Flow control for road vehicle drag reduction

Abstract

This thesis covers topics that span bluff-body aerodynamics, hybrid RANS-LES CFD methods, flow control and model-order reduction. These topics arise from investigating the flow past three geometries: the bullet shaped D-body, the canonical squareback Ahmed body and the commerical Nissan NDP. The study on the D-body was aimed at transitioning the research group from the restrictive block-structured formulated StreamLES solver to the more flexible OpenFOAM code that can use unstructured meshes. Linear feedback control for base pressure increase was applied as was done in the work by Dalla Longa et al. (2017). Identification of the plant, G(s), that represents the wake's response to forcing was completed and correlated well with the results from Dalla Longa et al. (2017). The same can also be said of the sensitivity based designed feedback control law, K(s). When applied in simulation, an attenuation of the base pressure fluctuations was, as desired, achieved, although the base pressure increased by 24.5% as opposed to the 38% achieved by Dalla Longa et al. (2017). In the study on the squareback Ahmed body, wall-resolving (WRLES) and wall-modelled (WMLES) large eddy simulation were successfully applied. First, a simulation setup that is both able to resolve wake bimodality, while remaining reasonable in computational resource use, was created. Subsequently, variants of this setup were used to identify a flow feature that plays a critical role in forcing wake bimodality events. More specifically, a heavily under-resolved WMLES simulation in which both the near-wall and part of the outer-region of the turbulent boundary layer are Reynolds-averaged did not capture the front recirculation bubble near the Ahmed body nose; neither did it resolve a bimodal wake switching event. Meanwhile, the simulations with a more refined near-wall mesh did capture the front separation bubble as well as bimodal switching events of the wake. This front separation bubble sends out powerful hairpin vortices that interact with the rear wake. Specifically, these vortices go on to produce significant amounts of TKE, which, upon convection to the rear of the Ahmed body, ultimately help trigger a bimodal event. The Ahmed body study also involved the application of linear feedback control for drag reduction as was done in the D-body study. In the short term, mean blowing did lead to a base pressure increase, but as the zero-net-mass-flux (ZNMF) jet settled, it oscillated around zero making its effects indiscernible. The final geometry analyzed was the Nissan NDP. This was done by performing benchmark wall-resolving LES (WRLES). First, the benefit of appending a rear cavity to an otherwise "squareback" geometry was assessed. It was concluded that the cavity allows the wake to move more freely about the rear base. Specifically, the wake is freed from its more restricted motion that is present with the "squareback" Nissan NDP. In doing so, the drag reduction achieved with the cavity appendage is about 13.6%. Work on the Nissan NDP also involved an assessment of a moving ground in the simulation. It was concluded that, in the stationary ground simulation, flow detachment at the ground where the flow exits from the underbody has an adverse drag effect. In other words, although moving ground simulations better replicate the real-world conditions, the stationary ground variant is in this case more conservative, as it returns slightly higher drag values.