Direct Numerical Simulation of Two-Phase Turbulent Reactive Flows

Direct numerical simulation (DNS) provides a viable means for simulating turbulent flows without using turbulence models. This is achieved by solving the governing equations on a very fine grid and is usually feasible on fast supercomputers only. As a result, these simulations are mainly limited to somewhat simple flows, and are primarily used to enhance the fundamental understanding of the underlying physics. To apply DNS to two-phase flows, in this project a large number of droplets are followed in a Lagrangian frame while the carrier phase is simulated by various spectral methods. With recent advances in computational resources, we are now able to apply DNS to more complex flow geometries using a new multi-domain spectral element method.

Probability Density Function Modeling of Two-Phase Turbulent Flows

In this project, we are taking a probabilistic approach to the problem of dispersion and polydispersity (i.e. size variation) of liquid fuel droplets in a turbulent flow. First, a transport equation for the phase space density of the relevant variables is derived through the application of the Liouville's theorem. The ensemble average of this equation over a large number of realizations then yields a transport equation for the probability density function (PDF) of the droplet properties such as velocity and temperature. The process of ensemble averaging, however, results in the appearance of unclosed terms, manifesting the well-known closure problem in turbulence. We are investigating the application of a variety of closure schemes to this problem, and the final goal is to provide "fluid-like" (continuum) transport equations for various statistics of the dispersed phase. These transport equations take the forms of partial differential equations (PDE's) and can be integrated using standard methods. The predictions of these equations are assessed against direct numerical simulation (DNS) results and laboratory data.

Stochastic Simulations of Liquid Fuel Combustors

The main idea in the application of stochastic methods to two-phase turbulent flows is to introduce a large number of droplets into the flow of interest and to generate a "synthetic" turbulence with the known statistical properties. These known properties can, in general, be obtained by solving any single-point closure schemes such as those developed by Reynolds averaging the Navier-Stokes equations for the carrier fluid. A large number of droplets are then introduced into the flow and their trajectories are simulated while updating various droplet properties, such as velocity and temperature, in time. It is important to realize that, the variables generated through the synthetic turbulence can only represent the physics of the real turbulence after averaging over a large number of droplets (i.e. samples) in a statistical sense. The statistical analysis on the droplets is performed by collecting samples from a small control volume within the flow field. This procedure is similar to that implemented in laboratory experiments where the droplets crossing a small illuminated volume are sampled. This project deals with development and practical application of various stochastic models in a liquid-fuel combustor.

Nanoparticle Dynamics in Dusty Plasmas

This project primarily deals with mathematical modeling and numerical simulation of coagulation and growth of nanoparticles, and the related phenomena such as particle charging, turbulence effects, anomalous heating, and thermophoresis effects, in dusty plasma. The formation and dynamics of these nanoparticles are not only interesting as a challenging interdisciplinary fundamental subject, but also extremely important in electronics, catalysis, and environmental control. For example, generation of even a few nanoparticles during the plasma-assisted chemical vapor deposition makes the electronics product virtually useless, whereas the nanoparticles deposited from the nonequilibrium plasma form surfaces with very high catalytic activity. This project is a multidisciplinary effort that is conducted through collaboration among three different groups with expertise in two-phase turbulence, plasma, and higher order numerical methods. The performance of the developed models will be assessed via comparisons with laboratory data provided through our collaboration with an experimental group.

Inter-phase Transfer in Oscillating Drops

With the increasing interest in analytical treatment of two-phase turbulent flows, there has been associated an increasing demand for analytical/empirical correlations describing the inter-phase transfer of mass, momentum, and energy at the surface of the drop. In nearly all of the previous applications only spherical droplets were considered. However, a large number of two-phase systems involve liquid drops that may undergo significant shape deformations while interacting with the carrier gas. In particular in combustion systems, the droplets that are generated by an atomization process, always undergo some oscillations before relaxing from the initial ligament to a final spherical shape. In this project, a Galerkin finite element method is used in conjunction with an interface-tracking scheme to simulate the simultaneous oscillation and evaporation of the droplets in a combustor environment.

CFD Calculations of the METHANE de-NOX Reburn Process

The "reburn" process describes an economic NOx reduction technology for solid fuel-fired stoker boilers (coal, biomass, municipal solid waste, etc.), which decreases NOx emissions by 50%-70% while reducing other emissions. This technology utilizes the injection of natural gas together with re-circulated flue gases (for enhanced mixing) to create an oxygen-deficient zone above the combustion grate. Overfire air is then injected at a higher furnace elevation to burn out the combustibles. In order to improve the performance of the reburn process, a numerical model has been developed. Under this project, we are implementing this model for extensive simulations of the reburn process, by considering different values for the relevant parameters.