![Figure 4. (a) Views of the air blowing experimental unit. (b) Schematic drawing of the experimental set-up. [AQ13]](/figures/figure-4-a-views-of-the-air-blowing-experimental-unit-b-1nnxg0i8.webp)
![Figure 1. [AQ30]Single station PET bottle production pneumatic scheme with air recovery system. [AQ5]](/figures/figure-1-aq30-single-station-pet-bottle-production-pneumatic-2gvizikz.webp)
![Figure 3. Computational grid along the axial direction of a non-tapered pipe for the HLL first-order scheme. [AQ9]](/figures/figure-3-computational-grid-along-the-axial-direction-of-a-2jlfbnsc.webp)
Original Article
Air recovery assessment on
high-pressure pneumatic systems
Jose
´
A Trujillo, Pedro J Gamez-Montero
and Esteban Codina Macia
`
v
Abstract
A computational simulation and experimental work of the fluid flow through the pneumatic circuit used in a stretch blow
moulding machine is presented in this paper. The computer code is built around a zero-dimensional thermodynamic
model for the air blowing and recycling containers together with a non-linear time-variant deterministic model for
the pneumatic three stations single acting valve manifold, which, in turn, is linked to a quasi-one-dimensional unsteady
flow model for the interconnecting pipes. The flow through the pipes accounts for viscous friction, heat transfer,
cross-sectional area variation, and entropy variation. Two different solving methods are applied: the method of charac-
teristics and the HLL Riemann first-order scheme. The numerical model allows prediction of the air blowing process and,
more significantly, permits determination of the recycling rate at each operating cycle. A simplified experimental set-up of
the industrial process was designed, and the pressure and temperature were adequately monitored. Predictions of the
blowing process for various configurations proved to be in good agreement with the measured results. In addition, a
novel design of a valve manifold intended for the polyethylene terephthalate (PET) plastic bottle manufacturing industry is
also presented.
[AQ1]
Keywords
Air recycling system, energy assessment, air blow moulding manufacturing process, pneumatic valve manifold
Date received: 6 November 2015; accepted: 10 March 2016
Introduction
Despite the many contributions linked to energy con-
servation in pneumatic systems, no publications report
the efficiency on high-pressure pneumatic applications.
[AQ2]In order to bring some light to this issue, it is
crucial to get into the patents published during the last
20 years. Amongst various industrial applications that
require high-pressure air, polyethylene terephthalate
(PET) stretch blow moulding machine manufacturers
have contributed significantly to enhancing energy effi-
ciency in that specific field of pneumatic systems.
In 1981 Air Products and Chemicals Inc.
1
pub-
lished a patent related to a process for the production
of blow moulded articles in which the blowing gas was
recovered and treated to be used in subsequent
moulding operations. A year later Robert Bosch
GmbH suggested recovering the compressed air used
in the moulding operation to feed other pneumatic
applications. A similar proposal was provided by
The Continental Group Inc.
2
in 1984, which was sub-
sequently taken as a reference by other blow moulding
bottle manufacturers. In 1995 Krupp Corpoplast
Maschinenbau GmbH
3,4
presented an invention that
recovered part of the air used for moulding a
container made of thermoplastic material. The high-
pressure blowing air was supplied to the low-pressure
air supply during a transitional phase by employing a
reversing mechanism. An invention that has been
cited by several blowing machine manufacturers is
the patent of Procontrol AG (1996),
5
which proposed
to produce the high-pressure air adiabatically while
the low-pressure air was generated isothermically,
thus enabling the entire blowing process to be carried
out with the smallest possible amount of energy. Over
the same period and based on the same principle,
A.K. Tech Lab Inc. (1997)
6
proposed recovering the
exhaust air into a tank that later supplied air to oper-
ate secondary pneumatic circuits. In order to compen-
sate for the difference between the recovered air and
that consumed by the installation, a compressor
LABSON, Campus Terrassa, UPC, Terrassa, Barcelona, Spain
Corresponding author:
Jose
´
A Trujillo, Escuela Tcnica Superior de Ingenieras Industrial y
Aeronutica de Terrassa, Campus de Terrassa, Edificio TR5, C/ Colom,
11 Terrassa, 08222 Spain.
Email: jose.trujillo@ast.es
Proc IMechE Part C:
J Mechanical Engineering Science
0(0) 1–12
! IMechE 2016
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954406216645823
pic.sagepub.com
provided sufficient air to balance the pressure in the
tank. Also, the proposal of Asahi Kasei Kogoyo
Kabushiki Kaisha (1993)
7
must be taken into account,
which added a recovery container from which the com-
pressed gas could be aspirated by a multistage com-
pressor. In 2003 Technoplan
8
published an invention
which targeted the optimization of the above-men-
tioned methods. A relevant improvement was the fact
that the recovered gas (17 bar) was expanded before
being used in the low-pressure air phase, which meant
that it did not have influence on the low-pressure air at
the time of its use. On the other hand, several proposals
were given to re-use the recovered air, such as actuating
the preform-stretching rams, actuating consumables
of the packaging-production machine, or even return-
ing the recycled gas to the compressed air network. The
method allowed around 20% to 45% of air recovery
and a reduction of electrical power consumption of
15% to 45%.
[AQ3]Based on the existing state of the art it may
be concluded that even though numerous attempts
have been made to improve the efficiency of air blow-
ing pneumatic systems, there are no previous publica-
tions which focused specifically on analysing the
complexity of this particular industrial field.
Therefore, this investigation aims to determine the
main constraints that limit the efficiency of a blow
moulding plastic PET bottle pneumatic circuit with
the help of a computational model which is able to
predict the maximum amount of recycled air that may
be ensured at each operating cycle. Moreover, this
tool will not only contribute to assessing the efficiency
of the air blowing machine but will also allow re-
designing of the regpneumatic lay-out to minimize
the energy losses.
9
Mathematical model of the air blow
moulding pneumatic system
[AQ4]
Due to the complexity of the air blow moulding
machine, the pneumatic circuit has been reduced to
the pneumatic scheme depicted in Figure 1, resulting
Figure 1. [AQ30]Single station PET bottle production pneumatic scheme with air recovery system. [AQ5]
2 Proc IMechE Part C: J Mechanical Engineering Science 0(0)
in three individual submodels, which are represented
by the fluid flow through the pipes, the charging and
discharging process from/to the vessels and the fluid
dynamics inside the valve manifold. The valve mani-
fold is supplied with two different pressures, and a
special cylinder is responsible for providing com-
pressed air to the plastic preform through a hollowed
stretching rod. From the patents mentioned in the
previous section we learned that once the plastic
bottle is produced, the air inside the container is par-
tially recycled while the remaining fluid is exhausted
to the atmosphere once the air level inside the recy-
cling chamber reaches a certain pressure. It must be
pointed out that the main scope of this study does not
take into account the deformation of the preform
during the blowing process, but the amount of air
that is needed to produce the bottle. As a matter of
fact the pressure characteristics inside the mould will
behave slightly differently in a real blow moulding
machine. On the other hand, for the sake of simplicity
the simulation will omit the components located
before the valve manifold, such as the filter and pres-
sure regulator.
The recycling stage always takes place after closing
V
1
(refer to Figure 1). At this point the air flows
through the pipe connecting the cavity chamber and
the manifold, and circulates through the valve mani-
fold until it reaches the recycling chamber. At a cer-
tain stage, the air in the recycling chamber equalizes
the pressure in the cavity chamber, being the point
when the recycling process ends, and the remaining
air in the cavity chamber is released to the atmos-
phere. As a matter of fact, the use of an additional
recycling process may be also considered at this point,
however, a different concept design of the valve mani-
fold should be used. It must be noted that the amount
of energy available in the cavity chamber drops as the
pressure decreases so an additional recycling stage
should be considered.
Mathematical model at the pipes
The flow through the pipes connecting the different
units has been considered quasi-one-dimensional and
the methods implemented in order to determine the
characteristics of the fluid flow have been the method
of characteristics (MOC)
10,11
and the HLL Riemann
solver
12–14
respectively. [AQ6] Both models were
implemented in Fortran and only differed in the way
that the governing equations were solved. The simu-
lations were run on a x86 (32-bit) architecture
Pentium processor with a dual Intel Core Quad
CPU 2.4 GHz processor and 3.0 GB memory.
Zero-dimensional thermodynamic volume
The performance of the recycling system is deter-
mined largely by the efficiency of the processes of
charging and discharging. The vessels have been
discretized by a zero-dimensional model, and the gov-
erning equations are as follows.
. Non-adiabatic charging:
dP
dt
¼
_
m
in
RT
V
T
in
T
v
2
in
2c
v
T
ð 1Þ
w
A
w
T
PV
1
T
w
T
ð1Þ
dT
dt
¼
_
m
in
RT
2
PV
T
in
T
1
v
2
in
2c
v
T
ð 1Þ
w
A
w
T
2
PV
1
T
w
T
ð2Þ
. Non-adiabatic discharging:
dP
dt
¼
_
m
out
RT
V
T
out
T
v
2
out
2c
v
T
w
A
w
T
PV
1
T
w
T
ð3Þ
dT
dt
¼
_
m
out
RT
2
PV
T
out
T
1
v
2
out
2c
v
T
ð 1Þ
w
A
w
T
2
PV
1
T
w
T
ð4Þ
where the suffix ‘in’ refers to the port where inflow
occurs, and the suffix ‘out’ refers to the port where
outflow occurs. It must be taken into account that the
equations above are only valid under the assumption
that a perfect mixing of the fluid to an equilibrium
state occurs, so the use of a single pressure and tem-
perature describe the state of the gas in the vessels.
Mathematical model of the valve manifold
The following discussion assumes that the spool valve
only moves in the axial direction. Therefore, the devi-
ation from the central position caused by unsteady
transverse flow forces was not taken into account.
The alignment of the spool valve with respect to the
valve body is a basic factor in avoiding possible eccen-
tricities which may cause a rotating movement of the
spool valve, that may consequently lead to the gener-
ation of a moment with respect to its central axis.
The control volume depicted in Figure 2 describes
the nature of F
s
, which is represented by the static
pressure force acting on the spool valve and the flow
force F
f
yielded by the flow passage across the valve
that originates a linear momentum change.
Therefore, based on the previous assumptions the
dynamics of each spool valve is given by
m
s
v
i
ð
€
z
v
i
þ g Þþc
f
_
z
v
i
þ k
v
i
ðz þ z
o
Þ
v
i
¼ F
f
v
i
þ F
s
v
i
ð5Þ
Trujillo et al. 3
where z is the instantaneous vertical displacement
referenced from the seat, k
v
i
is the spring rate,
F
s
v
i
and F
f
v
i
are the pressure forces acting on the
entire control surface and the flow forces respectively,
and v
i
is the index assigned to each spool valve.
The equation describing the dry friction force
between the contacting surfaces can be mathematic-
ally represented as follows:
15–19
F
c
¼
F
c
n
sgnð
_
zÞ if
_
z 6¼ 0
if jj5 F
c
0
if
_
z ¼ 0
F
c
0
sgnðÞ if jj5F
c
0
if
_
z ¼ 0
8
>
<
>
:
where F
c
n
is the nominal dry friction force on the spool
valve, F
c
0
is the initial dry friction force on the spool
valve, and ¼
P
n
i¼1
P
i
A
i
F
f
v
i
F
s
v
i
represents the
balance of forces acting on the spool valve body.
After applying the Navier–Stokes equations in vector
form in the control volumes shown in Figure 2, the
result will be as follows:
m
s
v
i
€
z
v
i
þ c
f
_
z
v
i
þ k
v
i
ðz þ z
o
Þ
v
i
¼ðA
p
P
p
Þ
v
i
þðA
s
P
s
Þ
v
i
ðA
u
P
u
Þ
v
i
ðA
l
P
l
Þ
v
i
ðA
n
P
n
Þ
v
i
þ m
s
v
i
g
@
@t
_
mðz þ z
o
Þ½
_
mv
out
v
in
ðÞ ð6Þ
The steady-state form of equation (6) is
k
v
i
ðz þ z
o
Þ
v
i
¼ðA
p
P
p
Þ
v
i
þðA
s
P
s
Þ
v
i
ðA
u
P
u
Þ
v
i
ðA
l
P
l
Þ
v
i
ðA
n
P
n
Þ
v
i
þ m
s
v
i
g
_
mv
out
v
in
ðÞ
ð7Þ
which can be manipulated in order to determine the
minimum force required to shift the valve from the
rest position,
A
p
P
p
Þ
v
i
5k
v
i
ðz þ z
o
Þ
v
i
ðA
s
P
s
Þ
v
i
ðA
u
P
u
Þ
v
i
ðA
l
P
l
Þ
v
i
ðA
n
P
n
Þ
v
i
þ m
v
i
g
ð8Þ
The mass flow through the spool valve openings
can be either subsonic or sonic depending on the
pressure ratio between inlet and outlet pressure.
Figure 2. (a) Static and flow fluid forces acting on the spool valve lower packing before and after opening. (b) Schematic view of the
main valve body and pilot ports of the pneumatic unit.
[AQ7]
4 Proc IMechE Part C: J Mechanical Engineering Science 0(0)
We, therefore, get
_
m
m2
v
i
¼
C
d
2
v
i
A
0
1
P
ðsÞ,v
i
ffiffiffiffiffiffiffiffi
T
ðsÞ,v
i
p
if
P
ðmÞ, v
i
P
ðsÞ, v
i
4P
cr
C
d
2
v
i
A
0
2
P
ðsÞ,v
i
ffiffiffiffiffiffiffiffi
T
ðsÞ,v
i
p
P
ðmÞ,v
i
P
ðsÞ,v
i
1=
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
P
ðmÞ,v
i
P
ðsÞ,v
i
ð1Þ=
r
8
>
>
<
>
>
:
9
>
>
=
>
>
;
if
P
ðmÞ, v
i
P
ðsÞ, v
i
4 P
cr
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
_
m
m3
v
i
¼
C
d
3
v
i
A
0
1
P
ðmÞ,v
i
ffiffiffiffiffiffiffiffiffi
T
ðmÞ,v
i
p
if
P
ðuÞ, v
i
P
ðmÞ, v
i
4P
cr
C
d
3
v
i
A
0
2
P
ðmÞ,v
i
ffiffiffiffiffiffiffiffiffi
T
ðmÞ,v
i
p
P
ðuÞ,v
i
P
ðmÞ,v
i
1=
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
P
ðuÞ,v
i
P
ðmÞ,v
i
ð1Þ=
r
8
>
>
<
>
>
:
9
>
>
=
>
>
;
if
P
ðuÞ, v
i
P
ðmÞ, v
i
4 P
cr
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
_
m
m4
v
i
¼
C
d
4
v
i
A
0
1
P
ðresÞ, v
i
ffiffiffiffiffiffiffiffiffiffi
T
ðresÞ, v
i
p
if
P
ðpÞ, v
i
P
ðresÞ, v
i
4P
cr
C
d
4
v
i
A
0
2
P
ðresÞ, v
i
ffiffiffiffiffiffiffiffiffiffi
T
ðresÞ, v
i
p
P
ðpÞ,v
i
P
ðresÞ, v
i
1=
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
P
ðpÞ,v
i
P
ðresÞ, v
i
ð1Þ=
r
8
>
>
<
>
>
:
9
>
>
=
>
>
;
if
P
ðpÞ, v
i
P
ðresÞ, v
i
4 P
cr
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
where C
d
v
i
is a non-dimensional discharge coefficient
referring to the corresponding spool valve seat, and
the subscript ‘(res)’ refers to the reservoir that supplies
air to the pilot port of the valve manifold. On the
other hand the stagnation pressure and temperature
of the fluid upstream and downstream of the restric-
tion will alternately vary depending on the flow dir-
ection, and this applies equally to the downstream
stagnation pressure. The following are constants
that depend on the specific heat ratio of the given
fluid:
1
¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R
2
þ 1
þ1
1
s
;
2
¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
Rð 1Þ
s
;
P
cr
¼
2
þ 1
1
The air flowing through the piloting channels, incor-
porated in the lower packing of the spool valves, is
assumed to be laminar,
20
and is determined by
_
m
m1
v
i
¼ %
av
d
4
c
128
av
P
l
c
ð9Þ
where d
c
and l
c
are the internal diameter and length of
the piloting channels,
av
and %
av
are the average
value of the dynamic viscosity and density of the
fluid, and P is the pressure drop between internal
volumes.
The flow entering and exiting each valve port
_
m
0=5
v
i
will be calculated by the results obtained at the
boundary conditions applied to the pipe ends.
[AQ8]On the other hand, the flow through any
narrow annular clearance, where a sealing component
is located, was ignored. This assumption was experi-
mentally supported by ensuring that no internal leak-
age occurred when operating the unit.
Boundary conditions
The procedure used to determine the variable values
at the boundaries has been based on solving the gov-
erning equations through a convergent nozzle.
According to the discretization shown in Figure 3,
the internal cavities of the valve manifold located
immediately after the pipe ends were taken as small
control volumes inside which the physical properties
of the fluid could be determined under certain
assumptions. Contrary to what occurs when consider-
ing the boundary conditions near a high-volume res-
ervoir the speed of the fluid cannot be disregarded,
and therefore the stagnation pressure will be influ-
enced by the kinetic energy of the fluid at each specific
control volume. The different cases that must be taken
into account are as follows.
Figure 3. Computational grid along the axial direction of a non-tapered pipe for the HLL first-order scheme. [AQ9]
Trujillo et al. 5