Professional Documents
Culture Documents
Majidi 2016
Majidi 2016
Majidi 2016
www.springerlink.com/content/1738-494x(Print)/1976-3824(Online)
DOI 10.1007/s12206-016-0313-4
(Manuscript Received April 29, 2015; Revised October 12, 2015; Accepted November 26, 2015)
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract
A ghost fluid based computational tool is developed to study a wide range of compressible multiphase flows involving strong shocks
and contact discontinuities while accounting for surface tension, viscous stresses and gravitational forces. The solver utilizes constrained
reinitialization method to predict the interface configuration at each time step. Surface tension effect is handled via an exact interface
Riemann problem solver. Interfacial viscous stresses are approximated by considering continuous velocity and viscous stress across the
interface. To assess the performance of the solver several benchmark problems are considered: One-dimensional gas-water shock tube
problem, shock-bubble interaction, air cavity collapse in water, underwater explosion, Rayleigh-Taylor Instability, and ellipsoidal drop
oscillations. Results obtained from the numerical simulations indicate that the numerical methodology performs reasonably well in pre-
dicting flow features and exhibit a very good agreement with prior experimental and numerical observations. To further examine the
accuracy of the developed ghost fluid solver, the obtained results are compared to those by a conventional diffuse interface solver. The
comparison shows the capability of our ghost fluid method in reproducing the experimentally observed flow characteristics while reveal-
ing more details regarding topological changes of the interface.
Keywords: Compressible multiphase flow; Ghost fluid method; Level-set; Surface tension; Viscous flows
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
In Volume of fluid approach the indicator variable is con- The main objective of the present study is to develop a
sidered to be zero for one phase and one for the other. In VOF ghost fluid solver applicable to simulate compressible multi-
algorithm, phase variable is first updated at each time step by phase flows possessing a diverse range of physical characteris-
solving a volume fraction advection equation. Then, the inter- tics. These compressible multiphase flows include highly con-
face is reconstructed based on the new distribution of the indi- vective flows with strong shocks and interfaces with high-
cator variable [15-19]. Despite its inherent mass preservation, acoustic impedance mismatch, gravity-driven flows, and sur-
easy extension to multidimensional problems, and automatic face tension and viscosity dominated flows. The GFM pre-
handling of topology change [20, 21], VOF method requires sented in this study is a generalization of the algorithm intro-
specific treatments in interface reconstruction [22, 23] and duced earlier to simulate convection dominated multiphase
curvature estimation [24, 25] due to its sharp distribution over flows especially supersonic liquid jets [45]. In the present
the interface. paper, our ghost fluid method is extended to include numerical
Level set method is another major category in interface cap- schemes accounting for surface tension, viscosity, and gravity.
turing approach which defines the phase indicator variable as A comprehensive assessment of the method is conducted
a distance function from the interface [26]. Consequently, in through several test problems in one and two dimensions. To
this method, the interface is represented by the zero value the extent of our knowledge, no other ghost fluid algorithm
isosurface of the level set variable which has a continuous has been tested against a set of benchmark problems, as di-
distribution over the interface [27]. Therefore, interface curva- verse and extensive as those utilized in the present study. The
ture and normal vectors are calculated more accurately com- predictions of the flow features and interface motion are in
pared to VOF. Here, interface orientation at each time step is good agreement with the results reported in the literature.
determined directly from levelset variable distribution without Therefore, the computational methodology appears to be a
requiring reconstruction. However, the distance function char- promising approach for numerical investigation of multiphase
acteristic of the level set variable is not maintained during compressible flow simulations. Also, to evaluate the accuracy
advection of the level set variable, necessitating regular reini- of the presented GFM solver, our ghost fluid results are com-
tialization of the level set variable [28]. Mass loss is deemed pared to those obtained using a diffuse interface solver. The
to be a major disadvantage of the levelset algorithm causing comparison indicates that our ghost fluid approach does a
the appearance of hybrid levelset algorithms such as combined wonderful job in maintaining the interface capturing quality
levelset-VOF [14], particle levelset [29], refined levelset grid and retrieving the details of interfacial evolutions.
method [30], coupled levelset-VOF [31], and Level contour
reconstruction method (LCRM) [32]. Moreover, several reini-
2. Governing equations and numerical scheme
tializaition methods have been introduced to maintain the
constancy of the interface location in reinitialization process The governing equations for non-reacting compressible two
and therefore minimize the mass loss [33-35]. phase flow without phase change are
Generally, interface capturing methods do not give any in-
formation on the flow status of each phase adjacent to the ¶U ¶ ( F - Fv ) ¶ (G - Gv )
+ + =S (1)
interface. Hence, in sharp interface framework context, trans- ¶t ¶x ¶y
interface communication and data transfer are usually estab-
lished through other algorithms called Ghost fluid methods where
(GFM) which are coupled to the main interface capturing
é r ù é ru ù é ru ù
scheme. Fedkiw et al. [36] proposed a GFM, where ghost ê ú ê 2 ú ê ú
r u r u + P r uv
values are replaced by corresponding values of the opposite U = ê ú ,F = ê ú ,G = ê ú
phase in that interfacial cell. This method, although simple ê rv ú ê r uv ú ê ru + P ú
2
ê ú ê ú ê ú
and easy to code, is not robust in simulating flows with large ëê r E ûú êëu ( r E + P ) úû êëv ( r E + P ) úû
density ratios, and phases with different equations of state. é 0 ù é 0 ù
Approximate Riemann solvers were applied at the interface to ê
t ú ê
t ú
Fv = ê ú ,G = ê ú
xx xy
construct the ghost values in order to prevent the defects of the
ê t xy ú v ê t yy ú
original GFM [37-39]. Sambasivan and Udaykumar [40] de- ê ú ê ú
veloped a GFM suitable for numerical simulation compressi- êëut xx + vt xy úû êëut xy + vt yy úû
ble multiphase flows as well as fluid-solid interaction [41]. é 0 ù
Later, they applied Local mesh refinement (LMR) [42] in their ê ú
ê ¶f ú
GFM to improve the quality of interface capturing and mass r g x - sk
ê ¶x ú
conservation of the scheme. Houim and Kuo [43] introduced a S = êê ¶f ú
ú (2)
ghost fluid algorithm capable of simulating reacting flows r g y - sk
ê ¶y ú
with phase change. Recently, Bo and Grove [44] have intro- ê ú
ê -usk ¶f ¶ f
duced a new VOF based GFM implemented in a sophisticated - vsk + r ug x + r vg y ú
êë ¶x ¶y úû
simulation code xRage.
S. Majidi and A. Afshari / Journal of Mechanical Science and Technology 30 (4) (2016) 1581~1593 1583
with E = e+0.5(u2+v2) and e as specific internal energy. where K = L, R and U*K is the intermediate state of the con-
Stiffened gas equation of state describes pressure as a func- servative variables and is calculated as
tion of internal energy and density
é1 ù
P = r e ( g -1) - g P¥ (3) ê ú
æ S - u K ö ê S*,K ú
U *,K = r K ç K ÷ê ú
è S K - S* ø êvK ú (9)
where γ is the specific heat ratio and P∞ is a constant which êë EK + ( S* - u K ) M úû
accounts for the material stiffness. The deviatoric and normal
PK
components of viscous stress tensor are given as: M = S* + .
r K ( S K - uK )
æ ¶v ¶u ö 2 æ ¶u ¶v ö
t xy = m ç + ÷- mç + ÷ With the intermediate fluxes (Eq. (8)) at hand, the numeri-
è ¶x ¶y ø 3 è ¶x ¶y ø
cal fluxes used in Eq. (5) are defined as
¶v 2 æ ¶u ¶v ö
t yy = 2m - mç + ÷ (4)
¶y 3 è ¶x ¶y ø
ì F (U L ) 0 £ S L
¶u 2 æ ¶u ¶v ö ï
t xx = 2m - mç + ÷. ï F*,L S L £ 0 £ S*
¶x 3 è ¶x ¶y ø F hllc =í (10)
ï F*,R S* £ 0 £ S R
ï F (U ) 0 ³ S
Also, vector S in Eq. (2) includes gravitational and surface î R R
F*,K = FK + S K (U K - U *,K ) (8) then, tagging variable of the cell (i,j) is defined as below.
1584 S. Majidi and A. Afshari / Journal of Mechanical Science and Technology 30 (4) (2016) 1581~1593
R ( PIm ; u L , PL , r L , u R , PR , r R ) =
(15)
GL ( PIm ; PL , r L ) + GR ( PIm ; PR , r R ) + u R - u L = 0.
æ aK ö
GK ( PIm ; PK , r K ) = ç ÷ ( PI - PK ) (16)
(a) Configuration of the interface ç PK + bK ÷
è ø
where
2
aK = ,
(g K + 1) r K
g K -1
bK = PK , (17)
gK +1
PK = PK + P¥ ,K
(b) Ghost cell layer for (c) Ghost cell for the liquid phase
the gas phase
and the respective relation for expansion waves is defined as
Fig. 1. Schematic representation of ghost cell layer encompassing the
interface. In images (b) and (c) the yellow cells represent the interfacial g K -1
cells and the green cells depict the ghost cell layer of each phase. Blue æ ö
m
æ 2cK ö ç æ PI ö 2g K ÷.
and Red cells show the interior cells for liquid and gas phase, respec- GK ( PI ; PK , r K ) = ç ÷ ç çç ÷÷ - 1 ÷ (18)
tively. g
è K - 1 P
øç è K ø ÷
è ø
ì0 if fi , j ´ fnb > 0 In the above equation, cK is the sound speed on the side k of
ï the interface.
ï1 if fi , j ´ fnb < 0
c ( i, j ) = í (13) It is possible that convergence is not achieved within a pre-
ï2 if fi , j ´ fn ( i , j ) < 0
determined number of iterations in Newton’s root finding
ï3 else .
î algorithm [49]. In this case, the numerical algorithm switches
to an approximate interface Riemann solver to recalculate the
In the above equation, nb represents indices of a neighbor- state vectors of the physical variables at cell edges intersecting
ing cell to the cell with indices (i,j). It is worth mentioning that the interface [45]. In the approximate procedure, the interme-
if c(i,j) = 0, then the cell (i,j) is an interior cell of phase A. The diate state of variables is given as below
value one of c(i,j) indicates that cell (i,j) is an interfacial cell
but still belonging to the phase A. When c(i,j) is set equal to PI = PK + r K ( S K - u K )( S* - u K ) ,
two, the cell (i,j) is located in the ghost cell layer of phase A. æ S K - uK ö
Value three for c(i,j) indicates that the cell belongs to the r I ,K = r K ç ÷, (19)
phase B and is located out of the p node thick ghost cell layer. è S K - S* ø
Fig. 1 shows distribution of the tagging variable for each u I = S* ,
phase in a computational zone including the phase interface.
After generation of the ghost cell layer, interface Riemann where index I indicates the intermediate state of the variables
problems are solved between every couple of adjacent interfa- and K = L,R. Also we have
cial cells. To this end, the exact Riemann solver presented by
Rallu [48] is used, in which interfacial pressure is obtained by S R = max(u R + cR , u% + %c),
the Newton’s iterative root finding process. (20)
S = min(u - c , u% - %c).
L L L
r L uL + r R uR
where PI is pressure at the interface and function R is defined u% = h ( u ) = , (21)
as rL + rR
S. Majidi and A. Afshari / Journal of Mechanical Science and Technology 30 (4) (2016) 1581~1593 1585
%c 2 = h ( Y ) + h ( G ) éêh æ P ö + L ùú .
ç ÷
êë è r ø úû
2
0.25
1 æ ( u - u L )( r L r R ) ö
L= ç R ÷ '
2ç rL + rR ÷
è ø
(a) Gaseous phase (b) Liquid phase
¶P
Y= ' (22) Fig. 2. Direction of multidimensional extrapolation for gas and liquid
¶r e
phase.
1 ¶P
G= .
r ¶e r neighboring the interface
t nn = t xx nx 2 + t yy n y 2 + 2t xy nx n y
In the above equation, κ is the interface curvature defined as
t nt = ( nx 2 - n y 2 )t xy + (t xx - t yy ) nx n y
(29)
k = Ñ.n (25) un = unx + vn y
ut = vnx - un y .
and n is computed as
3. Results Table 1. Error norms obtained using sequentially refined uniform grids.
In order to assess the capabilities of our developed ghost Grid number Grid spacing L1 error norm
fluid method, several compressible multiphase problems are 128 0.078125 5597.986768
considered. Convection dominated shock-interface interaction 256 0.0390625 2786.203767
problems include one-dimensional tin-air shock tube, conver- 512 0.01953125 1394.123768
gent shock-bubble interaction, air bubble collapse in water,
1024 0.009765625 697.5438371
and under-water explosion. The performance of our numerical
algorithm in predicting viscous and surface tension effects is
evaluated through investigating Rayleigh-Taylor instability
and oscillating drop. All two-dimensional convective and
viscous dominated test cases are also simulated using a sepa-
rately developed diffuse interface solver for comparison pur-
poses.
(a) 55 μs
(b) 135 μs
passed by air ideal gas with the same pressure (Fig. 5). The
atmospheric air is separated from the post-shock air region by
(c) 187 μs
a left-moving shock wave of Mach 1.72, initially situated at a
50 mm distance from the bubble center. The initial conditions
of the problem as well as thermodynamic constants are given
in below.
(r , u , v, P;g , P¥ ) = (0.991,0,0,3.059;5.5,1.505 ´ 104 ). (33) Fig. 8. Density and pressure distribution along the horizontal center-
line. The black line represents the GFM results, while the DIM results
are shown by red line. The symbols represent the results of Bo and
All sides of the domain are modeled as walls and the prob- Grove [44].
lem is simulated using a uniform 200×200 grid. The results
from the developed ghost fluid solver are exhibited in Fig. 7.
We can see that the algorithms predict a radially outward with a speed of 681.5 m/s (Fig. 9). Non-reflecting boundary
moving shock in the liquid phase, an inward running rarefac- condition is imposed on the left and right sides of the compu-
tion wave in the gaseous phase and the circular interface posi- tational domain, while top and bottom sides are assumed to be
tioned in between these two waves. Moreover, as witnessed in slip walls.
Fig. 8, density and pressure variations along the horizontal Initial conditions for this problem are given below.
centerline obtained using both ghost-fluid and diffuse inter-
face solvers agree well with the results of Bo and Grove [44]. ì(1323 kg/m3 , 1.9 GPa) post-shock water
However, the density gradient across the interface is noticea- ï
(r , P ) = í(1000 kg/m3 , 1 atm) pre-shock water (34)
bly smeared in the results obtained using the diffuse interface
ï(1.2 kg/m3 , 1 atm) air bubble .
method compared to those of the ghost fluid approach. î
The last problem upon which we test our ghost fluid solver,
involves the oscillations of a two-dimensional ellipsoidal drop
due to the presence of surface tension. This test case is a
benchmark problem to evaluate the accuracy of multi-phase
flow solvers accounting for surface tension [61, 62].
An ellipsoidal liquid drop with major and minor axes of 3.0
and 2.0 is initially at rest in gaseous medium. Surface tension
coefficient is set equal to 342 N/m and physical properties of
the liquid phase are
3 -3
ïì(100 kg/m ; 1.8 ´ 10 Pa.s) water
(r ; m ) = í 3 -5
. (35)
ïî(1 kg/m ;1.93 ´ 10 Pa.s). air
(b) Diffuse interface method
Nomenclature------------------------------------------------------------------------
ρ : Density
(c) t = 0. 5τc (d) t = τc u,v : Cartesian velocity components
Fig. 15. Volume Fraction contours and velocity vectors shown for E : Specific total energy
oscillating 2D drop during a time period. P : Pressure
f : Volume fraction
e : Specific internal energy
γ,P∞ : Thermodynamic constants
c : Sound speed
U : Vector of conservative variables
F,G : Vectors of convective fluxes
Fv,Gv : Vectors of viscous fluxes
S : Vectors of source terms
References
[1] J. C. Hermanson, Dynamics of supersonic droplets of vola-
tile liquids, AIAA Journal, 45 (3) (2007) 730-733.
Fig. 16. Kinetic energy versus time for oscillating drop problem simu-
[2] R. Saurel and O. Lemetayer, A multiphase model for com-
lated using a 256×256 grid.
pressible flows with interfaces, shocks, detonation waves and
cavitation, Journal of Fluid Mechanics, 431 (2001) 239-271.
the computed solutions. The relative error in the numerically [3] E. Johnsen and T. Colonius, Shock-induced collapse of a gas
computed time period is 6.9% for the 128×128 grid and 5% bubble in shockwave lithotripsy, The Journal of the Acoustical
for a 256×256 grid. Society of America, 124 (4) (2008) 2011-2020.
[4] A. Murrone and H. Guillard, A five equation reduced model for
compressible two phase flow problems, Journal of Computa-
4. Conclusion
tional Physics, 202 (2) (2005) 664-698.
A ghost fluid solver was developed and applied to a variety [5] R. Saurel, F. Petitpas and R. A. Berry, Simple and efficient
of compressible multiphase flows such as shock-wave domi- relaxation methods for interfaces separating compressible fluids,
nated problems, shock-interface interaction, gravity driven cavitating flows and shocks in multiphase mixtures, Journal of
viscous flows, and interfacial flows with surface tension ef- Computational Physics, 228 (5) (2009) 1678-1712.
fects. A comprehensive assessment of the method was con- [6] F. Gibou, L. Chen, D. Nguyen and S. Banerjee, A level set
ducted through several test problems with different flow char- based sharp interface method for the multiphase incompressible
acteristics including multiphase shock-tube problem, shock- Navier–Stokes equations with phase change, Journal of Com-
bubble interaction, air bubble collapse in water, and underwa- putational Physics, 222 (2) (2007) 536-555.
ter explosion, Rayleigh-Taylor instability, and oscillating drop. [7] A. Prosperetti and G. Tryggvason, Computational methods for
To further evaluate the interface resolution capabilities of our multiphase flow, Cambridge University Press (2007).
ghost fluid method, most of the test cases were also simulated [8] I.-L. Chern, J. Glimm, O. McBryan, B. Plohr and S. Yaniv,
1592 S. Majidi and A. Afshari / Journal of Mechanical Science and Technology 30 (4) (2016) 1581~1593
Front tracking for gas dynamics, Journal of Computational refined grids modeled with VOF, staggered finite volumes, and
Physics, 62 (1) (1986) 83-110. spline interpolants, Journal of Computational Physics, 166 (2)
[9] J. B. Bell, P. Colella and M. L. Welcome, Conservative front- (2001) 302-335.
tracking for inviscid compressible flow, Proc. in 10th Compu- [25] M. M. Francois, S. J. Cummins, E. D. Dendy, D. B. Kothe, J.
tational Fluid Dynamics Conference: AIAA, Honolulu, Hawaii, M. Sicilian and M. W. Williams, A balanced-force algorithm
USA (1991) 814-822. for continuous and sharp interfacial surface tension models
[10] S. O. Unverdi and G. Tryggvason, A front-tracking method within a volume tracking framework, Journal of Computational
for viscous, incompressible, multi-fluid flows, Journal of Com- Physics, 213 (1) (2006) 141-173.
putational Physics, 100 (1) (1992) 25-37. [26] S. Osher and R. Fedkiw, Level set methods and dynamic im-
[11] J. Glimm, J. W. Grove, X. L. Li, K.-M. Shyue, Y. Zeng and Q. plicit surfaces, Springer Science & Business Media, 153 (2003).
Zhang, Three-dimensional front tracking, SIAM Journal on [27] M. Gorokhovski and M. Herrmann, Modeling primary atomi-
Scientific Computing, 19 (3) (1998) 703-727. zation, Annual Review of Fluid Mechanics, 40 (2008) 343-366.
[12] G. Tryggvason, B. Bunner, A. Esmaeeli, D. Juric, N. Al- [28] J. Sethian and P. Smereka, Level set methods for fluid inter-
Rawahi, W. Tauber, J. Han, S. Nas and Y.-J. Jan, A front- faces, Annual Review of Fluid Mechanics, 35 (1) (2003) 341-
tracking method for the computations of multiphase flow, 372.
Journal of Computational Physics, 169 (2) (2001) 708-759. [29] D. Enright, R. Fedkiw, J. Ferziger and I. Mitchell, A hybrid
[13] R. Scardovelli and S. Zaleski, Direct numerical simulation of particle level set method for improved interface capturing,
free-surface and interfacial flow, Annual Review of Fluid Me- Journal of Computational Physics, 183 (1) (2002) 83-116.
chanics, 31 (1) (1999) 567-603. [30] M. Herrmann, A balanced force refined level set grid method
[14] M. Sussman, P. Smereka and S. Osher, A level set approach for two-phase flows on unstructured flow solver grids, Journal
for computing solutions to incompressible two-phase flow, of Computational Physics, 227 (4) (2008) 2674-2706.
Journal of Computational Physics, 114 (1) (1994) 146-159. [31] B. Ningegowda and B. Premachandran, A Coupled Level Set
[15] W. J. Rider and D. B. Kothe, Reconstructing volume tracking, and Volume of Fluid method with multi-directional advection
Journal of Computational Physics, 141 (2) (1998) 112-152. algorithms for two-phase flows with and without phase change,
[16] J. E. Pilliod and E. G. Puckett, Second-order accurate volume- International Journal of Heat and Mass Transfer, 79 (2014)
of-fluid algorithms for tracking material interfaces, Journal of 532-550.
Computational Physics, 199 (2) (2004) 465-502. [32] S. Shin, Direct numerical simulation of multiphase flow for
[17] Y.-Y. Tsui, S.-W. Lin, T.-T. Cheng and T.-C. Wu, Flux- arbitrary geometry using level contour reconstruction method,
blending schemes for interface capture in two-fluid flows, In- Journal of Mechanical Science and Technology, 23 (7) (2009)
ternational Journal of Heat and Mass Transfer, 52 (23) (2009) 1795-1800.
5547-5556. [33] E. Olsson, G. Kreiss and S. Zahedi, A conservative level set
[18] A. Cervone, S. Manservisi and R. Scardovelli, An optimal method for two phase flow II, Journal of Computational Phys-
constrained approach for divergence-free velocity interpolation ics, 225 (1) (2007) 785-807.
and multilevel VOF method, Computers & Fluids, 47 (1) [34] D. Hartmann, M. Meinke and W. Schröder, The constrained
(2011) 101-114. reinitialization equation for level set methods, Journal of Com-
[19] C. Wu, D. Young and H. Wu, Simulations of multidimen- putational Physics, 229 (5) (2010) 1514-1535.
sional interfacial flows by an improved volume-of-fluid method, [35] E. Brun, A. Guittet and F. Gibou, A local level-set method
International Journal of Heat and Mass Transfer, 60 (2013) using a hash table data structure, Journal of Computational
739-755. Physics, 231 (6) (2012) 2528-2536.
[20] X. Jiang, G. Siamas, K. Jagus and T. Karayiannis, Physical [36] R. P. Fedkiw, T. Aslam, B. Merriman and S. Osher, A non-
modelling and advanced simulations of gas–liquid two-phase oscillatory Eulerian approach to interfaces in multimaterial
jet flows in atomization and sprays, Progress in Energy and flows (the ghost fluid method), Journal of Computational Phys-
Combustion Science, 36 (2) (2010) 131-167. ics, 152 (2) (1999) 457-492.
[21] H. K. Myong, Numerical simulation of multiphase flows with [37] T. Liu, B. Khoo and K. Yeo, Ghost fluid method for strong
material interface on an unstructured grid system, Journal of shock impacting on material interface, Journal of Computa-
Mechanical Science and Technology, 26 (5) (2012) 1347-1354. tional Physics, 190 (2) (2003) 651-681.
[22] Y. Renardy and M. Renardy, PROST: a parabolic reconstruc- [38] X. Y. Hu and B. C. Khoo, An interface interaction method for
tion of surface tension for the volume-of-fluid method, Journal compressible multifluids, Journal of Computational Physics,
of Computational Physics, 183 (2) (2002) 400-421. 198 (1) (2004) 35-64.
[23] K. Yokoi, A practical numerical framework for free surface [39] T. Liu, B. Khoo and C. Wang, The ghost fluid method for
flows based on CLSVOF method, multi-moment methods and compressible gas–water simulation, Journal of Computational
density-scaled CSF model: Numerical simulations of droplet Physics, 204 (1) (2005) 193-221.
splashing, Journal of Computational Physics, 232 (1) (2013) [40] S. K. Sambasivan and H. Udaykumar, Ghost fluid method for
252-271. strong shock interactions Part 1: Fluid-fluid interfaces, AIAA
[24] I. Ginzburg and G. Wittum, Two-phase flows on interface Journal, 47 (12) (2009) 2907-2922.
S. Majidi and A. Afshari / Journal of Mechanical Science and Technology 30 (4) (2016) 1581~1593 1593