Ensemble-averaged dynamics of premixed, turbulent, harmonically excited flames
MetadataShow full item record
Increasing awareness of the negative impacts of pollutant emissions associated with combustion is driving increasingly stringent regulatory limits. In particular, oxides of nitrogen, generally referred to as NOx, now face strict limits. These restrictions have driven development of cleaner burning combustion systems. Because NOx formation increases significantly at elevated temperatures, one method to reduce NOx emissions is to burn the fuel at lower temperatures. By premixing the fuel and oxidizer prior to combustion significantly lower flame temperatures can be achieved, with corresponding reductions in NOx emissions. Unfortunately, premixed combustion systems are generally more prone to potentially problematic feedback between the unsteady heat release from the flame and unsteady pressure oscillations. This self-excited feedback loop is known as combustion instability. Because these oscillations are associated with unsteady pressure fluctuations they can degrade system performance, limit operability, and even lead to catastrophic failure. Understanding combustion instability is the primary motivation for the work presented in this thesis. The interaction of quasi-coherent and turbulent flame disturbances changes the spatio-temporal flame dynamics and turbulent flame speed, yet this interaction is not fully understood. Therefore, this thesis concentrates on identifying, understanding, and modeling these interactions. In order to address this topic, two primary avenues of research are followed: development and validation of a flame position model and experimental investigations of predicted ensemble-averaged flame speed sensitivity to flame curvature. First, a reduced order modeling approach for turbulent premixed flames is presented, based on the ensemble-averaged flame governing equation proposed by Shin and Lieuwen (2013). The turbulent modeling method is based on the G-equation approach used in laminar flame position and heat release studies. In order to capture the dependence of the ensemble-averaged turbulent flame speed on the ensemble-averaged flame curvature, the turbulent flame model incorporates a flame speed closure proposed by Shin and Lieuwen (2013). Application of the G-equation approach in different coordinate systems requires the inclusion of time-varying integration limits when calculating global flame area. This issue is discussed and the necessary corrections derived. Next, the reduced order turbulent modeling approach is validated by comparison with three-dimensional simulations of premixed flames, for both flame position and heat release response. The reduced order model is the linearized, allowing development of fully analytical flame position and heat release expressions. The use of the flame speed closure is shown to capture nonlinear effects associated with kinematic restoration. Second, the development of and results from a novel experimental facility are described. This facility has the capability to subject premixed flames to simultaneous broadband turbulent fluctuations and narrowband coherent fluctuations, which are introduced on the flame using an oscillating flame holder. Mie scattering images are used to identify the instantaneous flame edge position, while simultaneous high speed PIV measurements provide flow field information. Results from this experimental investigation include analysis of the ensemble-averaged flame dynamics, the ensemble-averaged turbulent displacement speed, the local ensemble-averaged area and consumption speed, and the dependence of both the displacement speed and consumption speed on the ensemble-averaged flame curvature. Finally, the flame speed sensitivity to curvature is quantified through calculation of the normalized turbulent Markstein displacement and consumption numbers. The results show that the amplitude of coherent flame wrinkles generally decreases with both downstream distance and increasing turbulence intensity, providing the first experimental validation of previous isothermal results. The average displacement and consumption speeds increase with downstream distance and turbulence intensity, reflecting the increasing wrinkled flame surface. The ensemble-averaged, phase dependent displacement and consumption speeds demonstrate clear modulation with the shape of the ensemble-averaged flame. Specifically, these turbulent flame speeds increase in regions of negative curvature. For both the displacement and consumption speed, the magnitude of the normalized turbulent Markstein length increases with ratio of the turbulent flame wrinkling length to the coherent wrinkling length when u'/SL0 >2.5 . For u'/SL0 < 2.5 the trends are less clear due to the presence of convecting disturbances which introduce additional fine scale wrinkles on the flame. Together the results presented in this thesis provide a foundation for modeling turbulent flames in the presence of quasi-coherent disturbances. The flame position can be modeled using the ensemble-averaged governing equation with the dynamical flame speed closure, and the corresponding heat release can be calculated from the turbulent consumption speed closure. The turbulent Markstein numbers and uncurved flame speed may be extracted from experimental or numerical data.