Stochastic noise source methods for efficient computation of aeroacoustic nearfield phenomena

This academic thesis delves into the topic of broadband aeroacoustics, with a particular focus on the development of efficient numerical prediction methods. The study introduces a stochastic noise source approach based on the Fast Random Particle-Mesh method within an aeroacoustic nearfield framework. The realized velocity fluctuations satisfy prescribed one-point and two-point turbulence statistics by a weighted convolution of a spatio-temporal white-noise field with filter kernels of predefined correlation lengths. The hydrodynamic nearfield pressure, obtained through the conversion of solenoidal velocity fluctuations with a Poisson equation, serves as a crucial metric for the aeroacoustic source term. This source term is integrated into a hydrodynamic/acoustic splitting approach based on the Perturbed Convective Wave Equation. The research systematically evaluates the method's performance across various characteristic external aeroacoustic test cases. Starting with a forward-facing step at Reynolds number Re_h=8000 and validating the method to highly resolved Large Eddy Simulations in both farfield and nearfield as well as amplitudes, coherences and phase relations, it progresses to Re_h=30000. The prescription of the integral length scale as foundation of the local correlations of the synthesized vortices is worked out to be of significant importance for the absolute levels of the acoustic pressure. However, it is also demonstrated that relative differences within the investigated limits are captured when absolute deviation to reference data is given. This holds for geometrical modifications such as step heights and chamfered steps. The computed trends in the RANS input data are sufficiently well represented for the Fast Random Particle-Mesh method to react in differences of the acoustic amplitude. The investigation extends to the application of the Fast Random Particle-Mesh method in the context of automotive external aeroacoustics, focussing on a side mirror configuration and subsequently explores its limitations in underbody noise simulations. The side mirror reveals further importance on the quality of the input data as a significant flow phenomenon in the form of a separation bubble on the inner side of the mirror is crucial for the local production of turbulent kinetic energy. This feature is only captured in one specific turbulence model resulting in high-quality aeroacoustic levels. The underbody is well-behaved due to one singular source mechanism for acoustic contribution, namely edge noise of the wheel house surroundings. Various turbulence models compute satisfying acoustic patterns. The hydrodynamic pressure however shows first inaccuracies due to its locality and higher dependency on input data. The theoretical limits assumed in the derivation of the Fast Random Particle-Mesh method employing local isotropy in one-point and two-point turbulence statistics are reached. The results show the robustness, accuracy and consistency of the Fast Random Particle-Mesh method, particularly when dealing with geometric variations and changes in turbulence models. The thesis emphasizes the importance of careful consideration of input data quality and highlights the adaptability of the Fast Random Particle-Mesh method to modifications in integral length scales. The successful application of the method in predicting acoustic Deltas for diverse geometries including challenges posed by rounded step configurations and underbody noise scenarios, underscores its potential for early-stage development in aeroacoustic optimization frameworks. The computational advantage of the method with a speed-up of factor 7-8 which is shown to be consistently given for each test case contributes to the overall performance.