MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Warpage Prediction of Optical Media

1
Bingfeng Fan
Department of Mechanical & Industrial Engineering
University of Massachusetts Amherst
Amherst, MA 01003

the relatively cold mold cavity under high pressure, a thin layer of
ABSTRACT
the melt solidifies. As the part experiences further cooling, the
solidification develops inward. Along the thickness direction of
A numerical simulation of the warpage of injection-
the part, the temperature distribution is highly non-uniform.
compression molded optical media such as CD’s and DVD’s due
Consequently, each material point through the thickness solidifies
to asymmetric cooling is developed. A thermal viscoelastic
at a different time, under different pressure, and experiences
constitutive model is employed to calculate the thermally induced
different shrinkage. The temperature and pressure history, coupled
stress. The stress is not symmetric with respect to the mid-plane
with the viscoelastic properties of the polymer during the rapid
of the disk because of the differential mold temperatures
cooling process results in the development of the thermally
employed. Before demolding, the out-of-plane deformation is
induced residual stresses. While the flow induced stresses are the
constrained by the mold walls. After ejection, the disk is
main sources of molecular orientations and hence the sources of
quenched in the air, and is free to deform, causing warpage to
anisotropic mechanical and optical properties between the flow
develop in the disk. A finite element analysis code is developed
and cross-flow directions, they are generally lower than the
using axisymmetric plate elements to simulate the warpage of the
thermally induced stresses by one to two orders of magnitude [3].
disk after demolding. The complete finite element formulation is
As such, the warpage of the disk is mainly caused by the
described. Simulation results of warpage for CD-R moldings are
thermally induced stresses.
compared with experimental observations under different
processing conditions such as melt temperature, mold Frozen Layer Melt Front
temperature, and packing pressure using an optical grade of
polycarbonate. Good agreement of the simulation results and
experimental observations is obtained. The comparison of the
simulation and experiment reveals that the processing conditions Thickness

have complicated effects on the warpage, and mold temperature

has the greatest effect on the warpage in the current study.

INTRODUCTION Velocity Profile Fountain Flow

Optical media such as CD’s and DVD’s are typically Fig. 1. Schematic of polymer flow during mold filling
injection-compression molded using optical grade of Ideally, if the temperatures of the mold walls of the stationary
polycarbonate (PC). One critical quality of the mass produced and moving mold halves are exactly the same, the stress
optical media is warpage caused by internal residual stresses, distribution across the thickness direction will be symmetric over
which reduces the dimensional stability of the disks, and is the mid-plane of the disk, and there will be no warpage in the
particularly detrimental for DVD’s because it hinders the bonding molded disk. However, it is common to find a temperature
of multiple substrates. There are many causes of the warpage of difference between the two mold halves in practice due to the
injection/injection-compression molded parts. Paro, et al. offers location of the gates, the difference of the coolant temperatures of
various reasons for warpage such as uneven cooling, flow both sides, or the stochastic deviation of the actual temperature
orientation effect, uneven packing, and gate effect, with schematic from the control parameters set on the machine. As a result, the
illustrations [1]. In the current study, the warpage induced by gapwise distribution of the residual stress will no longer be
asymmetric thermal stress due to differential mold temperatures is symmetric. Before ejection of the disk, the out-of-plane
studied, and simulation results of CD-R moldings are compared deformation is constrained by the mold walls. After ejection, the
with experimental observations under different processing disk experience free quench in the air, and is free to deform. Due
conditions. to the asymmetry of the internal stress, the disk will tend to warp
There are two types of residual stresses [2]: flow induced in order to achieve a lower energy state.
stress and thermally induced stress. The first is due to the shear
flow during the filling and packing stages, and the shear and PROCESS SIMULATION
normal stresses developed are partly frozen-in due to the rapid The simulation of the injection-compression molding process
cooling of the melt. The second type of residual stresses arises is previously described [4]. The energy equation for the process
during the cooling process of the injection-compression molding. before ejection is:
As illustrated in Fig. 1, when the hot polymer melt is injected into

1
© Copyright by Bingfeng Fan 2002. All rights reserved.
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

 ∂T ∂T  ∂ ∂T FINITE ELEMENT FORMULATION

ρC p  +u  = (k ) + ηγs 2 (1)
 ∂t ∂x  ∂z ∂z
After demolding, the warpage of the disk is modeled by a
where η is the viscosity, P is the pressure, T is the temperature, finite element analysis (FEA). The elements chosen are 2-node 1-
D axisymmetric thin plate elements discretized along the radial
γ is the shear rate, Cp and k are the specific heat and thermal direction as shown in Fig. 2. A cylindrical coordinate system
conductivity of the polymer respectively. After ejection, the ( r , θ , z ) is employed. There are 3 degrees of freedom for each
energy equation simplifies to:
node:
∂T ∂ ∂T
ρC p = (k (2) D = [u w ϕ ]
T
)
∂t ∂z ∂z
The energy equation is solved by a finite difference method to corresponding to radial and vertical displacement as well as
obtain the transient temperature along both the radial direction rotation.
and the thickness direction throughout the whole process.

CONSTITUTIVE MODELS
Assuming that time-temperature superposition applies, a
linear thermo-viscoelastic constitutive model [5] is given by:
σ = − ph I + σ d (1)

1 t α 1 
p h = − tr (σ ) = ∫  TD − tr (εD ) dτ (2)
3 0
κ κ 
m
σ d = ∑ 2 ∫ g i e −(ξ ( t ) −ξ (τ ) ) /θi εD d dτ
t
(3)
0
i =1

where σd and εd are the deviatoric stress and strain tensors; g i and
θi are the shear moduli and the corresponding relaxation times, Fig. 2. Schematic of disk discretization. From [7]
h Each element along the radial direction of the disk is divided
and m is the number of modes; T is the temperature; p is the
hydrostatic pressure and I is the unit tensor. The temperature into a number of layers along the thickness direction to calculate
dependent thermal volume expansion coefficient, α , and the the displacements and thermal stresses as a function of z. The
isothermal compressibility coefficient, κ , in the expression of
thickness discretization is the same as in the calculation of the
gapwise temperature distribution by the finite difference method.
hydrostatic pressure are defined as:
Due to the mold filling process, the temperature distribution of the
1  ∂v  1  ∂v  (4) disk is not homogeneous along the radial direction. Therefore,
α=   , κ =−  
v  ∂T  p v  ∂P  T temperature and thermal stress are both functions of r and z.

which are obtained from the p-v-T relationship. The pseudo-time MAIN ASSUMPTIONS OF FEA
ξ is defined as:
1. There is no transverse shear deformation. Therefore, normals to
τ 1
ξ (τ ) = ∫
the mid-plane before deformation remain straight and normal to ds (5)
0 aT (T )
the mid-plane after deformation. This assumption is from the
classical Kirchhoff thin plate theory. Since the radius to
where aT is the shift factor of the time-temperature superposition
thickness ratio of CD’s and DVD’s is generally greater than 50,
principle. the thin plate assumption is justifiable.
The shear modulus of the polymer is modeled by a set of dw ( r )
2. u = u − zϕ , where ϕ = is the rotation, u and w
discrete moduli g i and relaxation times θ i , determined from dr
dynamic mechanical testing. The material behavior and model are the displacements of the mid-plane of the disk.
parameters were characterized and found consistent, as previously 3. v = 0 , which is the axisymmetric condition.
described [6]. 4. w = w (r ) .

2
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

STRAIN-DISPLACEMENT RELATIONSHIP where σ is the stress tensor, ∆ε is the strain increment tensor, H
is the rigidity matrix, and h is the history vector. Detailed
Since there are six degrees of freedom in each element, six
derivations of H and h can be found in [9].
coefficients have to be taken in the displacement interpolation
polynomials. Assuming radial displacement varies linearly in x,
and vertical displacement varies as a cubic in x, where x is the STIFFNESS MATRICES AND LOAD VECTORS
local coordinate of the radial direction, the displacement within an Derived from the principle of virtual work, the set of linear
element is described as [8]: equations to be solved can be written as:
 u1 
 w  (6)
K∆D = R (11)
 1

 u  1 − ξ
6 z (ξ − ξ ) 2
z ( −1 + 4ξ − 3ξ 2 ) ξ
6 z ( −ξ + ξ ) 2

z ( 2ξ − 3ξ 2 )  ϕ 1  where K is the global stiffness matrix, ∆D is the displacement
 = s s   increment vector to be solved, and R is the right-hand side vector.
w  0 1 − 3ξ 2 + 2ξ 3 s(ξ − 2ξ + ξ ) 2 3
3ξ 2 − 2ξ 3
 u
s(ξ 3 − ξ 2 )   2 
 0
 w2 
  The element stiffness matrices can be written as:
ϕ 2 
d /2
where ξ = x/s, s is the length of the element, and k e = ∫ B T HBdV = ∫ ∫ B T HBdzdA
−d / 2 (12)
[u1 w1 ϕ1 u2 w2 ϕ 2 ] is the nodal displacement vector.
T
d /2
V
r2 2π
A
d /2 r2
= ∫ ∫ ∫ B T HBdθrdrdz = 2π ∫ ∫ B T HBrdrdz
− d / 2 r1 0 − d / 2 r1
The strain-displacement relationship in cylindrical coordinate
system is derived to be: where B is the strain-displacement vector defined in Eq. 8, H is
 1 6z z 1 6z z   u1  the viscoelastic rigidity matrix defined in Eq. 10, d is the
−s (1 − 2ξ ) (4 − 6ξ ) ( −1 + 2ξ ) ( 2 − 6ξ )  (7)
s2 s s s2 s  
 ε rr  1 − ξ 6 z z ξ 6z z   w1  thickness of the disk, and r1 and r2 are the radii of the two nodes
ε   (ξ − ξ )
2
( −1 + 4ξ − 3ξ )
2
( −ξ + ξ )
2
( 2ξ − 3ξ ) ϕ 
2
 θθ   r rs r r rs r  respectively.
  =  
1

κ rr 
6  0 2 6 2
(1 − 2ξ ) (2 − 3ξ ) 0 (2ξ − 1) (1 − 3ξ )   u2 
κ θθ   s2 s s2 s  w 
  2 The element right-hand side vector can be written as:
( 2ξ − 3ξ 2 )  ϕ 2 
6 1 6 1
 0 (ξ − ξ 2 ) ( −1 + 4ξ − 3ξ 2 ) 0 ( −ξ + ξ 2 )
 sr r sr r  d /2
R e = ∫ ( N T f − B T h)dV = ∫ ∫ ( N T f − B T h)dzdA (13)
−d / 2
where ε rr and ε θθ are the radial and hoop strains respectively, κ rr V A
d /2 r1 2π d /2 r2
and κ θθ are the radial and hoop curvatures respectively, and r is =∫ ∫∫ ( N T f − B T h)dθrdrdz = 2π ∫ ∫ ( N T f − B T h)rdrdz
− d / 2 r2 0 − d / 2 r1

the radius. Eq. 7 can be written in matrix form:

where N is the shape function matrix, f is the body force vector,
and h is the history vector defined in Eq. 10.
n
ε = ∑ B i ui (8)
i =1 The numerical integration over the thickness is implemented
where B i is the strain-displacement matrix. by the trapezoidal rule, and the numerical integration over the
radial direction is implemented by the Gauss quadrature for both
The incremental strain-displacement relationship can be derived element stiffness matrices and element right-hand side vectors.
in a similar manner as:
BOUNDARY CONDITIONS
n
∆ε = ∑ B i ∆ui (9)
In-mold:
i =1
1. σ zz = − p , where p is the pressure in the mold cavity.
The detailed derivation of the strain-displacement relationship is
listed in the Appendix. 2. ∆ε rr = ∆ε θθ = 0 , ∆ε zz ≠ 0 . There only non-zero strain
increment is the one in the thickness direction.
STRESS-STRAIN RELATIONSHIP Condition 2 is equivalent to a no slip boundary condition on the
mold walls. The radial and hoop strains are zero, only the
The viscoelastic constitutive equation is solved by a finite
displacement in the thickness direction is allowed. This condition
difference method in the time domain consistent with [5, 9]. In
is especially true for injection-compression molding, where the
this method, at time t n +1 , the stresses are evaluated based on the compression is in the z direction. In fact, any radial displacement
stress state at time t n , plus the temperature and pressure history would harm the micro-replication of the features on the disks.
Since ∆ε zz can be obtained directly, there is no need to calculate
up to t n +1 . The stress in matrix form is found to be: the displacements by the finite element analysis. The stress
components are directly calculated by σ = h .
σ = H∆ε + h (10)
Free quench:
1. σ zz = 0 , which is the plane-strain condition.

3
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

2. ∆ε rr ≠ 0 , ∆ε θθ ≠ 0 , ∆ε zz ≠ 0 . All strains are non-zero, and the the finite element analysis at the circular stack bump of the disk (r
disk is free to deform. = 18 mm), where the disks are in contact when stacked in a
spindle. The warpage of the molded disks is measured at every
The displacement field is solved by the finite element method. No millimeter of the radius from r = 27 mm to r = 57 mm. Five disks
boundary conditions should be applied during free quenching are measured for each DOE run, and the experimental data shown
except the displacement constraints to prevent rigid body in the following results are the average of the five disks.
movement.
The gapwise distributions of the thermal stress at a radial
RESULTS AND DISCUSSION location of r = 35 mm just before demolding and after 20 seconds
of quench in the air at room temperature when the temperature of
The simulation of injection-compression molding process, as the whole disk has dropped to the room temperature are shown in
well as related p-v-T, rheological, thermal, and viscoelastic Fig. 3. Before demolding, the stress is dominantly compressive
models have been previously described [4, 6]. Briefly, because of the high cavity pressure during the filling and packing
stages. The gapwise distribution of the thermal stress is not
• The flow rate is controlled by the screw speed, which is profiled
symmetric with respect to the mid-plane because of the
to be slow at the start of injection and increase with the disk
differential mold wall temperatures. The stress in the left hand
radius so as to best maintain constant melt front velocity during
side of the thickness is more compressive than that in the right
the filling stage. A 4-stage screw speed profile is listed in Table
hand side since the pressure decay of the left hand side is slower
1.
than the right hand side due to the higher temperature of the left
hand side. After demolding, the disk is free to deform, and the
Table 1. Screw speed as a function of position
asymmetric internal stress causes the warpage of the disk to
Position (mm) 31.50 29.00 25.00 6.00 3.50 develop. The stress distribution is less asymmetric after 20
seconds of free quench compared with just before demolding, and
Speed (mm/s) 80.0 100.0 120.0 80.0 more tensile stress develops in the core region of the disk during
• A 3-step packing pressure profile is used, which gradually the free quench in the air. The final disk has tensile stress at the
decreases the packing pressure in the cavity as the melt surfaces and in the core, and compressive stress near the surfaces.
solidifies. The profile is listed in Table 2. Note that the value of The integration of the stress over the thickness is zero, indicating
the packing pressure is the hydraulic pressure. The melt that the stress is in a self-equilibrium state.
pressure at the nozzle is calculated by the hydraulic pressure
multiplied by an intensity factor, which is 12.36 for the machine 8
used in this study.
6
Just Before Demolding
4 After 20 Seconds of Free Quench
Table 2. Packing pressures as a function of time
2

Time (s) 0.05 0.10 0.25

Stress (MPa)

2 -2
Pressure (kgf/cm ) 40.0 20.0 10.0
• A 3-step clamp force profile is used to allow the mold to open
-4

-6
(breathe) during the filling to promote flow then close during
-8
the packing to promote pressure uniformity and feature
replication. The profile is listed in Table 3. -10

-12
-1 -0.5 0 0.5 1
Table 3. Clamp force as a function of time z/d

Clamp Force (ton) 30 22.0 15.0

Fig. 3. Thermal stress gapwise distribution under
Clamp Time (s) 0.8 1.50 1.00 median processing conditions
Change Time (s) 0.2 0.10 0.50 The simulated warpage and the experimental data under all
The experiments also investigated the effects of melt temperature median processing conditions are shown in Fig. 4 as a benchmark
(between 300oC and 320oC) and mold temperature (between in the study of the effects of the processing conditions on the
100oC and 114oC) with a 3 level design of experiments. warpage. The magnitude of the simulation compares well with the
experimental data, and better agreement between simulation and
In order to study the warpage of injection-compression experiment occurs in smaller radii. Near the rim of the disk, the
molded disk due to asymmetric cooling, the temperature of the simulation slightly underpredicts the warpage.
stationary side of the mold in the simulation is set to be 5oC lower
than that of the moving side, which is the nominal mold coolant
temperature. Zero vertical displacement constraint is applied for

4
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

0 0
-10

-20 Exp. -20

Vertical Deviation (micron)

Exp.

Vertical Deviation (micron)

Sim.
-30 Sim.
-40 -40

-50
-60
-60

-70
-80
-80

-90
-100
-100
23 28 33 38 43 48 53 58
-120
Radius (mm)
23 28 33 38 43 48 53 58
Radius (mm)

Fig. 4. Warpage of a disk under median processing

(b)
conditions
Fig. 5. Warpage of sisks under different melt
EFFECT OF MELT TEMPERATURE o o
temperatures (a) Tmelt=300 C (b) Tmelt=320 C
In order to study the effects of melt temperature on the
warpage of the molded disks, low and high melt temperatures are
used with other processing conditions the same as the median 120
values. The experimental and simulated vertical displacements are
shown in Fig. 5 (a) and (b). As is observed from the figure, the Exp.
100
simulation results compare very well with the experimental data. Sim.

The main effect of melt temperature on warpage is shown in Fig.

Vertical Displacement (micron)

6. In the simulation, the trend is that the higher the melt 80

temperature, the higher the value of the warpage, but in the

experiment, the median melt temperature has the lowest warpage. 60

0
40

-20 Exp.
Vertical Deviation (micron)

20
Sim.
-40
0
295 300 305 310 315 320 325
-60 o
Melt Temperature ( C)

-80
Fig. 6. Main effect of melt temperature on warpage
-100
EFFECT OF MOLD TEMPERATURE
-120 The experimental and simulated vertical displacements with
23 28 33 38 43 48 53 58 low and high values of the mold temperature are shown in Fig. 7
Radius (mm) (a) and (b). Other conditions are the median values. The main
effect of mold temperature on the warpage is shown in Fig. 8.
(a) Both simulation and experiment show that higher mold
temperature leads to higher value of warpage.

5
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

EFFECT OF PACKING PRESSURE

0
The experimental and simulated vertical displacements with
-20 low and high packing pressures and median values for other
Vertical Deviation (micron)

conditions are shown in Fig. 9 (a) and (b). The main effect of
-40 packing pressure on the warpage is shown in Fig. 10. In the
Exp. experiment, the warpage at low and median packing pressures has
-60 Sim. almost the same values, and as the packing pressure increases to
the high value, the warpage increases significantly. While in the
-80
simulation, the median pressure has the highest value of warpage.
Higher and lower packing pressure both result in decreased
-100
warpage compared with median value. This phenomenon is
-120
different from the observations in [10] for injection molding,
23 28 33 38 43 48 53 58 where it is observed that the higher the packing pressure is, the
Radius (mm) lower the warpage will be.

(a) 0

0
Exp.
-20
Sim.
-20 Exp.

Vertical Deviation (micron)

Sim.
-40
Vertical Deviation (micron)

-40

-60
-60

-80
-80

-100
-100

-120
-120 23 28 33 38 43 48 53 58
23 28 33 38 43 48 53 58
Radiaus (mm)
Radius (mm)

(a)
(b)
0
Fig. 7. Warpage of disks under different mold
o o
temperatures (a) Tmold=100 C (b) Tmold=114 C -20 Exp.
Sim.
Vertical Deviation (micron)

120
-40

100
Exp. -60
Sim.
Vertical Displacement (micron)

80
-80

60
-100

40 -120
23 28 33 38 43 48 53 58

20 Radius (mm)

(b)
0
95 100 105 110 115 Fig. 9. Warpage of disks under different packing pressures
2 2
MoldTemperature ( C)
o
(a) 12kgf/cm maximum (b) 40kgf/cm maximum

Fig. 8. Main effect of mold temperature on warpage

6
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

120
as a cubic in x, where x is the local coordinate of the radial
dw ( x )
direction, and taking advantage of the relationship ϕ = :
100
Exp. dx
Sim.
Vertical Displacement (micron)


 u = α1 + α 2 x
80

 (14)
 w = α 3 + α 4 x + α5 x + α 6 x
2 3
60
 dw
 dx = α 4 + 2α 5 x + 3α6 x
2

40

The last equation in Eq. 14 comes directly as the derivative of the

20 second equation. Writing Eq. 14 in matrix form:
α1 
0
α 
0 10 20 30 40 50    2
 u  1 x 0 0 0 0  (15)
α 
2
Packing Pressure (kgf/cm )
  
 w  = 0 0 1 x x
2
x 3  3 

Fig. 10. Main effect of packing pressure on warpage  dw  0 0 0 1 2 x 3x 2  α 4 
    α 
 dx  5
 
CONCLUSIONS α
 6
A numerical simulation of thermally induced stress warpage By substituting the values of u , w , and ϕ at node 1 and node
of injection-compression molded disks using a thermal
2 of an element, the coefficients are solved to be:
viscoelastic material model is developed. Finite element approach
is employed using axisymmetric plate elements to simulate the
warpage of the injection-compression molded disks after
demolding. Simulation results of CD-R moldings under different  α1 = u1
 u2 − u1
processing conditions such as melt temperature, mold α2 =
temperature, and packing pressure are compared with  s
 α = w (16)
experimental observations. Good agreement of simulation results  3 1

and experimental data is observed. The main effect analysis α 4 = ϕ1

shows that processing conditions can have complicated effects on  ϕ2 ϕ w w
α 5 = − − 2 1 + 3 22 − 3 21
the warpage, and usually no monotonic relationships between  s s s s
warpage and processing conditions are observed.  α 6 = ϕ 2 + ϕ 1 − 2 w2 + 2 w1

 s2 s2 s3 s3
ACKNOWLEDGEMENT
where s is the length of the element.
The author would like to acknowledge Dr. David O. Kazmer,
Substituting the coefficients α1 ~ α6 into Eq. 14, one obtains:
Department of Mechanical and Industrial Engineering, University
of Massachusetts Amherst, Dr. Wit C. Bushko, GE Corporate  u1 
 w  (17)
Research and Development, and Dr. Richard P. Theriault and Dr.  1
Andrew J. Poslinski, GE Plastics for contributing to this article.  u  1 − ξ 0 0 ξ 0 0  ϕ 1 
 =  
The author is grateful to Prof. Ian Grosse for his guidance of this w   0 1 − 3ξ 2 + 2ξ 3 s(ξ − 2ξ 2 + ξ 3 ) 0 3ξ 2 − 2ξ 3 s(ξ 3 − ξ 2 )   u2 
 w2 
project. Thanks also extend to Dr. Sam Miller, GE Plastics, and  
Dr. Irene Dris, GE Corporate Research and Development for ϕ 2 
providing the material data, and colleagues in the Optical Media
u 
Development Center, GE Plastics, for enabling validation of the or  u  = n  N u (ξ ) 0 0  i  (18)
warpage simulation.   ∑  wi 
  i =1 
w 0 N w (ξ ) N ϕ (ξ )  ϕ 
 i
APPENDIX: STRESS-STRAIN RELATIONSHIP IN
AXISYMMETRIC PLATE BENDING FEA where ξ = x/s, N u (ξ ) , N w (ξ ) , and Nϕ (ξ ) are the shape functions
Using 2-node plate elements, there are 6 degrees of freedom for u , w , and ϕ respectively, and [u1 w1 ϕ1 u2 w2 ϕ 2 ] is
T

in each element. Therefore, 6 unknowns are taken in the nodal displacement vector.
interpolation polynomials for the displacements. Assuming radial
By taking the relationship:
displacement varies linearly in x, and vertical displacement varies

7
MIE 605 Finite Element Analysis Project Report, Spring 2002 Bingfeng Fan
_______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

 1   ∆u  (24)
dw   z  dw   6z z 1 6z z
u   −s (1 − 2ξ ) (4 − 6ξ ) ( −1 + 2ξ ) ( 2 − 6ξ )
u − z  u −   (19) s2 s2  1 
s  dξ  
∆ ε s s s
=   ∆w1 
dr  = 
 rr   1 − ξ
    ∆ε  
6z
(ξ − ξ 2 )
z
( −1 + 4ξ − 3ξ 2 )
ξ 6z
( −ξ + ξ 2 )
z
(2ξ − 3ξ 2 )  ∆ϕ 
  
w
 w      θθ   r
 =
rs r r rs r  1 
 w   ∆κ rr   0
6
(1 − 2ξ )
2
( 2 − 3ξ ) 0
6
(2ξ − 1)
2
(1 − 3ξ )   ∆u 2 
 2
s2   ∆w 
∆κ θθ   s s s
 2
the displacement anywhere in the thickness direction within an (2ξ − 3ξ 2 )  ∆ϕ 2 
6 1 6 1
 0 (ξ − ξ 2 ) ( −1 + 4ξ − 3ξ 2 ) 0 ( −ξ + ξ 2 )
 sr r sr r 
element is found to be:
 u1  In matrix form:
 w  (20)
 1 n

 u  1 − ξ
6 z(ξ − ξ 2 )
z( −1 + 4ξ − 3ξ 2 ) ξ
6 z( −ξ + ξ 2 ) 
z( 2ξ − 3ξ 2 ) ϕ 1  ∆ε = ∑ B i ∆ u i (25)
  = s s   i =1
w   0 1 − 3ξ 2 + 2ξ 3 s(ξ − 2ξ 2 + ξ 3 ) 3ξ 2 − 2ξ 3
 u
s(ξ 3 − ξ 2 )   2 
 0
 w2 
 
ϕ 2  REFERENCES
The strains are derived from displacements by employing the [1] Paro, H.T., C.P. Bosnyak, and K. Sehanobish, 1986,
following relationship in the cylindrical coordinate system: "Eliminating Warp in Molded Plastics". Machine Design.
58(19): pp. 69-73.
 du   1 du  [2] Isayav, A.I. and D.L. Crouthamel, 1984, "Residual Stress
 dr   s dξ   N u ,r − zN w ,r − zNϕ ,r  (21)
 ε rr   u   u  1 −z −z  Development in the Injection Molding of Polymers". Polymer-
ε      n  N Nw N ϕ   ui 
 θθ   r    u   Plastics Techynology and Engineering. 22(2): pp. 177-232.
  =  d 2w  =  1 d 2w  = ∑  0
r r r r  w
− N w ,rr − N ϕ ,rr   i 
κ rr   − 2  − 2  i =1  ϕ  [3] Hieber, C.A., Chapter 1. Melt Viscosity Characterization and
κ θθ   dr   s dξ 
2
 1 1  i 
  0 − N w , r − N ϕ ,r  Its Application to Injection Molding, in Injection and
 1 dw   1 dw  
− −
 r dr   sr dξ 
r r
  Compression Molding Fundamentals, A.I. Isayev, Editor.
1987, Marcel Dekker, Inc.: New York.
[4] Fan, B., et al., "Simulation of Injection-Compression Molding
where ε rr and ε θθ are the radial and hoop strains respectively, κ rr for Optical Media". Submitted to Polymer Engineering and
and κ θθ are the radial and hoop curvatures respectively, and r is Science.
[5] Baaijens, F.P.T., 1991, "Calculation of Residual Stresses in
the radius. The strain ε zz is to be determined by σ zz assuming
Injection Molded Products". Rheologica Acta. 30: pp. 284-
plane stress condition. 299.
Substitute Eq. 20 into Eq. 21, the strain-displacement [6] Fan, B., et al., "Birefringence Prediction of Optical Media".
relationship in cylindrical coordinate system is derived to be: Submitted to Journal of Polymer Science: Part B: Polymer
Physics.
 1 6z z 1 6z z   u1 
 − 2
( 1 − 2ξ ) ( 4 − 6ξ ) ( − 1 + 2ξ ) ( 2 − 6ξ )   (22) [7] Melerski, E.S., 1994, "Hybrid Finite Element and Modified
 ε rr  1 −sξ 6sz s s s2 s
z ξ 6z z   w1  Mixed Finite Element Solutions of Axisymmetric Circular
ε   (ξ − ξ )2
( −1 + 4ξ − 3ξ ) 2
( −ξ + ξ )2
( 2ξ − 3ξ ) ϕ 
2
 θθ   r rs r r rs r  1  Plates". Computers & Structures. 52(3): pp. 405-417.
 =
κ rr   0
6 2 6 2
(1 − 2ξ ) (2 − 3ξ ) 0 (2ξ − 1) (1 − 3ξ )   u2  [8] Rockey, K.C., H.R. Evans, and D.W. Griffiths, The Finite
κ θθ   s2 s s2 s  w 
 6 1 6 1 2 
2
 Element Method. 2nd ed. 1983, New York: John Wiley &
 0 (ξ − ξ 2
) ( −1 + 4 ξ − 3ξ 2
) 0 ( −ξ + ξ 2
) ( 2 ξ − 3ξ )  ϕ 2 
 sr r sr r  Sons.
[9] Douven, L.F.A., Towards the Computation of properties of
where ε rr and ε θθ are the radial and hoop strains respectively,
Injection Molded Products: Flow and Thermally Induced
κ rr and κ θθ are the radial and hoop curvatures respectively, Stresses in Amorphous Thermoplastics. 1991, Eindhoven
and r is the radius. Eq. 22 can be written in matrix form: University of Technology.
[10] Denizart, O., M. Vincent, and J.F. Agassant, 1995, "Thermal
n
Stresses and Strains in Injection Moulding: Experiments and
ε = ∑ B i ui (23)
i =1
Computations". Journal of Materials Science. 30(2): pp. 552-
560.
where B i is the strain-displacement matrix.

The incremental formulation of strain-displacement can be

derived in a similar way to be: