(522b) A Comparison of One Dimensional Turbulence (ODT) and Direct Numerical Simulation (DNS) of Non-Premixed Flames with Extinction
AIChE Annual Meeting
Wednesday, November 10, 2010 - 3:40pm to 4:05pm
The modeling and simulation of turbulent reacting flows with strong finite rate chemical kinetic effects is a significant challenge. Flows of this type are common in combustion involving pollutant formation (e.g., NOx, and soot), and flame extinction and reignition phenomena. This complexity arises from the wide range of turbulent length and timescales, which cannot be fully resolved in high Reynolds number simulations, along with the highly nonlinear nature of chemical reaction rates, the presence of complex flame and flow structures, and multicomponent transport effects. In order to develop more advanced and capable combustion models, two important and related types of information are required: (1) fundamental information of physcal processes of transport-chemistry interactions in turbulent flows; and (2), data for model validation. Direct numerical simulation (DNS) is an unparalleled tool for probing the details of reacting flows, and is continually improving in its physical realism and range of accessible flows as computational resources expand. Modern DNS, however, require millions of CPU-hours for detailed combustion simulations at low-to-moderate Reynolds number. These simulations are typically computed in isolation, or perhaps including a few parametric extensions to a baseline configuration. This high cost precludes detailed investigation of parameter sensitivity, or problem scoping. The one-dimensional turbulence (ODT) model is a simulation approach in which high computational costs are mitigated through solution in only a single dimension, while turbulent advection is modeled stochastically by spatial maps whose size, location, and frequency are implemented using turbulent scaling laws based on the local instantaneous velocity field. These maps increase scalar gradients in a manner consistent with correponding turbulent advection. Computations are resolved in physical space with reactive-diffusive evolution of scalar fields. Hence, the ODT model may provide useful information for model validation and parametric investigation of turbulent flows at a cost orders of magnitude lower than that of DNS. While it is not implied that ODT is a replacement for DNS, there is a relatively large gap between DNS with full resolution and large eddy simulation (LES), which requires modeling of all fine, unresolved flow and chemical structures. In this sense, ODT may play an important intermediate role, by resolving structures in one dimension, with stochastically-modeled turbulent motions the compromise for cost efficiency. A comparison is made of the ODT model with a series of three parametric DNS simulations of turbulent, nonpremixed, planar ethylene jet flames. These three simulations occur at a constant Reynolds number, but varying Damkohler number (ratio of characteristic reaction rate to mixing rate), resulting in varying levels of flame extinction. Each simulation exhibits a high degree of turbulence-flame-chemistry interaction. The DNS simulations are presented, along with the ODT model. Comparisons between the models includes the spatial evolution of the jets, statistical spatial properties (mean and rms) of velocity, temperature, and composition profiles, and statistics of these properties conditioned on the nominal flame-normal coordinate (mixture fraction). Remakable agreement between the ODT and DNS models is observed with 3-4 orders of magnitude difference in computational cost.