Simulation and dynamics of hypersonic turbulent combustion

Abstract

To achieve sustained flight at over five times the speed of sound, advances in propulsion technology like the supersonic combustion ramjet (scramjet) will be required. Numerical simulation of these devices suffers from uncertainty due to the interaction between turbulent flow and chemical reactions, which cannot be computed precisely at full-scale and requires approximate modelling. This project considers the implications of this problem from a number of angles, using high-fidelity numerical modelling and advanced post-processing techniques to investigate the magnitude of this effect, the prospects for approximate modelling of it, and its effects on high velocity combustion in general.A review of the high-speed reacting flow literature has revealed no consensus on the best way to model Turbulent/Chemistry Interaction (TCI). Many simulations make contradictory assumptions in developing their models and some others ignore the effect completely, including some successful studies that recreate experimental measurements to within the estimated margins of error. These margins of error however, can still be quite large; so it is of interest to estimate the effect of TCI on high-speed combustion, to understand when and if the effect can be ignored.In this project, a calculation of the magnitude of hypersonic TCI is performed using Direct Numerical Simulation of a simple flow: A temporally evolving, reacting, supersonic, hydrogen-air mixing layer at elevated temperature. These results are compared with an Improved Delayed Detached Eddy Simulation (IDDES) where TCI errors are present. The simulations take place in a periodic, cubic domain with 2993 (DNS) and 993 (IDDES) cells, in which the turbulence could be sustained for approximately 100 µs. Comparison of the two simulations shows a minor acceleration in the IDDES's ignition process, of around 5 µs in duration, and some additional differences in the subsequent development of the mixing layer. These effects are attributed to the unmodelled turbulence/chemistry interaction that is the main difference between the two simulations. Extrapolating this result to a full size engine suggests the unmodelled TCI would cause an error in ignition location of perhaps several centimeters. Though this is a minor effect, the difference may be amplified as the combustion proceeds downstream, as it would in a more realistic, engine-like flow.To explore the effects of turbulence/chemistry interaction in an engine-like environment, this project also includes numerical simulations of a simplified model scramjet. The domain was a planar, double-ramp, inlet-injected, hydrogen fuelled scramjet at Mach 7, based on a recent experimental model. The simulation used IDDES, with detailed chemical kinetics (33 reactions and 13 species), enforced thermal equilibrium, and no explicit model for turbulence/chemistry interaction (quasi-laminar chemistry). In spite of the simplicity of the domain shape, the combustion in the model scramjet simulations displayed a number of complex features. These included multiple ignition sources, shock-vortex interactions at the entrance to the combustor, and a mixture of premixed and nonpremixed flow throughout. Analysis of the combustion regimes occurring in the flow predicted high turbulent Damkohler and turbulent Reynolds numbers, around 104 - 105. This places the combustion in a difficult to model section of the parameter space where the flame structure is strongly affected by the small turbulence scales, in addition to major distortion of the overall reaction sheet by the large scales. This is an area of (relatively) intense turbulent chemistry interaction, though its exact magnitude cannot be estimated from the analysis itself. Interestingly, the 1% most reactive points, those that account for 70% of the total heat release, defy this trend. They cluster near the laminar boundary in the parameter space, with much smaller values of both turbulent parameters. This indicates weak turbulence chemistry interaction among this subset of points, a finding which may explain for the success of many prior studies that have ignored TCI effects in their modelling. Different trends in the regimes were also obtained for the premixed and nonpremixed points, with the nonpremixed combustion tending to have higher turbulent Damkohler numbers. The physical interpretation of this kind of nonpremixed combustion is unclear however, and more simulation work is needed to illuminate the subgrid structure of such a flow. Application of the same regime analysis technique to the Mixing Layer flow produced consistent results between the DNS and IDDES, indicating that the method may be applied to quasi-laminar simulations.Based on these results, the most promising candidate models for hypersonic TCI are those that can handle mixtures of premixed and nonpremixed flow, and have internal structure that accounts for the thickening of reaction zones by the small turbulent structures. Some existing models may be able to handle these structures, but it seems likely that much additional development will be needed to achieve reliable predictive simulation tools for hypersonic turbulent combustion.