Analysis of in-situ

Thermal
Thermoplastic Composite Tape Laying
M. N. GHASEMI NEJHAD, R. D. COPE AND S. I. GÜÇERI
Department of Mechanical Engineering
and Center for Composite Materials
University of Delaware
Newark, DE 19716, USA
(Received December 3, 1990)

ABSTRACT: This paper presents analytical and numerical thermal analysis for melting
and consolidating impregnated composite tapes in the presence of a localized heat source.
This analysis also leads to the prediction of the processing window for a given tape-laying
configuration. Heat of melting/solidification is included in the form of a heat generation
term. A separation of variables method is employed to solve the governing equations ana-
lytically. In the numerical analysis, the governing equations are discretized using a non-
uniform mesh and are solved using a finite difference approach. The processing parame-
ters, such as consolidation speed, heat intensity, heat source width, etc., as well as
material properties are incorporated within the analysis. The results show large thermal
gradients in the vicinity of the consolidation point. The error between the analytical solu-
tion and the numerical result is found to be 3 % for the maximum temperature, and the
maximum error for the temperature over the entire domain is observed to be 7%. The
effects of processing speed, heat intensity, and the width of the local heat source are inves-
tigated, and the overall optimization of the process is discussed.

1. INTRODUCTION

&dquo;N’% ECENT DEVELOPMENTS IN high-temperature, advanced thermoplastic resins

ave opened new opportunities for their application in thermoplastic com-
posites for producing high-performance parts. The advantages of thermoplastic
composites over those based on thermoset resins are well known and include
superior fracture toughness, infinite shelf life and easier processibility. An ex-
’

cellent review of the current status of thermoplastic composites is presented by

Chang and Lees [1]. In thermoplastic-matrix composites, the research issues
center on the development of new material systems and the development of pro-
cessing techniques to produce parts in a cost-effective way. The processing of
thermoplastic composites differs completely from that of thermoset composites
in that it is based on the melting and solidification of the matrix which is based
on the transport of energy, often in the form of heat [1-4].
The tape-laying process for thermoplastic composites is a manufacturing

20 Journal of THERMOPLASTIC COMPOSITE MATERIALS, Vol. 4-,lanuary 1991

0892-7057/91/01 0020-26 $600/0
@ 1991 Technomic Publislung Co., Inc.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 21

method that appears to be suitable for producing parts with large surfaces and
mild curvatures, like those seen in aircraft wing skins. It offers the possibility of
combining the laydown, melting and consolidation steps in a continuous process,
thereby eliminating the curing steps that are necessary with thermosetting
materials. Also, whereas producing thick cross-section parts is very difficult us-
ing thermoset resins because of their exothermic reaction and heat diffusion limi-
tations, there is conceptually no limitation on the thickness of an in-situ consoli-
dated thermoplastic part. This is due to localized melting and solidification
during consolidation that prevents the build up of large residual stresses from the
high temperature gradients in thermoset curing or the large volumetric changes
in post consolidation of thick-section thermoplastic parts.
One of the major characteristics of composite materials is the strong coupling
between their microstructure and processing history. In semicrystalline ther-
moplastics, this coupling occurs most significantly in the final crystallinity of the
resin, which can range from low percentages in quenching-like processing condi-
tions, to more than forty percent for processes in which the rate of heat transfer
is very slow. General treatment of this subject is reported by Seferis [5], Velisaris
and Seferis [6], and Blundell et al. [7]. Recent studies by Motz and Schultz [8,9]
present an excellent analysis of crystal morphology and kinetics for Polyether-
etherketone (PEEK).
In addition to crystallimty, the prediction of residual stresses during processing
of thermoplastic composites also largely depends on the information about their
thermal history. In an in-situ thermoplastic tape-laying process, a localized heat
source generates a molten region around the laydown point. The prepreg tow is
consolidated through a melting diffusion process under the pressure of a consoli-
dation roller. The solidification of the material occurs as a result of heat removal
as it consolidates and moves away from the laydown point. The temperature field
is altered with the variation of the processing parameters, such as processing
speed, heat input, size of the heater, thermal boundary conditions. A parametric
study can be used to determine the processing window that indicates the bounds
on the heat flux at any given manufacturing speed and configuration.
In the current study, analytical and numerical thermal analysis for melting and
consolidation taking place during thermoplastic tape laying is presented. A con-
tinuous manufacturing process (in-situ) in the presence of a localized heat source
is considered. The primary objective of this study is to identify the dominant
processing parameters and observe their effects on the thermal history, as well as
to determine the overall feasibility of in-situ thermoplastic tape laying. An analyt-
ical model and a numerical technique are presented that allow for the prediction
of the temperature fields in the anisotropic composite. The results are assembled
in a way to develop a processing window that can be applied to a wide range of
material and processing configurations. The analytical solutions are compared
with the results obtained using the numerical method to assess the accuracy in
computations.
2. THERMAL ANALYSIS
A schematic description of a typical in-situ tape-laying process is depicted in

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

22

Figure 1. The geometry and Eulerian control volume (ECV) for a tape laying process.

Figure 1. The localized heat source of finite width is applied at the consolidation
point, and the tape lay-up is achieved by moving either the heat source or the
composite substrate. A descriptive temperature distribution around the consoli-
dation zone is also indicated in this figure. In a realistic tape-laying process, the
laminate is relatively long with respect to the width of the local heat source,
which has been a common assumption in some of the earlier studies by Beyeler
[10], Beyeler and Gugeri [11], Beyeler et al. [3], and Grove [12,13]. Therefore, a
two-dimensional heat transfer analysis becomes necessary to accurately represent
the thermal diffusion and its effects on the material properties.
Figure 2 depicts the geometry considered in the present analysis and indicates
the various boundary conditions. The formulation of the governing equations is
performed by assuming a local heat source fixed with respect to inertial coor-
dinates x and y, and the process-relative velocity Vx is assigned to the laminate
and the base plate as a whole. The length of the domain for the analysis is taken
as L, which is sufficiently greater (to validate the boundary conditions) than the
width of the heat source 2b. The portions of the substrate upstream and down-
stream from-the consolidation point are indicated as a and c respectively.
The tape laying is a transient heat diffusion problem in a Lagrangian frame-
work, i.e., from a material point of view. On the other hand, in a steady opera-
tion of continuous process, an Eulerian Control Volume (ECV) enveloping the
region influenced by the local heat source can be identified, and a steady-state
formulation can be developed (Eulerian approach). This steady-state approach
becomes possible since at every instant, neglecting the edge effects, the tempera-
ture distribution would remain unchanged with respect to a fixed coordinate
frame. However, the ECV must be chosen large enough to satisfy the boundary
conditions at the upstream and downstream ends with respect to the heat source.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 23

As shown in Figure 2, the temperature T, of the material entering the control

volume at the left side of the ECV can be assumed to be the composite/base-plate
temperature sufficiently far away from the local heat source. For the right side of
ECV, a gradient type boundary condition becomes applicable, indicating that the
primary heat transport is due to the motion of the substrate rather than thermal
diffusion. Considering the relatively shallow zone experiencing temperature vari-
ations and the high speed (relative to diffusion speed) of motion at this intersec-
tion, this assumption can be easily justified [11,14]. Moreover, it is assumed that
the thickness of the composite substrate at the instant when the analysis is per-
formed is at least an order of magnitude greater than the thickness of the prepreg
tow (approx. 0.125 mm) being consolidated. Therefore, the effect of the ply-drop
is assumed to be negligible, leading to a rectangular domain for the analysis as
shown in Figure 2. The accuracy of this assumption increases with increasing
substrate thickness. It is also assumed that the prepreg feed-tape is at a uniform
specified temperature.
A change of variable is introduced by defining an excess temperature as
T =
Tact ― T, where Tact and T, are the material and ambient temperatures, re-
spectively.qo indicates the amount of heat supplied by the localized heat source
and can be absorbed by the composite over the exposed area. It has a non-zero
value only in the interval a - b _< x <- a + b. A factor of A9/A is introduced
to modify the qo term at y = y2 boundary condition for the numerical analysis.
This term accounts for the effective area, on the surface of the composite, which
absorbs heat and can be found using a control volume approach. For example,

Figure 2. The anisotropic domain for the analysis and the boundary conditions used in gen-
erating the results.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

24

this fraction would be unity for nodes which are within the boundary of the area
beneath the local heat source and less than unity for the boundary nodes.
In light of the preceding discussion, the governing equation for the anisotropic
composite domain becomes:

where the terms within = KL and K22 KT represent diffusion in an anisotro-

=

pic domain. Since the laminate under consideration is unidirectional, a formula-

tion based on orthotropic diffusion is sufficient to model the process. General
treatment of the theoretical aspects of anisotropic heat transfer is given by Ozisik
[15]. The convective-like term that appears on the right-hand side accounts for the
motion of the laminate and base plate. The heat generation term, U, is included
to account for the heat absorption or heat release effects resulting from the melt-
ing and solidification that occurs during consolidation. The heat transfer taking
place during the phase change of most polymers is quite complex, and the heat
generation term in this expression captures only the essential effects of such a
phase transformation as it relates to the thermal analysis of the process [16]. A
perfect thermal contact is assumed to exist between the composite and base plate.

3. ANALYTICAL SOLUTION

Equation (1), which governs the heat transfer for an orthotropic (in this case,
transversely isotropic) domain, can be transformed to that for an isotropic do-
main by applying the transformations [15]:

where K is a reference conductivity. Note that the overall effect of these expres-
sions is to appropriately scale the directional coordinates to account for different
rates of thermal conductivity in different directions. With this change in indepen-
dent variables, the governing Equation (1) becomes:

If thequantities QC, and the heat generation term, defined on the basis of unit
volume, should have the same physical significance, then:

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 25

which implies that the fraction at the right-hand side of Equation (4) is unity, i.e.,
K = (KllK22)O 5. Therefore, Equation (3) can be rewritten as:

where

The boundary conditions also have to be transformed accordingly. For exam-

ple, the transformed boundary condition at y = y, becomes:

The governing Equation (5) is non-homogeneous due to the presence of the

heat generation term U. The following operation on the temperature is performed
to enable the homogeneous and non-homogeneous solutions to be separated:

where

set equal to zero for the present case. Equation (9) gives the
The constant A is
non-homogeneous solution, and the governing equation for the homogeneous
part is:

The boundary conditions are also transformed according to Equation (8) and they
become:

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

26

with

where the factor (A9/A) is the unity in the analytical solution.

It should be noted that there exist two non-homogeneous boundary conditions
at x = 0 and y = y,. Using the principle of superposition the problem on 0
can be divided into two separate problems assigning one non-homogeneous

boundary condition to each. It follows that:

Therefore, the 6x equation becomes

subject to the boundary conditions

and the 6y equation becomes

subject to the boundary conditions

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 27

A separation of variables approach becomes applicable to solve the system of

equations given above. Let

Substituting this expression into the governing Equation (17) and rearranging
yields;

The sign of the separation constant, X2, in Equation (28) is taken in such a way
that a Sturm-Liouville system exists on x-direction and is given by:

where the weighting function is exp( - )’x) and the primes denote the deriva-
tives with respect to x. The equation on y becomes:

where the primes denote the derivatives with respect to y. It can be shown that
the solution for the 6x system is of the form:

The X~’s are the eigenvalues that are the roots of the following transcendental
equation:

Dn, x&dquo; and Y, are constants and given in Appendix B, and the eigenfunctions are
found to be:

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

28

Similarly let

Substituting into the governing Equation (22) and rearranging, we obtain:

The sign of the separation constant, w2, in Equation (35) is taken in such a way
that a Sturm-Liouville system exists on y-direction and is given by:

Therefore, the equation on x becomes

It can be shown that the solution for 0; is of the form

The úJn’s are the eigenvalues which are the roots of the following transcendental
equation:

G&dquo;, Ho, ~2, and Ci,. are constants and given in Appendix D, and the eigenfunc-
tions are given as:

The results obtained for T(x,y) are then transformed back to the x - y do-
main using the transformations given in Equation (2).

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 29

3.1 Numerical Solution

NON UNIFORM MESH

The numerical solution of this problem allows for the implementation of more
complex geometrical configurations and boundary conditions. In discretizing the
governing equations for this problem, it becomes advantageous to use a non-
uniform mesh over the domain due to the existence of zones with large tempera-
ture gradients in the vicinity of the surface exposed to the heat source. The non-
uniform mesh allows for reducing the numerical errors without necessitating
excessive computational memory. While the governing equations can be solved
over a non-uniform mesh, it is also advantageous to transform them onto a com-

putational domain where a uniformly spaced mesh can be generated correspond-

ing to the non-uniform mesh of the physical domain. This approach eliminates
the necessity of using interpolations and reduces the errors that would result from
discretization with unequal spacing by building in the characteristic transforma-
tion features into new independent variables. Complete details on algebraic and
numerical transformations can be found in the literature [17,18].

TRANSFORMATION
In the present study, an algebraic transformation between the physical and
computational coordinates is sufficient. The transformed governing equation for
the composite becomes:

Appendix A gives the expressions used for the x and y variables for mapping
the current problem onto a computational domain. The expressions for the dif-
ferent coefficients can be found in Appendix C.
The transformations on the boundary conditions were also performed follow-
ing a similar procedure. For example, the boundary condition at y = y, trans-
forms as

These expressions are then discretized in the computational domain using central
differencing for the diffusion terms. For the heat transfer term induced by the
velocity at the right-hand side of Equation (41) (convective-like term), an upwind
differencing is applied where the discretization is performed by using a backward
differencing expression. This technique is inspired from computational fluid
dynamics where the thermal behavior of the fluid depends largely on the thermal

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

30

history of the upstream region (Gosman et al. [19], Anderson et al. [17], Gfqeri
[18]). The mesh parameters are modified using the Newton-Raphson method to
ensure that mesh points lie on the boundary of the area exposed to the heat
source. Additional details of the numerical procedure can be found in the study
of in-situ thermoplastic-matrix composite filament winding reported by Ghasemi
Nejhad et al. [14].
4. ACCURACY ASSESSMENT
For this study, APC-2 is taken as the composite material, and the average asso-
ciated properties used are those reported by Grove and Short [12], Cattenach and
Cogswell [20], and Blundell and Willmouth [21], with 60 percent fiber volume
fraction; i.e., T, = 550°C, Tm = 340°C, T, = 140°C, Kll = 6.0 W/m-’C,
K22 = 0.72 W/m-’C, Q = 1.56 x 103 kg/m3, Cp 1.425 x 103 J/kg-OC, and
=

U = 10° W/kg. U and Cp are taken as constants and obtained by averaging the
values from experimental data reported by Beyeler and Guqeri [11] and Ghasemi
Nejhad et al. [14]. The laminate is assumed to rest on a solid surface (base plate)
resulting in a convection coefficient at the bottom surface of 103 W/m2.oC and
the top surface is exposed to air with a convection coefficient of 5 W/ml-’C
(Beyeler and Guqeri [11]).
Due to the presence of the exponential terms within the analytical solution
(with the material properties appearing in the power) difficulties are faced in ob-
taining an analytical solution. However, to assess the accuracy of the numerical
code, the results obtained from the numerical method are compared with the ana-
lytical solutions for smaller values of the material properties. To achieve this
goal, all the material properties input into the problem were taken as those re-
ported for APC-2 in the literature except the value of the heat capacity which was
reduced to 10 J/kg’°C. The value of the heat generation was also reduced to 10,
accordingly. The dimensions and processing parameters taken in this section are:
a = 0.005 m, c = 0.01 m, y2 0.005 m, 2 b
= 0.002 m, T, =
100 ° C, T, = =

22 °C, Vx = 0.05
m/sec, and qo 4 x 105
= W/m’.
The temperature profile obtained analytically for the surface of the composite,
where the error is maximum, is given in Figure 3. The analytical results are ob-
tained using the first 20 terms of the series. However, it should be mentioned that
the difference between the results obtained using 10 and 20 terms is insignificant,
and they remain practically unchanged beyond 20 terms.
The result obtained from the numerical method is also shown in Figure 3 for
comparison with the analytical solution. A transformation on the x-axis (i.e.,
x’ = x - a) is performed, in this figure, to clearly show the location of the heat
source at x’ = 0. The relative error at the maximum temperature is 3 %, and the
maximum relative error is 7 %, which indicates excellent agreement between the
analytical solution and the numerical results.
5. PROCESSING WINDOW
In aneffort to understand the effect of typical processing parameters on in-situ
tape laying, a parametric study was conducted. Due to the limitations imposed by

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 31

Figure 3. Comparison between the analytical and numerical results for the temperature pro-
file at the top surface of composite (where the maximum error occurs).

the analytical solution concerning the material properties and realizing the close
agreement between the analytical and numerical approaches, the numerical ap-
proach was used exclusively in the following analysis. APC-2 was taken as the
composite material, and the associated properties are provided in the beginning
of Section 4. U and Cp are taken as functions of temperature employing the ex-
perimental data reported by Beyeler and Guqeri [11] and Ghasemi Nejhad et al.
[14]. For the process of in-situ tape laying, typical processing parameters were
identified to be the heat intensity, qo, the speed, VX, and the width of the heater,
2b (see Figure 2). These parameters must be adjusted so as to prevent the degra-
dation of the outermost surface while still assuring melt at the first ply interface.
For the modelled process, T, and T, are assumed to be 40°C and 22°C, respec-
tively. A typical mesh concentration in the vicinity of the local heat source used
in this study is shown in Figure 4. Figure 5 shows typical temperature contours
within the laminate during the process.

Figure 4. The non-uniform mesh used in the numerical analysis.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

32

Figure 5. Temperature contours in the vicinity of consolidation point during a tape laying
process.

Prior to running the parametric analysis, it was necessary to investigate the ef-
fect of the control volume length, specifically the value of &dquo;c,&dquo; on resulting tem-
perature predictions. Several cases were run, with a typical result provided in
Figure 6. This figure shows the outermost surface temperature along the control
volume for several values of &dquo;c&dquo; (VX = 0.01 m/sec, qo = 106 W/ml, and b =

Figure 6. Temperature profiles for several lengths of the control volume.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 33

0.005 m). As can be seen, the temperature profile is essentially unchanged as &dquo;c&dquo;
is varied from 0.02 to 0.05 m. The maximum temperature deviated by less than
2 percent with essentially no variation in the longitudinal location of the maxi-
mum temperature. Based on these results, all following studies used control vol-
ume dimensions of a = 0.015 m, c = 0.025 m, and Y2 0.005 m. =

With the dimensions of the control volume fixed, values of heat intensity,
heater width and relative speed were systematically varied to determine appro-
priate process conditions. As stated earlier, viable conditions provide melt at the
interface while not degrading the outer surface. Typical results of this study are
shown in the seven plots of Figure 7. For all of these plots, the heater width was
0.01 m (i.e., b = 0.005 m), while speed was varied from 0.01 to 0.10 m/sec and
heat intensity from 0.3 x 106 to 4.5 X 106 W/m2. As one would expect, increas-
ing heat intensity and decreasing speed results in increasing temperature values.
The large temperature gradient through the thickness is attributed to the low ther-
mal conductivity in that direction. The peak of temperature is at the rightmost
edge of the heat source, in the direction of the processing velocity in the analysis
(the opposite direction of the motion of the heat source). It should be noted that,
in Figures 6 and 7, the zero position on the x-axis is shifted along the axis to cor-
respond with the middle of the local heat source (i.e., x’ = x - a) to clearly
show the location of the center of the heat source at x’ = 0.
Examining the plots of Figure 7, no test case provides viable processing condi-
tions. However, the general trends suggest that viable conditions may exist for
qo = 0.9 x 106 W/Ml with a V, around 0.04 m/sec, and Vx = 0.05 m/sec with
a qo around 1.2 x 106 W/m2.

Although the plots of Figure 7 offer some insight, they fail to establish clear
bounds for the processing parameters. These parameters could be established by
performing many more trials to carefully bound the processing window. Al-
though this approach is viable, it is certainly not efficient. In an effort to more
directly determine the effects of each of the variables on the laminate tempera-
ture, plots of heat intensity, speed and heater width versus maximum temperature
were developed and are provided for the interface in Figures 8-10. In Figure 8,
each of the nine lines represents a distinct processing condition. Indeed, if one
were to select any of the heater width and speed combinations, appropriate values
for heat intensity could be directly read from the graph. Similarly, Figure 9 dis-
plays the speed dependence for a variety of heater width and heat intensity com-
binations, while Figure 10 displays the heater width dependence for three heat in-
tensity and speed combinations. Although Figures 8-10 apply only to the
maximum interface temperature, similar trends were also noted for the maximum
surface temperature plots.
Taken together, Figures 8-10 actually provide three views of a temperature sur-
face. For a given heater width, if the speed and heat intensity were plotted on the
horizontal axes and maximum temperature plotted as the elevation, the result
would be a temperature surface with its maximum value at the lowest Vx and
highest qo, and its minimum at the highest V,, and lowest qo. For the interface
temperature, viable conditions would exist only when the temperature exceeded
the melt temperature. In an analogous fashion, for the outer surface temperature,

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 Figure 7. Results for a representative overall parametric analysis ( lN = 0.01 m).

!A
34

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

Figure 7 (continued). Results for a representative overall parametric analysis (W =

0.01 m).

35

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

36

Figure 7 (continued). Results for a representative overall parametric analysis (W =

0.01 my.

viable conditions would exist only when the temperature was below the degrada-
tion level. Those speed and heat intensity values which satisfy both the interface
and surface condition define the processing window.
From Figure 8, it appears that over a reasonable range of values, the tempera-
ture surface is approximately linearly dependent on heat intensity. From Figure
9, the maximum temperature is seen to be inversely related to speed with a non-
linear effect present. Finally, Figure 10 indicates that the maximum temperature

Figure 8. Effect of heat intensity on maximum mterface temperature.

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

Figure 9. Effect of processmg speed on maximum interface temperature.

Figure 10. Effect of heater width on maximum interface temperature.

37

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

38

is related to heater width raised to some power less than one. In an attempt to
match the above noted trends, a variety of equation forms were fitted to the nu-
merical data. For each equation form attempted, an iterative scheme using a Tay-
lor series expansion of the particular equation was used to determine the &dquo;best,&dquo;
in a least squares sense, values of all unknown coefficients. Surface and interface
maximum temperatures were examined separately, although a common form was
used for both. Equations with up to 10 unknown coefficients were investigated,
with a total of 27 numerical trials used as the basis to determine the coefficient
values. Ultimately, the form that was found to provide the best fit of the numerical
maximum temperature data was:

where W is the full heater width or W = 2b. By the nature of its determination,
Equation (43) is valid only over the interpolated region of the input data (see Fig-
ures 11 and 12). This corresponds to 1500C :5 Tmax :’5 1000°C in Equation
(43). For the 27 temperature distributions determined using the numerical heat
transfer analysis, Equation (43) was able to fit all of the maximum interface and
surface temperature values within 3 %. The optimum values of the coefficients are
supplied in Table 1. These values are based on a tape thickness of 0.005 inches.
Once the &dquo;best fit&dquo; coefficients were determined, Equation (43) was then used to
provide those values of heat intensity, heater width and speed which produced the
desired maximum temperatures at the interface and at the surface. Once again,
for the interface the minimum condition is T~rte,,,,, equal to Tm, while at the sur-
face the maximum condition is Tsurface equal to T,. Replacing Tmax in the interface

Figure 11. Processing windows for various heater widths IN = 0 002 m, 0.010 m, and
0.020 m. Note that the processing window is bound by polymer degradation (upper limits)
and insufficient melting (lower limits).

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 39

Figure 12. Processing windows for various processing speeds V 0.01 m/sec, 0.05 =

m/sec, and 0. 10 mlsec. The processmg wndows are bound by the polymer degradation (up-
per limits) and msufficlent meltmg (lower limits).

and surface forms of Equation (43) with T&dquo;, and Td, respectively, provides the
lower and upper bounds of the processing window. This approach resulted in the
typical plots shown in Figures 11 and 12 where the dashed line indicates the mini-
mum process conditions and the solid line represents the maximum conditions.
The processing window is the zone above the dashed line and below the solid
one. It should be realized that Figures 11 and 12 are two views of the process sur-
faces.
As can be seen in Figure 11, for a heater width of 0.002 m (i.e., b = 0.001 m),
no viable process conditions exist for speeds greater than 0.01 m/sec. If the
curves are extrapolated, a narrow processing window does exist for speeds below
0.01 m/sec. At the other extreme, for a heater width of 0.02 m (i.e., b = 0.01 m)
viable processing conditions can be found for speeds up to approximately 0.12
m/sec, with heat intensities ranging up to approximately 1.4 x 106 W/m2.
Figure 7 shows that the temperature decreases with increasing processing

Table 1. Coefficients for Equation (43).

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

40

speed for a given heat intensity and heater width; however, to achieve sufficient
melting the heat intensity should increase. It is found that the temperature gradi-
ent through the thickness increases with increasing processing speed and heat in-
tensity. Therefore, the height of the processing window decreases with increasing
processing speed for a given heat source width. Also, Figure 11 shows that for a
given heat intensity, an increase in the heat source width not only increases the
area of the processing window but also increases the processing speed which, in

turn, increases the production rate. Of some interest is the fact that for the heater
widths shown, no viable process conditions exist for heat intensities in excess of
about 1.4 x 101 W/m2, and the maximum viable heat intensity drops only
slightly with increasing width.
Examining Figure 12, the dependence on heater width becomes evident. For
speeds in excess of 0.1 m/sec, the interpolated results show limited process condi-
tions. However, at speeds of 0.01 and 0.05 m/sec a range of conditions is possible.
Of interest is the anticipated inverse relationship between heater width and heat
intensity. From a process development viewpoint, there is a diminishing value in
using progressively wider heat zones. Except for the leftmost edge of the process-
ing window, the process appears to be relatively insensitive to changes in heater
width, but allows only small variations in heat intensity.
From a practical viewpoint, Figures 11 and 12 are the key to the development
of a viable process. Although the data presented here was interpolated only from
the initial numerical analysis results, it appears that to maximize output, i.e., to
maximize velocity, one should use a heater width of at least 0.02 m. This 0.02 m
heater width would allow for speeds above 0.1 m/sec; however, in this range the
processing window is very small. Clearly, well-focused lasers are not at all
suitable for the example developed herein. Unless a very high degree of process
control is employed, one would be wise to operate below the maximum allowable
velocity in order to &dquo;widen&dquo; the processing window, which allows for some varia-
tion in the absorbed heat intensity without adversely affecting part quality. Fi-
nally, regardless of the heater width and speeds used, heating devices which pro-
vide absorbed intensity less than 0.25 x 106 W/M2 and greater than 1.4 x 106
W/ml are of little to no value.
The effect of the consolidation roller at the top surface of the composite was
taken into account by modifying the top surface boundary condition. This effect
was found to be insignificant due to the fact that the contact area is extremely
small compared with the total top surface.
The number of nodes taken in the analysis is 61 x 21 in the x and y directions,
respectively. Doubling the number of the nodes had an insignificant effect on the
results obtained. Since the computational model relied on experimental measure-
ments of thermal properties which do not have similar accuracies, the results
were considered to be satisfactory. The Gauss-Seidel iteration method is em-

ployed in the numerical analysis. The convergence tolerance (i.e., the relative
error between the consecutive values of the iterations) was taken as 10-4. Succes-
sive over/under relaxation did not enhance the convergence in this study. Addi-
tional information on the numerical study will be reported in [22].
The above example was developed to demonstrate the usefulness of the numeri-

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 41

cal technique in developing and analyzing a given fabrication process. All param-
eters are easily adjusted to suit any alternate process. Typical run times for a sin-
gle case require 1 to 5 minutes of cpu time on a VAX 11/780. Unusual geometries
and drastic temperature variations may require considerably more time.

6. CONCLUSIONS

Analytical and numerical thermal analysis for in-situ thermoplastic tape laying
using a local heat source are presented. The results show large temperature gradi-
ents in the vicinity of the local heat source. The accuracy of the numerical
method is assessed by comparing its results with those obtained from the analyti-
cal analysis. Very good agreement exists between the two methods. Using the
more general numerical technique, the influence of the processing speed, heat
flux and heater width on the nature of the overall heat transfer is demonstrated.
A methodology is presented to characterize all viable processing conditions
based on a finite set of numerical results of the general problem. The processing
window is shown to vanish as heater width decreases, speed increases or heat in-
tensity increases. Indeed, bounds on all three conditions are indicated.
The availability of such an analysis is expected to provide the basis for future
studies on heating/cooling rates, microstructure, and state of residual stresses. It
also forms a bridge linking processing, microstructure, and material perfor-
mance, which are strongly coupled in the case of composite materials.

NOMENCLATURE
A9/A nodal area fraction for the surface exposed to the heat source
a upstream distance from the heat source, m
b half-width of the heat source, m
c downstream distance from the heat source, m
C heat capacity, J/kg’oC ..

D a constant in the analytical solution -’ -

&dquo;

G a constant in the analytical solution ..

-
-
,

ho,h2 convection coefficients, W/m2.oC -

’ ,

h maximum length in numerical transformations . 1.

&dquo;°
Kll conductivity in x-direction, W/m-’C
K22 conductivity in y-direction, W/m’oC ~,

qo absorbed heat supplied by the heat source, W/ml , ,

T temperature, ° C
U heat generation, W/kg ~ .

K~ processing speed, m/sec

W width of the local heat source, m
x,y Cartesian coordinates, m
x’I x-axis shifted to the location of the heat source, m
x,y transformed Cartesian coordinates, m
°

x, x at the right-hand side boundary, m

y, y at the top-surface boundary, m
Xb&dquo;&dquo;Xbp x at the left and right edges of the local heat source, m

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

42

y symbolic length in mesh transformation, m

ac a constant in the mesh refinement formulae
{3 mesh refinement parameter in y-direction
ly a constant in the analytical solution
6 transformed temperature
x eigenvalue with respect to x-direction
e density, kg/m’
T mesh refinement parameter in x-direction
a eigenvalue with respect to y-direction
Subscripts:
0 bottom surface of laminate
2 top surface of laminate
c control
d degradation
E environment
g glass transition
L longitudinal
m melt
n number of eigenvalues
T transverse
x expansion on x

y expansion on y

APPENDIX A
Transformation on &dquo;X&dquo; ( T : refinement parameter)

where

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 43

Transformation on &dquo;Y&dquo; (0: refinement parameter, a =

0)

APPENDIX B
The Dn coefficient in Equation (31),.k, in Equation (32), and y,

where:

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 44

APPENDIX C
The coefficients of Equation (41),

APPENDIX D
The Gn, Ho, H2, and Cg, coefficients of Equations (38) and (39):

where:

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015

 45

REFERENCES
1. Chang,I Y and J K Lees 1988 "Recent Development in Thermoplastic Composites: A Re-
viewof Matrix Systems and Processing Methods," Journal of Thermoplastic Composite Mate-
, 1 277-295
rials
2. Griffiths, G R , W D Hillier and J A S Whiting 1988. "Thermoplastic Composite Manufac-
turing Technology for a Flight Standard Tailplane," in Proceedings, 33rd International SAMPE
Symposium, Anaheim, CA, 1988, volume 33, pp 308-316
3. Beyeler, E P , W Phillips and S I Gu&ccedil;eri 1988. "Experimental Investigation of Laser-Assisted
Thermoplastic Tape Consolidation," Journal of Thermoplastic Composite Materials
, 1 107-121.
4 Gilmore, S D 1991 "Processing of Thermoplastic Matrix Composite Materials," Ph.D thesis,
Dept Mech. Eng., Univ Delaware, Newark, DE.
5 Seferis, J C "Polyetheretherketone (PEEK) a Matrix Semicrystalline Polymer for High Per-
formance Composites," Techmcal report, National Academy of Athens, Athens, Greece, 1984
6. Velisaris, C N and J C Sefens 1988 "Heat Transfer Effects on the Processing-Structure Rela-
tionships of Polyetheretherketone (PEEK) Based Composites," Polymer Eng. and Sci , 28(9):
583-591
7. Blundell, D J.,J M. Chalmers, M W. Mackenzie and W. F. Gaskin 1985. "Crystalline Mor-
phology of the Matrix of PEEK-Carbon Fiber Aromatic Polymer Composites,1 Assessment of
Crystallinity," SAMPE Quarterly, 16(4) 22-30
8 Motz, H. and J. M. Schultz 1989 "The Solidification of PEEK Parti Morphology," Journal
, 2:248-266
of Thermoplastic Composite Materials
9. Motz, H. and J M Schultz 1989 "The Solidification of PEEK, Part 11 Kinetics," Journal of
, 2 267-280
Thermoplastic Composite Materials
10. Beyeler, E P 1987 "Melting and Solidification in Processing and Applications to Thermoplastic
Matrix Composites Manufacturing," Ph D thesis, Dept Mech Eng , University of Delaware,
Newark, DE
11. Beyeler, E. P and S I Gu&ccedil;eri 1988. "Thermal Analysis of Laser-Assisted Thermoplastic-
Matrix Composite Tape Consolidation," Journal of Heat Transfer
, 110:424-430.
12. Grove, S and D Short "Heat Transfer in APC Fabrication," Technical report, ICI PLC Petro-
chemicals and Plastics Division, Wilton, DE, 1984
13. Grove, S M 1988. "Thermal Modelling of Tape Laying with Continuous Carbon Fiber-
Remforced Thermoplastic," Composites, 19(5):367-375
14 Ghasemi Nejhad, M N , R D Cope and S I Gu&ccedil;eri. "Thermal Analysis of in-situ Thermo-
plastic-Matrix Composite Filament Winding," Journal of Heat Transfer (in press, 1991).
15 &Ouml;zi&scedil;ik, M N. 1979 Heat Conduction New York. Wiley Interscience.
16 Erhun, M and S G Advani. 1990 "A Predictive Model for Heat Flow during Crystallization of
Semi-Crystalline Polymers," Journal of Thermoplastic Composite Materials
, 3:90-109
17 Anderson, D A , J. C. Tannehill and R H Pletcher. 1984. Computational Fluid Mechanics and
Heat Transfer. New York Hemisphere Pub Co
18 G&uuml;&ccedil;eri, S. I 1989. "Finite Difference Solution of Field Problems," in Computer Modeling for
Polymer Processing. C L. Tucker, ed Munchen-New York. Hanser Pub.
19 Gosman, A. D., W M. Pun, A. K Runchal, D B Spalding and M Wolfshtein. 1969. Heat and
Mass Transfer in Recirculating Flows. London Academic Press.
20. Cattenach, J B. and F N Cogswell 1985. "Processing with Aromatic Polymer Composites," in
Developments in Reinforced Plastics G Pritchard, ed New York. Applied Science Publishers.
21 Blundell, D J. and F. M. Willmouth. 1986 "Crystalline Morphology of the Matrix of PEEK-
Carbon Fiber Aromatic Polymer Composites," SAMPE Quarterly, 17(2):50-57.
22. Ghasemi Nejhad, M N "Three-Dimensional Thermal and Residual Stress Analysis of in-situ
Thermoplastic Composite Filament Winding," Ph.D. thesis, Dept. Mech. Eng., Umv. Delaware,
Newark, DE (in progress, 1991)

Downloaded from jtc.sagepub.com at CONCORDIA UNIV LIBRARY on July 3, 2015