Linear aeroacoustic solver in OpenFOAM

Abstract

Due to rapid advances in computer technology the field of computational aeroacoustics has become an alternative to empirical and experimental methods for the prediction of sound production and propagation. However, in order to provide numerical aeroacoustic predictions for realistic applications a hybrid approach is still often the only option. OpenFOAM is a popular open-source CFD package in the academic world and its usage is increasing in industry as alternative for expensive commercial software. Currently there is no acoustic module in OpenFOAM. Following a hybrid methodology, the user would therefore need to transfer saved ow data to another software package to perform the acoustic calculation. This requires high data storage and makes the method (unnecessarily) complicated and cumbersome. In this thesis an aeroacoustic method based on the linearized Euler equations (LEE) has been implemented in OpenFOAM. Instead of using the standard available OpenFOAM schemes, a different numerical method was implemented. This involves solving a Riemann problem to determine the convective terms. The code was veriffed on a series of benchmark problems for which analytical or reference solutions are available. The considered problems were one-dimensional wave propagation, the propagation of an acoustic pulse in uniform ow, the radiation by elementary acoustic sources and source radiation in two types of non-uniform ow. In all cases the found numerical solution was very close to the reference solution. The Riemann based solver was compared for these problems with a LEE solver that uses the readily available discretization schemes of OpenFOAM. Both implementations were shown to be able to correctly account for convection and refraction effects. The Riemann based solver has as advantage that it propagates waves in a less dispersive manner than the standard solver, but has as disadvantage that it is more di_usive. As validation case the sound production by a pair of co-rotating vortices was chosen. In addition to reference data, a direct simulation was carried out to also serve as reference solution. The hybrid method was tested with the LEE and with Lighthill's analogy. If the average acoustic source is not included as source term in the LEE, the LEE and Lighthill solutions were found to be very similar. Compared to the direct simulation, both hybrid methods showed a clear phase difference, which for this case can be attributed to the start-up phase of the hybrid simulation. Finally, it was demonstrated that the ow calculation can be done on small domain, followed by a mapping of the obtained acoustic source onto the larger acoustic grid, after which the acoustic equations are solved for that ow time step. This leads to considerable savings in computational power and disk space usage.