An extensive numerical investigation is conducted in order to assess the effect of Rayleigh-Taylor instability on fuel consumption rate (or flame speed). Two geometries are used for this investigation, viz., a high pressure high-g (HPHG) cavity stabilized combustor and a curved duct with a backward facing step. The former geometry is a more practical combustion system that contains liquid fuel injectors with operating conditions that mimic gas turbine cycles, whereas the latter is a canonical combustor used to study turbulent premixed flames. Reynolds averaged Navier-Stokes (RANS) and large eddy simulations (LES) are used. RANS is used for the practical combustor, and both RANS and LES are used for the canonical combustor. The combustion models used are the flamelet generated manifold (FGM) and the two-step species transport for the practical and canonical combustor, respectively.
The HPHG combustor is designed to induce bulk rotational flow in the cavity, inducing centrifugal acceleration. The centrifugal force acts from the high-density reactants towards the low-density products creating a Rayleigh-Taylor instability (RTI). Rayleigh-Taylor instabilities are expected to increase the turbulent flame speed and reduce the size of the combustor by increasing the flame wrinkling and/or corrugation. Simulations at two different levels of centrifugal acceleration, and, consequently, dissimilar Rayleigh-Taylor instability were performed. It was found that the nominal g-loads are overestimating the local g-loads from the simulation because thermal expansion is not taken into account. From these simulations it was not possible to discern the effect of RTI on fuel consumption rate due to the complex physical-chemical process inherent to this combustor such as fuel vaporization, molecular mixing, spray-turbulence interaction, turbulence-chemistry interaction, and partial premixing.
Therefore, gaseous premixed turbulent flames were simulated in a curved duct with a backward facing step. Two radius of curvature were used, viz., an infinite (straight duct) and a finite radius of curvature (curved duct). These combustors were operated at low and high Reynolds number (3,200 and 32,000). The computational results are compared with broadband chemiluminescence and shadowgraph images reported in the literature for similar conditions and geometries. Both RANS and LES results are in general agreement with measurements. Both experiments and simulations show that increasing the Reynolds number in both straight and curved canonical combustor the flame cannot withstand the Karlovitz number effects and the flame is positioned behind the backward-facing step. In addition, the LES results indicate that at high Reynolds number the flame blows out for the straight channel while it remains stabilized for the curved channel. This result is in agreement with the blowout data reported in the literature. On the other hand, RANS over predict the flame stabilization for the straight channel. Consequently, RANS should not be used in research involving RTI-induced blowout.
In conclusion, RTI interacts with a turbulent premixed flame and its overall effect is to extend the conditions under which turbulent premixed flames can be stabilized. This improved flame stabilization is a direct manifestation that the fuel consumption rate (or flame speed) has been enhanced in order for the flame to withstand higher Karlovitz number effects induced by high Reynolds number. However, the mechanism through which RTI works on the turbulent premixed flame is not clear. A new hypothesis is proposed. The increase in RTI should increase the turbulent length scale as well as increase the Karlovitz number. The corrugated flame would withstand the higher Karlovitz number because RTI temporarily and periodically reverses the turbulent energy cascade by minimizing the potential energy of the stratified flow.