Adaptive mesh refinement based simulations of three-dimensional detonation combustion in supersonic combustible mixtures with a detailed reaction model

Abstract

Detonation combustion initiated with a hot jet in supersonic H2-O2-Ar mixtures is investigated by large-scale three-dimensional (3D) simulations conducted in Tianhe-2 computing system with adaptive mesh refinement method. The reactive Euler equations are utilized as governing equations with a detailed reaction model where the molar ratio of the combustible mixture is 2:1:7 under the condition of pressure 10 kPa and temperature 298 K. Results show that the Mach stem surface which is formed after the shock surface reflection on the upper wall is actually a local overdriven detonation. The side walls in 3D simulations can play an important role in detonation initiation in supersonic combustible mixtures, because they can help realize triple lines collisions and reflections during the initiation process. The width of the channel has an important influence on the strength of side-wall reflections, and under certain condition there might exist a critical width between the front and back sides of the channel for the successful initiation. Both the two-dimensional (2D) and 3D detonations are overdriven and have constant but different overdrive degrees after their complete initiations. Although the overdrive degree of the 3D detonation is smaller than that of the 2D case, more complex and irregular detonation fronts can be observed in the 3D case compared with the 2D detonation, which is likely because of the propagation of transverse waves and collisions of triple lines in multi-directions in 3D detonations. When the hot jet is shut down, the newly formed 2D Chapman-Jouguet (CJ) detonation has almost the same characteristic parameters with the corresponding 3D case, indicating that the 2D instabilities can be perfectly preserved in 3D simulations. However, the slapping wave reflections on the side walls in the 3D detonation result in the second oscillation along with the main one, which presents stronger instabilities compared with the 2D case. The inherent stronger 3D instabilities are also verified through the quantitative comparison between the 2D and 3D cases where the 3D result always shows stronger fluctuations than the 2D case.