www.elsevier.com/locate/flowmeasinst
Abstract
Generally, flow conditioner efficiency is experimentally evaluated on the basis of the effects that these devices produce on a
particular type of flowmeter. This approach, beside being very expensive, has certainly obstructed the development of a general
theory and consequently the optimal design of flow conditioners. The present paper presents a comparison of the performance of
different flow conditioners, independently from their effects on particular flowmeters. Therefore, the authors propose the use of
several efficiency parameters to evaluate and compare flatness, uni-directionality and axi-symmetry of the velocity profiles produced
downstream from the conditioners. In order to compare the main conditioners, a brief description of their characteristics is reported.
Then, the numerical analysis is carried out using the finite volume technique. From the results obtained it was possible to calculate
the efficiency of the conditioners under low-level disturbance conditions, such as those produced by an out of plane double elbow
configuration. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: CFD; Finite volumes; Flow conditioners; Installation effects
1. Introduction
Flow measurement is strongly influenced by the velocity profile. Since flowmeters are calibrated and characterised under completely developed flow conditions, perturbations such as swirl, cross-flow, and asymmetry can
produce relevant systematic errors [1].
In practical applications, the fully developed conditions can hardly be obtained, in fact fluid-dynamic perturbations are caused by the elements of the piping itself,
such as elbows, joints and valves. Theoretically, it could
be possible to reduce the influence of such perturbations
using an adequate segment of straight pipe between the
disturbance and the instrument. In practice, due to the
reduced dimensions of the piping, it is useful to use proper flow conditioners [2].
Generally, flow conditioner efficiency is not based on
the velocity profile produced downstream from the
device itself but on the effects produced on a particular
0955-5986/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 5 - 5 9 8 6 ( 0 2 ) 0 0 0 1 7 - 1
Nomenclature
A
D
k
Ka
Kf
Kfm
Kv
Kvm
m
p
r
R
Re
U
Um
U0
Urif
V
W
x
xs
y
ys
z
eKi
m
r
fixed costs of the plant, related to the type of installation and the piping dimensions (length of the
straight conduits upstream and downstream the flow
conditioner)
variable costs, related to pressure losses, maintenance,
and product uncounted by the meter.
In literature, different flow conditioner classifications
exist that generally refer to either the fluid-dynamic performance [14] or the shape of the conditioner [2]. Table
1 presents the main characteristics of the flow conditioners available on the market.
The first classification, based on the functioning
characteristics, distinguishes between turbulent mixing,
whirling action and hybrid behaviour flow conditioners.
Turbulent mixing conditioners remove swirl effects
and asymmetries of velocity profiles through a turbulent
zone that extends up to two diameters downstream from
the conditioner.
In whirling action conditioners the swirl effects and
the velocity profile distortions are removed by the whirl
action due to fluid flow through the grid of the device,
Table 1
Main flow conditioner classifications
Conditioners
Description
PLa
ULb
DLc
12
13
2
3
9
11
25
23
25
1015
10
22
510
23
Tubes diameter 0.25 D, with intermediate space 1/16 of the transversal section. The tubes 5
with length 10 D are symmetrically displaced
Tubes diameterwith length 24 D. The tubes 9 are designed with an incoming angle (45) to 5
facilitate the flow
The most utilized model comprehends 22 tubes, displaced on concentric rows with smaller
1
diameter toward the external
13
20
22
20
22
2 places, 2 D long, designed in a grid configuration. The combs have a square section with
510
side 0.075 D and 0.45 D depth
Different number of radial vanes divide the conduit in combs. ISO 5167 recommends a length 23
of 2 D
4 little wings and 4 vortex generator elements orthogonally disposed on the internal surface of 1
the conduit and extending till 1/5 D in the transversal section and till 2 D in the axial direction
45
23
1
4
7
3
12
7
3
510
Turbulent mixing
Perforated plates
Mitsubishi
Laws
K-Lab
Spearman
Nova
Sprenkle
Kinghorn
New Zanker
Tube bundle
AGA 3
ISO 5167
Stuart
Whirling action
Screens
AMCA
Etoile
Bluff body
Swirl vor tab
Hybrid action
Multi-body
Zanker
Gallagher
Laws vanes
X4X
a
b
c
1
PL: Pressure losses=p / rU2m
2
UL: Upstream length in D
DL: Downstream length in D
whose length (14 diameters), gives rise to non-negligible pressure losses. Compared to the turbulent mixing,
whirling action flow conditioners require a shorter
straight pipe upstream (0.78 diameters), but do not have
the same effectiveness in producing a fully developed
profile.
The hybrid behaviour conditioners are designed to
Kf
r(U2U2rif)dA
prU2mR2
(1)
r(U2U2rif)rdA
Kfm
prU2mR3
(2)
r|UV|dA
Kv
(3)
prU2mR2
Kvm
r|UV|rdA
(4)
prU2mR3
y2s
2
s
(5)
x dm
xs
y dm
ys
(6)
mass conservation
U 0
(7)
momentum conservation
1
(UU) p 2mDR
r
(8)
The so-called ke model [24] was found to give satisfactory results for the simulation of low level disturbance
and some flow conditioners [9,10].
For all the problems studied, with reference to Fig.
1a, the following boundary conditions are assumed:
walls:
U 0,
V 0, W 0
inlet:
(x y)2 n
(n 1)(2n 1)
1
,
U U0
2n2
D
1
V 0, W 0,
k3/2
3
k [Um0.16(Re)1/8]2, e 0.164
2
0.07D
outlet:
(U,V,W) 0
z
where n is the well-known function of Reynolds number
[22], while and e can be evaluated from the turbulent
intensity and length scale [25].
For the simulation of a viscous layer near the walls,
simplified models are used, the so-called wall functions,
in which a velocity profile is assumed, on the base of
well known empirical relations. The equations for the
turbulent flow are solved outside this layer [25].
The code adopted for the numerical solution of the
problem described above (FLUENT 5.0) is based on the
finite volumes technique [25,26]. Finite volumes allow,
respect to traditional finite difference schemes, modeling
of complex geometry by using unstructured grids. This
is essential for the simulation of geometry, such as that
of the flow conditioner.
The discretized algebraic equations, obtained from the
finite volume procedure, are solved using a semi-implicit
algorithm, SIMPLEC, derived from that originally
devised by Patankar [27]. The second order up-winding
scheme is used for all velocity terms in the momentum
equation, and second order interpolation was also used
for pressure.
The geometry of the flow conditioners is discretized
using unstructured grids. The mesh used for the flow
conditioners is generated using the advancing front type
of procedure; Fig. 1 shows the geometry studied (a) and
the grid used for the inlet section (b). The flow conditioners studied are positioned immediately downstream
from the second elbow, where the origin of the coordinate system is placed (Fig. 1a). About half a million cells
have been used for each of the three computational
domains, but still the mesh sensitivity analysis shows
that a finer mesh would slightly improve the results. An
Fig. 3. Axial (left) and tangential velocity profiles for the single elbow configuration with no conditioner: comparison with the experimental results
[4] 2.6 diameters downstream from the disturbance.
Fig. 4. Axial (left) and tangential velocity profiles for the double elbow configurations (no conditioner): comparison with the experimental results
[4] 2.7 diameters downstream from the disturbance.
Fig. 5. Flatness number Kf (a) and its relative efficiency Kf (b) calculated downstream from the three flow conditioners.
Fig. 6. Flatness number Kfm (a) and its relative efficiency Kfm (b)
calculated downstream from the three flow conditioners.
Fig. 7. Axial vortex number Kv (a) and its relative efficiency eKv (b)
calculated downstream from the three flow conditioners.
Fig. 8. Swirl number Kvm (a) and its relative efficiency eKvm (b) calculated downstream from the three flow conditioners.
6. Conclusions
Numerical methods represent a very useful tool for
the evaluation of flow conditioner efficiency, and can
also be employed in the development of a more general
theory for the evaluation of their efficiency. The work
presented describes a comparison of different types of
flow conditioners under low level perturbation, based on
numerical simulation. Absolute and relative efficiency
10
Fig. 9. Asymmetry number Ka (a) and its relative efficiency eKa (b)
calculated downstream from the three flow conditioners.
11