GPU-accelerated LES framework for aeroacoustic predictions of integrated systems

To achieve the targets of efficient fuel burn, low pollutant emissions and low noise radiation, new engine design concepts have been developed for next generation aircraft. These designs leverage a more compact engine-airframe integration to reduce drag and distributed turbofans with smaller diameters to control noise. These tightly coupled systems have posed a greater challenge to numerical prediction tools, especially for problems such as aeroacoustics that requires high-fidelity simulations. While high fidelity simulations (e.g., wall modeled/wall resolved large eddy simulations) have demonstrated predictive capabilities for separated boundary layers, transitions, and turbulent wakes, they are often infeasible in design contexts due to their high computational costs, and only limited to component level analysis. This SBIR project will incorporate recent technical advances Cascade has made in low dissipation, compressible flow discretizations and wall modeling for large eddy simulations, as well as an innovative approach for treating stationary-moving interface which is conservative, numerically stable, and computationally efficient. The proposed framework is implemented to exploit cost-effective throughput afforded by modern accelerated architectures (e.g., GPUs). This will finally result in an approach for these turbomachinery flow and aeroacoustic calculations that are both accurate and feasible (with less than 1 day wall clock turnaround times). This affordable approach will eventually provide a high-fidelity computational solution for system level designs.