Anda di halaman 1dari 14

c  



        
  !!"   

a a a
 
 aa   
 

 

Vyacheslav M. Kraev and Dmitry S. Yanyshev*


Moscow Aviation Institute (State Technical University), 125993 Moscow, Russia
(* Corresponding author: dyanishev@list.ru)

  The survey of the current state of the problem of analysis of hydrodynamically
unsteady turbulent flow in channels is presented. Main approaches to the problem and analysis of some
obtained results are given. The results of calculations made in MAI using modern approaches to
turbulence modeling are provided.

 
a

Ô · channel diameter, m
F ± cross-sectional area of channel, m2
G - mass flow rate, kg/s
Ho ± time parameter, dimensionless
Ho ± time parameter referring to the total time of the process, dimensionless.
Ho - comparative time parameter, dimensionless
 * - hydrodynamic unsteadiness criteria, dimensionless
pst ± static pressure, Pa
qv ± volume heat generation, W/m3
R ± channel radius, m
r ± radius, m
Sij ± strain rate tensor, 1/s
w ± mean flowrate velocity, m/s
X ± mass force density, m/s2
İq ± eddy heat conductivity, W/m2 K
İȞ ± eddy viscosity, Pa s
Ȝ ± heat conductivity coefficient of the gas, W/m2 K
ȟ ± friction coefficient, dimensionless
IJ · time, s


 ± quasisteady value
i, j, k ± tensor indices
m ± modified value
max ± maximum
min - minimal

 
- Reynolds time averaging operation
¶ ± pulsation part of a parameter
c   

        
  !!"   

 

The problem of system development under unsteady conditions is still remaining one of the critical
ones, and cannot be fully resolved yet. So analysis of the unsteady thermal and hydrodynamic
processes is one of the most important ones while designing advanced devices in such spheres like:
aviation, astronautics, shipbuilding, cryogenics, chemical industry, etc.

By now, considerable test data had been obtained for unsteady single-phase flows in round tubes,
flat ducts, and in shaped channels. It allows carrying out one-dimensional analysis in a variety of
heat-stressed structures and heat exchangers. Experimental researches and theoretical estimations
show that values of heat transfer and friction coefficients may be three or even four times larger
than the results of so-called quasisteady analysis, i.e. the analysis, in which formulas obtained for
steady conditions are used.

Experimental researches carried out in Moscow Aviation Institute (State Technical University)
showed that unsteadiness of mass flow rate at the inlet of the channel influences greatly on the flow
structure. According to this, flow hydrodynamic unsteadiness should considerably affect on heat
transfer. This influence may be explained by specific changes in turbulent structure [Kalinin E.K.,
Dreitser G.A., 1994).

The main issue here is that the effects of unsteady conditions on processes in heat exchangers and
other heat-stressed devices could be great, and quasisteady approach neglecting these effects is
often wrong in case of designing of such devices. So general engineering methods, which will take
unsteady effects, are to be developed.
In this paper, analysis of turbulent structure and physicals phenomena of unsteady flows and several
approaches to modeling of such flows are presented.

 


  !"# "$%!&% ' # &!()!(*Method of measurement of temperature and velocity
turbulent pulsations was based on TSI thermo-anemometer equipment.
The researches were carried out in tube with inner diameter of 42 mm under flow acceleration and
deceleration. Non-isothermal conditions were provided by channel wall heating up toO O1,18.
Experimental facility let to realize a hydraulic unsteadiness in range of mass flow changing rate of
|´£´ |,2 kg/sec2. The time of unsteady process varied from 2 up to 5 sec.

Main parameter characterizing hydraulic unsteadiness was the following:

´£ 1 Ô
   (1)
´ £ 

where £ · mass flow rate, kg/sec;  · time, sec; Ô · channel diameter, m;  · free fall acceleration,
m/sec2.

It was varied in the range from ·,111 up to ,111.

Time parameter was chosen as follows:


c   

        
  !!"   

a
a  (2)
a 

where a   ± time parameter corresponding to the current time of unsteady process and a 
± the one corresponding to the total time of the process.

Research methods and experimental facility are stated in detail in the work of Dreitser G.A., et al
[1994].

( &)($+ &% $&%$&$)"*%$%*&!#& !%&  A plenty of unsteady


turbulent flow researches [e.g. Markov S.B,1973; Koshkin V.K. et al., 197; V.Kraev et al., 22]
showed that velocity and eddy viscosity distributions on the channel radius considerably depend on
flow acceleration and deceleration.

At flow acceleration axial velocity profile becomes more µfilled¶, and value of eddy viscosity
coefficient in near-wall region increases.

Figure 1. Qualitative variations of the velocity profiles at flow acceleration and deceleration
[Kalinin and Dreitser, 1994]

At flow deceleration, opposite structure changes were found - axial velocity profile becomes less
µfilled¶, and value of eddy viscosity coefficient in near-wall region decreases. It was also found that
near-wall zone, where the most significant changes took place, becomes narrower than in flow
acceleration case (see Figure 1 and 2).

Figure 2. Qualitative variations of the eddy viscosity profiles at flow acceleration and deceleration
[Kalinin and Dreitser, 1994]
c   

        
  !!"   

Changes in eddy structure of turbulent flow in unsteady conditions affect on pressure losses. The
deviations of the friction coefficient ȟ from the quasisteady values ȟ for different flow regimes are
shown in Figures 3-6

Figure 3. Flow acceleration influence on friction coefficient ½ 31 93, O O1, a- 
.111, b-   .88.

Figure 4. Flow acceleration influence on friction coefficient Re62 187, Tw/Tf1, a- Kg *max
.59, b- Kg *max .35.

Figure 5. Flow deceleration influence on friction coefficient ½ 31 93, O O1, a-   -
.111, b-    -.44.
c   

        
  !!"   

Figure 6. Flow deceleration influence on friction coefficient ½ 62 187, O O1, a- 





As it could be seen from the figures above, the increasing of the Reynolds number decreases the effects
from hydrodynamic unsteadiness. So when Reynolds number is great enough we can use quasisteady
approach.

( & )($+   &%  $&%$& $) )($+ &$& ,$" #!(* !&% "*%$%*&!#
& !%&     Maximal increase of turbulent structure parameters was found, like in isothermal
case, in the middle of unsteady process.

At flow acceleration axial velocity profile becomes more µfilled¶, like in isothermal case, but
unsteady influence on the turbulent structure is more significant.

It was also found that in non-isothermal conditions turbulent structure parameters come to the
steady value slower, than in isothermal conditions.


Figure 7. Flow acceleration influence on Nusselt number ½ 31 93 ½ 31 93,
O O1.18, a-  .111, b-   .88.


c   

        
  !!"   

Figure 8 Flow acceleration influence on Nusselt number Re62 187, Tw/Tf1, a- Kg *max
.59, b- Kg *max .35.

Figure 9. Flow deceleration influence on Nusselt number ½ 31 93 ½ 31 93,
O O1.18, a-  -.111, b-   -.88.



Figure 1. Flow deceleration influence on Nusselt number ½ 62 187, O O1, a-  


 


c   

        
  !!"   


Flow acceleration/deceleration influence on heat transfer coefficient Į is significant. Test data for
Nusselt number Nu change under the conditions of hydrodynamic unsteadiness is shown in Fig 7-
1.

The influence of Reynolds number on heat transfer is the same as on friction coefficient.

 - &* !&!(* $) )($+   !!#      It is a well known fact [e.g. Kraev V.M.,
25] that results of frequency analysis are very useful for turbulence phenomena understanding.
The analyses and estimations showed that generation of large (with pulsation frequency of 3-7
Hz) and medium (7-2 Hz) vortexes takes place in the wall region. These vortexes disintegrate to
smaller ones (2-5 Hz) on their way to the flow core. Wall heating in steady conditions comes
to turbulence generation intensification.

Hydrodynamic unsteadiness (like non-isothermal conditions) has a very strong influence on


turbulent spectrum. At flow acceleration, large vortexes disintegrate much faster than in steady
conditions. At flow deceleration opposite effect was found ± vortexes disintegrate slower.

"*!( #$% ( $) & !%* )($+( &   After analysis of experimental research
results and frequency analysis, we suggest that non-isothermal conditions and hydraulic
unsteadiness influence great on physics of processes taking place in the near-wall region.

Corino E.R., Brodkey R.S. [1969] showed that near-wall region is the main source of turbulent
vortexes. Turbulence energy production also depends on interaction of near-wall region with the
main flow. Under conditions of flow acceleration an axial velocity profile becomes more µfilled¶,
that comes to µcompression¶ of near-wall region. That then comes to more intensive interaction
between these two regions and high increase of turbulent vortexes ejection into main flow. Further,
when the vortex comes into flow zones with higher (in comparison with steady case) stress gradient,
due to flow acceleration, and destroy they self on smaller vortexes faster than in steady conditions.
These processes lead to eddy viscosity and eddy heat conductivity coefficients significant
increasing.

The opposite effect is caused by flow deceleration.



   

 

& ,%# &$&!( !$!"    The results of test data mentioned above were generalized into
empirical formulas for Nusselt number x and friction coefficient m. With these formulas, one-
dimensional calculations of turbulent flows in round tubes with transient mass flow at the inlet may
be now performed.

One-dimensional form of the equations of motion, continuity and energy for calorically perfect gas
flow in round tubes is the following:

£´ ´  ´ ^ 2
£    ^½ · ·m (3)
´ ´  ´ Ô 

´^ ´ £
   (4)
´ ´
c   

        
  !!"   

£ ´ O  ´ O
´
 £
´
Ú
  x  O ·O  (5)

The empirical formulas for friction coefficient and Nusselt number generalizing experimental data
obtained in Moscow Aviation Institute are given bellow.

The formulas for the friction coefficient are the following.

Isothermal accelerating flow:

- a -.33
m  m  (.212  .126 a ) (.25 - .146  *) -.5 (6)

Isothermal deceleration flow:


.48 a -.4
m  m . (.819  .484   *) 2  .5 (7)

Non-isothermal accelerating flow:

- a -.3
m  m (.312  .379a ) (.1- .6  *)-.2 (8)

Non-isothermal decelerating flow:

.789 a -.4
m  m . (.557  .481   *)  .68 (9)

In these formulas ȟ is quasisteady friction coefficient and it is calculated with Blasius formula:

m   .3164 (1)
Re .25

Accuracy of the test data concerning friction coefficient is not very high (the error may reach the
value of about 2), which is connected with the difficulty of velocity gradient measuring and with
some design features of the test facility.

However, test data obtained for Nusselt number is more accurate ± the error is about 8.

Formula for Nusselt number is the following:

 .23   ,8  ,4 (11)

where ½  is modified Reynolds number.

For accelerated flow it is the following:

] - a -.5
     .9391.176 a .4791.374 * 49.7 (12)
Ú 
 
c   

        
  !!"   

For decelerated flow we have:

 ½ ] 2.66  1.27 a
Ú - a -.5
½ .25  2.27 *  224.42 (13)
  

These formulas were obtained for the following range of parameters:

Re31...28, K*g-.111 .111, O O1 1.18

Actuating medium is air.

The comparison of results obtained with quasisteady calculation and calculation using Eq. (6-13)
are presented at Fig. 11-14

Fig. 14 Friction coefficient calculated using a ± quasisteady approach, b ± Eq. (11-13).


½ 31...93, O O1.18,  .1

Fig. 12 Friction coefficient calculated using a ± quasisteady approach, b ± Eq. (11-13).


½ 31...93, O O1.18,  -.1
c   

        
  !!"   

Fig. 13 Nusselt number calculated using a ± quasisteady approach, b ± Eq. (11-13). ½ 31...93,
O O1.18,  .1



Fig. 14 Nusselt number calculated using a ± quasisteady approach, b ± Eq. (11-13). ½ 31...93,
O O1.18,  -.1


Besides stated empirical formulas one can use a very simple formula, derived from the eq. (3)
assuming that pressure gradient is quasisteady, X, and velocity gradient in the direction of channel
axis is zero:

 1 
m  m   2  
  (14)
 Ú 

Eq. (14) is not very accurate, but it could be useful for preliminary designing calculations.
c   

        
  !!"   

 & !(. % " ,%# &$&!( !$!"    Sometimes carrying out tests is very difficult and
expensive. In such cases multidimensional approach is used.
This approach implies solving of Navier-Stokes equations. For Newtonian fluid they can be written
(using Cartesian tensor notation and Einstein¶s convention of summation over repeated indices) as:

´^ ´ ( ^  )
  (15)
´ ´ 

´ (^  ) ´ (^   ) ´ ´ ] ´  ´ 
 ·       ^  (16)
´ ´  ´  ´     ´   ´  

´ (    ) ´ (    ) ´ ] ´ 
      (17)
´ ´  ´   ´   

Here pressure includes thermodynamic, or static, pressure  and a term proportional to the trace
of the strain rate tensor :

   23 ö   (18)

1 ´  ´  
    (19)
2  ´   ´  

It is a well known fact that equations (15)-(17) have smooth solution only in case of laminar flow.
While the flow is turbulent it is very hard to solve them directly (although this approach sometimes
could be used). One of the most popular way of solving Navier-Stokes equation in case of turbulent
flows is Reynolds Averaging approach [see e.g. Willcox, 1994).

The system of averaged Navier-Stokes equations is usually called RANS (Reynolds Averaged
Navier-Stokes) equations. For the incompressible fluid they are the following:

´ 
 (2)
´ 

]´ (  ) ´ (   ) ´  ´ ] ´ 
^  ·   · ^ '  '   ^   (21)
 ´ ´    ´  ´    ´   

]´ O ´O ´ ] ´O
^       · ^   ' O '    (22)
 ´ ´    ´    ´   

Equations (2-22) contain nine additional variables: '  '  and  'O ' . So they need additional closure
equations.

!*$)($ &! $)& !%*)($+&  Nowadays, the most popular approach is still
the Boussinesq hypothesis, which says that turbulent additional stresses and heat flows may be
c   

        
  !!"   

expressed as a function of corresponding gradients. This leads to the conception of eddy transport
coefficients, i.e. to eddy viscosity and eddy conductivity.

 ´ 
   ½ (23)
´ 

´O '
· ' O '    (24)
´ 

This approach can be used only in case of isotropic turbulence. Turbulence in pipe flow is usually
considered as isotropic. This assumption is acceptable for unsteady flow too.

Analysis performed in Moscow Aviation Institute showed that most popular and widely used k-
epsilon turbulent model cannot be applied in this particular case, because standard k-epsilon model
belongs to the group of high-Reynolds models [see e.g. Wilcox, 1994] so it does not reflect the
complex process of wall layer restructuring in flows with inconstant mass flow rate at inlet. Thus it
needs to be modified.

In the works of Derevitch I.V. [25] and Valueva E.P. and Chen Ley [2] authors use modified
two-parameter k-epsilon model. They considered only the part of the flow where axial velocity
gradient is zero. As a boundary condition they took standard von Karman velocity profile [e.g. see
Wilcox D.C., 1994]. However, their results do not go with the experimental data. Moreover, their
results are directly opposite to the experimental data, viz the flow acceleration leads to Nusselt
number and friction coefficient decreasing and vice versa.

Derevitch I.V. [25] remarked that his theory could match with the test data in case of artificial
deformation of the Karman velocity profile at the inlet. In the light of it he supposed that this
deformation was connected with some design features of the feeding device.

Nevertheless, we suggest that the deformation should not be considered as µartificial¶. Kalinin E.K.
and Dreitser G.A. [1994], and Kraev V.M. [25] showed that unsteady velocity profile differs
significantly from the steady one. So the reason of discrepancy between estimated by Derevitch I.V.
[25] Valueva E.P. and Chen Ley [2], and observed data is that these authors considered only
the self-similar part of the flow.

Calculations performed in Moscow Aviation Institute considered the channel with the geometry of
the real experimental facility used by Dreitser G.A., Neverov A.S., Bukharkin V.B., Kraev V.M.
[1994]. The boundary conditions were the following: mass flow rate as function of time at inlet,
ambient pressure at outlet and slip-conditions on the wall.

To solve the RANS equations system, Finite Volumes Method was used. The problem (because of the
symmetry) was considered in 2-d. Channel was subdivided into 1x3 finite volumes so that y+”1.

Two-parametric Menter¶s Shear Stress Transport (SST) model was used as a turbulence model. This
model is of low-Reynolds group, so it does not need wall functions.

The results go with the test data qualitatively. Difference between test data and calculation results is
about 25 (see Figure 15). We suppose that for more accurate calculations SST turbulence model
should be somehow modified.
c   

        
  !!"   

Figure 15. Calculations using SST model in comparison with experimental data of Kraev V.M. et al
[22]; Re31 93, Tw/Tf1, Kg*max .88. m is calculated with Blasius formula for pipes.

Another thing to be discussed here is the question of flow self-similarity. The results of calculation
showed that hydrodynamic unsteadiness may disturb it.

This effect is especially strong if we consider temperature profile self-similarity. It could be seen if we
trace the value of flow temperature along the axis of the channel. It is convenient here to use so-called
µreduced temperature¶:

O ·O
a (24)
O · O  

ș distribution along the channel axis is presented on Figure 16.


Figure 16. Calculated ș distribution along the channel axis, o  1, ½ 31 93, Kg * .88.
c   

        
  !!"   

We should remark that the effect of self-similarity disturbance has not been verified experimentally
yet.

 
a 

As it can be seen, the main problem of CFD calculations of the unsteady turbulent flows is that we
need very fine meshes to take into account unsteady effects. The approach of self-similar flow
sometimes may be inconsistent so calculation in many cases will need much more CPU time.

In Moscow Aviation Institute we are currently trying to develop a new turbulence model for general
engineering applications for flows with inconstant mass flow rate.

  

Corino E.R., Brodkey R.S. [1969], A visual investigation of the wall region in turbulent flow,

 Ô   37(1), pp. 1-3
Derevitch I.V. [25], About unsteady hydrodynamics modeling in case of turbulent flow in pipes,
a O     Vol 43, ʋ2, pp. 231-248
Dreitser G.A., Neverov A.S., Bukharkin V.B., Kraev V.M. [1994], Experimental study of the
structure of turbulent unsteady gas flow in tubes,         Ô   
a      O     a   O    pp. 223-227
Kalinin E.K., Dreitser G.A. [1994], Unsteady Convective Heat Transfer in Channels, Ô  
a  O , Vol.25., pp.1-15.
Koshkin V.K., Kalinin E.K., Dreitser G.A., Galitseyski B.M., Izosimov V.G. [197], Experimental
study nonsteady convective heat transfer in tubes, 
 a  Ô  O  Vol. 13, ʋ 8,
pp. 1271-1281.
Kraev V.M. [25], Heat Exchange and Hydrodynamics of Turbulent Flows under Conditions of
Hydrodynamic Nonstationarity, ½      x   x       ,
3, pp.39-42.
Kraev V.M., Dreitser G.A. and Bukharkin V. B. [22], Experimental study of the infuence of
hydrodynamic unsteadiness on a turbulent tube gas flow structure,   Ô   aO
!
a  Ô  O      Ô " x  a  Ô  O       # 
Ô pp. 263-268.
Markov S.B. [1973], Experimental structure and hydraulic r esistance research in unsteady
turbulent flows,  Ô Ô £   ʋ 2. pp. 65±75.
Menter F.R., [1993] Zonal two equation k-Ȧ turbulence models for aerodynamic flows, 
 , N93-296, 21p.
Valueva E.P., Chen Ley [2], Numerical modeling of heat transfer and hydrodynamic processes
of unsteady turbulent flow with variable properties in pipe,    ! 
  $   ʋ 6, p.38.
Wilcox D.C. [1994], Turbulence modeling for CFD,  %Ô California, 537p.

Anda mungkin juga menyukai