Large Eddy Simulation of flame stabilisation dynamics and vortex control in a lifted H₂/N₂ jet flame

Abstract

Flame stabilisation in (highly) preheated mixture is common in several industrial applications. When the reactants are injected separately in the device (usually at high-speed), the flame is lifted so that the fuel and oxidant first mix to give an ignitable mixture. If the temperature of the mixture is adequate, it auto-ignites stabilizing the flame. Here we focus on an academic lifted jet flame and Large Eddy Simulation (LES) is used to capture the flame and auto-ignition dynamics. Comparisons with experimental data show that LES simulates accurately high OH fluctuation levels at the stabilisation location. The vortex dynamics linked to these fluctuations is analyzed and it is found that small scale coherent structures play a vital role in the auto-ignition process. These structures are axial vorticity tubes (braids) and are located relatively far (in the radial direction) from the shear-layer. As a consequence, the lift-off height varies dramatically in time leading to OH fluctuations of the order of the mean OH concentration. This scenario is monitored in the compositional space highlighting the simultaneous evolution of OH, HO ₂ and temperature. Further, different strategies for open-loop control of the flame lift-off height are tested. In order to anchor the flame at different positions downstream of the nozzle, the vortex dynamics in the shear-layer was modified. Promoting successively vortex ring and braids, the auto-ignition region was moved significantly. In particular, modified nozzle geometries impacted the formation of braids and ensured a good premixing very close to the nozzle. As a consequence, it was possible to reduce significantly the lift-off height and stabilise the flame few diameters downstream of the nozzle.

Keywords:

• turbulent flame
• lifted flame
• auto-ignition
• Large Eddy Simulation
• vortex dynamics
• flame and vortex control

Acknowledgments

This work was partly supported by the Swedish Research Council (VR). The computations were performed using the super-computing facilities at Lund University (LUNARC). The routine used to generate the time dependent inflow boundary condition was kindly provided by Markus Klein.