Combustion noise has become a significant contributor to the overall noise emitted by modern aero-engines. This development is attributed to reduced noise sources in other components due to design improvements and the introduction of premixed combustors that burn more unsteadily and hence emit more noise. These next generation combustion systems are more prone to acoustic instabilities and thus require improved methods for the prediction of combustion noise. This paper presents a numerical approach which exploits the different scales prevalent in these combustion systems by computing the flow field and acoustics separately and coupling both simulations in real time. To demonstrate its capabilities, the methodology is successfully applied to a premixed and pressurized propane flame which was experimentally investigated by CNRS [1] and for which pressure measurement data are available. In the present numerical model, the separation of the physical phenomena facilitates the application of the most suitable numerical schemes and governing equations to both sets of problems. Due to the low Mach numbers governing typical combustors, the flow field can be adequately described by the incompressible Navier-Stokes equations. These are solved by an implicit finite volume flow solver which is supplemented by a tabulated chemistry approach to account for the combustion processes. For the acoustics, the three dimensional Acoustic Perturbation Equations (APE) are solved using a state of the art, low dispersion Discontinuous Galerkin CAA tool. Both codes are run in parallel and exchange fields on-line to maintain the highest possible temporal resolution and data transfer rates. The different natures of both phenomena require an elaborate coupling scheme that comprises temporal and spatial interpolation and filtering. Since the incompressible formulation of the flow solver allows for considerably larger time steps and the acoustics solver employs much simpler governing equations, the overall computational costs of this approach are up to 10 times lower than those of a compressible simulation with similar fidelity.

This content is only available via PDF.
You do not currently have access to this content.