Beruflich Dokumente
Kultur Dokumente
A. P. Szilas, E. B o b o k , a n d L. N a v r a t i l
With 6 figures and 2 tables
(Received February 9, 1981)
There is lack of sufficient experimental data The time-averaged velocity field is steady and
obtained b y / n situ measurements of crude oil axisymmetric. Its only component v lies in the
pipelines in a wide range of Reynolds-numbers. direction of the pipe axis. It is assumed that the
On the other hand, determination of the fric- distributions of the velocity fluctuations u' and
tion factor equations is based on doubtful w' are axisymmetric, too. When using these re-
velocity measurements. strictions the equation of momentum balance is
In the present paper we show another pos- as follows:
sible way to determine the friction factor rela-
tionship for turbulent flow of pseudoplastic oil pgJr
2 ~1' d_~r n - pu' w' = 0. [1]
in rough pipes at the "partially rough wall"
region, where friction factor is affected by both
relative roughness and Reynolds number. The first term of the equation is the driving
In order to get such a relationship: force of motion, in which J i s the dimensionless
- w e deduced a theoretical expression, in hydraulic gradient. The second is the resistance
which approximate assumptions play a negli- due to viscous shear stress, in which t/' is the
gible role; index of consistency and n is the so-called flow
- we made a great number of in situ measure- index. The third term is the resistance due to an
ments on a 305 mm diameter crude oil additional virtual stress, caused by momentum
pipeline in a wide range of Reynolds num- flux due to turbulent fluctuations.
ber; It is necessary to distinguish between two
- we compared the accuracy of our (BNS) for- regimes when solving the above differential
mula with the previous equations. The ac- e q u a t i o n , in accordance with experimental
curacy and utility of the BNS equation are data.
clearly demonstrated. In the viscous sublayer turbulent fluctuations
practically disappear, thus the third term of the
2. F o r m u l a t i o n o f the p r o b l e m equation may be omitted. We get the following
simple form:
We consider a homogeneous, isotropic turbu-
lent flow in a cylindrical tube of radius R and of pgJr _ ~, dv n.
infinite length. The fluid is incompressible, and dr [21
its rheological behaviour is pseudoplastic. On
the pipe wall a thin viscous (often called
"laminar") sublayer occurs, in which turbulent Since the viscous sublayer is very thin, an
fluctuations are negligible. On the other hand, additional assumption may be accepted that the
in the core flow viscosity is assumed to have a shear stress r is constant throughout the
negligible effect on the random turbulent fluc- sublayer:
tuations. A cylindrical coordinate system is
chosen as it is shown in figure 1.
where
Turbu[ent
core f[ow has the dimension of velocity. It will be called
friction velocity. Thus
which after integration leads to a linear velocity arbitrary additional term. i t is obvious that
distribution velocities are equal on both sides of the
bounding surface of the core flow and the sub-
layer. Thus from eqs. [61 and [10] we get:
This is a formula for the velocity distribution means the so-called Metzner-Reed Reynolds
of the core flow, but with Vmax as an unknown number (7), D = 2R is the diameter of the pipe.
490 Rheologica Acta, Vol. 20, No. 5 (1981)
+ [22]
v, R27r o v , \ R e * t? (2-n)/2n 3.71 D /
Substituting v/v, from eq. [10], and Vmax/V , where fl consists of the last three terms of eq.
from eq. [15] we get after integration: [20]. This is a generalization of the Colebrook-
formula and clearly demonstrates how friction
c _ 1 In Re* 1 factor depends on the pseudoplastic Reynolds-
V, K K number, the flow index and the relative rough-
[181 ness of the pipe. The BNS-eqution is a friction
. 548 + In a + factor relationship is valid for turbulent flow of
non-newtonian fluids flowing in rough pipe at
\ 240 n n
the transition region between the wholly rough
It is clear that the quantity c / v , depends only and the smooth pipe character.
on the Reynolds number. On this base the
friction factor can be deduced similarly to the 3. Experimental results
newtonian case:
A significant series of experimental verifica-
__1 _ __0"8141lg(Re,21_(n/2)) + 0.7532__n-2 tions of the BNS-equation were made on the
]/2 nK 2nK Algy-Szzhalombatta crude oll pipeline having
a diameter of 305 mm and a length of 161.3 km.
The temperature of pseudoplastic oll decreased
[19]
in the direction of flow, thus the corresponding
(~;~6n+~3 n
value of flow index also decreased with tem-
perature. Test stations were designed and equip-
ped using a number of precision instruments for
pressure, temperature and flow rate measure-
Substituting n = 1 into this equation we may ments along the pipeline at eight several points
evaluate constants a and x from the Prandtl- (see tble 1). Flow rate was controlled by gate
Nikuradse equation. Thus we get valves, thus Reynolds numbers varied w i t h i n
the range: 10 4 < Re* < 105.
= 2.087, K = 0.407. Measurements were repeated five times ac-
cording to the different temperatures of the
In this way the friction factor equation is ground (1). At the same time rheological prop-
obtained for hydraulically smooth pipes: erties of the oil were measured by a rotational
viscometer at the corresponding temperature.
1 2 lg(Re*21-(n/2)) + 1.511/n The interpretation of the laboratory measure-
ments involved the calculation of points to
(;07)40,,
.0"__ + 2.12
/7
[20]
1.057 .
determine the flow curve of the oil at the
adequate temperature. The parameters t/' and n
were determined from the flow curves. Know-
ing the flow rate and the rheological parameters
of the oil, the pseudoplastic Reynolds-number
It is obvious, that ;, can be determined for Re* can be determined. The interval between
rough wall turbulence as 10 4 and 10 indicated the transition nature of
the flow.
[21] Pressure loss per unit length was determined
from pressure measurements, as it is demon-
strated in figure 2. The series of measured
In the transition region between the smooth points were interpolated by a linear function
and rough wall case may propose the-following using the method of the least squares fitting.
Szilas et al., Determination o f turbulent pressure loss o f non-newtonian oil f l o w in rough pipes 491
0,5
Ap
L
[~r 1 0,4
kmJ
--o ~
0,3 ~0 r mO
0,2
0,1
I I
0 50 100 150
L [ kml
Fig. 2. Measured and calculated flow gradients
Pressure gradients of the five test series are de- factors obtained by the Dodge-Metzner,
monstrated in table 1. The temperature distri- Shaver-Merrill, Tomita, Clapp, as weil as the
bution was also linear along the pipe length, BNS-equation. Scattering, absolute and relative
thus is had been convenient to take the arith- mean errors were determined for every case. At
metic mean for any tested section of the pipe- the end it appears that BNS-equation provides
line. Friction factor values were determined by the best results. It is interesting to see that the
the Weisbach-equation: Clapp-formula is successful to predict 2 values
for only one value of relative roughness. As is
2D Apv shown in figure 3, the curve corresponding to
pc 2 L the Clapp-formula intersects some k / d =
const, curves on the M o o d y ' s diagram (assum-
ing the case n = 1). It can be evidenced that
There were seven tested sections of the pipe- such a relationship is obtained evaluating the
line, thus 35 measured friction factor values constant of the smooth pipe friction factor
were obtained f r o m the five test series. These equation f r o m experimental data of rough
values were compared with computed friction pipes.
492 Rheologica Acta, Vol. 20, No. 5 (1981)
Table 2. Measured and calculated friction factors at Algy-Szzhalombatta crude oil pipeline
Friction factor
0,0350
0,0300
2~
0,0 250
~ ~ ~ = -
0,0150
104 2 3 4 5 6 7 8 9 105
. Re ~
Fig. 3. Friction factor cnrves according to C]app (1) and Bobok-Navratil-Szi]as (2) presented on the transi-
tional part o f M o o d y - d i a g r a m
0,50 n = 0,60
r~' = 0,0117 Ns/m z
bp = 820 kglm 3
L D = 0,3m
EN]
0,40
1. BNS, k/D=lO 3 1
2. Tomita /2
3. BNS~ k/D=5,1 4 //
0,30
~. 0-
5 CLAPP //
II 13
6. BNS~ k/D=3.1 5
7. S - M
0,20
6
0,10
favourable than the others from the point of to be expected for commerical steel pipes (1).
view of both scattering and absolute and The flow indices of the oil are n = 0.6 in the
relative errors. In this special case BNS-equa- case demonstrated in figure 4, and n -- 0.9 in
tion is not very much better than Clapp's rela- figure 5. Especially in figure 4, it is conspicious,
tionship. In spite of this, BNS equation is valid that deviation of pressure gradients determinat-
for arbitrary relative roughnesses, at which the ed by different methods increases with the flow
error of Clapp's formula can be important. rate (thus with the Reynolds number). From a
comparison of figures 4 and 5 it is evident that
4. Accuracy of pressure loss determination
~deviation of pressure gradients decreases with
In the following the accuracy of pressure loss :increasing flow index (i.e. approaching the
determination will be demonstrated at any sup- newtonian behaviour).
posed transport conditions (flow rate, oil prop- It is also evident that earlier friction factor
erties) as well as the error to be expected using equations without reference to relative rough-
the earlier friction factor equations. We con- ness may lead to major errors. By chance it may
sider the inner diameter of the pipe and the flow be possible to obtain a good result as is near the
curves of the psendoplastic oll to be given. Cha- only point of Tomita's curve, in which it is in-
racteristic parameters are shown in figures 4 tersected by the BNS curve of k / d = 10 -3. In
and 5 where pressure gradient A p / L is plotted figure 6 BNS pressure gradients are compared
versus flow rate q. The three supposed different with Clapp's curves depending on the flow
relative roughnesses for the BNS-equation are index n, at some constant flow rate. It is
0,50 n = 0,90
r~ = 0,0117 Nsn/m 2
Ap
= 820 kg/m 3
L D = 0,3 m
[ bar]
kmJ
0,40
1. BNS, k / D = 1 0 -3
2. Tomita
3. BNS, k / D = 5 ' I 4
4. D - M
0,30
5. C L A P P
6 BNS, k/D=3.1 s
7 S-M
0,20
0,10
0 I ] i i I
Fig. 5. Calculated friction gradient
0 100 200 3OO curves after formulas of different
q [m3/h] authors at great flow index
Szilas et al., Determination of turbulent pressure loss of non-newtonian oil flow in rough pipes 495
0,50
1. k/D= 3.10
2 k/D = 5.1 4 st BNS
3. klo = l 3
t~p 300 m3/h
L
[bar]
-~mJ
0,40
0,30
zoo M/h
0,20
0,10
obvious that Clapp's curve may approach a cer- (i) Dodge-Metzner formula:
tain BNS pressure gradient referring to some
relative roughness, for example the k / d = 3 1_ 2 / / ~ \ , (,,2q
10 -5 curve at flow rate 300 m3/h. In other situa- B nO.7~lg~Re, k__~) j n120"2"
tions, however, the discrepancy may be im-
portant. Here it can also be seen that the dif- (ii) Shaver-Merrill equation:
ference increases with growing flow rate q and 0.316
2-
with decreasing magnitude of n. n 5(Re*)m
BNS equation is slightly more complicate where
than earlier relationships, but this seems to be
2.63
quite unimportant when using digital com- m-
1.05 n "
puters. At is customary the necessary knowl-
edge of relative roughness for pipeline designing (iii) Tomita's formula:
can be taken from performance data of existing
similar pipelines, like in the newtonian case for B - 2lg e* - 0.2.
using the Colebrook formula.
(iv) Clapp's equation:
Appendix 2.27 [ //,& \1 (/2)1
1 _ 1.35__ l ' 4 8 + - - l g [ R e * [ k - 4 - B J
The investigated friction-factor relationships for B n n
time-independent visous non-newtonian fluids are as
follows. All equations are valid for hydraulically 5n - 8
+ 0.34--
smooth pipes. n
496 Rheologica Acta, Vol. 20, No. 5 (1981)