Numerical Investigation of the Flow Around a Generic Car Using Dynamic Mode Decomposition

Abstract

The drag of a car is highly dependent on the topology of its complex wake system. Small changes in the shape of the car, that do not have a big effect when considered separately, can lead to significant changes in the total drag when the vortex systems of the changed part of the car body interact with the wake vortices. To understand these interferences, a method is necessary that decomposes the flow based on dynamic information. In this paper, the feasibility of using the Dynamic Mode Decomposition (DMD) to analyze the dynamic behavior of the wake flow of a car is investigated. The DMD is found to extract useful information from the flow when applied to three dimensional velocity vector fields. The CFD simulations are validated by yet unpublished experimental results from experiments in two different wind tunnels.

Investigation of the Unsteady External and Underhood Airflow of the DrivAer model by Dynamic Mode Decomposition Methods

“…Before performing DMD analysis, a low-pass filter with 10Hz of cut off frequency is applied on the wall force for the temporal direction by forward-backward filtering with 2nd order Butterworth filter, which results in 4th order zero-phase filter, in addition to filtering performed by SVD. These preprocessing techniques are conducted according to Peichl, et al (4) (Please also refer to Schmid (1) , and Duke, et al (20) regarding sensitivity of dataset on DMD.) Total time length of the dataset used for DMD is 3.5 sec in total and snapshots of the flow field are sampled with ∆ = 0.005sec.…”

“…In the field of applied aerodynamics, Muld, et al (3) analyzed the unsteady flow field around a high-speed train by using DMD, and identified the flow structure related to slip-stream. Peichl, et al (4) applied DMD on the flow field around a road vehicle and discussed the coherent flow structures which are dominant in the unsteady flow field around the vehicle through the results of DMD. In these researches, DMD was applied on the velocity field, but DMD can be applied on any multi-dimensional data.…”

“…the Ahmed body, that have mostly been investigated in academic research and real production car geometries which are developed and investigated in industry but to which academia has only limited access (7) . The geometry data is publicly available and the DrivAer has therefore been well respected and is the subject of multiple investigations and publications both numerically and experimentally (4), (8 -10) . Since the model comes with three different rear end types, a notchback, an estate back and a fastback type, it offers a variety of possible insights into the external flow around a vehicle.…”

Online Dynamic Mode Decomposition Methods for the Investigation of Unsteady Aerodynamics of the DrivAer Model (First Report)

“…For example of DMD analysis in the field of applied aerodynamics, Muld, et al (3) identified the flow structure related to slip-stream behind a highspeed train by applying DMD on the unsteady flow field. Peichl, et al (4) discussed the dominant flow structures in the unsteady flow field around a road vehicle by using DMD.…”

Online Dynamic Mode Decomposition Methods for the Investigation of Unsteady Aerodynamics of the DrivAer Model (Second Report)

“…For vehicle aerodynamics, DMD has been successfully applied to examine CFD data with various objectives. Peichl et al (3) presents the application of conventional DMD on flow field data from a Detached Eddy Simulation of the DrivAer reference body. They suggest preprocessing the flow field data, which filters out numerical noise and high frequency components, in order to generate a Reduced Order Model (ROM).…”