| dc.description.abstract | Challenges in turbulence modeling are often associated with a trade-off between accuracy of flow physics and computational effort. For complex industry-relevant flows, high computational costs require engineers to adopt statistical closures, e.g., Reynolds-averaged Navier-Stokes (RANS) models. These models, however, are unable to capture large-scale multi-point correlations in the flow. On the other hand, high fidelity simulations, e.g., large-eddy simulations (LES) and direct numerical simulations (DNS) are often associated with large computational effort and are currently limited to simple geometries at low Reynolds numbers. Scale-resolving simulations (SRS) have emerged in recent years as physics-based multi-point closures that provide ‘accuracy-on-demand’. Partially-averaged Navier-Stokes (PANS) is a bridging-SRS strategy that offers computational accuracy ranging from RANS to DNS as a function of commensurate computational effort by specification of implicit filter parameters (fk , fε ). These filter parameters dictate the range of scales resolved in the turbulent flow thereby controlling the accuracy. The effect of unresolved scales on the resolved field is captured using two-equation turbulence closures, e.g. k − ε, k − ω models. This allows for the cut-off filter to be placed in inertial subrange and only dynamically relevant scales may be resolved reducing overall computational effort compared to LES. Thus, PANS models are well positioned to provide multi-fidelity simulations for numerous marine applications, specifically numerical wave tanks (NWTs). Two major challenges for application of turbulence models in accurate simulations of NWTs include efficient near-wall modeling and reliable representation of wakes. Therefore, this dissertation focuses on advancing the PANS strategy in: (i) near-wall closure and (ii) quantitative analysis of coherent structures in three-dimensional wakes. The first study develops a PANS two-layer turbulence modelling approach. This closure aims to provide a simple and robust computational strategy to alleviate numerical challenges posed by steep gradients near-wall. This model allows for the near-wall region to be computed with one-equation model while the outer high-Re region is resolved with the PANS ku − εu model. This wall-resolved PANS (WR-PANS) two-layer model is analyzed in turbulent channel flow for Reτ = 180 − 950. The results establish the model as an accurate and computationally feasible approach with inherent ease of application. For bluff body wakes displaying multi-point coherence, it is imperative to ensure the reliability of coherent structures predicted by PANS simulations. In Study 2, a quantitative framework for assessment of the large-scales structures is proposed in the wake of a sphere at Re = 3700 using proper orthogonal decomposition (POD). The simulations are performed using a WR-PANS approach with the PANS ku−ωu closure for the unresolved scales. The results reveal that multi-point correlations in the resolved flow are accurately captured using WR-PANS closures. In an effort to further reduce the computational burden of resolving the near-wall region, a wall-modeled PANS (WM-PANS) strategy is investigated in Study 3. This model allows for a spatially varying filter in order to compute the near-wall region using RANS and resolve the wake region using high-resolution PANS. The commutation residue arising from the spatial variation of filter parameter is systematically incorporated using energy conservation principle. The theoretical foundation and computational outcome of the model is examined in turbulent channel at Reτ = 950 and 8000. The model is further investigated in flow past a sphere at Re = 3700. The results are highly encouraging. In the final study, the WM-PANS strategy is applied to flow over a sphere at very high Reynolds number, Re = 1.14×106 . The results further confirm that the WM-PANS strategy is a computationally efficient approach for simulating practical flows. Moreover, quantitative assessment of coherent structures based on the approach outlined in Study 2 demonstrates that reasonably accurate large scales are resolved in this supercritical regime using WM-PANS. | en |
