Direct Numerical Simulation Of Coupled Multi-physics In Supersonic Turbulent Combustion

Supersonic turbulent combustion is a complex phenomenon including multiphysical problems such as turbulence, shock wave, combustion wave and their interactions. In the present thesis, based on the theory of high fidelity direct numerical simulation (DNS), a massively parallel computing platform has been developed for DNS of supersonic turbulent gaseous and spray combustion. In order to reveal the underlying coupling mechanism among turbulence-shock-flame-evaporation in supersonic turbulent combustion, problems concerning shock-isotropic turbulence interaction (SITI), detonation-isotropic turbulence interaction (DITI), supersonic gaseous and spray jet flame have been analyzed.DNS of shock-isotropic turbulence interaction has been first conducted to investigate the interactions between turbulence and shock wave. It has been found that velocity and vortex variance both have been amplified after compressed by shock. The isotropic turbulence first transits to anisotropic and then recovers toward isotropic again downstream of the shock. Turbulent eddies have been compressed and length scales decreased. The shock wave becomes corrugated due to turbulent eddies. And this has been found to be quite related to the upstream velocity fluctuations. Effects of different inflow turbulent intensity on the dynamic structure of shock wave have been analyzed. It has been found that the amplification of Reynolds stress increase as the turbulent intensity increase. The reduce amount of turbulent length scale decrease. And the recovering toward isotropic becomes faster. As the increase of the inflow turbulent intensity, shock wave becomes more corrugated, and even broken down to a "hole" structure.DNS of detonation-isotropic turbulence interaction has then conducted to analyze the heat release effects on shock-turbulence interactions. Comparable studies of different inflow turbulent fluctuations interacting with a Mach3detonation have been conducted. The inflow conditions including no turbulent fluctuation, vorticity fluctuation, entropy fluctuations. The canonical triple point structure has been found in detonation. The triple point is the intersection point of Mach stem, Incident shock, Transverse wave. The historical trajectory of the triple point has been recorder as the cellular structure of detonation, shown as the smoke-foil inscription. The triple structure has been changed due to the inflow forcing. However, the period of the collision of triple points has been delayed and the length of the cellular structure increase a little. The collisions of the triple points produce high-pressurized regions, resulting the formation of large vertical structure. A pair of clockwise and counterclockwise circulation inside the central jet, extracts the unburnt materials into its rolling zone. Compared with the shock-turbulence interaction without heat release, the amplification of Reynolds stress increases a lot. No significant changes of the cellular structure can be found for different inflow vertical fluctuations. The turbulent fluctuations downstream of detonation mainly induced by the large scale generated vortex by collision of triple points. When Mach number equals4, detonation becomes unstable. The cellular structure becomes irregular. And the effects of the inflow forcing present to be more complicated.Turbulent combustion models such as flamelet model, conditional moment closure (CMC) have been investigated with the DNS data of a spatially developing three dimension supersonic gaseous jet flame. The probability density functions of mean conditional and unconditional scalar dissipation rate prove to qualitatively agree with the presumed log-normal distribution, while a little skewed to the higher scalar dissipation rate in the sonic mixing layer. The conditional mean scalar dissipation rate presents to be radial dependent at the flame base, especially in the fuel lean mixture. The DNS results show good agreement with the trends of the flamelet calculations; however, the amplitudes of temperature are far lower than the corresponding flamelet statistics due to finite rate reaction and expansion of the high speed reacting flow. Submodels for the unclosed terms in the CMC approach are assessed with DNS results. Beta pdf of mixture fraction can well capture the mixing space of the high speed reacting flow. The linear model exhibits a good performance for the axial velocity predictions. Girimaji’s model for scalar dissipation rate preforms well at upstream, while the AMC model presents better further downstream. The first order closure for the conditional reaction rate deviates a lot from the DNS extracted results. Second-order corrections made to temperature only or to the two rate-limiting reaction steps induce improvement, still with much discrepancy. Second order closure considering fluctuations of all the reacting species and temperature can accurately reproduce the DNS results.In order to reveal the underlying interacting mechanism among turbulence-flame-evaporation, DNS of supersonic spray jet flame has been conducted, including the nonreacting case NR, reacting case Rl and R2with double fuel injection. Compared with the gaseous jet flame, the spray flame surface contains many local low temperature regions due to evaporation. It can be found many belt structure and some distributed burning pockets in the spray flame. Premixed and diffusion flame coexist in the spray flame. The heat release contribution from premixed combustion increase downstream to more than50%. Comparative study of the non-reacting and reacting case indicates that combustion amplifies the evaporation rate of droplet. The velocity of droplet has been found to be increased compared with the droplet at the same position in non-reacting flow. Combustion also affects the turbulent intensity of the flow field. Reynolds stress has been found to be larger in the mixing layer with intense heat release, and smaller in the jet central region. In the mixture fraction space, the fluctuation of temperature, species concentration has been found to be amplified in the reacting flow. The scalar dissipation rate also increases. The DNS data is also adopted to the priori study of CMC models. The probability distribution function of mixture fraction has been found to be generally consistent with βfunction. The conditional scalar dissipation rate presents to be bi-mode and cannot be well modeled by the AMC model. The conditional evaporation rate can be modeled by the linear model. Conditional temperature and species mass fraction on mixture fraction fluctuate to a larger extent, and higher order closures are needed to be adopted for the reaction source closure. Or the double conditional moment closure model can be adopted. It has been found the doubly conditioned reaction source termed can be modeled by the first order scheme.