Pollutants are emitted by different sources and it is worthwhile to note that ground vehicles are one of the main contributors to their high concentration levels in urban areas. Among them, ultrafine particles (UFP) with diameter below 100nm and gas such as nitrogen oxides (NOx) are due to engine combustion meaning that this is a key issue for automotive engineering.

UFP may be inhaled by breathing and penetrate into the respiratory system. Thus, car passengers, cyclists and pedestrians are exposed to those UFP that have strong impacts in terms of health. UFP increase mortality and diseases leading to huge costs which have been estimated at up to 2.6% of GDP in France (WHO).

Some recent studies have put in evidence that UFP concentrations may be larger inside the cabin than outside. Decreasing the infiltration rates and thus exposure of car passenger during commuting time is then a major goal. Having that in mind, understanding the nanoparticle dispersion in the wake of a car since their emission from the tailpipe is paramount. Characterizing their interactions with the flow will help to improve our knowledge regarding their dynamics. To achieve this goal, the first step is to provide an accurate description of the flow dynamics downstream of a car. This is the target of the present paper.

Measurements are conducted in a wind tunnel whose test section length, height and width are Ls=1m, Hs=0.3m and ls=0.3m, respectively. Velocity measurements are recorded using a 2D LDV system mounted on a 2D traverse system. Three simplified car models known as Ahmed bodies and representative of flow topologies developing downstream of real cars are used, with rear slant angles of φ=0°, 25° and 35°. Their length/height/width are Lc=0.196m, Hc=0.054m and lc=0.073m, respectively. Vehicles are fixed on the bottom of the test section through four circular supports (height 15mm). The blockage coefficient is then below 5% in agreement with the literature to avoid wall effect. For all experiments, the ratio between velocities of the upstream flow and the exhaust gas one is similar to that of a real car moving in an urban area. It leads to a constant upstream velocity of U∞=14.3m/s, the corresponding Reynolds number based on the model height being Re=49,500.

A preliminary detailed calibration of the test section enables us to identify a homogenous region with a boundary layer thickness less than 12mm and a turbulence intensity level always below 1%. Achieving optimum flow seeding for LDV measurements is known as a key parameter to obtain successful and representative results. Due to the irregular inter-arrival time between measurements, specific data treatment methods must be taken into account. Different methods are available in the literature that avoid velocity statistics bias occurring in the case of homogeneous spatial concentration of seeding. However, there is a lack of information about non-homogenous situations that causes bias through burst seeding occurrences. Here, an innovative data analysis method has been developed to provide reliable and repeatable results whatever the seeding conditions are. According to that, the boundary layer development on the roof has been characterized showing a separation downstream the front edge of models. Then, wake turbulence fields developing downstream of our three models in the same wind tunnel are assessed. Flow topology, recirculation lengths (Lr), boundary layer detachment on rear slants and shear-layer properties are characterized and compared to literature. Among others, our results show that Lr is maximum (Lr=1.5Hc) for φ=0° and minimum (Lr=0.6Hc) for φ=25°. The maximum of turbulence intensity is about 30% for all models whereas their profiles are very sensitive to the rear slant angle.