Cerfacs Enter the world of high performance ...

PhD Defense: Franchine NI – Accounting for complex flow-acoustic interactions in a 3D thermo-acoustic Helmholtz solver

  Monday 24 April 2017 at 14h00

  Phd Thesis       Cerfacs, Salle de conférence Jean-Claude ANDRÉ    


Environmental concerns have motivated turbine engine manufacturers to create new combustor designs with reduced fuel consumption and pollutant emissions. These designs are however more sensitive to a mechanism known as combustion instabilities, a coupling between flame and acoustics that can generate dangerous levels of heat release and pressure fluctuations. Combustion instabilities can be predicted at an attractive cost by Helmholtz solvers. These solvers describe the acoustic behavior of an inviscid fluid at rest with a thermoacoustic Helmholtz equation, that can be solved in the frequency domain as an eigenvalue problem. The flame/acoustics coupling is modeled, often with a first order transfer function relating heat release fluctuations to the acoustic velocity at a reference point.
One limitation of Helmholtz solvers is that they cannot account for the interaction between acoustics and vorticity at sharp edges. Indeed, this interaction relies on viscous processes at the tip of the edge and is suspected to play a strong damping role in a combustor. Neglecting it results in overly pessimistic stability predictions but can also affect the spatial structure of the unstable modes. In this thesis, a methodology was developed to include the effect of complex flow-acoustic interactions into a Helmholtz solver. It takes advantage of the compactness of these interactions and models them as 2-port matrices, introduced in the Helmholtz solver as a pair of coupled boundary conditions: the Matrix Boundary Conditions. This methodology correctly predicts the frequencies and mode shapes of a non-reactive academic configuration with either an orifice or a swirler, two elements where flow-acoustic interactions are important.
For industrial combustors, the matrix methodology must be extended for two reasons. First, industrial geometries are complex, and the Matrix Boundary Conditions must be applied to non-plane surfaces. This limitation is overcome thanks to an adjustment procedure. The matrix data on non-plane surfaces is obtained from the well-defined data on plane surfaces, by applying non-dissipative transformations determined either analytically or from an acoustics propagation solver. Second, the reference point of the flame/acoustics model is often chosen inside the injector and a new reference location must be defined if the injector is removed and replaced by its equivalent matrix. In this work, the reference point is replaced by a reference surface, chosen as the upstream matrix surface of the injector. The extended matrix methodology is successfully validated on academic configurations. It is then applied to study the stability of an annular combustor from Safran. Compared to standard Helmholtz computations, it is found that complex flow-acoustic features at dilution holes and injectors play an important role on the combustor stability and mode shapes. First encouraging results are obtained with surface-based flame models.


Yves AUREGAN       Research Director – Laboratoire d’Acoustique de l’Université du Maine                    Referee

Jonas MOECK          Associate Professor – TU Berlin ISTA                                                                        Referee

Aimee MORGANS    Associate Professor – Imperial College London                                                         Member

Yoann MERY             Research Engineer – Safran Aircraft Engine                                                              Member

Thierry POINSOT      Research Director –  Institut de Mécanique des Fluides Toulouse                             Advisor

Franck NICOUD        Professor – Université de Montpellier  IMAG                                                              co Advisor







Data assimilation

From Monday 24 June 2024 to Thursday 27 June 2024






Machine learning for data science

From Monday 7 October 2024 to Thursday 10 October 2024






Fundamentals to understand and analyse high fidelity compressible Large Eddy Simulations

From Monday 14 October 2024 to Wednesday 16 October 2024

  Training       Cerfacs, Toulouse, France