Professional Documents
Culture Documents
Liu, Z., Wang, X., Cheng, Z., Sun, R., & Zhang, A. Simulation of Construction Ventilation in Deep Diversion Tunnels Using Euler-Lagrange Method
Liu, Z., Wang, X., Cheng, Z., Sun, R., & Zhang, A. Simulation of Construction Ventilation in Deep Diversion Tunnels Using Euler-Lagrange Method
a r t i c l e i n f o a b s t r a c t
Article history: Removal of harmful components is a major concern in the construction ventilation of deep diversion tun-
Received 12 March 2014 nels with heat exchange. The intersecting tunnels and air leakage along the long duct system increase the
Received in revised form 25 August 2014 difficulty of smoke extraction. A 3D unsteady Euler–Lagrange two-phase turbulence model is utilized
Accepted 4 September 2014
with the aim of understanding the pollutant discharge problem caused by the storage heat effects, the
Available online 16 September 2014
intersecting tunnels and air leakage along the long duct system. The effects of air leakage, the gas–solid
heat exchange, the gas–solid interaction forces and the particle collision are considered. A grid-sensitivity
Keywords:
study was performed to obtain a reasonable mesh resolution. The model is validated based on the in-situ
Deep diversion tunnels
Construction ventilation
ventilation test data. The construction ventilation for the diversion tunnels of the Jinping Second-cascade
Euler–Lagrange turbulence model Hydropower Station in China is taken as a case. The air field distribution, the distribution of heat transfer
Heat exchange coefficient, the air leakage rate per 100 m, the particle collision force and dust particles migration tracks
Dust migration are analyzed. The dust particles collisions have a negative influence on ventilation and dust exhaust. Most
dust particles move toward the tunnel outlet along the tunnel bottom with the effects of airflow and den-
sity flow, whereas part of the dust particles move upwards under the effect of buoyancy.
Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compfluid.2014.09.016
0045-7930/Ó 2014 Elsevier Ltd. All rights reserved.
Z. Liu et al. / Computers & Fluids 105 (2014) 28–38 29
CFD models to analyze and compare the behavior of gas dilution where q0 is the density of the flow (kg/m3), b is the thermal expan-
after blasting in underground mines. Fewer studies have focused sion coefficient (K1), and DT is the temperature difference (K).
on construction ventilation in hydropower engineering. Wang The trajectory of a discrete phase particle is predicted by inte-
et al. [2] proposed a 3D unsteady quasi-single phase model to opti- grating the force balance on the particle, which is written in a
mize the ventilation time with different tunneling lengths. The Lagrangian reference frame. This force balance equates the particle
model was confirmed by the experimental values reported by inertia with the forces acting on the particle, and can be written as
Nakayama [17] and was applied to simulate diversion tunnel con-
dup gðqp qÞ
struction ventilation for the Xin Tangfang power station in China. ¼ F D ðu up Þ þ þ F p þ F ss ð2Þ
Wang et al. [3] used a three-dimensional renormalization-group dt qp
(RNG) model for forced ventilation to the working face of a long
where FD(u up) is the drag force per unit particle mass and Fp is
diversion tunnel, taking into account the effects of air leakage
the force acting on the particles, including the drag force, the ther-
and the frictional resistance along the tunnel. The case study
mophoretic force, the pressure gradient force, and the virtual mass
involved the working face during TBM construction of the Xinjiang
force. The relevant equations can be found in the Fluent manual
81 Daban long diversion tunnel in China.
[22].
Because of the smoke extraction difficulties caused by the
The influence from collision force between particles is taken
storage heat effects and the pollutant discharge difficulties with
into account in the force balance equation of the particle, which
cross distribution of tunnels, the lack of knowledge about gas–solid
is based on molecular kinetic theory. The particle collision force
two-phase construction ventilation for deep tunnels remains an
Fss can be determined as follows [25]:
obstacle. There are few two-phase flow investigations focused on
underground construction ventilation in mine tunnels. Klemens 2
F ss ¼ rJ ¼ ð1 þ eÞrðC s qs hv 02
s iÞ ð3Þ
et al. [18] established a two-dimensional laminar model to simu- 3
late the dust deposition in a rectangular coal mine roadway with
where e is the collision restitution coefficient, set to 1.0; Cs is the
upper obstacles. The influence of gas–solid two-phase convective
volume coefficient of the particles; v 0s is the particle fluctuating
diffusion was not considered in the model. Cannoo [19] used a
velocity given by hv 02
s i ¼ jks =ms ; ms is the particle mass (kg); ks is
Eulerian approach to simulate the dust and methane distribution
instantaneous kinetic energy (N m); j is Boltzmann constant,
in a mine tunnel. Toraño et al. [20] performed a two-phase numer-
1.3806504 1023 J K1. The particle turbulent dispersion is only
ical simulation of airflow and dust in mining roadways with the
scarcely caused by fluid pulsation [25]. ks can be expressed as
Euler–Lagrange method. There are few existing studies examining
follows:
construction ventilation simulation of a single tunnel in hydro-
power projects. Zhang et al. [21] employed a three-dimensional srs 1
transient two-phase Eulerian model to simulate velocity fields ks ¼ 1 þ kl ð4Þ
st
and dust diffusion of the heading face in a diversion tunnel. rffiffiffi 3=4
Overall, most of the aforementioned studies of diversion tun- 2 C l kc
st ¼ ð5Þ
nels have primarily concentrated on the single-phase flow simula- 3 ec
tion of a single diversion tunnel. Only a small number of them have qs d2s
srs ¼ ð6Þ
analyzed construction ventilation by focusing on the isothermal 18l
two-phase flow construction ventilation process itself. In this
paper, to examine smoke extraction difficulties caused by storage where srs is the particle dynamic response time (s); st is the air-
heat effects, pollutant discharge difficulties with cross distribution phase pulsation time (s); ds is particle diameter (m); qs is the parti-
of tunnels and air leakage along a long duct system, an cle density (kg/m3); l is the air dynamic viscous coefficient (N m/s);
Euler–Lagrange model is utilized for deep diversion tunnels. The kc is the air turbulent kinetic energy (m2/s2); ec is the dissipation
heat transfer coefficient and dust migration are analyzed. A rate of turbulent kinetic energy (m2/s3).
grid-sensitivity study was performed to obtain a reasonable mesh
resolution. In addition, the model was verified by the on-site 3. Computational model
ventilation test data.
3.1. Computational geometry and grid
2. Governing equations
The construction ventilation of the Jinping Second-cascade
In the construction ventilation of deep diversion tunnels, dust Hydropower Station deep diversion system, with four diversion
particles are convected and diffused, forming a dispersion system tunnels and a length of approximately 16.7 km from intake slope
suspended in the air medium. Therefore, an Euler–Lagrange model to upper tank, is taken as a case study. The design section is a
is applied. The 3D unsteady RANS equations in conjunction with horseshoe with a diameter of 13 m. The maximum burial depth
energy equation and the standard k e turbulence model are of the diversion system is approximately 2525 m. Therefore, the
solved [22]. The standard k e turbulence model proposed by diversion tunnels have the characteristics of deep cover, long dis-
Jones and Launder [23] has been largely validated by the scientific tances and large diameter, which lead to the pollutant discharge
community. For brevity, only the model constants are given here: difficulty. Forced ventilation is adopted. The construction ventila-
Cl = 0.09; C1e = 1.44; C2e = 1.92; rk = 1.0, re = 1.3 [24]. They are tion of the western No. 1 and No. 2 diversion tunnels is simulated.
the default values in Fluent 6.3. A transverse passageway is excavated every 500 m between the
The viscous dissipation terms are included in the energy equa- No. 1 and No. 2 diversion tunnels. To avoid a contaminated air
tion, which describe the thermal energy created by viscous shear in cycle among the transverse passageways and diversion tunnels,
the flow [22]. The Boussinesq model is used to solve the thermal the former transverse passageway is sealed off after the next trans-
buoyancy in the tunnels. This model treats density as a constant fixion. The actual distance of the drilling and blasting construction
value in all solved equations, except for the buoyancy term in is 7000 m, which we obtained from the construction data. CFD
the momentum equation [22]: technology is used for the simulation of the construction ventila-
tion at locations 550 m, 1500 m, 3000 m, 4000 m, 5000 m,
ðq q0 Þg ffi q0 bDTg ð1Þ
6000 m and 7000 m along the No. 1 and No. 2 diversion tunnels.
30 Z. Liu et al. / Computers & Fluids 105 (2014) 28–38
Fig. 1. The geometrical model and the calculation grid model of the diversion tunnels.
Fig. 1(a) shows the geometrical model of the diversion tunnels. The Because the overall grid model is so big that it is difficult to show
lengths of the No. 1 and No. 2 construction branch tunnels are the grid resolution, the detailed view near the working face of the
664 m and 820 m, respectively. Two axial flow fans are located at three grids are illustrated in Fig. 2. The three grids are compared
the exit of the No. 1 construction branch tunnel. An air duct, with based on the jet axial velocity from air duct outlet to the working
a diameter of 2 m, is located at the top of the tunnel. After the air- face, and the results are shown in Fig. 3(a). The average difference
flow arrives at the working face, it reverses to the tunnel outlet. between the coarser grid (grid A) and the fine grid (grid B) is 9.98%,
Small holes are open as air leakage ports at 100 m intervals along while it is only 2.81% between the fine grid (grid B) and the finer
the air duct. To reduce grid size, the blocked transverse passage- grid (grid C). For uniform reporting of the grid-convergence study,
ways are not built into the geometrical model. The total number we adopted the concept of the grid-convergence index (GCI) by
of cells is 1,662,283 for tunneling lengths of 550 m. The grid model Roache [26,27]. The result on selected grid with indication of band
of the diversion tunnels is shown in Fig. 1(b). of 1.25 grid-convergence index [28] is shown in Fig.3(b). The
A grid-sensitivity analysis of 550 m length diversion tunnel is results show that grid sensitivity is most obvious within
performed by conducting simulations on the fine grid with 10–30 m away from air duct outlet. Above analysis also shows that
1,662,283 cells (grid B), a coarser grid with 765,810 cells (grid A) the fine grid is a suitable grid. Therefore, the fine grid was selected
and a finer grid with 3,044,521 cells (grid C). The grids were for further analysis with the calculation accuracy and the compu-
obtained by coarsening the fine grid with twice about a factor 2. tational efficiency consideration.
Z. Liu et al. / Computers & Fluids 105 (2014) 28–38 31
the diameter (m) [29,30]. The pressure outlet boundaries are con-
sidered to be in the tunnel outlet (see Fig. 1). The gauge pressure is
set to 0. The outlet temperature is equal to the external
temperature.
The standard wall functions by Launder and Spalding [29] are
used with the sand-grain based roughness modification by Cebeci
and Bradshaw [31]. Because of the heat storage effect of the deep
tunnels and the strong heat exchange between the tunnel wall
and heat flow in the tunnel, the convective heat transfer boundary
conditions are considered. The thermal wall function is introduced
to determine the heat exchange coefficient [22].
In the construction ventilation gas–solid two-phase flow Euler–
Lagrange simulation of deep diversion tunnels, the dust particle
source is set in the working face. Dust particles escape to the tun-
nel outlet, so capture boundary conditions are set for the solid-
phase outlet boundary. In addition to the tunnel outlet, the rest
Fig. 2. View near the working face of the three grids for grid-sensitivity analysis.
of the walls are provided with reflecting boundary conditions.
Assuming that particle collision is elastic, the wall collision elastic
recovery factor is set to 1.
The initial discrete dust is mainly distributed near the working
face. The initial dust concentration is 400 mg/m3 [1]. The working
face is set as the dust injection source (see Fig. 1), the parameters
are shown in Table 1. The mass flow rate is calculated by cvA,
where c is the dust concentration (kg/m3); v is the air velocity in
the tunnel (m/s); A is the tunnel sectional area (m2).
The CFD model has been implemented in the commercial finite-
volume CFD code FLUENT. The time-dependent terms are handled
through an implicit second-order backward differentiation in time.
The second-order upwind scheme is used to approximate the con-
vective terms at the faces of the control volumes. The pressure-
implicit with splitting of operators (PISO) algorithm is employed
as the pressure velocity coupling method [32]. Convergence is
assumed to be obtained when the scaled residuals [22] reach
103 for all equations except the energy, for which the criterion
is 106.
dissipation rate are given by the equations kin ¼ 3=2ðU in Iin Þ2 and Particle distribution Uniform
ein ¼ C 0:75 1:5 Particle diameter (m) 5.0 106
l kin =lm . The turbulence intensity Iin is equal to
Initial particle velocity (m/s) 0
0.16(Re)1/8, the mixing length lm is lm = 0.07L, where Cl is an Particle mass flow rate (kg/s) 0.0098
empirical constant, 0.09; Re the Reynolds number, 1.06 105; L
32 Z. Liu et al. / Computers & Fluids 105 (2014) 28–38
Table 2
Comparison of the temperature simulated values and the measured values.
Measurement position P1 P2 P3 P4 P5
Measured values (m/s) 18.5 18.4 17.7 17.4 17.5
Simulation values (m/s) 21.1 20.3 19.4 19.0 18.5
Average relative error value (%) 14.1 10.3 9.6 9.2 5.7
Table 3
Comparison of the simulation values of average airflow velocity and the measured
values.
Measurement position P1 P2 P3 P4 P5
Fig. 5. The grid model for validation. Measured values (m/s) 0.498 0.791 0.577 0.474 0.453
Simulation values (m/s) 0.466 0.766 0.617 0.459 0.413
Average relative error value (%) 6.4 3.2 6.9 3.2 8.8
Fig. 8. Distribution of the basically stable airflow near the working face (unit: m/s).
Fig. 11. Distributions of the leakage rates per 100 m for different tunnel lengths.
Fig. 9. The air velocity distribution near the air leakage points at different locations
(unit: m/s).
Fig. 10. The air volumetric flow rate along the air duct at different tunnel lengths.
tive heat exchange between the tunnel wall and thermal airflow
under the effect of temperature difference. At a ventilation time
of 1800 s, the heat transfer coefficient of the tunnel wall increases, Fig. 12. Distribution of surface heat transfer coefficient near the working face
as shown in Fig. 12(b). The heat and temperature difference at the (unit: W/(m2 K)).
working face is maximal, which results in convective heat
exchange, a maximal heat transfer coefficient (10.1 W/(m2 K)). As
the partial enlarged area shows in Fig. 12(b), the heat transfer coef- field measurement of convective heat transfer coefficient in deep
ficient inside the tunnel (y = 6 m to 6 m) is basically lower than diversion tunnels. The heat transfer coefficient is solved using
1 W/(m2 K) as well as that near the tunnel wall (y = 6 m to the method of thermal wall function. The reliability of simulation
6.5 m and y = 6 m to 6.5 m) is 4 W/(m2 K), which indicates stron- results are proved by comparing with the empirical values that
ger convective heat exchange between the tunnel wall and airflow. are obtained using the empirical formula of tunnel wall heat trans-
With existing empirical formulas, it is difficult to obtain the law fer coefficient from the measured data. The empirical formula [34]
of convective heat transfer. There are presently few reports on the is as follows:
Z. Liu et al. / Computers & Fluids 105 (2014) 28–38 35
0:045k Re0:8
h¼ ð8Þ
Lf
toward the opposite direction of the temperature gradient of the the No. 2 transverse passageway is low. Thus, dust particles are
tunnel section, which causes thermophoresis and goes against detained here and hard to discharge, which causes dust in the
the effects of ventilation and dust exhaust. No. 2 transverse passageway to be retained for a longer time, and
The tracks of the dust particles near the working faces are the mean retention time is 1300 s (t = 21.67 min), as shown in
shown in Fig. 16. As affected by both the attached jet and density
flow, the dust around the working face travels along the wall
toward the tunnel bottom and moves forward in the control of air-
flow in a backflow zone. When arriving at the eddy zone, rotation-
ally entrained by eddy, part of the dust moves back and the
phenomenon of secondary suspension occurs. At the same time,
because of the high speed of airflow near the eddy zone, which
accelerates the dust movement, the movement of the dust particles
is accelerated, as shown in Fig. 16 (Area C and D red dashed frame).
Because of the carrying role of airflow, most dust moves toward the
tunnel outlet, except some dust moves back. When the dust arrives
near the No. 2 transverse passageway, the buoyancy greatly affect
the dust, instead of the airflow in the jet zone. As the partial
enlarged area E shows in Fig. 16(b), most dust particles move along
the bottom of the channel under the effect of density gravity flow,
but a portion of the dust particles move upwards under the effect
of buoyancy.
The distribution of dust particle trajectories and residence time
are shown in Fig. 17. The dust particle residence time increases
successively from the working face to tunnel outlet. Dust particles
move from the working face to tunnel outlet with the effect of air-
flow, the retention time near the working face is short, and the
mean value is 300 s (t = 5 min). Some of the dust particles move
to the No. 2 transverse passageway, and the airflow velocity in
Fig. 16. The tracks of the dust particles near the working faces (unit: m/s). Fig. 17. Distribution of dust particle trajectories and residence time (unit: s).
Z. Liu et al. / Computers & Fluids 105 (2014) 28–38 37
[18] Klemens R, Kosinski P, Wolanski P, Korobeinikovb VP, Markovb VV, Menshov [27] Roache PJ. Quantification of uncertainty in computational fluid dynamics.
IS, et al. Numerical study of dust lifting in a channel with vertical obstacles. J Annu Rev Fluid Mech 1997;29:123–60.
Loss Prevent Proc 2001;14(6):469–73. [28] Ramponi P, Blocken B. CFD simulation of cross-ventilation for a generic
[19] Cannoo B. STAR-CD digs miners out of trouble. In: CD adapco dynamics, isolated building: impact of computational parameters. Build Environ
fall. New York: CD Adapco Group; 2004. p. 27–8. 2012;53:34–48.
[20] Toraño J, Torno S, Menendez M, Gent M. Auxiliary ventilation in mining [29] Launder BE, Spalding DB. The numerical computation of turbulent flows.
roadways driven with roadheaders: validated CFD modelling of dust behavior. Comput Methods Appl Mech Eng 1974;3:269–89.
Tunn Underground Space Technol 2011;26(1):201–10. [30] Rodi W. Influence of buoyancy and rotation on equations for the turbulent
[21] Zhang J, Wang XL, Chen HC, Liu XP, Sun YF. Simulation on ventilation and dust length scale. In: 2nd symposium on turbulent shear flows, vol. 1; 1979. p. 10.
diffusion on heading face of the diversion tunnel. J Hydroelectr Eng [31] Cebeci T, Bradshaw P. Momentum transfer in boundary layers. Hemisphere
2008;27(1):111–7. Publishing Corporation; 1977.
[22] Fluent Inc.. Fluent 6.3 user’s guide. Lebanon: Fluent Inc.; 2006. [32] Issa R. Solution of implicitly discretized fluid flow equations by operator
[23] Jones WP, Launder BE. The prediction of laminarization with a two-equation splitting. J Comput Phys 1986;62:40–65.
model of turbulence. Int J Heat Mass Tran 1972;15:301–14. [33] The Co., Ltd. of the 13th Bureau Group of China Railway, Southwest Jiaotong
[24] Launder BE, Spalding DB. Lectures in mathematical models of University. Application research and in-situ report on construction
turbulence. London: Academic Press; 1972. coordination ventilation technology of the Jinping second-cascade
[25] Wu YL, Tang XL, Liu SH. Hydraulic machinery cavitations and solid-liquid two- hydropower station extra long tunnels. China; 2011.
phase hydrodynamics. Beijing: China Water Resources and Hydropower Press; [34] Mou LQ. Computation and application of air through tunnel
2007. cooling. Beijing: China Building Industry Press; 1982.
[26] Roache PJ. Perspective: a method for uniform reporting of grid refinement
studies. J Fluids Eng 1994;116:405–13.