KMITL Sci. J. Vol.8 No.

1 January – June, 2008

NUMERICAL INVESTIGATION OF TURBULENT FLOW

THROUGH A CIRCULAR ORIFICE
∗
Smith Eiamsa-ard 1, Artit Ridluan1, Prachya Somravysin1, and Pongjet Promvonge2

1
Department of Mechanical Engineering, Faculty of Engineering
Mahanakorn University of Technology, 51 Cheum-Sampan Rd., Nong Chok
Bangkok 10530, Thailand, Email:smith@mut.ac.th
2
Department of Mechanical Engineering, Faculty of Engineering
King Mongkut’s Institute of Technology Ladkrabang, Chaelong Krung Rd.,
Bangkok 10530, Thailand

ABSTRACT

This paper describes the flow through circular orifice by using computational fluid dynamics (CFD) with
various turbulence modeling. Effects of orifice diameter ratios (d/D = 0.5, 0.6, and 0.8) on flow field
characteristics is extensively investigated. To study the influence of turbulence model on the predicted
results, the standard k-ε turbulence model was employed to compare with the Reynolds Stress Model, RSM.
From calculation results, it can be found that the calculated solutions are in good agreement with
experimental data [1] which the RSM leads to perform better that those found that the standard k-ε
turbulence model. Moreover, influence of the orifice diameter ratio on physical flow is also reported.

KEYWORDS : Orifice plate, Turbulent flow, RSM, k-ε turbulence model

1. INTRODUCTION

Uncertainty in orifice metering accuracy is estimated to account for mischarges on order of 250 million
Dollars annually in the natural gas industry. Consequently, attempts are now being made to improve the
accuracy of existing orifice metering facilities. Most prior efforts toward this end have concentrated on
determination of discharge coefficients [1]. The orifice meter is a device commonly used for measuring
fluid flow in industrial processes such as metering flow in the natural gas industry. The popularity of the
orifice meter can be attributed primarily to its simplicity, relatively low cost and little maintenance
requirements in comparison with other fluid meters. The orifice plate becomes the essential part of a fluid
flow meter when installed in a pipe such that the fluid stream must negotiate the constriction. By far the
most common orifice plate installation is that of the concentric round orifice plate. In this type of
arrangement the orifice is round and the plate is mounted between pipe flanges. The plate is positioned
perpendicular to a fully developed pipe flow while at the same time the circular orifice is concentric with
respect to the (circular) pipe interior. Other types of orifice plate exist, such as square orifices, series and
non-concentric. Also, the inner edge of the orifice is machined in one of several different ways. Some
orifice plates are square edged while others are rounded and beveled. This study is concerned with the
concentric, round, beveled orifice plate. Most of the work thus far on orifice meters has focused almost
entirely on the determination of discharge coefficients. There have been a handful of attempts to study in
detail the flow field in the vicinity of the orifice plate. It is believed that knowledge concerning details of
the orifice flow field will lead the way to improvements in metering accuracy [1]. These improvements
could come via improved determination and prediction of discharge coefficients. This research paper deals
with the simulation of turbulent flow through an orifice plate with a view to increasing the knowledge of
orifice meter flow. The mathematical model including the k-ε turbulence model and the Reynolds Stress
Model (RSM), and other computational details is described. Comparisons of the calculated gas axial
velocity with 3D LDV measured data [1] are made to evaluate the turbulence models used. Finally,
influence of orifice diameter ratio is also studies.

∗
Corresponding author. Tel:662-9883666, Fax:662-9883666,
E-mail: smith@mut.ac.th
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

2. PROBLEM CONSIDERATION
A flow in tube with a circular orifice of Nail [1] was employed in the present simulation. A schematic
configuration of the tube orifice is shown in Fig. 1 below. The tube with a length of 9D has a diameter of
25.4 mm and the 3.2-mm thick orifice diameter is 12.7 mm. Profile measurements of centerline axial
velocities, wall-static pressure, Reynolds stresses, and wall shear stresses were measured by using Laser
Doppler anemometer (LDA). The flow had the Reynolds number (Re) of 1.84 x 104 and rate of mass flow
of 1.356 x 10-2 kg/s with temperature at 300K (see Table 1 for specifications).

o
45
m& B /2

D
D/2
B

2D 7D x

Figure 1 Geometry of circular tube fitted with the orifice.

3. MATERIALS AND METHOD

3.1 Governing Equation

The ensemble-averaged, steady, incompressible Navier-Stokes equations in Cartesian tensor notation can be
written in the following form.
Continuity equation:
∂
(ρui ) = 0 (1)
∂xi
Momentum equation:
∂ ( ρui u j ) ∂p ∂
=− + ( t +τ ) (2)
∂x j ∂xi ∂x j ij ij
where ρ is density, ui is mean velocity tensor, p is mean pressure and coordinate tensor respectively. The
mean viscous stress tensor tij is approximated as:
⎛ ∂u ∂u j ⎞
tij = µ ⎜ i + ⎟ (3)
⎜ ∂x j ∂xi ⎟
⎝ ⎠
where µ is laminar viscosity. The time-averaged Reynolds stress tensor, τij = - ρ ui' u 'j , in the above equation
is not known and, thus, models are needed to express it in terms of the solution variables. In the present
study, two turbulence models are used, namely the standard k-ε turbulence model and a Reynolds stress
model (RSM). The k-ε turbulence model has been reviewed in references [2, 3] and it will be described
briefly. The standard k-ε turbulence model relates the turbulent eddy viscosity to the turbulence kinetic
energy (TKE), k and the dissipation rate, ε through Boussinesq’s approximation as:
2 ∂u ∂u j
τ ij = − δ ij (ρk ) + µ t ( i + ) (4)
3 ∂x j ∂xi
where µ t = ρC µ k 2 ε is the turbulent eddy viscosity and ε is the dissipation rate of TKE. The modelled
equation of TKE, k is given by:
⎛ µ e ∂k ⎞
∂
∂x j
(ρu j k = )∂
∂x j
⎜ ⎟ + G − ρε
⎜ σ k ∂x j ⎟
(5)
⎝ ⎠

44
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

in which µe = µt + µ is effective viscosity. Similarly the dissipation rate of TKE is given by the following
equation:
∂ ⎛⎜ µ e ∂ε ⎞⎟ ε
∂
∂x j
(
ρu j ε = ) + (Cε 1G − Cε 2 ρε )
∂x j ⎜⎝ σ ε ∂x j ⎟⎠ k
(6)

where G is the rate of generation of the TKE while ρε is its destruction rate. G is given by:
⎡⎛ ∂u ∂u j ⎞ ∂ui ⎤
G = µ e ⎢⎜ i + ⎟ ⎥ (7)
⎢⎣⎝ ∂x j ∂xi ⎟⎠ ∂x j ⎥⎦
⎜
The boundary values for the turbulence quantities near the wall are specified using the wall
function approach [4]. Cµ = 0.09, Cε1 = 1.44, Cε2 = 1.92, σk = 1.0, and σε = 1.3 are empirical constants [2, 3]
in the turbulence transport equations.

3.2 Reynolds Stress Model (RSM)

For an RSM, the Reynolds-averaged transport equations are solved for the Reynolds stress tensor,
τij, [3], the modeled equations for which are:
∂τ ij ∂ ( u kτ ij )
+ = −Gij − Φ ij + Dij + ε ij (8)
∂t ∂xk
⎛ ∂u j ∂u ⎞ ρε ⎛ ' ' 2 ⎞ ⎛ 2 ⎞
where Gij = ρPij = −⎜⎜ ρ u'i u'k + ρ u'j u'k i ⎟⎟ , Φij = −C1 ⎜ ui u j − kδ ij ⎟ − C2 ⎜ Gij − Gδ ij ⎟ ,
⎝ ∂ x k ∂ x k ⎠ k ⎝ 3 ⎠ ⎝ 3 ⎠

∂ ⎛⎜ µ e ∂ ui u j ⎞⎟
' '
2
Dij = − ( ) , ε = ρεδ ij ,
∂xk ⎜⎜ σ T ∂xk ⎟⎟ ij 3
⎝ ⎠
in which C1 = 2.5, and C2 = 0.55 are model constants.

In equation (8), from left to right, we have the time rate of change of the Reynolds stress at a fixed
point, the net convection of Reynolds stress by the mean flow to the fixed point, local production (Gij) of
Reynolds stress, local pressure strain (Φij), net diffusive transport (Dij) of Reynolds stress to a fixed point,
and local dissipation tensor. More details can be seen in [2, 3] equation (7) provides an expression for each
of the six Reynolds stresses. These six simultaneous equations for stress are to be solved along with the
equations of turbulence kinetic energy dissipation rate.

3.3 Common Form for Equation

All the governing equations can be re-organized and expressed in a standard form that includes the
convection, diffusion, and source terms for 2-D flows as follows:
∂
(ρuφ ) + ∂ (ρvφ ) − ∂ ⎛⎜ Γφx ∂φ ⎞⎟ − ∂ ⎛⎜⎜ Γφy ∂φ ⎞⎟⎟ = Sφ (9)
∂x ∂y ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠
where φ may stand for any variable including the velocity components, Γφx and Γφy are the exchange
coefficients for φ, and Sφ is the source term.

3.4 Solution Procedure

In the present computation, the time-averaged Navier-Stokes equations, the TKE equation, and the
TKE dissipation rate equation are solved numerically by a control-volume finite-difference method [4]
together with the turbulence model equations, equation (4) for the k-ε model or equation (8) for the RSM.
All equations are in a generalized form of equation (9). The SIMPLE algorithm is utilized for pressure-
velocity de-coupling and iteration. The first order upwind (FOU), SOU and QUICK schemes [5] were used
for discretising convection and diffusion transports on a staggered grid. The under-relaxation iterative
TDMA line-by-line sweeping technique is used for solving the resultant finite-difference equations. The
computation was carried out using a personal computer. About 10,000 iterations were needed for
satisfactory convergence criteria when overall mass and momentum residual reach a value of less than
5x10-5 in each case.

45
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

(a) centerline axial velocity

(b) wall-static pressure

Figure 2 Comparison between measurements and predicted profiles with different numerical schemes by
the k-ε turbulence model.

4. RESULTS AND DISCUSSION

The comparisons between the predicted results by RSM and the k-ε turbulence model with different
numerical schemes and the measured data of the flow are presented in Figs. 2 through 3. The predicted gas
pressure and centerline axial velocities are compared with the measurements, where solid or dash curves are
represented for the calculated results while open circles for the measured data. The computational results
are based on a 70x30 non-uniform grid with refinement in the vicinity of the orifice. Grid independence of
the numerical results was verified with a 90x50 finer grid. It is found that the differences for both the base
grid and the finer grid in local flow properties are marginal. This suggests that grid independent solutions
can be obtained with a 70x30 grid, which is used throughout the computations. In order to reduce
uncertainties in the inlet profiles of the mean flow field, the inlet boundary conditions were specified at x/D
= -2.0 for which measured data was available, apart from the radial velocity v which is set to zero.

46
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

(a) centerline axial velocity

(b) wall-static pressure

Figure 3 Comparison between measurements and predicted profiles with different numerical schemes by
RSM.

4.1 Effects of Numerical Schemes

The distributions of centerline axial velocity predicted with the FOU, SOU and QUICK schemes
are compared with the measured data in Figs. 2a and 3a. A closer examination reveals that predictions with
all three schemes are in good agreement with the measurements only in the upstream of the orifice and in
the downstream from x/D = 5. However, for prediction with the three schemes, under-predicted results are
seen in the downstream orifice region in comparison with experimental data. The use of the SOU and
QUICK leads to slight improvement for this flow as can be seen from x/D = 0 to x/D = 5. Figure 2b
compares the profiles of static pressure along the wall using the FOU, SOU and QUICK schemes with the
measurements. It is found that there is a high-pressure drop across the orifice. Predictions with three
numerical schemes show favorable agreement in upstream region of the orifice. However, immediately
after the orifice plate, the calculated wall pressure rises faster than the experiment shows. At downstream
regions from the orifice, the FOU, SOU and QUICK results are over-predicted. Again, the use of the SOU
and the QUICK results in slight improvement for this flow as can be viewed in Fig. 3b. Therefore, only the
SOU is used in the next evaluation of the turbulence models.

4.2 Effect of the Turbulence Models

The distributions of centerline axial velocity and wall-static pressure predicted by both turbulence
models are compared with the measured data in Fig. 4a and 4b, respectively. It is worth noting that
predictions with the RSM are generally in good agreement with the measurements. At downstream regions
from the orifice, the RSM results mimic experimental data very well while under-predicted centerline axial
velocity and over- predicted wall-static pressure profiles are seen for using the k-ε model. The major
difference between calculations with the two turbulence models is most clear in the downstream orifice
region. The use of the RSM leads to substantial improvement for this flow as can be seen in Fig. 4.

47
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

(a) centerline axial velocity

(b) wall-static pressure

Figure 4 Comparison of measurements with profiles predicted by different turbulence models.

Streamlines predicted with the k-ε model and the RSM are shown in Figs. 5 and 6. A recirculation
zone is found at the downstream of the orifice. The size of the recirculation zone calculated by the RSM is
larger than that by the k-ε model. The center of recirculation predicted by the RSM is at about x/D = 1.25
and r/D = 0.35 while at about x/D = 0.6 and r/D = 0.35 is seen for the k-ε model. The reattachment length,
an important measure of the quality of numerical results, is well predicted.

Figure 5 Streamlines predicted by the k-ε model.

Figure 6 Streamlines predicted by the RSM.

48
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

Figure 7 Velocity profiles of flow through the circular orifice.

4.3 Effect of diameter ratio

Figures 7 and 8 display contour plots of velocity profile and path-line for d/D = 0.5, 0.6 and 0.8
respectively. Figure 7 shows the velocity profiles of flow in a duct with orifice plate at 7 stations, namely,
x/D = -0.8, -0.4, 0.03 (orifice plate location), 0.4, 0.8, 1.2, and 1.6 mm, at Re = 18400. at the entrance (x/D
= -0.8, -0.4) the velocities in the vicinity of the wall are very small compared with that in the middle as a
result of shear and friction forces from the wall. The mean velocity profile at x/D = 0.03 shows a high
velocity over the orifice plate due to a narrow channel area. It can be visible that the separation flows were
observed at the positions x/D = 0.4, 0.8, and 1.2 mm because reverse flows occur. It indicates that the mean
velocity profiles at the downstream of orifice plate become to be fully developed in slowly movement in
position of x/D = 1.6 mm. It can be observed that higher reduction of the orifice diameter (d/D = 0.5) results
in a smaller contraction area near the orifice plate and smaller contraction area. This is because of influence
of great velocity flow rate that gives more shear force. In figure 8, it is interesting to note that several
features are observed from these plots. First, the path-lines are considerably distorted in the tube due to
presence of the orifice plate. The velocity distribution is somewhat uniform at both the entrance and exit of
the channel. However, this distribution changes rapidly as the fluid turns the corner of the orifice plate.
Downstream from the orifice plates, it takes much longer to return to rather uniform distribution for d/D =
0.5. The size of the recirculation increases as the orifice diameter ratio decreases from 0.8 to 0.5. The core
flow introduces a shear layer in the fluid driving the re-circulating flow, just as a moving plate over an
enclosed cavity. There is a very steep velocity gradient in the fluid as it turns the corner around the orifice
plate. This is similar to a secondary boundary layer starting from the left corner of the orifice plate which it
affects on the discharge coefficient.

(a) d/D = 0.8

(b) d/D = 0.6

49
 KMITL Sci. J. Vol.8 No.1 January – June, 2008

(c) d/D = 0.5

Figure 8 Streamlines predicted by the RSM.

5. CONCLUSIONS
The numerical simulations with various turbulence models and numerical differencing schemes have been
conducted. The predicted centerline axial velocity and wall-static pressure profiles which the SOU scheme
gives slightly better results than other schemes. In addition, predictions of the flow through orifice plates
using RSM and the k-ε model are generally in good agreement with measurements.

Table 1 Data for flow through an orifice plate [1].

Parameter Magnitude
Test section characteristics
Pipe diameter (D), m 0.0254
Orifice diameter (d), m 0.0127
pipe length (L), m 1.8
Inlet fluid properties (air)
Mean axial velocity, m/s 5.6
Temperature, K 300
Reynolds number 1.84x104

6. REFERENCES
[1] Nail, G.H. 1991 A Study of 3-Dimensional Flow Through Orifice Meters, Ph.D. Dissertation, Texas
A&M University.
[2] Gatski, T.B. 1996 Turbulent Flows: Model Equations and Solution Methodology, in Handbook of
Computational Fluid Mechanics, Edited by Roger Peyret, Academic Press Ltd, London.
[3] Wilcox, C.D. 1993 Turbulent Modelling for CFD, DCW Industries, Inc., California.
[4] Versteeg, H.K. and Malalasekera, W. 1995 An Introduction to Computational Fluid Dynamics: The
Finite Volume Method, Longman Scientific & Technical, Longman Group Limited, Essex, England.
[5] Patankar, S.V. 1980 Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington, D.C.

50