Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Contents lists available at ScienceDirect

Journal of Wind Engineering & Industrial Aerodynamics


journal homepage: www.elsevier.com/locate/jweia

Dispersion of ultrafine particles in the wake of car models: A wind


tunnel study
Romain Rodriguez a, Frederic Murzyn a, *, Amine Mehel b, Frederique Larrarte c
a
ESTACA West Campus, Department of Mechanical Engineering, Air Quality and Depollution Group, Rue Georges Charpak, 53000, Laval, France
b
ESTACA Paris Saclay Campus, Department of Mechanical Engineering, Air Quality and Depollution Group, Avenue Paul Delouvrier, 78066, Saint Quentin en Yvelines,
France
c
University Gustave Eiffel, IFSTTAR, Department of Geotechnics, Environment, Natural Risks and Earth Sciences, 14-20 Boulevard Newton, 77447, Marne La Vallee
Cedex 2, France

A R T I C L E I N F O A B S T R A C T

Keywords: Worldwide around 7 million annual deaths are due to air pollution. Among all pollutants, Ultrafine Particles
Dispersion (UFP) cause strong adverse effects. In this paper, the dispersion of UFP is studied in the wake of car models
Ultrafine particle characterized by different rear slant angles (ϕ). Velocities and UFP concentrations are collected in a wind tunnel.
Particle number concentration
The influence of the flow topology on the dispersion of these UFP is discussed. The results indicate that its
Ahmed body
Wake flow
structure strongly influences their dispersion. Whatever the rear slant angle is, the size of the recirculation region
Recirculation region is a key parameter governing the dispersion of these UFP. For ϕ ¼ 0 , the flow is almost symmetric and con-
Wind tunnel centration levels are higher and homogeneous in the close wake. The recirculation region is the largest one. The
LASER Doppler velocimetry dispersion is enhanced in both horizontal and spanwise directions. For ϕ ¼ 25 , the flow is attached on the rear
slant leading to a strong downwash effect. The volume of the recirculation region is the smallest. Longitudinal
vortices develop from the edges of car model entrapping particles. Particle Number Concentration field is no more
symmetric. For ϕ ¼ 35 , results are almost similar to those obtained for ϕ ¼ 0 . Comparisons with previous studies
are discussed and possible applications are suggested.

1. Introduction they are issued from agriculture, transports, industries, heating as well as
green waste burning … Particles can be classified according to their size
Air quality improvement is a key issue with important consequences as coarse particles (diameter between 2.5 μm and 10 μm), fine particles
in terms of public health and environmental issues. In France, the at- (diameter between 0.1 μm and 2.5 μm) and ultrafine particles (known as
mospheric pollution often reaches peaks above recommended thresholds UFP, diameter below 100 nm). Among all UFP emissions sources,
in big cities such as Paris, Lyon or Lille leading to traffic restrictions transportation systems contribute to up to 90% of the total emissions in
(Mehel et al., 2019). There are huge economic related challenges as well number (Mejia et al., 2007; Manigrasso et al., 2019). As it will be
as health concerns. As an example, in France, the annual cost of air explained below, these smallest particles are more toxic than the larger
pollution has been estimated to 101 Billions of Euros. This is twice as ones. They are not dangerous because of their mass but because of their
tobacco. Per year, 650000 days of work stoppage are recorded due to air number. Furthermore, they can infiltrate the car cabin increasing com-
pollution. At a larger scale, the economic shortfall in the world is about muter’s exposure (Airparif, 2007; Hudda et al., 2011; Hudda et al., 2012;
200 Billions of Euros (Daycard-Heid, 2019). It is worthwhile to note that Joodatnia et al., 2013; Knibbs and de Dear., 2010; Knibbs et al., 2010;
other consequences are linked with this issue such as global warming or Knibbs et al., 2011; Mehel et al., 2019; Morin et al., 2009; Polednik et al.,
fouling of buildings. Overall, it greatly impacts the total cost for hu- 2018; Xu and Zhu, 2013). It is well-known that Diesel engines play an
manity as well as for the environment. The surrounding air contains important role (ADEME, 2018) for primary particles and nitrogen oxide
gaseous pollutants (nitrogen oxides, sulfur dioxide or Volatile Organic emissions. There are other sources of emission such as brakes, tyre wear
Compounds (VOC) such as benzene) and particles. According to the and roadworks (dust). Such particles are also harmful because of their
French Environment and Energy Management Agency (ADEME, 2018), small sizes and their chemical nature. Nevertheless, they are out of the

* Corresponding author.
E-mail address: frederic.murzyn@estaca.fr (F. Murzyn).

https://doi.org/10.1016/j.jweia.2020.104109
Received 26 August 2019; Received in revised form 19 January 2020; Accepted 20 January 2020
Available online 31 January 2020
0167-6105/© 2020 Elsevier Ltd. All rights reserved.
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

scope of the present paper. For a Diesel engine, at the exit of the tailpipe, of Large Eddy Simulation (LES) to predict flow features (Tunay et al.,
the particle size distribution (PSD) exhibits two peaks. The main one is 2016). From wind tunnel investigations, the sensitivity to experimental
the first that corresponds to particles having a diameter around 10 nm conditions was discussed by Thacker et al. (2012). On the one hand, a
(nuclei mode). They are numerous but their contribution to the total drag discrepancy of 10% was depicted for the same rear slant angle (ϕ ¼
mass is weak. The second one is depicted around 200 nm (accumulation 25 ) depending on the flow separation on the rear slant (sharp versus
mode). In that size range, particles are less numerous but heavier. rounded edge). On the other hand, a lot of data (velocities and pressure)
In the past, it has been clearly demonstrated that UFP can reach the was collected and the main flow properties such as length of the recir-
alveolar region of the human lung with greater efficiency than larger culation region, the positions of vortices and their centres were clearly
particles. They can deposit in alveoli due to their rapid diffusion. They depicted for 0 <ϕ < 40 and different Reynolds numbers. Then, 2D and
can also damage pulmonary cells and enter blood and respiratory systems 3D maps of mean and turbulent flow properties are available in the
(Buzea et al., 2007). Recently, Valentino et al. (2016) stated that literature. Rodriguez (2018) presented a detailed review of the
“maternal exposure to diluted diesel engine exhaust alters placental state-of-the-art. Based on that, it has been shown that, depending on the
function and induces intergenerational effects in rabbits” meaning that rear slant angle (ϕ), different flow topologies can be generated repre-
UFP can cross placenta barrier exposing foetus. According to Buzea et al. senting real situations. Working with such car models is then relevant to
(2007), nanoparticles are able to contaminate any part of the human discuss UFP dispersion. Nevertheless, the correlation between PNC and
body including the brain. Air pollution is responsible for aggravation of flow topology has never been studied to date. Based on this statement, a
cardiopulmonary and respiratory diseases such as bronchitis, asthma, thorough work was undertaken for different shapes of vehicles.
lung cancer and exacerbation of allergies (Araujo et al., 2008; Delfino In this paper, an overview of results on that topic issued from wind
et al., 2005; Diaz-Sanchez et al., 2003; Manigrasso and Avino, 2012; tunnel investigations is addressed. First, the experimental facilities, the
Manigrasso et al., 2019; Pope et al., 2002; Silverman et al., 2012; Sioutas instrumentation, data acquisition modes and analysis methods are
et al., 2005; Tissot, 1999; Valberg, 2004; Verrier et al., 2002; Zweiman detailed. The three reduced scale and simplified car models (known as
et al., 1972). Furthermore, it can be responsible for other adverse effects Ahmed bodies) with different rear slant angles (ϕ ¼ 0 , 25 and 35 ) are
such as visibility reduction for drivers (known as particle fog) which presented. In the third section, the most significant results are exposed
increases the risk for accidents or a decline in agricultural yields (Bell for both velocity and PNC fields. Correlations between them are dis-
et al., 2004). As an example, the decrease of the latter point was assessed cussed. The fourth section aims at discussing our experimental approach
to be about 10% around Paris for wheat by INRA (Daycard-Heid, 2019). with some previous studies. In the last section, a conclusion suggests
In 2019, a new study revealed even more alarming conclusions (Lelieveld further studies that may be undertaken in a close future.
et al., 2019). The surrounding pollution may kill twice compared to
previous estimations. The worsening of air quality due to human activ- 2. Experimental facilities, instrumentation and data acquisition
ities may be responsible for about 8.8 millions of death per year in the and analysis
world, 2.8 millions of them being in China and 67000 in France (against
48000 commonly considered). In France, that means that air pollution 2.1. Wind tunnel tests
kills as much as tobacco (72000) and more than alcohol (49000).
Worldwide, previous data suggested around 4.5 millions of death. Measurements were conducted in the wind tunnel at ESTACA West
Furthermore, Lelieveld et al. (2019) attribute 40%–80% of untimely Campus in Laval. It is an open circuit tool manufactured by Deltalab. The
deaths to cardiovascular diseases linked to air pollution. test section is 1 m in length (Lwt) and 0.3 m in both width (Wwt) and
Having that in mind, it has become crucial to get a better under- height (Hwt). The maximum wind speed is Umax ¼ 40 m/s. A detailed and
standing of pollutant’s dynamics in the wake of a vehicle. In the present accurate calibration of the experimental facility with an empty test sec-
article, attention is drawn to particles emitted from the tailpipe. Once tion was performed before the present study (Rodriguez, 2018) and the
released in the atmosphere, they can either disperse in the surrounding most relevant properties of the incoming flow were assessed (Rodriguez,
environment or infiltrate the car cabin of the following vehicles 2018). That is the boundary layer (thickness and type), the velocity
increasing commuter’s exposure. Previous studies have shown that this gradient and the flow homogeneity. The most important results tend to
infiltration can be more or less important depending on ventilation set- indicate that the turbulence level was below 1% on the streamwise di-
tings (Mehel et al., 2019) but only a few were interested in the pollutant rection (out of the boundary layer). Furthermore, our boundary layer was
dispersion from their source (Richards, 2002; Gosse, 2005; Kanda et al., turbulent with a maximum thickness (δ) of 12 mm at the exit. Lastly, the
2006; Carpentieri et al., 2012; Mehel and Murzyn, 2015). Most of them velocity gradient was low (<4% over the whole length). Overall,
considered gaseous pollutant and passive scalar (Richards, 2002; Gosse, the conditions were partially developed. Fig. 1 presents the wind tunnel.
2005; Kanda et al., 2006; Carpentieri et al., 2012). Consequently, they The white arrow indicates the flow direction.
did not provide PNC fields that would be more relevant. Then, there is
still a lack of knowledge about real PNC exposure that prevents epide-
miologic studies to be reliable. This is the reason why it has been hard to
fix any recommendation about UFP exposure in the early 2000s ac-
cording to World Health Organization (WHO).
Mehel and Murzyn (2015) performed a preliminary study using UFP.
They measured PNC in the wake of simplified car model showing the
strong influence of the flow topology on the UFP dispersion. From a fluid
mechanics point of view, Gillieron and Kourta (2011) recalled that the
governing parameter of the flow developing in the wake of a car is the
rear slant angle. In the earlier 1980s, Ahmed et al. (1984) proposed a
simplified geometry of a vehicle called the Ahmed body. Since that, it has
been widely investigated both numerically (Corallo et al., 2015; Guil-
mineau, 2008; Tunay et al., 2016) and experimentally (Gosse, 2005;
Lienhart and Becker, 2003; Lienhart et al., 2002; Thacker et al., 2012;
Tunay et al., 2014, 2016; Rodriguez, 2018; Watkins and Vino, 2008).
From the numerical point of view, the influence of turbulence models, Fig. 1. Overview of the test section with LASER Doppler Velocimettry (LDV)
aspect ratio and stilts were discussed showing, for instance, the accuracy system and a car model.

2
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

2.2. Reduced-scale car models: Ahmed bodies

The flow regime associated with the wake dynamics is linked with the
drag force. This is particularly true for Ahmed bodies. Many studies
pointed out the relation between the rear part of the car model with the
structures of the wake flow (Leclerc, 2008; Thacker, 2010; Barros
(2015)). They mentioned that the drag due to the rear part of the car
represents almost 30% of the total drag. Considering the experimental
results giving the drag coefficient as a function of the rear slant angle (ϕ)
(Ahmed et al., 1984), three behaviors were clearly depicted:

- ϕ<~10 : the flow is 2D out of the walls and considered as square


back. The streamlines reveal a toric shape of the recirculation region.
In the centreline of the car model, two counter rotating vortices
develop in the close wake;
- ~10 <ϕ<~30 : the wake flow is defined as fastback. The flow is 3D
due to the interaction of the turbulent structures developing in the
wake. In this case, it is sometimes partially detached on the rear slant.
Longitudinal vortices appear coming from the lateral sides of the
vehicle. They lead to an increase of the drag, this drag being
maximum for ϕ~30 . In the close wake the toric shape of the recir- Fig. 2. Sketches of the car models: below, rear and side views (lengths in mm).
culation region is still observed. Its length is smaller compared to that
measured for ϕ < 10 . Overall, the structure of the wake flow is much assessed in the context of air quality in transportation systems. This issue
more complex compared to the first case. The sensitivity to experi- is more crucial in cities where people are mainly exposed to air pollution.
mental conditions is also more effective; As a consequence, the flow developing in the wake of a car in an urban
- ϕ>~30 : the situation is closed to that observed for ϕ<~10 . The city environment was simulated. Then, according to French regulations,
flow is then considered as square back. After reaching its peak, the the speed of the incoming flow must be 50 km/h (13.9 m/s), which is the
drag suddenly decreases (drag crisis) and recovers a level closed to speed limit in cities in France. Considering the size of the wind tunnel,
that measured for ϕ ¼ 0 . On the rear slant, the flow is fully detached. the Reynolds similitude can not be achieved unless working with unre-
Out of the walls, it can be considered as a 2D flow. alistic velocities. Then, to select this upstream velocity, the ratio between
the speed of the car (U0) and the velocity of the exhausted particles at the
In the present experiments, these three flow topologies were exam- end of the tailpipe (Utp) was kept constant. Equation (2) gives this
ined. Then, measurements were conducted in the wake of three Ahmed relation.
bodies having rear slant angles ϕ given by ϕ ¼ 0 , 25 and 35 . Scale of    
U0 U0
models was 0.19 with respect to the original one developed by Ahmed ¼ (2)
et al. (1984). In terms of dimensions, they were 0.196 m in length (L), Utp wind tunnel Utp real situation

0.054 m in height (h) and 0.073 m in width (l). The models were fixed on
Here, Utp is the ratio between the flow rate (m3/s) at the tailpipe exit and
the floor of the test section by a cylindrical rod (diameter 5 mm) and four
its corresponding cross section (m2).
stilts. The height (hs) and diameter (ds) of these stilts were 15 mm and 6
According to Comite des Constructeurs Français d’Automobiles
mm, respectively. This height was chosen to be larger than the boundary
(2017), a representative car has an average stroke volume Vcyl ¼ 1.486 L.
layer thickness. Accordingly, the dimensionless ground clearance was Hs
Following Roberge et al. (2006), the flow rate at the tailpipe Qtailpipe,
¼ hs/h ¼ 0.28. To avoid any correction of wall effects, the models were
vehicle is given by Equation (3):
designed so that the blockage coefficient was below 5% (West and Apelt,
1982). This coefficient is given by the ratio between the frontal area of 1
the car and the area of the test section (Wang et al., 2013) (Eq. (1)). Qtailpipe; vehicle ¼ eVcyl Ωm (3)
2
h*l Where e is the efficiency of the engine and Ωm the speed of the engine.
B¼ (1)
Wwt *Hwt According to Heywood (1988) and Hancke (2009), e ¼ 0.9 and Ωm ¼
Three sketches of the models with below, rear and side views are 2200 RPM (Rotation Per Minute) for a vehicle in an urban environment.
shown on Fig. 2. Considering a tailpipe diameter of 0.055 m, then Qtailpipe, vehicle is 0.0245
O (0, 0, 0) is the origin of the coordinate system. It is located at the m3/s. This gives Utp ¼ 10.3 m/s and U0/Utp~1.35 for real conditions.
bottom of the rear face of the car and on the centreline of the wind tunnel. Taking into account all relevant parameters of the experimental facilities
x is the streamwise direction (positive downstream), y is the vertical (for instance the diameter of the tailpipe in the wind tunnel and the
direction (positive upwards) and z is the spanwise direction (positive characteristics of the particle generator detailed in section 2.4.2), the
from right to left when looking from the back of the vehicle). The tailpipe incoming air flow velocity was constant and fixed to the value U0 ¼ 14.3
has an outer/inner diameters of 6 mm and 4 mm, respectively. These m/s (Rodriguez, 2018). This ensures U0/Utp~1.35 for the experiments.
settings were chosen taking into account the boundary layer thickness Compared to previous studies (Richards, 2002; Gosse, 2005; Kanda et al.,
and the averaged velocity of the UFP at the exit of the tailpipe. The exit of 2006; Carpentieri et al., 2012), this is a novelty.
the tailpipe (center) is located at X ¼ x/h ¼ 0, Y ¼ y/h ¼ 0.06 and Z ¼ Finally, the corresponding Reynolds number based on the height of
z/h ¼ 0.31. One can refer to Rodriguez (2018) for more details. the car model (Eq. (4)) was 49500.

U0 *h
Re ¼ (4)
2.3. Determination of experimental conditions υ

Where ν is the kinematic viscosity of air.


As presented above, the goal of the present study was to analyse It is acknowledged that the present Reynolds number is one order of
particle dispersion in the wake of car models. This phenomenom was

3
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

magnitude lower than it would be in real situation. Nevertheless, Barros the highest level to ensure highest SNR (Signal to Noise Ratio) for PNC
(2015), Lahaye (2014), Gosse (2005) and Ahmed et al. (1984) indicated measurements. This gave a mass flow rate of particles of about 6.5 mg/h.
that the main flow structures remain the same for a large range of Rey- Similarly, the nitrogen flow rate was set at 8 L/min. Note that the
nolds numbers as well as the velocity profiles in the wake. The recircu- recommendation of the manufacturer at that spark frequency was 6.5 L/
lation region was also quite stable in these studies with Reynolds min (minimum). Taking into account the inner diameter of the exhaust
numbers up to 106. This was confirmed by comparing the present find- pipe (4 mm), the exit velocity of the particles was Utp ¼ 10.6 m/s.
ings with previous results covering a large range of Reynolds numbers
show similar trends, making the present study meaningful. Lastly, for ϕ 2.4.3. Electrical low pressure impactor (ELPI)
¼ 25 , Leclerc (2008) showed the same behavior for the pressure dis- For PNC measurements (number of UFP per cm3 for each given po-
tributions on the rear slant for 67000<Re < 670000. sition), an ELPI was used. This device enabled measurement of real-time
PSD and concentration in the size range of 6 nm–10 μm with 1 Hz
2.4. Instrumentation sampling rate. It has been developed by the University of Tampere
(Finland). Its features include real-time stand-alone operation, wide
2.4.1. LASER Doppler Velocimetry (LDV) sample concentration range, wide particle size range and robust structure
Velocity measurements were recorded with a 2D LDV system manu- for operation even in harsh conditions. Considering all features, the use
factured by DANTEC dynamics (model 2D Flow Explorer). The two pairs of the ELPI was well-suited for our measurements. It was previously used
of LASER beams have wavelengths of 660 nm and 785 nm. Fringe (Mehel and Murzyn, 2015) showing its satisfying performances. Overall,
spacings were 5.45 μm and 6.40 μm in the longitudinal and vertical di- the operating principle can be divided into three major parts:
rections, respectively. Diameter and length of the measuring volume in
the z-direction are 168 μm and 2.81 mm respectively for the first 1) Particle charging;
component (horizontal) and 200 μm and 3.34 mm for the second 2) Size classification throughout a cascade impactor;
component (vertical). The focal length is 500 mm and Bragg cell fre- 3) Electrical detection by sensitive electrometers.
quency shift is 80 MHz. The LDV system was fixed on a 2D displacement
table. It was controlled by BSA Flow Software v5.03.00. The fog used in The particles were first charged into a known charge level in the
this study for seeding was the SAFEX inside Nebelfluid Extra Clean Corona charger. After that, the particles entered a cascade low-pressure
provided by DANTEC. The fog generator model was SAFEX S 195 G. It impactor with 14 electrically insulated collection stages. The particles
had an adjustable flow rate. The mean diameter of the generated droplets were collected in the different impactor stages according to their inertia
was 1.068 μm. According to Algieri et al. (2005), Rodriguez et al. (2018) (aerodynamic diameter). The electric charge carried by particles into
and Rodriguez et al. (2019), the accuracy of the measurements is then each impactor stage was then measured in real time by sensitive elec-
ensured with respect to the present experimental conditions. Note that trometers. This current signal was directly proportional to the particle
the lowest point for LDV measurements was 20 mm above the bottom. number concentration per stage (size range). Measured current signals
Below, the optical path of the LASER beams associated with the vertical were converted to particle size distribution using particle size dependent
component of the velocity was blocked by this bottom. relations describing the properties of the charger and the impactor stages.
The result was particle number concentration and size distribution in
2.4.2. PALAS real-time. For more details, one can refer to Rodriguez (2018). In order to
Particles were generated by a PALAS DNP 2000. It is a nano-scale test satisfy the isokinetic condition, the velocity of the aerosol suction must
generator of aerosols from monolithic graphite. The resulting carbon be equal to the flow velocity at the measuring point to avoid any diver-
agglomerates are similar to Diesel soots (Evans et al., 2003) with respect gence of the streamlines. Nevertheless, this could not be ensured during
to particle size distribution (Fig. 1). This system required nitrogen as the experimentations. So, based on the upstream velocity, this condition led
carrier gas. This gas caused virtually no change in the density of the to a sampling probe diameter of ~3.86 mm. Consequently, a diameter of
exhaust gas being measured. From a technical point of view, PALAS was 4 mm for the sampling probe was chosen. At this stage, for experimental
used to generate a jump spark between two graphite electrodes under constraints, the sampling probe was bended. It is worthwhile to note that
high voltage. It then ripped tiny amount of graphite material from the two models of such probes were used with lengths of 25 and 50 mm. A
electrodes at high temperatures. The graphite material that was vapor- preliminary calibration study proved the robustness of this choice
ized by this spark then condensed to form extremely tiny particles. The (Rodriguez, 2018). Fig. 3 shows the devices used for PNC measurements
high number concentration can result in the coagulation of these very as well as and the experimental arrangement. Similarly, the lowest point
small particles into agglomerates. By adding mixed air, the aerosol was for PNC measurements was 13 mm above the bottom due to the bending
able to be diluted, enabling the defined adjustment of the agglomerate of the sampling probe.
formation. The energy converted in each spark remained constant due to
the constant sparkover voltage. This constant energy in each individual
spark guaranteed stable particle size distribution. A technically sophis-
ticated control of the distance between the electrodes during burn-off
ensured very high long-term stability. The mass flow rate was quickly
and easily adjusted within a wide range using the spark frequency. The
digital regulation of the frequency and the continuous regulation of the
voltage guaranteed a more specific regulation of the distance between
the two electrodes. This enabled a higher constancy of the particle size
distribution and the mass flow. Due to its excellent reproducibility and
high level of functional reliability, the DNP digital 2000 was especially
well suited for our experiments (Evans et al., 2003; Oberd€ orster et al.,
2004; Price et al., 2014; Mehel and Murzyn, 2015). Indeed, as mentioned
above, the generated aerosol distribution was very similar to the distri-
bution of Diesel soot particles from a combustion engine. In terms of PSD,
a preliminary study showed that more than 93% of the particles have a
diameter between 30 and 109 nm (Rodriguez, 2018).
For the present study, the spark frequency was set at 200 Hz. This is Fig. 3. Experimental arrangement for ELPI and PALAS during measurements.

4
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

2.5. Data acquisition and analysis the spanwise direction, 5 positions were studied (Z ¼ 0.62, 0.31, 0,
0.31 and 0.62) for X < 2.89. For X > 2.89, 7 positions were studied (Z ¼
Preliminary studies were undertaken as a calibration work. First, the 1.85, 1.24, 0.62, 0, 0.62, 1.24 and 1.85). This was done to cover a
convergence of the LDV data and their statistical accuracy were ensured. wider area to get a better assessment of the dispersion in the far field.
Based on it, measurements lasted 90 s with 3000 samples as a minimum Overall, PNC were measured for 1926 positions for the three car models.
value (Rodriguez, 2018). Measurements took place in the wake of the Fig. 5a (side view) and Fig. 5b (top view) show the measuring points for
three car models. The covered domain spread as follow: 0.09<X ¼ x/h < PNC.
5.65, 0.27<Y ¼ y/h < 1.3 and Z ¼ z/h ¼ 0, 0.23, 0.45 and 0.68
where X, Y and Z are the dimensionless distance in the x,y and z direction 3. Results
respectively (X ¼ x/h, Y ¼ y/h and Z ¼ z/h). For ϕ ¼ 25 and 35 ,
measurements were also conducted above the rear slant. The character- In this section, both mean and turbulent properties of the wake for the
ization of the flow detachment or attachment on it was then possible. In three car models are of interest. First, an overview of the mean flows is
terms of mesh size, the step between 2 points was 10 mm in the presented. In a second time, the fluctuating part of the flows is depicted.
streamwise direction and 5 mm in the vertical direction for 0.09<X < U corresponds to the horizontal component of the velocity vector while V
1.94. For 1.94<X < 5.65, it was set to 50 mm in the streamwise direction is the vertical one. u’ and v’ are the corresponding RMS values.
and 5 mm in the vertical one, respectively. For a given (xy) 2D map, 195
points were considered (out of the boundary layer and the rear slant). 3.1. Overview of the wake flows
Fig. 4a (side view) and Fig. 4b (rear view) show these measuring points
for LDV. For more information, one can refer to Rodriguez (2018). For Firstly, the results dealing with the mean flow characteristics are
the three car models, the total amount of measuring points for the LDV presented according to the rear slant angles. Fig. 6 is dedicated to the 2D
part was nearly 5800 including rear slants, wakes and boundary layers. maps of the velocity vectors in the wake of the 3 models. The dimen-
Regarding the data treatment, the Interarrival Time and Transit Time sionless modulus of the velocity is shown as a color map. Same scales are
(ITTT) method was applied (Rodriguez, 2018; Rodriguez et al., 2018). used to make the comparisons easier.
This new technique was developed to take into account the specificities At this stage, the tailpipe is not installed. Nevertheless, its influence as
of our experimental setup and arrangement. It particularly tackled the well as that of the exhaust flow will be discussed later. These preliminary
issue of non-homogeneous seeding and ensured reliable results. results are mandatory to validate the whole experimental methodology
For PNC measurements, based on a preliminary study, data acquisi- by comparison with the existing data available in the literature.
tion lasted 80 s (Rodriguez, 2018). Over this period of time, three steps First, the recirculation length (Lrec) is considered. It is defined as the
were defined. The first one corresponded to the acquisition of the largest distance between the rear face of the car and the position for
ambient level. The duration of this phase was about 6 s. No particle was which a negative component of the horizontal velocity (U) is detected.
injected to get access to the ambient level. Then, as a second stage, The results are indicated in Table 1. This is in agreement with previous
particles are emitted from the tailpipe. PNC increased for over 5–10 s. findings (Tunay et al., 2014; Lahaye, 2014; Wang et al., 2013) for rela-
Then, a stabilized phase was observed which lasted at least 50 s. During tively similar experimental conditions.
this stage, levels were almost constant. Nevertheless, a filter was applied Furthermore, the existence of two contra-rotating vortices for each
for the data analysis: data above  3 Root Mean Square (RMS) from the case is pointed out. Depending on the rear slant angle, they are more or
mean level were eliminated. Then, data corresponding to the ambient less developed. The upper one is clockwise while the lower one is counter
level were substracted to the ones corresponding to the stabilized phase. clockwise. Their positions are given in Table 2.
All details regarding the conversion current/concentration levels and the For the case ϕ ¼ 0 (Fig. 6, (a)), the properties of U and V are
data treatment method are given in Rodriguez (2018). PNC were recor- respectively almost symmetric and antisymmetric according to Y ¼ 0.5.
ded using two sampling bent probes. They had lengths of 25 and 50 mm, For ϕ ¼ 25 (Fig. 6, (b)), the general trend of the flow is pointing
respectively. Regarding the mesh size for the measurements, it was downward except in the closest part of the rear face of the car model due
different depending on the position (near wake versus far wake). For to the presence of the lower vortex. The attachment of the flow on the
0.57<X < 2.89 (near-wake), it was 25 mm in the streamwise direction rear slant is observed for this configuration. As a consequence, the ver-
and 5 mm in the vertical direction starting just below (2 mm) the level of tical component of the velocity is mostly negative. Note that the position
the tailpipe. Between X ¼ 2.89 and X ¼ 7.06 (far wake), it was set to 25 of the lower vortex was not possible as velocity measurements were not
mm in the horizontal direction and to 10 mm in the vertical direction. In allowed closed to the bottom of the wind tunnel for technical reasons. For

Fig. 4. a. Measuring points for LDV in the wake of the car models (side view). b. Measuring points for LDV in the wake of the car models (rear view).

5
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Fig. 5. a. Measuring points for PNC in the wake of the car models (side view). b. Measuring points for PNC in the wake of the car models (top view).

Table 1
Comparison of recirculation length with the literature for similar experimental
conditions.
Lrec/h ϕ ¼ 0 ϕ ¼ 25 ϕ ¼ 35

Present study 1.39 0.58 1.06


Literature 1.50 (Lahaye, 0.57 (Tunay et al., 1.07 (Wang et al.,
2014) 2014) 2013)
Deviation from the 7% 2% 1%
literature

Table 2
Comparison of the positions (X, Y, Z) of upper and lower vortices in the symmetry
plane Z ¼ 0 with the literature for similar experimental conditions (N/A: Not
Available).
Positions of ϕ ¼ 0 ϕ ¼ 25 ϕ ¼ 35
vortices

Present study Upper: (0.62, Upper: (0.19, Upper: (0.21,


0.84, 0) 0.37, 0) 0.70, 0)
Lower: (0.70, Lower: N/A Lower: N/A
0.14, 0)
Literature Upper: (0.93, Upper: (0.17, Upper: (0.28,
0.84, 0) 0.30, 0) 0.69, 0)
Fig. 6. 2D maps (XY) of the velocity vectors and dimensionless velocity Lower: (0.68, Lower: (0.35, Lower: (0.44,
magnitude in the wake of the 3 car models (a: ϕ ¼ 0 , b: ϕ ¼ 25 , c: ϕ ¼ 35 ) in 0.20, 0) 0.05, 0) 0.11, 0)
the symmetry plane Z ¼ 0. Lahaye (2014) Tunay et al. Tunay et al.
(2014) (2014)

ϕ ¼ 35 (Fig. 6, (c)), the flow on the rear slant is detached. As a conse-
quence, the downwash effect of the flow is less important compared to ϕ (6), respectively.
¼ 25 . As for ϕ ¼ 25 , the position of the lower vortex was not deter-
mined. For comparison, results from previous studies are also provided in 1 u’2 þ v’2
TKE* ¼ (5)
Table 2 showing good agreements. 2 U 20
In terms of fluctuations, the dimensionless Turbulent Kinetic Energy
(TKE*, Eq. (5)) and Reynolds Shear Stresses (SS*, Eq. (6)) are presented u’v’
on Fig. 7 and Fig. 8. In 2D, they are given by Equation (5) and Equation SS* ¼ (6)
U 20

6
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Fig. 7. 2D maps (XY) of TKE* in the wake of the 3 car models in the symmetry
plane Z ¼ 0 (a: ϕ ¼ 0 , b: ϕ ¼ 25 and c: ϕ ¼ 35 ). Fig. 8. 2D maps (XY) of SS* in the wake of the 3 car models in the symmetry
plane Z ¼ 0 (a: ϕ ¼ 0 , b: ϕ ¼ 25 and c: ϕ ¼ 35 ).
The dotted lines on Figs. 7 and 8 correspond to the boundary of the
recirculation region while A and B account for the center positions of the On the other way, SS are positive in the lowest part of the flow where both
upper and lower vortices, respectively. A and B are determined from the U and V are positive. For ϕ ¼ 0 , SS were almost symmetric with respect to
streamlines deduced from the velocity measurements. In the following Y ¼ 0.5 until X~3. In terms of magnitude, they are slightly larger in the
figures, large black dotted lines refer to the boundary of the recirculation upper part compared to the lower one. For ϕ ¼ 25 , high shear stress re-
region obtained from velocity measurements while tiny black dotted gions were located in the lower part of the flow. It shows that the flow
lines refer to the boundary of the recirculation region deduced from flow seems much more unstable in this lower region for this configuration.
symmetry. In some cases, B can not be determined for technical reasons Nevertheless, this 2D map is not complete for the closest part of the bot-
(too close to the bottom of the wind tunnel). tom. This configuration was associated with the smallest recirculation
These plots provide some useful information regarding the flow dy- region. For ϕ ¼ 35 , this was an intermediate situation between ϕ ¼ 0 and
namics. In the literature, it is often admitted that an increase in the ϕ ¼ 25 . The largest values are found in the lower region.
Reynolds Shear Stresses (SS) leads to an increase of Turbulent Kinetic Table 3 details all these information.
Energy (TKE). In other words, high gradient regions for turbulent kinetic Finally, this preliminary study allowed us to validate our experi-
energy often correspond to high shear stress regions. According to mental setup. Indeed, these results were in good agreements with some
Lienhart et al. (2002, 2003), peaks of TKE are located at the same posi- previous studies in similar experimental conditions. Two different re-
tions as peaks of turbulent shear stresses. Overall, these results were in gions were clearly identified. One accounts for the near wake (0<X<~3,
agreement with those findings. For ϕ ¼ 0 , high TKE levels are found in depending on the rear slant angle) and a second one which corresponds
the upper half (Y > 0.50) over a relatively large area. For ϕ ¼ 25 and 35 , to the far wake. In the latter one, the flow gradually recovered its initial
highest values are mostly in the lower half (Y < 0.50) and spread over a (upstream) state. Furthermore, one must notice that some complemen-
relatively smaller area. For ϕ ¼ 0 , 25 and 35 , in the symmetry plane (Z tary measurements highlight important results. On the one hand, the
¼ 0), the corresponding values of the peaks of TKE are given in Table 3. In properties of the flow discussed above are almost symmetric with respect
terms of magnitude, the corresponding maximum turbulence intensities to Z ¼ 0 (Rodriguez, 2018). On the other hand, Particle Imagery Veloc-
(Ix ¼ u’/U0) are 29%, 29% and 31%, respectively. Similar conclusions imetry (PIV) measurements realized at the University of Rouen (France)
were revealed for the vertical component of the velocity (Iy ¼ v’/U0). in similar experimental conditions (same models and Reynolds number)
Overall, these results are in good agreement with those of Tunay et al. provided access to the 3rd component of the velocity vector. Then, the
(2014) for ϕ ¼ 25 and 35 . presence of two counter-rotating vortices in the horizontal plane was
The shear stresses represent the correlation between the fluctuations of demonstrated meaning that the structure of the recirculation region in
the horizontal and vertical components of the velocity vector. On the one the wake of the models was toric (Rodriguez, 2018). Longitudinal
hand, SS are negative in the upper part of the wake flow where the vertical vortices were depicted for ϕ ¼ 25 (Rodriguez et al., 2018) but not for ϕ
component V was negative and the horizontal component U was positive. ¼ 0 and 35 .

Table 3
Properties of TKE* and SS* with the corresponding positions (X, Y, Z) in the wake of the 3 car models.
Rear slant angle ϕ ¼ 0 ϕ ¼ 25 ϕ ¼ 35

TKE*max Position 0.07 (0.99, 0.83, 0) 0.07 (0.26, 0.09, 0) 0.08 (0.49, 0.09, 0)
SS*max Position 0.026 (1.05, 0.18, 0) 0.035 (0.32, 0.09, 0) 0.044 (0.82, 0.09, 0)
SS*min Position 0.043 (1.22, 0.83, 0) 0.021 (0.65, 0.36, 0) 0.020 (0.05, 0.93, 0)

7
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

3.2. Influence of the tailpipe system and exhaust flow on the wake flow
properties

Before injecting particles, a study was conducted to analyse the in-


fluence of the tailpipe system and the exhaust flow on the wake flow
properties. This is new and innovative as the previous studies compared
mass concentration fields with flow topologies without tailpipe. Fig. 9
and Fig. 10 show the results with (dots) and without (lines) the tailpipe
and the exhaust flow for ϕ ¼ 25 for 2 different vertical planes Z (Z ¼
0.23 in blue and Z ¼ 0.45 in red). Fig. 9 is for U (mean) and Fig. 10 is
for u’ (RMS of U).
These figures indicate that the results for both mean and RMS values
for the horizontal component of the velocity are similar with and without Fig. 10. Comparison of the RMS horizontal velocity (u’) with (WT) and without
the tailpipe. It means that the presence of the exhaust system does not (WOT) the tailpipe system for Z ¼ 0.23 and Z ¼ 0.45 (ϕ ¼ 25 ).
significantly affect the mean and turbulent flow dynamics in the wake of
this car model. It can be explained by the fact that the velocity difference
between the upstream flow and the exhaust one is small. Then, the deficit
of linear momentum of the latter one is easily counterbalanced by the
surrounding underbody upcoming flow. It is important to note that
comparable results were obtained for ϕ ¼ 0 and 35 for mean and RMS
values of U and V. Precisely, for the mean velocities, the difference be-
tween with and without tailpipe are below 0,04U0 while it only reaches
0,02U0 for the RMS values.

3.3. UFP concentration in the wake of the Ahmed body (ϕ ¼ 0 )

Measurements were collected in a wide 3D domain downstream of


the three car models. In this section, attention is focused on 2D maps of
PNC in the wake of the 3 car models. Dimensionless PNC (PNC*, Eq. (7))
are presented. All concentrations were divided by a reference level (Cref)
which corresponds to the PNC measured at the exit of the tailpipe
(Cref~6.96.107 part/cm3). That is:

PNC PNC
PNC* ¼ ¼ (7)
Cref 6:96*107

For the next figures, black dots correspond to the positions where
measurements were done.

a) Case ϕ ¼ 0

Fig. 11 show the results in three different vertical planes (XY) for the
squared back model (Z ¼ 0.31, 0 and 0.31). The first one (Z ¼ 0.31)
was aligned with the tailpipe while the second one (Z ¼ 0) corresponded
to the symmetry plane. The highest levels of dimensionless PNC are
found for Z ¼ 0.31. In this case, measurements were aligned with the
tailpipe exit. High levels of PNC* were measured over a wide area in both Fig. 11. 2D maps (XY) of PNC* in the wake of the Ahmed body (ϕ ¼ 0 ) for Z ¼
streamwise and vertical directions. The peak for PNC* (~0.33) was 0.31 (a: tailpipe plane), Z ¼ 0 (b: symmetry plane) and Z ¼ 0.31 (c).
measured at (X ¼ 0.57; Y ¼ 0.04; Z ¼ 0.31). That is one third of the
level measured at the tailpipe exit. As X increases, PNC* decreases down 0.01 spreads up to y/h ¼ 0.90 even if ejection occurs at Y < 0. This was
to 0.1 (X ¼ 1.27) and 0.01 (X ¼ 3.46). In the vertical direction, PNC* ¼ almost the full height of the vehicle. It can be explained by the flow to-
pology described in the previous section. As soon as they are ejected from
the tailpipe, the particles are entrapped within the recirculation region
which is the largest for this configuration (Lrec/h ¼ 1.39). Their motions
are mostly governed by the two vortices depicted by the LDV measure-
ments in the close wake. Overall, it is still obvious that most of the par-
ticles are found in the lower part of the flow for Z ¼ 0.31. For Z ¼
0 (symmetry plane), the PNC* levels are significantly lower compared to
those measured for Z ¼ 0.31. PNC* is always below 0.10. Nevertheless,
a homogeneous distribution of PNC* is depicted for Z ¼ 0 over the total
height of the car model. In the close wake, PNC* reaches 0.03 and levels
were between 0.01 and 0.10 up to X ¼ 3.95. The toric shape of the tur-
bulent structure appearing in the recirculation region explains this
finding (Rodriguez, 2018). The combination of the lower and upper
Fig. 9. Comparison of the mean horizontal velocity (U) with (WT) and without vortices was capable of capturing particles and carrying them from the
(WOT) the tailpipe system for Z ¼ 0.23 and Z ¼ 0.45 (ϕ ¼ 25 ). bottom to the upper part of the flow. Similarly, the horizontal turbulent

8
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

structures disperse the particles in the spanwise direction. Overall, for Z Fig. 13 present three vertical maps of PNC* corresponding to the same
¼ 0, the PNC* map is almost symmetric with respect to Y ¼ 0.50. This dimensionless positions Z ¼ 0.31, 0 and 0.31 for the case ϕ ¼ 25 .
was comparable to the velocity field given for the same dimensionless As seen for ϕ ¼ 0 , the highest levels for PNC* are measured for Z ¼
position Z. Lastly, for Z ¼ 0.31, PNC* levels are significantly lower 0.31 (PNC*max ¼ 0.14 at X ¼ 0.57, Y ¼ 0.06 and Z ¼ 0.31). PNC* are
compared to those depicted for Z ¼ 0.31 and 0. Practically, PNC*<0.01 larger than 0.01 up to X ¼ 3.20. This is a shorter distance compared to ϕ
for X>0.80. Nevertheless, some particles are still conveyed in the span- ¼ 0 . The vertical spreading is weaker as well: for Y < 0.34, PNC* does
wise direction by the turbulent structures of the flow. Finally, not exceed 0.01. Above, levels are even smaller. One more time, this is
PNC*>0.001 everywhere in the 2D map. Lastly, the vertical mixing is caused by the flow topology downstream of the car model: the recircu-
efficient, as the corresponding 2D velocity map tended to be symmetric. lation region was less volumic (Lrec/h ¼ 0.58) as the flow was attached on
For a full description of the dispersion of the particles in the wake of the rear slant leading to a more pronounced downwash effect compared
the car model, additional 2D horizontal maps (XZ) are presented for to ϕ ¼ 0 . Similarly, PNC* are between 0.001 and 0.01 up to X ¼ 2.50.
different dimensionless distances above the bottom of the wind tunnel Lastly, for this flow configuration, longitudinal vortices arise from the
(Fig. 12). That is for Y ¼ 0.04, 0.33 and 0.70. Y ¼ 0.04 is located in edges of the car model and develop downstream. Out of the recirculation
the vicinity of the tailpipe exit which is less than 2 mm below. For Y ¼ region, they become predominant. Then, they catch and carry particles
0.04, there is no black dotted line as there is no recirculation in this part away from the symmetry plane of the model, in the same side of the
of the flow. tailpipe. As a consequence, PNC* levels are very low for Z ¼ 0 and 0.31.
The highest levels are obviously measured for Y ¼ 0.04 and Z ¼ This is a typical feature associated with this rear slant angle. For Z ¼ 0,
0.31, aligned with the tailpipe exit. Up to X ¼ 1.60, PNC* are larger PNC* are always below 0.01. The downwash effect is still strong and the
than 0.01. In this case, most of particles are kept in the half width maximum values for PNC* are always found in the lower part of the flow
(tailpipe side) of the body. Above (Y ¼ 0.33), in the recirculation region, for X < 2. Further downstream, PNC* increases again to reach more than
horizontal mixing occurred and concentration homogenization was then 0.01 for Y ¼ 0.40. Then, the distribution of PNC* is not homogeneous for
strongly enhanced. As a consequence, the PNC* field became more Z ¼ 0 while it was the case for ϕ ¼ 0 . For Z ¼ 0.31, PNC* is always below
symmetric with respect to Z ¼ 0 in the far wake. It was then the mixing 0.001 even in the close wake. This behavior differs from that observed for
caused by the recirculation region that governed the particle dynamics in ϕ ¼ 0 . For ϕ ¼ 25 , the longitudinal vortices tend to play a key role in
the close wake. The same results were observed for Y ¼ 0.70. Particles are particle entrapment.
brought in the upper part of the flow by the recirculation region and the Fig. 14 show 2D (XZ) maps of the PNC* for ϕ ¼ 25 for 3 dimen-
horizontal mixing led to a symmetric field of PNC* with respect to Z ¼ 0. sionless positions Y. As for ϕ ¼ 0 , for Y ¼ 0.04, there is no black dotted
line as there is no recirculation in this part of the flow. Similarly, for Y ¼
(b) Case ϕ ¼ 25 0.70, it is above the recirculation region. So there is not a black dotted
line.
From the results corresponding to Y ¼ 0.04, it is confirmed that
longitudinal vortices influence the particle dynamics. Most of the UFP

Fig. 12. 2D maps (XZ) of PNC* in the wake of the Ahmed body (ϕ ¼ 0 ) for Y ¼ Fig. 13. 2D maps (XY) of PNC* in the wake of the Ahmed body (ϕ ¼ 25 ) for Z
0.04 (a: tailpipe level), Y ¼ 0.33 (b) and Z ¼ 0.70 (c). ¼ 0.31 (a: tailpipe plane), Z ¼ 0 (b: symmetry plane) and Z ¼ 0.31 (c).

9
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Fig. 15. 2D maps (XY) of PNC* in the wake of the Ahmed body (ϕ ¼ 35 ) for Z
¼ 0.31 (a: tailpipe plane), Z ¼ 0 (b: symmetry plane) and Z ¼ 0.31 (c).

corresponding to this position (Z ¼ 0.31) exhibit a strong correlation


Fig. 14. 2D maps (XZ) of PNC* in the wake of the Ahmed body (ϕ ¼ 25 ) for Y with those presented for ϕ ¼ 0 at the same dimensionless position Z. Up to
¼ 0.04 (a: tailpipe level), Y ¼ 0.33 (b) and Z ¼ 0.70 (c). X ¼ 0.88 (3.07, respectively), PNC* are above 0.10 (0.01, respectively).
The dispersion is also important in the vertical direction (PNC*>0.01 up to
emitted from the tailpipe are entrapped in the half width of the tailpipe Y ¼ 0.66 in the close wake). These findings confirm that the size of the
because of the longitudinal vortex which arises from the edge of the car recirculation region is of primary importance regarding particle disper-
model. Particle paths are diverted from the symmetry plane (Z ¼ 0) to- sion. Furthermore, due to the detached flow over the rear slant, there is no
wards the external part of the flow (Z < 0). Furthermore, at this more strong downwash effect compared to the case ϕ ¼ 25 . For Z ¼
dimensionless distance above the bottom, the recirculation region is not 0 (symmetry plane), PNC* are below 0.1. Nevertheless, the distribution
crossed yet. On the opposite side (Z > 0), PNC* are very low. The mixing seems to be homogeneous over the height of the car model in the close
on the spanwise direction can not take place. Above the tailpipe level (Z wake. This is due to the size of the recirculation (Lrec/h ¼ 1.06) region
¼ 0.33), the PNC* field is still not symmetric at all while it was for ϕ ¼ 0 . which is larger than that measured for ϕ ¼ 25 (Lrec/h ¼ 0.58). From X ¼
This result confirms the combined influence of the longitudinal vortices 2.23, PNC* decrease below 0.01. Far downstream, the homogeneity of
above mentioned and the very small recirculation region due to the PNC* is lost in the vertical direction and particles tend to accumulate in the
downwash effect. Furthermore, at this stage, the recirculation region is lower part (Y < 0.50). For Z ¼ 0.31, lower levels are measured
not large enough to catch particles and drive them on the other side of the (PNC*<0.01). Nevertheless, the recirculation region allows a mixing in the
car model (Z > 0): the length of the recirculation region was only 0.58 h horizontal direction in the close wake.
for ϕ ¼ 25 (it was 1.39 h for ϕ ¼ 0 ). For Z > 0, PNC* is less than 0.001 Fig. 16 exhibit 2D horizontal maps (XZ) for 3 different dimensionless
while it reaches 0.1 for Z < 0. Z ¼ 0 can then be associated with a virtual distances above the bottom of the wind tunnel (Y ¼ 0.04, 0.33 and
boundary between the longitudinal vortices emitted from the right and 0.70). Once again, for Y ¼ 0.04, there is no black dotted line as there is
left edges of the car model, respectively. In the horizontal map (Y ¼ no recirculation in this part of the flow. At the tailpipe level (Y ¼ 0.04),
0.70), very low levels are measured. This is mostly due to the attachment concentrations are relatively high. For Z < 0, PNC* are larger than 0.1 for
of the flow on the rear slant. For Z > 0, above the recirculation region, X < 2 while they are roughly 0.001 for Z > 0. Beyond X ¼ 2, differences
mixing can not occur and particles can not be conveyed from the tailpipe are weaker as PNC* levels are about 0.01. In this case, the PNC* distri-
to this upper region. bution is not strictly symmetric with respect to Z ¼ 0. More particles are
found for Z < 0 (tailpipe side). Above (Z ¼ 0.33), the recirculation area is
c) Case ϕ ¼ 35 crossed. This enables the enhancement of the horizontal mixing even if
PNC* is still three times larger in the tailpipe side compared to the levels
Fig. 15 present the PNC* field for the last car model (ϕ ¼ 35 ) for the observed on the other side (Z > 0). Finally, at the highest dimensionless
same planes Z ¼ 0.31, 0 and 0.31. As for ϕ ¼ 0 and 25 , the highest distance above the ground level (Z ¼ 0.70), the distribution of PNC* is
levels of PNC* are located in the tailpipe axis, that is for Z ¼ 0.31. more homogeneous (PNC*~0.01) in the close wake as we are still in the
PNC*max is 0.20 at X ¼ 0.57, Y ¼ 0.04 and Z ¼ 0.31. Overall, the results recirculation region.

10
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Table 4
Experimental conditions for Richards (2002), Gosse (2005), Kanda et al. (2006)
and Carpentieri et al. (2012).
Study Reynolds Car model and rear slant Pollutant
number angle (ϕ)

Present study 49500 Ahmed bodies, ϕ ¼ 0 , 25 Particles


and 35
Richards (2002) 335000 MIRA Gas tracer
Gosse (2005) 5600 and Ahmed bodies, ϕ ¼ 5 , 25 Heated
17600 and 40 air
Kanda et al. (2006) 7100 and Passenger car and truck Gas tracer
14000
Carpentieri et al. 12000 and Vauxhall Astravan Gas tracer
(2012) 48000

 For ϕ ¼ 0 , the recirculation region is the largest one and the flow is
also almost symmetric vertically and horizontally leading to higher
and more homogeneous PNC* levels in the close wake of the car
model;
 For ϕ ¼ 25 , the attachment of the flow on the rear slant leads to a
smaller recirculation region and prevents from a homogeneous mix-
ing over the whole height of the car model. Furthermore, the longi-
tudinal vortex on the side of the tailpipe and developing from the
edge of the car model is capable of trapping the particles in the cor-
responding half width of the model. On the other side, because of a
weaker mixing, concentrations are lower;
 For ϕ ¼ 35 , some similitudes are found with the case ϕ ¼ 0 as the
flow is detached on the rear slant angle. Nevertheless, the recircula-
tion length is smaller for ϕ ¼ 35 than for ϕ ¼ 0 .

The present results are discussed with respect to previous experi-


mental studies undertaken in wind tunnels. Four studies are considered:
Richards (2002), Gosse (2005), Kanda et al. (2006) and Carpentieri et al.
(2012). They were selected as they all dealt with pollutant dispersion in
the wake of vehicles. Table 4 summarizes their experimental conditions.
It underlines the novelty of the present study considering that previous
Fig. 16. 2D maps (XZ) of PNC* in the wake of the Ahmed body (ϕ ¼ 35 ) for Y
¼ 0.04 (a: tailpipe level), Y ¼ 0.33 (b) and Z ¼ 0.70 (c). works mostly used gas tracer or heated air while particles were used here.
Similarly, PNC fields were depicted in the present work while in the past
mass concentration fields were described.
4. Discussion
In the first one, Richards (2002) developed a numerical model to
assess pollution dispersion in the near wake of a vehicle. As part of it, she
These 2D horizontal and vertical maps of PNC highlight new insights
worked in a wind tunnel with the MIRA (Motor Industry Research As-
about particle dispersion in the wake of vehicles. Overall, they show that:
sociation) car model. She measured the concentration of a tracer gas in
the near-wake (X < 1) of her 33% scaled model by FID (Flame Ionisation
 At a distance of X ¼ 0.50 downstream of the tailpipe, PNC is rapidly
Detector). Among her most significant results, she firstly pointed out the
divided by a factor 3 to 14 depending on the rear slant angle;
influence of the recirculation region and its contra-rotating vortex
 For the three rear slant angles, highest levels of PNC are found in line
structure. Secondly, she also depicted a general upward dispersion of the
with the tailpipe exit;
tracer gas confirming that pollutant was drawn up into the recirculation
 In the close wake, the volume of the recirculation region is a key
region by the lower internal vortex. Thirdly, she stated that a rapid
parameter governing mixing and dispersion of particles in both ver-
dispersion occurred away from the source with concentration values
tical and spanwise directions;
falling to 7% of the maximum value within 50 mm of the source. Her
 Particles are emitted below the recirculation region. Nevertheless, the
findings exhibit some similarities with the present results meaning that
shearing revealed from the LDV measurements allows the carriage of
the recirculation region has a major influence on the pollutant dynamics;
UFP within the low pressure region associated with lower vortex;
let it be gas or particles.
 In a second time, another part of these UFP is shifted vertically to the
Gosse (2005) was interested in the dispersion of a passive scalar
upper part of the flow due to the interaction with the upper vortex;
(temperature) in the close wake of Ahmed bodies with three rear slant
 Meanwhile, a portion of the UFP in the recirculation region are
angles (ϕ ¼ 5 , 25 and 40 ). Scale of his models was 1/50 compared to
conveyed horizontally on both sides of the symmetry plane through
that of Ahmed et al. (1984). It is worthwhile to note that our corre-
the same turbulent system;
sponding scale is 1/20. Heated air was ejected from a tailpipe with a
velocity comparable to that of the incoming flow. His results showed a
For all conditions (ϕ ¼ 0 , 25 and 35 ), the flow dynamics is strongly
Reynolds dependence of the passive scalar field for ϕ ¼ 25 while it is
influenced by the toric vortex developing in the recirculation region. For
almost 2D for ϕ ¼ 5 and 40 whatever the Reynolds number. In the close
ϕ ¼ 25 , the longitudinal vortices emitted from the edges of the car
wake, the influence of the recirculation region on the passive scalar field
model are crucial. The turbulent structure of the wake flow plays a key
was underlined as well as the role of the longitudinal vortices for ϕ ¼ 25
role in the particle dispersion in the wake of the car models. By looking
further downstream. He also suggested that the mixing occurred rapidly
more carefully at each experimental conditions, the present results point
in the wake with a favoured dispersion in the spanwise direction. This
out that:

11
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

finding was mostly pronounced for ϕ ¼ 25 . Similarly, these results tend conditions. For ϕ ¼ 25 , the influence of the longitudinal vortices emitted
to indicate a different feature for ϕ ¼ 25 due to these longitudinal from the edges of the car was also underlined.
vortices. Overall, this paper brings new insights regarding the dispersion of
Kanda et al. (2006) worked in a wind tunnel to assess exhaust gas particles (UFP) in the wake of cars. To date, studies dealing with this
dispersion from two models of road vehicles (a sedan and a small-size topic are quite limited. Most of them used passive tracers to address this
truck, respectively). Their scale was 1/20 compared to real size vehi- issue leading to mass concentration fields. For the first time, UFP having a
cles. Reynolds numbers were 7100 and 14000, respectively. They PSD similar to that of a Diesel engine were used. The present results can
mentioned that the presence of the car enhanced the dispersion in a be used to improve existing numerical models as well as to provide data
significant way. They also suggested that the position of the tailpipe was for real PNC exposure rates for UFP epidemiologic studies. Combined
important. Indeed, for the small size truck (similar to a squared back with on board measurements, it is believed that they may contribute to
vehicle), they stated that the exhaust gas diffusion would have been some recommendation in terms of ventilation settings, driving modes
suppressed for an exhaust pipe on the top of the truck. Furthermore, they and/or air intake and tailpipe positioning to limit pollutant infiltration in
affirmed that the most affected region should be in the near-wake where the car cabin.
mixing rapidly occurs. In the present study, the role of the near-wake Nevertheless, further investigations are required to identify the role
with its recirculation region is pointed out as well as the tailpipe posi- of parameters such as the yaw angle and the incoming velocity on the
tion on the PNC field downstream of a vehicle. dynamics of particles in the wake of cars. Studying more realistic car
Carpentieri et al. (2012) carried out wind tunnel measurements shapes, changing the tailpipe’s position, size and/or angle would be
downstream of reduced scale car models (similar to 2004 Vauxhall Astra interesting as well. Furthermore, working on the experimental setup
Van). They worked with a 1/5 scale model for the near-wake charac- could be done to allow measurements closed to the bottom. Last but not
terization and a 1/20 scale model for far-wake investigation. A passive least, the improvement of analytical models (non Gaussian) would be
gas tracer was used with an exit velocity as low as possible to provide a possible.
passive release. They highlighted the effect of the recirculation region
and the corresponding vortices. According to their measurements, the Declaration of competing interest
plume was not Gaussian, especially in the near-wake. Their study was
presented as a “first step” in order to assess nanoparticle dispersion in the The authors declare that they have no known competing financial
wake of a vehicle. In that sense, this experimental work may be consid- interests or personal relationships that could have appeared to influence
ered as a further step. Furthermore, they identified some differences the work reported in this paper.
between results provided by experiments and those given by a numerical
model (CAR BUILD) which assumes a Gaussian plume. Concentrations CRediT authorship contribution statement
measured in line with the tailpipe in the wind tunnel were larger than
those predicted by the model. Indeed, this model was based on a com- Romain Rodriguez: Methodology, Validation, Formal analysis,
plete mixing assumption. From the experimental measurements, this de
Investigation, Data curation, Visualization. Fre ric Murzyn: Concep-
hypothesis must be obviously revised and the position of the tailpipe was tualization, Methodology, Validation, Investigation, Writing - original
then a relevant parameter which was not taken into account in the nu- draft, Writing - review & editing, Supervision. Amine Mehel: Supervi-
merical model. Finally, they emphasized that nanoparticle dispersion in sion. Frederique Larrarte: Writing - review & editing, Supervision.
the wake of a real car was a complex phenomena as “emitted particles
undergo a range of very fast transformation processes just after their Acknowledgements
release from the tailpipe while the passive tracer gas is affected only by
dilution”. Although difficult, it should be taken into account for wind The authors acknowledge the financial support of ESTACA (Ecole
tunnel investigations. Superieure des Techniques Aeronautiques et de Construction Automo-
bile) and Laval Agglomeration. They both funded the PhD research
5. Conclusions and perspectives project of Romain Rodriguez.

In the present paper, the dispersion of UFP in the wake of three References
Ahmed bodies having different rear slant angles (ϕ ¼ 0 , 25 and 35 )
ADEME, 2018. La pollution de l’air en 10 questions, Comment respirer un air de meilleure
was studied. This is an innovative approach as UFP were used for the first
qualite. Cles pour agir, p. 27. Septembre 2018.
time. Experimental results were obtained in a wind tunnel with reduced Ahmed, S.R., Ramm, G., Faltin, G., 1984. Some salient features of the time-averaged
scale models. The incoming velocity was U0 ¼ 14.3 m/s corresponding to ground vehicle wake. In: SAE Technical Paper Series, Paper 840300 (SP569), p. 30.
a Reynolds number based on the height of the vehicle of 49500. These Airparif, 2007. Mesures dans le flux de circulation. Etude exploratoire, Research Report.
Airparif, 35 pages (in french).
conditions were defined to be representative of a vehicle in an urban Algieri, A., Bova, S., De Bartolo, C., 2005. Experimental and numerical investigation of
environment. Velocities were recorded with a 2D LDV system. Particles the effects of the seeding properties on LDA measurements. J. Fluid Eng. 127 (3),
were generated with a PALAS DNP 2000 and PNC were measured using 514–522.
Araujo, J.A., Barajas, B., Kleinman, M., Wang, X., Bennett, B.J., Gong, K.W., Mohamad
an ELPI. PSD of these UFP corresponds to that of a Diesel engine. A Navab, M., Harkema, J., Sioutas, C., Lusis, A.J., Nel, A.E., 2008. Ambient particulate
particular attention was dedicated to data acquisition and treatment pollutants in the ultrafine range promote early atherosclerosis and systemic oxidative
method. A first set of experimental data were acquired to characterize the stress. Circ. Res. 102 (5), 589–596.
Barros, D., 2015. Wake and Drag Manipulation of a Bluff Body Using Fluidic Forcing. PhD
flow dynamics downstream of the three Ahmed bodies without the Thesis. ISAE-ENSMA, p. 97.
exhaust system. Basic flow features were characterized (recirculation Bell, M.L., Davis, D.L., Fletcher, T., 2004. A retrospective assessment of mortality from the
length, vortex center positions, turbulent kinetic energy and Reynolds London smog episode of 1952: the role of influenza and pollution. Environ. Health
Perspect. 112 (1), 6–8.
shear stresses). The results indicated recirculation lengths of 1.39 h, 0.58
Buzea, C., Pacheco, I.I., Robbie, K., 2007. Nanomaterials and nanoparticles: sources and
h and 1.06 h (h being the height of the car) for ϕ ¼ 0 , 25 and 35 , toxicity. Biointerphases 2 (4), 55. MR17.
respectively. It was shown that the presence of the exhaust system as well Carpentieri, M., Kumar, P., Robins, A., 2012. Wind Tunnel Measurements for Dispersion
Modelling of Vehicle Wakes, vol 62. Atmospheric Environment, pp. 9–25.
as the exhausted flow did not significantly affect the former results. Then,
Comite des Constructeurs Français d’Automobiles, 2017. L’industrie automobile française
UFP concentrations were assessed in the wake of the three car models. : analyse et statistiques, p. 98, 2017.
The strong correlation between the flow topology and the UFP concen- Corallo, M., Sheridan, J., Thompson, M.C., 2015. Effect of aspect ratio on the near-wake
tration fields was demonstrated. The key role of the two vortices devel- flow structure of an Ahmed body. J. Wind Eng. Ind. Aerod. 147, 95–103.

oping in the recirculation region was highlighted for all experimental

12
R. Rodriguez et al. Journal of Wind Engineering & Industrial Aerodynamics 198 (2020) 104109

Daycard-Heid, S., 2019. Quel est le coût de la pollution atmospherique ? Decod’actu, Oberd€ orster, G., Sharp, Z., Atudorei, V., Elder, A., Gelein, R., Kreyling, W., Cox, C., 2004.
Saison 2. https://education.francetv.fr/matiere/actualite/premiere/video/quel-est-l Translocation of inhaled ultrafine particles to the brain. Inhal. Toxicol. 16 (6–7),
e-cout-de-la-pollution-atmospherique. november 6th. 437–445.
Delfino, R.J., Malik, S., Sioutas, C., 2005. Potential role of ultrafine particles in Polednik, B., Piotrowicz, A., Pawlowski, L., Guz, L., 2018. Traffic-related particle
associations between airborne particle mass and cardiovascular health. Environ. emissions and exposure on an urban road. Arch. Environ. Protect. 44 (2), 83–93.
Health Perspect. 113 (8), 934–946. Pope III, C.A., Burnett, R.T., Thun, J.M., Calle, E.E., Krewski, D., Ito, K., Thurston, G.D.,
Diaz-Sanchez, D., Proeitti, L., Polosa, R., 2003. Diesel fumes and the rising prevalence of 2002. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine
atopy: an urban legend? Curr. Allergy Asthma Rep. 3, 146–152. particulate air pollution. J. Am. Med. Assoc. 287, 1132–1141.
Evans, D.E., Harrison, R.M., Ayres, J.G., 2003. The generation and characterization of Price, H.D., Stahlmecke, B., Arthur, R., Kaminski, H., Lindermann, J., Dauber, E.,
elemental carbon aerosols for human challenge studies. J. Aerosol Sci. 34 (8), Asbach, C., Kuhlbusch, T.A., Berube, K.A., Jones, T.P., 2014. Comparison of
1023–1041. instruments for particle number size distribution measurements in air quality
Gillieron, P., Kourta, A., 2011. Aerodynamique automobile pour l’environnement, le monitoring. J. Aerosol Sci. 76, 48–55.
design et la securite. Cepadues Editions, p. 299. Richards, K., 2002. Computational Modelling of Pollution Dispersion in the Near Wake of
Gosse, K., 2005. Etude experimentale de la dispersion d’un scalaire passif dans le proche a Vehicle. PhD. Thesis. University of Nottingham, p. 258.
sillage d’un corps d’Ahmed,. PhD Thesis. University of Rouen, p. 178 (in french). Roberge, B., Menard, L., Turcotte, A., Beaudet, Y., Lazure, L., 2006. Evaluation d’un
Guilmineau, E., 2008. Computational study of flow around a simplified car body. J. Wind systeme d’evacuation des gaz d’echappement lors de l’entretien des chariots
Eng. Ind. Aerod. 96, 1207–1217. elevateurs au propane. Institut de Recherche Robert Sauve en sante et en securite du
Hancke, C., 2009. Le systeme de production PSA et l’expertise des metiers au service des travail (intern report, in french), p. 45.
objectifs du groupe: voies d’amelioration de performances moteurs essence et diesel. Rodriguez, R., 2018. Etude experimentale de la dispersion de particules ultrafines dans le
Societe des Ing
enieurs Automobiles (SIA) (in french). sillage de modeles simplifies de vehicules automobiles. PhD Thesis. Ecole Centrale de
Heywood, J.B., 1988. Internal Combustion Engine Fundamentals. Mc GrawHill series in Nantes, p. 270. pages (in french).
Mechanical Engineering, p. 930. Rodriguez, R., Murzyn, F., Aubry, J., Mehel, A., Larrarte, F., 2018. An innovative LDV
Hudda, N., Eckel, S.P., Knibbs, L.D., Sioutas, C., Delfino, R.J., Fruin, S.A., 2012. Linking data processing method for statistical error corrections. In: Application to
in-vehicle ultrafine particle exposures to on-road concentrations. Atmos. Environ. 59, Homogeneous and Non-homogeneous Seeding, Flow Measurement and
578–586. Instrumentation, vol 60, pp. 67–77.
Hudda, N., Kostenidou, E., Sioutas, C., Delfino, R.J., Fruin, S.A., 2011. Vehicle and driving Rodriguez, R., Murzyn, F., Mehel, A., Larrarte, F., 2019. Ultrafine particle dispersion in
characteristics that influence in-cabin particle number concentrations. Environ. Sci. the wake of a squareback vehicle model. In: 23rd Transport and Air Pollution
Technol. 45 (20), 8691–8697. Conference, Paper 16053, Thessaloniki (Greece), 15-17 may (10 pages).
Joodatnia, P.M., Kumar, P., Robins, A., 2013. The behaviour of traffic produced Silverman, D.T., Samanic, C.M., Lubin, J.H., Blair, A.E., Stewart, P.A., Vermeulen, R.,
nanoparticles in a car cabin and resulting exposure rates. Atmos. Environ. 65, 40–51. Coble, J.B., Rothman, N., Schleiff, P.L., Travis, W.D., Ziegler, R.G., Wacholder, S.,
Kanda, I., Uehara, K., Yamao, Y., Yoshikawa, Y., Morikawa, T., 2006. A wind-tunnel study Attfield, M.D., 2012. The Diesel Exhaust in Miners study: a nested case-control study
on exhaust gas dispersion from road vehicles – Part 1: velocity and concentration of lung cancer and diesel exhaust. J. Natl. Cancer Inst. 104 (11), 55–68.
fields behind single vehicles. J. Wind Eng. Ind. Aerod. 94, 639–658. Sioutas, C., Delfino, R.J., Singh, M., 2005. Exposure assessment for atmospheric ultrafine
Knibbs, L.D., de Dear, R.J., 2010. Exposure to ultrafine particles and PM2.5 in four particles (UFP) and implications in epidemiological research. Environ. Health
Sydney transport modes. Atmos. Environ. 44 (26), 3224–3227. Perspect. 113 (8), 947–955.
Knibbs, L.D., Cole-Hunter, T., Morawska, L., 2011. A review of commuter exposure to Thacker, A., 2010. In: Universite d’Orleans (Ed.), Contribution experimentale a l’analyse
ultrafine particles and its health effects. Atmos. Environ. 45 (16), 2611–2622. stationnaire et instationnaire de l’ecoulement a l’arriere d’un corps de faible
Knibbs, L.D., de Dear, R.J., Morawska, L., 2010. Effect of cabin ventilation rate on allongement. Universite d’Orleans, pp. 1–210.
ultrafine particle exposure inside automobiles. Environ. Sci. Technol. 44 (9), Thacker, A., Aubrun, S., Leroy, A., Devinant, P., 2012. Effects of suppressing the 3D
3546–3551. separation on the rear slant on the flow structures around an Ahmed body. J. Wind
Lahaye, A., 2014. Caracterisation de l’ecoulement autour d’un corps de Ahmed a culot Eng. Ind. Aerod. 107–108, 237–243.
droit. PhD Thesis. University of Orleans, p. 159. pages (in french). Tissot, S., 1999. Toxicite des particules emises par la circulation automobile: suivi et
Leclerc, C., 2008. In: INP Toulouse (Ed.), Reduction de la traînee d’un vehicule synthese bibliographique. Final Report, INERIS (in french).
automobile simplifie  a l’aide du contr^
ole actif par jet synthetique. Institut National Tunay, T., Sahin, B., Ozbolat, V., 2014. Effects of rear slant angles on the flow
Polytechnique de Toulouse, Toulouse, pp. 1–330. characteristics of Ahmed body. Exp. Therm. Fluid Sci. 57, 165–176.
Lelieveld, J., Klingmuller, K., Pozzer, A., Poschl, U., Fnais, M., Daiber, A., Munzel, T., Tunay, T., Yaniktepe, B., Sahin, B., 2016. Computational and experimental investigations
2019. Cardiovascular disease burden from ambient air pollution in Europe reassessed of the vertical flow structures in the near wake region downstream of the Ahmed
using novel hazard ratio functions. Eur. Heart J. Fast Track Clin. Res. 1–7, 0. vehicle model. J. Wind Eng. Ind. Aerod. 159, 48–64.
Lienhart, H., Becker, S., 2003. Flow and turbulence structure in the wake of a simplified Valberg, P.A., 2004. Is PM more toxic than the sum of its parts? Risk assessment toxicity
car model. In: SAE Technical Paper Series, 01-0656 (SP-1786), p. 12. factors vs. PM-mortality “effect functions”. Inhal. Toxicol. 16 (Suppl. 1), 19–29.
Lienhart, H., Stoots, C., Becker, S., 2002. Flow and turbulence structure in the wake of a Valentino, S.A., Tarrade, A., Aioun, J., Mourier, E., Richard, C., Dahirel, M., Rousseau-
simplified car model (Ahmed model). In: New Results in Numerical and Experimental Ralliard, D., Fournier, N., Aubriere, M.-C., Lallemand, M.-S., Camous, S., Guinot, M.,
Fluid Mechanics III, Berlin, Heidelberg, pp. 323–330. Charlier, M., Aujean, E., Al Adhami, H., Fokkens, P.H., Agier, L., Boere, J.A.,
Manigrasso, M., Avino, P., 2012. Fast evolution of urban ultrafine particles: implications Cassee, F.R., Slama, R., Chavatte Plamer, P., 2016. Maternal exposure to diluted
for deposition doses in the human respiratory system. Atmos. Environ. 51, 116–123. diesel engine exhaust alters placental function and induces intergenerational effects
Manigrasso, M., Protano, C., Martellucci, S., Mattei, V., Vitali, M., Avino, P., 2019. in rabbits. Part. Fibre Toxicol. 13 (39), 14.
Evaluation of the submicron particles distribution between mountain and urban site : Verrier, R.L., Mittleman, M.A., Stone, P.H., 2002. Air pollution: an insidious and pervasive
contribution of the transportation for defining environmental and human health component of cardiac risk. Circulation 106, 890–892.
issues. Int. J. Environ. Res. Publ. Health 16, 1339, 14 pages. Wang, X.W., Zhou, Y., Pin, Y.F., Chan, T.L., 2013. Turbulent near wake of an Ahmed
Mehel, A., Murzyn, F., 2015. Effect of air velocity on nanoparticles dispersion in the wake vehicle model. Exp. Fluid 54, 1490, 19 pages.
of a vehicle model: wind tunnel experiments. Atmosph. Pollut. Res. 6, 612–617. Watkins, S., Vino, G., 2008. The effect of vehicle spacing on the aerodynamics of a
Mehel, A., Murzyn, F., Cuvelier, P., Deville Cavellin, L., Baudic, A., Joly, F., Patte representative car shape. J. Wind Eng. Ind. Aerod. 96, 1232–1239.
Rouland, B., Varea, E., Sioutas, C., 2019. Caracterisation et Analyse des Polluants West, G.S., Apelt, C.J., 1982. The effects of tunnel blockage and aspect ratio on the mean
issus du Transport automobile s’Infiltrant dans les Habitacles des Vehicules, Rapport flow past a circular cylinder with Reynolds numbers between 10000 and 100000.
final. Projet CAPTIHV, ADEME, p. 136 pages (in french). J. Fluid Mech. 114, 361–377.
Mejia, J., Morawska, L., Mengersen, K., 2007. Spatial variation in particle number size Xu, B., Zhu, Y., 2013. Investigation on lowering commuters’ in-cabin exposure to ultrafine
distributions in a large metropolitan area. Atmos. Chem. Phys. 8 (5), 1127–1138. particles. Transport. Res. Transport Environ. 18, 122–130.

Morin, J.-P., Gouriou, F., Preterre, D., Bobbia, M., Delmas, V., 2009. Evaluation de Zweiman, B., Slavin, R.G., Feinberg, R.J., Falliers, C.J., Aaron, T.H., 1972. Effects of air
l’exposition aux polluants atmospheriques des conducteurs de vehicules automobiles pollution on asthma: a review. J. Allergy Clin. Immunol. 50 (5), 305–314.
par la mise en œuvre de mesures dynamiques dans l’habitacle du vehicule. Arch.
Maladies Prof. Environnement 70 (2), 184–192 (in french).

13

You might also like