Swirler effects on combustion instabilities analyzed with measured FDFs, injector impedances and damping rates

Abstract

The influence of the injection system on combustion instabilities is investigated on a laboratory-scale combustor equipped with a single injector that is weakly-transparent to acoustic waves. The combustor is fed with liquid heptane delivered as a spray by a hollow cone atomizer. Experiments are carried out with three swirlers having similar geometries but different pressure losses and swirl numbers. Self-sustained oscillations (SSOs) corresponding to these swirlers feature differences in oscillation frequency and amplitude for a given chamber length. These observations do not match with standard modeling predictions. Therefore, a low-order analytical model is derived, representing the effect of the acoustically weakly-transparent injection system using an impedance at the injector outlet. This quantity and the flame describing function (FDF), both determined experimentally, are used together with damping rate estimates as model inputs. It is shown that the FDF can only be determined by suitably selecting the position for the measurement of incident velocity disturbances at the injector outlet and that plenum-based velocity measurements cannot be used for this purpose. It is also assumed that the OH*-chemiluminescence intensity can be used as a proxy for the heat release rate. This admittedly strong assumption for spray flames is discussed in detail and justified by showing that the equivalence ratio modulations are relatively weak for the particular spray flames considered in this study. Results from the model indicate that the injector impedance (that depends on the swirler characteristics) shifts the classical bands of instability and modifies the growth rate magnitude compared to a generic combustor with an acoustically transparent injector. Using the proposed model, the stability of the system can be rated along with a prediction for growth rate and frequency of oscillation. Predictions generally agree with experimental observations with some limitations. The model combined with damping rate estimates is finally used to predict limit cycle oscillation amplitudes with the aid of the FDF framework.