Gardner, B., Mejdi, A., Musser, C., Chaigne, S. et al., "Coupled CFD and Vibro-Acoustic Modeling of Complex-Shaped Mufflers Accounting for Non-Uniform Mean Flow Effects," SAE Technical Paper 2015-01-2313, 2015, doi:10.4271/2015-01-2313.
Flow strongly affects the propagation of acoustics wave transmission within a duct and this must be addressed by the vibro-acoustic modelling of duct systems subject to non-uniform flow. Flow impacts both the effective sound propagation speed in a duct and refracts the sound towards or away from the duct walls depending on whether the acoustic waves are propagating in the direction of the flow or against the flow. Accurate modeling of the acoustic propagation within a duct is crucial for design and “tuning” of muffler systems that need to strongly attenuate narrowband acoustic sources from the engine. Muffler systems that may avoid matching acoustic resonances to engine narrowband sources when no flow is present may experience shifting of resonances to frequencies that match engine sources and cause problems when the flow during a real operating condition is present. Therefore accounting for detailed flow effects on the acoustic propagation and the modal characteristics of the muffler system is essential for effective design. However, for real-life, complex geometries, flow patterns have no analytical solution and have to be determined by means of computational fluid dynamics (CFD). This paper describes an automated process for coupling acoustic finite element (FE) muffler vibro-acoustic design to an open source computational fluid dynamics (CFD) solver whose predicted results are automatically returned and used to generate muffler noise attenuation predictions accounting for flow effects. A validation case study showing the ability for a non-CFD expert modeler to obtain accurate muffler predictions for cases of non-uniform flow is presented. This approach can support a broad range of muffler designs and can represent a significant savings of testing and development time. Limitations and future steps for this modeling approach are discussed.