Journal of Composite

Materials http://jcm.sagepub.com/

Winding Trajectory and Winding Time in Robotized Filament Winding of Asymmetric

Shape Parts
W. Polini and L. Sorrentino
Journal of Composite Materials 2005 39: 1391
DOI: 10.1177/0021998305050431

The online version of this article can be found at:

http://jcm.sagepub.com/content/39/15/1391

Published by:

http://www.sagepublications.com

On behalf of:

American Society for Composites

Additional services and information for Journal of Composite Materials can be found at:

Email Alerts: http://jcm.sagepub.com/cgi/alerts

Subscriptions: http://jcm.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://jcm.sagepub.com/content/39/15/1391.refs.html

>> Version of Record - Jul 11, 2005

What is This?
Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014
 Winding Trajectory and Winding Time
in Robotized Filament Winding
of Asymmetric Shape Parts

W. POLINI AND L. SORRENTINO*

Dipartimento di Ingegneria Industriale, Università degli Studi di Cassino,
via G. di Biasio 43, 03043 Cassino, Italy

(Received March 23, 2004)

(Accepted October 13, 2004)

ABSTRACT: Robotized filament winding is a technology to deposit a roving on a

die according to a planned robot trajectory and to polymerize the wound part in
a furnace for thermoset resin matrix. In this way, it is possible to wind asymmetric or
complex parts. The winding trajectory is a critical aspect of the robotized filament
winding process, since the mechanical performances and the winding time of the
manufactured composite part change according to winding trajectory geometry. The
present work discusses original criteria for the determination of a proper winding
system trajectory and its influence on the winding time. This work studies the
influence of the geometric parameters, that characterize the trajectory covered by
the deposition system to wind the roving on the die, on winding time. The obtained
results may help to identify the values of the geometric parameters that minimize the
winding time and, at the same time, that keep constant the winding tension in order
to assure an acceptable quality of composite parts. This work represents the first step
to optimize the planning of the winding trajectory.

KEY WORDS: winding trajectory, winding time, robotized filament winding,

asymmetric-shape parts, composite parts.

INTRODUCTION

ILAMENT WINDING TECHNOLOGY allows to design and to manufacture workpieces of

F high performances, since it gives the possibility to place fibers impregnated by resin
(known as roving) along the directions of stresses the workpiece is submitted to in exercise.
Thus, filament winding favors rigidity and strength of a workpiece along some preferential
directions.
The possibility to design the material performances, by acting on the fibers and/or on
the matrix, is the most interesting property of a composite. Robotized filament winding

*Author to whom correspondence should be addressed. E-mail: [email protected]

Journal of COMPOSITE MATERIALS, Vol. 39, No. 15/2005 1391

0021-9983/05/15 1391–21 $10.00/0 DOI: 10.1177/0021998305050431
ß 2005 Sage Publications

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1392 W. POLINI AND L. SORRENTINO

is the evolution of traditional filament winding, since it tries to manufacture workpiece

shapes that, otherwise, will not be manufactured by traditional technology. A robotized
filament winding system is composed of an industrial robot, opportunely equipped with a
feed and deposition system, and a winding die. Coordinating the relative motions of these
two parts, it is possible to describe a 3D deposition trajectory of the roving on the winding
die. The process involves the roving deposition, according to the planned robot trajectory,
and its polymerization, that may be accomplished in a furnace for thermoset resin matrix.
The robotized filament winding cell replaces the human operator representing the only
system that was previously used to manufacture asymmetric parts with a full section.
It represents an advantage as regards process control, repeatability, and manufacturing
times.
The family of asymmetric parts with a full section is complex to manufacture
by robotized filament winding technology. The deposition system should move along
a critical winding trajectory. It is a trajectory, where bending and sharp direction shifts
cause fiber unfastenings or speed variations during the roving fiber winding that
provoke remarkable dishomogeneities of a manufactured part [2]. The winding trajectory
should take account of roving tension loosening and roving location inaccuracy due to
winding speed. The roving tension, that has to be as constant as possible, influences
directly the compaction and the alignment of the fibers [1]. The composite resistance
against loads applied along the roving deposition direction is due to both its fibers, the
higher the fiber percentage in a unitary volume, the greater the workpiece mechanical
resistance, and the presence of defects in the workpiece. Therefore, it is possible to adjust
the amount of fibers for a unitary volume by means of roving tension that influence fiber
compactness.
When the robot performs a winding trajectory, it approximates a path for points; the
deposition system, which represents the robot end effector, moves along a sequence of
points and performs the winding trajectory. The locating error of the deposition system in
every point of the path increases with the winding speed. Therefore, the manufactured part
will have fibers distribution that will be different from the designed one. Consequently,
there will be a deterioration of the structural characteristics. The winding trajectory
may amplify the winding problems, such as fiber bridging, slippage tendency, and fiber
twisting. Fiber bridging is due to the concavity surface of the part. In fact, the tension,
which is applied to the roving wound around a concave surface of the die, produces a fiber
bridge, since the fiber does not adhere perfectly to the die surface. If the surface is concave
in one direction, the problem can be solved by changing the winding trajectory. The
slippage tendency depends on the type of the trajectories that are being performed.
Generally, with a curved surface there is a geodetic path, joining two surface points with
the smallest distance. A roving, located along a geodetic path, is stable on the surface.
If not, a roving slippage toward the nearest geodetic or a balance position happens. In the
last case, the roving direction will no more be coincident with the designed one, even
though small variations are tolerated. A necessary condition to reduce the roving slippage
is to guarantee a constant roving tension during winding. Another critical factor on
manufacturing quality is the twisting of the roving around its longitudinal axis. During
the deposition, because of the relative motion between the feed system and the deposition
one, and between the deposition system and the deposition die, rotations can be produced,
that cause the wear and the twisting of the roving around its longitudinal axis. This occurs
especially while sharp curvatures, that characterize the complex parts, are presented
implying an inversion of the winding direction. This phenomenon represents a material

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1393

fault and can be tolerated if it occurs without influencing the part working. A possible
solution could be to locate the fault, due to roving twisting, in the least stressed areas of
the component by changing the winding trajectory.
The planning and the optimization of the winding trajectory, associated with the
optimization of other technological parameters, such as winding speed and roving tension,
would allow to minimize many of the winding problems previously described.
The planning of the winding trajectory is a critical aspect of the robotized filament
winding process, since the mechanical performances and the winding time of the
manufactured composite part change according to the winding trajectory geometry.
Different alternative CAD/CAM software packages may be found in the literature [3–11],
while there is no method able to optimize the generation of the winding trajectory and,
therefore, there is no software able to generate the optimized winding trajectory or to
compare alternative winding trajectories in order to identify the best.
The first step toward trajectory optimization is the identification of those geometric
parameters, which characterize the winding trajectory by changing its shape and
perimeter’s value and, therefore, that influence the time spent by the deposition system
to cover the trajectory. The winding time strongly depends on the acceleration and
deceleration ramps due to the direction changes of the deposition system. The same
geometric parameters may involve loosening of tension and, therefore, a decrease of the
mechanical performances of the wound composite parts.
This work studies the influence of the geometric parameters that characterize the
trajectory covered by the deposition system to wind the roving on the die, on winding time.
It starts from the geometric parameters introduced to compare alternative winding
trajectory in [12] and from the study of the value of the winding tension that gives the best
performances in terms of roving alignment and compactness [13,14]. They show how an
increase of the mechanical performances of the composite parts due to the constancy of
tension value along winding implies, in the same time, an increase of the winding time.
It uses an experimental approach by means of the robotized cell available at the
Manufacturing Laboratory of Engineering Faculty in Cassino, Italy.
The obtained results may help to identify the values of the geometric parameters that
minimize the winding time and, at the same time, that keep constant the winding tension in
order to assure an acceptable quality of composite parts. This work represents the first
step to optimize the planning of the winding trajectory.
First, the considered kind of composite parts is introduced. This work aims to identify
the geometric parameters, of the trajectory covered by the deposition system to wind
the roving on the die, that influence winding time but at the same time to preserve the
constancy of the winding tension, thus assuring an acceptable quality of the manufactured
composite parts.

COMPOSITE PARTS WITH A COMPLEX SHAPE

A part obtained thanks to the extrusion of a full section along a closed, not auto-
intersecting path is the family of workpieces that have been considered in the present
work. This kind of part is impossible to manufacture by traditional technology, while it is
possible, even if complex, to obtain by robotized filament winding technology. The
closed path is formed by 3D curves, the part is automatically produced through a
sweeping operation. The composite roving has an approximate rectangular section smaller

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1394 W. POLINI AND L. SORRENTINO

than part section. Therefore, the part section filling can be obtained through a continuous
location of the roving section. A grid is positioned on the straight section of the part: it has
cells with dimensions equal to the roving ones in the straight section, positioned over it
(see Figure 1). The cells represent the various coils of the roving, necessary to define the
layers of the whole part section. A continuous deposition is realized starting from one
place (such as 1) and ending, after a whole winding round, in a new one (such as 2) from
which a new winding coil starts. The support gives the supporting walls for the most
external of the reinforcement, and because of the shape of the section, it is possible to take
two or three retaining surfaces into consideration. It is possible to define two main
directions: the first one Y (versor U1) determines the growth of each single layer by
approaching single coils, while the second main direction X (versor U2), which is
perpendicular to the first one, constitutes the direction where the multiplication of the
layers takes place. The two directions are associated to the supporting surfaces and, in
particular, to the sides of the product section to be analyzed.
A possible deposition strategy is the continuous layering that allows to align layers that
are generated by the deposition of the filament according to opposite directions along the
Y-axis, as shown in Figure 2.
This strategy allows to avoid that voids and bridges are present inside the part.
Independently by layering, the path along which the roving has to be wound must be

Y
Y

... ... ... 3

... ... ... 2

n ... ... 1
X
X
Z

Figure 1. Irregular ring section.

Y
Y

... 9 4 3

... 8 5 2

n 7 6 1
X X
Z
Figure 2. Continuous layering.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1395

generated; it is called base path. Then, the trajectory along which the deposition system
should move in order to wind the roving along the defined base path needs to be defined.
The trajectory of the deposition system is a set of points ordered in space; it represents the
image of the points of the base path. The existence of a relation of biuniqueness between
the points of the base path and the deposition system trajectory ones, is absolutely
necessary. It is possible to generate the winding coils through extrusion along the base
path of the cells belonging to the straight section of the workpiece. To guarantee an
accurate winding of the roving on die, the deposition system trajectory must be tangential
to the roving trajectory along each contact point between the winding roving and the
already wound roving. Moreover, the roving tension must be as constant as possible, the
deposition system and the robot arms should be moved on collision-free trajectories,
the free roving must not interfere with the support and the whole environment. A control
volume is generated to calculate the impacts, in order to keep the path outside the
considered volume. The control volume is a positive offset of the solid model related to
part-support assembly. The deposition system has to respect the safety distance (d ), which
is defined by the user, the bearing, and by the deposited fiber.

INFLUENCE OF WINDING TRAJECTORY ON WINDING TIME

The winding trajectory is constituted by the sequence of points, ordered in space, along
which the deposition system moves in order to deposit the composite roving on the die.
It represents the image of the points of the base path, i.e., point A, A1, A2, . . . in Figure 3.
This means that when the deposition system, i.e., the robot end effector,
performs the 
winding trajectory, it approximates the continuous path by points A0 , A01 , A02 , . . . .
The number of points (n) used to discretize the winding trajectory influences the regularity
of the deposition system movement. In fact, an increase of the number of points makes the
movement of the deposition system more continuous during winding, since it avoids
the sudden change in head’s direction. A more and more continuous movement of

Deposition head trajectory or

winding trajectory

A'1 A'2

A'n
An
A2
A'
A1
Winding die

Base path or
deposition trajectory

Figure 3. Discretized points.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1396 W. POLINI AND L. SORRENTINO

the deposition system makes less probable the occurrence of tension loosen during
winding and it increases the accuracy and the repeatability in performing the winding
trajectory.
The deposition system moves from one point to the following one of the trajectory
during winding. The angle that the vector of the deposition system movement from point
to point forms with the roving direction is very critical for winding. It is called the
trajectory angle and is indicated by
. The trajectory angle is responsible for roving tension
control during winding. It aims to avoid decrease in the tension value of the roving during
winding, i.e., roving loosens. Figure 4 shows the deposition of the roving from point A1 to
point A2 on the winding die: on the left the roving is placed on point A1, while on the right,
it is on A2. To deposit the roving between points A1 and A2, the deposition system moves
from point A01 to point A02 . During its movement from A01 to A02 , the trajectory of the
deposition system A01 A02 has to form with the roving direction A1 A01 a
-angle greater or
equal to 90
, in order to satisfy the condition A1 A02  A1 A01 , that avoids roving loosenings.
This
-angle condition is valid when the process parameters assume nominal values, i.e.,
the winding speed is constant among the different trajectory points, the unwinding of
roving from the spool is uniform, and the robot inertia may be neglected. Those three
factors may be considered as process troubles. If the process goes away the nominal
conditions, the value of the
-angle should increase in order to compensate the influence of
process troubles.
The deposition system moves along the trajectory points by keeping at the safety
distance (d ) from the die in order to avoid collisions with the die during winding. An
increase of the safety distance may avoid collisions between the deposition system or
the robot arms and the winding die during winding, especially for small parts. The value of
the safety distance strongly depends on the value of the trajectory angle. If the value of the
safety distance does not allow to satisfy the condition on the value of the trajectory angle
previously introduced (
 90
), the value of the safety distance should be increased as far
as the trajectory angle satisfies its constraint. In fact, during its movement from A01 to
point A02 along the control volume in Figure 5 on the right, the trajectory of the deposition
system A01 A02 does not form with the roving direction A1 A01 a angle greater or equal to 90
,
such as happens when the deposition system moves from point A0 A01 in Figure 5 on the left.

Figure 4. Trajectory angle.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1397

Figure 5. Safety distance.

Deposition head
trajectory
Ws_2
A'1 A'1 A'2
l2
1
s_
W

l1

A2 A2 A2
A' A' A'
A1 A1 A1
Winding die Winding die Winding die

A A A
Base path or Base path or Base path or
deposition trajectory deposition trajectory deposition trajectory

Figure 6. Winding trajectory length and winding speed.

Therefore, the safety distance has to be increased to d 0 >d in order to have a trajectory
angle
at least equal to 90
.
We may conclude that the winding trajectory of the deposition system is characterized
by the three geometric parameters previously described: the number of discretized points,
the trajectory angle, and the safety distance [12].
The length of the winding trajectory

is easily calculated
 by the distance of each couple of
points constituting the trajectory A0 , A01 , A02 , . . . . Figure 6 shows the winding speed (Wsi)
by which the deposition system moves from point A0 (on the left) to point A01 (in the
middle) and, then to point A02 (on the right). The resulting winding time may be calculated
by dividing the trajectory length by the winding speed of the deposition system, but the
obtained result may be very distant from the theoretical one, since the path along which
the deposition system accelerates or decelerates may significantly influence the resulting
time needed to wind the workpiece. Therefore, it is possible to decrease the winding time

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1398 W. POLINI AND L. SORRENTINO

by choosing the value of the geometric parameters that characterize the winding
trajectory. At the same time, the chosen values of the geometric parameters have to assure
an acceptable quality of the composite parts. The influence of the geometric parameters
characterizing the winding trajectory on the winding time and the help to choose the
suitable value of those parameters that allow both to satisfy the composite quality
constraints and to decrease the winding time are described here.

EVALUATION OF WINDING TIME

Experimental tests have been carried out by means of design of experiment (DOE)
techniques to evaluate the influence of geometric parameters of the winding trajectory on
the winding trajectory time. The trajectory angle, the number of discretized points, the
safety distance, and the nominal speed of the deposition system have been considered as
variable parameters, while the winding tension has been kept fixed at 70 N. A full section
workpiece belonging to the family introduced in the section on ‘‘Composite Parts with a
Complex Shape’’ has been used as a benchmark as described under the following heading.
Moreover, the composite material and the robotized cell are presented together with the
trajectories planned by changing the value of the previously described geometric
parameters.

Benchmark

The benchmark is an irregular ring, shown in Figure 7. It is commonly used by a

well-known Italian aeronautic company to test alternative composites manufacturing
technologies and systems.
The material used for the experimental tests is carbon roving impregnated by epoxy
resin, conformed to MIL-R-9300 requirements. The slip roving consists of 12 thousand
(12K) filament count tows. Polyacrilonite (PAN) precursor graphite fibers are used. The
slip roving has a 3.2  0.8 mm width and a 0.76–0.85 g/m yield.
The part usually requires about 90 revolutions around the supporting die. This is the
number of revolutions required to deposit the amount of fibers needed to guarantee
the composite strength, that was defined during the design stage, once the roving tension
that influences fiber compactness is fixed.

Winding die Carbon roving

Figure 7. Irregular ring benchmark.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1399

The Robotized Filament Winding Cell

The manufacturing cell is composed of an anthropomorphic robot, opportunely

equipped with an unique and innovative device and a winding die [13], shown in Figure 8.
The robot is an anthropomorphic Kuka, with 6 d.o.f., payload of 45 kg, maximum reach
of 2041 mm, work envelope volume of 24 m3, repeatability <0.15 mm.
The winding device, already described in previous works [15,16], has been designed and
built on the basis of compactness, structural lightness, stiffness, and functionality
principles, in order to guarantee both the maximum dexterity of the robot, to minimize the
probability of crashes between the winding die and the components of the cell, and to
improve the control of the process parameters for accuracy and repeatability. The feeding
device shows a modular structure, shown in Figure 9, constituted by four critical sub-
groups or modules: the main frame, the roving guide system, the roving tensioner, and the
deposition system.

Figure 8. Robotized filament winding cell.

Figure 9. Feed deposition device.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1400 W. POLINI AND L. SORRENTINO

The winding die is mounted on a circular plate by means of a tie rod. The plate allows
to mount the tie rod in different locations as required by the shape of the winding die. The
length of the tie rod is reduced in order to avoid collisions during winding.

Planning of Winding Trajectory

For the planning of the winding trajectory, the nominal speed of the deposition system,
the trajectory angle, the number of discretized points, and the safety distance have been
considered as variable parameters, while the winding tension has been kept fixed. The
nominal speed (S) has been fixed to 50, 75 and 100% of the maximum value of the robot
linear speed (S*) which is 2 m/s. A set of experimental tests has been designed by means of
a factorial experimental plan (Table 1). We have twenty-four possible trajectories, that are
obtained by combining the two values of the trajectory angle with the three values of the
discretized points and with the four values of the safety distance, for each considered value
of the robot speed. However, we have obtained 20 trajectories instead of twenty four, since
for some combinations of the geometric parameters (trajectory angle, discretized points
and safety distance), the software does not converge toward a valid trajectory. A trajectory
is considered valid if it satisfies the constraints described in ‘‘Influence of Winding
Trajectory on Winding Time’’. Each trajectory has been replicated three times for each
value of the robot speed, yielding a total of 180 tests.
The winding trajectory has been planned by means of a CAD/CAM software designed
for robotized filament winding [12,17–19]. The software starts from the 3D model of the
composite part where the deposition system can not enter during winding to generate the
control volume, i.e., the safety volume, to plan the base path, i.e., the path of the roving on
the die, and finally to plan the corresponding winding trajectory. The software takes into
account the values of the geometric parameters, previously identified, in order to generate
the trajectories. The control volume depends on the safety distance, while the winding
trajectory depends on the number of discretized points and the trajectory angle. Figure 10
shows the benchmark to be wound on the left and the corresponding control volume on
the right; in the same figure, the base path and the winding trajectories are presented. The
process parameters associated to twenty trajectories are reported in Table 2. Figures 11–16
show the shape of the winding trajectory as related to the value of the geometric
parameters. An increase in the safety distance removes the trajectory from the winding die,
while an increase in the number of discretized points associated with a greater trajectory
angle (
>90
) reduces the sudden changes in deposition system direction by making the
path more and more regular.

Table 1. Experimental plan.

Process variables Number of levels Values

, Trajectory angle [
] 2 90–100
n, Discretized points [#] 3 14–30–44
d, Safety distance [mm] 4 50–70–90–150
S, Robot speed [%] 3 50–75–100
Replications [#] 3
Manufactured benchmarks [#] 180

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1401

(a) (b)

Figure 10. Base path and winding trajectory with: the benchmark on the left (a) and the control volume on the
right (b).

Table 2. Winding trajectories.

Trajectory [#] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Points n [#] 14 14 14 14 14 14 30 30 30 30 30 30 30 44 44 44 44 44 44 44
Safety distance 50 50 70 70 90 150 50 50 70 70 90 90 150 50 50 70 70 90 90 150
d [mm]
Trajectory 90 100 90 100 90 90 90 100 90 100 90 100 90 90 100 90 100 90 100 90
angle
[
]

The obtained winding trajectory has been implemented by means of the robotized cell
for different values of the nominal speed (S). The winding time has been measured by a
chronometer for the different trajectories covered by the deposition system with different
values of speed. The obtained values of the winding time1 are shown in Table 3.

DISCUSSION

To evaluate the influence of geometrical parameters on winding time, nonparametric

tests (Kruskal Wallis and Mood’s median) have been adopted. By these tests, the length of
the trajectory perimeter has been proved to depend significantly on the safety distance,
while it seems not to vary by changing the number of discretized points and the trajectory
angle, as shown in Figure 17. Figure 18 shows how the discretized points (n) and trajectory
angle do not affect the trend of the trajectory length. Increasing the safety distance
value from 50 mm to 150 mm, the trajectory length increases significantly for all the
considered conditions. The twenty trajectories that are shown along the x-axis
correspond to the values of the process parameters reported in Table 2. The mathematical
relationship among the trajectory perimeter (indicated as p in mm) and the safety distance

1
This value is equal to the average winding time associated to 90 revolutions around the winding die.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1402 W. POLINI AND L. SORRENTINO
400
d=50mm d= 70mm
d=90mm d= 150mm
Benchmark
300

200

100

Z [mm]
0
−400 −300 −200 −100 0 100 200 300 400

−100

−200

−300

−400

X [mm]

Figure 11. Winding trajectories for n ¼ 14 and
¼ 90
.

400
d=50mm d=70mm
Benchmark
300

200

100
Z [mm]

0
−400 −300 −200 −100 0 100 200 300 400

−100

−200

−300

−400

X [mm]

Figure 12. Winding trajectories for n ¼ 14 and
¼ 100
.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1403
400
d=50mm d=70mm
d=90mm d=150mm
Benchmark
300

200

100

Z [mm]
0
-400 -300 -200 -100 0 100 200 300 400

-100

-200

-300

-400
X [mm]

Figure 13. Winding trajectories for n ¼ 30 and
¼ 90
.

400
d=50mm d=70mm
d=90mm Benchmark
300

200

100
Z [mm]

0
-400 -300 -200 -100 0 100 200 300 400

-100

-200

-300

-400
X [mm]

Figure 14. Winding trajectories for n ¼ 30 and
¼ 100
.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1404 W. POLINI AND L. SORRENTINO
400
d=50mm d=70mm
d=90mm d=150mm
Benchmark
300

200

100

Z [mm]
0
-400 -300 -200 -100 0 100 200 300 400

- 100

- 200

- 300

- 400

X [mm]

Figure 15. Winding trajectories for n ¼ 44 and
¼ 90
.

400
d=50mm d=70mm
d=90mm Benchmark
300

200

100
Z [mm]

0
-400 -300 -200 -100 0 100 200 300 400

-100

-200

-300

-400
X [mm]

Figure 16. Winding trajectories for n ¼ 44 and
¼ 100
.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1405

Table 3. Winding time measured for the different winding trajectories covered by the
deposition head with different values of speed.
n [#] d [mm]
[
] p [mm] tS¼100% [s] tS¼75% [s] tS¼50% [s]
Discretized Safety Trajectory Trajectory Winding Winding Winding
Trajectory points distance angle length time time time
1 14 50 90 1045 7.23 9.59 14.46
2 14 50 100 1109 7.67 10.25 15.29
3 14 70 90 1220 7.45 9.78 14.74
4 14 70 100 1200 8.17 10.95 16.37
5 14 90 90 1425 7.35 9.65 14.59
6 14 150 90 1887 7.64 10.25 15.28
7 30 50 90 1002 11.00 15.23 22.44
8 30 50 100 1069 11.50 15.60 23.59
9 30 70 90 1139 11.31 15.28 22.42
10 30 70 100 1238 11.32 14.86 22.56
11 30 90 90 1403 10.93 14.30 22.08
12 30 90 100 1419 11.40 14.91 23.00
13 30 150 90 1894 11.33 14.80 22.48
14 44 50 90 982 14.31 19.05 28.32
15 44 50 100 1121 14.10 18.93 27.57
16 44 70 90 1118 14.45 18.93 28.89
17 44 70 100 1312 14.00 18.44 28.30
18 44 90 90 1245 14.73 19.46 29.00
19 44 90 100 1518 13.36 17.70 26.71
20 44 150 90 2105 13.58 18.03 27.22

n [#] d [mm] teta [°]

1900

1700
p [mm]

1500

1300

1100

14 30 44 50 70 90 15
0 90 10
0

Figure 17. Main effects plot of trajectory length vs geometric parameters.

(indicated as d in mm) has been obtained by regression analysis (see Figure 19):

p ¼ 582:31 þ 9:14  d ð1Þ

An increase in the safety distance causes an increase in the length of the trajectory
perimeter.
The average value of the winding speed of the deposition system to cover its trajectory
(called Ws) has been calculated by dividing the perimeter length for the measured winding
time. This value depends significantly on the number of discretized points, the safety

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1406 W. POLINI AND L. SORRENTINO
2500

2000
Trajectory length p [mm]

1500

1000

50
0

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Trajectory [#]

Figure 18. Length of the 20 winding trajectories.

S= 75.7824 R-Sq = 94.0% R-Sq(adj) = 93.9%

2000
p [mm]

1500

1000

50 60 70 80 90 100 110 120 130 140 150

d [mm]

Figure 19. Regression equation of trajectory length vs safety distance.

distance, and the nominal speed of the deposition system (see Figure 20). An increase in
the number of discretized points involves an increase in the number of times the head
accelerates and decelerates; thus reducing the average value of winding speed. Differently,
an increase of the safety distance allows to increase the time interval between a transitory
and the next one; thus allowing to increase the speed between two points (Wsi). An
increase in nominal speed (S) causes an increase in winding speed according to linear
model.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1407
n [#] d [mm] teta [°] S [%]

140

120
Ws [mm/s]

100

80

60
14 30 44 50 70 90 15
0 90 10
0 50 75 10
0

Figure 20. Main effects plot of winding speed (Ws) vs geometric parameters.

300.00

S=100% S=75% S=50%

250.00
Winding speed "Ws" [mm/sec]

200.00

150.00

100.00

50.00

0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Trajectory [#]

Figure 21. Winding speed for the 20 considered trajectories.

Figure 21 shows how the nominal speed of the deposition system (S) does not affect the
trend of the winding speed due to the number of discretized points; passing from 14 to 44
discretized points, the speed Ws decreases significantly for all the considered conditions.
The twenty trajectories that are shown along the x-axis correspond to the values of the
process parameters reported in Table 2. The nominal speed amplifies or reduces this trend
by translating the trend up or down.
Finally, the winding time depends significantly on the number of discretized points and
on the robot linear speed (see Figure 22). An increase in the number of discretized points
causes an increase in the winding time, but at the same time it involves a more continuous
path of the deposition system with a small probability of loosening of tension. In fact, each

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1408 W. POLINI AND L. SORRENTINO
n [#] d [mm] teta [°] S [%]
22,0

19,5

t [s] 17,0

14,5

12,0

14 30 44 50 70 90 15
0 90 10
0 50 75 10
0

Figure 22. Main effects plot of winding time vs geometric parameters.

point of the winding trajectory implies a deceleration and a subsequent acceleration of

the deposition system; therefore, an increase in the number of discretized points involves
a longer path, where the deposition system accelerates and decelerates, and a larger time
to wind the workpiece. The dependence by the safety distance of both the trajectory length
and the nominal speed of deposition system does not influence the winding time since
it changes the trajectory length and the speed in the same direction: it increases or
decreases both the length and the speed. A simple equation has been identified by
regression analysis to express the relationship among the winding time, and both the
number of discretized points and the winding speed:

t ¼ 9:36 þ 0:31  n þ 1114  1=S ð2Þ

Once the value of the nominal speed (S), has been fixed, the winding time depends on the
number of discretized points as shown in Figure 23, where the time value varies from 5 s
to 15 s with the increase in the number of points from 14 to 44 with a nominal speed of
2 m/s (S ¼ 100%). The twenty trajectories that are shown along the x-axis correspond to
the values of the process parameters reported in Table 2.
We can conclude that to plan the winding trajectory, it is desirable to fix the value of the
trajectory angle >90
to preserve the constancy of the winding tension and to decide
the value of the optimal safety distance in order to avoid possible collisions. Then, it is
possible to choose the number of discretized points by solving the trade-off between the
need to avoid loosening of tension, that advises to increase this number, and the will to
reduce the manufacturing time, that advises to decrease this number. When the number of
discretized points is increased in order to keep the movement of the deposition system
continuous, it is possible to increase the winding speed in order to keep the value of the
winding time low.

CONCLUSIONS

The present work has shown how to plan the trajectory of the deposition system in
order to decrease the winding time, but at the same time to preserve the constancy of the
winding tension, thus assuring an acceptable quality of the manufactured composite parts.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1409
2500

Trajectory length "p" [mm] Winding time "t" [s*100]

2000

1500

1000

500

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Trajectory [#]

Figure 23. Trajectory length and winding time for the twenty trajectories at a winding speed of 2 m/s
(S ¼ 100%).

The trajectory angle should be >90
in order to avoid loosening of tension, while it seems
not to significantly influence the winding time. An increase in the number of the
discretized points may assure a more continuous movement of the deposition system
during winding and, therefore, avoids loosening of tension, but at the same time it may
sensibly increase the time needed to complete the winding. An increase in the safety
distance, when combined with suitable values of both the trajectory angle and the number
of discretized points, may favor a decrease in winding time, since even if it involves a
longer winding trajectory, it allows to increase the speed of the deposition system among
the different points of the trajectory.
A suitable compromise between part quality and winding time may address the choice
of the number of discretized points and where the result is characterized by many points,
it is possible to increase the winding speed in order to reduce the time.
Moreover, this work shows how a set of simple geometric parameters, i.e., the number
of discretized points, the safety distance, and the trajectory angle, may address the
planning of deposition system trajectory by taking into account both the tension
constancy and the winding time.
This work represents the first step toward the optimization of the algorithm to generate
both a short trajectory and a constant tension of the roving during winding.

ACKNOWLEDGMENTS

This work was carried out with the funding of the Italian M.I.U.R. (Ministry of
University and Research) and CNR (National Research Council of Italy). The authors are
grateful to Anagni Agusta-Westland, especially to Dr. F. Natalizia, Eng. E. Anamateros

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

1410 W. POLINI AND L. SORRENTINO

and G. Paris for supporting this work. Special thanks to Eng. A. Viperino and Mr. P. Held
for supporting the experiments.

NOMENCLATURE

n ¼ number of points used to discretize the winding trajectory [#]

¼ trajectory angle [
]
d ¼ safety distance [mm]
li ¼ length of winding trajectory joining

P two  consecutive points [mm]
p ¼ length of winding trajectory ¼ ni¼1 li [mm]
S* ¼ maximum value of robot linear speed [m/s]
S ¼ nominal speed of deposition system [%]
Wsi ¼ actual speed of deposition system between two points of winding trajectory
[mm/s]
P
Ws ¼ winding speed ¼ average actual speed of deposition system ¼ ni¼1 li 
Wsi =pÞ [mm/s]

REFERENCES

1. Chan, S., Munro, M. and Fahim, A. (1996). Accuracy-speed Relationships of a Robotic

Filament Winding Cell, Robotics and Computer-Integrated Manufacturing, 12(1): 3–13.
2. Imamura, T., Kuroiwa, T., Terashima, K. and Takemoto, H. (1999). Design and Tension
Control of Filament Winding System, In: IEEE International Conference on Systems, Man and
Cybernetics, 2, pp. 660–670.
3. Scholliers, J. and van Brussel, H. (1994). Computer-integrated Filament Winding: Computer-
integrated Design, Robotic Filament Winding and Robotic Quality Control, Composite
Manufacturing, 5(1): 15–23.
4. Scholliers, J. and van Brussel, H. (1996). Design and Off-line Programming of a Robotic Tape
Winding Cell, Robotics and Computer-Integrated Manufacturing, 12(1): 93–98.
5. Seereeram, S. and Wen, J.T.Y. (1991). An All-geodesic Algorithm for Filament Winding of a
T-shaped Form, IEEE Transactions on Industrial Electronics, 38(6): 484–490.
6. Liang, Y.D., Zou, Z.Q. and Luo, G., (1996). Non-Axisymmetric Filament Winding for Elbows
Based on the Quasi-Geodesics, In: Proceedings of International Sampe Technical Conference,
Vol. 29, pp. 366–374.
7. Liang, Y.D. and Luo, G. (1996). Two-dimensional Non-axisymmetric Filament Winding
Patterns, In: Proceedings of International Sampe Technical Conference, Vol. 29, pp. 505–514.
8. Liang, Y.D., Luo, G., Chen, Y. and Qiu, K.X. (1996). Graphical Representation and
Calculation of Interweaving Degrees of Filament Winding Patterns, In: Proceedings of
International Sampe Symposium and exhibition, Vol. 41, pp. 486–495.
9. Liang, Y.D., Shou, H.H. and Luo, G. (1997). On the Definition and Graphical Representation
of Non-axisymmetric Filament Winding Patterns, In: Proceedings of International Sampe
Symposium and Exhibition, Vol. 42, pp. 354–367.
10. Liang, Y.D. and Luo, G. (1999). The New Generation Computer Controlled Filament Winding
Machines Developed in China, In: Proceedings of International Sampe Symposium and
Exhibition, Vol. 1, No. 44, pp. 216–226.
11. Steiner, K.V. and Wiest, W. (1993). Algorithms for Computer-Aided Robotic Filament
Winding, In: Proceedings of International Sampe Technical Conference, Vol. 25, pp. 725–735.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014

Robotized Filament Winding of Asymmetric Shape Parts 1411

12. Carrino, L., Polini, W. and Sorrentino, L. (2004). Method to Evaluate Winding Trajectories in
Robotized Filament Winding, Journal of Composite Materials, Ed. Sage, 38(1): 41–56.
13. Carrino, L., Polini, W. and Sorrentino, L. A New Robotized Filament Winding Cell to
Manufacture Complex Shape Parts, SME Technical Papers 2003, ID: TP03PUB226, Paper No:
EM03-324.
14. Carrino, L., Polini, W. and Sorrentino, L. (2002). Experimental Validation of a New Fiber
Deposition Device for a Robotized Filament Winding Cell, In: ECCM10 ‘‘10th European
Conference on Composite Materials’’, Brugge, Belgium, June 3–7, Abstract 311.
15. Carrino, L., Polini, W. and Sorrentino, L. (2003). Modular Structure of a New Feed-Deposition
Head for a Robotized Filament Winding Cell, Composites Science and Technology , Ed. Elsevier
Science, Volume: 63, Issue: 15, November, pp. 2255–2263.
16. Carrino, L., Polini W. and Sorrentino, L. (2002). Design of a New Feed-deposition Head for
Robotized Filament Winding, In: Proceedings ASME ‘‘7th Annual Design for Manufacturing
Conference’’, Montreal, Canada, September 29 – October 2, DETC2002/DFM-34157.
17. Carrino, L., Landolfi, M., Moroni, G. and Di Vita, G. (1996). CAM for Robotized Filament
Winding, Advanced Manufacturing System and Technology, In: CISM Courses and Lectures, E.
Kulianic (Ed.), Springer Verlag, Vol. 372, pp. 601–608.
18. Carrino, L., Moroni, G. and Turchetta, S. (1998). CAD/CAM for Robotic Filament
Winding Process Design, In: European Conference on Composite Materials, Napoli, Italy,
pp. 615–622.
19. Carrino, L., Polini, W. and Turchetta, S. (2000). Computer Aided Filament Winding System,
Advancing with Composites, Milano, Italy, pp. 107–114.

Downloaded from jcm.sagepub.com at GEORGIAN COURT UNIV on November 3, 2014