You are on page 1of 11

Composites Part B 177 (2019) 107284

Contents lists available at ScienceDirect

Composites Part B
journal homepage: www.elsevier.com/locate/compositesb

Numerical analysis of screw swirling effects on fiber orientation in large


area additive manufacturing polymer composite deposition
Zhaogui Wang, Douglas E. Smith *
Department of Mechanical Engineering, School of Engineering and Computer Science, Baylor University, Waco, TX, 76798, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Large Area Additive Manufacturing (LAAM) polymer deposition employs a single screw extruder to deliver
Large area additive manufacturing pelletized feedstock resulting in significantly higher flow rates as compared to conventional filament-based
Swirling motion extrusion additive processes. Swirling kinematics in LAAM melt flow that result from the screw rotation
Non-Newtonian viscoelastic flow
generate unique particle alignment patterns within the fiber-filled polymer during deposition processing. This
Reduced strain closure model
paper investigates the effect of the single screw swirling motion on the resulting fiber orientation in a short fiber
polymer composite extrudate. An axisymmetric non-Newtonian viscoelastic flow is simulated with the finite
element method, where the flow nearby the extruder screw tip, within the printing nozzle, and a short section of
free extrudate compose the flow domain. Fiber orientation tensors within the flow domain are evaluated using
the Wang-O’Gara-Tucker Reduced Strain Closure (RSC) fiber orientation diffusion model with the orthotropic
fitted closure. The results indicate that swirling kinematics yield a longer flow path for fibers to travel and
orientate within the flow domain, yielding orientation tensor results that are notably different as compared to a
non-swirl event. The predicted principal elastic constant from the swirling flow and non-swirling flow models
exhibit 21% difference, and those from the swirling model using the RSC closure shows a good agreement with
reported experimental data for a similar material system.

1. Introduction single screw extruder to melt and extrude pelletized feedstock onto a
print bed which results in a significant increase in the material output
Fused Filament Fabrication (FFF) is a popular method of Additive rate, enabling the fabrication of larger structures in less time. Previous
Manufacturing (AM) that has experienced widespread application literature demonstrated the single screw rotation has a direct effect on
owing to its high degree of manufacturing freedom and affordability [1]. the properties of an extruded composite material. For example, Ke, et al.
The FFF process melts and deposits polymer filament onto a plate sur­ [8] used simulation to show that the viscosity dropped initially and then
face, layer-by-layer, to form a 3D object from digital data. Unfortu­ tended to recover in single screw extruder flows. This trend appeared to
nately, the inherently weak meso-structure of an FFF printed part results be periodic and related to the geometry of the screw. Canevarolo, et al.
in an 11%–37% reduction in modulus and 22%–57% decrease in [9] explored the effects of various screw element type on the degrada­
strength as compared to the same size injection-molded part [2]. Fillers tion of Polypropylene (PP). Five different screw designs were employed
such as chopped carbon fibers are added to neat polymer feedstock to to show that increasing the screw number as well as the screw profile
enhance the material properties of the printed structure [3–6]. Rein­ aggressiveness yielded a noticeable reduction in the average molecular
forcing fibers within printed composite parts become a vital structural weight of PP. Kelly, et al. [10] experimentally measured the melt tem­
component that greatly enhances the mechanical and perature profiles in single screw extrusion flow using various screw
thermo-mechanical properties of solidified structures [7]. geometries, and Ramani, et al. [11] investigated the fiber length
FFF-type AM produced structures have seen recent shifts in appli­ degradation in a twin-screw extruder. The reported data indicated that
cation from rapid prototyping of moderate-size structures to the fabri­ the injection molding process did not benefit from the screw-yielded
cation of large-scale end use parts and tooling [1,7]. Unlike traditional wet-out and dispersion of fiber reinforcements. More recently, Duty,
FFF processes, Large Area Additive Manufacturing (LAAM) utilizes a et al. [7] measured the structural performances of fiber reinforced

* Corresponding author.
E-mail address: Douglas_E_Smith@baylor.edu (D.E. Smith).

https://doi.org/10.1016/j.compositesb.2019.107284
Received 22 February 2019; Received in revised form 2 August 2019; Accepted 11 August 2019
Available online 13 August 2019
1359-8368/© 2019 Elsevier Ltd. All rights reserved.
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

polymers printed with Big Area Additive Manufacturing (BAAM). The homogenization method [19–21] with the computed steady-state
measured stiffness of short carbon fiber-filled Acrylonitrile Butadiene orientation tensors at the flow domain exit. Swirling flow results are
Styrene (ABS) beads showed that the printed material exhibited a high compared to non-swirling flow results for a similar flow domain geom­
degree of anisotropy owing to the fiber orientation within the extrudate, etry and processing conditions. The polymer composite considered in
where the elastic modulus in the print direction is much higher than this paper is 13% Carbon Fiber reinforced ABS (CF-ABS), which enjoys
those normal to the axis of the printed bead. In addition, their work widespread application in LAAM related research [7,22,23]. Our
revealed that screw design significantly affected the resulting elastic approach that assumes fiber orientation does not affect flow kinematics
properties of the printed materials, such that the principal modulus (i.e., has proven to be effective in narrow gap, shear dominate flow applica­
that aligned with the print direction). Duty showed that the elastic tions such as injection molding and some extrusion scenarios [17]. The
modulus of a 20% glass fiber reinforced ABS bead printed with a con­ fully-coupled flow-fiber orientation problem has been solved by Ver­
ventional screw was 5.67 GPa, while that using retrofit screw was Weyst and Tucker [24] and Dinh and Armstrong [25], among others, for
3.33 GPa (c.f. Fig. 10 in Ref. [7] for screw geometric designs). internal flows. However, the fully coupled approach for free surface
Fiber orientation during polymer composite deposition has recently flows such as those seen in LAAM are left for future studies.
become of interest. Nixon, et al. [12] computed fiber orientation dis­
tributions in printer nozzles with various internal geometries using the 2.1. Material rheology
Moldflow software (Moldflow Corporation, Framingham, MA). Their
results indicated that a convergent nozzle geometry yields higher fiber In this paper, we apply the Phan-Thien-Tanner (PTT) model to
alignment than that of a divergent designed nozzle. Heller, et al. [13] simulate the polymer melt rheology in LAAM deposition as in Wang and
extended Nixon’s work by including the fiber orientation computation Smith [16]. The constitutive equation of the PTT model is [26]
within the die swell of a short section of polymer composite extrudate. A � � h� ξ� ξ i
Newtonian creeping flow model was employed in the finite element exp
ελ
trðT1 Þ T1 þ λ 1 þ Tr þ Tr ¼ 2η1 D; (1)
suite COMSOL (Comsol Inc., Burlington, MA) where fiber orientation η1 2 1 2 1
tensors were computed from the computed flow fields. Their results
quantified the high fiber alignment that occurs along the primary flow where
direction. More recent work by Heller, et al. [14] included the deposi­ DT1
tion of a polymer composite bead onto a moving platform using a 2D TΔ1 ¼ þ T1 ⋅ðrvÞT þ rv⋅T1 ; (2)
Dt
planar flow model in COMSOL. Russell, et al. [15] computed the effec­
tive thermal-structural properties of a single large-scale AM deposited and
2D planar bead using the orientation homogenization method with the DT1
computed fiber orientation, solved from the flow domain computed by Tr
1 ¼ T1 ⋅rv ðrvÞT ⋅T1 ; (3)
Dt
Heller, et al. [14] where both elastic constants and the Coefficient of
Thermal Expansion (CTE) are reported. In addition, Wang and Smith In the above, D and v are the rate-of-deformation tensor and velocity
[16] evaluated the effects of assumed polymer rheology on the predicted tensor, respectively; where D is written as
fiber orientation in a nozzle flow and free extrudate. Generalized 1� �
Newtonian Fluid (GNF) models (power law and Carreau-Yasuda law) D¼ ðrvÞ þ ðrvÞT ; (4)
2
and a viscoelastic fluid model (Phan-Thien-Tanner (PTT) model) were
In the above, the extra stress tensor T is decomposed into a visco­
each applied to model the rheology properties of molten polymer melt
elastic component T1 and a purely viscous term T2 ¼ 2η2 D. The su­
flow. Their results revealed that the non-Newtonian rheology effects of
perscript ‘T’ indicates matrix transpose and η2 represents the viscous
the polymer melt significantly varied the predicted fiber orientation
viscosity. Noted by the user guide of Polyflow [27], we set η2 as 1/9 of
pattern, and in turn affected the result of a predicted elastic constants. It
the η1 of the fastest relaxation time (e.g. mode i ¼ 1 in Table 1).
was shown that results obtained using the PTT model yielded lower
We obtain the rheological properties of 13% CF-ABS through a fre­
averaged principal fiber orientation than those from both the GNF and
quency sweep test using the rotational rheometer HAKEE MARS 40
Newtonian models.
(Thermo Fisher Scientific, Waltham, MA) at 215 Celsius, which is a
common temperature for printing of this material. Results obtained from
2. Materials and methods
an oscillatory rheometer frequency sweep over a range of 0.01–100 Hz
are converted to shear rate using the Cox-Merz rule [28]. The applica­
Flow-induced orientation is one of the most important attributes in
tion of the Cox-Merz rule for filled polymer systems can influence the
short fiber polymer composite processing that determines the mechan­
predicted rheological properties. Prior experimental work [29] shows
ical properties of a solidified part [17]. This paper provides a
that the deviation between Cox-Merz-rule-estimated data and the same
numerical-based approach to evaluate the influence of the rotating
measured by a capillary rheometer can be as high as one order of
single screw on predicted the fiber orientation in LAAM polymer com­
magnitude (e.g., ~10 � ). While the presence of fiber reinforcements
posite deposition. To address this issue, we simulate the polymer melt
may introduce error into the results appearing in Fig. 1, it is expected
flow in the extruder-nozzle region using the ANSYS Polyflow Release
that the swirling kinematics of the relatively low fiber content polymer
19.1 (ANSYS, Inc., Canonsburg, PA, USA) finite element software. The
flow would not be greatly affected as compared to other factors, such as
finite element flow domain is composed of a short section of the extruder
the inlet flow rate or screw RPM. An alternative approach is to employ a
that connects the screw tip, the nozzle portion, and a short strand of a
fully-coupled fiber orientation-flow analysis (e.g., see VerWeyst and
vertical extrudate. Specifically, the edge of the screw tip is included in
the nozzle flow domain which introduces swirling kinematics into the
modeled flow fields. Second order fiber orientation tensors are then Table 1
Fitted parameters for the five-mode PTT model.
computed throughout the flow domain using the velocity and velocity
gradients fields solved in the melt flow simulation, assuming the fiber i λi ðsÞ ηi ðPa ⋅sÞ
presence does not affect flow rheology. We employ the Wang-O’Gar­ 1 0.0032 719.238
a-Tucker Reduced Strain Closure (RSC) orientation diffusion model [18] 2 0.024 2009.14
to capture the slow orientation kinetics of the reinforcing short fibers 3 0.17 6812.35
within the polymer melt. Effective elastic constants of a printed com­ 4 1.37 11716.6
5 10.00 124392
posite extrudate bead are also computed using an orientation

2
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

Fig. 1. Fitted rheology data of 13% CF-ABS using PTT model: (A) shear viscosity; (B) dynamic shear moduli.

Tucker [24]) that more directly incorporates the effects of fiber re­ The flow domain considered here includes the lower section of the
inforcements into the computation of the flow. As stated above, the extruder near the tip of the rotating screw, the nozzle, and a short sec­
fully-coupled scheme is expected to be employed in the future work. tion of free extrudate just outside the nozzle exit as shown in Fig. 3.
PTT model parameters are computed using ANSYS Polymat [30] Axisymmetry of the flow domain is assumed which is modeled with a 2-
where curve fit results of the PTT model appear in Fig. 1 for the PTT D finite element mesh to reduce computational expense. Gravity is not
parameters given in Table 1. Values given here for η, G’ and G’’ refer to considered in the simulation. In addition, the length of free extrudate in
the shear viscosity, storage shear modulus and loss shear modulus, the model is one-inch which allows for the fiber orientation to reach
respectively. In addition, the PTT model constants ξ and ε control the steady state beyond the nozzle exit. As the design of the single screw is
shear thinning viscosity and extensional viscosity behaviors (c.f. [27]), not provided by the manufacturer and thus herein we adapt the geom­
respectively. Fitted results for η, G’ and G’’ appearing in Fig. 1 assume etries as referred to Duty, et al. [7], where the investigated large-scale
ξ ¼ 0.15 and ε ¼ 0.5. Note that the length of each error bar appearing in polymer deposition system was employed with a 25-mm (~1-inch)
Fig. 1 is twice of the absolute value between the measured data and the diameter screw. In addition, Duty has improved the design of a screw in
fitted data at that shear rate. It can be seen that the fitted rheology order to achieve higher material feeding rate, where a 50-mm (~2-inch)
properties show a good agreement to those obtained from the frequency longer screw is employed to their extruder. Based on this prior work, we
sweep test, except at extreme shear regions (e.g., very small or very large assume a 1-inch distance between the screw tip to the nozzle entrance as
shear rates). shown in Fig. 3. Additionally, the angle of the screw tip is assumed as
shown, which is not expected to yield significant effects on the resulting
flow fields and fiber orientation.
2.2. Finite element polymer flow Flow simulations are performed using the finite element suite,
ANSYS-Polyflow, which is well-suited for polymer melt flow analysis in
To evaluate fiber orientation in the LAAM process, the velocity and complex geometries. The melt flow velocity is computed assuming an
velocity gradients within the polymer melt flow through the extruder isothermal incompressible steady creeping flow, and the transient and
nozzle and within the die swell just outside the nozzle exit are computed. inertia effects are ignored as in similar prior studies [13–16]. The con­
The flow domain of interest is created based on the Strangpresse Model- tinuity and momentum equations for flow models that include swirling
19 extruder nozzle appearing in Fig. 2, which is common in LAAM flow in the θ direction (c.f. Fig. 2C) can be written as [31]
polymer deposition additive manufacturing systems. We have chosen to
ignore the nozzle valve that is included in the Strangpresse nozzle design 1 ∂ðρrvr Þ ∂ρvz
þ ¼ 0; (5)
for simplicity. In addition, screw flights on the extruder screw are not r ∂r ∂z
included in the model of flow domain and thus any turbulence or added
and,
shear stresses within the polymer melt caused by the screw flights are
neglected in this study.

Fig. 2. Strangepresse Model 19 extruder nozzle: (A) external view of nozzle; (B) nozzle dimensions; (C) cylindrical coordinates system.

3
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

Fig. 3. The boundary condition labels of the swirling flow domain (lengths in inches). Note, the shadow area is the flow domain computed in Polyflow. (See online
publication for full color image.)

� � � �
∂vr v2θ ∂vr ∂P 1 ∂ τθθ ∂τzr 2D free surface problems [28].
ρ vr þ vz ¼ ðrτrr Þ þ ; (6)
∂r r ∂z ∂r r ∂r r ∂z In addition to the swirling flow defined above, a straight flow model
is also considered for comparison. The straight flow model shares the
� � � �
∂vθ vr vθ ∂vθ 1 ∂ 2 � τθr τrθ ∂τzθ same nozzle geometry, flow rate and melt flow rheology with the
ρ vr þ vz ¼ r τrθ þ þ ; (7)
∂r r ∂z r ∂r
2 r ∂z swirling flow, except that no rotational kinematics are assigned to Γ3 , i.
e., vθ ¼ 0 with a no-slip wall boundary condition. Evaluating the poly­
� � � �
∂vz ∂vz ∂P 1 ∂ ∂τzz mer melt flow and fiber orientation in the straight flow models provides
ρ vr þ vz ¼ ðrτrz Þ þ ; (8)
∂r ∂z ∂z r ∂r ∂z a means to directly assess the effects of swirling kinematics on predicted
outcomes. Note that, although, both swirling flow and straight flow
where P is pressure, ρ is the density of the polymer, τ is the shear stress models are created in 2D, where all partial derivatives with respect to θ
tensor, and r; z; θ refer to the three major directions in a cylindrical co­ are zero, the vθ component in the swirling flow is a computed non-zero
ordinate system as shown in Fig. 2C. Equations (5)-(8)) may be obtained solution variable. In the straight flow model, vθ and all derivative with
from the more general Navier-Stokes equations [31] for steady flows respect to θ are set to zero, such that Equation (6) reduce to
having no inertia or gravity, and setting all derivatives with respect to θ � � � �
to zero. ∂v ∂v ∂P 1 ∂ ∂τzr
ρ vr r þ vz r ¼ ðrτrr Þ þ ; (9)
The boundary conditions for the flow domain appear in Fig. 3 as ∂r ∂z ∂r r ∂r ∂​z​
follows: In this formulation, Equation (7) no longer applies, and Equation (8)
remains as given above.
� Γ1 : Flow inlet with the prescribed volumetric flow rate Q. The inlet
flow is assumed to be a fully developed where the inlet velocity
2.3. Fiber orientation computation
profile is computed by ANSYS-Polyflow based on Q and the selected
rheology. In this paper, Q ¼ 1000 mm3 =s based on an 80 RPM screw
Fiber orientation evaluation in polymer melt flow has mostly
speed of the Strangpresse Model 19 extruder or approximately 8 lbs./
developed from the foundational work of Jeffery who first derived the
hour of 13% CF-ABS.
motion of a single rigid massless ellipsoid in a purely viscous fluid [32].
� Γ2 : No slip wall boundary where vs ¼ vn ¼ 0.
Following the work of Jeffery, the direction of a fiber is defined by the
� Γ3 : Screw barrel edge, where vs ¼ vn ¼ 0; vθ ¼ 260 πrn ¼ 8:4r for a
unit vector p, which aligns with the longitudinal axis of the suspended
screw RPM of n ¼ 80. Note that r is the radial distance in the rigid particle. The equation of motion for a suspended ellipsoid is given
axisymmetric model as defined in Fig. 3. as [33]
� Γ4 : Axis of symmetry where Fs ¼ vn ¼ 0.
� Γ5 : Free surface where v ⋅n ¼ 0. p_ ¼ W⋅p þ γe ðD⋅p þ D : pppÞ; (10)
� Γ6 : Flow domain exit where Fn ¼ vs ¼ 0.
where D and W are the rate-of-deformation tensor and vorticity tensor,
In the above, Fs is the tangential force, Fn is the normal force, vs is the respectively, which are the symmetric and antisymmetric parts of the
tangential velocity, vn is the normal velocity, v is the velocity vector on velocity gradient of the flow such that,
the free surface, and n is a unit normal vector on the free surface. The die 1� �
swell of the free surface is computed using the methods of spines in W¼ ðrvÞ ðrvÞT ; (11)
2
ANSYS Polyflow, which is an efficient remeshing rule often applied to

4
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

and D is given in Equation (4). Also, the factor γe relates to the hydro­ which are written in terms of αm and σ m i ,the m-th eigenvalue and the i-th
dynamic aspect ratio of the particle. For an ellipsoidal fiber with axis component of the m-th eigenvector of A, respectively. In the above, κ
ratio aer (i.e., the hydrodynamic aspect ratio for the ellipsoid), γ e ¼ serves as a strain reduction factor limiting the growth rate of fiber
ðaer 2 1Þ =ðaer 2 þ1Þ [19,21]. Note that Jeffery’s work only applies to alignment. Upon comparing Equation (13) to Equation (12), the addi­
single particles in dilute suspensions, limiting its direct use in short fiber tional terms in the Wang-O’Gara-Tucker model of Equation (13) appear
polymer composites flow simulations. However, Folgar and Tucker [33] as a closure approximation, thus Equation (13) is referred to as the
extended Jeffery’s work to include fiber-fiber interaction in concen­ Reduced Strain Closure (RSC) model. Note that when κ ¼ 1, the RSC
trated suspensions by introducing a fiber orientation distribution func­ model in Equation (13) reduces to Advani-Tucker IRD model in Equation
tion and fiber interaction into the fiber orientation equation of motion (12). In flow fields that promote changes in fiber orientation, such as
(c.f. Equation (10)). Unfortunately, the Folgar-Tucker model is compu­ those found in LAAM nozzle flow, computation of A depends on the
tationally prohibitive for flows involving a large number of fibers such as value of κ in the RSC model.
those found in polymer deposition AM [13]. It is important to note here that we employ a different coordinate
Advani and Tucker defined fiber orientation tensors to quantify the system in the polymer flow equations (c.f., Equations (5)-(9)) which are
orientation state for concentrated suspensions with far fewer indepen­ written in terms of a cylindrical coordinate frame as defined through
dent variables as compared to the Folgar-Tucker model [19]. Their Fig. 2C), than in the computation of the second order orientation tensor
approach has enjoyed widespread application in polymer composite field (c.f., Equations (10)–(13)) which are in terms of a Cartesian coor­
molding where the moments of the fiber orientation distribution func­ dinate frame). Hence, the velocity gradients computed in our polymer
tion ψðpÞ define a second order orientation tensor A ¼ Aij ¼ flow calculation are transformed to the Cartesian system for fiber
H
pi pj ψðpÞdS, and a fourth order orientation tensor A ¼ Aijkl ¼ orientation computation through (see e.g., Ref. [38])
HS 2 3 2 dv dv dv 3
pi pj pk pl ψðpÞdS [19]. Note that integrations are performed over the dvr 1 dvr vθ dvr 1 1 1
S 6 ∂r r dθ r dz 7 6 6 dx 1 dx2 dx 37
7
6 7 6
surface S of the unit sphere and that the integral of the probability 6 7 6 dv dv dv 7
6 dv 1 dvθ vr dvθ 7 6 2 27
(14)
2
density function ψðpÞ over the entire sphere equates to unity which rv ¼ 6 θ
6 dr r dθ
þ 7¼6
7 dx dx dx
7;
7
r dz 7 6 1 37
forms a normalization condition in A. It follows that the trace of A is one
2
6 6 7
4 dvz 1 dvz dvz 5 4 dv dv dv 5
and A is symmetric; both of which reduce the number of independent 3 3 3
∂r r dθ dz dx1 dx2 dx3
components in A to five [21]. Therefore, the second order fiber orien­
tation tensor can be written in a compact form as A ¼ Aij ¼ ½A11 ; A12 ;
where we use subscript 1, 2, 3 to represent the coordinate directions in
A13 ; A22 ; A23 �, which provides an efficient approach for computing fiber
the 3D Cartesian frame. Spatial locations are transformed between (r,θ,
orientation in polymer melt flows. It is common to approximate the forth
z) and (x1, x2, x3) in the usual manner where we designate x3 as the axial
order fiber orientation tensor A as a function of A where numerous
direction of the nozzle, z, and x1 corresponds to r direction in the cy­
closure approximations have been proposed including the hybrid
lindrical coordinates. Note that effective elastic properties computed as
closure [34], the natural closure [35], the invariant-based fitted closure
shown in the following section is also performed in the Cartesian coor­
[36], and the orthotropic closure [37]. Our study employs the ortho­
dinate frame.
tropic closure due to its proven numerical stability in various prior
studies ([13–16]).
2.3.2. Orientation homogenized effective stiffness
Starting with Jeffery’s work, Advani and Tucker [19] derived the
It is important to quantify the effect of suspended fibers and fiber
equation of motion for the second order orientation tensor A as
orientation on elastic properties within the extruded bead of the LAAM
DA process. Development of micromechanics models provides analytical
¼ ðA⋅W W⋅AÞ þ γ e ðD⋅A þ A⋅D 2A : DÞ þ 2CI γ_ ðI 3AÞ; (12)
Dt approximations for the stiffness tensor components of a unidirectional
aligned fiber reinforced polymer (see e.g., Refs. [20,39–41]). These
D
where Dt donates the material derivative. The last term of Equation (12) unidirectional models serve as a basis for orientation homogenization
imposes an isotropic rotary diffusion which eliminates the tumbling methods (see e.g., Advani and Tucker [19]. and Jack and Smith [21])
motion in Jeffery’s work where the rotary diffusion is proportional to that yield orientation averaged elastic properties for a short fiber poly­
scalar magnitude of D, appearing as γ_ in Equation (12). The fiber-fiber mer composite. The local orientation average stiffness tensor C ~ ijkl may be
interaction coefficient CI is a phenomenological factor that may be written as [19]
adjusted to accommodate the interaction between suspended fibers. The � �
Isotropic Rotary Diffusion (IRD) term proposed by Folgar and Tucker ~ ijkl ¼ M1 Aijkl þ M2 Aij δkl þ Akl δij þ M3 Aik δjl þ Ail δjk þ Ajl δik þ Ajk δil
C

assumes a steady orientation state under large strains [33]. The model þ M4 Aij δkl þ M5 Aik δjl þ Ail δjk ;
appearing in Equation (12) is often referred to as the Advani-Tucker IRD (15)
model (or IRD model).
where material constants MI, I ¼ 1, …,5 are computed from
2.3.1. Reduced Strain Closure (RSC) fiber orientation diffusion model
More recently, fiber orientation measurements in injection molded M1 ¼ C11 þ C22 2C12 4C66 ; M2 ¼ C12 þ C22 ; M3
plaques showed that fiber alignment occurs slower than that simulated ¼ C66 þ 1=ð2C22 þ 2C23 Þ; M4 ¼ C23 ; M5 ¼ 1=ð2C22 2C23 Þ (16)
using Equation (12) [18]. In response, Wang, et al. modified the
In the above, δij is the Kronecker delta [42], and the Aij are orien­
Advani-Tucker orientation evaluation equation by introducing a term
tation tensor components from the solution of either Equation (12) or
that reduces the growth rate of fiber orientation owing to the strain
13. It is common to compute the Aijkl using the orthotropic closure
effects. The resulting Wang-O’Gara-Tucker model is written as [18]
approximation [37] and we employ the orthotropic fitted closure of
DA VerWeyst for this purpose [24]. The Cij appearing in Equation (16) are
¼ ðA⋅W W⋅AÞ þ γe ðD⋅A þ A⋅D 2½A þ ð1 κÞðL M : AÞ � : D Þ
Dt components of the stiffness tensor for the associated unidirectional fiber
þ κCI γ_ ð2I 6AÞ; filled composite written in contracted notation, which we compute using
(13) the Tandon-Wang micromechanics model [20]. It is shown by Tucker
and Liang [43], as well as other researchers [21], that the Tandon-Wang
P3 m m m m
P3 m m m m
where L ¼ Lijkl ¼ m¼1 αm σ i σ j σ k σ l , M ¼ Mijkl ¼ m¼1 σ i σ j σ k σ l , equation yields one of the most accurate estimates for elastic properties

5
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

over the range of fiber aspect ratios found in the short fiber composites,
such as those considered in this paper.
In our work, values of orientation tensor components Aij are calcu­
lated along flow streamlines with Equation (13). Then, the C ~ ijkl are
computed at the end of each streamline (i.e., at the flow domain exit Γ6 )
which is considered to be the steady-state orientation state of the
extruded bead. Extrudate cross-section averaged values form the effec­
tive bead stiffness tensor Cijkl written as
Z 2π Z ro
1 �
Cijkl ¼ ~ ijkl ⋅ r drdθ
C (17)
πr20 0 0

which are obtained through numerical integration using the trapezoidal


rule as in Ref. [42]. In the above, ro is the radius of the free extrudate at
the flow domain exit. Evaluating elastic constants from the stiffness
tensor Cijkl is performed in the usual manner which is omitted here for
conciseness [44].

3. Results and discussions

Computed values of flow velocity, fiber orientation tensors, and


elastic properties obtained from the LAAM nozzle simulation approach
describe above are given in this section. To better evaluate the effect of
Fig. 4. 3D streamlines of the melt flow computed with the swirling flow model
swirling flow on the computed results, two finite element models are (length values in meters). (See online publication for full color image.)
defined: 1) the first is identified here as the ‘swirling flow’ model using
Equations (5)–(8) with vθ 6¼ 0 and 2) the second is denoted as the
a distinguishable difference between the straight flow and swirling flow
‘straight flow’ model using Equations (5), (8) and (9) with vθ ¼ 0. Both
fiber orientation tensors near the screw tip, and downstream from the
flow models assume all partial derivatives with respect to theta are zero.
screw tip as well.
The swirling flow model imposes the boundary condition vθ ¼ 260 πrn along

the surface of the screw tip to simulate steady rotating motion as


described above. While the flow simulation for both models is performed 3.2. Fiber orientation tensors
within the r-z plane, the swirling flow model requires streamlines and
fiber orientation tensors be computed in three-dimensional space. With In this paper, fiber orientation tensors are computed by integrating
the exception of the detail around the screw tip, both the swirling flow Equation (13) along a streamline using an adaptive fourth-order Runga-
and straight flow model share the same nozzle geometry, flow rate, and Kutta scheme, where an initial condition is required at the flow inlet (i.
polymer melt flow properties. In both flow models, fiber orientation e., Γ1 in Fig. 3). Heller, et al. assumed the fiber orientation state up­
tensors are computed with the RSC model in Equation (13) and elastic stream of the flow inlet reaches steady state based on the upstream
properties are computed for the same extruded 13% CF-ABS polymer homogeneous flow field [13]. We adopt the same assumption by first
composite bead using Equations (15)-(17). computing the fiber orientation of a sufficiently long extruder tube (not
shown here) for the same flow rate Q evaluated above starting with a
random alignment at its inlet. Computed orientation tensor values at the
3.1. Polymer melt flow kinematics
exit of this tubular flow domain are used to define the inlet initial
condition in the following fiber orientation computations. The inner and
The weakly coupled formulation adopted in our study solves the
outer radii of the tubular geometry used to obtain these initial condi­
fiber orientation tensor equation of motion (see e.g., Equation (13))
tions are the same as those appearing in Fig. 3 for the nozzle at z ¼ 4.491
from velocity and velocity gradient values along streamlines in the flow
inch (i.e., at Γ1 ). We set vθ ¼ 0 in the steady flow initial conditions
domain. Melt flow streamlines computed in our swirling flow and
simulation for the straight flow, and define vθ based on the extruder
straight flow simulations appear in Figs. 4 and 5. Streamlines computed
screw diameter and RPM for the swirling flow in a similar manner as that
with the swirling flow model appear in Fig. 4 where the out-of-plane
described above for our nozzle flow simulations.
coordinates of the 3D streamline are obtained by computing the angle
Selection of the RSC modelling parameters is required for its appli­
of rotation θ from
cation to our nozzle flow problem. Wang, et al. [18] proposed that the
Z t
RSC model was effective for computing fiber orientation with CI ¼ 0.01,
θ¼ _
θdt; (18)
0 κ ¼ 0.1, and ξ ¼ 1 for transient simple shear flows such as that seen in the
injection molding process and some extrusion scenarios. Our computa­
where θ_ ¼ vrθ , and vθ and r are obtained along surface projection tions assume the same values of the RSC parameters.
streamline. Equation (18) is calculated through a trapezoidal rule nu­ Results appearing in Figs. 6 through 9 are the A11 and A33 compo­
merical integration approach [42]. nents of computed fiber orientation tensors along the flow domain
Fig. 5 shows the surface projection streamlines as computed by streamlines of the straight flow and swirling flow models. In this section,
Polyflow for both the straight flow and swirling flow simulations. It can “Nozzle CZS” and “Nozzle CZE” refer to Nozzle Convergent Zone Start
be seen from Fig. 5 that the swirling flow model yields much different and Nozzle Convergent Zone End. Our results show that the initial
path through the flow domain than that of 2D straight flow. Note that steady orientation state of the straight flow has more fibers that align
the swirling flow streamlines appear closer to the center of the flow than along the principal flow direction (e.g., x3 ), while that of the swirling
those of straight flow in the region between the screw tip and nozzle flow align normal to the flow direction due to the swirling effects see
entrance. In a steady state flow problem such as the simulations involved Fig. 7.
in our study, the streamlines are the same as the pathlines that particles It is generally known that shear dominated flows tend to align fibers
would follow throughout the flow domain. We would, therefore, expect along flow direction while elongational flows promote fiber alignment

6
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

Fig. 5. Surface projections of the flow streamlines: (A) straight flow and (B) swirling flow simulation. (See online publication for full color image.)

Fig. 6. Fiber orientation tensor A11 component through the straight flow
domain (color indicates streamline as defined in Fig. 5A). (See online publi­ Fig. 7. Fiber orientation tensor A33 component through the straight flow
cation for full color image.) domain (color indicates streamline as defined in Fig. 5A). (See online publi­
cation for full color image.)
in the principle stretching direction. Heller, et al. [13] illustrates the
correlation between the strain rate fields and the resulting orientation v � �
tensors for 2D axial extrusion nozzle flow, which is similar to the straight et ¼ ¼ e1t ; e2t ; e3t ; (19)
jjvjj
flow considered in this study (cf. Fig. 5A). To better illustrate the effect
of the velocity gradient fields on the fiber orientation kinetics in 3D where v is the velocity vector at the point of interest given in global
swirling streamlines, Polyflow computed velocity gradients results are Cartesian coordinates and jjvjj indicates the magnitude of the velocity
rotated into a tangential-normal coordinate system for each point along vector. In this calculation procedure, velocity gradients rv are first
swirling streamline, as shown in Fig. 10. computed in cylindrical coordinates in Polyflow which are transformed
The tangential-normal coordinate system at each point along a to global Cartesian coordinates using Equation (14). These transformed
streamline is defined by first computing the tangential unit vector (et ) of velocity gradients are further rotated to the tangential-normal co­
the point as ordinates using the rotation matrix

7
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

Fig. 10. Tangential-normal coordinates of a point on streamline 15 of the


swirling flow simulation.
Fig. 8. Fiber orientation tensor A22 component through the swirling flow
domain (color indicates streamline as defined in Fig. 5B). (See online publica­
tion for full color image.) Notice, the x1 axis in the Cartesian coordinates is rotated to the di­
rection of et , and then the x2 , x3 axes represent the directions of en . Here,
dvt
det
refers to the normal strain rate field. In addition, it is important to
note that the choice of en is not unique but must remain in the plane
normal to the et . Consequently, the magnitude of de dvt
n
is computed as
�� �� vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
�� �� u ! !2
��dv �� u dv 2 dvt
�� t �� t
(23)
t
�� �� ¼ þ ;
��den ��
�� �� de1n de2n

to represent the variance of the shear strain field along the flow domain.
Following Equations (19)-(23)), the strain rate tensors components
along streamlines 1, 8 and 15 (c.f. Fig. 5B) of the swirling flow results are
computed and plotted in Figs. 11–13. These streamlines are chosen as
examples since they represent the core (streamline 1), intermediate
(streamline 8) and boundary (streamline 15) regions of the flow domain.
It is seen that the strain rate fields initially experience variations in the
flow upstream of the screw tip. At this point, the shear strain rate for all
three streamlines is much higher than that of the normal strain rate.
Hence, the resulting shear-dominate flow directs the principle orienta­
tion direction near the screw ending position as appearing in Figs. 8 and
9. In addition, it can be seen that the A33 component of the swirling flow
Fig. 9. Fiber orientation tensor A33 component through the swirling flow in Fig. 9 decreases considerably and then recovers to high value as
domain (color indicates streamline as defined in Fig. 5B). (See online publica­ polymer melt passes through the shear dominated nozzle convergent
tion for full color image.) zone. Specifically, the larger than shear strain is, the more significant
changes the diagonal components of the orientation tensor experiences.
1
2
0 0
32
cosðβ2 Þ sinðβ2 Þ 0
3 For instance, the magnitude of the shear strain rate of streamline 15 is
R ¼ 4 0 cosðβ1 Þ sinðβ1 Þ 54 sinðβ2 Þ cosðβ2 Þ 0 5; (20) higher than that of streamline 8, and thus the change (e.g., decreasing)
0 sinðβ1 Þ cosðβ1 Þ 0 0 1 of A33 component along streamline 15 is much more significantly than
that along streamline 8, as appearing in Fig. 9. Alternatively, the normal
where strain rate of streamline 1 is higher than its shear strain rate magnitude,
� � � � and thus the relatively higher elongational effects of the flow increases
β1 ¼ cos 1
e3t ; β2 ¼ π 2 sin 1
e2t sinðβ1 Þ (21) the A33 component of streamline 1. Finally, as the flow exits the nozzle,
Then final rotated rv equates is evaluated in the tangential-normal the normal strain rate of streamline 1 and 8 increase significantly while
coordinate at each point from that of streamline 15 decreases. The positive elongation flows in
streamline 1 and 8 increase the fiber alignment along flow direction
2 dvt dvt 3
dvt while the negative elongation decreases the A33 component in stream­
6 det de1n de2n 7 line 15. As the melt flow exits the nozzle, the shear strain fields (e.g,
6 7 �� ��
6 7 �� dv ��
6 dv1 dv1n dv1n 7 �� t ��) of all three selected streamlines experience a significant incre­
6 7
e1n
(22) ��den ��
2
rvet en
¼ R ðrvÞR ¼ 6 n
T
7;
6 det de1n de2n 7
6
6
7
7 ment and then vanish to zero, which causes the diagonal components of
4 dv2
n dv2n dv2n 5 the orientation tensor. This variation causes the A33 component of the
det de1n de2n swirling flow at the nozzle exit decreases greatly and then recovers back.
In detail, the change of A33 component along streamline 15 is most

8
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

significant among the three selected streamlines due to its biggest numerically and experimentally. For example, simulations by Heller,
magnitude of strain rate change as seen through Figs. 11B, 12B and 13B. et al. [14] predicted the E33 of a deposited bead as 6.9 GPa using a
Once fiber orientation values are computed downstream of the Newtonian creeping flow simulation and classical Advani-Tucker model.
nozzle at the flow exit, steady state fiber orientation tensors may be Russell, et al. [15] computed a value of 4.82 GPa using the same flow
computed for both the straight and swirling flow models. Fig. 14 shows model as employed in Heller’s work and RSC orientation model with
computed values of the orientation tensor diagonal components for each same RSC model parameters as in our study. Wang and Smith calculated
streamline at the flow end. Note, the fiber orientation is expected to property value of E33 of a vertical extrudate as 6.57 GPa using the PTT
reach a steady state orientation prior to reaching the flow end which is, rheology model and Advani-Tucker orientation model [16]. In the cur­
therefore, assumed to represent the fiber orientation within the solidi­ rent study, our swirling-flow-predicted E33 value is 8.13 GPa with the
fied extruded polymer composite [13–16]. From the data appearing in PTT rheology model and the RSC-model. It can be seen that our result is
Fig. 14, it can be seen that the fiber alignment along the flow direction is higher than the previously numerical analyses. On the other hand, Duty,
higher than those along transverse directions for both the straight flow et al. [7] reported that the elastic modulus of a 13% vol. CF-ABS
and swirling flow. which agrees with trends seen in prior experimental deposited bead is 7.24 GPa through their experimental tests. Kunc ,
and numerical studies [7,13–16,22,23]. However, the fiber orientation et al. and Love, et al. measured the elastic modulus of a 13% vol. CF-ABS
pattern differs along the radial direction of the bead when comparing deposited bead experimentally and found that the mechanical tensile
the swirling flow and the straight flow results. For the swirling flow modulus loaded parallel to the printing direction of a 13% vol. CF-ABS
results, it is seen that the A33 component increases gently from the outer deposited bead is 8.15 GPa [22] and 8.91 GPa [23], respectively. Our
boundary toward the core of the flow, and the results of the straight flow reported E33 value obtained for the swirling flow model is 8.13 GPa
decreases initially after leaving the outer boundary and then increases which shows good agreement with most of the experimental data.
after the intermediate region of the flow. In addition, the A33 component Although, our numerical method does not match the entire experimental
of the swirling flow is higher than that of the straight flow, over the procedure and sampled materials in referred literature [7,22,23]
entire flow exit. exactly, the overall favorable comparison with prior corresponding
Finally, the effect of the swirling flow on the material properties of an experimental work supports our proposed algorithm.
extruded bead is considered. The material stiffness of a CF-ABS com­ Finally, our finite element swirling flow model is meshed in 2100 4-
posite is evaluated based on the steady-state orientation tensor com­ node quadrilateral elements with 2217 nodes. Additional models having
ponents appearing in Fig. 14 using the orientation homogenization two different meshes of the same flow domain (not shown here) were
method described through Equations (15)–(17)). The elastic properties also studied. These alternate models included a finer mesh (4745 nodes
of the fiber and matrix phase materials are given in Table 2, where E and and 4567 elements) and a coarser mesh (1347 nodes and 1240 elements)
ν refer to the Young’s modulus and the Poisson’s ratio. The fiber volume as compared to our intermediate mesh that was used to compute all of
faction of the composite is assumed to be 13%, which is commonly seen the results given above. It is found that the computed results of effective
in large area additive manufacturing applications [7,22,23]. The fiber
elastic constants E11 , E22 , and E33 using the flow kinematics solved with
aspect ratio used in the stiffness computation is 15 which is the same as
the intermediate mesh and the fine mesh are within 1% absolute relative
that used in other studies [13,14]. To obtain a clear view of the property
difference [42]. Therefore, we consider the intermediate mesh, that we
enhancement for the composite extrudate, we pull out the effective
are currently using, is computationally efficient and reliable for our
elastic constants from the computed stiffness and the results are given
study, such that a finer mesh will not increase much accuracy but yield
through Table 3 (c.f. [44]. for the transformation between material
considerate computational cost and numerical instability.
stiffness and elastic constants). The computed data shows that the elastic
constant E33 is much higher than those along transverse directions for 4. Conclusions
both the swirling flow and the straight flow simulation cases which is
due to the high principal fiber alignment parallel to the flow direction. This paper provides a numerical approach to investigate the screw
Specifically, the E33 value predicted by applying a swirling flow exhibits swirling effects on the predicted fiber orientation and associated effec­
a significant increment of 21% comparing to the same property pre­ tive elastic properties of a polymer composite extrudate in a LAAM
dicted by using a straight flow, which demonstrates that the screw polymer deposition process. A weakly coupled analysis is performed to
swirling has a considerable effect on fiber alignment pattern within an simulate the polymer melt flow and fiber orientation through a typical
extruded polymer composite strand. LAAM extrusion nozzle. The rheology properties of a molten 13% CF-
In addition, prior literature reported the tensile modulus of a 13% ABS melt is fitted with a 5-mode viscoelastic PTT fluid model. The
CF-ABS in the print bead direction (e.g., E33 appearing in Table 3) both finite element method is employed to simulate the polymer melt flow

Fig. 11. Swirling flow strain rate fields in tangential-normal frame for streamline 1: (A) normal strain rate; (B) magnitude of shear strain rate.

9
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

Fig. 12. Swirling flow strain rate fields in tangential-normal frame of streamline 8: (A) normal strain rate; (B) magnitude of shear strain rate.

Fig. 13. Swirling flow strain rate fields in tangential-normal frame of streamline 15: (A) normal strain rate; (B) magnitude of shear strain rate.

Table 2
Elastic properties of the phase materials of a 13% vol. CF-ABS.
Material E (GPa) G (GPa) ν

ABS matrix 2.25 0.83 0.35


Carbon fiber 230 96 0.2

Table 3
Computed effective elastic constants of a 13% vol. CF-ABS extrudate.
Flow model E11 (GPa) E22 (GPa) E33 (GPa) G12 (GPa) G13 (GPa)

Swirling Flow 3.19 3.43 8.13 1.13 1.31


Straight Flow 3.46 3.54 6.73 1.25 1.52

Flow model G23 (GPa) ν12 ν13 ν23

Swirling Flow 1.50 0.44 0.39 0.38


Straight Flow 1.51 0.40 0.40 0.38

streamlines and elongational flows that align fiber particles traveling


Fig. 14. Steady state fiber orientation tensor diagonal components at stream­ along pathlines through the flow domain which differs from results
lines ends.\ (See online publication for full color image.) computed for non-swirling flows. Additionally, the evolution of the
second order orientation tensors through the flow domain exhibits
including the screw tip, nozzle, and a strand of material beyond the significantly different trends depending on if the consideration of
nozzle exit where die swell occurs. The Wang-O’Gara-Tucker RSC fiber swirling kinematics is included. In addition, the final orientation pattern
orientation model is applied to solve the fiber orientation state with of the extruded composite at the flow domain exit is different locally
computed kinematics of the uncoupled flows. over the diameter of the bead when comparing swirling flow and
It is found that the swirling motion of the flow yields rotational straight flow models. Based on the computed steady-state orientation
tensors with a prescribed orientation modelling parameter set, the

10
Z. Wang and D.E. Smith Composites Part B 177 (2019) 107284

effective elastic constants of a 13% vol. CF-ABS are computed. The [15] Russell T, Heller B, Jack DA, Smith DE. Prediction of the fiber orientation state and
the resulting structural and thermal properties of fiber reinforced additive
predicted principal tensile modulus from the swirling flow is 21% higher
manufactured composites fabricated using the Big area additive manufacturing
than that yielded by a straight flow. Moreover, the swirling-flow pre­ process. J. Compos. Sci. 2018;2(2):26.
dicted data show more favorable agreement with previously published [16] Wang Z, Smith DE. Rheology effects on predicted fiber orientation and elastic
experimental data on a similar material system than the results pre­ properties in large scale polymer composite additive manufacturing. J. Compos.
Sci. 2018;2(1):10.
dicted by the straight flow. [17] Tucker III CL. Flow regimes for fiber suspensions in narrow gaps. J Non-Newtonian
Fluid Mech 1991;39(3):239–68.
Acknowledgements [18] Wang J, O’Gara JF, Tucker III CL. An objective model for slow orientation kinetics
in concentrated fiber suspensions: theory and rheological evidence. J Rheol 2008;
52(5):1179–200.
The authors would like to thank the financial support offered by [19] Advani SG, Tucker III CL. The use of tensors to describe and predict fiber
Baylor University during this study. We would also like to thank orientation in short fiber composites. J Rheol 1987;31(8):751–84.
[20] Tandon GP, Weng GJ. The effect of aspect ratio of inclusions on the elastic
Strangpresse for the donation of a Model 19 extruder that allows us to properties of unidirectionally aligned composites. Polym Compos 1984;5(4):
study the large-scale polymer deposition both through lab experiments 327–33.
and fluid modelling. We would also like to acknowledge many beneficial [21] Jack DA. Advanced analysis of short-fiber polymer composite material behavior.
PhD Thesis. University of Missouri–Columbia; 2006.
discussions with Dr. David Jack at Baylor University on fiber orientation [22] Love LJ, et al. The importance of carbon fiber to polymer additive manufacturing.
modelling and simulation, and experimental methods in polymer J Mater Res Sep. 2014;29(17):1893–8.
processing. [23] Kunc. Advances and challenges in large scale polymer additive manufacturing. In:
15th SPE automot compos conf novi MI; 2015. p. 27.
[24] VerWeyst BE, Tucker III CL. Fiber suspensions in complex geometries: flow/
Appendix A. Supplementary data orientation coupling. Can J Chem Eng 2002;80(6):1093–106.
[25] Dinh SM, Armstrong RC. A rheological equation of state for semiconcentrated fiber
Supplementary data to this article can be found online at https://doi. suspensions. J Rheol 1984;28(3):207–27.
[26] Thien Nhan Phan, Tanner Roger I. A new constitutive equation derived from
org/10.1016/j.compositesb.2019.107284. network theory. J Non-Newtonian Fluid Mech 1977;2(4):353–65.
[27] ANSYS, Inc, “ANSYS Polyflow user’s guide Release 15.0.”.
References [28] Cox WP, Merz EH. Correlation of dynamic and steady flow viscosities. J Polym Sci
1958;28(118):619–22.
[29] Xiao Karen, Tzoganakis Costas. Rheological properties of HDPE-wood composites.
[1] Brenken B, Barocio E, Favaloro A, Kunc V, Pipes RB. Fused filament fabrication of In: Annual technical conference - ANTEC, conference proceedings, nashville, TN;
fiber-reinforced polymers: a review. Addit. Manuf 2018;21:1–16. 2003.
[2] Huang B, Singamneni S. Raster angle mechanics in fused deposition modelling. [30] ANSYS, Inc, “ANSYS Polymat user’s guide Release 15.0.” .
J Compos Mater 2015;49(3):363–83. [31] Baird Donald G, Collias Dimitris I. Polymer processing: principles and design. John
[3] Gray RW, Baird DG, Helge Bøhn J. Effects of processing conditions on short TLCP Wiley and Sons; 2014.
fiber reinforced FDM parts. Rapid Prototyp J 1998;4(1):14–25. [32] Jeffery GB. The motion of ellipsoidal particles immersed in a viscous fluid. Proc R
[4] Gray RW, IV, Baird DG, Bøhn JH. Thermoplastic composites reinforced with long Soc Lond 1922;102(715):161–79.
fiber thermotropic liquid crystalline polymers for fused deposition modeling. [33] Folgar F, Tucker III CL. Orientation behavior of fibers in concentrated suspensions.
Polym Compos 1998;19(4):383–94. J Reinf Plast Compos 1984;3(2):98–119.
[5] Masood SH, Song WQ. Development of new metal/polymer materials for rapid [34] Advani SG, Tucker III CL. Closure approximations for three-dimensional structure
tooling using fused deposition modelling. Mater Des 2004;25(7):587–94. tensors. J Rheol 1990;34(3):367–86.
[6] Zhong W, Li F, Zhang Z, Song L, Li Z. Short fiber reinforced composites for fused [35] De Frahan HH, Verleye V, Dupret F, Crochet MJ. Numerical prediction of fiber
deposition modeling. Mater Sci Eng A 2001;301(2):125–30. orientation in injection molding. Polym Eng Sci 1992;32(4):254–66.
[7] Duty CE, et al. Structure and mechanical behavior of Big area additive [36] Chung DH, Kwon TH. “Fiber orientation in the processing of polymer composites.
manufacturing (BAAM) materials. Rapid Prototyp J 2017;23(1):181–9. Korea-Aust. Rheol. J. 2002;14(4):175–88.
[8] Ke Z, Zhongqi H, Shupan Y, Wanghua C. Numerical simulation for exploring the [37] Cintra Jr JS, Tucker III CL. Orthotropic closure approximations for flow-induced
effect of viscosity on single-screw extrusion process of propellant. Procedia Eng. fiber orientation. J Rheol 1995;39(6):1095–122.
2014;84:933–9. [38] Heller BP. Effects of nozzle geometry and extrudate swell on fiber orientation in
[9] Canevarolo SV, Babetto AC. Effect of screw element type in degradation of fused deposition modeling nozzle flow. Master’s Thesis. Baylor University; 2015.
polypropylene upon multiple extrusions. Adv. Polym. Technol. J. Polym. Process. [39] Mori T, Tanaka K. Average stress in matrix and average elastic energy of materials
Inst. 2002;21(4):243–9. with misfitting inclusions. Acta Metall 1973;21(5):571–4.
[10] Kelly AL, Brown EC, Coates PD. The effect of screw geometry on melt temperature [40] Halpin JC. Stiffness and expansion estimates for oriented short fiber composites.
profile in single screw extrusion. Polym Eng Sci 2006;46(12):1706–14. J Compos Mater 1969;3(4):732–4.
[11] Ramani K, Bank D, Kraemer N. Effect of screw design on fiber damage in extrusion [41] Laws N, McLaughlin R. The effect of fibre length on the overall moduli of
compounding and composite properties. Polym Compos 1995;16(3):258–66. composite materials. J Mech Phys Solids 1979;27(1):1–13.
[12] Nixon J, Dryer B, Chiu D, Lempert I, Bigio DI. Three parameter analysis of fiber [42] Chapra SC. Applied numerical methods. With MATLAB for engineers and scientists.
orientation in fused deposition modeling geometries. Annu. Tech. Conf. - ANTEC 2012.
Conf. Proc. Jan. 2014;2:985–95. [43] Tucker III CL, Liang E. Stiffness predictions for unidirectional short-fiber
[13] Heller BP, Smith DE, Jack DA. Effects of extrudate swell and nozzle geometry on composites: review and evaluation. Compos Sci Technol 1999;59(5):655–71.
fiber orientation in Fused Filament Fabrication nozzle flow. Addit. Manuf. Oct. [44] Gurtin ME. “The linear theory of elasticity,” in Linear theories of elasticity and
2016;12:252–64. thermoelasticity. Springer; 1973. p. 1–295.
[14] Heller Blake P, Smith Douglas E, Jack David A. Planar deposition flow modeling of
fiber filled composites in large area additive manufacturing. Additive
Manufacturing 2019;25:227–38.

11

You might also like