Vous êtes sur la page 1sur 17

Problem

Taking the density as =1000 kg/m3 and viscosity as =2 10-3 kg/ms the
Reynolds number calculated as
V avg d
= =100

Where V avg is the average inlet velocity and the value of pipe diameter and inlet
velocity given as 0.2 m and 1 m/s respectively.
Ansys CFX has been used here to solve this laminar pipe flow problem and Tecplot
360 has been used to plot and compare results for different meshes in terms of
centerline velocity, exit velocity and skin friction coefficient.
Introduction: Governing equations

D
+ p u=0
Dt

Du
= 2 u P+ g
Dt

Those two above equations are Navier Stokes equation for continuity and
momentum respectively.
For friction factor in a laminar flow we can solve Navier stokes equation analytically
and obtain the following expression
F=16/Re
The Reynolds number given for this problem is 100. So by putting this value on the
above equation we get the skin friction coefficient as
F=16/100=0.16 in the fully developed region
Solving Navier stokes equation for momentum in Z direction we get the velocity
profile as

r2
=2 V avg (1 2
)
R

Where R is the pipe radius and V avg is the average velocity. In this problem the
average velocity given as 1 m/s
2
r
=2(1 2
)
R
Fig1: Analytical exit velocity profile
Figure2: Velocity contour at Z = 1 and Z = 7
The figure above shows that the velocity is 0 m/s at the wall for different location
along Z-plane. So we can say that the boundary condition is met in Ansys CFD
simulation.
If the value of r is taken as 0 (zero) we get the maximum velocity
max=2 V avg

Here in this problem the average velocity is given as 1 m/s


max=2 1=2 m/ s

Boundary Conditions:
The boundary condition applied is as follows

inlet Velocity inlet(1 m/s)


outlet Pressure outlet (1 atm)
Wall No slip condition

Number of Nodes and ELements for all meshes:

Mesh Name Node Numbers Element Numbers


Arbitrary 171897 214878
Intermediate 313474 390274
Fine 528945 698361

Arbitrary Mesh:
For an arbitrary mesh the sweep number was taken as 64 distributed uniformly
along the length of the pipe. we took the body spacing as 0.005 m, sphere radius as
0.075 m and sphere element size as 0.005 m. The total number of nodes was
171897 and number of elements was 214878 and is shown in the figure below.

Fig3: Arbitrary mesh normal view (left) and closed look view(right)
Refined meshes with decreasing spacing and will give the accurate results on the
desired location.
A finer mesh is needed to get the accurate results since from the figure it can be
seen that for arbitrary mesh the skin friction deviates from the analytical result in a
huge margin.

Intermediate Mesh:
For the intermediate mesh, there are 72 layers distributed uniformly along the
length of the pipe. The sphere radius and sphere element size taken as 0.075 m and
0.0035 m respectively. Plot of the mesh cross section is shown below.
Fig4: Intermediate mesh
Here,the default body spacing taken as 0.0035 m for a total number of 390274
elements and 313474 nodes.
Fine Mesh:
For the fine mesh, there are 72 layers distribute uniformly along the length of the
pipe. The sphere radius and sphere element size taken were 0.08 m and 0.0025 m
respectively. The number of nodes taken here as 528945 and the total number
elements as 698361. Plot of the mesh cross section is shown below.
Fig5: Refined mesh
The fine mesh takes much less time to give the solutions and the results show us a
major advantage over arbitrary mesh.
Figure below shows us the comparison among the three meshes in terms of
centerline analytical velocity.
Fig6: Centerline velocity comparison for three meshes
From the above figure we can see that the intermediate and fine mesh are clearly
on top of each other while the arbitrary mesh has a slightly different sloping. The
error obtained for fine mesh is much more less compared to the arbitrary mesh and
the result gives a way to assessing results for this laminar pipe flow problem.
Figure below shows us the comparison among these three meshes in terms of skin
friction.

Fig7: Skin friction comparison for three meshes


The skin friction curves for both fine and intermediate are almost impossible to
separate but for arbitrary mesh the curve follows a different sloping. Though the
percentage of error for fine mesh is much less compared to the other two meshes
still the percent is huge. That means a much more fine mesh is required to get the
precise results.
The comparison among arbitrary, intermediate and fine mesh in terms of exit
velocity are plotted below.
Fig8: Exit velocity comparison for three meshes
The plotted figure above shows us that, the exit velocity for these three meshes and
analytical solution almost identical though a much closed look will give us an idea
that the arbitrary mesh is not as accurate as the intermediate and fine mesh.
Effect of convergence:
For skin friction, exit velocity and centerline velocity the effect of convergence is
discussed below and we took the convergence as 1E -6,1E-4,1E-3,1E-2.
!E-5 is identical to 1E-6. So while we have not show 1E-5 as we have shown results
for convergence criteria for 1E-6.
The centerline velocity results of fine mesh are plotted blow for different
convergence.

Fig9: Centerline velocity convergence comparison


For the lowest convergence (1E-6) the centerline velocity is almost closed to the
analytical solution .For 1E-3 and 1E-2 the center velocity results move far away from
the analytical solution though for 1E-4 the centerline velocity is identical and closed
enough to the value obtained for 1E -6.

Convergence 1E-6 1E-4 1E-3 1E-2


criteria
% of error 0.5 1.1 18 25

From the above table we can clearly say that the convergence has great influence
on centerline velocity except for 1E-6 and 1E-4
The skin friction results of fine mesh are plotted blow for different convergence.
Fig10: Skin friction convergence comparison
For the wider convergence like 1E-2 and 1E-3 , the skin friction did not decrease much
enough to comply the analytical solution obtained which is 0.16. However, when
the convergence reduces to 1E-4, the results shown were almost identical to 1E -4,1E-6
and analytical solution.

Convergence 1E-6 1E-4 1E-3 1E-2


criteria
% of error 6.25 6.25 23.75 31.25
From the above table it can be said that the percentage of error for convergence
criterion 1E-2 and 1E-3 are very high. For 1E-4, the error percent still remains high
enough which is quite a few times than for 1E -6. And the percentage of error for 1E -6
is 6.25. So, since the effect of convergence shows us a greater percentage of error,
it can be said that a finer mesh is needed since the higher convergence value
doesnt increase the accuracy much more.
The exit velocity results of fine mesh are plotted blow for different convergence.

Fig11: Exit velocity convergence comparison


The lowest convergence criteria results in almost identical to the analytical solution
and for 1E-4 , the solution is closed to the value obtained for 1E -6. The same thing
seemed to be happened for both friction co-efficient and centerline velocity as well.

Convergence 1E-6 1E-4 1E-3 1E-2


criteria
% of error 0.1 0.19 20 28.75

So, from these comparisons, we can clearly say that the effect of convergence has a
profound influence on the results for the laminar pipe flow problem.
Result and Discussion for fine Mesh
Figure shows the comparison of analytical solution with the fine mesh.
Fig12: Analytical and fine mesh exit velocity comparison for fine mesh
In case of exit velocity the fine mesh result is identical to the analytical. This is due
to less discretization error due to higher number of nodes and due 1E-6
convergence criteria linearization error is also very less.
Fig13: Analytical and fine mesh centerline velocity comparison for fine mesh

We have considered fully developed flow at the inlet while calculating the centerline
velocity. But in case of CFX simulation the flow is fully developed after the intel at
Z= 2. As the flow become fully developed the analytical solution and the fine mesh
solution become identical due to near zero discretization error and linearization
error.
Fig14: Analytical and fine mesh skin friction comparison for fine mesh
The calculated skin friction is constant at Reynold number 100 that is 0.16. The CFX
simulation shows when the flow is fully developed in case of fine mesh the friction
matches with the analytical solution. This is because we have assumed fully
developed flow while solving Navier-Stokes equation analytically.

Conclusion:
The computation time to refine mesh is much longer but it took much less time to
converge them into solutions. This is mainly because of that the flow we have taken
here as laminar and effect of turbulence doesnt need to be counted so as the
temperature change. The inflation layers, high node numbers, line control
contributes to increase the improved results immensely. Though the fine mesh takes
more computation time and cost, the results it provide pave the way to assess
result as accurate as possible to the analytical solution.

Vous aimerez peut-être aussi