Search Results

This paper discusses an emerging area of applying machine learning (ML) methods to augment traditional Computational Fluid Dynamics (CFD) simulations of road vehicle aerodynamics. ML methods have the potential to both reduce the computational effort to predict a new geometry or car condition and to explore a greater number of design parameters with the same computational budget. Similar to traditional CFD methods, there exists a broad range of approaches. In particular, the accuracy and computational efficiency of a CFD simulation vary greatly depending on the choice of turbulence model (DNS, LES, RANS) and the underlying spatial and temporal numerical discretizations. Similarly, the end-user must select the correct ML method depending on the use-case, the available input data, and the trade-off between accuracy and computational cost. In this paper, we showcase several case studies using various data-driven ML methods to highlight the promise of these approaches.

Computational Fluid Dynamics (CFD) is widely used in vehicle aerodynamics development today, but typically used to study one vehicle shape at a time. In order to be used for aerodynamic shape exploration and optimization the CFD simulation process has to be able to study a large set of design alternatives (vehicle shape variants) within the short period of time typically available in the overall aerodynamics development process. This paper reports the development and testing of a process, referred to as the 50:50:50 Method, which is developed to study a large set of design alternatives in a highly automated way, while ensuring that each design alternative is simulated with a high fidelity CFD simulation.

Thorough design exploration is essential for improving vehicle performance in various aspects such as aerodynamic drag. Shape optimization algorithms in combination with computational tools such as Computational Fluid Dynamics (CFD) play an important role in design exploration. The present work describes a Free-Form Deformation (FFD) approach implemented within a general purpose CFD code for parameterization and modification of the aerodynamic shape of real-life vehicle models. Various vehicle shape parameters are constructed and utilized to change the shape of a vehicle using a mesh morphing technique based on the FFD algorithm. Based on input and output parameters, a design of experiments (DOE) matrix is created. CFD simulations are run and a response surface is constructed to study the sensitivity of the output parameter (aerodynamic drag) to variations in each input parameter.

To-date the primary challenge in conducting aerodynamic CFD simulations of actual vehicles with realistically complex geometry has been the construction of a computational mesh. The CAD-to-Mesh processes used to-date have been laborious, often requiring many weeks of engineering time. In this paper we present a new technique to greatly expedite the CAD-to-Mesh process. The fundamentals of this technique are discussed followed by case studies that show that this technique can reduce the engineering time required for the CAD-to-Mesh process to just a few hours.

Computational Fluid Dynamics simulation of aero-acoustics requires a high fidelity mesh. For Direct simulations, a very good quality and reasonably refined mesh is required in the entire domain encompassing the source and receiver of the sound. A usual practice so far has been to use structured grid to mesh the geometries. For complex geometrical shapes, such as throttle body, creating a fully structured mesh becomes very tedious and could consume a lot of time. Once the computational model is in place, obtaining meaningful solution also takes a long time since the solution has to be run for quite the long time in order to capture reasonably accurate sound pressure data. The current paper focuses on both of these time-consuming aspects. A comparative study of three different mesh types in a throttle body geometry is considered.

The purpose of present study is to use Immersed Boundary (IB) method in flow field simulations of a simplified generic ground transportation system (GTS) at 0° yaw. The IB method is usually employed in conjunction with a body non-conforming Cartesian grid. Thus, grid generation is greatly simplified. This plays an important role in reducing the cost and time in design process. This paper demonstrates the ease of use of IB method compared to body fitted mesh method and possible use of IB method to automate the external aerodynamics simulations. Also in order to assess the accuracy, the results are compared with corresponding experimental data reported in literature.

This paper examines the aerodynamic forces on the open hood of a stationary vehicle when another large vehicle, such as an 18-wheel semi-truck, passes by at high speed. The problem of semi-truck passing a parked car with hood open is solved as a transient two-vehicle aerodynamics problem with a Dynamic Moving Mesh (DMM) capability in commercial CFD software package FLUENT. To assess the computational feasibility, a simplified compact car / semi-truck geometry and CFD meshes are used in the first trial example. At 70 mph semi-truck speed, the CFD results indicate a peak aerodynamic force level of 20N to 30N on the hood of the car, and the direction of the net forces and moments on the hood change multiple times during the passing event.

A numerical method of predicting aeroacoustic performance of HVAC ducts is presented here. The method comprises of two steps. First, the steady state flow structure inside a duct is simulated using computational fluid dynamics (CFD). A k-epsilon based turbulence model is used. In the second step broadband noise source models are used to estimate the sound power generation within the duct. In particular, models estimating dipole and quadrupole sound source strengths are studied. A baseline generic duct geometry was studied with 3 additional design variations. The loudness rankings of these three designs were determined numerically. Simultaneously, the sound generated by these three designs was measured on a flow bench with a microphone kept downstream of the duct outlet. The numerically predicted loudness rankings were compared with experimentally determined rankings and the two are found to be in agreement, thus validating the numerical method.

A numerical simulation of the flow structure around an idealized automotive A-pillar rain-gutter and the sound radiated from it is reported. The idealized rain-gutter is an infinitesimally thin backward facing elbow mounted on a flat plate. It is kept in a virtual wind-tunnel with rectangular cross-section. The transient flow structure around the rain-gutter is described and time-averaged pressure distribution along the base plate is provided. Time-varying static pressure was recorded on every grid point on the base-plate as well as the rain-gutter surfaces and used to calculate sound pressure signal at a microphone held above the rain-gutter using the Ffowcs-Williams-Hawkings (FWH) integral method was used for calculating sound propagation. Both the transient flow simulation as well as the FWH sound calculation were performed using the commercial CFD code FLUENT6.1.22.

Minor geometric features in the intake manifold airflow path with side-branch cavities are often responsible for unusual noise due to the complex air flow structure and its interaction with the internal acoustic field. Although airflow bench tests are faster to evaluate various alternate design geometries, understanding the mechanism of such noise generation is necessary for developing an effective design. A 2D computational fluid dynamics (CFD) simulation was performed on a baseline geometry, which produced a distinct whistle, and on a modified geometry, which suppressed the whistle. These 2D models were able to simulate the flow-acoustic coupling responsible for the whistle generation and hence clearly predicted the presence or the absence of a distinct whistle peak as observed in the experimental measurements.

The present study is aimed at studying the use of Detached Eddy Simulation (DES) in simulating truck aerodynamics. A computational procedure based on DES implemented within the Finite Volume Method (FVM) framework is developed. Detailed descriptions of various aspects of the procedure are provided here including mesh generation, solution procedure and post-processing guidelines. The computational procedure is applied to study aerodynamics of a generic Ground Transportation System (GTS) at 0° yaw. This is a largely simplified ⅛th scale model of a tractor-trailer truck. Time-average and transient surface pressures, skin friction coefficients, and wake velocity structures are reported. To assess the accuracy of the present procedure, these are compared with corresponding experimental data reported in literature. Such comparison shows that the present procedure predicts drag coefficient accurately well within the bounds of experimental uncertainty.

Buffeting is a wind noise of high intensity and low frequency in a moving vehicle when a window or sunroof is open and this noise makes people in the passenger compartment very uncomfortable. In this paper, side window buffeting was simulated for a typical SUV using the commercial CFD software Fluent 6.0. Buffeting frequency and intensity were predicted in the simulations and compared with the corresponding experimental wind tunnel measurement. Furthermore, the effects of several parameters on buffeting frequency and intensity were also studied. These parameters include vehicle speed, yaw angle, sensor location and volume of the passenger compartment. Various configurations of side window opening were considered. The effects of mesh size and air compressibility on buffeting were also evaluated. The simulation results for some baseline configurations match the corresponding experimental data fairly well.

In the present work computational fluid dynamics (CFD) simulations of the flow field around a generic tractor-trailer truck are presented and compared with corresponding experimental measurements. A generic truck model was considered which is a detailed 1/8th scale replica of a Class-8 tractor-trailer truck. It contained a number of details such as bumpers, underbody, tractor chassis, wheels, and axles. CFD simulations were conducted with wind incident on the vehicle at 0 and 6 degree yaw. Two different meshing strategies (tet-dominant and hex-dominant) and three different turbulence models (Realizable k-ε, RNG k-ε, and DES) are considered. In the first meshing strategy an unstructured tetrahedral mesh was created over a large region surrounding the vehicle and in its wake. In the second strategy the mesh was predominantly hexahedral except for a few narrow regions around the front end and the underbody which were meshed with tetrahedral cells owing to complex topology.

Side window buffeting is simulated for a passenger car using unstructured mesh and a finite volume based CFD solver. We first provide a description of the analysis method. Two vehicle configurations are considered: front window open and rear window open. The accuracy of RNG k-ε and LES turbulence models is evaluated for this application by comparing predicted buffeting frequency spectrum with corresponding experimental measurements made in a wind tunnel. Further, the effects of several parameters on buffeting frequency and amplitude are studied. They include vehicle speed, yaw angle, inlet turbulent intensity, observer location inside the passenger compartment, presence of exhauster and side view mirror design. Simulation results prove to follow the trends observed in the experiments.