Post-processing enhancement: feature detection and evaluation of unsteady/steady flows

The exponential growth in computational capabilities, and the increasing reliability and precision of current simulation solvers, has fostered the use of Computational Fluid Dynamics (CFD) in the analysis of highly non-linear and complex flow problems. The nature of these flows usually involves a large number of scales and flow features, which makes it very challenging to achieve a clear understanding of the inherent problem. Additionally, current numerical simulations produce large amounts of raw data that needs to be evaluated. However existing post-processing tools are unable to extract with accuracy and efficiency all the valuable information contained in it.

Searching for meaningful structures through the entire dataset, by means of classical visualization techniques, might result in a fruitless, or at least inefficient, effort. Alternatively, with the use of flow feature detection and data decomposition techniques, the identification of the relevant features becomes much more straightforward, allowing more accurate visualizations and faster analysis, with lower uncertainties. In this thesis, these two promising post-processing approaches are studied, and applied to problems of physical and industrial relevance: a three-dimensional open cavity flow, and a Counter-Rotating Open Rotor (CROR) engine. On the one hand, amongst current feature detection techniques, Region-based (RB) vortex detection methods can delimit rotating regions in the flow, while Line-based (LB) ones are capable of reconstructing the imaginary center lines of the vortices. On the other hand, the Dynamic Mode Decomposition (DMD) is a recent tool used to decompose oscillatory dominated flows into spatial modes, with the advantage of associating each extracted dynamic mode to a single frequency.

At first, the DMD technique is employed to investigate the dynamics of saturation inside a rectangular open cavity. Previous experiments and linear stability analysis of the problem completely described the flow in its onset, as well as in a saturated regime, characterized by coherent three-dimensional centrifugal modes. The morphology of the modes observed in the experiments matched the ones predicted by linear analysis, but with a shift in frequencies for the dominant oscillating modes. This work presents a detailed numerical simulation of the entire saturation process, from 2D to 3D flow, shedding some light on the main mechanism that produces the discrepancies encountered between both approaches. The capability of the DMD to analyze the underlying dynamics inside the cavity is demonstrated in this thesis, enabling to explain the main reason for the aforementioned differences in frequency.

Finally, some vortex detection algorithms are applied to the particular case of CROR, aiming the monitoring and visualization of the trajectory of the vortices generated at the tip of the front rotating blades. This is of critical importance to understand and prevent vortex-blade interaction with subsequent stages, as this non-linear flow topology strongly influences the aerodynamic performance and acoustic footprints of the engine. The suitability and performance of four typical Region-based (RB) vortex detection criteria, and one Line-based (LB) method, are firstly evaluated. Then, two new methodologies are introduced that improve the original assortment of seeds required by the tested LB method, as they increase the probability of the selected seeds to grow into a tip vortex line, providing faster and more accurate answers during the design-to-noise iterative process.