulence to be resolved. Figure 4 and the linked animation shows the high level of detail captured by this technique. The contours are a grey-scale representing density gradients within the thin boundary layer. This is a common way of visualising turbulence.
Andrew's work will be focused on the simulation of hydrogen combustion within the turbulent domain. The work is particularly computationally intensive due to two factors:
- the high resolution computational mesh required to spatially resolve all turbulent structures; and
- the small time-steps required to capture all turbulent motions.
Andrew is using 3-dimensional code on shared memory machines at the APAC National Facility at ANU using MPI. Typical calculations use between 40 and 60 processes while accessing 2Gb of memory per process. Andrew is hoping to carry out studies at three grid resolutions (coarse, medium and fine), for two experimental conditions.
Contacts
Joseph Tang,
Rowan
Gollan,
Andrew Denman, Dr Peter Jacobs
Compressible Flow Computational Fluid
Dynamics group
Centre for Hypersonics, University
of Queensland
Publications
R. J. Goozée, P. A. Jacobs and D. R. Buttsworth, "Simulation of a complete reflected shock tunnel showing a vortex mechanism for flow contamination", Shock Waves (2006) 15(3-4):165-176.
R. Morgan, T. McIntyre, R. Gollan, P. Jacobs et al., "Radiation measurements in nonreflected shock tunnels.", AIAA Paper
V. Wheatley, H.S. Chiu, P.A. Jacobs et. al. (2004), "Rarefied, superorbital flows in an expansion tube", International Journal for Numerical Methods for Heat and Fluid Flow (2004) 14(4):512-537.
Written by R. Gollan, A. Denman, J. Tang and T. Curtis, June/October 2006