Engineering, 2013, 5, 268-276
http://dx.doi.org/10.4236/eng.2013.53037 Published Online March 2013 (http://www.scirp.org/journal/eng)
Particles Removal from a Moving Tube by Blowing
Systems: A CFD Analysis
Giovanni Lombardi, Rubens Curatola
Department of Civil and Industrial Engineering, Pisa, Italy
Email: lombardi@ing.unipi.it
Received October 1, 2012; revised January 28, 2013; accepted February 5, 2013
ABSTRACT
The use of an air jet to clean the products during the manufacturing is usual in an industrial environment. In this paper
the problem of the removal, from the inner of the tube, of the waste arising from the cutting operation by means of an
air blowing system is analysed. In order to obtain indications about the real importance of the involved parameters and
their effects, a numerical procedure, based on CFD analysis trough the code STAR CCM+, has been settled. The effects
of the different parameters were highlighted. The results are congruent with the experience, indicating that the adopted
solution is not suitable for the small tubes. Therefore, a different blowing solution is investigated, with the object to
improve the cleaning capabilities of the system. The results for the new blowing solution provide a significant im-
provement of the capability to remove the waste arising from the cutting operation can be expected.
Keywords: Fluid Dynamics; Industrial Application; Internal Flow; CFD; Moving Mesh
1. Introduction
The use of an air jet to clean the products during the
manufacturing is usual in an industrial environment, [1].
In this paper, in particular, the production of glass tubes
is considered.
The specific problem is the removal, from the inner of
the tube, of the waste arising from the cutting operation.
Typically, the tube is transported by a conveyor belt
through the various tools. In the considered problem,
along the path it meets a sector with some nozzles, which
generate air jets to remove the particles remained in the
tube after the cutting operation.
The considered problem is affected by several pa-
rameters, and it is very difficult a quantitative evaluation
of their effects. Therefore, in order to obtain indications
about the real importance of the involved parameters and
their effects, a numerical procedure, based on CFD
analysis, has been settled. For the numerical solution the
code STAR CCM+ is used. With the support of the nu-
merical analysis it is possible to highlight the effects of
the parameters, and, therefore, to suggest modification to
improve the effectiveness of the system.
Finally, a different blowing solution is proposed and
analysed.
2. Overview of the Problem
The basic scheme consists in a nozzles row, placed be-
low the cutting machine, on one side of the conveyor
belt.
The fundamental parameters, see Figure 1, are:
The internal diameter of the glass tube, .
The length of the tube, L.
The translation velocity of the tube.
The gap between the nozzle exhausts and the tube
internal side, w.
The number, shape and dimensions of the jet nozzle.
The total span of the jets, s.
The supply pressure of the jets, Δp (on respect to the
atmospheric pressure).
Figure 1. Sketch of the geometrical parameters.
C
opyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA 269
All the solutions presented in this paper are for a
length tube L = 1570 mm and a distance between the
nozzle exhausts and the tube inner side, w, of 40 mm.
Two different diameters of the tube are considered: the
large tube, with internal diameter = 14 mm (external
diameter 16 mm) and the small tube, with internal di-
ameter = 6.35 mm (external diameter 8.15 mm). The
translation velocity of the tube is fixed by non-aerody-
namics considerations; for the large tube it is fixed at 60
mm/s, while for the small tube it is fixed at 135 mm/s.
Two working pressures Δp, are analysed: 101,325 and
303,975 Pa.
A first configuration with a row of five circular jets,
with a diameter of 4 mm and s = 252 mm, is considered.
Its blowing surface is 62.8 mm2.
3. The CFD Procedure
3.1. Conceptual Approach
The problem under consideration is obviously a non-
steady one, and it is necessary the simulation of the rela-
tive movement between two parts: this implies the up-
date of the calculation grid at each time step, [2]. Fur-
thermore, it is necessary to set the time step sufficiently
low, in order to have a good representation of the flow
time history; a sensitivity analysis to this computational
parameter was carried out in a preliminary phase.
The computational domain is divided in two separate
domains, one containing the blowing zone, the other
containing the tube. The motion is assigned only to the
region containing the tube. With this approach, the inter-
face between the two regions is the only part that needs
to be update at each time-step.
Finally, because a particularly accurate evaluation of
the tangential stresses into the tube is required, a very
refined representation with prism layers is created at its
internal wall. The sensitivity analysis to grid refinement
will be discussed in the following.
The solution gives a complete description of the flow
field, at any time step. The analysis of the flow field is
clearly interest, and a complete understanding of the pro-
blem is available from it. Nevertheless, the amount of
data is so large that it is difficult a simple interpretation.
Therefore, some “integral” results are derived, in order to
have a more clear and immediate idea of the performance
of the system.
For this purpose, the tube is divided into 50 sectors, in
order to evaluate the distribution along the tube of the
quantities of interest. In particular are evaluated the mass
flow rate [Kg/s], the integral of the shear stress over each
sector, indicated with τint [N], and its value for unit length,
referred as θ [N m].
Furthermore, to have an indication of the capabilities
of the system to clean the tube, the following integral
quantities are considered:
The mass flow through the end section of the tube,
defined as the integral value over the simulation time.
It is assumed as index of the available “energy” to
remove the particles.
The integral shear stress over the inner surface of the
tube, defined as the integral value over the total inner
surface and the simulation time. It is assumed as in-
dex of the “traction” force to remove the particles.
The maximum value of τint, over time and space, as-
sumed as index of the capability to “start” the move-
ment of a particle.
3.2. The Computational Grid
The surface mesh was generated with ANSA v. 13.01.
The volume mesh, generated with Star CCM+, is com-
posed by polyedrical elements, except the zone close to
the wall, represented by prism layers.
The dimensions of the computational domain are re-
ported in Table 1.
The volume grid is characterised by two local refine-
ment zones: one in the ending zone of the tube, the other
in the interface zone between the two domains.
Grids up to about 8 millions of cells were analysed in a
preliminary phase. The analysis of the results shows that
they are essentially stable when the number of grid cells
is higher than 3 millions. Therefore, a grid refinement
with about 4 millions of cells was settled; the grid pa-
rameters chosen for the simulations are shown in Table 2.
Some details of the grid are shown in Figure 2.
3.3. Set-Up of the Numerical Solution
The numerical simulation can be divided into two
phases:
Table 1. Domain dimensions.
Length 2250 mm 9 jets row spans
Width 8230 mm 5.25 tube lengths
Height 1000 mm 62 “large” tube ext. diameters
Table 2. Volume mesh parameters.
Max. Skew 0.57
Surface cells 400.000
Volume poly cells 4 × 106
Poly dens. 1.1
Poly grow factor 0.9
Prism layer thickness 5 mm
Nr of prism layers 7
Prism layer grow rate 1.1
Copyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA
Copyright © 2013 SciRes. ENG
270
(a) (b)
(c) (d)
Figure 2. Details of the volume grid. (a) Inner zone of the tube; (b) Outer zone of the tube; (c) Cross section of the tube; (d)
Nozzle inlet.
Table 3. Computational set-up.
1) A steady phase to develop the jet flow. The move-
ment of the tube is turned off. Steady phase Unsteady phase
Time Steady Implicit unsteady
Flow Coupled flow Coupled flow
Turbulence model k-ε k-ε
Wall treatment Two layer all y+ wall Two layer all y+ wall
2) The movement of the tube is turned on and the flow
regime became unsteady.
The set-up of the calculation is summarised in Table
3.
The energy equation is activated to take into account
the different temperature between ambient and walls tube.
In particular, the ambient temperature was settled at 28˚C,
while the temperature of the tube surfaces was settled at
200˚C.
The analysis is considerably challenging in terms of
computational costs. For each evaluation the time re-
quired is about one week, by using a Linux cluster with
16 SUN Fire X4100, each server with 2 AMD Opteron
285 (Dual Core) processors and 4 GB RAM (64 proc-
esses).
Several analyses were carried out to choose the time
step values to use. The best solution appears to consider a
time step variable during the flow evolution, depending
on the position of the tube. In particular, a greater value
of the time step at the beginning of the simulation, when
the air jet is evolving and the tube is far from the nozzles,
is acceptable; on the contrary, when the tube is in the jets
zone, a very small value of the time step is necessary,
depending on the translational speed of the tube.
4. Analysis of the Results
4.1. The “Large” Tube
In a first phase the “large” tube, within the described jets
set-up, is analysed.
For each simulation the time histories of the monitored
variables and frames of the aerodynamic field at each
time step are analysed. Data are then processed in order
to obtain the quantities described in the previous section.
The analysis of the videos, obtained by the frames
captured at each time step (Δt = 0.005 s), shows that the
flow behaviour is similar for both the applied Δp, and
G. LOMBARDI, R. CURATOLA 271
that the air regularly flows into the tube; as an example,
the flow fields at four times, at different interference
level between the jet and the tube, are shown in Figure 3.
In order to have quantitative indications on the per-
formance, it is interest to analyse the integral data, re-
ported in Table 4. As previously said, the mass flow
through the end section of the tube is assumed as index
of the available “energy” to remove the particles and,
clearly, higher values are preferred. From this point of
view, an increase in the applied pressure at the nozzles
gives a significant increase, although not proportional to
the increment in Δp. The integral shear stress over the
inner surface of the tube, assumed as the “traction” force
to remove the particles, is essentially proportional to Δp,
such as the maximum τint, indicating the capability to
“start” the movement of a particle.
Summarising, from data reported in Table 4 it is evi-
dent that increasing the applied pressure increase the
capabilities of the system to clean the tube.
It is also interest to analyse the variation in time of the
mass flow rate at different section along the tube, x,
shown in Figure 4. It is evident that, for each of the four
considered sections and for both the applied Δp, very
slight differences in the five periods are present. There-
fore, the flow appears oscillatory but well stabilised since
the first jet. For the first two sections (x/L = 0 and x/L =
0.33) a reflux, negative mass flow rate, is present when
the tube moves from a jet to the next one. It is evident
that a part of the air forced from the jets into the tube is
not able to through up the end of the tube, and this means
that the involved energy is not completely used to re-
move the particles. Furthermore, in a reflux zone the par-
ticles, previously moved towards the exit of the tube,
tends to move in the opposite direction, with a relevant
negative effect on the efficacy of the cleaning process.
Another interest aspect is the variation in time of the
longitudinal distributions, along the tube, of the value of
the shear stress integrated over a cross section, θ, shown
in Figures 5 and 6 for a period (data are represented any
0.13 s). At time t1 the air start to flow into the tube, and a
peak in tangential stress is present at the tube inlet. In-
creasing time the tangential stress peak moves down-
stream and the total tangential stress increases. At time t5,
the peck is close to the end of the tube; at this time the
tube is at the end of the jet, and a reflux begin in the tube
(part b of the Figures 5 and 6), with a significantly re-
duction of the tangential stress. From a qualitative point
of view, the situation is very similar for both Δp, with
higher values for the higher Δp.
4.2. The “Small” Tube
The analysis of the videos, obtained by the frames cap-
tured each time step (Δt = 0.002 s), shows that the flow
behaviour is similar for both the applied Δp, but, on re-
spect to the large tube, the air flows along the internal of
the tube with less regularity, and a significant reflux oc-
curs when the tube left the jet. As an example, the flow
fields at four times, at difference interference level be-
tween the jet and the tube, are shown in Figure 7, for the
case with Δp = 101325 Pa.
The integral data are reported in Table 5.
The mass flow through the end section increases of
only 30%, with a Δp that is three times. This behaviour,
significantly different from the large tube, indicates that
the flow in close to be “blocked” and, therefore, with the
small tube the increase of Δp do not appears a possible
solution to improve the performance of the system, be-
cause the energy is far to be completely transferred to the
flow.
Table 4. Integral results for the large tube.
Applied Δp 101,325 Pa 303,975 Pa
Mass flow [Kg] 0.0116 0.0218
Integral shear stress [N s] 0.6675 2.0165
Max τint [N] 0.0161 0.0419
(a) (b) (c) (d)
Velocity: Magnitude (m/s)
0 40 80 120 160 200
Figure 3. Example of the flow field; large tube, Dp = 101,325 Pa. (a) Initial interference; (b) Complete interference; (c) Final
interference; (d) No interference.
Copyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA
272
Δp = 101,325 Δp = 303,975
x/L
Figure 4. Variation in time of the mass flow rate at different section; large tube.
(a) (b)
Figure 5. Longitudinal distribution of the shear stress in a cross section; large tube, Δp = 101,325. (a) No reflux condition; (b)
Reflux condition.
(a) (b)
Figure 6. Longitudinal distribution of the shear stress in a cross section; large tube, Δp = 303,975. (a) No reflux condition; (b)
Reflux condition.
Copyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA
Copyright © 2013 SciRes. ENG
273
(a) (b) (c) (d)
Velocity: Magnitude (m/s)
0 40 80 120 160 200
Figure 7. Example of the flow field; small tube. (a) Initial interference; (b) Complete interference; (c) Final interference; (d)
No interference.
Table 5. Integral results for the small tube.
Applied Δp 101,325 Pa 303,975 Pa
Mass flow [Kg] 0.0019 0.0025
Integral shear stress [N s] 0.0996 0.2848
Max τint [N] 0.0124 0.0260
Therefore, a different blowing solution was investi-
gated, with the object to improve the cleaning capabili-
ties of the system, particularly for the small tube. The
results will be described in the next section.
4.3. A Proposed Different Blowing System
As previously shown, in the original solution the flow
has a pulsed trend, due to the non-continuity of the air jet.
This leads to a phase with a reverse flow, with a signifi-
cant reduction of the efficacy of the system.
On the contrary, the integral shear stress along the in-
ner surface of the tube, assumed as “traction” force to
remove the particles, appears essentially proportional to
Δp, but the maximum τint, indicating the capability to
“start” the movement of a particle, is far to be propor-
tional to Δp, as occurs for the large tube. From data re-
ported in Table 5 it is evident that, for the small tube,
increasing the applied pressure is not a suitable method
to increase the capabilities of the system to clean the tube.
The variation in time of the mass flow rate at different
section along the tube, x, are shown in Figure 8.
To improve the situation, the proposal is to replace the
nozzles with a single blade, as shown in Figure 11. The
analysed configuration has a span s = 150 mm and height
h = 0.5 mm. Therefore, the blowing surface is 75 mm2,
slightly higher (19.4%) on respect the original solution.
The study is carried out for the small tube (the critical
one), at the higher Δp, 303,975 Pa.
The analysis of the videos, obtained by the frames cap-
tured each time step (Δt = 0.005 s), shows a completely
different flow, as can be seen from Figure 12. It is evi-
dent that, with this blowing system, the flow appears
completely regular during the entire blowing phase.
For each of the four considered sections and for both
the applied Δp, significant differences in the five periods
are present, indicating that the flow is not well stated.
Furthermore, the reflux appears really important, signifi-
cantly higher than for the large tube. Therefore, it is evi-
dent that the cleaning capabilities for the small tube are
significantly reduced.
The integral data are reported in Table 6. The increase
in the system efficacy appears evident. The mass flow
through the end section of the tube, assumed as index of
the available “energy” to remove the particles, shows a
significant increase, about 65%, while the integral shear
stress along the inner surface of the tube, assumed as
“traction” force to remove the particles, increases of 68%.
On the other side, the maximum τint, assumed as index of
the capability to “start” the movement of a particle is
reduced. This because the flow is now regular, while the
high peak in the basic solution appears related to transi-
tional phase of the flow.
The variation in time of the longitudinal distributions,
along the tube, of the value of the shear stress integrated
over a cross section, are shown in Figures 9 and 10 for a
period (data are represented any 0.058 s). The behaviour
is similar to the large tube, but the values are signifi-
cantly lower and the reflux, that appears significantly
higher, occurs at any time.
The variation in time of the mass flow rate at different
section along the tube, x, are shown in Figure 13. It is
clear that the behaviour is now completely different. The
In conclusion, the results confirm the experience, in-
dicating that the adopted solution appears suitable for the
large tube, but not for the small one.
G. LOMBARDI, R. CURATOLA
274
Δp = 101,325 Δp = 303,975
x/L
Figure 8. Variation in time of the mass flow rate at different section; large tube.
(a) (b)
Figure 9. Longitudinal distribution of the shear stress in a cross section; small tube, Δp = 101,325. (a) Initial phase of the pe-
riod; (b) Final phase of the period.
(a) (b)
Figure 10. Longitudinal distribution of the shear stress in a cross section; small tube, Δp = 303,975. (a) Initial phase of the
eriod; (b) Final phase of the period. p
Copyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA 275
Figure 11. Sketch of the geometrical parameters.
(a) (b)
(c) (d)
Velocity: Magnitude (m/s)
0 40 80 120 160 200
Figure 12. Example of the flow field, blade solution. (a) No
interference; (b) Initial interference; (c) Complete interfer-
ence; (d) Regime interference.
flow appears almost constant up to the end of the inter-
ference between the tube and the blowing (the interfer-
ence time, at the considered translational velocity of the
tube, is 1.1 s). Furthermore, the reflux appears only be-
hind the end of the blowing phase.
To highlight the different behaviour of the two solu-
tions, in Figure 14 the time history of the mass flow rate
at the outlet section are shown. It is evident that the blade
solution shows a higher continuity in the flow, with an
Table 6. Integral results for the two proposed blowing so-
lutions.
Scheme Basic Blade
Mass flow [Kg] 0.0025 0.0041
Integral shear stress [N s] 0.2848 0.4785
Max τint [N] 0.0260 0.0202
Figure 13. Variation in time of the mass flow rate at dif-
ferent section; blade blowing.
Figure 14. Outlet mass flow rate for the two blowing solu-
tions.
expected significant increase oh the cleaning capabilities.
Finally, the variation in time of the longitudinal dis-
tribution, along the tube, of the value of the shear stress
integrated over a cross section is shown in Figure 15
(data are represented any 0.15 s, therefore the total time
is significantly higher than in the corresponding Figure
10). The behaviour is completely different from that of
the jets solution. The tangential stress appears almost
constant over all the tube, and no reflux appears up to
final time, when the tube is completely out from the
blow.
Copyright © 2013 SciRes. ENG
G. LOMBARDI, R. CURATOLA
276
Figure 15. Longitudinal distribution of the shear stress in a
cross section; blade solution.
In conclusion, all the results indicate that the cleaning
capabilities would be significantly improved with the
blade blowing system.
5. Conclusions
The problem of the removal, from the inner of the tube,
of the waste arising from the cut operation by means of
an air blowing system was analysed. In order to obtain
indications about the real importance of the involved
parameters and their effects, a numerical procedure,
based on CFD analysis trough the code STAR CCM+,
has been settled.
With the support of the numerical analysis the effects
of the parameters were highlighted. The results confirm
the experience, indicating that the adopted solution ap-
pears suitable for the large tube, but not for the small one.
Therefore, a different blowing solution was investigated,
with the object to improve the cleaning capabilities of the
system.
The results indicate that, in order to remove the waste
arising from the cutting operation, a significant im-
provement of the capability can be expected with the new
blowing solution.
REFERENCES
[1] A. Guha, R. M. Barron and R. Balachandar, “An Experi-
mental and Numerical Study of Water Jet Cleaning Proc-
ess,” Journal of Materials Processing Technology, Vol.
211, No. 4, 2011, pp. 610-618.
[2] M. M. Pror, “Automated Moving Mesh Techniques and
Re-Meshing Strategies in CFD Applications Using Mor-
phing and Rigid Motions,” CRS4 Technical Report, 2012.
Copyright © 2013 SciRes. ENG