Low-Order Modeling of Combustion Noise in an Aero-Engine: The Effect of Entropy Dispersion

摘要

The present work studies the effect of entropy dispersion on the level of combustion noise at the turbine outlet of the Rolls-Royce ANTLE aero-engine. A new model for the decay of entropy waves, based on modeling dispersion effects, is developed and utilized in a low-order network model of the combustor (i.e., LOTAN code that solves the unsteady Euler equations). The proposed model for the dispersion of entropy waves only requires the mean velocity field as an input, obtained by Reynolds-averaged Navier-Stokes (RANS) computations of the demonstrator combustor. LOTAN is then coupled with a low-order model code (LINEARB) based on the semi-actuator disk model that studies propagation of combustion noise through turbine blades. Thus, by combining LOTAN and LINERAB, the combustion noise and its counterparts, direct and indirect noise, generated at the turbine exit are predicted. In comparison with experimental data, it is found that without the inclusion of entropy dispersion, the level of combustion noise at the turbine exit is overpredicted by almost 2 orders of magnitude. The introduction of entropy dispersion in LOTAN results in a much better agreement with the experimental data, highlighting the importance of entropy wave dispersion for the prediction of combustion noise in real engines. In more detail, the agreement with the experiment for high and low frequencies was very good. At intermediate frequencies, the experimental measurements are still overpredicted; however, the predicted noise is much smaller compared to the case without entropy dispersion. This discrepancy is attributed to (i) the role of turbulent mixing in the overall decay of the entropy fluctuations inside the combustor, not considered in the model developed for the decay of entropy waves, and (ii) the absence of a proper model in LINEARB for the decay of entropy waves as they pass through the turbine blade rows. These are areas that still need further development to improve the prediction of low-order network codes.