Jtam.dvi


JOURNAL OF THEORETICAL

AND APPLIED MECHANICS

49, 2, pp. 301-311, Warsaw 2011

MEASUREMENT OF FLOW RATE IN SQUARE-SECTIONED

DUCT BEND

Kazimierz Rup
Cracow University of Technology, Faculty of Mechanical Engineering, Kraków, Poland

e-mail: krup@riad.usk.pk.edu.pl

Łukasz Malinowski
ABB Corporate Research, Kraków, Poland

e-mail: lukasz.malinowski@pl.abb.com

Piotr Sarna
Cracow University of Technology, Faculty of Mechanical Engineering, Kraków, Poland

e-mail: piotr@inloop.pl

In this paper the authors describe an attempt to utilise installed square-
sectioned elbows in order to measure the fluid flow rate. In order to
practically accomplish the measurement of the volumetric flow rate of
the air, a special researchstandhasbeenbuilt, and square shapedelbows
have been installed (80×80mm in dimension). The numerical compu-
tations were carried out using the software package FLUENT 6.2. The
obtained resultswere compared to correspondingones coming fromorifi-
cemeasurements and from experimental work available in the literature.
The comparative analysis of the obtained numerical and experimental
results evidenced a high degree of their conformity.

Key words: measurement of flow rate, inverse coefficient problem

1. Introduction

Standardized methods of fluid flow rate measurement are based on the inte-
gration of the velocity profile. They are basically the only methods that can
be applied to a rectangular cross section according to Bean (1971) and Spit-
zer (2001). The above-mentioned methods are characterised bymulti-discrete
measurement of local velocities at points that have been appropriately distri-
buted on a lateral stream cross section. In order to obtain a satisfactory result
of measurement, accuracy is indispensable in providing a large number of me-
asuring points. Consequently, such a method is time-consuming and may be
quite tedious. Constant registration of the flow intensity is also impeded.



302 K. Rup et al.

In the case when a classical measuring orifice is used to measure the flow
rate, it is necessary to install longpipeline sectionswith a circular cross section
as described by Bean (1971), Spitzer (2001) and in compliance with PN-EN
ISO(2004). Itappears that circular-shapedelbowflowmeters,whichhavebeen
installed on the pipeline, enable one to obtain a signal that corresponds to the
entire volumetric rate of the fluid flow as suggested by Bean (1971), Spitzer
(2001) and Rup and Malinowski (2005, 2006). The principle that pertains
to operation of the elbow flow meter is based on the usage of the ratio of
volumetric rateofflowto thedifferencebetweenpressuresmeasuredat extreme
points of the elbow arc secant. In practice, engineers sometimes use a solution
where thepressure ismeasured throughpulse apertures that are situated along
the plane, which is not aligned with the secant. In this case, the pressure
difference that is measured attains lower values according to Bean (1971),
Spitzer (2001) and Rup andMalinowski (2005, 2006).

The standardmeasurement is carried out in the followingway: thepressure
difference is measured and, subsequently, the relevant volumetric rate of flow
value is assigned based on the previously worked out characteristic of the
instrument. At times, the instrument characteristics are substituted by simple
algebra dependencies, which have been devised on the basis of experimental
results.

Recently, a new measurement method for usage of elbow flow meters in
circular-sectioned pipes has been elaborated, see Rup and Malinowski (2005,
2006). The idea of this particularmethod is based on selecting such a numeric
solution of fluid equations of motion (including turbulence model equations)
which fulfills the given accuracy, the balance condition that has beenmeasured
and calculated at extreme points of elbow flow meter secant. This brand new
method eliminates the necessity to calibrate elbow flowmeters.
Themain goal of this paper is an extension of themethod of identification

or control of the volumetric flow rate of the fluid inside circular sectioned
pipes, elaborated earlier by Rup and Malinowski (2005, 2006), for channels
where quadratic arc elbows have been installed.

2. Test stand

To determine the value of themeasured pressure difference (∆pexp) necessary
to carry out the described method, an appropriate test stand was built. A
sample section of the stand is shown in Fig.1. Themeasuring set displayed in
Fig.1 consists of a rectangular pipelinewith the following dimensions: a×a=



Measurement of flow rate in square-sectioned duct bend 303

=80mm×80mm, installed elbow arc (90◦) (1), with the average radius equal
Rs = 160mm. The length of the straight line section of the pipeline with
a quadratic cross section, that was placed in front of the elbow arc equals
L3 = 20Dh = 1.6m, whilst the length of the appropriate section beyond the
arc equals L4 = 25Dh = 2.0m. For control purposes, a measuring orifice
was also installed at station (2). The inner diameter of the measuring orifice
is equal to dorif = 78mm. The measurement of pressure decrease on the
orifice was carried out using a water manometer (3). In order to obtain a
correct measurement of the volumetric flow rate via the orifice, a straight line
section of the pipeline (L1,L2) was installed in front and behind the orifice as
suggested by Spitzer (2001) and according to PN-EN ISO (2004).

Fig. 1. Test stand; 1 – elbow arc with quadratic cross section, 2 – measuring orifice,
3 – water manometer, 4 – transition from circular to square-section pipe,

5 – pressure converter and digital meter, 6 – Askania differential micro-manometer,
7 – temperature sensor, 8 – sheet metal cone, 9 – centrifugal fan, 10 – thyristor

frequency converter, 11 – frequency regulator

Between the circular and square-section pipe, transition segment (4) was
installed. Two pulse apertures have been drilled at the elbow arc secant in
order to measure the pressure difference. The mentioned pressure difference
wasmeasuredwith the use of the pressure converter and digitalmeter (5). For
control purposes, Askaniamicro-manometer (6) has been attached diagonally
to the ends of pressure receivers. The temperature of flowing air wasmeasured
using resistance-type temperature sensor (7). Based on the temperature of the
flowing air that has been measured, it was possible to determine thermo-
physical properties of the fluid. As shown in Fig.1, the pipeline was placed on



304 K. Rup et al.

the suction side of the fan so that flow disturbances originating from the fan
could be eliminated. The tail end of the pipelinewas linkedwith suction flange
ofMPB500T (9) fan by applyingmetal sheet cone (8). The desired volumetric
streamflowwas regulated by thyristor frequency converter (10)which enabled
smooth control of the fan RPM.

3. Determination of the fluid flow rate according to the idea of

the applied method

Considering the turbulent nature of the examined flows, the time averaged
Navier Stokes equations (RANS) with continuity equation were used. Assu-
ming that the flows were incompressible, in statistically stationary flow the
time averaged equations resulting from themomentum andmass conservation
equations have the following form

∂vi
∂xi
=0

(3.1)

ρ
∂

∂xj
(vivj)=

∂

∂xj

[

−pδij +µ
(∂vi
∂xj
+
∂vj
∂xi

)

−ρv′iv
′

j

]

The last term on the right side of equation (3.1) expresses the components of
the Reynolds stress tensor

r
(t)
ij =−ρv

′

iv
′

j (3.2)

The appearance of the additional term in equation of turbulent flow (3.2) cau-
ses that the system of equations (3.1) becomes an opened one. For a unique
description of fluid motion, additional relations determining the components
of turbulent stress tensor (3.2)must be taken into consideration. Todetermine
these components the 7-equation turbulence model (RSM) available in FLU-
ENT was used as described in Fluent User Manual (2008). The RSM model
is recommended especially in the case of the fluid flow in curved channels. To
obtain a unique description of the investigated flow, the following boundary
conditions were formulated: given value of static pressure on the inlet of the
duct, zeroing of the velocity vector and turbulent kinetic energy on the walls,
assumed value of mean flow velocity, turbulence intensity Tu0 on the inlet
of the duct and length scale l. The turbulence intensity and the length scale
were defined by the following relations

Tu0 =0.16Re
−1/8 l=0.07Dh (3.3)



Measurement of flow rate in square-sectioned duct bend 305

Thementioned uniqueness of the description of the examined flow requires
that a precise geometrical model is created. This model contains an exact 3D
geometry of the bendalongwith corresponding straight ducts on the inlet and
outlet as presented in Fig.2. Themodel wasmeshed inGAMBIT, a FLUENT
pre-processor.

Fig. 2. Geometry dimensions andmesh in the boundary layer of the volume

To achieve high accuracy of the results, a finemeshwas created, especially
in areas where large velocity gradients exist, that is near the duct walls. The
ratio of the smallest element adjacent to the wall to the hydraulic diameter
δy1/Dh = δz1/Dh = 0.00075 and the growth factor (ratio of the preceding
element to the next element) is 1.3 for 16 elements. The size of the last element
is δy16/Dh = δz16/Dh)= 3.992/80=0.04991.

The longitudinal dimensions of the finite volumes increased from
δx′1/Dh = 4/80 = 0.05 at the inlet of the duct (x

′/Dh = −1600/80) to
δx′28/Dh = 8/80 = 0.10 at two hydraulic diameters downstream from the in-
let (x′/Dh =−1440/80) and then remained constant (δx/Dh =0.10) to two
hydraulic diameters from the bend (x′/Dh =−160/80). From this point, the
length of the volumes decreased to a value of δx′216/Dh = 4/80 = 0.05. The
length of the volumes within the bend was ∆ϕ′ = 4/80 = 0.05 on the outer
wall of the bend and decreased towards the inner wall. The finite volumes in
the outlet section of the duct increased from δx/Dh =0.05 to δx/Dh =0.10
at four hydraulic diameters downstream from the bend. The lengths of the
remaining volumes were constant (δx/Dh = 0.10). The total number of seg-
mented control volumes in the model shown in Fig.2 is 1034775, while the
number of grid nodes is 1083392. The obtained grid utilises only prismatic



306 K. Rup et al.

elements based on squares. Figure 2 presents a sample mesh at the inlet sec-
tion of the duct.

It is worth mentioning that the generated mesh is characterised by a very
small skewness coefficient of the finite volumes. The maximum value of the
skewness coefficient in the presented examples does not exceed 0.5.

The simulation was carried out in FLUENT, a commercially available so-
lver.The obtained resultswere comparedwith the corresponding results found
in the reference elaborated by Sudo et al. (2001). In this case, the geometrical
model was such that the parameters describing the fluid flow and its physical
properties were identical as the ones described by Sudo et al. (2001).

In the reference, mentioned above, the development of steady turbulent
flow of air was examined in a curved 80mm×80mm square duct. The mean
bend radius was Rs =R/Dh =2.0. Tomeasure the longitudinal components
of velocity a hot-wire anemometer was used.

Figure 3 represents the longitudinal component of average velocity profi-
le of the air flow in the section of the curved square duct described by the
following coordinates: x′/Dh =−1.0 and crosswise z/Dh =0. The Reynolds
number: Re = 40000. The points connected with lines represent numerical
simulation values for RSMmodel. The discrete points represent experimental
values obtained by Sudo et al. (2001).

Fig. 3. Comparison of velocity profiles obtained with RSMmodel with experimental
results for x′/Dh =−1.0, z

′/Dh =0

Figures 4 and 5 represent similar profiles measured for ϕ = 60◦ and
z/Dh =0 and for x/Dh =10.0 and z/Dh =0, respectively. It can be clearly
seen in Figures 3 to 5 that numerical results and experimental results show
a deformation of the velocity profile caused by the curvature of the duct as
presented by Sudo et al. (2001).



Measurement of flow rate in square-sectioned duct bend 307

Fig. 4. Comparison of velocity profiles obtained with RSMmodel with experimental
results for ϕ=60◦ (z/Dh =0)

Fig. 5. Comparison of velocity profiles obtained with RSMmodel with experimental
results for x/Dh =10.0, z/Dh =0

This paper concerns amethod designed tomeasure the flow rate of fluid in
square-sectionedbends installed inducts.Themethodof indirectmeasurement
of the flow rate in circular-sectioned pipes, applied in the study, was earlier
developed by Rup andMalinowski (2005).

The essence of this method is based on the measurement of pressure dif-
ference in stationary flow conditions on two opposite walls of the bend angle
secant. In the next phase of the measurement, a comparison of the value of
measured difference (∆pexp = pi − p0) with the corresponding value obta-
ined through numerical solution (∆pnum) is carried out. The numerical value
(∆pnum) is obtained through solution of mass and momentum equations and
equations describing the selected turbulencemodel; the examined value of vo-



308 K. Rup et al.

lumetric (mass) flow is adjusted so as to satisfy the assumed accuracy of the
condition of equality of both values being compared

F(Re)= |∆pexp−∆pnum| ¬ ε (3.4)

where ε is a given value which equals in the example to the inaccuracy of
pressuremeasurement.
In numerical computations, an exact geometry of the flow space is taken

into consideration along with the measured temperature and pressure of flu-
id on the base of which its thermo-physical properties are described. In the
first step of computations, two extreme values of the Reynolds number (Re1
and Re2) are assumed corresponding to theminimumandmaximumvalues of
volumetric flow rate being measured. In the next phase of computations, the
unknown value of the the volumetric flow rate is numerically determined by
minimizing the expression of the difference between measured value (∆pexp)
and the corresponding calculated value (∆pnum) (3.4). The process of mini-
mizing of thementioned pressure difference (measured and calculated values)
is carried out by the method of secants. In other words, the principal of this
method is to select such a value of theReynolds number – used as a coefficient
in the partial differential equations of fluid motion – for which the computed
pressure difference (∆pnum) satisfies condition (3.4). The field of velocities
and pressures in the fluid is being determined numerically for the assumedRe
number values. The numerically found values of the pressure drop ∆pnum dif-
fer in both extreme cases from the pressure dropmeasured in the same points
∆pexp = p0−pi (Figs.1, 2). It is evident that the assumed extremely different
values of the Reynolds number Re1 and Re2 do not meet condition (3.4) in
both cases, thus they cannot constitute its radicals (solutions). In order to
find the next approximation of condition (3.4), the mentioned secant method
is being applied. Following the idea of themethod of secants described byTa-
ler and Duda (2006), the next n-th approximation of the radical of condition
(3.4) is being determined as follows

Ren =Ren−1−
F(Ren−1)(Ren−1−Ren−2)

F(Ren−1)−F(Ren−2)
for n> 2 (3.5)

The process endswhen condition (3.4) is satisfied. The volumetric rate of flow
of the elbow flowmeter in question is determined from the formula

Q=
πDhν

4
Re (3.6)

whereRe is the last calculated valueof theReynoldsnumber, ν – thekinematic
viscosity.



Measurement of flow rate in square-sectioned duct bend 309

Because of the lack of optimizing procedures in the FLUENT package, it
was necessary to elaborate a suitable procedure coupling that package with
thementioned numerical procedurewhich implements the idea of themethod
of secants. That procedure was written in C++ language. The consecutive
iterations used journal files of the package FLUENTaccording to Fluent User
Manual (2008).

Table 1.Tabulation of measured and calculated values

∆porf Qorf ∆pexp ∆pnum Renum Qelb
Qorf−Qelb
Qorf

[Pa] [m3/s] [Pa] [Pa] [–] [m3/s] [–]

58.7 0.03584 20.7 20.6 29484.4 0.03446 0.038

166.4 0.05965 68.0 67.9 53526.1 0.06255 −0.049

362.2 0.08735 147.0 146.9 78736.1 0.09201 −0.053

440.6 0.09615 178.5 178.6 86814.2 0.10145 −0.055

587.4 0.11070 236.6 236.7 99941.1 0.11679 −0.055

704.9 0.12103 275.5 275.4 107805.3 0.12598 −0.041

802.8 0.12897 322.5 322.5 116662.2 0.13633 −0.057

969.2 0.14138 394.5 394.4 129010.4 0.15076 −0.066

1096.5 0.15013 444.3 444.2 136925.9 0.16001 −0.066

1292.3 0.16261 532.3 532.4 149890.3 0.17516 −0.077

At the end of calculations for the last value of Re, the procedure compares
the values of pressure differences, and then starts next calculations assuming
the newly found value of the Reynolds number or stops iteration when condi-
tion (3.4) is satisfied. The maximum number of iteration steps carried out in
those investigations did not exceed 10. The values of the flow rates found by
using the orifice Qorf are given in Table 1. The measurement results and the
Reynolds number Re determined according to the idea of the appliedmethod
are given inTable 1 aswell. The relative deviations of experimentally determi-
ned values of Qorf and the corresponding ones determined by implementing
the applied method (Qelb) shown in Table 1 are small.

The maximum relative deviation of the flow rate in the discussed elbow
flow meter is ∆=7.7%, and concerns the greatest rate of flow in that elbow
flow meter. It is to be stressed that the mentioned relative deviation can be
considerably minimized by using ducts with a lesser size or elbows with a
smaller curvature radius in the construction of elbow flow meters for smaller
flow rates. On the other hand, in the case of flows with a Reynolds number
Re > 200000, it is advisable to apply flow meter constructions with greater
size of ducts in order to eliminate, among other things, intense vibrations



310 K. Rup et al.

of the duct which affect the accuracy of the measurement of the pressure
difference ∆pexp.

4. Conclusions

The indirect method used to measure the volumetric flow rate of a fluid is
characterised byhighaccuracy and repeatability. Thehighaccuracy is possible
due to a very realistic mathematical model of the complex flow in the curved
duct. The indirect method eliminates the necessity of frequent calibration of
the flow meter. The discussed elbow flow meter, implementing the extended
intermediate measuringmethod, can be applied to determine the flow rate of
gases as well as liquids and their suspensions. In the performed calculations,
the fluid was treated as incompressible due to the fact that the maximum
value of flow average velocity in experimental measurements was lower than
40m/s.
The experimental investigations carried out indicate that the measured

pressure differences ∆pexp at the end points of the secant of the elbow are
nearly 3 times smaller than the corresponding pressure differences measured
on the orifice at the same flow rate.

Acknowledgements

The researchwas financed from scientific grants in the years 2007-2010 conducted

as a research project.

References

1. BeanH.S., 1971,FluidMeters; Their Theory andApplication, 6edEd.,ASME

2. Fluent 6.2, 2008,User’s Guide, Fluent Inc.

3. Malinowski Ł., Rup K., 2008, Measurement of the fluid flow rate with use
of an elbow with oval cross section, Flow Measurement and Instrumentation,
19, 358-363

4. PN-EN ISO 5167-1, 2004,Measurement of fluid flow by means of pressure dif-
ferentia devices inserted in circular cross-section conduits running full – part 1:
General principles and requirements, and part 2: Orifice plates

5. Rup K., Malinowski Ł., 2005, Metoda pomiaru strumienia objętości płynu
w zastosowaniach do przepływomierzy kolanowych,PAK, 5, 35-37



Measurement of flow rate in square-sectioned duct bend 311

6. RupK.,MalinowskiŁ., 2006,Fluidflow identificationonbaseof thepressure
difference measured on the secant of a pipe elbow,Forschung im Ingenieurwe-
sen, 70, 199-206

7. Spitzer D.W., 2001,Flow Measurement – Practical Guides for Measurement
and Control, 2nd Ed., ISA

8. Sudo K., Sumida M., Hibara H., 2001, Experimental investigation on tur-
bulent flow in a square-sectioned 90-degree bend, Experiments in Fluids, 30,
246-252

9. Taler J., Duda P., 2006, Solving Direct and Inverse Heat Conduction Pro-
blems, Springer, Berlin

Pomiar strumienia płynu w kanale z łukiem kolana o przekroju

kwadratowym

Streszczenie

W pracy podjęto próbę wykorzystania zainstalowanych w kanałach przepływo-
wych łuków kolan o przekroju kwadratowymdo pomiaru strumienia objętości płynu.
W tym celu wykorzystano opracowaną wcześniej metodę pomiaru pośredniego dla
rurociągów o przekroju kołowym. Dla praktycznej realizacji pomiaru wspomnianego
strumienia przepływu zbudowano stanowisko badawcze za łukiem kolana o przekroju
poprzecznym w kształcie kwadratu (80× 80mm). Obliczenia wykonano za pomocą
pakietu FLUENT 6.2. Uzyskane rezultaty porównano z odpowiednimi zmierzonymi
za pomocą kryzy pomiarowej oraz z innymi wynikami doświadczalnymi dostępnymi
w literaturze. Z analizy porównawczej wynika wysoki stopień zgodności otrzymanych
rezultatów pomiaru pośredniego.

Manuscript received July 5, 2010; accepted for print November 22, 2010