Journal of Materials Processing Technology 229 (2016) 587–595

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Contribution of thermo-mechanical parameters and friction to the

bonding of thermoplastic tapes in the tape winding process
Mikhael Tannous a,∗ , Anaïs Barasinski a , Christophe Binetruy a , Benoît Courtemanche b
a
GéM, Ecole Centrale de Nantes, France
b
CETIM, Ingénierie Polymères & Composites, France

a r t i c l e i n f o a b s t r a c t

Article history: Ensuring a good inter-laminar bonding while accelerating the production in the tape winding process of
Received 6 June 2014 thermoplastic tapes is a key question. Finite element models and numerical simulation help to predict
Received in revised form 28 April 2015 the influence of process parameters on the product quality. This article addresses the finite element mod-
Accepted 9 October 2015
eling of the thermal and mechanical phenomena involved in the tape winding process of thermoplastic
Available online 23 October 2015
tapes. It highlights through simulation and experimental validations the influence and necessity to pro-
mote the friction between the different components on the product quality and the process itself. This
Keywords:
parameter is neglected in the literature which focuses mainly on thermal parameters and occasionally
Filament winding
Inter-laminar bonding
on the mechanical parameters (roller deformations, etc.), excluding the friction.
Friction role © 2015 Elsevier B.V. All rights reserved.
Finite element model
Experimental validation

1. Introduction
Tape winding has an increasing value because it reduces the
cost of manufacturing composite structures while preserving a
high quality as indicated by Sun et al. (1995) and eliminating the
autoclave needs as claimed by Barasinski et al. (2011). Advances
in robotics and manipulators design made the tape winding pro-
cess affordable even to small companies as mentioned by Koussios
et al. (2012). Tape winding technology is in continuous progress
and is optimized such a way that production is faster and the prod-
uct quality is enhanced as insisted by Schmitt and Witte (2012).
The common physical mechanism to all thermoplastic composite
processes is the fusion bonding that is responsible for the cohe-
sion between yarns and plies and is the main factor that leads to a
good product quality as described by Schell et al. (2009). The fusion
bonding, first of all, is a thermal mechanism extensively studied
in the literature because the laser assisted filament winding tech-
nology started with the fiber placement technology and consisted
to heat two layers of thermoplastic tapes and put them into con-
tact. In absence of roller pressure, or tape tension, obtaining a good
inter-laminar bonding reduces to a thermal problem. Therefore, the
heating source was a major concern in the literature. Modeling the
process required an appropriate modeling of the heating source.
∗ Corresponding author.
E-mail address: mikhael.tannous@ec-nantes.fr (M. Tannous).
There are many technologies available: Gennaro et al. (2011) used
an infrared heating source, Toso et al. (2004) used a Hot Gas Torch,
Stavrov and Bersee (2005) used an electromagnetic radiation and
more precisely a resistance welding, while Ageorges et al. (2001)
used friction as a heating source. Other heating sources exist and
are listed by Shih (1997): thermal conduction from an external
source such as contact with hot platens, microwave, radiofre-
quency, or laser (see Wang and Lou (2003) for example). The laser
heating was first discussed in Beyeler and Güçeri (1988) and later
adapted by Mazumdar and Hoa (1993). Laser-assisted fiber place-
ment machines were enhanced for mass production and for large
industrial products such as submarine structures studied by Sharp
et al. (1995), and is nowadays frequently used in the fusion bond-
ing process of thermoplastic composites as done in Casalino and
Ghorbel (2008). The interface temperature prediction was the main
key to get a correct model of the process, and controlling this tem-
perature became a strong objective: Tierney and Gillespie (2006)
used a closed loop control system to control the temperature of the
interface between the top deposited tape and the substrate in order
to achieve a good inter-laminar bonding quality.
However, the interface temperature could not be precisely pre-
dicted without accurately modeling the interply contact resistance.
The thermal conductance between two contacting bodies was
introduced a few decades ago. Cooper et al. (1968) is among the
first works in the literature to address this issue, which made there-
after the subject of many research works that aimed to describe it
theoretically and find an appropriate physical law as done by Mikic

http://dx.doi.org/10.1016/j.jmatprotec.2015.10.013
0924-0136/© 2015 Elsevier B.V. All rights reserved.
588 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Fig. 1. Laser assisted tape winding machine (AFPT).
(1972). Nowadays, interply thermal contact resistance is still inves-
tigated in the simulation and studies of the tape winding process
or the tape placement technology. In fact, interply thermal contact
resistance was the main concern in Barasinski et al. (2013) for the
study of an automated tape placement process.
With the introduction of the roller to the fiber placement tech-
nology, the roller pressure and the distribution of the pressure
became mechanical parameters that influence the bonding quality.
Rigid rollers were replaced by deformable ones for a better pressure
distribution. The effect of pressure on the welding of two compos-
ites ribbon was addressed by Hinkley et al. (1997) then the bonding
became a thermo-mechanical problem. The bonding is achieved
when a sufficient pressure is applied on two bonding tapes hav-
ing reached a given temperature for a sufficient period of time.
Achieving a good inter-laminar bonding quality with the lowest
temperature, pressure and exposure time were the main concern in
Schell et al. (2009). The only mechanical parameter that contributes
to the good bonding quality is the roller pressure and distribution.
Therefore, deformable rollers allow a better pressure distribution.
The influence of rotation and tape tension to the bonding qual-
ity should be considered. There are very few papers on the effect
of tape tension and winding kinematics on the bonding quality.
The consolidation pressure and winding speed for carbon thermo-
plastic composite were studied by Colton and Leach (1992). It was
found that high speed and consolidation pressure reduced void in
the product and thus enhanced the inter-laminar bonding quality.
However, the literature still focuses on the thermal problem as the
main key behind the interply defaults and the bad bonding that
occurs during the process.
This article deals with the laser-assisted winding of car-
bon thermoplastic tapes on a technologically advanced machine
(see Fig. 1) where the manufacturing process combines sev-
eral mechanical, kinematic and thermal parameters. It presents a
thermo-mechanical model in which all mechanical, kinematic and
thermal parameters are taken into account in order to evaluate the
quality of the adhesion between the deposited tapes. The friction
between the different components, and especially the roller/tape
and roller/mandrel friction are taken into account. Note that the
friction is considered as a pure mechanical parameter. The heat
generated by the friction in this study case, and for the considered
machine leads to a negligible temperature increase. Therefore, the
friction is not a thermal parameter as in Ageorges et al. (2001),
where friction is the heating source. Controlling pure mechanical
parameters and especially the friction for a better inter-laminar
bonding is, to the best of our knowledge, not adressed in the
literature. Studies today focus on the roller’s behavior and deforma-
tion, which affect the pressure distribution, and thus the literature
neglects the friction role.
This research work highlights via simulation and experimental
studies the importance of the friction.
The proposed approach relies on commercial finite element
software (Abaqus) and can be implemented by engineers with
a reasonable computational cost. Note that the computational
effort could be decreased by taking advantage of model reduction
methods such as the PGD (Proper Generalized Decomposition) for
solving the thermo-mechanical problem in Chinesta et al. (2014).
This article aims at highlighting the contribution of mechani-
cal parameters and more specifically, friction coefficients between
the roller, tape, and mandrel, on the inter-laminar bonding quality.
For that reason, no results concerning the thermal problem will be
shown. This later has been extensively addressed in the literature.
The article is structured as follow. The process is described in
a first part. Secondly, the production limits and defaults due to
neglecting the friction are presented. Then, the role and importance
of friction are assessed through experiments before developing a
thermo-mechanical finite element model to have a better under-
standing of the process and to predict the bonding quality with
respect to the different thermal, kinematic and mechanical param-
eters. Some conclusions are drawn. Future applications based on
the results of this research work are proposed.
2. The process and its different parameters
Fig. 1 shows the tape winding machine (Advanced Fiber Place-
ment Technology (AFPT)) used in this study. It is a robotized tape
winding machine that consists of a rotating mandrel on which ther-
moplastic tapes are placed by the robot. On the left hand side of
Fig. 1, the tapes are placed such a way they form a 45◦ angle with
respect to the rotation axis, while on the right hand side of Fig. 1
the tapes are placed at a 90◦ angle (circumferential winding). The
main applications and illustrations of this article focus on the cir-
cumferential winding.
The tapes are placed a pressure given by a roller and are sub-
jected to a controlled tension. A laser source is used to heat the
tapes up and the temperature is controlled by an infrared camera.
Fig. 3 illustrates a 90◦ filament winding process. Although the
application examples of this article concern mostly a cylindrical
mandrel of Rmandrel = 100.5 mm, the mandrel can have several forms
and diameters. Larger diameters are also of interest. The linear
placement velocity can vary in the [3, 30] m/min interval which
is equivalent to a [0.5, 5] rd/s rotational velocity on the 100.5 mm
mandrel radius. Polyamide thermoplastic tapes (Celstran CFR-TP
PA6 CF60-01) are considered, with carbon fiber occupying 60% of
the tapes mass and 48 % of the tapes volume. The tapes are 1.2 mm
wide and  = 0.19 mm thick and have the orthotropic characteristics
reported in Tables 1 and 2. Note that 1 indicates the carbon fiber
direction, 2 and 3 correspond to the two other directions orthogonal
to 1.
The tape is under tension and has an attack angle ˛tape = 45◦ . The
tape tension is ensured via a braking motor that imposes a force
opposite to the tape’s motion and depends on the winding veloc-
ity. It can be up to 300 N. The pressure in the roller’s actuators can
range from 0.1 MPa to 0.6 MPa. The roller itself is removable and can
be rigid (metal roller) or deformable (silicone roller). One can also
change the roller’s dimensions. However, the one used in this study
has a 40 mm external radius, a 20 mm internal radius and a 20 mm
width. The roller pressure can cause large deformation to the sili-
cone one. The later is used in this study. The silicone is an isotropic
material and has the characteristics listed in Table 3. An experimen-
tal test aiming at determining the Young Modulus of the silicone
roller is performed on the latter and the results are shown in Fig. 2.
It is obvious that the silicone has a non linear behavior for  ≥ 0.3.
However, during the winding process, the roller deformation does
 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595 589

Table 1
Mechanical characteristics of the tapes.

Density Young modulus (GPa) Shear modulus (GPa) Poisson ratio

 (kg/m3 ) E1 E2 = E3 G12 = G13 G23 12 = 13 23

1460 112 5.7 2.81 3.7 0.25 0.1

Table 2 not exceed this limit and thus a constant Young Modulus of 3 MPa
Thermal characteristics of the tapes.
is assigned to the silicone roller.
Specific heat (J/kg K) Conductivity (W/m K) The laser diode applies a power ranging from 500 W to 1500 W
at an angle ˛laser = 23◦ . This attack angle can be adapted. An infrared
C k11 k22 k33
camera monitors and controls the tape temperature and adjusts the
1400 5.556 0.338 0.338
laser power if needed.
The main challenge in tape winding modeling is defining the
different parameters leading to a good inter-laminar bonding while
increasing the production efficiency. One cannot increase the rota-
tional velocity of the mandrel without adapting the laser power.
The roller pressure and the tape tension must be adapted as shown
in this paper.

3. Production defaults: description and possible causes

Fig. 4 shows the cross-section of a ±  thermoplastic compos-

ite tube made by laser assisted tape winding. One can observe air
pockets, of different sizes indicating a poor inter-laminar bonding.
In the tape winding process, users consider the tape tension as the

Fig. 2. The stress versus the strain curve of the silicone roller.

Fig. 3. Schematic and notations of the circumferential winding process.

Fig. 4. Examples of bad inter-laminar bondings.

Table 3
Mechanical and thermal characteristics of the silicone roller.

Density Young modulus Poisson ratio Specific heat Conductivity

1090 (kg/m3 ) 3 (MPa) 0.25 1400 (J/kg K) 0.27 (W/m K)

590 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Fig. 5. Interest of friction for a ± orientation tape winding.
main mechanical parameter and do not give a special care to the
roller pressure and deformation. They tend to increase the tension
for a better inter-laminar bonding, thus forcing the tape to stick to
the newly deposited tape layers. As long as circumferential wind-
ing is considered, increasing the tape tension has no detrimental
effects as far as the required tension does not lead to tape damage.
However, too much tape tension can lead to defects as those
found Fig. 4, for example, for dome windings or winding at an ori-
ented angle ± as shown in Figs. 5 and 6. In that cases a high tension
T will increase Ts the projection of the tangential force Tp in the
orthogonal direction to the trajectory. These configurations are not
achievable with a metallic roller, and a silicone one must be used.
It is not the roller deformation that contributes to a better inter-
laminar bonding when the silicone roller is used but the increased
friction forces induced but a higher friction coefficient due to the
silicone roller. Moreover, increasing the tape tension in such cases
reduces the inter-laminar bonding quality.
In order to support this thesis, let us consider a ± tape winding
process as the one illustrated in Fig. 5. The tape tension T is applied
according to the tape’s direction. A  angle exists between the ten-
sion’s direction and the main axis of rotation of the mandrel. The
tape tension is reported to the roller to mandrel contact zone and
possesses three main components: N, Tn and Ts . (Tn , Ts ) are two
orthogonal vectors of (Ptang ), which represents the plan that is tan-
gent to both the mandrel and the roller. N is orthogonal to (Ptang ).
The tape tension in the roller to mandrel contact zone reads:
T = Tn + Ts + N
N (in the case of Fig. 5) and T keeps the tape in contact with
the mandrel. However, Ts drives the tape to slide on the man-
drel and generates spurious strains in the newly bounded zone
and can even lead to tape stick up. Obviously, this cannot be
Fig. 6. Interest of friction for spherical geometries.
compensated by increasing the tape tension as was the case in the
circumferential tape winding example. Increasing the tape tension
contributes to increasing Ts and thus creates spurious stresses in
the newly bonded area and may lead to defects. The roller/tape
friction compensates this effect and prevents the tape from mov-
ing in the direction of Ts , i.e., parallel to the main rotational axis of
the mandrel.
The same problem is encountered when winding doubly curved
shapes, as the domes of a cylindrical tank winded by Zu et al. (2010).
Fig. 6 illustrates the winding of a dome. The tape tension when
reported to the newly bonded area presents, similarly to the case
of Fig. 5, a normal component which is not presented for sake of
simplicity, and two components in the plan tangent to both the
roller and the mandrel. Ts causes a sliding between the two tape
layers. Only the roller/tape friction can compensate this effect.
4. Experimental investigation of the roller to tape friction
This section assesses the friction role and supports the hypothe-
ses and conclusions of the previous sections through simple
experimental tests.
4.1. Quantifying the roller to tape friction
The main objective of this experimental test is to highlight the
presence of the friction between the tape and the roller, denoted
froller/tape thereafter. Tape to mandrel friction and roller to mandrel
friction are denoted ftape/mandrel and froller/mandrel , respectively. This
latter depends on whether the direct contact between the roller
and the mandrel exists or not. A compliant roller is prone to large
deformations and can be in contact with the mandrel.
Fig. 7 schematizes the experimental test. The roller is subjected
to a controlled pressure P. The rotation of the mandrel around its
main rotational axis is blocked. The tape, that passes between the
mandrel and the roller, has one free end and the other one is sub-
jected to a tension T. This later is increased from zero to the value
at which the tape starts sliding. At this moment, the tape tension is
constant and its value is measured with a force sensor. The higher
this tension is, the higher global the friction coefficient will be. As
described by the Coulomb friction law, in Eq. (2), the tape tension
(1)
Fig. 7. Experimental set up to highlight effect of mandrel/tape friction and
tape/roller friction in the winding process. In that set up, the mandrel is held motion-
less while the roller is free to rotate when the tape is being pulled.
 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595 591

Table 4 Table 5
Quantifying the roller to tape friction. Friction role experimental tests.

Roller Steel Silicone P (MPa) ω (m/min) L0 (cm) T (N) Lf (cm) Lf − L0

P (MPa) 0.3 0.6 0.1 0.3 0.6 Silicone roller

T (N) 47 75 80 160 220 Test 1 0.3 3 32.8 Tmin 152 119.2
Test 2 0.3 3 27 Tmax 149 122

Steel roller
Test 3 0.3 3 4.5 Tmin 18.4 13.9
T is proportional to the normal contact forces F applied to the tape. Test 4 0.3 30 4.5 Tmin 19.3 14.8
These forces result from the roller’s pressure P. Test5 0.3 30 4.7 Tmax 4.7 ≈0

T = F
Table 4 reports the tension values with respect to the applied
pressure and the roller type. When the steel roller is used, small
tension is needed to set the tape’s motion. This is due to the small
friction coefficient between the roller and the tape, and the small
contact surface due to the high roller stiffness.1 When the silicone
roller is used, one can note the high tension required to pull the
tape. This later is generated by the friction and may, in the case of
a tape winding process, be larger than the tape tension.
This simple experimental tests validates the need to take the
friction into account in the thermo-mechanical model of the pro-
cess, as done and described in Sections 5.1 and 5.2.
4.2. The friction role
Fig. 8 shows a simple experimental test performed on the wind-
ing machine of Fig. 1. The left hand side of Fig. 8 shows a tape having
one free end, while the second is subjected to tension T. The tape
passes between the roller and the rotating mandrel. At t = 0, the dis-
tance between the free end of the tape and the point of contact with
the roller and the mandrel is L0 . If the friction between the roller and
the tape is negligible, the tension in the tape will cause this later to
move in the same tension direction. In other words, Lf ≤ L0 after a 
(rd) rotation of the mandrel. However, with a sufficient friction the
tape will be driven by the mandrel rotation in the direction oppo-
site to the tape tension and thus Lf ≥ L0 after a  (rd) rotation of the
mandrel. If no sliding at all occurs, then Lf − L0 =  × Rchuck . In this
case, the tape tension is dominated by the friction forces.
In the following, the same cylindrical mandrel of radius Rmandrel
= 100.5 mm is considered and two cylindrical rotations are per-
formed for each test case. If no sliding occurs, the total unrolled
tape length is:
Lf − L0 = (2 ×  × R) × 2 = 126.3 cm
(2)
Table 5 reports several experimental results of tests according
to the basics of Fig. 8. A silicone and a steel roller are considered.
The silicone roller is deformable and the friction coefficient with
the tape and the mandrel is high, giving rise to high friction forces.
The steel roller is rigid and has a much lower friction coefficient
with the tape and the mandrel. The exact values of the tape tension
are not known. However, the winding machine imposes the tape
tension once the tape moves. The minimum tape tension is about
20 N and the maximum reaches about 300 N.
Results in Table 5 show that the silicone roller holds firmly the
tape and makes it follow the mandrel rotations with minor sliding.
Increasing the tape tension does not affect the results. When the
steel roller is used, with much lower friction forces, the slipping
rate is very high. This later seems not to depend on the rotational
velocity as shown by tests 3 and 4. However, for a high tape tension
and low friction (test 5), the tape does not move at all.
These tests highlight the contribution of the friction forces to the
tape winding process. The tape tension is a much less influencing
parameter.
5. Numerical simulations
A numerical simulation model is developed for a better under-
standing of the process, and to predict the influence of the different
process parameters on the the bonding quality. The results confirm
the conclusions drawn from the experiments.
5.1. A thermo-mechanical finite element model of the process
As previously mentioned in the introduction, most of the lit-
erature focuses on the influence of the thermal parameters on the
bonding quality. The roller role is limited to apply and distribute the
pressure to ensure intimate contact between tapes. When the roller
is deformable, a better pressure distribution is achieved. Therefore,
(3) the literature focuses on the thermal parameters and pressure dis-
tribution to simulate and assess the inter-laminar bonding quality.
The behavior of the roller and its interaction with the tape is limited
to the pressure distribution and no friction is considered. In the fol-
lowing, the proposed finite element model includes the mechanical
and thermal interactions and friction between the roller and the
tape, the roller and the mandrel as well as between the two bonding
tapes. Friction modeling shows to be relevant and even necessary
for well describing the winding process. Section 4.1 presents an
experimental test that justifies the need for taking the friction into
account in the thermo-mechanical model of the process.
The mechanical and thermal phenomena concerns each part of
the system (roller, tape or mandrel) and are described by Eqs. (4)
and (5) respectively:

MÜ + CU̇ + KU + Fcont = F (4)

1
Deformable roller with frictionless contact will lead to zero tension. It is the
Fig. 8. Experimental highlighting of the roller to tape friction role. friction forces that are highlighted in such a test.
592 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Fig. 9. Contact conductance.
∂T
C(T) − ∇ .(K(T)∇ T) = Q + Qi
∂t
M, C and K denote respectively the mass, damping and stiffness
matrices of the given part of the system (roller, tape or mandrel).
U, U̇ and Ü represent respectively the displacements, velocities and
accelerations vectors of that same part. F is the external load and
Fcont the contact forces. It is obvious that the tape winding process
from a mechanical point of view involves three unconnected parts.
These parts interact via the contact and friction forces only. The
tape winding process is, therefore, mainly influenced by the way
contacts develop.
Eq. (4) describes the tape winding process from a dynamic
point of view. The tapes have an orthotropic elastic behav-
ior, while the roller has an homogenous elastic behavior. Large
displacements/rotations are considered with small deformations
hypothesis.
Eq. (5) describes the energy balance of the winding process. T is
the temperature vector at nodes. C(T) and K(T) represent respec-
tively the heat capacity matrix and the conductivity matrix, both
are temperature dependent. Since the transient stage is not of con-
cern here, the main objective is to study the steady state regime,
hence constant capacity and conductance matrices are considered.
Q and Qi represent the external and internal heat sources, respec-
tively. The external heat source is the laser heating device and the
main internal heat source is the friction. denotes the fraction of
contact energy transformed to heat, and thus the heat energy due
to friction and denoted Qf can be expressed by:
s cally under the external vertical pressure. The roller is mounted on
Qf =
t
where is the frictional stress that depends on the contact pres-
sure, the friction coefficient  and the temperature on either sides
of the surface. The effects of radiation and convection are found to
be negligible. This is often the case on laser assisted tape winding
machines (cf. Grouve (2012)). The heat flux is transferred by con-
ductance through each material (roller, tape, mandrel) and from
one material to the other. At each contact interface between two
materials, a contact conductance law is needed. Actually, as illus-
trated in Fig. 9, the contact interface between two bodies is never
perfect. This induces a given resistance Rc to the heat flux between
the two bodies as described by Jeevanashankara et al. (1990) and
formulated in Eq. (6). This contact resistance Rc depends on several
factors, but mostly the surface roughness.
1
Qc = T (6)
Rc
Eq. (5) is written to each part involved in the process. Heat
transfer occur between the different parts and are dominated by
the contact resistance between the contacting surfaces. The ther-
mal problem, like the mechanical problem, is mainly affected by
contacts between the moving parts.
Moreover, in the bonding area the contact conductance
increases when the two layers get closer. In this article and as the
results of Section 5.3 show, the role of mechanical parameters and
mostly the influence of friction to the inter-laminar bonding qual-
ity is highlighted. For this reason, the contact conductance is not
the main focus of this study and a simplified approach is used to
describe it.
The coupled thermo-mechanical problem means that the tem-
perature influences the stress solution, and this later influences the
temperature distribution. In such a case, the two problems should
be solved at each time step simultaneously. The coupled non lin-
ear problem is solved via Newton’s method and leads to the non
symmetric system:
    
(5) KUU KUT U RU
= (7)
KTU KTT T RT
U and T are the respective corrections to the incremental
displacement and temperature vectors, and RU and RT are the
respective mechanical and thermal residual vectors. KUU , KUT , KTU
and KTT are sub-matrices of the fully coupled non symmetric Jaco-
bian matrix. Note that when the coupling is weak, one may consider
KUT = 0 and KTU = 0.
This model is simulated on the commercial finite element soft-
ware Abaqus as described in Section 5.2.
5.2. Simulation on Abaqus
In this section, a finite elements model of the tape winding
process described is presented in Section 2 and Fig. 3. The main
objective of the simulation is to predict the inter-laminar bond-
ing quality with respect to the different mechanical and thermal
parameters of the process. As previously mentioned, a good bond-
ing quality between two tapes at a given point is achieved when this
point reaches a minimum temperature during a sufficient period of
time and under a minimum pressure are applied (see Schell et al.
(2009) for details).
Fig. 10 shows the finite elements model on Abaqus of the model
described in Fig. 3. The model is composed of three main compo-
nents: the roller, the tape and the mandrel. The roller is free to
rotate around its main rotational axis and free to translate verti-
a rigid metal axis providing a non deformable internal surface. This
is reproduced in the simulation by imposing a MPC-Beam (multi-
ple point constraint of beam type) between a node belonging to the
beam representing the rotational axis of the roller, and the inter-
nal roller surface. The roller itself is deformable. It is meshed with
3D quadratic elements. The main interest of the 3D elements is to
well describe the contact and avoid faulty interpenetrations (see
Fig. 11).
The two other components of the model, i.e., the tape and the
mandrel, are meshed with 3D shell elements. One end of the tape is
tied to the mandrel, and the other one is forced to follow a trajectory
inclined of 45◦ with respect to the horizon to simulate a 45◦ tape
placement. This same end is subjected to tape tension.
 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Fig. 10. Abaqus finite elements thermo-mechanical model.
Fig. 11. Interest of quadratic elements.
The mandrel is supposed to be infinitely rigid. When the simu-
lation begins, it is supposed that 10 tape layers are already placed
on the mandrel. The mandrel and the 10 tape layers are modeled by
a cylindrical shell. The later is rigid, but has the thermal properties
(thermal conductivity and capacity, etc.) of the tapes. The bottom
side of the shell represents the mandrel which is considered rigid
and is assigned a constant temperature (20◦ ) during the process.
The tape winding process is a stationary thermo-mechanical
dynamic problem. However, the simulation model passes through
a transient phase. The rotational velocity, roller pressure and
tape tension should be progressively established to avoid artifi-
cial numerical oscillations. For that reason the tape tension and the
rotational velocity are increased linearly along a Tt period that rep-
resents 1% of the simulation period. However, to accurately take
into account the roller deformation due to the applied pressure,
a static step should precede the dynamic problem. That way the
roller does not perform parasite oscillations or rebounds that may
lead to miscalculations.
Simulating the tape winding process cannot be achieved if
the interactions are not suitably modeled. The contact has three
components: normal contact and tangential contacts and thermal
contact. The normal contact modeling makes the tape take the form
of the mandrel and enables the simulation of the roller’s defor-
mation. The thermal contact modeling enables to model the heat
transfer between the interacting bodies. The heat transfer in the
heating zone is necessary to determine whether the temperature at
the heating location reached the value required for an inter-laminar
adhesion. Therefore, the good bonding quality, from a thermal point
of view is highly dependent on the thermal interactions model-
ing. The third component of the contact is the friction. This later
has not appeared yet in the literature. The friction leads to local
heating. However, this remains negligible in our study cases since
the rotational velocities are small. The main role of the friction
force is its mechanical contribution. The normal and tangential con-
tact (friction) are dealt with Lagrange multipliers that guaranty no
593
Fig. 12. Comparison of the external forces applied on the tape and its behavior at t
= 0.051 s with or without friction modeling.
interpenetration (cf. Belytschko et al. (2000)) with a master-slave
formulation. Moreover, implicit integration techniques are used to
solve the coupled thermo-mechanical problem.
5.3. Simulation results and conclusions
Multiple study cases and parameters values have been investi-
gated. However, for sake of clarity, only the results of simulations
that provided novel conclusions and contributions are shown. In
fact, defaults occur when a given combination of mechanical and
thermal parameters is not suitable for obtaining a good inter-
laminar bonding. This study focuses mainly on the mechanical
parameters. In the following, a combination of the mechanical
parameters that leads to a poor inter-laminar bonding is presented.
The later was mostly affected by the combination of two mechani-
cal parameters: the tape tension, and the friction. The friction being
the main error source as the following example case shows.
Let us consider the problem described in Section 2 and solved
for a 0.2 (s) simulation time. The tape winding is performed with a
small tape tension (30 N), a 0.35 MPa roller pressure and a negligible
friction between the different elements. The rotational velocity of
the mandrel is fixed to 5 rd/s.
The simulation results show that in this case, the roller is not
actually driven by the mandrel nor the tape, since this motion can
not be achieved without friction. However, this is not an issue since
its effect on the adhesion quality is negligible. The main drawback
of negligible friction modeling is that the tape performs random
motion and does not bond to the mandrel, as illustrated in Fig. 12.
This figure shows the comparison between the tape behavior in
case of a frictionless contact problem (Fig. 12a) and in case of a
 = 0.2 Coulomb friction (Fig. 12b). The results are presented at
594 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Fig. 13. Interest of friction in concave geometries.
t = 0.051 s. It is obvious that the tape sticks up from the mandrel at
the newly bounded area. At this point, the frictionless contact gen-
erates reverse forces that can be seen on the tape in Fig. 12b. These
reverse forces are not compensated by the friction forces, which
prevent tape’s free motions, and thus the tape sticks up from the
mandrel as Fig. 12a shows, in spite of a high compaction. Comparing
the external forces applied to the tape, in case of a frictionless con-
tact (Fig. 12a) and that of a contact with friction (Fig. 12b), shows
that the contact forces dominate and the reverse forces are absent
or negligible, when the friction is taken into account.
Note that the contact is not considered adhesive in order not to
hide the tape’s tendency to stick up from the mandrel in the absence
of friction. However, the tape should not stick up since bonding
already occurred when the two tape layers were heated and put in
contact under pressure. Although usually it is true, one has to bear
in mind that the reverse forces will lead to parasite stresses in the
newly bonded layers responsible of defaults and poor inter-laminar
bonding.
From a practical point of view, it is not possible to have a negli-
gible friction between the roller and the tape (or the roller and the
mandrel) as the roller is made from silicone material. Therefore, not
modeling the friction leads to a non-realistic model and wrong con-
clusions. A realistic tape winding process simulation should include
the friction phenomena.
This study case highlights also the physical importance of the
friction. The froller/tape (and froller/mandrel ) friction coefficient can be
low in the case of a metal roller, leading to small friction forces. In
addition, due to the limited deformation of the metallic roller, the
tape to roller contact surface is also small.
Modeling the friction eliminates the errors shown in
Figs. 12a and 13. The froller/mandrel 2 as well as the froller/tape make the
roller rotate in the opposite direction as that of the mandrel. The
froller/mandrel and ftape/mandrel are the two main forces that drive the
tape. They eliminate its free displacement and are like pliers that
hold the tape firmly and make it follow the mandrel movement.
Thus, the tape tension is not a must if high friction is available.
Since high friction depends on the roller material and contact
surface, then in the case of a silicone roller, the friction forces are
not negligible because the friction coefficient is important, and
the roller deformation, especially when the roller pressure is high,
ensures high contact surface. However, it is the friction and not the
roller deformation that attributes this property. Even a high roller
deformation without friction can still lead to tape random motion
as shown in Fig. 12a.
In few words, the simulations results and the experimental tests
lead to the same conclusion: the friction helps holding the tape
2
The roller width is higher than the tape’s width. Thus, the roller is in contact
with the mandrel.
firmly between the roller and the mandrel and makes it follow
the trajectory which is assigned to it. This results in a better
inter-laminar bonding quality by avoiding (even infinitesimal) tape
sliding and subsequently air pockets formation.
In the study cases shown in this article, the tape sticking up illus-
trated in Fig. 12a can be eliminated by increasing the tape tension.
In fact, increasing the tape tension in the simulation to its max-
imum (300 N) forced the tape to stick to the mandrel. Increasing
the roller pressure and deformation can not adjust such a situation.
One can, therefore, propose to neglect the friction modeling and
increase the tape tension. The drawbacks of such a simplified mod-
eling technique is overestimating the tape tension required for a
good inter-laminar bonding. Excessive tape tension and low fric-
tion are the primary factors yielding to significant defaults shown
in Fig. 4. Moreover, and as explained in Section 3, tape tension can
not compensate a lack of friction for winding domes and ±  tape
winding applications.
6. Importance of the friction for future industrial
applications
In case of a high frictional contact, tape winding can be
performed without or with a small tape tension. This may be inter-
esting, for example, for winding the internal corners of tanks and
for shapes such the one presented in Fig. 13. In fact, the tape tension
creates sticking up forces that can lead to local defects. The friction
forces are sufficient to making the tape follow the roller’s trajectory
and are the main key of success for such applications. Although the
shape presented in Fig. 13 is not yet frequent in industrial appli-
cations, tape winding applications are in constant expansion and
with the advances in robots technology, such applications are not
far from being the next generation of tape winding applications.
7. Conclusions
In this article, a finite element model is proposed for the
simulation of the thermo-mechanical tape winding process. The
simulation, performed on Abaqus, takes into account the different
thermal, kinematic and mechanical parameters that are involved
in the process. The simulation and the experimental tests show
that mechanical factors and mostly the friction play an important
role in the process. In fact, the tape is well hold between the roller
and the mandrel, thanks to the friction forces. This makes the tape
follow the mandrel and roller motion, and thus execute the tra-
jectory that is assigned to it without creating parasite forces that
affect the bonding quality and may even lead to the debonding of
the newly bonded zone in the deposited tape. The friction is a phys-
ical phenomenon that is a must to the winding of semi-spherical
shapes, concave shapes as well as ± tape winding processes, where
the loss of friction can not be compensated by the tape tension.
Whereas the literature attributes the good inter-laminar bonding
quality only to the thermal parameters and roller’s pressure distri-
bution, this study highlights the need to a wider view of the process
and to modeling the friction between the different elements for a
better understanding of the tape winding process and production
quality.
References
Ageorges, C., Ye, L., Hou, M., 2001. Advances in fusion bonding techniques for joining
thermoplastic matrix composites: a review. Composites A 32, 839–857.
Barasinski, A., Leygue, A., Soccard, E., Poitou, A., 2011. An improvement in thermal
modelling of automated tape placement process. In: International Conference on
Advances in Materials and Processing Technologies (AMPT2010). AIP Conference
Proceedings, vol. 1315, pp. 185–190.
Barasinski, A., Leygue, A., Soccard, E., Poitou, A., 2013. Identification of non uniform
thermal contact resistance in automated tape placement process. Int. J. Mater.
Form., 1–8.
 M. Tannous et al. / Journal of Materials Processing Technology 229 (2016) 587–595
Belytschko, T., Liu, W., Moran, B., 2000. Nonlinear Finite Elements for Continua and
Structures. John Wiley and Sons.
Beyeler, E., Güçeri, S., 1988. Thermal analysis of laser-assisted thermoplastic-matrix
composite tape consolidation. ASME J. Heat Transf. 110, 430–524.
Casalino, G., Ghorbel, E., 2008. Numerical model of co2 laser welding of thermoplastic
polymers. J. Mater. Process. Technol. 207, 63–71.
Chinesta, F., Leygue, A., Bognet, B., Ghnatios, C., Poulhaon, F., Bordeu, F., Barasinski, A.,
Poitou, A., Chatel, S., Maison-Le-Poec, S., 2014. First steps towards an advanced
simulation of composites manufacturing by automated tape placement. Int. J.
Mater. Form. 7, 81–92.
Colton, J., Leach, D., 1992. Processing parameters for filament winding thick-section
peek/carbon fiber composites. Polym. Compos. 13, 427–434.
Cooper, M., Mikic, B., Yovanovich, M., 1968. Thermal contact conductance. Int. J. Heat
Mass Transf. 12, 279–300.
Gennaro, R., Montagna, F., Maffezzoli, A., Fracasso, F., Fracasso, S., 2011. On-line con-
solidation of commingled polypropylene/glass roving during filament winding.
J. Thermoplast. Compos. Mater. 24, 789–804.
Grouve, W., 2012. Weld Strength of Laser-assisted Tape-placed Thermoplastic Com-
posites. (Ph.D. thesis), University of Twente, Enschede, the Netherlands.
Hinkley, J., Messier, B., Marchello, J., 1997. Effect of Pressure in Thermoplastic Ribbon
Thermal Welding. Tech. Rep. NASA Langley Research Center.
Jeevanashankara, C., Madhusudana, C., Kulkari, M., 1990. Thermal contact conduc-
tances of metallic contacts at low loads. Appl. Energy 35, 151–164.
Koussios, S., Zu, L., Wentzel, C., Elswijk, M., 2012. Filament winding: process
overview and novel developments. In: International SAMPE Technical Confer-
ence, Baltimore.
Mazumdar, S., Hoa, S., 1993. Experimental determination of process parameters
for laser assisted processing of peek/carbon thermoplastic composites. In: 38th
International SAMPE Symposium, pp. 189–203.
595
Mikic, B., 1972. Thermal contact conductance; theoretical considerations. Int. J. Heat
Mass Transf. 17, 205–214.
Schell, J., Guilleminot, J., Binetruy, C., Krawczak, P., 2009. Computational and exper-
imental analysis of fusion bonding in thermoplastic composites: influence of
process parameters. J. Mater. Process. Technol. 209, 5211–5219.
Schmitt, R., Witte, A., 2012. Control of a thermoplastic tape winding process with
optical in-line metrology. In: Proceedings in Manufacturing Systems, ISSN 7, pp.
2067–2923.
Sharp, R., Holmes, S., Woodall, C., 1995. Material selection/fabrication issues for
thermoplastic fiber placement. J. Thermoplast. Compos. Mater. 8, 2–14.
Shih, P.-J., 1997. On-line Consolidation of Thermoplastic Composites. (Ph.D. thesis),
Virginia Polytechnic Institute and State University.
Stavrov, D., Bersee, H., 2005. Resistance welding of thermoplastic composites – an
overview. Composites A 36, 39–54.
Sun, L., Mantell, S., Cohen, D., 1995. Thermoset winding process model and para-
metric study. Proc. ASME Mater. Div. 69, 11–23.
Tierney, J., Gillespie, J., 2006. Modeling of in situ strength development for
the thermoplastic composite tow placement process. J. Compos. Mater. 40,
1487–1506.
Toso, Y., Ermanni, P., Poulikakos, D., 2004. Thermal phenomena in fiber-reinforced
thermoplastic tape winding process: computational simulations and experi-
mental validations. J. Compos. Mater. 38, 107–135.
Wang, X., Lou, D., 2003. Simple thermal model for online laser curing of thermoset
composites in filament winding. Int. J. Energy Res. 27, 377–388.
Zu, L., Koussios, S., Beukers, A., 2010. Design of filament-wound domes based
on continuum theory and non-geodesic roving trajectories. Composites A 41,
1312–1320.