Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Building and Environment 105 (2016) 1e12

Contents lists available at ScienceDirect

Building and Environment


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

Coupling CFD and analytical modeling for investigation of monolayer


particle resuspension by transient flows
Carine Habchi a, Kamel Ghali b, Nesreen Ghaddar b, *
a
Mechanical Engineering Department, Lebanese American University, Beirut, Lebanon
b
Mechanical Engineering Department, American University of Beirut, P.O. Box 11-0236, Beirut 1107-2020, Lebanon

a r t i c l e i n f o a b s t r a c t

Article history: One of the main goals of public health studies is to determine the different factors affecting the
Received 12 March 2016 breathable air quality (BAQ). Particle resuspension can constitute significant threat to human health since
Received in revised form it results in the re-introduction of contaminated particles in the air after their deposition on surfaces.
19 May 2016
Particle resuspension by transient flows involves complex flows physics challenging its modeling.
Accepted 19 May 2016
Available online 20 May 2016
In the current work an analytical model is developed for modeling the probabilistic aspect of
monolayer particle resuspension by transient flows coupled with computational fluids dynamics model
predicting the stress variation on the surface. The developed analytical model incorporates the effect of
Keywords:
Monolayer particle resuspension
turbulent bursts, surface roughness in addition to the flow and particle/surface properties on monolayer
Transient flows particle resuspension. The coupling of the analytical and computational fluids dynamics models for
Turbulent bursts investigation of monolayer particle resuspension by transient flows was validated by published literature
Surface roughness data. Good agreement was obtained between literature data and models results revealing that the
Particle surface properties coupling between the analytical and computational fluid dynamics model is able of capturing the
principal physics affecting monolayer particle resuspension.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction particle resuspension and determining the different factors


affecting it. The influencing parameters can be divided into several
Resuspension is a physical mechanism by which deposited categories [17e23]: particle properties (size, shape, density, and
particles on surfaces are reintroduced to air [1]. Particle resus- charge); surface properties (material type, surface roughness, and
pension constitutes a threat to human health since surfaces acting charge); airflow characteristics (flow acceleration, friction velocity,
as particle sinks, by the process of deposition, become particulate fluid velocity, and turbulent level of the flow).
matter sources [2,3] by different human activities [4e7] or turbu- Several modeling studies were conducted to set the conditions
lent airflows [8e10]. As air quality is one of the main concerns of for particle resuspension progressing from static models to kinetic
public health studies [11e13], the resuspension process should be models and then to dynamic models [16]. The first modeling
well understood because it contributes to people infection through approach was static assuming that resuspension occurs when the
inhalation of contaminated resuspended particles [14,15]. It is the particle adhesion is overcome based on a force or moment balance
result of an unbalance between competing adhesion and removal rupture. However, earlier static approaches did not consider the
forces exerted on particles [1,3]. Several forces contribute to the net effect of turbulent bursts on particle resuspension under-predicting
adhesive force and are of different types: forces at distance (as resuspension rates for highly turbulent flows [16]. The impact of
gravity, electrostatic forces, and van der Walls force), and forces of turbulent bursts was found to significantly reduce the threshold
contact (surface tension forces) [1,3,16]. Removal forces are needed velocity necessary for resuspension [24,25] constituting an
for particles dislodgment from a surface, and can be categorized important contributor to particle removal.
into hydrodynamic, mechanical, vibrational and centrifugal forces To consider the effect of turbulence bursts, kinetic models were
[1]. Identifying these forces allowed modeling the conditions for developed. Reeks et al. [24] presented a new, energy balance kinetic
approach to model resuspension by which a particle will detach
once it accumulates sufficient vibrational energy, imparted to it
* Corresponding author. through the turbulent energy of the fluid flow related to the
E-mail address: farah@aub.edu.lb (N. Ghaddar).

http://dx.doi.org/10.1016/j.buildenv.2016.05.025
0360-1323/© 2016 Elsevier Ltd. All rights reserved.
2 C. Habchi et al. / Building and Environment 105 (2016) 1e12

Nomenclature u mean velocity


u0 fluctuating velocity
u* friction velocity
Symbols uþ normalized velocity
a particle/surface contact area radius u0max maximum fluctuating velocity
a∞ free stream acceleration u0rms root mean square (rms) fluctuating velocity
Cc Cunningham factor to account for the slip effect on the u0þ
rms root mean square (rms) normalized fluctuating
drag force velocity
Cr roughness ratio ut stream component of Ut
CFD computational fluid dynamics Ut magnitude of the velocity vector at the particle center
DF detachment fraction parallel to the wall
dp particle diameter U∞ free stream velocity
Ep particle Young modulus UDF user-defined function
Es surface Young modulus w transverse velocity
f1 drag force correction factor to account for the wall wt transverse component of Ut
effect W particle weight
f2 stress moment correction factor to account for the wall Wad thermodynamic work of adhesion
effect y distance from the wall
Fad,JKR adhesive force modeled by the JKR theory Greek symbols
FD drag force l molecular mean free path of air
FL lift force tw wall shear stress
Fp pulleoff force for rough surface r fluid density
Fp0 pulleoff force for smooth surface n fluid kinematic viscosity
k stiffness constant m fluid dynamic viscosity
L sample location from the wind tunnel inlet mf static coefficient of friction
LES Large eddy simulation z randomly distributed number following the Gaussian
Ms stresses moment about the center of the particle distribution
Ndet number of particles detached from the sampling d elastic deformation between the particle and the
surface surface
N number of particles remaining on the sampling surface sp particle Poisson ratio
RSM Reynolds stress model ss surface Poisson ratio
t time 4r reduction rate of the pull-off force
u streamwise velocity Dt time step

fluctuating component of the lift force (referred to as the RRH particles are resuspended [23] and particle and flow hydrody-
model). By accounting for resonant energy transfer, RRH predicts namics close to surfaces have been modeled under variable con-
particle resuspension at lower velocities compared to models based ditions to assess particle resuspension [23,26]. However, knowing
on force/moment balances. However, the RRH model provides ac- the threshold velocity for particle resuspension does not allow the
curate predictions of resuspension in case of large particle di- tracking of the variation in time of particle resuspension. To our
ameters sensitive to vibration energy and not small particles [16]. knowledge, published modeling approaches have not proposed
Furthermore, RRH does not account for the drag force influence on transient particle resuspension modeling that considers the prob-
particle resuspension. Reeks and Hall [25] refined the RRH model abilistic effect of velocity fluctuation, surface roughness and parti-
by developing the ‘Rock‘n’Roll’ (RnR) model to include the effect of cle diameter variability. In addition, in the case of transient flows,
the drag force and to be valid for small particle diameters. Never- the time variation of surface friction velocity and its resulting effect
theless, the RnR is a 2-D model with the assumption of symmetric on mean and fluctuating components of the flow velocity near the
two-point asperity contact which limits its application to particular surface were not tackled by modeling although they are important
surface geometry. Besides, both RRH and RnR consider steady flows parameters for computing the variation of the resuspended fraction
with constant mean forces. with time. Thus, previous models cannot capture the different
A recent static model by Ibrahim et al. [23] overcame the ge- experimental observations of resuspension during transient flows
ometry and steady state constraints of the RnR, and discussed the [27,28].
possibility of three modes of particle resuspension involving the Modeling particle resuspension by transient flows is challenging
effect of turbulence bursts: direct lift-off, sliding, and rolling. The since it involves variation of the surface friction map with time in
dominant mode is reported to be the one requiring the less addition to the probabilistic effect created by turbulent bursts’
threshold friction velocity to occur [23]. events and surface roughness distribution. Therefore, proper
In ideal situation when factors affecting resuspension are not combination of analytical and numerical modeling should be used.
changing (e.g. turbulent flow fluctuations are absent, particles are In fact, particle resuspension is driven by removal forces and mo-
of uniform size, and surface roughness is constant), particle ments which are correlated empirically to surface shear magnitude
detachment occurs at one free stream velocity [23]. Nevertheless, in or friction velocity. However, surface shear stress distribution with
reality and because of the variability of these factors, resuspension position and time induced by transient flows involves complex
occurs over a range of velocities [23]. The threshold velocity for physics requiring substantial numerical computations using
particle detachment is defined as the velocity for which 50% of computational fluid dynamics (CFD). On the other hand, the
C. Habchi et al. / Building and Environment 105 (2016) 1e12 3

particle resuspension condition using force and moment balances


on the particle and the probabilistic factors affecting this condition
should be modeled analytically. Elhadidi and Khalifa [29,30]
coupled CFD with a particle detachment model to assess the ef-
fect of flow acceleration on particle resuspension by falling objects.
In this work, the CFD modeling approach is coupled with particle
probabilistic analytical force and moment balance model to
investigate monolayer particle resuspension by transient flows. The
ability of the coupled models in capturing the flow physics near the
surface and predicting the variation of resuspension fraction with
time is validated using the published data of Ibrahim et al. [23,27].
A parametric study is then conducted to assess the effect of the
different influential variables on monolayer particle resuspension
by transient flows.
This work considers not only variation of turbulent bursts but
also probabilistic distributions in particle diameter and surface
roughness improving monolayer particle resuspension predictions.
Furthermore, the analytical detachment model is coupled with CFD
simulations to predict resuspension rates and velocities very close
to the wall using empirical correlations reducing significantly the
CFD computational cost. Literature experimental data [23,27] is
limited for certain ranges of variation of the involved parameters
affecting resuspension while the parametric study covers a wider
range of variable conditions allowing a deeper understanding of the
different physics involved in the resuspension mechanism.

2. Methodology

Particle resuspension is driven by airflow removal forces and


Fig. 1. Flowchart summarizing the research methodology.
moments which are correlated to surface shear stress. The analyt-
ical resuspension model requires that the shear distribution on the
surface to be known which cannot be computed analytically for the in the developed CFD model. Direct numerical simulation (DNS) is
case of transient flows. For this reason, the prediction of shear the most accurate method to solve for turbulence by direct reso-
variation with time and position on the surface needs CFD lution of the Navier Stockes equations but the computational time it
modeling. Then, the predicted CFD shear variation with time is used requires is extremely large [31]. ANSYS Fluent offers different tur-
as an input to the analytical monolayer resuspension model which bulence models. Large eddy simulation (LES) resolves by DNS the
predicts the near wall flow variation and sets the resuspension large eddies and by modeling the small eddies reducing the
condition that incorporates the effect of turbulent burst and surface computational cost of DNS [31]. However, in the near wall region of
roughness in addition to many other factors affecting monolayer interest small eddies are dominant therefore using LES for the
particle resuspension. resuspension problem increase the computational cost without
The experimental data of Ibrahim et al. [23,27] is used to vali- increasing the accuracy. Among the turbulence models with lower
date the coupling between CFD and analytical resuspension computational cost compared to LES, the Reynolds stress model
models. The validated resuspension model is then used to conduct (RSM) is the most accurate model in investigating particle resus-
a parametric study and a sensitivity analysis to assess the effect of pension since it takes into consideration the anisotropy of velocity
the various flow and particle parameters on particle resuspension. fluctuations near the wall [31]. Therefore the RSM turbulence
Fig. 1 represents a flowchart summarizing the followed model is selected as a compromise between accuracy and
methodology.

2.1. CFD modeling

Shear stress profiles for transient flows can’t be obtained by


simple analytical solutions requiring computational fluid dynamics
(CFD). The commercial software ANSYS Fluent is used to develop a
CFD model simulating complex flows situations. The particle
seeded surface of interest is discretized into small squares to
resolve for the shear stress variation with position. For each cell, the
variation of shear stress magnitude tij ðtÞ can be determined from
CFD and used in the developed resuspension model involving
empirical correlations relating friction velocity to near wall flow
velocity as described in the following section to predict the possi-
bility of monolayer particle resuspension from the corresponding
cell.
As turbulent fluctuations of the velocity play an important role
in particle resuspension appropriate turbulence model accounting Fig. 2. a) Front view of moments and forces acting on the particle; b) Top view of the
for the anisotropy in turbulence near the wall should be considered velocity and force diagrams.
4 C. Habchi et al. / Building and Environment 105 (2016) 1e12

computational cost.
u
The transient flow pattern are imposed as boundary condition to uþ ¼ (1b)
the developed CFD model using user-defined functions (UDF). A u*
robust CFD model that can accurately predict the airflow field and and u* represents the friction velocity given by
corresponding shear close to surfaces around deposited particles is
rffiffiffiffiffiffi
essential since it affects largely particle resuspension rates. There- tw
fore, the flow physics should be accurately modeled to capture u* ¼ (1c)
r
essentially turbulence effects and boundary layers development
near the surfaces and determine the corresponding shear stress where tw is the wall shear stress and r is the fluid density. The shear
distribution. The shear stress distribution constitutes a primordial distribution on the surface obtained from CFD is used as an input
factor since it represents the effect of the airflow pattern on particle for the analytical resuspension model.
resuspension. A tetrahedral unstructured grid was generated for its In the viscous sub-layer the mean normalized velocity can be
ability to adapt locally to fine flow scale structures in the devel- modeled [35] as
oping flow boundary layer region near surfaces of interest while
remaining coarse in regions of large flow structures away from the u* y
surfaces reducing the computational cost compared to structured
uþ ¼ yþ ¼ (1d)
n
grid, requiring a higher number of elements for equivalent mesh
refinement near the surfaces [32,33]. Fine element sizing and where y is the distance from the wall and n is the kinematic
inflation were used as mesh treatment for the surfaces because the viscosity.
control of the grid resolution close to the surfaces is essential to Within the viscous sublayer, the root mean square normalized
insure yþ values of the order of unity for resolution of the devel- fluctuating velocity u0þ
rms can be expressed as mentioned by Soltani
oping boundary layers. Enhanced wall treatment is selected with and Ahmadi [26]:
the turbulent RMS model to resolve accurately the developing
boundary layer. Mesh adaptation is performed to obtain yþ values u0þ þ
rms ¼ 0:25y þ 0:0325y
þ2
(1e)
in the required range (order of unity) for variable aerodynamics
The ratio of the maximum fluctuating velocity u0 max to the root
conditions to investigate accurately the flow physics for different
mean square fluctuating velocity was experimentally determined
airflow accelerations. For instance, a higher flow velocity requires a
by Johnson and Alfredsson [36] to be
larger mesh refinement near surfaces to maintain yþ values within
the recommended range. u0max
The Eulerian approach will be used to simulate air flow field ¼ 2:9 (1f)
u0rms
which can be considered as a continuous fluid [34]. ANSYS Fluent
software solves for the Navier-Stokes equations transformed to The velocity fluctuation at a certain position and time is variable
algebraic equations with different possible discretization schemes. and its occurrence can be modeled using a stochastic approach [31]
A second-order upwind discretization scheme is adopted for the as follows:
energy, momentum, turbulent intensity, and dissipation rate
equations. For pressure, the “PRESTO!” staggered scheme is u0 ¼ z u0rms (1g)
selected [31] and the SIMPLE algorithm couples the velocity and
pressure fields [31]. where z is a randomly distributed number following the normal
Gaussian distribution and u0rms is the root mean square (rms) fluc-
tuating velocity [31].
2.2. Analytical resuspension model The flow near the wall is dominated by eddies in the transverse
direction perpendicular to the stream wise flow leading to the
The surface shear map obtained from CFD was used as an input deviation of the flow within the viscous-sublayer from the main
for the developed analytical resuspension model. In fact, investi- flow direction [37]. Therefore, the vector of magnitude (Ut) parallel
gation of monolayer particle resuspension requires the determi- to the wall has two components ut (stream component) and wt
nation of the flow velocity and fluctuations at the particle center. As (transverse component). Hence Ut is given by
particles diameter range of interest for indoor air quality (IAQ) is in qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
micrometers, using only CFD modeling to resolve for velocity Ut ¼ u2t þ w2t ¼ u* uþ2 t þ wt
þ2
(1h)
variation at the particle center requires very high computational
time due to the needed very fine meshing which might not be According to Soltani and Ahmadi [26], the stream and transverse
realizable for sub-micrometers particle diameters. Furthermore, components u and w are correlated by the following equation:
many probabilistic factors affect monolayer particle resuspension
and require appropriate modeling. For this reason, the developed wþ ¼ 0:54uþ (1i)
CFD model was coupled with an analytical model to capture the
Replacing (1i) in (1h) gives
different physics affecting monolayer particle resuspension.
Ut ¼ 1:1365uþ u (1j)
2.2.1. Modeling of near wall velocity field
In order to account for turbulent bursts, flow fluctuations and
their effect on removal forces should be involved in the analytical
model. The stream wise velocity u can be divided into a mean u and 2.2.2. Forces and moments affecting resuspension
fluctuating u0 component as follows: Particle adhered to the surface is subjected to variable
competing forces and moments as shown in Fig. 2:
uþ ¼ uþ þ u0þ (1a)
a) The adhesive forces between the particle and the surface
where uþ represents the normalized stream-wise velocity given by b) The particle weight
C. Habchi et al. / Building and Environment 105 (2016) 1e12 5

c) The removal forces created by the flow which are mainly the position is essential for predicting resuspension rates and provided
drag and lift forces as input to the resuspension analytical model from the CFD model.
d) The surface stresses moment about the center O of the Once the shear stress variation on the surface is known, the friction
particle velocity, yþ values, velocities, forces and moments can be deter-
mined using Eqs. 1(aej) and 2(aee).
Micro-particles are embedded completely within the viscous
sublayer therefore their Reynolds number is relatively small so the 2.2.3. Modeling of resuspension condition
drag force FD can be modeled as reported by Stokes [23,26] as The condition of detachment depends on the mode of resus-
follows: pension (rolling, sliding, or lifting).

3 p f1 m dp a) Detachment by rolling:
FD ¼ Ut (2a.1)
Cc  
 
0:5dp  d FD þ aFL þ Ms  a Fad;JKR þ W (3a)
2l  
Cc ¼ 1 þ 1:257 þ 0:4e1:1dp =2l (2a.2)
dp where FD is the drag force, FL is the lift force, a is the particle/surface
contact area radius, d is the elastic deformation between the par-
where m is the fluid viscosity, f1 the correction factor found by ticle and the surface which can be neglected since (d <<< a) [26], Ms
O’Neill [38] to account for the wall effect, Cc the Cunningham is the surface stresses moment about the center of the particle, and
correction factor to account for the slip effect [31], dp the particle W is the particle weight.
diameter, l the molecular mean free path of air (equal to 0.07 mm The particle/surface contact area radius is given by
under normal pressure and temperature conditions) and Ut the
 . 1=3
magnitude of the velocity vector at the particle center O parallel to
a ¼ 1:5p Wad d2p k (3b)
the wall (Fig. 2).
The surface stresses moment about the center of the particle is
given by Ref. [32]: where k is the stiffness constant depending on the particle/surface
materials and is given by
Ms ¼ 2 p m f2 Ut d2p (2b) 0 
 11
4 @ 1  sp
2
1  s2s A
where f2 is the correction factor found by O’Neill [38] to account for k¼ þ (3c)
3 Ep Es
the wall effect on the stress moment.
The lift force is given according to Saffman [39] by
where Ep, Es are the Young modulus of the particle and the surface
u *2 Utþ respectively; and sp ; ss are the Poisson ratio of the particle and the
FL ¼ 0:975r d3p (2c) surface respectively.
n
The adhesive force Fad,JKR between a particle and a smooth sur- b) Detachment by sliding:
face is modeled by the JKR theory [40] as
 
FD  mf Fad;JKR þ W (3d)
Fad;JKR ¼ 0:75p dp Wad (2d)

where dp is the particle diameter, and Wad is the thermodynamic where mf is the static coefficient of friction.
work of adhesion which depend on the particle surface
combination. c) Detachment by lifting:
Surface roughness reduces the adhesive force and generally its  
surface distribution obeys to the normal Gaussian distribution. The FL  Fad;JKR þ W (3e)
roughness ratio Cr represents the ratio of the rough-to-smooth
surface pull-off forces which is equal to the rough-to-smooth ad-
hesive forces [23]. Cheng et al. [41] developed a theory to investi-
gate the effects of surface roughness on microparticle adhesion and 2.2.4. Probabilistic behavior of particle resuspension
represented graphically their main findings including the variation For each resuspension condition the threshold velocity, Ut,th can
graph of the reduction rate of the pull-off force 4r with the surface be determined. The lowest threshold velocity correspond to the
roughness standard variation. The reduction rate of the pull-off
force 4r is related to the roughness ratio Cr by the Eq. (2e.1)
allowing the determination of the Cr variation with surface
roughness:

Fp0  Fp Fp
4r ¼ ¼1 ¼ 1  Cr (2e.1)
Fp0 Fp0

where Fp0 and Fp are the pulleoff forces for smooth and rough
surfaces respectively.
Therefore, the reduced adhesive force Fad is given by

Fad ¼ Cr *Fad;JKR (2e.2)

Friction velocity and/or shear stress variation with time and Fig. 3. Probability distribution of surface roughness.
6 C. Habchi et al. / Building and Environment 105 (2016) 1e12

dominant resuspension mode. Once Ut,th is determined, the prob- interaction with CFD, experimental literature information is needed
ability of resuspension Pr over the sample surface at a certain time t on transient monolayer particle resuspension and the effect of the
is driven by the surface roughness distribution probability pi, the particle and surface properties and airflow aerodynamics. For this
particle diameter distribution probability pj and the burst occur- reason, the published data of Ibrahim et al. [23,27] are used for the
rence resuspension probability pb,i,j which depend on the particle current model validation where they studied the effect of flow ac-
diameter and surface roughness combination expressed as celeration and particle and surface properties during transient
monolayer particle resuspension.
X
m X
n The experiments of Ibrahim et al. [23,27] were conducted in a
Pr ðtÞ ¼ pj *pi *pb;i;j ðtÞ (4a) wind tunnel consisting of a contraction section, an inlet section
j¼1 i¼1
(1 m long and a section area of 0.203 m  0.203 m), a test section
(1 m long) and a diffuser section. A glass substrate (0.105 m
where n and m are respectively the number of sub-intervals of
long  0.1 m width  0.00127 m thickness) of 17 Å average
surface roughness and particle diameter covering the total rough-
roughness standard deviation seeded with monolayer particle de-
ness and diameter ranges. The parameters pi and pj are the prob-
posit was placed in the wind tunnel test section [23,27]. Two par-
abilities of having the surface roughness and particle diameter
ticle types (stainless steel and glass) were used to investigate the
respectively in the interval of center i/j. The value of the probability
effect of different particle/surface combinations on transient
is equal to the hatched area under the curve as shown in Fig. 3
monolayer particle resuspension for a flow acceleration of 0.18 m/s2
which illustrates the probability distribution of surface roughness
[23].
as an example.
Ibrahim and Dunn [27] investigated by experimentation the
For each combination of surface roughness i and particle
effect of flow acceleration (varied from 0.01 to 23 m/s2) on particle
diameter j corresponds an adhesive force resulting in a threshold
detachment. They observed that as the flow acceleration increased
velocity Ut,th,i,j. The probability pb,i,j of resuspension by the flow is
above 0.3 m/s2, the detachment fraction decreased. To provide an
equal to the probability of having Ut higher than Ut,th,i,j and is given
explanation for their findings, Ibrahim and Dunn [27] measured the
by
instantaneous flow velocity at 80 mm from the surface for different
free stream accelerations. They noticed a delay of turbulence onset
pb;i;j ¼ 0 if Ut;th;i;j > Ut;max;i;j (4b)
with increased flow acceleration resulting in reduced occurrence of
turbulent burst events within the boundary layer. Table 1 sum-
pb;i;j ¼ 1 if Ut;th;i;j < Ut;min;i;j (4c) marizes the different boundary conditions from the experimental
set-up of Ibrahim et al. [23,27] that are used as inputs to our current
coupled CFD and analytical models for validation.
UZ
t;max;i;j
Simulations were performed of the fluid flow at the conditions
pb;i;j ¼ fb ðxÞ dx if Ut;min;i;j < Ut;th;i;j < Ut;max;i;j (4d) of the experimental set-up of Ibrahim and Dunn [27]. Fig. 4 illus-
Ut;th;i;j trates the CFD mesh treatment. A grid independence test was
conducted for appropriate mesh selection resulting in nearly
where Ut,max,i,j, Ut,min,i,j represent respectively the maximum and 4,034,000 elements. The wind tunnel bottom surfaces particularly
minimum velocity components parallel to the wall at the particle the sampling test surface were meshed with fine elements sizing of
center and fb(x) the Gaussian distribution function of Ut due to the 2 mm. Furthermore, five-layer inflation was used at surfaces and
occurrence of turbulent bursts events. mesh adaptation was performed to trap the flow physics for the
The detachment fraction DF represents the ratio of the number different accelerations studied.
of detached particles to the initial number of particles on the The ability of the current developed models in capturing the
seeded surface and its variation with time t can be computed ac- flow physics and transient boundary layer development near the
cording to the following expression: wall is assessed by comparing the near wall velocity variation with

Ndet ðt þ DtÞ Ndet ðtÞ þ Pr ðt þ DtÞ  NðtÞ Ndet ðtÞ Pr ðt þ DtÞ  ðNð0Þ  Ndet ðtÞÞ
DF ðt þ DtÞ ¼ ¼ ¼ þ
Nð0Þ Nð0Þ Nð0Þ Nð0Þ (4e)
¼ DF ðtÞ þ Pr ðt þ DtÞ  ð1  DF ðtÞÞ

time with CFD predictions. Then, Ibrahim et al. data [23] is used to
where Dt is the time step, Ndet(t) is the number of particles de- validate the predictions of the coupled CFD and analytical models of
tached from the sampling surface at time t, N (t) the number of the transient resuspension fraction from rough surface for different
particles remaining on the sampling surface at time t, and Pr(t þ Dt) particle/surface combinations by incorporating the probabilistic
the probability of resuspension over the sample surface at time factors affecting particle resuspension.
t þ Dt.

3.1.1. Validation of the near-wall velocity field


3. Results and discussions The experimental set-up of Ibrahim and Dunn [27] is simulated
in CFD. The variation of the mean velocity magnitude with time at
3.1. Literature validation of the coupling between CFD and 80 mm from the surface is determined from coupling the CFD and
analytical models analytical models for two different accelerations (a ¼ 0.01 m/s2 and
a ¼ 23 m/s2). For the considered CFD mesh, a time step of 0.001 s is
In order to validate the proposed resuspension model and its used to insure convergence and scale residuals below 105.
C. Habchi et al. / Building and Environment 105 (2016) 1e12 7

Table 1
Boundary conditions from the experimental set-up of Ibrahim et al. [23,27] used for model validation.

Boundary conditions from the experimental set-up of Ibrahim and Dunn [16,20]

Wind tunnel inlet section dimensions 1 m long; 0.203 m  0.203 m cross section area
Wind tunnel test section dimensions 1 m long; 0.203 m  0.203 m cross section area
Substrate dimensions 0.105 m long; 0.1 m width; 0.00127 m thickness
Airflow acceleration range [0.01 m/s2e23 m/s2]
Substrate material Glass
Average ratio of the rough to smooth surface 0.01
Standard deviation of the surface roughness height 17 Å
Type of deposits Monolayer deposition on glass substrate
Particles’ material Stainless steel/Glass
Stainless steel average diameter 70 mm
Glass average diameter 72 mm
Stainless steel density 8000 kg/m3
Glass density 2420 kg/m3
Stainless steel Young modulus 215 GPa
Glass Young modulus 80.1 GPa
Stainless steel poisson ratio 0.28
Glass poisson ratio 0.27
Adhesion energy of stainless steel on glass 0.15 J/m2
Adhesion energy of glass on glass 0.4 J/m2

In comparison with Ibrahim and Dunn findings [27], good increase of flow acceleration in accordance with the observation of
agreement was obtained for both accelerations as shown in Fig. 5 Ibrahim and Dunn [27].
with a relative error of the order of 10% showing that the CFD The transient shear stress variation on the surface is obtained
model is able to predict the effect of flow acceleration on sample from CFD allowing the determination of the friction velocity vari-
surface shear stress variation with free-stream Reynolds number ation with time from Eq. (1c). As the sample testing area was very
(Re ¼ rLU ∞
n ). Note that the free-stream Reynolds number is propor- small, no significant variation of the shear stress was recorded with
tional to time since the modeled flow has constant acceleration position. Then using Eqs. (1d)e(1f), the mean velocity, rms fluctu-
(Re ¼ rLan∞ t ). Thus, by proper mesh adaptation the CFD model can ating velocity, and maximum fluctuating velocity can be deter-
accurately predict the aerodynamics of the flow for variable ac- mined respectively as shown in Fig. 6.
celerations. A delay in the onset of turbulence is observed with the The instantaneous flow velocity is obtained by the summation
of the mean and fluctuating velocity components (Eq. (1aeb))

Fig. 5. Comparison of the mean velocity variation at 80 mm from the surface measured
by Ibrahim and Dunn [27] with the results obtained by coupling CFD and analytical
Fig. 4. CFD mesh treatment: a) 3D view; b) profile view; c) front view. models for a free stream acceleration of: a) 0.01 m/s2; and b) 23 m/s2.
8 C. Habchi et al. / Building and Environment 105 (2016) 1e12

current model and those obtained from published experimental


data [27] with a relative error of the order of 10%. Furthermore, the
velocities fluctuations in the turbulent flow region are nearly
within the fluctuations limits (represented by the bars in Fig. 7)
found by the resuspension model. The methodology followed to
determine the instantaneous velocity is summarized in a flowchart
(Fig. 8).

3.1.2. Validation of the transient resuspended fraction profile


Once the velocity profile at the particle center is computed,
forces and moments acting on the particle are determined using
Eqs. (2aee). Then as shown in the flowcart of Fig. 8, the resus-
pension condition given by Eqs. (3aee) on the threshold detach-
ment velocity is combined with the probabilistic behavior of
particle resuspension (see Eqs (4aed)) which account for the
gaussian distribution of particle diameter, surface roughness, and
velocity fluctuation. Hence, the probability of particle resuspension
is determined and the variation of resuspension fraction versus
time is obtained.
The developed model prediction of the resuspension fraction
variation with the stream velocity for different particle/surface
combinations (stainless steel/glass and glass/glass) for a flow ac-
celeration of 0.18 m/s2 is compared with the findings of Ibrahim
et al. [23] as illustrated in Fig. 9. As the particle/surface work of
adhesion increases, the adhesive force between the particle and the
surface is increased leading to a delay in the initiation of resus-
pension occurring at higher free stream velocity (Fig. 9). Therefore,
the rate of particle resuspension variation with time is decreased
with higher work of adhesion. Good agreement was obtained be-
Fig. 6. Variation at 80 mm from the surface of: a) the mean velocity; b) rms fluctuating
tween the experimental and predicted values of resuspended
velocity; and c) maximum fluctuating velocity for a free stream acceleration of 23 m/s2. fraction of particles with a relative error of the order of 10% and
hence validating the accuracy of the developed resuspension model
in predicting the transient resuspension behavior for different
particle surface combinations.

3.2. Parametric study

In order to determine the most affecting parameters on mono-


layer particle resuspension a parametric study is needed involving
airflow characteristics, surface and particle properties. The same
set-up of Ibrahim and Dunn [27] is used to conduct a parametric
study to examine the effect of the different independent variables
(particle diameter, particle density, surface roughness standard
deviation, surface Young modulus, and flow acceleration) on par-
ticle resuspension. Since the studied variables are independent, the
variation of one of them will not affect the others. Each parameter is
varied in a realistic physical range while the others are kept fixed to
conditions similar to the ones reported by Ibrahim and Dunn [27]
(Stainless steel particle diameter of 70 mm, particle density of
Fig. 7. Comparison of the instantaneous velocity variation at 80 mm from the surface
8000 kg/m3, surface roughness standard deviation of 17 Å, Glass
measured by Ibrahim and Dunn [27] with the results obtained by coupling CFD and surface Young modulus of 80 GPa, flow acceleration of 0.3 m/s2).
analytical models for a free stream acceleration of 23 m/s2. Table 2 represents the different parameters studied and their range
of variation.
0
varying in the interval [umean  u0 max; umean þ u max] and according 3.2.1. Particle characteristics effect
to a probability function following a gaussian distribution as was 3.2.1.1. Particle diameter. The particle diameter is varied between 1
given in Eq. (1g). Fig. 7 shows the instantaneous flow velocity in and 200 mm covering fine and coarse particles’ modes and its effect
comparison with Ibrahim and Dunn findings at 80 mm from the on the resuspension fraction with time is assessed. Fig. 10a shows
surface for a free stream acceleration of 23 m/s2 [27]. The transition the variation of the detachment fraction with the free-stream ve-
from laminar to turbulent flow was captured by a sudden slope locity (that is varying proportionally with time as the acceleration is
increase of the mean flow velocity as shown in Fig. 5b. Instanta- constant) for different particle diameters. As the particle diameter
neous velocities during the laminar flow period and intervals of increases, the ratio of removal to adhesive forces and moments
velocity variation (represented by the bars in Fig. 7) due to turbu- increases. For instance, the adhesive force is proportional to dp (Eq.
lence effects during the turbulent flow period were computed. (2d)) while the surface stress moment is proportional to d2p (Eq.
Good agreement was observed between predicted values by (2b)), and the lift force to d3p (Eq. (2c)). Furthermore, the moment of
C. Habchi et al. / Building and Environment 105 (2016) 1e12 9

Fig. 8. Flowchart summarizing the analytical model methodology.

the drag force is proportional to dp while the moment of the ad-


hesive force is proportional to d2/3
p (Eqs. (3a) and (3b)). Thus with
increased particle diameter, resuspension starts at smaller free
stream velocity and the rate of particle resuspension with time is
increased requiring lower range of velocities to reach 100% resus-
pension. For particle diameters smaller than 1 mm, the threshold
velocity significantly increased such that no particle detachment is
observed for stream velocities lower than 12 m/s under the
experimental conditions of Ibrahim and Dunn [20] (Fig. 10a).

3.2.1.2. Particle density. Most materials have specific gravity values


ranging between 0.2 and 20 [42]. Fig. 10b illustrates the variation of
the detachment fraction versus free-stream velocity for variable
particle densities. As observed in Fig. 10b, particle resuspension is
insensitive to the particle density within this range. In fact, particle
density mainly affect the weight of the particle which is negligible
compared to the removal and adhesive force which make particle
detachment fraction insensitive to the particle weight and thus to
the particle density (Fig. 10b).

3.2.2. Surface characteristics effect


3.2.2.1. Surface roughness standard deviation. The surface rough-
ness standard deviation is varied between 1 and 25 Å and its effect
on the resuspension fraction with time is investigated. Fig. 11a

Table 2
Parameters and their ranges for the parametric study or/and sensitivity analysis and
their range of variation.

Particle diameter [1 mm; 200 mm]


Particle density [200 kg/m3; 20,000 kg/m3]
Surface roughness standard deviation [1 Å; 25 Å]
Surface Young modulus of elasticity [50 GPa; 90 GPa]
Fig. 9. Comparison of the model and Ibrahim et al. results [23] of the resuspended
Final stream velocity [9.8 m/s; 28.2 m/s]
fraction for different particle/surface combinations: a) stainless steel/glass; b) glass/
Flow acceleration [0.3 m/s2; 46 m/s2]
glass.
10 C. Habchi et al. / Building and Environment 105 (2016) 1e12

Fig. 10. Effect of a) the particle diameter and b) particle density on the variation of the resuspended fraction with free-stream velocity.

shows the variation of the detachment fraction with respect to the stream velocity (Fig. 11a). Furthermore, the rate of particle resus-
free-stream velocity for variable surface roughness standard devi- pension variation with time is increased with reduced adhesive
ation. As the surface roughness standard deviation increases, sur- forces leading to narrower velocities range to reach 100%
face average roughness is increased and thus the mean adhesive resuspension. For surface roughness standard deviation lower than
force is reduced (Eq. (2e)) leading to higher resuspension rates 1 Å, the threshold velocity significantly increased such that no
for the same free-stream velocity. Furthermore, higher the surface particle detachment is observed for stream velocities lower than
roughness standard deviation, higher the surface maximum 12 m/s at the experimental conditions of Ibrahim and Dunn [27]
roughness explaining the initiation of resuspension at lower free (Fig. 11a).

Fig. 11. Effect of the a) surface roughness standard deviation and b) surface Young modulus of elasticity on the variation of the resuspended fraction with free-stream velocity.
C. Habchi et al. / Building and Environment 105 (2016) 1e12 11

3.2.3. Final stream velocity and flow acceleration effects


The final free stream velocity is varied (9.8 m/s, 14.1 m/s, 28.2 m/
s) for the same time of exposure (0.613 s corresponding respec-
tively to accelerations of 16 m/s2, 23 m/s2, and 46 m/s2) and its
effect on the resuspension fraction with time is assessed. Fig. 12
illustrates the variation of resuspension fraction with time for
different final stream velocities. The higher the final stream velocity
is, the higher the shear stress value becomes along the exposure
time which explains the initiation of resuspension at earlier time
(Fig. 12). At any time, the resuspended fraction is highest for the
largest stream velocity. However, the threshold velocity decreases
with flow acceleration corresponding to higher final stream ve-
locity for same time exposure due to the delay in turbulence onset.

3.3. Sensitivity analysis

Fig. 12. Effect of the flow acceleration on the variation of the resuspended fraction From the performed parametric study, the most affecting pa-
with time. rameters on particle resuspension are: the particle diameter, sur-
face roughness and flow acceleration. One widely used parameter
to characterize particle resuspension is the threshold velocity
3.2.2.2. Surface Young modulus of elasticity. The surface Young defined as the velocity required for resuspending 50% of the par-
modulus of elasticity is varied in the range of Young modulus for ticles. Fig. 13 shows the variation of threshold velocity with the
glass [50 GPa; 90 GPa] [43] and its effect on the resuspension increase of particle diameter, surface roughness, and flow acceler-
fraction with time is investigated. Fig. 11b displays the variation of ation which are the factors that govern the resuspension condition.
resuspension fraction with free stream velocity for different surface A sensitivity analysis is done on each of these factors to assess its
Young modulus. As the particle Young modulus of elasticity in- effect on particle resuspension.
creases, the contact radius between the particle and the surface is Fig. 13a illustrates the variation of the threshold velocity with
reduced (Eqs. (3b) and (3c)) and thus the moment of adhesive force particle diameter. It is observed that the rate of increase of
is reduced (Eq. (3a)) facilitating particle resuspension. However, as threshold velocity is accentuated with reduced particle diameter.
observed in Fig. 11b, particle resuspension is nearly insensitive to For instance, the threshold velocity is much more sensitive to the
the glass surface Young modulus within its variation range. variation of particle diameter for diameters below 50 mm (Fig. 13a).
Therefore, the sensitivity of resuspension to fine particle diameters
is higher than its sensitivity to coarse particle diameters.
Fig. 13b represents the threshold velocity variation with respect
to surface roughness standard deviation. The rate of increase of
threshold velocity is strengthened with reduced surface roughness.
In fact, it is shown in Fig. 13b that threshold velocity sensitivity to
surface roughness is significantly increased for roughness standard
deviation lower than 6 Å. Hence, the sensitivity of particle resus-
pension to surface roughness is much higher for relatively
smoother surfaces.
Fig. 13c shows the effect of flow acceleration on the variation of
threshold velocity. The rate of decrease of threshold velocity is
higher for smaller flow acceleration. For example, the slope of
decrease of threshold velocity becomes significant for flow accel-
eration below 2 m/s2. Thus, the sensitivity of particle resuspension
to the flow acceleration is more pronounced for smaller
acceleration.

4. Conclusion

An analytical approach is followed to model the probabilistic


behavior of monolayer particle resuspension by transient flows and
the analytical model is coupled to a CFD model for computing the
surface shear stress variation. The resuspension model involves the
effect of turbulent bursts, surface roughness, airflow, and particle/
surface properties. The developed coupled models’ ability in
capturing the flow physics and predicting the transient resus-
pension fraction was validated by published literature data. Good
agreement was obtained between literature data and models with a
relative error of the order of 10% underlining the accuracy of the
developed models. Then, a parametric study and sensitivity anal-
ysis were performed for assessment of different variables effect on
Fig. 13. Effect of the a) particle diameter; b) Surface roughness standard deviation; and particle resuspension.
c) flow acceleration on the threshold velocity. With the increase of the particle diameter, the ratio of removal
12 C. Habchi et al. / Building and Environment 105 (2016) 1e12

to adhesive forces and moments increase. Therefore, resuspension Health Organ. 78 (9) (2000) 1078e1092.
[12] J.D. Spengler, K. Sexton, Indoor air pollution: a public health perspective,
starts at smaller free stream velocity and the variation in time of the
Science 221 (4605) (1983) 9e17.
rate of monolayer particle resuspension increases requiring lower [13] J. Sundell, On the history of indoor air quality and health, Indoor Air 14 (s7)
range of velocities to reach 100% resuspension. The same findings (2004) 51e58.
are observed with the increase of surface roughness standard de- [14] L. Morawska, Droplet fate in indoor environments, or can we prevent the
spread of infection? Indoor Air 16 (5) (2006) 335e347.
viation resulting in the reduction of the surface adhesive forces. On [15] L.L. Christianson, Building Systems, Room Air and Air Contaminant Distribu-
the other hand, particle density mainly affects the weight of the tion, American Society of Heating, Refrigerating and Air-Conditioning, 1989.
particle which is generally negligible compared to the removal and [16] C. Henry, J.P. Minier, Progress in particle resuspension from rough surfaces by
turbulent flows, Prog. Energy Combust. Sci. 45 (2014) 1e53.
adhesive forces. Thus monolayer particle resuspension is nearly [17] P. Lengweiler, P. Nielsen, A. Moser, P. Heiselberg, H. Takai, Deposition and
insensitive to particle density. resuspension of particles: which parameters are important?, in: Proceedings
As the final stream velocity increases, a higher shear stress value of Roomvent, 6th International Conference on Air Distribution in Rooms,
Stockholm, Sweden, vol. 1, 1998, pp. 317e323.
occurs along the same exposure time explaining the earlier initia- [18] D.J. Phares, G.T. Smedley, R.C. Flagan, Effect of particle size and material
tion of resuspension and the larger resuspended fraction for a given properties on aerodynamic resuspension from surfaces, J. Aerosol Sci. 31 (11)
period of exposure. Nevertheless, the threshold velocity decreases (2000) 1335e1353.
[19] G. Ziskind, M. Fichman, C. Gutfinger, Resuspension of particulates from sur-
with flow acceleration corresponding to higher final stream ve- faces to turbulent flowsdreview and analysis, J. Aerosol Sci. 26 (4) (1995)
locity for same time exposure due to the delay in turbulence onset. 613e644.
The most influential parameters on monolayer particle resus- [20] J.A. Garland, Some recent studies of the resuspension of deposited material
from soil and grass, in: H.R. Pruppacher, R.G. Semonin, W.G.W. Slinn (Eds.),
pension are: the particle diameter, surface roughness and flow
Precipitation Scavenging, Dry Deposition and Resuspension, Elsevier, New
acceleration. It has been shown that the sensitivity of resuspension York, NY, 1983, pp. 1087e1097.
to fine particle diameters is higher than its sensitivity to coarse [21] C. Mukai, J.A. Siegel, A. Novoselac, Impact of airflow characteristics on particle
particle diameters and its sensitivity to surface roughness is much resuspension from indoor surfaces, Aerosol Sci. Technol. 43 (2009)
1022e1032.
higher for relatively smoother surfaces. Furthermore, the sensi- [22] J. Qian, A.R. Ferro, Resuspension of dust particles in a chamber and associated
tivity of particle resuspension to the flow acceleration is more environmental factors, Aerosol Sci. Technol. 42 (2008) 566e578.
accentuated for smaller flow accelerations. [23] A.H. Ibrahim, P.F. Dunn, R.M. Brach, Microparticle detachment from surfaces
exposed to turbulent airflow: controlled experiments and modeling, J. Aerosol
In conclusion, the developed model was shown to be accurate in Sci. 34 (2003) 765e782.
predicting the variation of particle detachment with time by [24] M.W. Reeks, J. Reed, D. Hall, On the resuspension of small particles by a tur-
capturing the main physics affecting it. The validated developed bulent flow, J. Phys. D. Appl. Phys. 21 (1988) 574e589.
[25] M.W. Reeks, D. Hall, Kinetic models for particle resuspension in turbulent
model will be used in future work involving resuspension by flows: theory and measurement, J. Aerosol Sci. 32 (2001) 1e31.
oscillatory transient flows which have many engineering [26] M. Soltani, G. Ahmadi, On particle adhesion and removal mechanisms in
applications. turbulent flows, J. Adhes. Sci. Technol. 8 (7) (1994) 763e785.
[27] A.H. Ibrahim, P.F. Dunn, Effects of temporal flow acceleration on the detach-
ment of microparticles from surfaces, J. Aerosol Sci. 37 (10) (2006)
References 1258e1266.
[28] B.E. Boor, J.A. Siegel, A. Novoselac, Development of an experimental meth-
[1] Y. Kim, A. Gidwani, B.E. Wyslouzil, C.W. Sohn, Source term models for fine odology to determine monolayer and multilayer particle resuspension from
particle resuspension from indoor surfaces, Build. Environ. 45 (8) (2010) indoor surfaces, ASHRAE Trans. 117 (1) (2011).
1854e1865. [29] B. Elhadidi, H.E. Khalifa, Aerodynamic resuspension of particles due to a falling
[2] J.A. Bernstein, N. Alexis, C. Barnes, I.L. Bernstein, A. Nel, D. Peden, P.B. Williams, flat disk, Part. Sci. Technol. 31 (1) (2013) 35e44.
Health effects of air pollution, J. Allergy Clin. Immunol. 114 (5) (2004) [30] H.E. Khalifa, B. Elhadidi, Particle levitation due to a uniformly descending flat
1116e1123. object, Aerosol Sci. Technol. 41 (1) (2007) 33e42.
[3] B.E. Boor, J.A. Siegel, A. Novoselac, Monolayer and multilayer particle deposits [31] ANSYS Software: ANSYS Inc. http://www.ansys.com/ .
on hard surfaces: literature review and implications for particle resuspension [32] K.E. Jansen, A stabilized finite element method for computing turbulence,
in the indoor environment, Aerosol Sci. Technol. 47 (8) (2013) 831e847. Comput. Methods Appl. Mech. Eng. 174 (3) (1999) 299e317.
[4] I. Goldasteha, Y. Tianb, G. Ahmadia, A.R. Ferro, Human induced flow field and [33] K. Jansen, Large-eddy simulation of flow around a NACA 4412 airfoil using
resultant particle resuspension and transport during gait cycle, Build. Environ. unstructured grids, Annu. Res. Briefs (1996) 225e232.
77 (2014) 101e109. [34] Z. Zhang, Q. Chen, Comparison of the Eulerian and Lagrangian methods for
[5] N. Serfozo, S.E. Chatoutsidou, M. Lazaridis, The effect of particle resuspension predicting particle transport in enclosed spaces, Atmos. Environ. 41 (2007)
during walking activity to PM10 mass and number concentrations in an in- 5236e5248.
door microenvironment, Build. Environ. 82 (2014) 180e189. [35] M. Fichman, C. Gutfinger, D. Pnueli, A model for turbulent deposition of
[6] R.C. Oberoi, J.I. Choi, J.R. Edwards, J.A. Rosati, J. Thornburg, C.E. Rodes, Human- aerosols, J. Aerosol Sci. 19 (1) (1988) 123e136.
induced particle re-suspension in a room, Aerosol Sci. Technol. 44 (3) (2010) [36] A.V. Johansson, P.H. Alfredsson, On the structure of turbulent channel flow,
216e229. J. Fluid Mech. 122 (1982) 295e314.
[7] X. Zhang, G. Ahmadi, J. Qian, A. Ferro, Particle detachment, resuspension and [37] T.J. Hanratty, L.G. Chorn, D.T. Hatziavramidis, Turbulent fluctuations in the
transport due to human walking in indoor environments, J. Adhes. Sci. viscous wall region for Newtonian and drag reducing fluids, Phys. Fluids 20
Technol. 22 (5e6) (2008) 591e621. (10) (1977) S112eS119.
[8] B. Bin Zuo, K. Ke Zhong, Y. Yanming Kang, An experimental study on particle [38] M.E. O’Neill, A sphere in contact with a plane wall in a slow linear shear flow,
resuspension in a room with impinging jet ventilation, Build. Environ. 89 (July Chem. Eng. Sci. 23 (1968) 1293e1298.
2015) 48e58. [39] P.G. Saffman, The lift on a small sphere in a slow shear flow, J. Fluid Mech. 22
[9] I. Goldasteh, G. Ahmadi, A.R. Ferro, Wind tunnel study and numerical simu- (02) (1965) 385e400.
lation of dust particle resuspension from indoor surfaces in turbulent flows, [40] K.L. Johnson, K. Kendall, A. Roberts, Surface energy and contact of elastic
J. Adhes. Sci. Technol. 27 (14) (2013) 1563e1579. solids, Proc. R. Soc. A 324 (1971) 301e313.
[10] S. Wang, B. Zhao, B. Zhou, Z. Tan, An experimental study on short-time par- [41] W. Cheng, R.M. Brach, P.F. Dunn, Surface roughness effects on microparticle
ticle resuspension from inner surfaces of straight ventilation ducts, Build. adhesion, J. Adhes. 78 (2002) 929e965.
Environ. 53 (2012) 119e127. [42] M. De Podesta, Understanding the Properties of Matter, CRC Press, 2002.
[11] N. Bruce, R. Perez-Padilla, R. Albalak, Indoor air pollution in developing [43] S.J. Schneider, Engineered Materials Handbook, in: Ceramics and Glasses, vol.
countries: a major environmental and public health challenge, Bull. World 4, 1991.

You might also like