You are on page 1of 13

Additive Manufacturing 43 (2021) 102006

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Research Paper

A statistical homogenization approach for incorporating fiber aspect ratio


distribution in large area polymer composite deposition additive
manufacturing property predictions
Zhaogui Wang a, Douglas E. Smith b, *, David A. Jack b
a
Department of Mechanical Engineering, Naval Architecture and Ocean Engineering College, Dalian Maritime University, Dalian, Liaoning 116026, China
b
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: Assessing the material stiffness of fiber reinforced polymer composites deposited in Large Area Additive
Polymer composites Manufacturing (LAAM) is needed to define the process-structure-property mapping for the LAAM technology.
Large area additive manufacturing While the screw-extrusion-based LAAM systems yield a distribution of fiber aspect ratio (i.e., length to diameter
Fiber aspect ratio distribution
ratio for cylindrical inclusions) within deposited beads, most composite micromechanical models ignore the fiber
Fiber orientation homogenization
geometry variation and instead assume a single value of fiber aspect ratio. This paper presents a statistics-based
Elastic properties
homogenization approach for including the fiber aspect ratio distribution in the prediction of the elastic prop­
erties of an extruded polymer composite bead. The fiber length distribution of a 13 wt% Carbon Fiber reinforced
Acrylonitrile Butadiene Styrene (CF-ABS) processed through a LAAM deposition system is measured using high
resolution optical microscopy. The Weibull probability distribution function is employed to statistically describe
the measured values. The fitted probability density function is then incorporated into a fiber orientation ho­
mogenization approach to compute the variability of the elastic properties of the extruded composite. Elastic
properties predicted by our proposed method are shown to differ from those presented in prior studies that
ignored fiber length variability. The fiber aspect ratio reduction from fiber length attrition during the LAAM
single screw-extrusion is shown to decrease the predicted flow-direction effective elastic modulus by 7%. Elastic
moduli computed using our measured fiber aspect ratio distribution and proposed homogenization approach
compare well to reported data from previously associated experimental studies.

1. Introduction a nonuniform fiber aspect ratio distribution. Hausnerova et al. [2]


showed that the high shear stress imposed on the polymer composite
Chopped carbon fiber reinforced polymer composites have emerged melt by the single screw rotation has a direct effect on degrading the
as a popular choice for the Large Area Additive Manufacturing (LAAM) geometry of the reinforced fibers. In addition, Aigner et al. [3] studied
polymer deposition process, due in part to their superior mechanical and fiber breakage of a glass fiber polymer composite using X-ray computed
thermal properties as compared to virgin thermoplastics. In a typical tomography and found that the single screw extruder compounding and
LAAM process, thermoplastic pellets (filled or unfilled) are melted, extrusion process resulted in a nearly 50% reduction in the maximum
extruded, and deposited onto a heated platform following a computer- fiber length compared to the manufacturer’s specification sheet. Several
aided-designed deposition path, layer-by-layer, to form a three- other studies have investigated the degradation of fiber length of long
dimensional object. LAAM deposition has a distinct advantage fiber reinforced composites in an injection molding application where
compared to other polymer composite processing in that the bead di­ the effects of various processing conditions (e.g., melting temperature
rection can be defined (within limits) to achieve a desired fiber orien­ [4], residence time [5], and molding pressure [6]) were considered. It is
tation during part fabrication [1]. As the pelletized feedstock is not surprising that short fiber polymer composites processed with the
processed through the LAAM single screw extruder, carbon fiber LAAM system produce a nonuniform distribution of fiber aspect ratio.
breakage is known to occur causing a reduction in fiber aspect ratio and Models of fiber length attrition have been used to study the effect of

* Corresponding author.
E-mail addresses: zhaogui_wang@dlmu.edu.cn (Z. Wang), douglas_e_smith@baylor.edu (D.E. Smith), david_jack@baylor.edu (D.A. Jack).

https://doi.org/10.1016/j.addma.2021.102006
Received 23 November 2020; Received in revised form 2 April 2021; Accepted 12 April 2021
Available online 29 April 2021
2214-8604/© 2021 Elsevier B.V. All rights reserved.
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

Fig. 1. (a) LAAM process with a single screw extruder; (b) molten feedstock extruded through extruder nozzle.

processing on the state of fibers in the final product. Jin and Wang [7] homogenization approach where the fiber orientation state is obtained
and Phelps et al. [8] proposed models to predict the fiber length through melt flow simulation of the LAAM extrusion process. By the
reduction in Long Fiber Thermoplastics (FLT) during the injection proposed method, elastic properties of the sample composite materials
molding process. Results obtained by Phelps show a good agreement with nonuniformly distributed fiber aspect ratios are computed.
with corresponding experiments on glass-fiber/polypropylene FLT
molding. Bereaux et al. [9] conducted numerical simulations on fiber 2. Methodology
breakage in a single screw extrusion process and modeled the sensitivity
of fiber length to different screw processing parameters such as the Orientation homogenization is widely used to compute elastic
material viscosity, barrel temperature, and screw geometry. Unfortu­ properties of short fiber polymer composites (e.g., see [12–16]). This
nately, models from these earlier works have yet to be applied to approach incorporates specific flow-induced fiber orientation within the
describe fiber length reduction in LAAM processes. composite material which is homogenized using micromechanical
The length of fiber fillers has a significant influence on the material models that are commonly employed to predict the material stiffness of
properties of the polymer composite processed through screw extrusion. unidirectional aligned short fiber composites [16]. The methodology
Bayush et al. [10] studied the fiber length distribution of a natural fiber presented herein is based on this same orientation averaging theory.
reinforced polypropylene and concluded that maintaining a critical fiber Additionally, we consider nonuniformly distributed fiber aspect ratios,
length and minimizing the fiber breakage enhances mechanical and which extends the more common approach used in micromechanical
dynamic properties of the overall compound. Similarly, Gamon et al. models that typically assume a single value for fiber aspect ratio as an
[11] showed that longer fibers improve the flexural strength of the input. This section presents the polymer composite melt flow domain
extruded composite. In addition, Inoue et al. [12] reported that fiber used in this study, the method used to achieve the fiber orientation state
length directly impacts the mechanical properties of the mixed com­ of the flow field, the statistical model for describing the fiber aspect ratio
posite and the effect of the screw design on fiber breakage and disper­ distribution, and our enhanced orientation homogenization method.
sion. Additionally, Hausnerova et al. [2] showed that the shear viscosity
and die swell of filled polymers would decrease in high shear rate screw 2.1. Flow-induced fiber orientation
extrusion due to fiber length reduction and polymer matrix degradation.
Accordingly, it is important to consider the effect of fiber aspect ratio LAAM systems process polymer composite pellets via a single screw
distribution on the prediction of elastic properties of LAAM-fabricated extruder (cf. Fig. 1a). As molten polymer composite is extruded verti­
composites, especially since prior literature that predicts the material cally through the nozzle of the single screw extruder, (see e.g., Fig. 1b),
properties of LAAM deposited beads assumes a constant fiber aspect the kinematics of the flow depend on the design of the nozzle orifice
ratio [12–15]. among other things. The resulting velocity gradients within the nozzle
As mentioned above, prior literature addressed the effects of fiber flow cause the suspended fibers to undergo a change in orientation. The
length attrition during single screw extrusion processing on the material fiber orientation within the processed composite extrudate is important
properties of extruded composites [12–15]. Compared to injection since the direction of fiber alignment within the extrudate has a signif­
molding, LAAM extrusion does not include processing features such as icant influence on the mechanical properties of the printed material
screw motion and high pressure flow through runners and gates which [18]. Note that the material deposition process shown in Fig. 1a also
can cause fiber breakage. However, a specific quantitative method for changes the fiber orientation state within the composite as a shearing
evaluating the influence of fiber aspect ratio distribution (i.e., resulted force is applied during deposition when the deposited material contacts
from fiber length attrition during processing) on elastic properties of the substrate. This work focuses on the polymer melt flow within the
polymer composites fabricated by extrusion deposition additive extruder nozzle as well as the die swell of the composite extrudate, but
manufacturing has yet to be considered, to our best knowledge. In this ignores the turning flow during deposition. Analysis of the fiber orien­
paper we propose a statistics-based homogenization approach to quan­ tation state as the polymer turns during deposition (as discussed in a
tify the effect of fiber aspect ratio distribution of LAAM-processed separate work by the authors [15]) is beyond the main scope of the
composites on predicted elastic properties. To demonstrate our meth­ presented study.
odology, we measure the fiber length distribution within 13 wt%
CF-ABS pellet feedstock as well as the same in a LAAM-deposited bead. 2.1.1. Flow domain of interest
By assuming a constant carbon fiber diameter within the polymer Wang and Smith [16] reported that the unique swirling kinematics
composite feedstock, we obtain fiber aspect ratio distributions directly generated from the single screw LAAM extruder significantly affects
from measured fiber length distributions. The Weibull probability dis­ fiber orientation within a LAAM nozzle-extrudate flow. For the work
tribution function statistical model is applied to characterize the vari­ presented here, we adopt the same model as in [16] for the flow simu­
ability in the experimental aspect ratio data. The resulting fitted fiber lation. The geometry of the flow domain includes a short section of the
aspect ratio distribution model is incorporated into our fiber orientation extruder that connects the screw tip, the nozzle portion, and a short

2
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

Fig. 2. (a) Material extrusion with a Strangpresse Model 19 single screw extruder nozzle; (b) cross-section area of the extruder-nozzle connect region; (c) simplified
axisymmetric flow model.

where W and D are the flow vorticity and rate of deformation tensors,
respectively, which are the symmetric and antisymmetric parts of the
velocity gradient fields (∇v) computed in our Polyflow flow simulation
(i.e., W = 12(∇v − ∇vT ), and D = 12(∇v + ∇vT )). In the above, γ̇ is the
2
scale magnitude of the rate of deformation tensor, ζ = h( α) − 1
h(α)2 +1
is a
parameter that accounts for the geometrical effect of the fiber inclusions,
where h is the hydrodynamic aspect ratio and α = L/d is the geometric
aspect ratio of the cylindrical inclusion of diameter d and length L [20].
Further, CI is the fiber interaction coefficient for Folgar-Tucker isotropic
rotary diffusion model [21]. Components of the second and fourth order
fiber orientation tensors, A and A, respectively, in Eq. (2) are defined as
Fig. 3. Vector p(θ, ϕ) in a three-dimensional Cartesian coordinate. ∮
Aij = pi pj U(φ, ϕ)dS and
strand of a vertical extrudate immediately outside the nozzle exit (cf. ∮S (3)
Fig. 2a). Due to the axisymmetric nozzle geometry and assumed flow (cf. Aijkl = pi pj pk pl U(φ, ϕ)dS
S
Fig. 2b), we simplify the flow domain as a 2D axisymmetric model (cf.
Fig. 2c), which saves significant computational expense. As in prior where U(φ, ϕ) is the fiber orientation probability distribution function
related studies (e.g., see [12–15]), an isothermal, incompressible and S is surface of a unit sphere. It is common to use only the second
creeping flow is assumed within the flow domain, where the inertia order orientation tensor (A with components Aij ) to describe the fiber
effects, thermal effects and gravitational effects are ignored. In another orientation state within a composite part and approximate A with a
work by Wang and Smith [14], the effects of various rheological fluid closure approximation which expresses A as a function of A [17]. In this
flow models on predicted fiber orientation in a similar nozzle flow were study, we use the orthotropic fitted closure approximation from Ver­
studied, where the Phan-Thien-Tanner (PTT) model was found to yield Weyst and Tucker [22]. Due to the symmetry of the A tensor as well as
favorable estimations as compared to experimental work [19]. Hence, the normalization condition (i.e., trace of A is unity [17]), resulting in
the polymer melt flow in this paper is also modeled as a PTT fluid. The five independent components in the second order orientation tensor.
velocity field within the flow domain is computed with the finite Furthermore, fiber orientation measurements in injection molded
element suite, ANSYS-Polyflow. Additional detail on the flow simula­ plaques show that fiber alignment occurs slower than that simulated
tion, such as the mesh size, convergence, etc., can be found in [16]. using Eq. (2). In response, Wang et al. [23] modified the Advani-Tucker
orientation evolution Eq. (2) by introducing a strain reduction term as
2.1.2. Fiber orientation tensor [23].
To evaluate fiber orientation in the LAAM nozzle, we first consider
DA
the motion of a single rigid inclusion represented by the three- = (A∙W − W∙A) + ζ(D∙A + A∙D − 2[A + (1 − κ)(L − M:A)]:D)
dimensional unit vector p(θ, ϕ) (cf. Fig. 3 for the definition of angles θ Dt
and ϕ), written as + κCI γ̇(2I − 6A),
⎧ ⎫ (4)
⎨ sinθcosϕ ⎬
(1) whereL = Lijkl =
∑3 m m m m
∑3 m m m m
m=1 εm ni nj nk nl , M = Mijkl = m=1 ni nj nk nl ,
p(θ, ϕ) = sinθsinϕ ,
⎩ ⎭
cosθ m
which are written in terms of εm and ni , the m-th eigenvalue and the i-th
Advani and Tucker [17] proposed a widely used computationally component of the m-th eigenvector of A, respectively. In the above, κ
efficient method to describe the statistical behavior of a set of fibers by serves as a strain reduction factor which slows the growth rate of fiber
expressing the fiber direction through fiber orientation tensors A and A alignment. In flow fields that exhibit significant changes in fiber orien­
which evolve in the flow as tation, such as those found in LAAM nozzle flows, the computation of A
is strongly dependent on the value of κ in the Wang-O’Gara-Tucker
DA
= (A∙W − W∙A) + ζ(D∙A + A∙D − 2A:D) + 2CI γ̇(I − 3A), (2) model.
Dt

3
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

properties which would, unfortunately, lead to a tedious and almost


impossible task. Alternately, we employ statistics tools to model the
distribution of the length variation which significantly reduces compu­
tational cost in comparison. Ularych et al. [25] described the fiber length
distribution for a glass fiber polypropylene using a two-parameter
Weibull probability distribution function β which can be written as
(see also Fu et al. [26])
( )k− 1 [ ( )k ]
k L L
β(L|k, λ) = exp − , (5)
λ λ λ
Fig. 4. Inter-relations between the 3-D cylindrical coordinates applied for the
flow simulation and 3-D Cartesian coordinates applied for the fiber orienta­ where L is the fiber length, and k and λ are shape and scale parameters
tion evaluation. for the Weibull distribution function, respectively. In this study, we use
the ‘wblfit’ function in Matlab 2018a [27] (The Mathworks, Natick,
The fiber aspect ratio is an important factor in the computation of the Massachusetts, USA) to compute the parameters of a Weibull distribu­
second-order orientation tensor through the parameter ζ appearing in, e. tion function for sets of experimentally measured fiber length data.
g., the Wang-O’Gara-Tucker model (cf. Eq. (4)). Fortunately, as the fiber Weibull distribution parameters computed with ‘wblfit’ are considered
aspect ratio α becomes large, ζ→1 (e.g., α > 20, ζ > 0.99) and the effect to be the best-fit result for our measured data. We note that the goodness
of ζ becomes somewhat trivial when computing fiber orientation tensors of Weibull fit has a direct effect on the accuracy of the predicted elastic
with Eq. (4) [23]. In this study we assume that the effect of the fiber properties computed in this work. Further research addressing the
aspect ratio distribution on computed fiber orientation with Eq. (4) is quality of the Weibull fit is beyond the main scope of this research.
negligible. This is supported by our experimental data, where the mean Nguyen et al. [28] found that weight distribution of fiber length is
value of our measured fiber length to diameter ratio data (i.e., appear in more suitable than the number distribution expression (i.e., apply Eq.
results section: Table 3) is above 40. Therefore, we assume that the so­ (5) directly to the fiber length measurements) since most micro­
lution at the flow end of our computed flow domain as obtained from mechanical models employs volume averaging schemes and the weight
Wang and Smith [16], using ζ = 1 (as in [21,22]). distribution function is proportional to the volume of fibers, such that
The fiber orientation evaluation adopted here assumes a weakly- /∫ ∞
coupled formulation, where the flow fields are solved first ignoring βw (L) = β(L) β(L)Ldx, (6)
0
the effects of any fiber presence and the fiber orientation equation is
then computed based on the calculated flow kinematics. In our where βw (L) is referred as the weight averaged Weibull distribution
approach, finite element results of flow velocity and velocity gradient function and β(L) denotes the number averaged Weibull distribution
are interpolated along streamlines from the nodal solutions. Then the function. In addition, the number-averaged fiber length (i.e., Ln ) and the
Wang-O’Gara-Tucker fiber orientation equation (cf. Eq. (4)) is solved weight-averaged fiber length (i.e., Lw ) of a set of measurement can be
using an adaptive step size sixth order Runge-Kutta numerical integra­ written as [28].
tion scheme which provides values of the second order orientation ∫∞
tensors along all of flow streamlines. It is important to note here that we Ln = β(L)Ldx, (7)
employ a cylindrical coordinate system for solving the flow kinematics 0

(e.g., velocity tensor, v, and the velocity gradient, ∇v), while the
computation of the second order orientation tensor field (e.g., A in Eq. and
(4)) is solved under a 3-D Cartesian coordinate frame. In the following, ∫ ∞ /∫ ∞

subscripts 1, 2, 3 designate the 3D Cartesian reference frame when Lw = β(L)L2 dL β(L)LdL, (8)
0 0
defining orientation tensor components (e.g., A11 , A22 ) which relate to
coordinate directions in the cylindrical frame as shown in Fig. 4 (cf. 2.3. Enhanced orientation averaging homogenization method
Wang and Smith [16] for more detail as needed). Specifically, we
designate x3 as the axial direction of the nozzle, z, and x1 corresponds to The orientation homogenization approach (e.g., see [16,28,29]) is
r in the cylindrical coordinates. Note that effective elastic properties used here to predict the mean stiffness of a short fiber polymer com­
computations as shown in the following section are also performed in posite that has a known fiber orientation distribution (e.g., one which
the Cartesian coordinate frame. was obtained from the LAAM process simulation). This orientation ho­
Computed flow streamlines within the flow domain of interest and mogenization approach decomposes the composite into N subdomains,
fiber orientation tensor components appear in Figs. 5 and 6, respec­ each representing a uniaxially aligned fiber composite. Table 1 sum­
tively, where A33 indicates fiber alignment in the x3 direction. The marizes the various stiffness tensor definitions in our derivations that
location of the flow streamlines used in fiber length distribution calcu­ follows.
lation below appear in Fig. 5b (i.e., flow streamline ends indicate a Following the Voigt procedure (where the mean strain is assumed
steady state fiber orientation of the extruded composite). The numerical constant over a subdomain), Jack and Smith [30] write the sample mean
values of the computed A tensor in the polymer melt flow at the end of
material stiffness tensor Y
̃ ijkl of each subdomain from the corresponding
the extrudate where γ̇= 0 appear in Appendix A for completeness.

Table 1
2.2. Fiber length statistical model Definitions of stiffness tensors appearing in this section.
Variable Definition
Suspended fibers within the thermoplastic composite break during Cuni Stiffness tensor of a reference unidirectional fiber composite
qrst
extrusion due to shear stresses produced by the extruder screw causing a uni
Cqrst The mean Cuni
qrst over distributed values of α
reduction in fiber length [24], which is expected to have a significant
Čijkl Material point mean stiffness tensor of a subdomain inside a composite,
effect on the material properties of extruded composites. One approach that has a specific orientation distribution function
for assessing the effect of fiber length variation may be to evaluate the H
Čijkl The mean Čijkl over distributed values of α
contribution of each fiber within the processed composite on the desired

4
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

fiber stress field as where the fiber orientation state for a subdomain is characterized by the
N [ ] orientation tensor components, Aij and Aijkl . The material constants MI , I
1 ∑
̃ ijkl
Y =
uni
Qqi (θn , ϕn )Qrj (θn , ϕn )Qsk (θn , ϕn )Qtl (θn , ϕn )Cqrst ψ (θn , ϕn ) , = 1,…,5 in Eq. (11) are computed from
N n=1
M1 = C11 + C22 − 2C12 − 4C66 ,
(8)
M2 = C12 − C23 ,
M3 = C66 + (C23 − C22 )/2, (12)
where i,j,k,l,q,r,s,t ∈ {1,2,3} and summation is implied over the repeated
M4 = C23 ,
indices q,r,s,t, and θ and ϕ are defined as in Fig. (3). In Eq. (8), Qij (θ, ϕ) is M5 = (C22 − C23 )/2,
the rotation tensor written as
⎡ ⎤ In the above, the Cij are components of the stiffness tensor for the
sinθcosϕ sinθsinϕ cosθ associated unidirectional fiber filled composite (i.e., Cuni
qrst ) written in
Qij (φ, ϕ) = ⎣ − sinϕ cosϕ 0 ⎦ (9)
cosθcosϕ cosθsinϕ sinθ contracted notation. In the present study we compute Cij using the
modified Tandon-Weng micromechanics model, which has proven to be
uni
In the above, the fourth-order tensor Cqrst is the material stiffness an effective and sufficiently accurate formula in estimating the elastic
tensor of a reference unidirectional composite computed using the properties of discontinuous fiber reinforced composites [32]. Note that,
Tandon-Weng Equation [31] and Qij is used to rotate a sample fiber from the Tandon-Weng model assumes an averaged constant fiber aspect ratio
uni in the material stiffness prediction, such that
its local coordinates defined by (θ, ϕ) where the Cqrst is computed to the
global coordinate system. In this approach the sample mean material Cuni uni
ijkl = C ijkl (α), (13)
stiffness tensor Y
̃ ijkl is an unbiased estimator of the population mean for a
given distribution (see, e.g., [30]). For a specific set of angles (e.g., where α is the geometric fiber aspect ratio. The analytical form used in
{n ,ϕn }), as N→∞, the sample mean stiffness tensor approaches the ex­ this work for the modified Tandon-Weng model is given in Appendix B
pected value, or the point-wise mean stiffness tensor of a subdomain, for completeness. Bay and Tucker [33] stressed the importance of the
fiber length geometry in numerically evaluating the elastic properties of
Čijkl , given as
short fiber composites through the orientation homogenization method.

ˇ Furthermore, Nguyen et al. [28] showed that the stiffness matrix of an
(10)
uni
Cijkl = Qqi (θ, ϕ)Qrj (θ, ϕ)Qsk (θ, ϕ)Qtl (θ, ϕ)Cqrst ψ (θ, ϕ)dS
S2 aligned fiber composite with suspended fibers with a fiber aspect ratio of
given statistical distribution can be evaluated from
Jack and Smith [30] derived the point-wise mean stiffness tensor Čijkl
∫ ∞ uni
appearing in Eq. (10) which can be evaluated from fiber orientation Cijkl (α)βw (α)dα
(14)
uni
tensors (i.e., also proposed by Advani and Tucker [17]) as Cijkl = 0 ∫ ∞ w
0
β (α)dα

( ) ( ) ( )
Čijkl = M1 Aijkl + M2 Aij δkl + Akl δij + M3 Aik δjl + Ail δjk + Ajl δik + Ajk δil + M4 Aij δkl + M5 Aik δjl + Ail δjk , (11)

Fig. 5. Flow streamlines: (a) over the entire flow domain; (b) detail of the free vertical extrudate.

5
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

we apply same fiber length distribution function to all material points


over a cross section of the extrudate.

3. Results and discussion

Computed values of the mean stiffness tensors for a LAAM-extruded


polymer composite including the effect of fiber aspect ratio distribution
are presented below. The measured fiber length distribution of carbon
fibers in the 13 wt% CF-ABS appears first. Then the Weibull distribution
function is employed to model the statistical nature of the fiber length
measurements. The elastic properties of the composite extrudate are
then predicted through the averaging method (cf. Eq. (15)) presented
above.

3.1. Fiber aspect ratio distribution

The geometric aspect ratio of cylindrical fibers suspended within a


Fig. 6. Steady state fiber orientation tensor diagonal components at streamline thermoplastic polymer composite material is the length-to-diameter
ends. Note, r is the radial distance of each streamline at z = 0 and r0 is the ratio (i.e., α = Lf /Df , where Lf and Df are the length and diameter of
radius of the nozzle exit. a cylindrical fiber, respectively). For short rigid fibers, prior research [1,
2] considered the degradation of fiber length as a result of damages
incurred while the fiber suspension travels through the single
where βw (α) is a weight averaged probability distribution function that
screw-extruder. Russell and Jack [35] measured the fiber diameter of
can be expressed in terms of numbers or weight of fibers. From the
the same 13% CF-ABS used in this study which was processed through
typical two-step homogenization approach (i.e., the composite is parti­
the same single screw extruder. They found that the variation on the
tioned into pseudo-grains, each grain having the same volume fraction
measured diameter data is statistically insignificant. Hence, we assume
of fibers as the overall composite [34]), Eq. (11) gives the point-wise
here the fiber diameter is a constant in the feedstock pellets provided by
mean stiffness tensor Čijkl . Combining Eqs. (11) and (14) (i.e., co­ the supplier (PolyOne Corp., Avon Lake, OH, USA) and does not change
efficients MI are determined using the results of Eq. (14)), the effects of as fibers travel through the LAAM screw extrusion process. For constant
fiber aspect ratio distribution and the fiber orientation distribution of a fiber diameter, the fiber aspect ratio can be obtained by dividing
LAAM-processed composite can be included in the prediction of its measured fiber lengths obtained from prepared samples with the single
elastic properties as [28], measured mean diameter. Specifically, we set the fiber diameter as 7 μm
as given by Russell and Jack [35].

( ) ( ) ( )
(15)
H
Čijkl = M 1 Aijkl + M 2 Aij δkl + Akl δij + M 3 Aik δjl + Ail δjk + Ajl δik + Ajk δil + M 4 Aij δkl + M 5 Aik δjl + Ail δjk ,

uni
where MI are computed from Cijkl through Eq. (12). In this study, sample polymer composite materials were taken from
both feedstock pellets and deposited beads deposited using a Strang­
As described above, this study considers distribution functions of two
presse Model-19 extruder (mounted in the Baylor-LAAM system).
statistical variables associated with fiber reinforcements, namely the
Measured differences between the two material sets are used to assess
fiber aspect ratio distribution and the fiber orientation distribution,
fiber length attrition during the LAAM fabrication process. Fiber lengths
which are included in the homogenization of material properties of each
are measured with a Keyence VR-3000 wide-area 3D microscopy mea­
pseudo-grain. As such, it is important to determine if the sequence of the
surement system (Keyence Corp., Osaka, Japan). Fiber length distribu­
two homogenizations has an influence on the computed properties. In
tions are obtained in this study by measuring the length of 990 fibers in
our construction, the fiber orientation evaluation is independent from
ten separate trials (i.e., 99 fibers are measured in each trial, which is the
the geometric fiber aspect ratio (i.e., α appearing in Eqs. (13)–(15) is
maximum VR-3000 length-counting in one image). The detail of the
decoupled from (θ, ϕ) appearing in Eqs. (8)–(10)), which indicate that
imaging procedure and associated discussion are included in the
along any direction the probability of finding a fiber of a given length is
no different than in any other direction. The physical interpretation of
this decoupling assumption is that the orientation solution has negligible Table 2
dependence on the fiber aspect ratio for the material used in this study. Parameters fitted for the WPDFs.
2
As a result, we set ζ = h(α) − 1
h(α)2 +1
= 1 in our fiber orientation computations Sample λ k
which is justified for the relatively high aspect ratios measured in our Pellet 62.47 2.17
material and presented below, e.g., for h = 50, ζ = 0.9992 ≈ 1. Conse­ Bead 49.00 2.77
quently, altering the sequence of fiber orientation and fiber length dis­
tribution homogenizations would not yield an impact in the computed
elastic properties of the extruded composites. In addition, it is assumed Table 3
Mean and mode values of the fiber aspect ratio distribution functions.
that the probability distribution of the aspect ratio is independent of
location within the extrudate. Therefore, our computations assume that Sample αn αw
the entire bead shares the same fiber length distribution function, which Pellet 55.3 68.4
may likely be obtained due to an efficient compound mixing from the Bead 43.6 50.3
LAAM single screw extruder feeding mechanism. Under this assumption,

6
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

Fig. 7. Histograms of fiber aspect ratio data measured for fibers from: (a) pellet sample; (b) bead sample.

Fig. 8. Statistical modeling the fiber aspect ratio data for the pellet and bead samples.

dissertation by Wang [36]. The measured fiber aspect ratio distribution extrudate composite is taken from the simulation of LAAM nozzle flow
of pellet and bead samples appear in Fig. 7. Eqs. (5) and (6) are fitted to in Wang and Smith [16]. The elastic properties of the constituent ma­
the measured data to obtain two Weibull Probability Distribution terials for the 13% CF-ABS considered here are given in Table 4, where
Functions (WPDFs) separately describing the number averaged and E, G, ν, and ρ denote the axial modulus (i.e., Young’s Modulus), shear
weight averaged probability density functions of the fiber aspect ratios modulus, Poisson’s ratio and material density, respectively. Note, the
measured in both the pellet and bead samples, as shown in Fig. 8. Fitted Tandon-Weng Equations (cf. Appendix B) requires the volume fraction fv
parameters in the WPDFs are given in Table 2, where the given pa­ of fibers rather than weight fraction fw (i.e., given from the feedstock
rameters are for the number-averaged distribution functions (i.e., β(x) in manufacturer) which are related through
Eq. (5)) and weight-averaged parameter values are further transformed /[ ]
fv = (fw ρm ) fw ρm + (1 − fw )ρfi , (16)
through Eq. (6). In addition, the number-averaged and weight-averaged
fiber aspect ratios computed for each fitted WPDF using Eqs. (7) and (8)
where the subscripts “fi” and “m” indicate the fiber and matrix phases,
appear in Table 3. We note that the measured data is primarily used to
respectively. With information given in Table 4, the fiber volume frac­
test the proposed formula (i.e., Eq. (15)). A more rigorous measurement
tion of the applied composite system is 8.4%.
including details for quantifying the fiber length attrition in
To provide a clearer interpretation of our results, we calculate the
LAAM-processed composites can be found in Russell and Jack [35].

Table 4
3.2. Elastic properties estimation Elastic properties of the phase materials of the CF-ABS.
Material E (GPa) G (GPa) ν ρ(kg/m3 )
Elastic properties for 13 wt% CF-ABS extruded by the Baylor-LAAM
system are evaluated below where special attention is given to the effect ABS matrix 2.25 0.83 0.35 1040
Carbon fiber 230 96 0.2 1700
of fiber aspect ratio distribution. The fiber orientation state of the free

7
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

Fig. 9. Predicted axial moduli across the printed extrudate: (a) axial moduli; (b) shear moduli. Note, x3 direction refers to the direction of extrudate bead (cf. Fig. 4),
x1 is the radial direction, and x2 is the tangential direction.

flow direction of the bead.


Table 5
In addition to the material point evaluations, we integrate the
Predicted effective elastic constants of the 13 wt% CF-ABS with the bead sample
computed mean elastic constants over the cross-sectional area of the
measurement.
extrudate, from which the effective elastic properties of the extruded
Stiffness E11 E22 E33 G12 G23 G13 composite can be obtained for the bead. Due to the axisymmetry of the
3.12 3.53 10.30 1.11 1.69 1.30 assumed cylindrical extrudate geometry (cf. Fig. 2c), the bead averaged
Čijkl
αn

modulus, E33 for example, is computed from


Čijkl
αw 3.14 3.59 10.78 1.12 1.74 1.32
∫ 2π ∫ r s
H
Čijkl 3.13 3.55 10.44 1.11 1.70 1.31 1
E33 = 2 (E33 ∙r)drdθ, (18)
π rs 0 0

where rs is the radius of the swelled extrudate at the flow domain exit. In
Table 6 the above, r and θ refer to major axes of the associated cylindrical co­
Predicted effective elastic constants of the 13 wt% CF-ABS with different sample
ordinate system (cf. Fig. 4–1). Additionally, the integration appearing in
measurements.
Eq. (18) (as well as that in Eq. (15)) are numerically integrated through
Stiffness Sample E11 E22 E33 G12 G23 G13 the trapezoidal rule [38] in our work.
H Beads 3.13 3.55 10.44 1.11 1.70 1.31 In Table 5, we first compare our proposed averaging approach with
Čijkl
results computed from Eq. (11) where a single value of fiber aspect ratio
H
Čijkl Pellets 3.16 3.63 11.17 1.13 1.78 1.34
is applied. We separately consider both the number-averaged and
weight-averaged fiber aspect values of the fiber aspect ratio taken from
the bead sample in Table 3 and include superscripts αn and αw in the
equivalent mean elastic properties from the material stiffness matrix by
notation to respectively indicate these two results. Properties appearing
[37]. H
⎧ ⎫ in the row of “Čijkl ” refer to those solved by homogenizing both the fiber



1

ν21

ν31
0 0 0


⎪ orientation and fiber aspect ratio distributions. It can be seen that the
⎪ E
⎪ E E ⎪

⎪ ⎪
E33 properties exhibit most notable differences, where E33n is slightly
⎪ 11 22 33 ⎪ α

⎪ ν ⎪


⎪ 12 1 ν 32

⎪ H H

⎪ − − 0 0 0 ⎪
⎪ lower than E33 and E33 is 0.34 GPa higher than
αw
E33 .
⎪ E
⎪ E E ⎪

⎪ 11 22 33 ⎪


⎪ ν




⎪ Computed elastic constants appear in Table 6 which compares
⎪ ν23 1 ⎪

⎪−

13
− 0 0 0 ⎪⎪
⎬ computed elastic properties for the bead and pellet samples. Comparing
E E E
data in Table 6 provides a means to assess effect of fiber length attrition
− 1
(17)
11 22 33
̃
C =S= ̃
⎪ ⎪


⎪ 0 0 0
1
0 0 ⎪

⎪ during screw extrusion of the LAAM. For example, computed values of

⎪ G12 ⎪
⎪ H







⎪ E33 based on pellet fiber length distributions are 7% higher as compared
⎪ 1 ⎪


⎪ 0 0 0 0 0


⎪ the same property computed with bead sample data.
⎪ ⎪




G13 ⎪


⎪ Lastly, we see that the effective moduli evaluated by our proposed

⎪ 1 ⎪
⎪ H

⎪ 0
⎩ 0 0 0 0


⎭ method (i.e., elastic constants associated with the beads sample Čijkl )
G23
agree well with the experimentally reported values on the same material
system through a similar single screw extruder with slightly different
where S ̃ and C̃ are respectively, the compliance and stiffness matrices
processing parameters [19,39], which supports our proposed approach
written in contracted notation of a material. The subscripts {1,2,3} refer and associated results. We recognize the importance of an experimental
to the 3D Cartesian coordinate directions. study to validate our predicted data. However, experimental work to
Computed results of elastic constants at specific material points fabricate samples with controlled fiber length distributions would itself
across the extrudate of the CF-ABS composite appear separately in Fig. 9, be a challenging task beyond the scope of our research presented here,
where components of the fiber orientation tensor at the flow end cross which is to provide a numerical tool for establishing the
section appear in Appendix A and the weight-averaged fiber aspect ratio process-structure-property relationship of the emerging LAAM technol­
distribution function of the bead sample is taken from Fig. 8 (i.e., ogy. More rigorous validation could be addressed in separate in-depth
weight-averaged Weibull distribution: beads). It is seen that the pre­ studies.
dicted values of E33 are much higher than the transverse axial moduli,
which is a result of having a high proportion of the fibers align along the

8
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

4. Summary erties computed using pellet and bead fiber length distributions are
compared, identifying a difference of 7% between the respective
A statistics-based averaging scheme is incorporated into an orienta­ computed E33 values, which implies that decreases in fiber length during
tion homogenization method to compute the elastic constants of a 13 wt screw extrusion has a direct impact on the elastic properties of LAAM-
% CF-ABS extrudate for a large scale polymer deposition system. Here deposited composites. In addition, effective elastic properties evalu­
the focus is on the fiber aspect ratio distribution of the composite ma­ ated by the proposed method agree well overall with experimental data
terial and how variation in fiber aspect ratio affects the final predicted reported in related literature, however, validation of the statistical re­
results. The length of fibers in both pellet feedstock and LAAM-deposited sults presented here have yet to be performed.
bead are separately measured where the fiber diameter is an assumed
constant based on previous studies by the authors. A two-parameter
Weibull probability distribution function is employed to characterize Declaration of Competing Interest
the statistical trends of measured fiber length data. The fiber aspect ratio
distribution is then integrated into a fiber orientation homogenization The authors declare that they have no known competing financial
method to obtain the orientation-averaged elastic properties, where interests or personal relationships that could have appeared to influence
fiber orientation tensors of the composite extrudate are evaluated the work reported in this paper.
through an associated prior work. Due to the decoupling construction,
the sequence of homogenizing the fiber orientation distribution and the Acknowledgments
fiber aspect ratio distribution yields no effects on the predicted
properties. The authors would like to thank Department of Mechanical Engi­
Computed results of bead averaged E33 show a high dependence on neering at Baylor University for providing the financial support during
averaged values of fiber length distribution, which is due, in part, to the this research. We would also like to acknowledge Strangpresse Corpo­
high fiber alignment along the material extrusion direction. By ration for the donation of a Model 19 single screw extruder that allows
comparing effective elastic bead properties, it is seen that the property us to study the large-scale polymer deposition process both through lab
predicted by our proposed method is in between those predicted by experiments and fluid modeling. The first author also appreciates
single-fiber-aspect-ratio methods based on number-averaged and beneficial discussions with Mr. Timothy Russell from Baylor University
weight-averaged fiber aspect ratios. This result indicates that the single on fiber length measurement. Finally, we would like to acknowledge the
fiber aspect ratio method for predicting elastic properties may also be anonymous reviewers on our previous submission, whose critical and
acceptable in preliminary evaluations. Further, effective elastic prop­ knowledgeable comments greatly improve the quality of our study.

Appendices

A. Second order fiber orientation tensors at flow end

The second order fiber orientation tensor results at the end of flow streamlines in Fig. 3a as computed by Wang and Smith [16] appear in Table A.1.
Note, the radius of the swelled extrudate end is 1.910 × 10-3 m.

Table A.1
Flow-end second order orientation tensors results solved along flow streamlines.
r (m) A11 A12 A13 A22 A23 A33

3.619 × 10-4 0.1194 0.0075 -0.0160 0.0526 -0.0071 0.8280


4.838 × 10-4 0.0650 0.0168 0.0928 0.1008 0.0120 0.8342
6.022 × 10-4 0.0558 -0.0089 0.0965 0.1108 0.0028 0.8334
7.094 × 10-4 0.0569 -0.0144 0.1028 0.1138 0.0009 0.8293
8.355 × 10-4 0.0605 -0.0162 0.1169 0.1165 -0.0018 0.8230
9.560 × 10-4 0.0652 -0.0178 0.1305 0.1193 -0.0055 0.8155
1.065 × 10-3 0.0692 -0.0210 0.1373 0.1214 -0.0098 0.8094
1.178 × 10-3 0.0687 -0.0240 0.1309 0.1237 -0.0159 0.8076
1.290 × 10-3 0.0692 -0.0258 0.1279 0.1279 -0.0239 0.8029
1.393 × 10-3 0.0696 -0.0278 0.1232 0.1331 -0.0350 0.7973
1.500 × 10-3 0.0651 -0.0284 0.1019 0.1380 -0.0474 0.7969
1.608 × 10-3 0.0620 -0.0265 0.0807 0.1469 -0.0643 0.7911
1.716 × 10-3 0.0609 -0.0238 0.051 0.1613 -0.0886 0.7778
1.817 × 10-3 0.0655 -0.0203 0.0440 0.1876 -0.1169 0.7469
1.901 × 10-3 0.0712 -0.0056 0.0449 0.2041 -0.0943 0.7247

B. Tandon-Weng equations

Zhang [40] extended the Tandon-Weng micromechanical model for predicting the elastic constants of aligned short fiber composites, which has
been shown as a reliable numerical tool in numerical evaluations of the material properties of composite materials [12–15]. The model can be written
as [40].

9
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

Em
E11 = / , (B.1)
1 + vfrac (A1 + 2νm A2 ) A

and
Em
E22 = / , (B.2)
1 + vfrac [− 2νm A3 + (1 − νm )A4 + (1 + νm )A5 A] (2A)

and
⎡ ⎤
⎢ v ⎥
G12 = Gm ⎢
⎣1 + ( frac ) ⎥, (B.3)
Gm
Gf − Gm
+ 2 1 − vfrac S1212 ⎦

and
⎡ ⎤
⎢ v ⎥
G23 = Gm ⎢
⎣1 + Gm
( frac ) ⎥,
⎦ (B.4)
Gf − Gm
+ 2 1 − v frac S2323

and
νm A − νf (A3 − νm A4 )
ν12 = , (B.5)
A + νf (A1 + 2νm A2 )

and
E22
ν23 = − 1 + , (B.6)
2G23
For the above, fibers are aligned along x1 direction. E, G, ν are tensile modulus, shear modulus and Poisson’s ratio, respectively. vfrac refers to the
fiber volume fraction and subscripts m and f denote the matrix and fiber, respectively. In addition, the material constants A1 , A2 , A3 , A4 , A5 , and A,are
A1 = D1 (B4 + B5 ) − 2B2 , (B.7)

and
A2 = (1 + D1 )B2 − (B4 + B5 ), (B.8)

and
A3 = B 1 − D 1 B 3 , (B.9)

and
A4 = (1 + D1 )B1 − 2B3 , (B.10)

and
A5 = (1 − D1 )/(B4 − B5 ), (B.11)

and
A = 2B2 B3 − B1 (B4 + B5 ), (B.12)

and constants Bi are


B1 = vfrac D1 + D2 + (1 − vfrac )(D1 S1111 + 2S2211 ), (B.13)

and
B2 = vfrac + D3 + (1 − vfrac )(D1 S1122 + S2222 + S2233 ), (B.14)

and
B3 = vfrac + D3 + (1 − vfrac )(S1111 + S2211 + D1 S2211 ), (B.15)

10
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

and
B4 = vfrac D1 + D2 + (1 − vfrac )(S1122 + D1 S2222 + S2233 ), (B.16)

and
B5 = vfrac + D3 + (1 − vfrac )(S1122 + S2222 + D1 S2233 ), (B.17)

and constants Dj are written as


2(Gf − Gm )
D1 = 1 + (B.18)
Lf− Lm

and
L m ∓ 2Gm
D2 = , (B.19)
Lf− Lm

and
Lm
D3 = , (B.20)
Lf− Lm

where L f , and L m are Lame’s constants that are expressed as


Ef νf
Lf = , (B.21)
(1 + νf )(1 − 2νf )

and
Em νm
L m = , (B.22)
(1 + νm )(1 − 2νm )
In addition, the non-zero components of the Eshelby tensor (Sijkl ) are [87]
{ [ ] }
1 3(ar )2 − 1 3(ar )2
S1111 = 1 − 2νm + − 1 − 2νm + g , (B.23)
2(1 − νm ) (ar )2 − 1 (ar )2 − 1

and
[ ]
3 (ar )2 1 9/4
S2222 = + 1 − 2 νm − g, (B.24)
8(1 − νm ) (ar )2 − 1 4(1 − νm ) (ar )2 − 1

and
S3333 = S2222 , (B.25)

and
{ / [ ] }
1 (ar )2 2 3/4
S2233 = − 1 − 2νm + g , (B.26)
4(1 − νm ) (ar )2 − 1 (ar )2 − 1

and
S3322 = S2233 , (B.27)

and
[ ]
− 1 (ar )2 1 3(ar )2
S2211 = − 1 − 2 νm − g, (B.28)
2(1 − νm ) (ar )2 − 1 4(1 − νm ) (ar )2 − 1

and
S3311 = S2211 , (B.29)

and

11
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

[ ] [ ]
− 1 1 1 3/2
S1122 = 1 − 2νm + + 1 − 2νm + g, (B.30)
2(1 − νm ) (ar )2 − 1 2(1 − νm ) (ar )2 − 1

and
S1133 = S1122 , (B.31)

and
{ / [ ] }
1 (ar )2 2 3/4
S2323 = + 1 − 2νm − g , (B.32)
4(1 − νm ) (ar )2 − 1 (ar )2 − 1

and
S3232 = S2323 , (B.33)

and
{ [ ]}
1 (ar )2 + 1 g (ar )2 + 1
S1212 = 1 − 2νm − − 1 − 2νm − 3 , (B.34)
4(1 − νm ) (ar )2 − 1 2 (ar )2 − 1

and additionally
⎧ ⎫
ar ⎨ [ ]12 ⎬
g=[ 2
]3/2 ⎩ar (ar ) − 1 − acosh(ar ) ⎭ (B.35)
2
(ar ) − 1

References [15] B.P. Heller, D.E. Smith, D.A. Jack, Planar deposition flow modeling of fiber filled
composites in large area additive manufacturing, Addit. Manuf. 25 (2019)
227–238.
[1] J.R. Raney, B.G. Compton, J. Mueller, et al., Rotational 3D printing of damage-
[16] Z. Wang, D.E. Smith, Numerical analysis of screw swirling effects on fiber
tolerant composites with programmable mechanics, Proc. Natl. Acad. Sci. 115 (6)
orientation in large area additive manufacturing polymer composite deposition,
(2018) 1198–1203.
Compos. Part B: Eng. 177 (2019), 107284.
[2] B. Hausnerová, N. Honkova, A. Lengálová, T. Kitano, P. Saha, Rheology and fiber
[17] S.G. Advani, C.L. Tucker III, The use of tensors to describe and predict fiber
degradation during shear flow of carbon-fiber-reinforced polypropylenes, Polym.
orientation in short fiber composites, J. Rheol. 31 (8) (1987) 751–784.
Sci. Ser. A 48 (9) (2006) 951–960.
[18] C.L. Tucker III, Flow regimes for fiber suspensions in narrow gaps, J. Non-Newton.
[3] M. Aigner, D. Salaberger, T. Köpplmayr, B. Heise, S. Schausberger, A. Buchsbaum,
Fluid Mech. 39 (3) (1991) 239–268.
D. Stifter, J. Miethlinger, Determining the orientation, distributions, and
[19] C.E. Duty, V. Kunc, B. Compton, B. Post, D. Erdman, R. Smith, R. Lind, P. Lloyd,
concentration of glass fibers in polymer matrix using X-ray computed tomography
L. Love, Structure and mechanical behavior of Big Area Additive Manufacturing
and optical coherence tomography images, Conference on Industrial Computed
(BAAM) materials, Rapid Prototyp. J. 23 (1) (2017) 181–189.
Tomography-Non destructive Testing, 3D Materials Characterisation und
[20] D. Zhang, D. E. Smith, D. A. Jack, S. Montgomery-Smith, Numerical evaluation of
Dimensional Measurement, 2014, pp. 395–402.
single fiber motion for short-fiber-reinforced composite materials processing,
[4] S. Goris, Experimental study on fiber attrition of long glass fiber-reinforced
J. Manuf. Sci. Eng. -Trans. Asme 133 (2011) 5.
thermoplastics under controlled conditions in a couette flow, annual technical
[21] F. Folgar, C.L. Tucker III, Orientation behavior of fibers in concentrated
conference and exhibition-Society of Plastics Engineers, 2017.
suspensions, J. Reinf. Plast. Compos. 3 (2) (1984) 98–119.
[5] H. Zhuang, P. Ren, Y. Zong, G. Dai, Relationship between fiber degradation and
[22] B.E. VerWeyst, C.L. Tucker III, Fiber suspensions in complex geometries: flow/
residence time distribution in the processing of long fiber reinforced
orientation coupling, Can. J. Chem. Eng. 80 (6) (2002) 1093–1106.
thermoplastics, Express Polym. Lett. 2 (8) (2008) 560–568.
[23] J. Wang, J.F. O’Gara, C.L. Tucker III, An objective model for slow orientation
[6] R. Bailey, H. Kraft, A study of fibre attrition in the processing of long fibre
kinetics in concentrated fiber suspensions: theory and rheological evidence,
reinforced thermoplastics, Int. Polym. Process. 2 (2) (1987) 94–101.
J. Rheol. 52 (5) (2008) 1179–1200.
[7] X. Jin, J. Wang, Fiber breakage calculation for injection molded long fiber
[24] W.K. Chin, H.T. Liu, Y.D. Lee, Effects of fiber length and orientation distribution on
composites, 11th ANTEC Conference, 2011.
the elastic modulus of short fiber reinforced thermoplastics, Polym. Compos. 9 (1)
[8] J.H. Phelps, A.I.A. El-Rahman, V. Kunc, C.L. Tucker III, A model for fiber length
(1988) 27–35.
attrition in injection-molded long-fiber composites, Compos. Part A: Appl. Sci.
[25] F. Ularych, M. Sova, J. Vokrouhlecḱy, B. Turčić, Empirical relations of the
Manuf. 51 (2013) 11–21.
mechanical properties of polyamide 6 reinforced with short glass fibers, Polym.
[9] Y. Béreaux, J.-Y. Charmeau, M. Moguedet, Modelling of fibre damage in single
Compos. 14 (3) (1993) 229–237.
screw processing, Int. J. Mater. Form. 1 (1) (2008) 827–830.
[26] S.-Y. Fu, B. Lauke, Effects of fiber length and fiber orientation distributions on the
[10] T.L. Bayush, B. Thattai, S. Pillay, U. Vaidya, Processing and characterization of
tensile strength of short-fiber-reinforced polymers, Compos. Sci. Technol. 56 (10)
hemp fiber reinforced polypropylene composites, eccm15 - 15th european
(1996) 1179–1190.
conference on composite materials, Venice, Italy, 2012.
[27] I. The MathWorks, MATLAB and Statistics Toolbox Release, Natick, Massachusetts,
[11] G. Gamon, P. Evon, L. Rigal, Twin-screw extrusion impact on natural fibre
United States, 2018.
morphology and material properties in poly (lactic acid) based biocomposites, Ind.
[28] N. Ba Nghiep, S.K. Bapanapalli, J.D. Holbery, M.T. Smith, V. Kunc, B.J. Frame, J.
Crops Prod. 46 (2013) 173–185.
H. Phelps, C.L. Tucker III, Fiber length and orientation in long-fiber injection-
[12] A. Inoue, K. Morita, T. Tanaka, Y. Arao, Y. Sawada, Effect of screw design on fiber
molded thermoplastics—Part I: modeling of microstructure and elastic properties,
breakage and dispersion in injection-molded long glass-fiber-reinforced
J. Compos. Mater. 42 (10) (2008) 1003–1029.
polypropylene, J. Compos. Mater. 49 (1) (2015) 75–84.
[29] D.A. Jack, Advanced analysis of short-fiber polymer composite material behavior,
[13] B.P. Heller, D.E. Smith, D.A. Jack, Effects of extrudate swell and nozzle geometry
University of Missouri–Columbia, 2006.
on fiber orientation in Fused Filament Fabrication nozzle flow, Addit. Manuf. 12
[30] D.A. Jack, D.E. Smith, Elastic properties of short-fiber polymer composites,
(2016) 252–264.
derivation and demonstration of analytical forms for expectation and variance
[14] Z. Wang, D. Smith, Rheology effects on predicted fiber orientation and elastic
from orientation tensors, J. Compos. Mater. 42 (3) (2008) 277–308.
properties in large scale polymer composite additive manufacturing, J. Compos.
[31] G.P. Tandon, G.J. Weng, The effect of aspect ratio of inclusions on the elastic
Sci. 2 (1) (2018) 10.
properties of unidirectionally aligned composites, Polym. Compos. 5 (4) (1984)
327–333.

12
Z. Wang et al. Additive Manufacturing 43 (2021) 102006

[32] C.L. Tucker III, E. Liang, Stiffness predictions for unidirectional short-fiber [36] Z. Wang, Computational modeling of fiber reinforced composites melt flow in
composites: review and evaluation, Compos. Sci. Technol. 59 (5) (1999) 655–671. nozzle extrudatefor polymer deposition additive manufacturing. PhD Dissertation.
[33] R.S. Bay, C.L. Tucker III, Stereological measurement and error estimates for three- Baylor University, 2019.
dimensional fiber orientation, Polym. Eng. Sci. 32 (4) (1992) 240–253. [37] Y.-c Fung, A First Course in Continuum Mechanics, 1977, Prentice-Hall, Inc,
[34] Issam Doghri, Laurent Tinel, Micromechanics of inelastic composites with Englewood Cliffs, NJ, 1977, p. 351.
misaligned inclusions: numerical treatment of orientation, Comput. Methods Appl. [38] C.C. Steven, Applied Numerical Methods With Matlab: For Engineers And
Mech. Eng. 195 (13–16) (2006) 1387–1406. Scientists, Tata McGraw Hill Education Private Limited, 2007.
[35] T. Russell, A.J. David, Fiber Aspect Ratio Characterization and Stiffness Prediction [39] V. Kunc, Advances and challenges in large scale polymer additive manufacturing,
in Large-Area, Additive Manufactured, Short-Fiber Composites, SPE ACCE 2019, Proceedings of the 15th SPE Automotive Composites Conference, Novi, MI, USA,
Novi, Mi, USA, 2019. 2015, pp. 9–11.
[40] C. Zhang. Modeling of Flexible Fiber Motion and Prediction of Material Properties.
Master Thesis. Baylor University, 2019.

13

You might also like