Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/280747627

Numerical modelling for improvement of mechanical properties

Article · May 2006

CITATIONS READS

0 86

4 authors, including:

Woo-Suck Han Woo Il Lee


Mines Saint-Etienne Seoul National University
50 PUBLICATIONS 364 CITATIONS 137 PUBLICATIONS 4,139 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Woo-Suck Han on 09 March 2016.

The user has requested enhancement of the downloaded file.


NUMERICAL MODELLING OF COMPRESSION MOULDING
FOR IMPROVEMENT OF MECHANICAL PROPERTIES

Moo Sun KIMa,b, Woo Suck HANb, Alain Vautrinb and Woo Il LEEa

a
School of Mechanical and Aerospace Engineering, Seoul National University
Seoul Korea
b
Mechanical and Materials Engineering Department,
Ecole Nationale Supérieure des Mines de Saint-Etienne, Saint-Etienne, France
email: moosun@emse.fr, han@emse.fr, vautrin@emse.fr, wilee@snu.ac.kr

Abstract: The objective of this study is to develop a simulation tool of compression


moulding process to improve the mechanical performance of composite structures. Most
of the influences of processing parameters on mechanical properties come from the fibre
conditions, for example, the fibre volume fraction and fibre orientation, which can be
changed by the conditions of the process. Therefore, the accurate prediction of fibre state
and processing conditions are demanded to obtain the required mechanical properties for
the final product. In this study, we developed a processing simulation program as the
preliminary step for the prediction of mechanical properties. Copyright © 2002 IFAC

Keywords: fibres, manufacturing processes, mechanical properties, optimisation problem,


finite elements method.

1. INTRODUCTION state and the optimization of processing conditions


are demanded to obtain the requested mechanical
properties for a final structure.
Compression Moulding is the manufacturing process There are two models suggested for the flow analysis
of a product by compressing precharge in the mould. of compression moulding: the generalized Hele-
The precharge is composed by resin, reinforcing Shaw (GHS) model (Folgar et al., 1983) and the
fibres, filler and etc. It has some advantages over Barone-Caulk model (Barone et al., 1985, 1986). In
other processes. The most important of them is that the case of thin precharge, GHS model shows a good
its process time is very short to satisfy mass agreement with the real flow. The higher is the
production indispensable for automobile manu- thickness of precharge, the more accurate results
facturing. The mechanical properties of final product Barone-Caulk model shows. This is caused by the
is relatively fine and the process is simple. The slipping phenomenon between the mould wall and
process involves an interaction among a squeezing the precharge. In this research, the numerical analysis
flow, heat transfer and curing reaction. Compression of the flow of GHS model for 2-Dimension has been
moulding is composed of the four steps: the completed using the CV/FEM 5Control Volume
precharge preplacing, the compression, the curing or FEM). Due to the thin thickness, the flow in the
cooling and the releasing. The representative thermo- direction of thickness can be ignored. In the
setting material for compression moulding is SMC, CV/FEM, if a fixed mesh is used, this method can
which is in the form of thin plate. The charging save much the computation time necessary to
process, which is defined as to perform a flow of generate new meshes of flow.
material, gives the great influences to a final product. Fibre separation is caused by the fibre-matrix and
The flow of fibre-reinforced material process shows fibre-fibre interaction (Hojo et al., 1986, 1988; Yoo,
a different flow form with those of a single material 1997). Due to the fibre separation, the initial uniform
or other composite material. Fibre states, such as fibre volume fraction in a precharge shows an uneven
fibre volume fraction and fibre orientation which can distribution in the final structure. The distribution of
be changed by the conditions of the process, fibre separation is computed by using the definition
influence the mechanical properties of a final product. of the relative velocity of fibres to the matrix. And
Non-uniform distribution of fibre volume fraction, the relative velocity can be calculated by the network
due to the fibre separation or fibre orientation, gives force of fibres. Using the mass conservation law of
rise to the non-uniform mechanical properties in the fibres, fibre volume fraction can be obtained in the
product. Therefore, the accurate prediction of fibre same control volume for CV/FEM of flow model.
As well as fibre volume fraction, fibre orientation is h zdz
the main cause for uneven properties in the product. ∫ 0 η (6)
As the methods to represent the state of fibre λ=
h dz
orientation, the fibre orientation distribution function ∫
0 η
(Jackson, et al.,1986) or the fibre orientation tensors
(Advani et al., 1987, 1990) can be used. We used the u and v can be expressed as Equation (7) when the
fibre orientation tensor because of the superior no-slip condition is applied at z=0,
features to the fibre orientation distribution function
for the smaller range of computation. ∂P z ( z − λ )dz
∂x ∫0
Finally, based on the flow analysis of 2-Dimensional u=
η (7)
GHS model, the algorithms taking into account the
∂P z ( z − λ )dz
∂y ∫0
fibre separation and the fibre orientation are inserted. v=
η

2. MODELLING FOR FLOW AND FIBRE STATES Then, average velocity of u and v can be expressed
as

2.1 Generalized Hele-Shaw model S ∂P S ∂P (8)


u =− , v =−
h ∂x h ∂y
The major assumptions of GHS model are:
- the inertia of the material is negligible; where S is defined as “flow conductance” as below.
- the material is incompressible;
- the variant of shear stress in z-direction and the h ( z − λ ) 2 dz
variant of pressure are the dominant terms in the S=∫
0 η
momentum equation.

SMC is assumed isotropic and incompressible. The By substituting Equation (8) into Equation (1), the
flow in z-direction can be ignored because the continuity equation is presented as following
thickness in z-direction is much smaller than the flow
∂ ⎛ S ∂P ⎞ ∂ ⎛ S ∂P ⎞ h (9)
plane length (h≪L). The flow velocity is defined as ⎜ ⎟+ ⎜ ⎟=
∂x ⎝ h ∂x ⎠ ∂y ⎜⎝ h ∂y ⎟⎠ h
the in-plane velocity averaged through the thickness
of material. With the above assumptions, the
continuity equation is described as follow. The boundary conditions are

∂P
∂u ∂v h (1) at the mold wall : | wall = 0 (10)
+ =− ∂n
∂x ∂y h at the flow front : P | front = 0
where
1 h 1 h
h ∫0
u= udz , v = ∫ vdz (2) 2.2 Fibre separation
h 0
Motions of flow and fibres show each different
The momentum equation is defined as velocity. It can be said to be the relative velocity
between resin flow and fibre flow. Due to the relative
∂P ∂τ xz velocity of them, the initial homogeneous fibre
=
∂x ∂z (3) content (i.e. fibre volume fraction) in precharge
∂P ∂τ yz becomes nonhomogeneous and gives an effect to the
=
∂y ∂z final mechanical properties of a product. The relative
velocity is caused by the external force on fibres due
The shear stresses can be presented as to the pressure distribution dominantly. The network
force as frictional force on a fibre tow, which is
τ iz = ηγiz ( i = x, y ) (4) caused by the contacts of peripheral fibre tows, acts
as a resisting force. Then it can be assumed that the
drag force on fibre by the resin flow is balanced with
The integration of Equation (3) and Equation (4) the frictional force on a fibre tows.
becomes Figure 1 shows the definition of each velocity.
∂P
τ xz = (z − λ)
∂x (5)
uc
∂P F nw
τ yz = (z − λ) uf
∂y
us
in which λ is the z-coordinate value where shear
stress is zero. In the symmetry case, λ is equal to
Fig.1. Definition of velocity components and
h/2, and in the asymmetry case, λ can be obtained network force acts on the fibre
from the following equation The difference of velocity between fibre and matrix
owing to this phenomenon, can be determined by
using the equivalence of drag force and network analysis. In case of the random distribution of fibre,
force. Drag force is presented as the characteristic length of fibre tow is obtained from
following equation
1 (11)
FD = C D ρ m u s2 A p
2 θc
π /2 π /2 Df lf Df l f 1 + cot
2 2 2
π∫ π∫
l= l (θ )dθ = dθ = ln 2
0 0 D f sin θ + l f cos θ π D 2f + l 2f θc
C D , ρ m and A p mean the drag coefficient, the 1 − tan
2
(19)
density of matrix and the area where drag force act
respectively.
where
u s , the relative velocity of matrix and fibre, is lf (20)
θ c = tan −1
defined as Df

u s = uc − u f (12) When fibre distribution has the orientation, by using


fibre orientation distribution function ψ (θ ) , l is
where u c is the matrix velocity and u f is the fibre defined as following (Park et al., 2001)
velocity.
The equivalence of network force and drag force is π /2 π /2 Df lf
l=∫ l (θ )ψ (θ )dθ = ∫ ψ (θ )dθ
presented as −π / 2 −π / 2 D f | sin θ | +l f | cos θ |
(21)
1
C D ρ m u s2 A p = Fnw (13)
2 Network force of fibre consists of the frictional
forces due to the elastic bending and the pressure.
Drag coefficient is expressed as the equation of Generally, the frictional force due to the elastic
Reynold’s number as following bending can be ignored compared with that due to
the pressure.
C D = α Re − β (14) Frictional force of pressure Fn is defined as
l

xo + l f / 2
where α = 1.328, β = 0.5 for a flat plate, Fn = ∫ p( s ) D f ds (22)
xo − l f / 2

α = 24.0, β = 1.0 for a spherical body and


α = 6.474, β = 0.884 for a cylindrical body. Network force Fnw is the difference of frictional
In this study, it is assumed that fibre tow has the forces between upstream and downstream
shape of a cylinder.
Substituting Equation (14) into Equation (13) gives
Fnw = Fn ,up − Fn , dn (23)
1 (15)
Fnw = ρ m u s2 A p (α Re − β )
2 l Fn ,up and Fn ,dn are defined as Equations (24) and
(25)
Reynold’s number with the characteristic length of
fibre is xo
Fn ,up = κ ∫ p ( s ) D f ds (24)
xo − l f / 2

ρ mus l (16) Fn,dn = κ ∫


xo + l f / 2
(25)
Re = p( s) D f ds
l μm xo

Equation (17) is obtained by the combination of


Equations (15) and (16) where κ is a proportionality constant.

β Velocity distribution
⎛ μ ⎞ Fnw (17)
α⎜ m ⎟ = of other fibres
⎜ρ u l⎟ 1 u B , rel
⎝ m s ⎠ ρ m u s2 A p
2 u A , rel Sample fibre
us velocity
Organizing Equation (17) for u s , it is presented as o
u A , rel u B , rel
1
β 0.896
⎡2 l Fnw ⎤⎥
2− β ⎡ l Fnw ⎤⎥
0.884
Friction force to Friction force to
us = ⎢ = 0.349⎢ 0.116 0.884
⎢α ρ m1− β μ mβ (πD f l f ) ⎥ ⎢ ρ m μ m (πD f l f ) ⎥ upper stream down stream
⎣ ⎦ ⎣ ⎦
(18) Fig.2. Scheme of friction force by contacts between
fibres
Friction forces which acts on the fibre by contacts
Fibre velocity, u f is obtained from Equations (12) between fibres is shown in Figure 2.
and (18). u c is calculated from the fluid flow Finally, network force is defined as
To calculate components of aij, one must replace aijkl
= κ ⎛⎜ rup ∫ p ( x) D f ds ⎞⎟
xo + l f / 2 by a suitable closure approximation. Various closure
Fnw
xo
p ( s ) D f ds − rdn ∫ (26)
⎝ x o − l f / 2 x o ⎠ approximations in planar and three-dimensional flow
fields have been used extensively and it has been
where rup and rdn are correction parameters shown that a hybrid closure, which combines the
simple linear and quadratic forms of approximations
performs best. The present calculations use the
2.3 Fibre orientation hybrid closure given for planar orientations.
Advani and Tucker (1987) developed an efficient
a 4 = aijkl = (1 − f )aˆ ijkl + fa~ijkl (33)
method to represent fibre orientation using
orientation tensors. The technique reduced the
computational requirements. A compact and general where the linear closure is given by
description of fibre orientation state is provided by
1
the tensors defined as the following: aˆ 4 = aˆ ijkl = − (δ ij δ kl + δ ik δ jl + δ il δ jk )
35
1
aij = ∫ pi p jψ ( p )dp (27) + (aij δ kl + aik δ jl + ail δ jk + a kl δ ij + a jl δ ik + a jk δ il )
7
(34)
aijkl = ∫ pi p j p k plψ ( p)dp (28)
and the quadratic form is expressed as
where unit vector pi for two-dimensional orientation
a~4 = a~ijkl = a 2 a 2 = aij a kl (35)
is defined as p1 = cos θ p 2 = sin θ .
Note that aij is symmetric and its trace is equal to
unity. The advantage of using the tensor and the scalar measure of orientation for planar
representation is that only a few numbers are orientation field f is defined as
required to describe the orientation state at any point
in space. For planar orientations there are four f = 2aij a ji − 1 (36)
components of aij , but only two are independent.
Folgar and Tucker’s equation of single fibre motion
in a concentrated suspension can be combined with 3. NUMERICAL ANALYSIS
the equation of continuity to produce an equation of
change for the probability distribution function
and/or the orientation tensor. The result for second- 3.1 Computation of pressure and velocity
order orientation tensors is
As process material is charging the mould, the
Da ij 1 1 domain to calculate is varied. For the analysis of
= − (ω ik a kj − a ik ω kj ) + λ (γ ik a kj + a ik γ k − 2γ kl a ijkl )
Dt 2 2 fluid flow in this study, the fixed grid methods, also
+ 2C I γ (δ ij − α a ij ) known as the Eulerian methods, is applied. For
(29) definition of the domain to calculate, volume of fluid
(VOF) method, which uses the volume fraction of
where δ ij is the unit tensor and α is equal to 3 for each control volume, is used.
In actual flow, the flow front doesn’t exactly lie on
three-dimensional orientation and to 2 for planar the element borders. Therefore the flow front line
orientation. ω ij and γij are the vorticity and the rate divides the flow front element into two regions; the
regions occupied either by the process material or by
of deformation tensors, respectively. And λ is a the air. In the conventional fixed grid method, it is
parameter related to the geometry of the fibre and difficult to predict each different pressure gradient in
expressed by the following equation. a flow front element as can be seen in Figure 3. In
order to consider two different pressure gradients in
(a / b) 2 − 1 (30)
λ= one element, Kang et al. (1997) proposed the method
( a / b) 2 + 1 of floating imaginary nodes and elements (FINE)
which adds and uses the additional nodes in a flow
where a is the length of the symmetric axis and b the front element to calculate more exact pressure
length of the other axis of the fibre tow. ω ij and γij distribution. The FINE method is used in this study.
are defined in terms of velocity gradients:
P1 P1
Pamb Pamb
∂v j ∂vi (31)
ω ij = − 3
1 3 1
∂xi ∂x j Pamb Pamb
2 2
Flow front Flow front
∂v j
∂v (32)
γij = + i
∂xi ∂x j (a) Flow front in fixed grid method (b) Actual flow front
Fig.3. Pressure distribution in fixed grid method and
actual flow front
From the pressure information of each nodes, Figure 4 represents the transformed elements into 2-
velocity can be computed at the centroid of an dimension and the original arbitrary oriented element
element. Then the velocity of one node is evaluated z
by the average of those values computed at its
a’
c’
surrounding elements. G
b’ n
a c
3.2 Computation of fibre separation
β b
Mass conservation of fibre can be expressed as the y
θ
equation of fibre content x

∂φ G h Fig.4. Original arbitrary element and transformed


+ ∇ ⋅ (φv ) = − φ (37)
∂t h Element

The right term means mass generation due to


compression by moulds. Integration of Equation (37) 4. FLOW SIMULATION RESULTS
for control volume and divergence theorem under the
assumption of the constant fibre content and rate of
fibre content gives To verify the flow simulation program, numerical
analysis results are compared with the analytical
∂φ G h (38) solutions. For example, the mould size is 0.5 m × 1 m
Vc.v + ∫ φv ⋅ nˆ dΓ = − φVc.v
∂t c. s h and precharge size is 0.5 m × 0.3 m as Figure 5. As
processing conditions, the compression speed is
The distribution of fibre volume fraction is computed 0.001 m/sec and the precharge thickness 0.01 m and
on the same control volume used for flow the precharge has 0.6 as initial fibre volume fraction.
computation by applying Equation (38). Time- At this time, its viscosity is the constant value, 500
difference equation of Equation (38) can be given as Pa·sec.
follow.

⎛ φ old v f ⋅ nˆdΓ ⎞

φ new = φ old − ⎜⎜ Γ h
+ φ old ⎟⎟Δt
(39)
mold

⎜ Vc.v h ⎟ 1.0m
⎝ ⎠

precharge 0.3m
where v f is the fibre velocity.
0.5m
3.3 Computation of fibre orientation
Fig.5. A shape of mould and precharge for
Using Equation (29), fibre orientation tensors at each verification of numerical method
nodes can be computed. For closure approximation
of 4th order tensors to 2nd order tensors, hybrid The exact analytical filling time is 7 sec for this case.
closure approximation is used. When 3-dimensional Our simulation programme obtained 6.88 sec.
geometry is considered, 2-dimensional planar fibre When the mould filling is completed, the distribution
orientation tensors are extended to 3-dimension. And of fibre volume fraction is shown in Figure 6. The
3-dimensional velocity vectors are to be transformed fibre volume fraction on the initial precharged
to an equivalent 2-dimensional space (Reifschneider location has a similar or greater value than the initial
et al., 1994). value, and the late charged area shows relatively
Transformation matrix for velocity vectors is defined lower values due to the compensation of higher
as the following. values occurred on the initial precharged location.

{v'} = {Rθz }{R−yβ }{R−zθ }{v}


1
where Fiber fraction
0.616602
0.9 0.597487
{v'} : transforme d vector 0.8
0.578373
0.559258
0.540144

{v} : original vector 0.7


0.521029
0.501915
0.482801
0.463686
{R} : rotation matrix 0.6 0.444572
0.425457
0.5 0.406343
y

0.387228
0.368114
0.4
Tensors can be transformed in a similar way. 0.348999

0.3

Tij = α ik 'α jm'Tk 'm'


0.2

0.1

where 0
0 0.5 1
x

α : element of transforma tion matrix {Rθz }{R−yβ }{R−zθ }


Fig.6. Fibre fraction distribution after completion of
filling
Concerning the distribution of fibre volume fraction Barone, M.R. and D.A.Caulk (1986). A model for the
due to the fibre separation, it shows a half decay of flow of a chopped fiber reinforced polymer
fibre fraction in the ending filling part of geometry. compound in compression molding. Journal of
Figure 7 shows the fibre orientation tensor (a11) Applied Mechanics, Vol. 53, pp.361-371.
distribution.
Folgar, F. and C.L. Tucker III (1983). A Model of
Compression Mold Filling. Polymer Engineering and
1
got11 Science, Vol.23. No.2, pp.69-73.
0.787954
0.9 0.739196
0.690438
0.641681
0.8
0.592923
0.544165
Hojo, H., C. Nagatsuka, E.G. Kim and K. Tamakawa
0.7 0.495407
0.446649 (1986). Separation of matrix and fibers during
0.397891
0.6 0.349133
0.300375
molding of long fiber-reinforced thermoplastics.
0.5 0.251618
y

0.20286
0.154102
Proceedings of the Third Japan-U.S Conference on
0.4 0.105344
Composite Materials, pp.605-612.
0.3

0.2
Hojo, H., E.G. Kim, H. Yaguchi and T.Onodera
0.1
(1988). Simulation of compression molding with
0
0 0.5
x
1
matrix-fiber separation and fiber orientation for long
fiber-reinforced thermoplastics, Intern.Polymer
Fig.7. Fibre orientation tensor distribution (a11)
Processing III, pp.54-61.
It is shown that the fibres tend to align with the flow
Jackson, W.C., C.L. Tucker and S.G. Advani (1986).
direction and the fibres in flow front region have the
Predicting the orientation of short fibers in thin
perpendicular direction to the flow.
compression moldings. Journal of Composite
Materials, Vol. 20, pp.539-557.
5. CONCLUSION
Kang, M.K. (1997). A numerical and Experimental
Study on Mold Filling and Void Formation During
We developed a simulation code for the flow Resin Transfer Molding. Ph.D. Thesis, Seoul
analysis and fibre states such as fibre volume fraction National University, Seoul.
and fibre orientation. Prediction of fibre states
permits us to analyse mechanical properties of a Park, C.H., E.G. Kim, W.I. Lee and Y.E. Yoo (2001).
structure manufactured by compression moulding. It A study on fiber orientation in the compression
can be said that the mechanical properties deduced molding of fiber reinforced polymer composite
from the fibre states depend on the processing material, Journal of Materials Processing
condition, such as the precharge location, Technology, Vol.111,pp.233-239.
compression speed, and viscosity etc, which cause
non-homogeneous fibre contents and anisotropic Reifschneider, L.G. and H.U. Akay (1994).
fibre orientation. For the future work, the Application of a fiber orientation prediction
optimisation process will be added to take into algorithm for compression molded parts with
account simultaneously the required mechanical multiple charges, Polymer Composites, Vol.15,
properties and the change of mechanical properties pp.261-269.
during the manufacturing process so as to satisfy the
requirements on the final product. Yoo, Y.E. (1997). A Study on the analysis of
compression molding process of composite material
structures. Ph.D.Thesis, Seoul National University,
REFERENCES Seoul.

Advani, S.G. and C.L. Tucker III (1987). The use of


tensors to describe and predict fiber orientation in
short fiber orientation in short fiber composite.
Journal of Rheology, Vol. 31, No. 8, pp.751-784.

Advani, S.G. and C.L. Tucker III (1990). A


Numerical Simulation of Short Fiber Orientation in
Compression Molding. Polymer Composites, Vol.11,
No.3, pp.164-173.

Barone, M.R. and D.A.Caulk (1985). Kinematics of


flow in sheet molding compounds. Polymer
Composites, Vol.6, No.2, pp.105-109.

View publication stats

You might also like