Building and Environment 105 (2016) 358e368

Contents lists available at ScienceDirect

Building and Environment

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

Finite panel method for the simulation of wind-driven rain

Chongwen Jiang*, Chenhao Xu, Zhenxun Gao, Chunhian Lee
National Laboratory for Computational Fluid Dynamics, School of Aeronautic Science and Engineering, Beihang University, Beijing, China

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

Article history: In present numerical methods for the wind-driven rain (WDR) simulation, a large computational work
Received 19 February 2016 for raindrop trajectories is required to obtain the distribution of the catch ratio because of the
Received in revised form randomness of the initial conditions of the raindrops. A new method named Finite Panel Method (FPM)
8 April 2016
that is employed to acquire the catch ratio distribution is proposed to reduce the calculation amount and
Accepted 11 April 2016
diminishes the randomness. In the FPM, raindrops ending in the corners of the panels which are divided
Available online 13 April 2016
on the surface are simulated, and then the distribution of catch ratio is acquired by a reconstruction
method. As a demonstration, catch ratio results by FPM in typical flow fields are presented, including the
Keywords:
Wind-driven rain
case of the three-dimensional flow field of power-law velocity profile, and cases of the flow around two-
Catch ratio and three-dimensional low-rise buildings. The FPM results are compared with these calculated by the
Numerical simulation classical method and literature. The results indicate that FPM can reduce the calculation amount of
Finite panel method raindrop trajectories with equivalent accuracy compared to the classical method.
Reconstruction © 2016 Published by Elsevier Ltd.

1. Introduction compared with semi-empirical methods. Hence, numerical

Studies of wind-driven rain (WDR) are important in the earth
science, building science and meteorology [1]. Predictions of the
WDR distribution can be used for the design of catchments, control
of water loss and soil erosion, as well as guidance for agriculture
and precaution against the forest fire [1]. In building science fields,
the WDR influences the hygrothermal performance of building
façades, and may cause surface soiling and mold growing, which
affect the durability of walls [2]. Moreover, the wetting of building
façades caused by WDR is also a boundary condition in building
heat-air-moisture (HAM) transport models [3].
Two parameters utilized to quantify the amount of WDR are the
specific catch ratio and catch ratio [4]. Measurements, semi-
empirical methods and numerical methods are main approaches
in building WDR researches. Blocken and Carmeliet [1] reviewed
the application of these methods. Compared with measurements,
numerical methods are easier to perform and consume less time.
Besides, numerical methods are not confined to weather condi-
tions, which are basic requirements for measurements. On the
other hand, numerical methods can provide more accurate results
* Corresponding author. National Laboratory for Computational Fluid Dynamics,
School of Aeronautics Science and Engineering, Beihang University, 37th Xueyuan
Road, Haidian District, Beijing 100191, China.
E-mail address: cwjiang@buaa.edu.cn (C. Jiang).
methods are widely used in building WDR researches.
Currently, numerical methods for WDR are based on the Euler-
Lagrange model or the EulereEuler model. The principle of the
Euler-Lagrange model is to calculate the motion of discrete rain-
drops in continuous wind flow field. Early in 1970s, researches have
been performed in simulating the raindrop trajectories in flow field
around buildings [5]. In the early 1990s, Choi [6,7] systematically
proposed the numerical method for building WDR simulation
based on the Euler-Lagrange model. This method is on the basis of
steady flow field, and the raindrop trajectories are calculated to
determine the end points in building façades. After that, the WDR
parameters on the wall are obtained based on the trajectories [6,7].
Blocken and Carmeliet extended Choi's method and established
numerical method for spatial and temporal distribution of WDR [4].
To validate the numerical methods, measurements of WDR on low-
rise buildings are conducted by Blocken and Carmeliet [8e10], Nore
et al. [11] and Abuku et al. [12]. Simulations of WDR under different
rainfall conditions are also performed, and the results validate the
numerical methods. At present, applications of the numerical
methods based on the Euler-Lagrange model on complex WDR
problems have been studied. Janssen et al. [3] combined WDR
numerical results with the HAM model. The spatial and temporal
distribution of WDR is calculated and used for HAM boundary
conditions. Briggen et al. [13] measured and calculated the WDR
distribution on a monumental tower. Van Hooff et al. [14] simulated
the WDR on 12 different generic stadium configurations, and

http://dx.doi.org/10.1016/j.buildenv.2016.04.010
0360-1323/© 2016 Published by Elsevier Ltd.
 C. Jiang et al. / Building and Environment 105 (2016) 358e368 359

explained the influence of configurations and roof slope on the area 2. Definitions of wind-driven rain parameters and the
of the stand on the catch ratio distribution. Foroushani et al. [15] numerical method
calculated WDR on a low-rise cubic building under various wind
and rain conditions and investigated the effects of the overhangs 2.1. The numerical method for wind-driven rain parameters
with various sizes.
Methods based on the EulereEuler model regard both the wind The specific catch ratio and catch ratio are two generally used
and rain as continuous phases, and the rain field is solved together WDR parameters [4]. The numerical method based on the Euler-
with the wind field. The Euler-Euler methods can provide the dis- Lagrange model to determine the WDR parameters in steady flow
tribution of volume fraction of rain. The EulereEuler model has field comprises 3 steps (shown in Fig. 1): firstly, the flow field
some advantages in determining the initial conditions of rain and around the building without raindrops is simulated; after that, the
calculating the specific catch ratio and catch ratio [16]. Huang and Li initial conditions of raindrops are determined, including the
[17] studied the distribution of catch ratio on building envelopes diameter, initial position and initial velocity, and the raindrop tra-
based on the EulereEuler model, and the results agreed well with jectories are calculated by solving the 3DOF (Degree of Freedom)
measurements. Kubilay et al. [16] calculated the WDR on a monu- mass-point equations of motion, and the force in equations is given
mental tower and compared the results with measurements and in terms of the acquired flow field; finally, the terminal position on
results of the Euler-Lagrange model. It is indicated that numerical the wall and the corresponding initial position of raindrops are
results agreed with measurements at the higher parts of the tower. statistically obtained. According to the definitions [4], the specific
Wang et al. [18] simulated the catch ratio distribution on an array of catch ratio of a sub-district is the ratio of the initial area and the
low-rise buildings under different oblique winds, and the results terminal area surrounded by the raindrops, and the distribution of
agreed well with experiments. the catch ratio can be acquired by combining the specific catch ratio
In real wind and rain situations, the volume fraction of rain is and the function of the volume fraction of rain.
less than 10%. The recorded maximum rainfall intensity is 38.1 mm/ Based on the theory of the Euler-Lagrange model, the raindrops
min [19], which equals to 2286 mm/h. However, the volume frac- are considered to be discrete particles in the trajectory calculation
tion of rain is only 4.2
105 in this circumstance based on the [21]. According to the Newton's Second Law, the 3DOF mass-point
raindrop size distribution function by Best [20]. Generally, the equations of motion of raindrops in Cartesian coordinate system are
Euler-Lagrange model is considered to be more suitable for the as follows:
cases where the volume fraction of the discrete phase is less than
10% [21]. dup 18m Cd Re

Nevertheless, the classical methods based on the Euler-Lagrange
¼ , ,cos 4, u  up
dt rp d2 24
model require a large computational work for raindrop trajectories
[1]. In the classical methods, the building façade is divided into sub- dx
¼ up
districts by end points of raindrops on the wall, and then the catch dt
ratio of each sub-district is calculated by raindrop trajectories [10]. dvp 18m Cd Re

The catch ratio of a sub-district can be regarded as an average of the ¼ , ,cos j, v  vp
dt rp d 24
2
value of all points in the district, thus a more accurate WDR result
requires a smaller area for each district. However, in the classical dy (1)
¼ vp
method the initial conditions of the raindrops are random, and thus dt
!
the terminal positions of the raindrops can not be determined
dwp 18m Cd Re
 r
before the simulation. Therefore, a large number of raindrops are ¼ , ,cos g, w  wp  g 1  a
dt rp d2 24 rp
required to ensure that sufficient raindrops can impinge on the
wall, thus enlarging the calculation amount. dz
In this paper, a new method named Finite Panel Method (FPM) ¼ wp
dt
for the WDR simulation based on the Euler-Lagrange model is qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r d
2
2
2
proposed. The goal of the proposed method is to reduce the Re ¼ a u  up þ v  vp þ w  wp
calculation amount for raindrop trajectories. In the FPM the target
m
surface is divided into a proper number of panels, and then calcu-
where rp and ra denote the density of water and air, respectively; x,
lates the raindrop trajectories ending on the corner points of the
y, z denote the coordinates of raindrops; u, v, w represent the wind
panels. Since the surface is divided, meaning that the terminal
velocity; up, vp, wp are the raindrop velocity; 4, j, g represent the
positions of the raindrops are determined, the randomness of the
initial conditions of the raindrops is diminished. Subsequently, the
catch ratio of each panel can be obtained, and the catch ratio dis-
tribution on the surface is acquired by a reconstruction function.
Compared with the classical Euler-Lagrange methods, the FPM can
reduce the calculation amount because only those raindrops
ending on the corners are calculated.
This paper is organized as follows. In Section 2, the numerical
method for WDR parameters based on the Euler-Lagrange model is
introduced, and the code employed in the present research is
validated. Then a new numerical method named Finite Panel
Method is proposed in Section 3. In Section 4 the FPM is applied to
three cases. The accuracy of the method proposed in the present
paper is validated, and the sensitivity of the panel division is
analyzed. Finally, in Section 5 the present work is summarized and
concluded.
Fig. 1. Procedures of the CFD method for wind-driven rain parameters.
360 C. Jiang et al. / Building and Environment 105 (2016) 358e368

angle between the rain to wind relevant velocity and the direction
of x, y, z, respectively; m is the air viscosity; d is the raindrop
diameter; Re is the Reynolds number based on the relevant ve-
locity; g is the gravity acceleration, which is set to 9.8 m/s2; Cd is the
drag coefficient of raindrops, and the formula yielded by Morsi and
Alexander is adopted in this paper [22]:

a2 a
Cd ¼ a1 þ þ 3 (2)
Re Re2

where the constants a1, a2, a3 are shown in Table 1.

2.2. Numerical method for wind flow field and validation

2.2.1. Numerical method for wind flow field

In common wind conditions, the flow around buildings is Fig. 2. Experiment setup.
incompressible. Besides, the Re of the flow around buildings in
atmosphere is over 107, which indicates that the flow is turbulent.
Hence, in this research the incompressible Reynolds Average
Navier-Stokes (RANS) [23] equations are solved for the flow field
around buildings.
In this paper, the numerical code IFAST (Incompressible Flow
Analysis and Simulation Tool) developed by National Laboratory for
Computational Fluid Dynamics (NLCFD) is employed for solving the
RANS equations. The solution strategy for the pressure-velocity
coupling is the SIMPLE scheme, and the SST k-u model is chosen
as the turbulent model.
To validate the numerical methods and the computation pro-
gram employed in this paper, the wind tunnel experimental results
of blunt bodies are utilized as reference, and the experiment cases
are simulated by the above methods.

2.2.2. Case of the experiment of a blunt body with the ground effect
In this section, the experiment conducted by Chiu and Squire
[24] is utilized to validate the numerical method for complex
configurations with concerning of the windward and leeward sides,
and the ground effect. The experiment setup is shown in Fig. 2. The
model consists of a head and a body. The height of the body is
H ¼ 0.125 m and the length of the head is 1.28 H. The gap between
the bottom of the model and the ground is 0.15 H. The definitions of
the coordinate and the angle are shown in Fig. 3. The x coordinate of
the front point of the head is 0, and the middle point of the cross Fig. 3. Definitions of the coordinate and the angle.
section in the windward surface is q ¼ 0 . Three sections depicted in
Fig. 3 (a) are x ¼ 1.0 H, x ¼ 2.5 H and x ¼ 7.5 H, where the pressure cross sections. The figure depicts that the numerical results agree well
coefficients are to be calculated and compared with the with the experimental results in both windward side (angle from 0 to
experiment. 60 and from 300 to 360 ) and leeward side (angle from 60 to 300 ).
In the experiment, the freestream velocity is 35 m/s, and the Re The computational results of the case indicate that accurate re-
based on the height of the model is 3
105. A trip wire is attached sults can be obtained by the numerical methods employed for the
in the corner in the windward direction, thus the flow around the flow around blunt bodies on the ground, and the surface pressure
model is turbulent [24]. In this case, the yaw angle of the freestream coefficients can be acquired correctly. Though the freestream con-
is 90 , meaning that the flow is perpendicular to the windward ditions are different from the generally adopted power law flow
surface of the model. profile in the simulation of the atmospheric boundary layer and the
Fig. 4 illustrates the pressure coefficients distribution of several height of the model is smaller than the common used reference
height, the present case simulates the situation with the effect of the
ground boundary layer and turbulence, and the flow field is
Table 1
Constants in the formula of drag coefficient for various Re complicated. Hence, the results could validate the appropriation of
the numerical code IFAST for the flow around buildings, and the
Re a1 a2 a3
program is to be employed in the following wind-driven rain
0e0.1 0 24.0 0 simulation.
0.1e1.0 3.69 22.73 0.0903
1.0e10 1.222 29.1667 3.8889
10e100 0.6167 46.5 116.67 2.3. Numerical method for wind-driven rain parameters and
100e1000 0.3644 98.33 2778 validation
1000e5000 0.357 148.62 4.75
104
5000e10000 0.46 490.546 5.787
105 2.3.1. Numerical method for wind-driven rain parameters
10000e50000 0.5191 1662.5 5.4167
106
To calculate the specific catch ratio and catch ratio, the terminal
 C. Jiang et al. / Building and Environment 105 (2016) 358e368 361

Fig. 4. Comparison of the pressure coefficient distribution in various sections.

velocity of raindrops and the volume fraction of rain should be

determined. In this research, the terminal velocity of raindrops is dFðdÞ
f ðdÞ ¼ (5)
determined by the fitting formula of the experimental results ob- dd
tained by Gunn and Kinzer [25]:

   2.3.2. Case of the two-dimensional low-rise building

Vt ðdÞ ¼ 9:4 1  exp 0:557d1:15
where Vt denotes the terminal velocity.
The raindrop size distribution function proposed by Best [20] is
adopted to calculate the volume fraction of rain:
(3) To validate the accuracy of the above methods for WDR pa-
rameters, the two- and three-dimensional low-rise building
models are simulated, and the catch ratio of a given point is
calculated. The model simulated by Blocken and Carmeliet [10] is
chosen as the two-dimensional low-rise building. As shown in

  n 
d
FðdÞ ¼ 1  exp  ; a ¼ AIp (4)
a

where F is the volume fraction of raindrops with diameter less than

d; A, n and p are constants, which are 1.30, 2.25 and 0.232,
respectively; I denotes the rainfall intensity. Differentiating Equa-
tion (4) with regard to the diameter yields the volume fraction of
rain: Fig. 5. Two-dimensional low-rise building model.
362 C. Jiang et al. / Building and Environment 105 (2016) 358e368

Fig. 5, the length of the model is 7.2 m, and the height is 3.95 m. The
catch ratio is calculated for the district between the heights from
3.65 m to 3.85 m in the windward surface. The freestream flow
profile represented by a power law can be written as:

 
h a
u ¼ u0 (6)
h0

where h0 denotes the reference height; u0 is the reference velocity;

a is the exponent. In this case, h0 ¼ 10 m, u0 ¼ 10 m/s and a ¼ 0.176.
The rainfall intensities are 1、2、3、4、5、6、8、10、12、15、
20、25、30 mm/h.
Fig. 6 indicates the numerical results as well as the results of the
literature. The numerical results show that the catch ratio decreases Fig. 7. Three-dimensional low-rise building model.
as the rainfall intensity increases, which corresponds with the
literature. The largest discrepancy between the numerical and
literature results is about 2.1%, when the rainfall intensity is 1 mm/
h.
1.2

1.15 Calculation Blocken & Carmeliet

Present
2.3.3. Case of the three-dimensional low-rise building
The cubic low-rise building model calculated by Blocken and
1.1
Carmeliet [26] is chosen as the three-dimensional model. Fig. 7 il-
lustrates the model to be simulated. The edge length of the cube is
10 m, and the target point is the middle point of the side in the 1.05
η

windward surface. The velocity profile of the freestream obeys the

power-law. The reference height is 10 m, the reference velocity is
10 m/s, and the exponent is 0.15. The direction of the freestream is 1

perpendicular to the windward surface. The rainfall condition is the

same as the two-dimensional low-rise building case in Section
0.95
2.3.2.
Fig. 8 depicts the catch ratio of the given point under different
rainfall intensities. It can be seen that, the numerical results indi- 0.9
0 5 10 15 20 25 30
cate the same trend as the literature [26]. A larger discrepancy
I (mm/h)
appears under smaller rainfall intensity, and the largest discrep-
ancy is about 4.3% when the rainfall intensity is 4 mm/h. Fig. 8. Catch ratio on the three-dimensional low-rise building under various rainfall
The numerical results of the two- and three-dimensional low- intensity.
rise buildings indicate that the numerical methods employed can
obtain a correct catch ratio result for a given point. In the following
research, the methods are to be utilized for the catch ratio
calculation. 3. Finite Panel Method for the wind-driven rain simulation

3.1. Basic principles of the Finite Panel Method

1.3

The classical method for calculating the specific catch ratio and
Calculation by Blocken & Carmeliet catch ratio requires a large amount of raindrop trajectories that can
Present
1.26 cover the whole wall to calculate the WDR parameters for each sub-
district, thus the computational work for trajectories is large. The
Finite Panel Method proposed in this paper aims at reducing the
1.22 calculation amount of trajectories and obtaining accurate distri-
bution of the catch ratio.
In wind and rain conditions, the wind flow field around build-
η

1.18
ings is continuous, and the raindrops are driven by the wind flow
field and impinge on the façades. When the number of raindrops
impinging on the wall is sufficient, the distribution of the catch
ratio can be considered continuous as well. However, in numerical
1.14
simulation the number of raindrops to be calculated is finite, thus
the simulation is actually discrete. The wall is discretized to several
sub-districts, and the specific catch ratio and catch ratio of the sub-
1.1
0 5 10 15 20 25 30 district represent the average of the value of each point in the area.
I (mm/h) As the number of raindrops increases, the area of single sub-district
decreases, which makes the catch ratio of the sub-district
Fig. 6. Catch ratio on the two-dimensional low-rise building under various rainfall
intensity.
approximate the value of the local point:
 C. Jiang et al. / Building and Environment 105 (2016) 358e368 363

Shi
hi ¼ lim (7)
Swi /0 Swi

where hi represents the catch ratio of the local point i, Swi, Shi are the
terminal area around the point i and the corresponding initial area.
To obtain a more accurate distribution of the catch ratio, the
classical method adopts the strategy of increasing the number of
raindrops, which can acquire a catch ratio closer to the local value.
In the CFD simulation of the boundary layer flow, a similar method
is employed. The scale of the grid normal to the wall in the near
wall region is decreased to obtain a precise flow profile in the
boundary layer. However, another strategy exists employing a wall
function to simulate the flow profile of the boundary layer, and in
this situation the scale of the grid in the near wall region can be
larger [27]. Based on the same thoughts, the distribution of the
catch ratio can be acquired by a reconstruction function which
describes the catch ratio distribution of a local panel. By employing
the reconstruction function, the area of a single panel can be larger,
and only a small number of raindrops need to be calculated to get
the catch ratio of each panel. When the reconstruction function is
continuous, the distribution of the catch ratio is then continuous. In
this paper, the Finite Panel Method is proposed based on the above
principles.

3.2. Rationale of the Finite Panel Method

In Finite Panel Method, the target façade is divided into

adequate panels. The catch ratio of each panel is calculated by the
definition. Whereafter, the catch ratio of the corner points of the
panels are obtained by a reconstruction method. The thoughts of
this method are similar to the Finite Element Method (FEM) [28],
which also divides a continuous field into discrete districts and
employs a function to express the distribution in each district.
Compared with the classical method, the Finite Panel Method
simulates less raindrop trajectories. Moreover, the randomness of
the initial conditions could also be diminished. In the classical
method, the positions and number of raindrops are determined
iteratively to acquire a continuous and accurate distribution on the
façade, especially for some areas like the corners of sides. Never-
theless, in the Finite Panel Method the division of panels actually
determines the initial conditions of raindrops, thus the randomness
is diminished to a certain extent.
Fig. 9. Procedure of the finite panel method.
reconstruction is conducted in a staggered grid comprising a virtual
grid consisting of the center points and the panels of the wall.
As shown in Fig. 10, the grid with solid lines represents the
panels of the wall, N1 ~ N9 denote the corners of the panels, and
C1 ~ C4 represent the center points of the panels. In the recon-
struction, the values of the corner points are obtained based on the
values of points C1 ~ C4 and the reconstruction function. Generally,
when the “domain” for reconstruction consists of n
n nodes and
the format of the reconstruction function is the Lagrange poly-
nomials, the reconstruction of the point (i, j) can be written as:

3.3. Procedures of the Finite Panel Method

According to the analysis above, the procedures of the Finite

Panel Method can be summarized as follows: (1) Solving the RANS
equations to obtain the flow field around the building; (2) Dividing
the wall into several panels; (3) For each panel, calculating the
raindrop trajectories that end on the corners of the panel; (4)
Calculating the specific catch ratio and the catch ratio based on the
terminal area and the initial area around by raindrops, and the
value represents that of the center of the panel; (5) Determining the
reconstruction function; (6) Applying the reconstruction function
to calculate the catch ratio of the corner points and the distribution
of the catch ratio based on the catch ratio of the centers of each
panel. Fig. 9 illustrates the procedure of the Finite Panel Method.

3.4. Reconstruction method

In the FPM, the catch ratio calculated based on the raindrop

trajectories of the corner points represents the value of the center
point, and the catch ratio of the corner points are calculated by
reconstructing the values of the center points. Hence, the Fig. 10. Reconstruction in the finite panel method.
364 C. Jiang et al. / Building and Environment 105 (2016) 358e368
P
n P
n
f ði; jÞ ¼ Lðp; qÞf ðp; qÞ
p¼1 q¼1
n x x z z
P
n P i l j k (8)
Lðp; qÞ ¼
l¼1 k¼1 xp  xl zq  zk
lsp ksq
where i and j are the numbers of the points on the wall; p and q are
the numbers of the nodes of the “domain”; f(i, j) and f(p, q) repre-
sent the values of the point (i, j) and (p, q), respectively; x denote the
coordinate in the i-direction and the p-direction, and the z denote
the coordinate in the j-direction and the q-direction. Based on the
theoretical analysis, the reconstruction above has an accuracy of
nth order.
4. Catch ratio results of typical flow fields by the Finite Panel
Method
Applying the Finite Panel Method proposed above, the catch
ratio on the outlet plane of the three-dimensional flow field of
power law velocity profile is calculated, together with the catch
ratio on the façade of the two-dimensional and three-dimensional
buildings. The distributions of the catch ratio of the above flow
fields are analyzed, and the accuracy of the FPM for the catch ratio
calculation is validated. According to the definition and numerical
procedures of the catch ratio, the rainfall intensity determines the
volume fraction of the raindrop with certain diameter but has no
effects on the distribution of raindrops on the façade. Hence, the
rainfall intensity is set to 1 mm/h in all of the three cases.
4.1. Case of the three-dimensional flow field of power law velocity
profile
In the simulation of wind-driven rain on buildings, the power
law profile is generally adopted as the freestream velocity profile.
Hence the catch ratio on the outlet plane of the three-dimensional
flow field of power law velocity profile is calculated to validate the
accuracy of the FPM in this flow field. In the calculation, the flow
field is undisturbed. The simulation domain is shown in Fig. 11,
where x 2 (0 m, 300 m), y 2 (0 m, 50 m), and z 2 (0 m, 200 m). The
Fig. 11. Computational domain of the three-dimensional flow field of power-law ve-
locity profile.
freestream is along the positive x-direction, and the z coordinate
denotes the height. The reference height of the freestream velocity
profile is 10 m, the reference velocity is 10 m/s, and the exponent is
0.176. The rainfall intensity is 1 mm/h.
The catch ratio distribution on the outlet plane is calculated by
both of the classical method and the FPM, and the district for
calculation is y 2 (20 m, 30 m) and z 2 (0 m, 10 m). In the classical
method, trajectories of 200 and 1000 raindrops are simulated.
While in the FPM, the district is divided into 8
8 panels (shown in
Fig. 12). According to the procedures of the FPM, trajectories of only
81 raindrops are required to be simulated.
A bilinear function is chosen as the reconstruction function, and
the “domain” has 2
2 nodes. For a point in the “domain”, the
reconstruction function can be expressed as:
X
2 X
2    
hðy; zÞ ¼ L yp ; zq h yp ; zq (9)
p¼1 q¼2
where
  X
2 X
2
y  yl z  zk
L yp ; zq ¼ (10)
yp  yl z q  z k
l¼1 k¼1
lsp ksq
The results of the classical method and the FPM are shown in
Fig. 13 and Fig. 14, respectively. Fig. 13 (a) and Fig. 13 (b) illustrate
the catch ratio distribution by the classical method with the
calculation of 200 and 1000 raindrops, respectively. It can be seen
that the catch ratio distribution may appear undulate when the
number of raindrops is not sufficient. As the number of raindrops
increases, the catch ratio distribution is clear and increases from the
bottom to the top. The minimum and maximum catch ratio
calculated by the classical method is 1.59 and 2.29, respectively.
Fig. 14 depicts the results by the FPM. It can be seen that the FPM
correctly demonstrate the trend of the catch ratio distribution
comparing to the classical method. Moreover, the minimum and
maximum catch ratio calculated by the FPM is 1.61 and 2.29
respectively, and the maximum discrepancy between the FPM and
the classical method is 1.2%. The comparison between the classical
method and the FPM indicates that the FPM can accurately obtain
Fig. 12. Panel division in the Finite Panel Method.
 C. Jiang et al. / Building and Environment 105 (2016) 358e368 365

Fig. 14. Catch ratio distribution in the outlet plane by the Finite Panel Method.

into 5 panels, and the trajectories of only 6 raindrops ending on the

corner points are simulated. The linear function is chosen as the
reconstruction function which can be written as:

y  y2 y  y1
hðyÞ ¼ hðy1 Þ þ hðy2 Þ (11)
y1  y2 y2  y1
Fig. 15 depicts the catch ratio distribution on the wall by both of
the classical method and the FPM. It can be seen that the distri-
bution by the FPM agrees well with that by the classical method.
Besides, the results indicate that the catch ratio on the two-
dimensional building façade nearly increases linearly with the
height in the power-law wind field.

4.3. Case of the three-dimensional building façade in the power-law

flow field

In the three-dimensional case, the catch ratio distribution on the

windward façade of the building is calculated. The model and

1.8

Classical method
Finite Panel Method

1.5
Fig. 13. Catch ratio distribution in the outlet plane by the classical method.

the catch ratio distribution on the outlet plane of the three-

1.2
η

dimensional flow field of power-law velocity profile.

4.2. Case of the two-dimensional building façade in the power-law

flow field 0.9

Considering the flow field around the building, the catch ratio
on the building façades is calculated. In the two-dimensional case,
the model and freestream conditions are the same as the case of 0.6
0 1 2 3 4
Blocken and Carmeliet [10] in Section 2.3.2, and the rainfall in-
y (m)
tensity is 1 mm/h.
In the classical method, trajectories of 150 raindrops impinging Fig. 15. Catch ratio distribution on the two-dimensional low-rise building in power-
on the wall are simulated. However, in the FPM the wall is divided law profile wind field.
366 C. Jiang et al. / Building and Environment 105 (2016) 358e368

freestream conditions are the same as the cubic low-rise building

case of Blocken and Carmeliet [26] in Section 2.3.3, and the rainfall
intensity is 1 mm/h. In the FPM, the wall is divided into 20
20
panels. The simulation domain and the division of the wall are
shown in Fig. 16. The reconstruction is chosen as expressed in
Equation (9).
Fig. 17 depicts the results by Blocken and Carmeliet [26] and the
FPM. In the vertical direction, the catch ratio increases as the height
increases, and in the horizontal direction the catch ratio is sym-
metrically distributed and increases from the center to the side. The
maximum catch ratio calculated in literature is 1.72, and the pre-
sent is 1.74. The discrepancy is 1.1%.
Comparison of the contour lines depicts that the contour lines
by the FPM agree well with that by the classical method, except for
the contour lines of 0.8 and 1.0. The discrepancy may be caused by
the difference of the flow fields calculated. The results indicate that
the FPM can correctly acquire the catch ratio distribution for three-
dimensional buildings.

Fig. 17. Catch ratio distribution on the three-dimensional low-rise building.

4.4. Sensitivity analyses of the panel division

The influence of the panel division on the catch ratio distribu-

Fig. 16. Computational model and panel division of the three-dimensional low-rise
building in power-law profile wind field.
tion is to be investigated by employing the three-dimensional
building case in Section 4.3. Three different divisions of panels
are adopted, which are 6
6 panels, 10
10 panels and 30
30
panels. Fig. 18 depicts the catch ratio distribution on the façade. It
can be seen that the distributions of the catch ratio are similar
among these cases of different panel divisions, and the difference
mainly appears near the sides and corners of the façade. Fig. 19
illustrates the catch ratio distribution along vertical and horizon-
tal center lines on the façade. Results show that a considerable
discrepancy exists between the division of 6
6 panels and 20
20
panels. The maximum discrepancy along the vertical center line
between the 6
6 panels and 20
20 panels is 9.8%, and 6.8% along
the horizontal line. As the number of panels increases, the
discrepancy diminishes.
The above results indicate that for the present building config-
uration a 20
20 panels is sufficient to obtain a correct catch ratio
distribution. Comparatively, in the classical method thousands of
raindrops need to be simulated. It can be seen that the FPM could
reduce remarkable calculation amount.
 C. Jiang et al. / Building and Environment 105 (2016) 358e368 367

1.4

6 × 6 panels
10 × 10 panels
20 × 20 panels
1.2 30 × 30 panels

η
0.8

0.6

0.4
0 2 4 6 8 10
z (m)

(a) Catch ratio distribution along the vertical center

line (y = 5.0m) on the façade
1.1

6 × 6 panels
10 × 10 panels
20 × 20 panels
30 × 30 panels
1

0.9
η

0.8

0.7
0 2 4 6 8 10
y (m)

(b) Catch ratio distribution along the horizontal center

line (z = 5.0m) on the façade
Fig. 19. Catch ratio distribution along center lines on the façade.

However, the above analyses also reveal that the panel division
influences the catch ratio distribution on the façade of a cubic
building. Therefore, when applying the FPM to a more complex
building, such as complicated façades or façades with a roof over-
hang, different panel divisions should be comparatively analyzed,
which is similar to the mesh independence analysis in the
Computational Fluid Dynamics (CFD). Moreover, the criterion of the
panel division for complex buildings needs to be studied.

5. Conclusions

Fig. 18. Catch ratio distribution of various panel divisions. A method based on the Euler-Lagrange model named Finite
Panel Method for calculating the wind-driven rain parameters on
building façades is proposed in this paper. The FPM can reduce the
number of raindrops required to be simulated comparing to the
classical method and obtain the distribution of the catch ratio. In
the FPM, the wall is divided into several panels, and the catch ratio
368 C. Jiang et al. / Building and Environment 105 (2016) 358e368
of each panel is acquired by calculating the raindrop trajectories
ending on the corner points of the panel. The catch ratio of the
corner points and the distribution of the catch ratio are obtained by
the reconstruction function.
The catch ratio of typical flow fields are calculated by the FPM. In
the case of the three-dimensional flow field of power law velocity
profile, trajectories of 1000 raindrops are simulated in the classical
method. While in the FPM, the district is divided into 8
8 panels
and only 81 raindrops need to be simulated, meaning that the
calculation amount has a more than 90% reduction in this case. The
maximum discrepancy between the FPM and the classical method
is 1.2%. In the calculation of the catch ratio on the two-dimensional
building façade, trajectories of 150 raindrops are simulated by the
classical method, but in the FPM the wall is divided into only 5
panels. The results indicate that the FPM can accurately obtain the
catch ratio distribution. Finally the catch ratio on the three-
dimensional building façade is calculated. The façade is divided
into 20
20 panels, and the discrepancy of the maximum catch
ratio between the FPM and the literature is 1.1%. Besides, three
more cases of different panel divisions are calculated, and the re-
sults show that panel division influences the obtained catch ratio
distribution. The cases of typical flow fields demonstrate that the
FPM can reduce the computational work for raindrop trajectories
and obtain the catch ratio distribution at the same time. Never-
theless, the criterion of the panel division for complex building
configurations need to be further studied.
References
[1] B. Blocken, J. Carmeliet, A review of wind-driven rain research in building
science, J. Wind Eng. Ind. Aerodyn. 92 (2004) 1079e1130.
[2] A. Kubilay, D. Derome, B. Blocken, J. Carmeliet, CFD simulation and validation
of wind-driven rain on a building façade with an Eulerian multiphase model,
Build. Environ. 61 (2013) 69e81.
[3] H. Janssen, B. Blocken, S. Roels, J. Carmeliet, Wind-driven rain as a boundary
condition for HAM simulations: analysis of simplified modeling approaches,
Build. Environ. 42 (2007) 1555e1567.
[4] B. Blocken, J. Carmeliet, Spatial and temporal distribution of driving rain on a
low-rise building, Wind Struct. 5 (2002) 441e462.
[5] G.G. Rodgers, G. Poots, J.K. Page, W.M. Pickering, Theoretical predictions of
rain drop impaction on a slab type building, Build. Sci. 9 (1974) 181e190.
[6] E.C.C. Choi, Simulation of wind-driven-rain around a building, J. Wind Eng.
Ind. Aerodyn. 46&47 (1993) 721e729.
[7] E.C.C. Choi, Determination of wind-driven-rain intensity on building faces,
J. Wind Eng. Ind. Aerodyn. 51 (1994) 55e69.
[8] B. Blocken, J. Carmeliet, High-resolution wind-driven rain measurements on a
low-rise building-experimental data for model development and model
validation, J. Wind Eng. Ind. Aerodyn. 93 (2005) 905e928.
[9] B. Blocken, J. Carmeliet, On the accuracy of wind-driven rain measurements on
buildings, Build. Environ. 41 (2006) 1798e1810.
[10] B. Blocken, J. Carmeliet, Validation of CFD simulations of wind-driven rain on a
low-rise building façade, Build. Environ. 42 (2007) 2530e2548.
[11] K. Nore, B. Blocken, B.P. Jelle, J.V. Thue, J. Carmeliet, A dataset of wind-driven
rain measurements on a low-rise test building in Norway, Build. Environ. 42
(2007) 2150e2165.
[12] M. Abuku, B. Blocken, K. Nore, J.V. Thue, J. Carmeliet, S. Roels, On the validity
of numerical wind-driven rain simulation on a rectangular low-rise building
under various oblique winds, Build. Environ. 44 (2009) 621e632.
[13] P.M. Briggen, B. Blocken, H.L. Schellen, Wind-driven rain on the façade of a
monumental tower: numerical simulation, full-scale validation and sensitivity
analysis, Build. Environ. 44 (2009) 1675e1690.
[14] T. Van Hooff, B. Blocken, M. Van Harten, 3D CFD simulations of wind flow and
wind-driven rain shelter in sports stadia: influence of stadium geometry,
Build. Environ. 46 (2011) 22e37.
[15] S.S.M. Foroushani, H. Ge, D. Naylor, Effects of roof overhangs on wind-driven
rain wetting of a low-rise cubic building: a numerical study, J. Wind Eng. Ind.
Aerodyn. 125 (2014) 38e51.
[16] A. Kubilay, D. Derome, B. Blocken, J. Carmeliet, CFD simulation and validation
of wind-driven rain on a building façade with an Eulerian multiphase model,
Build. Environ. 61 (2013) 69e81.
[17] S.H. Huang, Q.S. Li, Numerical simulations of wind-driven rain on building
envelopes based on Eulerian multiphase model, J. Wind Eng. Ind. Aerodyn. 98
(2010) 843e857.
[18] H. Wang, X.Z. Hou, Y.C. Deng, Numerical simulations of wind-driven rain on
building façades under various oblique winds based on Eulerian multiphase
model, J. Wind Eng. Ind. Aerodyn. 142 (2015) 82e92.
[19] H.N. Li, Y.M. Ren, H.F. Bai, Rain-wind-induced dynamic model for transmission
tower system, in: Proceedings of the 13th CSEE, 2007.
[20] A.C. Best, The size distribution of raindrops, Q. J. Roy. Meteor. Soc. 76 (1950)
16e36.
[21] H. Hangan, Wind-driven rain studies. A C-FD-E approach, J. Wind Eng. Ind.
Aerodyn. 81 (1999) 323e331.
[22] S.A. Morsi, A.J. Alexander, An investigation of particle trajectories in two-
phase flow systems, J. Fluid Mech. 55 (1972) 193e208.
[23] H. Lubcke, St Schmidt, T. Rung, F. Thiele, Comparison of LES and RANS in bluff-
body flows, J. Wind Eng. Ind. Aerodyn. 89 (2001) 1471e1485.
[24] T.W. Chiu, L.C. Squire, An experimental study of the flow over a train in a
crosswind at large yaw angles up to 90 , J. Wind Eng. Ind. Aerodyn. 45 (1992)
47e74.
[25] R. Gunn, G.D. Kinzer, The terminal velocity of fall for water droplets in stag-
nant air, J. Meteor 6 (1949) 243e248.
[26] B. Blocken, J. Carmeliet, On the validity of the cosine projection in wind-driven
rain calculations on buildings, Build. Environ. 41 (2006) 1182e1189.
[27] Z.X. Gao, C.W. Jiang, C.H. Lee, Improvement and application of wall function
boundary condition for high-speed compressible flows, Sci. China Tech. Sci. 56
(2013) 2501e2515.
[28] C. Mattiussi, The finite volume, finite element, and finite difference methods
as numerical methods for physical field problems, Adv. Imag. Electron Phys.
113 (2000) 1e146.