Boundary element and finite element coupling for aeroacoustics simulations

Nolwenn Balin, Fabien Casenave, François Dubois, Eric Duceau, Stefan Duprey, Isabelle Terrasse

Simulating how engine noise propagates through complex airflow is a critical challenge in aircraft design and certification. Researchers from Airbus and French academic institutions have developed a robust numerical framework that hybridizes two classical methods—the Finite Element Method (FEM) and the Boundary Element Method (BEM)—to create a tool capable of handling realistic "potential" flows near an engine while efficiently modeling the vast, uniform far-field.

The Computational Challenge of Aeroacoustics

Engineers typically face a difficult trade-off when modeling aircraft noise. Uniform flow models are computationally fast but fail to capture the nuances near the engine where air curves around the intake and nacelle. Conversely, full-volume models like FEM can handle complex geometries and varying flows but become prohibitively expensive when modeling sound radiating into the infinite far-field.

The core challenge lies in solving the convected Helmholtz equation, which describes sound waves moving through a fluid that is itself in motion. To address this, the authors propose a domain decomposition strategy that leverages the strengths of both FEM and BEM.

Methodology: A Hybrid FEM-BEM Approach

The researchers split the computational space into two distinct zones: an interior zone ($\Omega^i$) near the engine and an exterior zone ($\Omega^e$).

  • FEM for the Near-Field: Inside $\Omega^i$, the flow is "potential" (varying with space but irrotational). FEM is used here for its ability to handle complex geometries and spatially varying media.
  • BEM for the Far-Field: In $\Omega^e$, the flow is assumed to be uniform. BEM is ideal here because it only requires meshing the boundary of the domain, naturally handling the radiation of sound to infinity without a massive volume mesh.
Figure 1: Domain decomposition for the hybrid FEM-BEM approach, showing the interior potential flow region (Ωⁱ) and the exterior uniform flow region (Ωᵉ).

The Prandtl–Glauert Transformation

A key innovation in this framework is the application of the Prandtl–Glauert transformation. By "stretching" the coordinate system and shifting the phase of the waves, the authors map complex convected wave equations into standard Helmholtz forms. This allows them to use existing, highly optimized industrial BEM solvers (such as Airbus's ACTIPOLE) that were originally designed for stationary fluids. The two methods are linked at the interface using Dirichlet-to-Neumann (DtN) operators, ensuring continuity of pressure and velocity.

Validation and Industrial Application

The framework was validated using several benchmarks, including a sphere in a non-uniform medium and a cylindrical duct model. In the sphere test case, the coupling achieved a relative error as low as $7 \times 10^{-4}$ compared to analytical solutions.

Figure 2: Detailed view of the engine model domain (Ω_M) and boundary (Γ_M) used for industrial-scale validation.

To demonstrate industrial scalability, the team performed a full 3D simulation of a simplified turboreactor engine. The model involved 11.8 million tetrahedrons and 1.2 million degrees of freedom, with solve times ranging from 1.5 to 6.5 hours on a high-performance cluster.

Key Insights

  • Impact of Potential Flow: Simulations revealed that ignoring the potential flow near the engine (assuming uniform flow everywhere) leads to errors of approximately 1 dB in predicted noise levels. In the context of aircraft certification, a 1 dB margin is highly significant.
  • Efficiency through Symmetry: By carefully choosing integral operators, the authors maintained a symmetric linear system, which is faster to solve and more memory-efficient.
  • Decoupled Physics: The approach solves for the flow first and then uses it as an input for the acoustic simulation, simplifying the overall computation.

Limitations and Future Directions

While powerful, the current model assumes "potential" flow, meaning it cannot yet capture the turbulence and vortices found in engine exhaust. Future work may involve extending the framework to the Galbrun equation, which would allow for modeling more complex, rotational flows—such as the swirling air behind a fan—while retaining the efficient BEM far-field coupling.

Summary updated

Abstract

We consider the scattering of acoustic perturbations in a presence of a flow. We suppose that the space can be split into a zone where the flow is uniform and a zone where the flow is potential. In the first zone, we apply a Prandtl-Glauert transformation to recover the Helmholtz equation. The well-known setting of boundary element method for the Helmholtz equation is available. In the second zone, the flow quantities are space dependent, we have to consider a local resolution, namely the finite element method. Herein, we carry out the coupling of these two methods and present various applications and validation test cases. The source term is given through the decomposition of an incident acoustic field on a section of the computational domain's boundary.
BibTeX
@article{balin2015boundary,
  author = {Nolwenn Balin and Fabien Casenave and François Dubois and Eric Duceau and Stefan Duprey and Isabelle Terrasse},
  title = {{Boundary element and finite element coupling for aeroacoustics simulations}},
  journal = {Journal of Computational Physics},
  year = {2015},
  url = {https://openreview.net/forum?id=lWDqBfjbChX},
}
Figure