Professional Documents
Culture Documents
Mixing in Pipelines With Side and Opposed Tees
Mixing in Pipelines With Side and Opposed Tees
Numerical and experimental investigations of mixing in pipelines with side and opposed tees
are carried out. Cold water flowing in a main pipe is mixed with hot water flowing through a
tee. The temperature is measured experimentally to quantify the degree of mixing. The velocity
and temperature fields are also solved numerically. The effects of the mesh size, mesh-localized
refinement, dependence of the fluid physical properties on temperature, and turbulence model
on numerical results were examined. Experimental results show good agreement with corre-
sponding predictions of the numerical model over a relatively wide range of Reynolds number;
however, close agreement is harder to obtain in the vicinity of the jet through the tee. The pipe
length required to achieve 95% mixing is found to be a function of Uj/Um. The angle at which
the side jet is injected determines whether the jet impinges on the opposite wall and also affects
the pipe length required to achieve 95% mixing. This work recommends that industry should
not use 90° tees because of possible poor mixing at certain velocity ratios and hard impingement.
For pipe diameters ranging from 1 to 16 in., if dj/dm is kept constant, then for any velocity ratio
the 95% mixing is achieved at a distance corresponding to about the same number of the main-
pipe diameters. For opposed jets, numerical convergence was harder to obtain at high Reynolds
numbers. Some modifications, including the staggering of the two jets, made it easier for the
solution to converge.
Introduction
Mixing problems, such as the design and scale-up of
a mixer and quantification of mixing, have been tradition-
ally tackled by using empirical correlations. Although
this approach has proven to be satisfactory for many
applications, it is rather limited because it neglects the
Figure 1. Schematic diagram of a pipeline side tee.
complexity of flow in most mixing applications.
Applications where pipeline mixing with tees is used
include low-viscosity mixing such as wastewater treat-
ment and blending of some oils (injection of additives)
and petrochemical products. Other applications include
blending of fuel gas and mixing of feed streams for
catalytic reactors. A tee is formed by two pipe sections
joined at a right angle to each other. One stream passes
straight through the tee, while the other enters per-
pendicularly at one side as shown in Figure 1. This flow
arrangement is known as the side tee. However, other Figure 2. Opposed-tee mixer.
flow arrangements may be used, such as having the two
opposing streams enter coaxially and leave through a
pipe that is perpendicular to the entering direction. This
is known as an opposed tee and is shown in Figure 2.
Multiple tees could also be used as shown in Figure 3.
For all designs of pipe tees, mixing takes place in shorter
distances compared with distances required for mixing
in a pipe with undisturbed turbulent flow. Reviews of
pipeline mixing with tees have been presented by
Simpson,1 Gray,2 and Forney.3
More recent studies have been carried out by Liou et
Figure 3. Multiple-tee mixer.
al.,4 who experimentally investigated side-jet injection
near a rectangular duct entry with various angles. They angular channel have been done by Weber et al.5 and
obtained a reasonable agreement between laser-doppler Lam and Xia.6 Pan and Meng7 presented an experimen-
velocimetry measurements and numerical computa- tal investigation of turbulent flow in a round tee mixer.
tions, with the numerical model underpredicting. Other Moussa et al.8 and Crabb et al.9 investigated the flows
experimental studies of a jet issuing in an open rect- generated by a tee mixer. A survey of the literature
shows that simulation using computational fluid dy-
* To whom correspondence should be addressed. E-mail: namics (CFD) of mixing in a pipeline with tees has been
[email protected]. carried out by Cozewith et al.10 and Forney and Mon-
10.1021/ie0209935 CCC: $25.00 © 2003 American Chemical Society
Published on Web 09/10/2003
5334 Ind. Eng. Chem. Res., Vol. 42, No. 21, 2003
clova.11 Cozewith et al.10 simulated tee mixing charac- with tees. The dependence of the pipe length required
teristics with and without a reaction for a tee with d/D to achieve 95% mixing on the side- to main-velocity ratio
) 0.188 over a range of side-stream/main-stream veloc- and the side- to main-diameter ratio is investigated. The
ity ratios from 1.2 to 6.5. A three-dimensional model effects of the angle of the tee and the use of opposed
was constructed, and the k- model was used to model tees are also investigated.
turbulence. The numerical results of Cozewith et al.10
were in reasonable agreement with the experimental Model Equations
concentration trajectory found by Cozewith and Busko12
for x/D > 0.7. The concentration trajectory is defined The differential equations representing mass, mo-
as the locus of maximum concentration. Other compari- mentum, and energy conservation can be written in the
sons also showed a qualitative agreement between general form:
experimental and numerical results.
Forney and Monclova11 simulated pipeline side-tee ∂(Fφ)/∂t + div(FUφ - FΓφ grad φ) ) Sφ (1)
mixing quality with the commercially available fluid transient convection diffusion source
flow package PHOENICS. The k- model was used to
model turbulence. Their numerical results satisfactorily where φ is any conserved property, U is the velocity
matched the experimental results of Sroka and For- vector, Γφ is the exchange coefficient of φ in the phase,
ney.13,14 Sφ is the source rate of φ. Thus, the continuity equation,
Both of the above numerical models solved the for example, becomes
conservation equations for mass and momentum in
∂(F)/∂t + div(FU) ) 0 (2)
primitive variables for steady turbulent flow of a single-
phase fluid with an inert tracer introduced at the where F is the density of the fluid. The conservation of
injection point. The models also used a mixing criteria momentum for variable φ becomes
based on the standard deviation of the component mixed
and the mean value of the tracer over the pipe cross- div(FUφ - µeff grad φ) ) Sφ (3)
sectional area cj. The use of CFD, despite the above-
mentioned investigations, has still a lot to offer in where µeff is the effective viscosity and Sφ is the source
analyzing and understanding mixing in pipeline tees. of φ per unit volume. Similarly, the equation of energy
Simulation of variations of side- and opposed-tee mixers can be written.
has received limited attention in the literature.
Most of the research done on pipeline mixing has been Experimental Apparatus
for a configuration in which the jet is normal to the pipe-
line. Maruyama et al.15 carried out experiments for mix- The experimental setup used in the current study is
ing of two fluids at various angles. Feng et al.16 pre- shown in Figure 4. Hot water is injected through a 1/4-
sented an analytical solution using an asymptotic in. side tee (1/8 in. has also been used) and is mixed with
procedure to evaluate the tracer trajectory in a pipeline cold water flowing in a 1-in. main pipe. The temperature
with a side tee at various angles. The Feng et al.16 results is used as the measured variable to quantify mixing.
were in good agreement with those of Maruyama et al.15 Eight thermocouples are inserted at various positions
Feng et al.16 stated that, in many chemical engineering of the main and side pipes in order to measure the axial
processes, it is desirable to have the side jet impinge temperature of the flow and that of the incoming side
on the opposite wall in order to enhance mixing. flow. The thermocouples are connected via an OMEGA
The main interest of this paper concentrates around data-logging board to a PC as shown. Flow through the
the side and opposed tees shown in Figures 1 and 2. side tee passes through a heater that can raise the
Mixing of hot and cold fluids is investigated in pipelines temperature of the side stream significantly above that
Figure 4. Schematic diagram of the experimental rig used to investigate mixing of a pipeline with a side tee.
Ind. Eng. Chem. Res., Vol. 42, No. 21, 2003 5335
| |
c - cj At very low velocity ratios, the jet does not impinge
m) < 0.5 on the opposite wall and the high-pressure point is at
cj
the inlet of the tee. For higher velocity ratios, the side
where cj is the equilibrium concentration and c is the jet impinges on the opposite wall of the pipe and creates
concentration at any monitoring point at any time. a region of backflow. The high-pressure point is trans-
When the above condition is met at all monitoring points ferred to the opposite wall downstream of the axis of
in a cross-sectional plane of the main pipe, it can then the jet, while upstream of the jet, there is a zone of very
be said that the concentration at any point of the pipe low velocity.
after that length has reached 95% or more of the Before numerical results are compared with experi-
equilibrium concentration. mental results, the dependence of the numerical solu-
Other researchers defined uniformity criteria in terms tion on the grid size, turbulence model, and other factors
of the second and third moments about the arithmetic is first tested.
mean of an inert tracer at a given tube cross-sectional Dependence of Solution on the Grid Size
area. The second moment about a mean value is defined
for a given tube cross section A in the form The mesh size often has an impact on the accuracy
of the solution. The size has to be small enough in order
to properly resolve the fields that are solved for. To
m2 ) ∫A(c - cj)2u dA/∫Au dA quantitatively compare the results with different sizes,
mesh sizes of 4, 3, and 2 mm have been tested. The
In the present study, the flow in the main pipe before number of cells used for mesh sizes of 4, 3, and 2 mm
the jet inlet is set initially at a certain temperature. The are 18 610, 56 463, and 162 367 cells, respectively.
flow through the side tee is set at a higher but known The effects of the mesh size are tested by comparing
temperature. Because the flow rate and temperature of a plot of the temperature versus location along the
5336 Ind. Eng. Chem. Res., Vol. 42, No. 21, 2003
Figure 6. Velocity contours in a plane passing through the centerline for a mesh size of 2 mm.
Figure 7. Temperature contours in a plane passing through the centerline for a mesh size of 2 mm.
center line of the main pipe. Figure 8 shows such a using the Reynolds stress model (RSM) of turbulence
comparison for the three mesh sizes considered with for mesh sizes of 2 and 3 mm. A small difference is
turbulence modeled using the k- model. It is clear that observed mainly around the jet zone.
the solution shows a significant change when the mesh
size is reduced from 4 to 3 mm. The solution also Grid Refinement
changes when the mesh size is further reduced from 3
to 2 mm, but the difference between solutions of mesh The grid is locally refined especially in the area where
sizes of 3 and 2 mm is not very significant. The number the gradient of temperature and velocity is high. A
of cells for a mesh size of 2 mm is relatively very high; temperature gradient of 0.001 K/m was used as the
however, because the solution still shows some change, basis for mesh refinement. This refinement increased
a mesh size of 1 mm was attempted. This attempt could the number of cells from 162 367 to 183 161. A further
not be completed because the time required to perform refinement was introduced, and the number of cells
the meshing of the computational domain is prohibi- became 215 893 cells when a gradient of 0.0005 K/m was
tively excessive. Therefore, a mesh size of 2 mm was used.
used for all of the main runs in this study. An adapted grid is shown in Figure 10. More cells can
Figure 9 shows a comparison of the numerical results be seen along the boundary of the jet, where the
Ind. Eng. Chem. Res., Vol. 42, No. 21, 2003 5337
Turbulence Modeling
Figure 16. Temperature flow fields for tee angles of 30, 45, 60, and 90°, respectively.
velocity ratio of 17.1, the 95% mixing length was Experimental Results and Validation of the
reduced from 11 to 3 diameters. Numerical Model
On the basis of the above results, it is clear that there Experimental runs have been carried out with water
is very little incentive for the industry to use 90° tees flow rates through the main pipe of 7.02, 12.28, and
because these are likely to give the worst results in 19.30 L/min and water flow rates through the side pipe
terms of mixing and possible corrosion and erosion of 7.5, 5.0, and 3.0 L/min. The side-tee diameter is 1/4
problems. in. Details of these runs, together with the correspond-
Ind. Eng. Chem. Res., Vol. 42, No. 21, 2003 5341
Figure 17. Temperature profile along the axis of the main pipe
for tee angles of 30, 45, 60, and 90°.
Figure 20. 95% mixing length versus Uj/Um for all cases for an
1/ -in. right-angled side tee.
8
Figure 21. Pathline diagram of a side jet bending into the main
fluid as Uj/Um is increased: (a) low Uj/Um; (b) low to medium Uj/
Um; (c) high Uj/Um; (d) very high Uj/Um.
Table 2. Comparison of 1 in.-1/4 in. and 4 in.-1 in. Cases Keeping the Reynolds Number Constant
flow rate Re, Reynolds old Q flow velocity diameter scale-up new flow rate new velocity
kept constant number rates (L/min) (m/s) (in.) diameter (in.) (L/min) (m/s)
main 5842.0 7.01 0.23 1 4 28 0.06
jet 66802.0 19.99 10.52 0.25 1 80 2.63
ratio, jet/main 5.7 2.9 45.7 0.25 0.25 2.9 45.7
Table 3. Comparison of the Pipe Length Required for 95% for 1 in.-1/4 in. and 4 in.-1 in. Cases Keeping the Velocity,
Flow Rate, and Reynolds Number Constant with Base Case
length required, from entrance of side jet, for 95% mixing
case Dm-dj Remixed Uj/Um in. diameters of the main pipe
1 in.-1/4 in. base case 22 542.5 45.7 3 3
4 in.-1 in., velocity constant 90 170 45.7 9 2.25
4 in.-1 in., flow rate constant 5 639.3 45.7 9 2.25
4 in.-1 in., Reynolds number constant 22 542.5 45.7 9 2.25
the velocity ratio increases. However, for the same side- function of Uj/Um and dj/dm. The individual values of
to main-diameter ratios, the mixing length does not vary the velocities and pipe diameters did not have a
significantly. This means the length of the pipe required significant impact on the mixing length. For constant
for 95% mixing to be achieved in 1 in.-1/4 in., 4 in.-1 dj/dm and Uj/Um, the 95% mixing length is independent
in., and 16 in.-4 in. is about the same in diameters of of the value of the Reynolds number.
the main pipe. For velocity ratios smaller than 35, 95%
mixing increased marginally as the sizes of the pipe Acknowledgment
decreased. This trend was reversed for velocity ratios
larger than 35. The authors are grateful for the funding and support
On the basis of these results, it can be concluded that provided by King Fahd University of Petroleum &
the mixing length is a strong function of the velocity Minerals during the course of this work and the
ratio (Uj/Um) and the jet- to main-pipe-diameter ratio preparation of this paper.
(dj/dm). The actual size of the pipes does not have a
major impact on the 95% mixing length. Literature Cited
To investigate the issue of scale-up further, one case (1) Simpson, L. L. In Turbulence in Mixing Operations; Brod-
of 4 in.-1 in. was scaled up from 1 in.-1/4 in. Three key, R. S., Ed.; Academic Press: New York, 1975.
cases were considered with a constant velocity ratio. The (2) Gray, J. B. In Mixing: Theory and Practice; Gray, J. B.,
velocity ratio was kept constant either by (i) keeping Uhl, V. W., Eds.; Academic Press: New York, 1986; Chapter 13,
both velocities constant, which means that the values Vol. III.
of the Reynolds number were increased by a factor of (3) Forney, L. J. In Encyclopedia of Fluid Mechanics; Cher-
emisinoff, N. P., Ed.; Gulf Publishing Co.: Houston, TX, Chapter
4, (ii) keeping both flow rates constant, which means 25, Vol. II.
that the values of the Reynolds number were reduced (4) Liou, T.-M.; Liao, C.-C.; Chen, S.-H.; Lin, H.-M. Study on
by a factor of 4, or (iii) keeping the Reynolds number Side-Jet Injection near a Duct Entry with Various Injection Angles.
constant, which means that the flow rates were in- Trans. ASME 1999, 121, 580.
creased by a factor of 4 in order to keep Uj/Um constant. (5) Weber, L. J.; Schumate, E. D.; Mawer, N. Experiments on
Table 2 shows typical data for the above third case. The Flow at a 90° Open-Channel Junction. J. Hydraul. Eng. 2001, 127,
340.
95% mixing length for all three cases was found to be (6) Lam, K. M.; Xia, L. P. Experimental Simulation of a Vertical
constant as shown in Table 3. Round Jet Issuing into an Unsteady Cross-Flow. J. Hydraul. Eng.
A case with a main velocity of 1 m/s and a side velocity 2001, 127, 369.
of 17.1 m/s was also simulated for the 16 in.-1 in. case (7) Pan, G.; Meng, H. Experimental Study of Turbulent Mixing
for Uj/Um of 17.1. The 95% mixing length was found to in a Tee Mixer Using PIV and PLIF. AIChE J. 2001, 47, 2653.
be 9.5 diameters, which is the same for the base case of (8) Moussa, Z.; Trischka, J.; Eskinazi, S. The Near Field in the
Mixing of a round Jet with a Cross Stream. J. Fluid Mech. 1977,
Uj/Um of 17.1 where Um is 0.23 m/s and Uj is 3.933 m/s.
80, 49.
This shows that the 95% mixing length is affected by (9) Crabb, D.; Durao, D.; Whitelaw, J. H. A Round Jet Normal
the velocity ratio and not by the individual values of to a Cross Flow. J. Fluids Eng. 1981, 103, 143.
the velocities. (10) Cozewith, C.; Ver Strate, G.; Dalton, T. J.; Frederick, J.
W.; Ponzi, P. R. Computer Simulation of Tee Mixers for Non-
Conclusions Reactive and Reactive Flows. Ind. Eng. Chem. Res. 1991, 30, 270.
(11) Forney, L. J.; Monclova, L. A. Numerical Simulation of
Mixing of cold and warm water in pipelines with side Pipeline Tee Mixers: Comparison with Data. In Industrial Mixing
and opposed tees was experimentally and numerically Technology: Chemical and Biological Applications; Gaden, E. L.,
Tatterson, G. B., Calabrese, R. V., Penney, W. R., Eds.; p 141.
investigated. A good agreement between the experimen- (12) Cozewith, C.; Busko, M. Design Co-Relations for Mixing
tal and numerical results is observed. The numerical Tees. Ind. Eng. Chem. Res. 1989, 28, 1521.
and experimental values of the temperature distribution (13) Sroka, L. M.; Forney, L. J. Fluid Mixing with a Pipeline
and the distance required to achieve 95% mixing are Tee: Theory and experiment. AIChE J. 1989, 35, 212.
almost identical. The RSM and the k- model were used (14) Sroka, L. M.; Forney, L. J. Fluid Mixing in a 90° Pipeline
to account for turbulence and gave similar results except Elbow. Ind. Eng. Chem. Res. 1989, 28, 850.
(15) Maruyama, T.; Suzuki, S.; Mizushina, T. Pipeline Mixing
in the vicinity of the jet impingement region. between Two Fluid streams Meeting at a T-Junction. Int. Chem.
Results showed that the pipe length required to Eng. 1981, 21, 205.
achieve 95% mixing is a strong function of the ratio of (16) Feng, Z.; Wang, X.; Forney, L. J. Single Jet Mixing at
Uj/Um and the angle of injection. A jet centered along Arbitrary Angle in Turbulent Tube Flow. Trans. ASME 1999, 121,
the axis of the main pipe gives the shortest mixing 762.
length. The mixing length decreases as the velocity ratio (17) Jayanti, S. Hydrodynamics of Jet Mixing in Vessels. Chem.
Eng. Sci. 2001, 56, 193.
increases. The rate of change becomes slower at velocity (18) Zughbi, H. D.; Rakib, M. A. Numerical Simulations of
ratios higher than 30. For certain velocity ratios, the Mixing in a Fluid Jet Agitated Tank. Chem. Eng. Commun. 2002,
95% mixing length may start to increase if the velocity 189, 1038.
ratio is increased over a threshold value. The angle of (19) Patwardhan, A. W. CFD modeling of jet mixed tanks.
injection had a significant impact on the mixing length. Chem. Eng. Sci. 2002, 57, 1307.
A 90° jet angle showed the strongest impingememnt on
the opposite walls. For a velocity ratio of 17.1, the 95% Received for review December 6, 2002
mixing length was found to be the shortest for an angle Revised manuscript received July 7, 2003
Accepted July 10, 2003
of 45 or 165°. This optimum angle is likely to change
for a different velocity ratio. The mixing length is a IE0209935