Professional Documents
Culture Documents
Lagrangian and Eulerian
Lagrangian and Eulerian
Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv
h i g h l i g h t s
This study compares the results of two approaches of Lagrangian and Eulerian.
In low concentration of particles, the results of two approaches are different.
Lagrangian model converge to the Eulerian one by increasing simulation time.
a r t i c l e i n f o a b s t r a c t
Article history: Modeling the behavior of suspended particles in gaseous phase is important for diverse reasons; e.g.
Received 24 October 2013 aerosol is usually the main subject of CFD simulations in clean rooms. Additionally, to determine the rate
Received in revised form and sites of deposition of particles suspended in inhaled air, the motion of the particles should be
22 January 2014
predicted in lung airways. Meanwhile there are two basically different approaches to simulate the
Accepted 31 January 2014
Available online 1 February 2014
behavior of particles suspension, Lagrangian and Eulerian approaches. This study compares the results of
these two approaches on simulating the same problem. An in-house particle tracking code was devel-
oped to simulate the motion of particles with Lagrangian approach. In order to simulate the same
Keywords:
Lagrangian approach
problem with Eulerian approach, the solution to the transport equation with appropriate initial and
Eulerian approach boundary conditions was used. In the first case study, diffusion of particles, initially positioned homo-
Aerosol geneously on an infinite plane was modeled with both approaches and the results were compared and
Particle deposition the mismatch between Lagrangian and Eulerian approaches was analyzed for different concentrations. In
the second case study, airflow with parabolic velocity profile moving between two parallel plates was
modeled with two approaches. The airflow initially contained a homogeneous suspension of particles
and the plates were maintained at zero concentration. The concentration along the plates was compared
between the two approaches and the differences in the performance of each approach were investigated,
again for different initial concentrations. The overall results confirm that as particle concentration falls
below a minimum amount, approximately 105 m2, the results of the two approaches deviate consid-
erably from each other and hence the Eulerian approach cannot be taken as an alternative for Lagrangian
approach for low concentrations. For the third problem, we investigated the 3D particle flow in an
expanding lung alveolus. It is shown that when the number of total released particles increases, the
results of Eulerian approach can be used as an alternative to Lagrangian simulation. Since the number of
particles existing in the lung alveolus in normal condition is much lower than this value, we concluded
that Eulerian method cannot be applied to problems involving low concentration of particles. Although,
the results of the Lagrangian problem may converge to the Eulerian one by increasing simulation time,
but it is a hypothetical situation which not really exist in short time scale problems such as third case
study in this paper.
Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction
1352-2310/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.atmosenv.2014.01.069
200 M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206
example, one of the areas where study of aerosol behavior is sig- geometries will not raise an overall rule for the mismatch between
nificant is simulation of clean rooms. Any clean room is designed to the results of Lagrangian and Eulerian approaches for other studies
accommodate a manufacturing or scientific task in which even a and geometries. Zhao et al. (2008) evaluated particle dispersion in
limited number of aerosols may play a destructive role and result in ventilated rooms by three different approaches of Lagrangian, drift
erroneous results. Therefore, there is a vast pool of modern studies flux and mixture models. They concluded that Lagrangian models
on simulation of aerosol motion in clean rooms among which one agreed well with experimental data. The drift flux model is more
can point to numerical modeling in Sznitman et al. (2009), Gao and accurate near the wall while mixture model yield unacceptable
Zhang (2010) and combined numerical and experimental in- results for particle concentration. Although the result of drift flux
vestigations in Shih et al. (2011), Whyte et al. (2010). model agreed well with Lagrangian and experimental model, but
The industry of clean rooms is not the only field where studying they did not report number of tracked particles and simulation time
the behavior of fine particles seems necessary. Also airways of in their Lagrangian model which is the purpose of this article.
human lung, both acinar and conducting ones, are where transport Zhao et al. (2010) found penetration coefficient through a single
and deposition of fine particles should be modeled and analyzed. crack in a building envelope by three different approaches;
Drug delivery would be much more efficient if drug particles were Analytical, Eulerian and Lagrangian. They used Fluent software for
delivered straight to the intended tissue (Langers, 1998; Brain and Lagrangian simulation which independent results is obtained by
Valberg, 1979). The drug injected in the cardiovascular system increasing number of particles. They according to Aliabadia and
may have to pass from different organs like kidney, liver, etc. before Rogaka (2011), Zhao et al. (2011) concluded that Fluent
arriving at the intended tissue. So the possibility of administering Lagrangian method cannot be used to model Brownian motion of
drugs via acinar airways of the lung can contribute a great deal to fine particles correctly.
an efficient method of drug delivery. In order to fulfill such a task, a Most of these studies are constricted to the steady state prob-
given drug should be first aerosolized and its resulting physical and lems and they did not discuss the differences between the results of
pharmaceutical characteristics should then be determined. The two approaches in unsteady problems. Moreover, there is no dis-
next step is to determine the dynamics of the inhaled aerosol all the cussion on the dependency of this mismatch with concentration of
way to the deepest parts of respiratory system. This approach is, particles; although it is expected that for high concentrations the
however, not an easy task to achieve mainly due to lack of adequate results of Lagrangian and Eulerian calculations may be the same
in-vivo measuring techniques. Therefore the numerical simulations and as gradually the concentration is reduced, this mismatch
can be the only way to predict the behavior of particles in acinar should also become more significant. Therefore there is a need to
airways. bring up a quantitative comparison between Lagrangian and
In addition to drug delivery, the increasing rate of mortality and Eulerian results for different particle concentrations and simulation
morbidity because of inhaling fine particulates in the environment time.
reported in Schwartz and Dockery (1992), Wilson and Spengler It should be remarked that numerous studies such as Chibbaro
(1996) and the increasing threats of bio-terrorism reported in and Minier (2011), Sanjose et al. (2011) have compared
Harrington et al. (2006) are other issues which add to the impor- Lagrangian and Eulerian approaches for multiphase flow of bubbles
tance of the subject. in liquids, but these studies cannot be related to aerosols, as
The numerical simulations used in the above mentioned refer- different mechanisms play role in the motion of bubbles in liquids
ences follow two main separate approaches called Lagrangian and compared with the motion of particles in atmosphere.
Eulerian. Lagrangian approach deals with individual particles and In current work, we investigate three case studies to compare
calculates the trajectory of each particle separately, whereas the the difference between the results of Lagrangian and Eulerian ap-
Eulerian approach deals with concentration of particles and cal- proaches. The basic mechanism in these two case studies which
culates the overall diffusion and convection of a number of parti- makes particles move is Brownian motion of particles; or diffusion
cles. It is evident that when handling the same question, the of group of particles.
calculations in Lagrangian approach are quite more time
consuming than Eulerian approach since in Eulerian approach, an 2. Methods
average behavior of particles is investigated instead of the behavior
of each existing particle. 2.1. Lagrangian approach
Hu et al. (2002) applied CFD simulation for particle dispersion in
clean rooms and Zhao and Wu (2005) applied Eulerian approach to A particle tracking code was developed to analyze the motion of
study if the particles could be treated as passive gas pollutants. particles in atmosphere, capable of taking into account the effects
Zhao and Wu (2005) have modeled clean rooms using a well- of diverse forces such as Brownian, Saffman, Drag and gravity force.
known derivation of Eulerian approach called drift-flux model If the fluid flow domain is steady and simple and if flow velocity
has been used in simulation of indoor particle dispersion (Gao and components can be defined by simple functions, this code uses
Zhang, 2010). these functions to obtain the velocity of fluid at the position of
Although when discussing why Eulerian method has been particle; whereas if flow domain is rather complicated or if fluid
preferred over Lagrangian one, it is mostly argued that because: 1) flow is unsteady, this code can be used in conjunction with fluid
Eulerian method has less computational cost in comparison with flow solvers. Hence, at each time step of fluid flow this code in-
Lagrangian one and 2) instead of positions of particles, Eulerian teracts with flow properties computed by the solver and particle
method works with concentration of particles which is more positions will be advanced in time. This code solves the following
appropriate for engineering applications. However, the question is equation to obtain the particle trajectories:
that for what conditions the results of the two approaches are !
equivalent and when the two methods are different. Although dup r
¼ FD þ g 1 þ Fp þ Fm þ FBa þ Fb þ Fs (1)
much has been written on using these methods, less has been dt rp
written on the answer to this question in the area of aerosols.
Zhang and Chen (2007) compared Lagrangian and Eulerian ap- In which the right hand side of this equation is the summation of
proaches for two special geometries. However the shortcoming is forces on the particle including drag FD, gravity, pressure gradient
that modeling and analyzing two special cases on two special force FP, virtual mass Fm and Basset force FBa, Brownian force Fb, and
M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206 201
Saffman’s lift force Fs. Ratio of FP, Fm, FBa to FD expressed in Table 1 while the Lagrangian approach deals with the second law of
because of small air density to particle density 1according to Zhao Newton to compute the trajectory of individual particles, Eulerian
et al. (2004) are very small and can be neglected. Also the ratio of Fs approach solves the transport equation for the concentration of
to FD, indicates that Saffman’s lift force is important only for large particles as below:
particles in high shear flows.
vc
So, in three cases which we have studied only drag, gravity and þ ðu$VÞc ¼ D V2 c (6)
Brownian force are could be important. Since in our study the vt
prediction of particle motion by Lagrangian and Eulerian approach
is to be compared, the gravity force is excluded due to its similar in which D is the diffusion coefficient of species. This equation
effect on these two approaches. Therefore, we have only focused on together with the necessary initial and boundary conditions will
the drag and Brownian forces. determine the concentration of particles in space.
Since the diameters of particles under study are in the order of The connecting bridge between Lagrangian and Eulerian
microns, the fluid flow around particles can be supposed to be in approach in aerosol science is the diffusion coefficient. In other
the Stokes region due to low Reynolds number; therefore, referring words, the final equation obtained for Lagrangian approach is fully
to Hinds (1999), drag force is computed using the following simple dependent on properties of particles under investigation. The same
equation: properties will combine together, in the following format, to form
the diffusion coefficient which is used in the transport equation (Eq.
FD ¼ 3pmUrel dp (2) (6)) in Eulerian approach:
where m is the fluid viscosity, Urel is the relative velocity between kTCc
D ¼ (7)
particle and fluid and dp is the particle diameter. 3pmdp
In order to simulate the Brownian force, first the inverse of
With this in mind, the following paragraphs deal with com-
particle relaxation time is calculated as follows:
parison between the results of Lagrangian and Eulerian approaches
1 3pmdp in two case studies for the same particles, the same initial and
b¼ ¼ (3) boundary conditions. These results are expected to deviate from
s ðCc mÞ
each other as the particle concentration decreases.
where Cc is the slip correction factor and m is the particle mass.
Secondly, the spectral intensity is computed using the calculated
quantity: 3. Results and discussion
Table 1
Ratio of additional forces exerted on particles to drag force.
FD 1
ra a r
Fp rp ap zrp
Fm 1 r
2 rp
FBa 0:5
18rsR
CD prp t
Fs 9:69d C 0:5
in 2D flow : 18*y0:5p *pc du
dy
sgn du
dy
0:288dp Cc dij
in 3D flow : Fig. 1. The schematic of geometry for the first case study; a) particles arrangement at
y0:5 ðdlk dkl Þ0:25
t ¼ 0, b) t > 0.
202 M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206
Cðx; bÞ ¼ 0 (16)
Concentration of particles was set to zero at the plates.
Therefore, each particle which impacts on the plates will be
absorbed.
In dimensionless terms the Eq. (13) and its boundary conditions
can be written
v q v2 q
1 h2 ¼ (17)
vg vh2
with
qð0; hÞ ¼ 1 (18)
vq
ðg; 0Þ ¼ 0 (19)
Fig. 2. Error of Lagrangian and Eulerian methods with respect to each other. vh
M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206 203
Fig. 3. Concentrations at different positions in the first case study by Lagrangian and Eulerian approaches for different initial concentration; (a) 105 m2, (b) 104 m2, (c) 103 m2, (d)
102 m2.
were 1 mm and 0.05 m/s. For considering walls, it was assumed that
particles impact on walls will be trapped and will be excluded from
the remaining simulation. In order to maintain a constant con-
centration at the inlet, when a particle leaves the channel inlet
another particle was injected instead. Similar to first case by
dividing the space into equal horizontal bandwidths the average
concentration (q(g)) along the plates, is computed and compared
with the predicted ones by Eulerian approach as shown in Fig. 5.
Same as the results of the first case study the results reveal that
for initial concentrations lower than 102 m2 the accordance of the
results of two methods decreases. In order to observe this deviation
better, we can also compare the concentrations in y direction,
calculated by the two methods in a specific location along the
Fig. 4. The schematic of the geometry in the second case study and fully developed
plates as shown in Fig. 6.
velocity profile.
It can be shown that as long as the concentration of particles is
high enough, Eulerian and Lagrangian approaches reveal the same
results, although when concentrations fall below about 105 m2, as
it is depicted in Fig. 6, the results of the two approaches deviate
qðg; 1Þ ¼ 0 (20) from each other.
Fig. 5. Mean concentration variation along the plates in the second case study by Lagrangian and Eulerian approaches for different initial concentration; (a) 105 m2, (b) 104 m2, (c)
103 m2, (d) 102 m2.
then be described by the following sinusoidal kinematic displace- where L0 is the length scale at t ¼ 0, f ¼ 1/T is the breathing fre-
ment function: quency, and l(t) is the sinusoidal function defined in the brackets of
Eq. (21).We have assumed that T ¼ 4 s and the length scale
expansion factor b ¼ 0.077.
b b p The pressure inlet boundary condition is chosen so that moving
LðtÞ ¼ L0 1 þ þ sin 2pft ¼ L0 lðtÞ (21)
2 2 2 walls will cause air to flow in and out of the alveolus. In order to
compare the results of the two approaches, particle tracking and
Fig. 6. Concentration profile between the plates at g ¼ 1 for the second case study by Lagrangian and Eulerian approaches for different initial concentration; (a) 105 m2, (b)
104 m2, (c) 103 m2, (d) 102 m2.
M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206 205
Fig. 9. (a) Concentration contour resulted from Eulerian model, (b) Trajectory of 1-micron diameter particles after 2 s for 2000 released particle, (c) Trajectory of 1-micron diameter
particles after 2 s for 10000 released particle, (d) Comparison of Eulerian vs. Lagrangian for 2000 released particle and (e) Comparison of Eulerian vs. Lagrangian for 10000 released
particle.
206 M.S. Saidi et al. / Atmospheric Environment 89 (2014) 199e206
5. Conclusion Gao, Z., Zhang, J.S., 2010. Numerical analysis for evaluating the exposure reduction
effectiveness of room air cleaners. Building and Environment 45, 1984e1992.
Harrington, L., Prisk, G.K., Darquenne, C., 2006. Importance of bifurcation zone and
In this study, three different cases were analyzed. In each case, branch orientation in simulated aerosol deposition in the alveolar zone of the
exactly the same conditions were simulated by two different ap- human lung. Journal of Aerosol Science 37, 37e62.
proaches, Lagrangian and Eulerian. The findings of this study are: Hinds, W.C., 1999. Aerosol Technology, Properties, Behavior, and Measurement of
Airborne Particles. Wiley-Interscience, New York.
Hu, S.C., Chuah, Y.K., Yen, M.C., 2002. Design and evaluation of a minienvironment
1 In low concentrations, the results of Lagrangian and Eulerian for semiconductor manufacture processes. Building and Environment 37, 201e
modeling are not in satisfactory accordance with each other, and 208.
Langers, R., 1998. Drug delivery and targeting. Nature 391, 5e7.
Eulerian modeling may not look reasonable. Sanjose, M., Senoner, J.M., Jaegle, F., Cuenot, B., Moreau, S., Poinsot, T., 2011. Fuel
2 In steady state problems as found in case studies 1 and 2, despite injection model for Euler-Euler and Euler-Lagrange large-eddy simulations of
of long time of particle tracking in Lagrangian simulations an evaporating spray inside an aeronautical combustor. International Journal of
Multiphase Flow 37 (5), 514e529.
(200 s), the results of Eulerian method and Lagrangian one in Schwartz, J., Dockery, D.W., 1992. Increased mortality in philadelphia associated
low particle numbers are different. The minimum limit found in with daily air pollution concentrations. American Review of Respiratory Dis-
this study to coincide Eulerian results with Lagrangian ones is eases 145, 600e604.
Shih, Y.C., Yang, A.S., Lu, C.W., 2011. Using air curtain to control pollutant spreading
105. for emergency management in a clean room. Building and Environment 46,
3 Although, the results of the Lagrangian problem may converge 1104e1114.
to the Eulerian one by increasing simulation time, but it is a Sznitman, J., et al., 2009. Respiratoryþ þFlow phenomena and gravitational depo-
sition in a three-dimensional space-filling modelþ þof the pulmonary acinar
hypothetical situation which not really exist in short time scale
tree. Journal of Biomechanical Engineering Aerosol and Air Quality Research 10
problems. In other words, using Eulerian approach is not (131), 316e322.
reasonable for short time scale problems such as the third case Whyte, W., Hejab, M., Whyte, W.M., Green, G., 2010. Experimental and CFD airflow
study. In fact, when the particle concentration is low, the time studies of a clean room with special respect to air supply inlets. International
Journal of Ventilation 9, 197e209.
scale of problem should be long enough to be able to apply Wilson, R., Spengler, J.D., 1996. Particles in Our Air: Concentration and Health Ef-
Eulerian approach instead of time consuming Lagrangian fects. Harvard Univeristy Press, Boston, MA.
simulation. Zhang, Z., Chen, Q., 2007. Comparison of the Eulerian and Lagrangian methods for
predicting particle transport in enclosed spaces. Atmospheric Environment 41,
5236e5248.
Zhao, B., C., C., Lai, A.C.K., 2011. Lagrangian stochastic particle tracking: further
References discussion. Aerosol Science and Technology 45, 901e902.
Zhao, B., Chena, C., Yangb, X., Laic, A.C.K., 2010. Comparison of three approaches to
Ahmadi, G., Particle transport, depostion and removal course material. model particle penetration coefficient through a single straight crack in a
Aliabadia, A.A., Rogaka, S.N., 2011. Langrangian stochastic particle tracking. Aerosol building envelope. Aerosol Science and Technology 44, 405e416.
Science and Technology 45, 313e314. Zhao, B., Wu, J., 2005. Numerical investigation of particle diffusion in clean room.
Brain, J.D., Valberg, P.A., 1979. Deposition of aerosol in the respiratory tract. Indoor and Built Environment 14, 469e479.
American Review of Respiratory Diseases 120, 1325e1373. Zhao, B., Yang, C., Yang, X., Liu, S., 2008. Particle dispersion and deposition in
Chibbaro, S., Minier, J.P., 2011. A note on the consistency of hybrid Eulerian/ ventilated rooms: testing and evaluation of different Eulerian and Lagrangian
Lagrangian approach to multiphase flows. International Journal of Multiphase models. Building and Environment 43, 388e397.
Flow 37, 293e297. Zhao, B., Zhang, Y., Li, X., Yang, X., Huang, D., 2004. Comparison of indoor aerosol
Friedlander, S.K., 2000. Smoke, Dust, and Haze, Fundamentals of Aerosol Dynamics. particle concentration and deposition in different ventilated rooms by nu-
Oxford University Press, Los Angeles. merical method. Building and Environment 39, 1e8.