Simulation of Vehicle Soiling at Audi.

Dr. Moni Islam
Audi AG
Wind Tunnel Centre

With CFD methods now firmly established as a development tool for the prediction of the aerodynamic forces acting on a vehicle, the use of CFD for predicting the soiling characteristics of a vehicle have been investigated at Audi during the past several years. Of interest is how particulate matter from the environment – water, snow or dirt – is deposited, accumulated and transported on the surface of the car and in its vicinity. Common examples of these phenomena are the flow of water from the wind screen onto the side windows, the deposition of rain drops or dirt on the surface of a wing mirror or the accumulation of dirt on the rear surface of the boot lid. Until recently, activities have focused primarily on assessing whether or not the existing two-phase-flow models in STAR-CD are suitable and adequate for this application. The results presented in this article present some of the results of this work.

The primary challenges of simulating soiling phenomena are the accurate description of the physics of the liquid and solid phases and the correct modelling of the interaction of those phases with the air. Of particular relevance are, for example:
• size- and velocity-distributions of rain drops;
• modelling of snow and its properties;
• impact behaviour of rain drops on solid surfaces;
• deposition and accumulation of solid particulate (eg. dirt, snow) on solid surfaces;
• liquid-film formation and transport;
• primary break-up of liquid films due to aerodynamic forces;
• characterising the droplets generated at the road surface and at the wheels due to vehicle motion.
These issues, the last point in particular, present a major challenge to the successful modelling and simulation of vehicle soiling.

In the methodology applied at Audi, soiling is simulated using a one-way coupling between the gas and liquid/solid phases, namely that the liquid/solid phase is influence by the gas phase, but not vice versa. In this way, the computational overhead is kept relatively low, as calculations are performed on the liquid/solid phase only, with the gas phase “frozen”. The existing two-phase models in STAR – both two-phase Lagrangian and liquid film – are applied to vehicle configurations known from experiments, and the results of the simulations are compared with the experimental results in a qualitative way. In what follows, a number of examples are presented.

Figure 1: Experiment and simulation of rain soiling of the Audi A2

Figure 1 shows an example taken from investigations to assess the prediction of the interaction of rain drops with the vehicle's surface. The wind-tunnel experiments, seen on the left, show the existence of a cloud of droplets standing off the front bumper surface, arising from the interaction of the oncoming flow and the droplets rebounding off the surface. In the simulation, the droplets are modelled according to the standard Lagrangian formulation, with the initial/boundary conditions for size- and velocity-distribution obtained from experiments. Their interaction with the surface – whether they rebound, break--up, stick or smear – is based on the models developed by Bai available in STAR. Analysis of the simulation results shows that the experimentally observed phenomena are qualitatively predicted, and further investigations are underway to assess the predictions quantitatively and in more detail. Here, the formation of liquid films on the vehicle’s surface was not considered.

Figure 2: Experiment and simulation of film formation on the Audi A2

Studies of this phenomenon have also been carried out, as illustrated in Figure 2. In the simulations, as in the experiments, a liquid film is allowed to accumulate on the vehicle surface through the impingement of liquid droplets originating upstream. The film then evolves under the influence of aerodynamic forces and is transported across the vehicle surface accordingly. A number of qualitative features of the film distribution are captured fairly well by the simulation, such as the local film-thickness maxima between the headlamps and the inlet grille, as well as the separation bubble and accumulated film just downstream of the Audi rings on the bonnet. The film model represents a statistically averaged film and therefore cannot capture the water streaks observed in the experiments, which are stochastic in nature. Furthermore, these streaks are highly dependent on small local geometric features of the trim and gaskets and microscopic surface properties, not contained in the simulation model.

Figure 3: Aft soiling patterns on the Audi A2 for various particle sizes

Also of interest is the soiling of the aft portion of the vehicle arising from entrainment of particulate matter from the underbody and wheels into the vehicle’s wake. Figure 3 shows the soiling patterns on the rear hatch of the Audi A2, represented by a normalised number of particle impacts, for particles of two different diameters. As can clearly be seen, the smaller particles are entrained by the wake of the vehicle and impact on the surface of the rear hatch. The larger particles have a sufficiently high momentum that they are not deflected sufficiently by the air flow to land on the rear hatch.

Currently, these investigations only consider the unsteady aspect of the liquid phase, i.e. the transport and accumulation of droplets or particulate matter in the vehicle's vicinity. The interaction of the unsteady air flow, with its large-scale turbulent motions, and the droplets is clearly an aspect that must also be investigated in subsequent steps. Furthermore, essential for the accurate modelling of soiling phenomena is the correct description of two-phase boundary conditions for realistic road-like simulations, namely droplet atomisation from liquid films on road and tire surfaces and the two-phase wake of upstream vehicles. For these reasons, predicting vehicle soiling will remain a significant challenge for some time.

Acknowledgements: The assistance of Dr. L. Lührmann, Dr. M. Jaroch and Mr. H. Steinicke of Audi AG in carrying out the experimental investigations, and of Icon Computer Graphics Ltd. for the simulations, is gratefully acknowledged