A Thesis
Presented to
In Partial Fulfillment
Master of Science
Fengxia Ouyang
December, 2005
ABAQUS IMPLEMENTATION OF CREEP FAILURE
Fengxia Ouyang
Thesis
Approved: Accepted
________________________________ _______________________________
Advisor Dean of the College
Wieslaw Binienda George K. Haritos
________________________________ ______________________________
Faculty Reader Dean of the Graduate School
Pizhong Qiao George R. Newkome
________________________________ ______________________________
Department of Chair Date
Wieslaw Binienda
ii
ABSTRACT
applications for their light weight and durability. However, the numerical modeling of
these materials poses several challenges. This is primarily due to the highly anisotropic
nature of the creep exhibited by these materials above the glass transition temperature.
Also, the damage and failure of the material is of particular interest to designers using
the PMCs. Recently, a sustained effort has been to provide the designer with large
large number of internal variables and with the most general mathematical forms, for
use in structural design analysis. Although elegant and useful, such constitutive laws are
often expensive in implementation. Specially for early stages of the design, a quicker
way of estimating complicated PMC behavior is needed. In this work, the constitutive
material law by Robinson and Binienda (2001) [1,2] is utilized for such an approach.
Although the material law includes a single scalar parameter to describe the damage, it
iii
are obtained from experiments of thinwalled tubular specimens reinforced with
loading . The model correctly predicts the relation between logarithmic creep rate and
logarithmic stress. The user subroutine has robust convergence properties. The creep
strain rate and the effect of damage on the creep strain rate are presented for the
benchmark problem of a square plate with a circular hole at the center and pressure
vessel. The effect of fiber orientation on the durability of the square plate and pressure
iv
TABLE OF CONTENTS
Page
LIST OF TABLES…………………………………………………………….…...…..vii
LIST OF FIGURES........................................................................................................viii
CHAPTER
I. INTRODUCTION…………………………………………………….……...…1
v
VI. SHELL ELEMENT IMPLEMENTATION……………………………….…..49
VII. CONCLUSION……………………………………………………...….……..62
REFERENCES………………………………………………………….………….….64
APPENDICES…………………………………………………………………….…...66
APPENDIX A UMAT………………………………………….………......67
vi
LIST OF TABLES
Table Page
vii
LIST OF FIGURES
Figure Page
2.2 Thin wall tube under tension and torsion loading. (a) thin wall tube
with fiber orientation. (b) typical experimental creep data under
axial and shear loading……………………………………………………………..7
2.3 Nondimensional log creep rate vs. log stress for the complete
exploratory data set. Tensile and shear data are shown as indicated.
Fiber angles θ = 60 o and 45 o and shear data for θ = 90 o shifted to
form a master curve…………………………….……………………………...…..8
.
4.1 UMAT Loop ………………………………………………………………….….16
5.2 Comparison between power law and Robinson creep model for
isotropic case under a constant load of 45 MPa when the scalar
damage variable is not included in the Robinson creep model.
(a) Time evolution of creep strain in Y direction.
(b) Time evolution of creep strain in Y direction.
(c) Time evolution of creep strain in XY direction……………………….….…..29
viii
5.4 Time evolution of creep strain in the direction of loading (without
damage). (a) for different orientations of the fiber (0, 45, 90 deg)
at 45MPa load. (b) for 45 deg fiber orientation at varying loading
20, 45 and 100PMa…………………………………………………….….….…..31
5.8 Time evolution of creep strain with different orientations of the fiber
(0, 45, 90 deg) at 46MPa load in the direction of loading
(without damage). (a) time evolution of strain at 1 direction
(b) time evolution of shear strain ……………………………………………….36
5.9 Left, geometry and right, 2D quarter symmetry model for the
square plate with a hole problem…………………………………………….…...37
5.10 Contour plot of (a) stress distribution of the power law for isotropic
material after elastic step, (b) stress distribution of the Robinsons’
model law for isotropic material after elastic step, (c) stress
distribution of the power law for isotropic material after creep
step (5 hours), (d) stress distribution of the Robinsons’ model law
for isotropic material after creep step (5 hours)………………………..…...…..40
5.12 Contour plot of Creep strain distribution after 5 hours creep response.
(a) Robinson Damage Model and (b) Isotropic Power Law……..………....…..42
5.13 Creep strain along the hole (quarter circle) going in a counter
clockwise direction for the Robinson Damage Model and for
isotropic Power Law Creep model (after 5 hours)……….………………….…..43
ix
5.14 Comparison between Analytic solution and Robinson model (FEA)
in elastic step. (a) R=20mm and R/L=13%. (b) R=50mm and R/L=33%…........44
5.15 Creep / time response along the hole (quarter circle) going in a
counter clockwise with damage effect under axial stress 10MPa
with R=20mm, 50mm, 80mm, L=152.4mm, fiber orientation 45 deg……….....45
5.16 Creep / time response along the hole (quarter circle) going in a
counter clockwise with damage effect under axial stress 10MPa
with R=50mm, L=152.4mm, fiber orientation 45 deg and 90 deg………….…..46
5.17 Stress concentration zone with damage effect under axial stress
10Mpa fiber orientation 45 deg.
(a) stress concentration zone with R/L=0.1.
(b) stress concentration zone with R/L=0.3.
(c) stress concentration zone with R/L=0.5…………………………..……...….47
5.18 Stress compression zone with damage effect under axial stress
10Mpa fiber orientation 45 deg.
(a) stress compression zone with R/L=0.1.
(b) stress compression zone with R/L=0.3.
(c) stress compression zone with R/L=0.5……………………………...…….....48
6.1 Thin wall tube under tension and torsion with fiber orientation,…………....……53
6.2 Time evolution of creep strain under a constant tensile load 45 MPa
of thinwalled tube for different fiber orientations (0, 45, 90 deg),
(a) time evolution of maximum principle strain.
(b) time evolution of shear strain……………………………………………........54
6.5 Maximum Strain along path1 of the vessel with fiber orientation
0, 45, 90 deg under inside pressure 0.5Mpa with damage evolution
after 10 hours…………………………………………………………………......59
x
6.7 Time evolution of Maximum Principle Strain of the pressure vessel
with fiber orientation 60, 90 deg under pressure 1Mpa with damage
evolution. (a) fiber orientation 60 deg, (b) fiber orientation 90 deg……………....61

xi
CHAPTER I
INTRODUCTION
The present work details the development of a computational material model for
applications for their light weight and durability. However, the numerical modeling of
these materials poses several challenges. This is primarily due to the highly anisotropic
nature of the creep exhibited by these materials. Also, the damage and failure of the
material is of particular interest to designers using the polymer matrix composites. Over
recent decades, a sustained effort has been to provide the designer with large computer
state variables and the most general mathematical forms, for use in structural analysis in
support of design. Although this approach may be appropriate in the final stages of
design and where complex histories of stress and temperature are involved, it can be too
complicated for many design applications, particularly in the early stages of design, and
The constitutive material law based on a Norton /Bailey type of creep law by
Robinson and Binienda (2001) [1,2] is utilized for this computational model. The
and utility (on behalf of the designer), while still retaining the essence of the actual
material behavior.
When strongly reinforced PMC materials and structures operate in the creep range
of their polymer matrix (at or near Tg) they undergo timedependent deformation and
materials where long or continuous fibers are embedded in the polymer matrix. By
design, much of the load in such materials is carried by the strong fibers that creep very
little, if at all. Evidently, design engineers need quantitative tools for predicting the
designs, e.g., optimal fiber configurations. Although it is often observed that highly
Binienda (2001) [1], creep deformation models do not commonly include dependence
on hydrostatic stress; this dependence is a principal feature of the material law that is
implemented in this work. The model incorporates a dissipation potential function that
with transverse isotropy. Failure and damage are introduced in the model with a
2
CHAPTER II
The transversely isotropic viscoelasticity model of concern derives from the more
present model is an extension of that earlier work in the sense that hydrostatic stress is
taken into account in the deformation response. The extension follows Robinson et al.
(1994) and Robinson and Binienda (2001)b. The viscoelasticity model has the form
ε& ij 3 n −1 Γij 1
= Φ (2.2)
ε& o 2 σo ψ n
1 1
ψ
& =− ∆ν m (2.3)
( m + 1) t o ψ
where ε eij denotes the components of elastic strain, e& ij and ε& ij denote components of the
total and creep (viscous) deformation rates, respectively. σ ij are the components of
3
For transverse isotropy, the functions Φ and ∆ are taken to depend on the
following invariants of stress σ ij , deviatoric stress sij and a material orientation tensor
D ij
.
1
J2 = s ij s ji J o = D ij s ji I = σ ii I o = D ij σ ji J = D ij s jk s ki (2.4)
2
1 1
Φ= 3[J 2 − ξ(J − J 2o ) − (ζ − η)J 2o + (ζ − 4 η)I 2 ] (2.5)
σo 9
and
Γij = ∂Φ = sij − ξ ( Dik s kj + sik Dkj − 2 J o Dij ) −
∂σ ij
(2.6)
2(ζ − η ) J o ( Dij − δ ij ) + (ς − 4η )I∂ ij
1 2
3 9
various inequalities based on physical limitations that are specified in Robinson and
Binienda (2001) [2]. Also the isochronous damage function is specified as:
∆ = ∆ (σ ij , D ij ) = ∆ ( N , S) (2.7)
1/ p
⎡⎛ N ⎞ p ⎛ ⎞ ⎤
p
∆ ( N, S) = ⎢⎜ ⎟ + ⎜ αS ⎟ ⎥ (2.8)
⎢⎜⎝ σ o ⎟⎠ ⎜σ ⎟ ⎥
⎣ ⎝ o⎠ ⎦
in which, the invariant N specifies the maximum tensile stress normal to the local fiber
direction, and the invariant S denotes the local maximum longitudinal shear stress as:
1 1
N= (I − I o ) + J 2 + J 2o − J (2.9)
2 4
S = J − J o2 (2.10)
4
The angular brackets in (9) are the MacCauley brackets
Fig 2.1 illustrates the isochronous damage function in N S space for a general
power law form as well as the special case when the function is linear in N and S (P=1).
Fig 2.1 Illustration of the isochronous damage function in the normal stress (y axis).
P=1 is the linear form.
5
The material parameters are obtained from experiments of thinwalled tubular
under tensile and shear loading (Fig 2.2). The complete exploratory data set is plotted
in Fig 2.3. The correlation of the theoretical model and experimental data (solid lines)
and of creep response under two of the untested natural stress states (TS)(dotted line)
and (LN)(dashed line) validates the physics behind the proposed constitutive equation.
As a definitive measure of the strength of anisotropy, the creep rate under tensile stress
along the fiber direction (LN) is predicted as being less than one thirtieth (1/30) of that
6
2
T τ
F F
θ σ σ
1
T
3 τ
(a)
0.0040
axial and shear creep strain (%)
0.0035
0.0030
axial strain
0.0025
shear strain
0.0020
0.0015
0.0010
0.0005
0 1 2 3 4 5 6 7
time (hr)
(b)
Fig 2.2 Thin wall tube under tension and torsion loading. (a) thin wall tube with fiber
orientation. (b) typical experimental creep data under axial and shear loading.
7
1
log(ε /ε ΤΝ ) or log(γ / V 3 ε TN )
0
. _ .
1
90ten
60ten
45ten
. .
90shear
2 Correlations
(LN)  prediction
(TS)  prediction
3
0.4 0.2 0.0 0.2 0.4
Fig 2.3 Nondimensional log creep rate vs. log stress for the complete exploratory data
set. Tensile and shear data are shown as indicated. Fiber angles θ = 60 o and 45 o and
shear data for θ = 90 o shifted to form a master curve.
8
CHAPTER III
The powerlaw model can be used to do the creep calculation for isotropic material.
ε& cr = Aσ t m (3.1)
where
behavior and n must be positive and . Since total time is used in the
expression, such reasonable behavior also typically requires that small step times
compared to the creep time be used for any steps for which creep is not active in an
9
σ n n +1
ξ = ζ =η = 0
Now we obtained
Depending on the choice of units for either form of the power law, the value of A
may be very small for typical creep strain rates. If A is less than 10 −17 , numerical
difficulties can cause errors in the material calculations; therefore, use another system
10
CHAPTER IV
an extremely powerful and flexible tool for analysis. This chapter defines the interfaces
User subroutine UMAT can be used to define the mechanical constitutive behavior
of a material; it must update the stresses and solutiondependent state variables to their
values at the end of the increment for which it is called it must provide the material
Jacobian matrix, for the mechanical constitutive model; it can be used in conjunction
with user subroutine USDFLD to redefine any field variables before they are passed in.
interactions with external data files that are used in conjunction with user subroutines.
per increment, for use during the analysis; or output quantities that are accumulated
over multiple elements in COMMON block variables within user subroutines may need
11
to be written to external files at the end of a converged increment for postprocessing.
Such operations can be performed with user subroutine UEXTERNALDB. This user
interface can potentially be used to exchange data with another code, allowing for
User subroutines should be written with great care. To ensure their successful
Every user subroutine must include the statement. As the first statement after the
argument list. The file ABA_PARAM.INC is installed on the system by the ABAQUS
installation procedure. It specifies either IMPLICIT REAL*8 (AH, OZ) for double
precision machines or IMPLICIT REAL (AH,OZ) for single precision machines. The
ABAQUS execution procedure, which compiles and links the user subroutine with the
necessary to find this file and copy it to any particular directory; ABAQUS will know
1) User subroutines must perform their intended function without overwriting other
parts of ABAQUS. In particular, the user should redefine only those variables
which the user subroutine is the only complicated aspect of the model before
attempting to use them in production analysis work. If needed, debug output can
FORTRAN unit 6 to appear in the data (.dat) file; these units should not be
12
opened by the user's routines since they are already opened by ABAQUS.
FORTRAN units 15 through 18 or units greater than 100 can be used to read or
write other userspecified information. The use of other FORTRAN units may
interfere with ABAQUS file operations. These FORTRAN units must be opened
by the user; and because of the use of scratch directories, the full pathname for
3) Solutiondependent state variables are values that can be defined to evolve with
The equation of motion together with the constitutive law form system consisting
motion is solved with the help of a finiteelement package (ABAQUS), and the
constitutive law by a solver for ordinary differential equations. The relevant constitutive
loading as well as with the time increment ∆t , and an initial guess ∆ε n , for the strain
increment. The user subroutine UMAT has to supply ABAQUS with new Cauchy stress
derivative of stress with respect to the strain increment. With this information, a new
guess for the strain increment is calculated and the whole procedure is iterated until
detail introduction of definition of Jacobian Matrix in UMAT and the Newton Ralphson
The following diagram (Fig 4.3) indicates a certain step of UMAT working with
ABAQUS.
total strain increment ∆ε total (t n ) and total strain ε total (t n ) for the strain increment. And
also
σ (t n ) is calculated by previous increment. All these four values will pass to UMAT to
3 Γ
∆ε = Φ n −1 ∆t
2 σ0
Also, the relationship between total strain, creep strain and elastic strain is
∆ε total (t n ) = ∆ε e (t n ) + ∆ε (t n )
so
∆ε e (t n ) = ∆ε total (t n ) − ∆ε (t n )
In this way we can update Cauchy Stress tensor. The stress increment is
∆σ (t n ) = J ⋅ ∆ε e (t n )
σ (t n + ∆t ) = σ (t n ) + ∆σ (t n )
14
STEP 3: Strain Update
STEP 5: ABAQUS equilibrium iterations at new time t n+1 , the maximum iteration
number is set to 9 in ABAQUS and error tolerance is set to TOL 5e3. If less than 9
times iterations the error is less than TOL, it calls convergence, n=n+1 and move to
another increment. If the error is larger than TOL, ABAQUS will reduce the ∆t go to
15
step1
step n
ABAQUS supply information at time tn
Time increment ∆t
Total strain increment ∆ε (tn )
total
ε
Total strain n(t )
UMAT step2
3 Γ
Dr. Robinson’s creep law ∆ε ij = Φ n−1 ij ∆t
2 σo
∆ε total (tn ) = ∆ε e (tn ) + ∆ε (tn )
∆εije (tn ) = ∆εijtotal(tn ) − ∆εij (tn )
Stress update
∆σ (tn ) = J ⋅ ∆ε e (tn )
σ n+1 = σ (tn + ∆t ) = σ (tn ) + ∆σ (tn )
step3
ABAQUS update
Strain update
ε n+1 = ε
total total
(tn + ∆t ) = ε total (tn ) + ∆ε total (tn )
step4
step5
ABAQUS equilibrium
iterations at time tn+1
16
4.2.1 Jacobian Matrix for plane stress element and shell element
the free surface of a structural element, such as the surfaces of thinwalled pressure
vessels under external or internal pressure, the free surfaces of shafts in torsion and
beams under transverse load has one principal stress that is much smaller than the other
two. By assuming that this small principal stress is zero, the threedimensional stress
state can be reduced to two dimensions. Since the remaining two principal stresses lie in
Assume that the negligible principal stress is oriented in the zdirection. To reduce
the 3D stress matrix to the 2D plane stress matrix, remove all components with z
subscripts to get,
17
⎡σ x τ xy ⎤
⎢τ σ y ⎥⎦
⎣ yx
where τ xy = τ yx for static equilibrium. The sign convention for positive stress
components in plane stress is illustrated in the above Fig 4.1 on the 2D element.
Jacobian matrix of the constitutive model for plane stress and shell element,
∂∆σ / ∂∆ε , where ∆σ are the stress increments and ∆ ε are the strain increments can
Fig 4.3 Plane Stress Transformation from local coordinate (xy) to global coordinate (1
2)
18
According to the Fig 4.3 we can calculate the resultant force in 12 coordinate:
σ 1 = m 2σ x + n 2σ y + 2mnτ s (4.3)
where
We conclude that:
⎡ ⎤ ⎡ ⎤
⎢ ε1 ⎥ ⎢ εx ⎥ → [ε ]1, 2 = [T ][ε ]x , y (4.7)
⎢ ε ⎥ = [T ]⎢ ε ⎥
⎢1 2 ⎥ ⎢ y ⎥
⎢ γ6⎥ ⎢1 γ ⎥
⎣2 ⎦ ⎢⎣ 2 s ⎥⎦
matrix
19
⎡ ⎤ ⎡ ⎤
⎡σ 1 ⎤ ⎡Q11 Q12 0 ⎤ ⎢ ε1 ⎥ ⎡σ 1 ⎤ ⎡Q11 Q12 0 ⎤ ⎢ ε1 ⎥
⎢σ ⎥ = ⎢Q Q22 ⎥
0 ⎥ ε 2 ⇒ [T ]⎢⎢σ 2 ⎥⎥ = ⎢⎢Q12
⎢ ⎥ Q22 0 ⎥⎥[T ]⎢ ε 2 ⎥ ⇒
⎢ 2 ⎥ ⎢ 12 ⎢ ⎥ ⎢ ⎥
⎢⎣τ 6 ⎥⎦ ⎢⎣ 0 0 2Q66 ⎥⎦ ⎢ 1 γ ⎥ ⎢⎣ τ s ⎥⎦ ⎢⎣ 0 0 2Q66 ⎥⎦ ⎢ 1 γ ⎥
⎣⎢ 2 ⎦⎥ ⎣⎢ 2 ⎦⎥
6 s
⎡ ⎤
⎡ ⎤ ⎡σ x ⎤ ⎡Qxx Qxy 2Qxs ⎤ ⎢ ε x ⎥
⎡σ x ⎤ ⎡Q11 0 ⎤ ⎢ εx ⎥ ⎢σ ⎥ = ⎢Q ⎥
2Qys ⎥ ⎢ ε y ⎥
Q12
⎢ y⎥
[ ]
⎢σ ⎥ = T −1 ⎢Q
⎢ 12 Q22 0 ⎥⎥[T ]⎢ ε y ⎥ ⇒
⎢ ⎥
⎢ y ⎥ ⎢ yx
⎢⎣ τ s ⎥⎦ ⎢⎣Qsx
Qyy
⎢ ⎥
Qsy 2Qss ⎥⎦ ⎢ 1 γ ⎥
⎢⎣ τ s ⎥⎦ ⎢⎣ 0 0 2Q66 ⎥⎦ ⎢ 1 γ ⎥ ⎢⎣ 2 s ⎥⎦
⎢⎣ 2 s ⎥⎦
Eq. (4.8) is the formula to calculate the components in Jacobian Matrix for plane stress
algorithm that uses the first few terms of the Taylor series of a function f(x) in the
iteration, although in this work the latter term is reserved to the application of Newton's
method for computing square roots. For f(x) a polynomial, Newton's method is
essentially the same as Horner's method. The Taylor series of f(x) about the point
x = x0 + ε is given by
20
1 (4.9)
f ( x0 + ε ) = f ( x0 ) + f ' ( x0 )ε + f " ( x0 )ε 2 + ...
2
f ( x0 + ε ) = f ( x0 ) + f ' ( x0 )ε (4.10)
This expression can be used to estimate the amount of offset ε needed to land closer to
the root starting from an initial guess x0 . Setting f ( x0 + ε ) = 0 and solving (4.10) for
ε = ε 0 gives
f (x0 ) (4.11)
ε0 = −
f ' (x0 )
calculating a new ε 1 , and so on, the process can be repeated until it converges to a root
using
f ( xn ) (4.12)
ε0 = −
f ' ( xn )
extremum. However, with a good initial choice of the root's position, the algorithm can
f ( xn ) (4.13)
xn+1 = xn −
f ' ( xn )
For n=1, 2, 3, .... An initial point that provides safe convergence of Newton's method
= f ( xn ) (4.15)
εn −
f ' ( xn )
21
But
= 1 (4.16)
f ( x0 ) f ( xn−1 ) + f ' ( xn−1 )ε n + f " ( xn−1 )ε n2 + ...
2
= 1 (4.17)
f ' ( xn−1 )ε n + f " ( xn−1 )ε n2 + ...
2
so
1
f ' ( xn−1 )ε n + f " ( xn−1 )ε n2 + ...
f ( xn ) = 2 (4.19)
f ' ( xn ) f ' ( xn−1 ) + f " ( xn−1 )ε n + ...
1
f ' ( xn−1 )ε n + f " ( xn−1 )ε n2 .
≈ 2 (4.20)
f ' ( xn−1 )
ABAQUS sets the error tolerance TOL 5e3 and maximum iteration times 9.
22
CHAPTER V
In the previous chapter we introduce the theory of user subroutine UMAT and how
to implement user material law through UMAT. When developing user subroutines, test
them thoroughly on smaller examples in which the user subroutine is the only
complicated aspect of the model before attempting to use them in production analysis
We first investigate the oneelement tension test, given in Fig 5.1. The dimension
23
Y Y
u
σ
θ θ
(a) (b)
Fig 5.1 One element model problem set up with fiber orientation. (a) creep loading and
(b) constant displacement loading.
Fig 5.1 shows the schematic with boundary conditions for a single element model.
This particular test was done for the reduced integration plane stress elements with
linear and quadratic interpolation schemes ( ABAQUS element type CPS4R and
CPS8R). The elements were tested for response under constant load (creep) as well as
24
for this purpose are the ones obtained in the thin cylinder experiment described above
t 0 = 12.0hours
This particular set of parameters assumes that the isochronous damage function
Fig 5.2 shows comparison between power law and Robinson creep model for
isotropic case under a constant load of 45 MPa when the scalar damage variable is not
included in the Robinson creep model. The Fig 5.2.(a) indicates that at creep strain
evolution after 10 hours at 1 direction calculated by power law model and Robinson
creep model are agree with each other, the same to 2 direction (Fig. 5.2.(b)). But the
creep strains evolution after 10 hours for 12 direction are different for each material
n +1
σ n
ε&11 = ε&22 = γ&12 = ε&0 ( ) (1 − ξ ) 2
σ0
n +1
σ n
ε&11 = ε&0 ( ) (1 − ξ ) 2
σ0
n +1
σ n
ε&22 = ε&0 ( ) (1 − η ) 2
σ0
25
n +1
3τ
γ&12 / 3 = ε&0 ( ) n (1 − ξ ) 2
σ0
ξ = ζ =η = 0
In this case the Power Law equation reduced to
σ n
ε&11 = ε&22 = γ& = ε&0 ( )
σ0
σ n
ε&11 = ε&22 = ε&0 ( )
σ0
3τ
γ&12 / 3 = ε&0 ( )n
σ0
Fig 5.3 shows the response of a single element creep loading test under a constant
load of 46 MPa with fiber orientation 90 deg, the scalar damage variable is not included.
After compare FEA result with experimental date we observed that both experiment and
FEA calculation is with creep rate relatively constant at 0.013%/hr. We also compare
UMAT results with experiment data under different tensile stress with fiber orientation
45 and 90 deg. The UMAT results well coincide with experiment data (Table 1 and
Table 2).
Fig 5.4 shows the response of a single element creep loading test. The Fig 5.4.a
shows the time evolution of creep strain under a constant load of 45 MPa for different
fiber orientations (0 deg, 45 deg, 90 deg) when the scalar damage variable is not
included in the material model. It is observed that when the creep loading is along fiber
directions, the PMC has the strongest behavior. The creep strains are progressively
26
higher when the creep loading is at progressively higher angle to the fiber orientation.
For the 90 deg orientation, failure can be achieved in roughly 10 hours. Similarly, for a
given orientation, the creep strains increase progressively, as the creep load is increased
(Fig. 5.4.(b)).
Fig 5.5.(a) shows the time evolution of creep with fiber orientations 45 deg under
loads equal to 60MPa when the scalar damage variable is included. We find that when
include this damage factor the plot is getting curved and the value of the creep strain is
higher than that without this damage factor. Fig 5.5.(b) shows the time evolution of
creep with fiber orientations 45 deg under different loads (when the scalar da70Mpa,
75Mpa, 80Mpa) with damage variable is included. For creep load less than 70 MPa, the
evolution of creep strain with time is linear. But above a critical load, the increased
damage results in a larger creep strain. We also plot the time evolution of shear strain
with different fiber orientation 0 deg, 45 deg and 90 deg under shear loading 40 MPa
(Fig 5.5.(c)). It can be observed that 0 deg and 90 deg have the same strain evolution
behavior and 45 deg has more strain evolution than 0 deg and 90 deg.
27
Table 1 fiber orientation 90 deg under tension
28
.0025
.0002
.0015
.0004
ε11
ε11
.0010
.0006
.0005 .0008
0.0000 .0010
.0012
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Time(hr) Time(hr)
(a) (b)
4e17
Robinson model
power law
3e17
2e17
ε11
1e17
0 2 4 6 8 10 12
Time(hr)
(c)
Fig 5.2 Comparison between power law and Robinson creep model for isotropic case
under a constant load of 45 MPa when the scalar damage variable is not included in the
Robinson creep model. (a) Time evolution of creep strain in Y direction. (b) Time
evolution of creep strain in Y direction. (c) Time evolution of creep strain in XY
direction
29
0.10
UMAT
Experiment
0.08
0.04
0.02
0.00
0 2 4 6 8
time (hr)
Fig 5.3 Time evolution of creep strain in the direction of loading (without damage) for
fiber orientation of 90 deg at 45MPa load.
30
.0010
00
.0008 452
900
.0006
ε11
.0004
.0002
0.0000
0 1 2 3 4 5 6
Time(hr)
(a)
.0010
20Mpa
.0008 46Mpa
60Mpa
.0006
ε11
.0004
.0002
0.0000
0 1 2 3 4 5 6
Time(hr)
(b)
Fig5.4 Time evolution of creep strain in the direction of loading (without damage). (a)
for different orientations of the fiber (0, 45, 90 deg) at 45MPa load.
(b) for 45 deg fiber orientation at varying loading 20, 45 and 100PMa.
31
.0020
.0010
ε11
.0005
0.0000
0 2 4 6 8 10 12
Time(hr)
(a)
.005
70Mpa
.004 75MPa
80MPa
.003
ε11
.002
.001
0.000
Time(hr)
(b)
6e4
0 deg
5e4 45 deg
90 deg
4e4
3e4
12
ε
2e4
1e4
0 1 2 3 4 5 6
Time (hr)
(c)
the right boundary (u=0.1%). The dimension is still 10mm x 10mm. We compare stress
calculated by Robinson’s creep law with that by power law in isotropic case. We find
out that after 10 hours test the stress along loading direction calculated by Robinson
Creep model is lower than the result calculated by Power law mode, which is due to the
Power Law
Robinson Model
Fig 5.6: Time evolution of stress relaxation in the direction of loading (with damage).
33
Results for both the creep and relaxation tests were able to match the results from
an isotropic Power Law Creep Model when damage was not included. However, the
proposed model deviates from the Power Law model when creep anisotropy due to
orientation and damage is included. These element tests are achieved for a wide range
of loads and orientations, demonstrating the robustness of the numerical scheme. Also,
the tests show the usefulness of the constitutive model despite the use of a single scalar
One element tests are demonstrating the robustness of the numerical scheme of the
UMAT. In this part we extend our model to relatively complex model – a square plate
with multiple elements. The dimension of the plate is 20mm x 20 mm. This particular
test was done for the reduced integration plane stress element with linear and quadratic
interpolation schemes (ABAQUS element type CPS4R and CPS8R). Geometry and
boundary condition of the plate under tension is shown on Fig 5.7. We apply tensile
stress 46 Mpa stress along 1 direction with different fiber orientation (0 deg, 45 deg and
90 deg). Fig 5.8 shows the time evolution of creep strain with fiber orientations 0, 45
and 90 deg under tensile loads equal to 46MPa. Fig 5.8.(a) shows the 1 direction strain
distribution and Fig 5.8.(b) shows the shear strain distribution. All these results coincide
34
σ σ
20mm
35
.0008
00
450
0
.0006 90
.0004
ε11
.0002
0.0000
0 1 2 3 4 5 6
Time (hr)
(a)
4e5
00
450
3e5
900
2e5
ε12
1e5
1e5
0 1 2 3 4 5 6
Time (hr)
(b)
Fig5.8 Time evolution of creep strain with different orientations of the fiber (0, 45, 90
deg) at 46MPa load in the direction of loading (without damage). (a) time evolution of
strain at 1 direction. (b) time evolution of shear strain.
36
5.3 Plate with a hole in the middle
This particular test was done for the reduced integration plane stress element with
linear and quadratic interpolation schemes (ABAQUS element type CPS4R and
CPS8R). Geometry and boundary condition of the plate under tension and torsion is
shown on Fig 5.8. We apply tensile stress 10 MPa along 1 direction with different fiber
L=152.4mm
σ
R=80mm
Fig 5.9 Left, geometry and Right, 2D quarter symmetry model for the square plate with
a hole problem.
37
First of the study we choose isotropic material to compare Robinson creep Model
with Power law creep model. The Power law model is one of the creep model
ABAQUS use to do creep calculation for isotropic material. Fig 5.10 shows the stress
distribution of the Robinsons’ model law for isotropic material and comparative stress
distribution for a Power Law type model. It can be seen that stress distribution after
instantaneous elastic response for the proposed Creep Damage model (Fig 5.10.(b)) is
identical to a Power Law model (Fig 5.10.(a)). This is because in elastic step both
power law and Robinson model has the same constitutive equation. But due to the
damage effect we take into consideration in Robinson model we can observe some
difference existing after 5 hours creep response in Fig 5.10.(c) and Fig 5.10.(d).
To further compare these two models we plot stress distribution along the hole
(counter clockwise) because this area has the most complex behavior of the whole plate.
Fig 5.11 shows the distribution of Stress distribution along the hole, after elastic step (a)
and after 5 hours of creep load (b). Fig 5.119.(a) shows the plots along the hole for both
creep law are identical to each other after elastic step. Fig 5.119.(b) shows lower
stresses in the Robinson Model at the second half of the plot is a result of higher stress
relaxation from higher stress concentration area due to the damage evolution in
Robinson Model. Fig 5.12 shows the creep strain distribution for Power law and
Robinson creep Model after 5 hours creep response. Robinson Model has more creep
strain existing in the high stress concentration zone. Fig 5.13 shows the creep strain
distribution along the hole (counter clockwise), after 5 hours of creep load. We can
observe that at the first part of the path Robinson model is agree with Power law model.
At the second half of the path Robinson model undergoes more creep strain than power
38
law that is due to the damage factor we include in the calculation of creep strain in
Robinson Model. We also find the highest value located in different area in these two
models. The highest value for Power law located in the very end of the path while for
Robinson model highest value of stress redistribution is in the area close to the end of
the path which is due to the anisotropic constitutive equations of Robinson Model even
To verify the accuracy of the numerical scheme we compare the numerical results
with analytical solutions. We choose the radius of the plate 20mm and 50mm with fiber
orientation 90 deg (Fig 5.14). We compare each case with analytical solution. Both two
plots indicated the accuracy of the numerical result in positive value area. We can find
that there is big difference in negative area. Analytic solution has much lower value
than numerical solution in negative zone. The discrepancy is due to the reason that
Fig 5.15 shows the stress distribution along the hole (counter clockwise), after 5
hours of creep load 10 Mpa with R=20mm, 50mm, 80mm, L=152.4mm and fiber
orientation 45 deg. Fig 5.16 shows the stress distribution along the hole (counter
orientation 45deg and 90 deg. Fig 5.17 shows Stress concentration zone with damage
effect under axial stress 10Mpa fiber orientation 45 deg (a)R/L=0.1 (b)R/L=0.3 (c)
R/L=0.5. Fig 5.18 shows Stress compression zone with damage effect under axial stress
10Mpa fiber orientation 45 deg (a) R/L=0.1 (b) R/L=0.3 (c) R/L=0.5. All these results
indicates that the computational routine (UMAT) successfully describes the rather
39
78MPa
78MPa
50MPa
50MPa
38MP
38MPa
20MPa 20MPa
(a) (b)
40MPa
38MPa
20MPa 20MPa
(c) (d)
Fig 5.10 Contour plot of (a) stress distribution of the power law for isotropic material
after elastic step, (b) stress distribution of the Robinsons’ model law for isotropic
material after elastic step, (c) stress distribution of the power law for isotropic material
after creep step (5 hours), (d) stress distribution of the Robinsons’ model law for
isotropic material after creep step (5 hours).
40
100
Robinson model
80 powerlaw
60
σ11(MPa)
40
20
20
0 20 40 60 80 100 120 140
(a)
50
Robinson model
40 power law
30
σ11(MPa)
20
10
10
0 20 40 60 80 100 120 140
(b)
Fig 5.11 Stress distribution along the hole (quarter circle) going in a counter clockwise
direction for the proposed Creep Damage Model as and for isotropic Power Law Creep
model. (a) stress distribution after elastic step. (b) stress distribution after 5 hours.
41
0.003%
0.014%
0.033%
0.051%
0.073%
(a)
0.003%
0.014%
0.033%
0.051%
0.073%
(b)
Fig 5.12 FEA plot of Creep strain distribution after 5 hours creep response.
(a) Robinson Damage Model and (b) Isotropic Power Law.
42
.0012
Robinson model
.0010
power law
.0008
.0006
ε11
.0004
.0002
0.0000
.0002
0 20 40 60 80 100 120 140
Fig 5.13 Creep strain along the hole (quarter circle) going in a counter clockwise
direction for the Robinson Damage Model and for isotropic Power Law Creep model
(after 5 hours).
43
40
30 Robinson model
Analytic solution
20
σ11(MPa)
10
θ
10
20
30
0 20 40 60 80 100
(a)
40
Robinson model
30 Analytic solution
20
σ11(MPa)
10
10
20
30
0 20 40 60 80 100
(b)
Fig 5.14 Comparison between Analytic solution and Robinson model (FEA) in elastic
step. (a) R=20mm and R/L=13%. (b) R=50mm and R/L=33%.
44
60
50 R=80
R=50
R=20
σ11(ΜPa) 40
30
20
10
10
0 20 40 60 80 100 120 140
Fig 5.15 Creep / time response along the hole (quarter circle) going in a counter
clockwise with damage effect under axial stress 10MPa with R=20mm,50mm,80mm,
L=152.4mm, fiber orientate 45 deg.
45
40
450
900
30
σ11(MPa) 20
10
10
0 20 40 60 80 100
θ (deg)
Fig 5.16 Creep / time response along the hole (quarter circle) going in a counter
clockwise with damage effect under axial stress 10MPa with R=50mm, L=152.4mm,
fiber orientate 45 deg and 90 deg.
46
20MPa
25 MPa
(a)
20MPa
40Mpa
20MPa (b)
25 MPa
30Mpa
40Mpa
45Mpa
56MPa
(c)
Fig 5.17 Stress concentration zone with damage effect under axial stress 10Mpa fiber
orientation 45 deg. (a) stress concentration zone with R/L=0.1. (b) stress concentration
zone with R/L=0.3. (c) stress concentration zone with R/L=0.5.
47
0.5MPa (a)
2Mpa
0.5MPa
6MPa
(b)
4Mpa
3Mpa
2Mpa
0.5MPa
(c)
Fig 5.18 Stress compression zone with damage effect under axial stress 10Mpa fiber
orientation 45 deg. (a) stress compression zone with R/L=0.1. (b) stress compression
zone with R/L=0.3. (c) stress compression zone with R/L=0.5.
48
CHAPTER VI
Shell elements are surface representations of structures that are much thinner in
one direction than the other two (thinwalled structures). The elements are
geometrically defined by three or four sided surfaces, and are located in space at the
midplane of the solid they are representing. The user specifies the thickness of the
elements as an input to the software. These elements are used in modeling all types of
thinwalled structures, such as airplane and automotive bodies, pressure vessels, sheet
Shell elements have six active degrees of freedom per node, much like beam
elements. Because of commonality in degrees of freedom, beam and shell elements are
49
Shell elements are good for modeling structures that are thin. These elements are
usually formulated under the assumptions governing thin plate theory. If a structure is
too thick, the behavior of thin plates is no longer seen (shear stresses become large, etc.),
and shell elements should not be used. This limit is usually seen at a thickness to width
Shell elements generally have a lower limit on this ratio as well. At thicknessto
width ratios between 1/100 and 1/1000, thin plates begin to behave like membranes,
with no bending stiffness (like a string in tension, subjected to transverse load). Because
of this shell elements cannot be used to model very thin, flexible structures such as
valid for thick and thin shell problems. See below for a discussion of what constitutes a
“thick” or “thin” shell problem. This concept is relevant only for elements with
and accurate solutions to most applications and will be used for most applications.
obtained with the thin or thick shell elements; for example, if only small strains occur
Element type S4R is generalpurpose shell. These elements allow transverse shear
deformation. They use thick shell theory as the shell thickness increases and become
50
discrete Kirchhoff thin shell elements as the thickness decreases; the transverse shear
Element type S8R should be used only in thick shell problems. Thick shells are
interpolation is desired. When a shell is made of the same material throughout its
thickness, this occurs when the thickness is more than about 1/15 of a characteristic
length on the surface of the shell, such as the distance between supports for a static case
First, the computational model was applied to the thinwalled tube (7.5 in dia)
under a tensile stress of 0.5 MPa. This particular test was done for the reduced
integration shell elements with linear and quadratic interpolation schemes (ABAQUS
element type S4R and S8R). Geometry and boundary condition of thin wall tube under
tension and torsion is shown on Fig 6.1. We apply tensile stress along 1 direction with
different fiber orientation (0 deg, 45 deg and 90 deg). Because the stress and strain
distributions are evenly along the tube, we take one element to analyze the strain
development. Fig 6.2 shows the time evolution of creep strain under a constant tensile
load of 45 MPa for different fiber orientations (0 deg, 45 deg, 90 deg) when the scalar
damage variable is included in the material model. It is observed that when the creep
loading is along fiber directions, the PMC has the strongest behavior. The creep strains
are progressively higher when the creep loading is at progressively higher angle to the
51
fiber orientation. For the 90 deg orientation, failure can be achieved in roughly 6 hours.
Fig 6.3 shows the time evolution of creep strain under a constant shear load of 45 MPa
for different fiber orientations (0 deg, 45 deg, 90 deg) when the scalar damage variable
52
2
T
F F
θ
1
T
3
Fig 6.1 Thin wall tube under tension and torsion with fiber orientation
53
(a)
.0052
.0050 0 deg
45 deg
.0048 90 deg
.0046
.0044
ε12
.0042
.0040
.0038
.0036
.0034
.0032
0 2 4 6 8 10 12
Time (hr)
(b)
Fig 6.2 Time evolution of creep strain under a constant tensile load (F) 45 MPa of thin
walled tube for different fiber orientations (0 deg, 45 deg, 90 deg). (a) time evolution of
maximum principle strain. (b) time evolution of shear strain.
54
Further the computational model was applied to the pressure vessel (20 cm dia)
under a pressure of 0.5 MPa. This particular test was done for the reduced integration
shell elements with linear and quadratic interpolation schemes (ABAQUS element type
S4R and S8R). Fig 6.3 shows the geometry, mesh and path of the pressure vessel, the
thickness is 0.4.
First we plot the path along path 1 (Fig 6.3) with different fiber orientation. 0 deg
is the direction along the path 1, 90 deg is the direction perpendicular to the path1. It
can be observed that 0 deg has the smallest strain deformation along the path1, the
higher angle of fiber, the higher strain deformation occurs. We can also find the
movement of the strain distribution along path 1 with fiber orientation of 45 deg (Fig.
6.4) after 10 hours evolution. This phenomena is also obvious in other fiber orientation.
Fig 6.4 shows path plot of time evolution of Maximum Strain along path1 of the vessel
with fiber orientation 45 deg under inside pressure 0.5Mpa with damage evolution. Fig
6.5 shows Maximum Strain along path1 of the vessel with fiber orientation 0, 45, 90
deg under inside pressure 0.5Mpa with damage evolution after 10 hours.
Fig 6.6 shows path plot of time evolution of Maximum Strain along path2 of the
vessel with fiber orientation 45, 60, 90 deg under inside pressure 0.5Mpa with damage
evolution. It can be observed that the obvious difference in maximum principle strain
between 45, 60 and 90 deg is at the center part of the pressure vessel and the 90 deg has
Fig 6.7 shows the time evolution of Maximum Principle Strain of the pressure
vessel with fiber orientation 60, 90 deg under pressure 1Mpa with damage evolution.
And we set the failure maximum strain is 1%. We can observe that the 60 deg vessel
55
reached failure at 4 hours and 90 deg vessel reaches failure about 3 hours. The strain
distributions at the dome area for both vessels are close to each other. The obvious
discrepancy occurs in the straight part of the vessel and in this area the fiber orientation
is important for the strain evolution and 90 deg has the weakest behavior. So it takes
56
Path 2
Path 1
57
.0032 .0036
.0034
.0030 .0033
.0032
.0029
.0031
.0028
.0030
.0027 .0029
0 20 40 60 80 0 20 40 60 80
Path 1 Path 1
(a) (b)
.0037
.0036
Maximum Principle Strain
.0035
.0034
.0033
.0031
0 20 40 60 80
Path 1
(c)
Fig 6.4 Path plot of time evolution of Maximum Strain along path1 of the vessel under
inside pressure 0.5Mpa with damage evolution (a) fiber orientation 45 deg. (b) fiber
orientation 60 deg. (c) fiber orientation 90 deg.
58
.0038
.0036
Maximum Principle strain
.0034
.0032
.0030
.0028
45 deg
60 deg
90 deg
.0026
0 10 20 30 40 50 60 70
Path 1
Fig 6.5 Maximum Strain along path1 of the vessel with fiber orientation 0, 45, 90 deg
under inside pressure 0.5Mpa with damage evolution after 10 hours
59
.0040
45 deg  10 hours
45 deg  elastic step
.0035 60 deg  10 hours
Maximum Principle Strain 60 deg  elastic step
90 deg  10 hours
90 deg  elastic step
.0030
.0025
.0020
.0015
0 10 20 30 40 50 60
Path 2
Fig 6.6 Time evolution of Maximum Principle Strain along path2 of the pressure vessel
with fiber orientation 45, 60, 90 deg under pressure 0.5Mpa with damage evolution
60
0.3% 0.2%
0.4% 0.4%
0.6% 0.6%
0.8% 0.8%
1% 1%
Fig 6.7 Time evolution of Maximum Principle Strain of the pressure vessel with fiber
orientation 60, 90 deg under pressure 1Mpa with damage evolution. (a) fiber orientation
60 deg. (b) fiber orientation 90 deg.
61
CHAPTER VII
CONCLUSION
This paper demonstrates the computational utility of the anisotropic, creep damage
model presented in a paper by Robinson, Binienda and Ruggles (2002). The references
tubular specimens fabricated from a model PMC. Thinwalled tubes are used not for
their interest as structural components but because they are convenient specimens for
PMCs. The routine is used with a commercially available code. The present work
demonstrates the utility of the creep damage law in describing the essential physics
behind creep damage using a single scalar parameter. The model is successfully applied
A primary assumption in the damage model, cf., Robinson et al. (2002), is that the
shear traction acting at the fiber/matrix interface. Accordingly, the isochronous damage
Exploratory data are generated to partially define the isochronous damage curve
∆( N , S) = 1 for the model PMC; evidently, a more extensive data base is required to fully
62
define ∆( N , S) = 1 and to verify that this stress dependence correlates directly with creep
failure. This maybe of general interest and the present code needs to be extended to
Also, the code may easily be extended to calculate creep rupture life based on the
deformation rate and the damage variable calculations. These are left for future work.
63
REFERENCES
6. LISSENDEN, C.J., LERCH, B.A., ELLIS, J.R. AND ROBINSON, D.N. (1997).
“EXPERIMENTAL DETERMINATION OF YIELD AND FLOW SURFACES
UNDER AXIALTORSIONAL LOADING.” STP 1280, A.S.T.M, 92112.
64
10. ROBINSON, D.N., TAO, Q. AND VERRILLI, M.J. (1994). “A
HYDROSTATIC STRESSDEPENDENT ANISOTROPIC MODEL OF
VISCOPLASTICITY.” NASA TM 106525.
14. ROBINSON, D.N., KIM, K.J AND WHITE, J.L. (2002). “CONSTITUTIVE
MODEL OF A TRANSVERSELY ISOTROPIC BINGHAM FLUID.” J. APPL.
MECHANICS, 69,(1), 18.
65
APPENDICES
66
APPENDIX A
UMAT FILE
SUBROUTINE UMAT(STRESS,STATEV,DDSDDE,SSE,SPD,SCD,
* RPL,DDSDDT,DRPLDE,DRPLDT,STRAN,DSTRAN,
* TIME,DTIME,TEMP,DTEMP,PREDEF,DPRED,MATERL,NDI,NSHR,NTENS,
* NSTATV,PROPS,NPROPS,COORDS,DROT,PNEWDT,CELENT,
* DFGRD0,DFGRD1,NOEL,NPT,KSLAY,KSPT,KSTEP,KINC)
INCLUDE 'ABA_PARAM.INC'
CHARACTER*80 MATERL
DIMENSION STRESS(NTENS),STATEV(NSTATV),
* DDSDDE(NTENS,NTENS),DDSDDT(NTENS),DRPLDE(NTENS),
* STRAN(NTENS),DSTRAN(NTENS),TIME(2),PREDEF(1),DPRED(1),
* PROPS(NPROPS),COORDS(3),DROT(3,3),
* DFGRD0(3,3),DFGRD1(3,3)
C ELASTIC PROPERTIES
EMOD1=PROPS(1)
67
EMOD2=PROPS(2)
ENU=PROPS(3)
EG=PROPS(4)
THETA=PROPS(5)
KSI=PROPS(6)
ESI=PROPS(7)
ENTA=PROPS(8)
N=PROPS(9)
E0=PROPS(10)
SIG0=PROPS(11)
V=PROPS(12)
M=PROPS(13)
T0=PROPS(14)
EBULK=1/(1.0ENU**2*EMOD2/EMOD1)
C ELASTIC STIFFNESS
IF (KSTEP.EQ.1) THEN
DO K1=1,NTENS
DO K2=1,NTENS
DDSDDE(K2,K1)=0
END DO
ENDDO
68
C
Q11=EMOD1*EBULK
Q12=ENU*EMOD2*EBULK
Q21=ENU*EMOD2*EBULK
Q22=EMOD2*EBULK
Q66=EG
DDSDDE(1,1)=Q11*COS(THETA)**4.0
* +2.0*(Q12+2*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q22*SIN(THETA)**4.0
DDSDDE(1,2)=(Q11+Q224*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q12*(SIN(THETA)**4.0+COS(THETA)**4.0)
DDSDDE(2,1)=DDSDDE(1,2)
DDSDDE(2,2)=Q11*SIN(THETA)**4.0
* +2.0*(Q12+2*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q22*COS(THETA)**4.0
DDSDDE(NTENS,1)=(Q11Q122*Q66)*SIN(THETA)*COS(THETA)**3.0+
* (Q12Q22+2*Q66)*SIN(THETA)**3.0*COS(THETA)
DDSDDE(1,NTENS)=DDSDDE(NTENS,1)
DDSDDE(NTENS,2)=(Q11Q122*Q66)*SIN(THETA)**3.0*COS(THETA)+
* (Q12Q22+2*Q66)*SIN(THETA)*COS(THETA)**3.0
DDSDDE(2,NTENS)=DDSDDE(NTENS,2)
69
DDSDDE(NTENS,NTENS)=(Q11+Q222*Q122*Q66)*SIN(THETA)**2
* *COS(THETA)**2.0+
* Q66*(SIN(THETA)**4+COS(THETA)**4.0)
C1=DDSDDE(1,1)
C2=DDSDDE(1,2)
C3=DDSDDE(1,NTENS)
C4=DDSDDE(2,1)
C5=DDSDDE(2,2)
C6=DDSDDE(2,NTENS)
C7=DDSDDE(NTENS,1)
C8=DDSDDE(NTENS,2)
C9=DDSDDE(NTENS,NTENS)
DO K1=1,NTENS
DO K2=1,NTENS
STRESS(K2)=STRESS(K2)+DDSDDE(K2,K1)*DSTRAN(K1)
END DO
END DO
STATEV(1)=STATEV(1)+DSTRAN(1)
70
STATEV(2)=STATEV(2)+DSTRAN(2)
STATEV(3)=STATEV(3)+DSTRAN(3)
A1=STATEV(1)
A2=STATEV(2)
A3=STATEV(3)
D1=C1
D2=C5
D3=C2
END IF
IF (KSTEP.EQ.2) THEN
S11=STRESS(1)(STRESS(1)+STRESS(2)+STRESS(3))/3.D0
S22=STRESS(2)(STRESS(1)+STRESS(2)+STRESS(3))/3.D0
S33=STRESS(3)(STRESS(1)+STRESS(2)+STRESS(3))/3.D0
S12=STRESS(NTENS)
D11=COS(THETA)**2.D0
D22=SIN(THETA)**2.D0
D12=COS(THETA)*SIN(THETA)
J0=D11*S11+D22*S22+2.D0*D12*S12
J=D11*(S11**2.D0+S12**2.D0)+D22*(S12**2.D0+S22**2.D0)
* +D12*S12*(S11+S22)
71
J2=.5*(S11**2.D0+S22**2.D0+S33**2.D0)+S12**2
Q1=JJ0**2
Q2=J0**2
TEMP1=(ESI4.D0*ENTA)*(STRESS(1)+STRESS(2))**2/9.D0
TEMP2=(KSI)*Q1
TEMP3=(ESIENTA)*Q2
TEMP4=3.0*(J2TEMP2TEMP3+TEMP1)
PHI=SQRT(TEMP4)/(E0)
TA11=2.D0*S11*D11+2.D0*D12*S12
TA22=2.D0*S22*D22+2.D0*D12*S12
TA12=S12*(D11+D22)+D12*(S11+S22)
TB11=2.D0*J0*D11
TB22=2.D0*J0*D22
TB12=2.D0*J0*D12
TONE11=TA11TB11
TONE22=TA22TB22
TONE12=TA12TB12
TTWO11=2.D0*J0*(D111.D0/3.D0)
TTWO22=2.D0*J0*(D221.D0/3.D0)
TTWO12=2.D0*J0*D12
TAO11=S11KSI*TONE11(ESIENTA)*TTWO11+
* 2.D0*(ESI4.D0*ENTA)*(STRESS(1)+STRESS(2))/9.D0
TAO22=S22KSI*TONE22(ESIENTA)*TTWO22+
72
* 2.D0*(ESI4.D0*ENTA)*(STRESS(1)+STRESS(2))/9.D0
TAO12=S12KSI*TONE12(ESIENTA)*TTWO12
RSTRAN11=3.D0*PHI**(N1.D0)*TAO11/(2.D0*E0*SIG0*100)
RSTRAN22=3.D0*PHI**(N1.D0)*TAO22/(2.D0*E0*SIG0*100)
RSTRAN12=3.D0*PHI**(N1.D0)*TAO12/(2.D0*E0*SIG0*100)
DCRSTRAN11=RSTRAN11*DTIME
DCRSTRAN22=RSTRAN22*DTIME
DCRSTRAN12=RSTRAN12*DTIME
STATEV(4)=DCRSTRAN11+STATEV(4)
STATEV(5)=DCRSTRAN22+STATEV(5)
STATEV(6)=DCRSTRAN12+STATEV(6)
c DAMAGE calculation
I=STRESS(1)+STRESS(2)+STRESS(3)
I0=D11*STRESS(1)+2*D12*STRESS(NTENS)+D22*STRESS(2)
RE1=J2+.25*J0**2.0J
IF (RE1.LE.0) THEN
RE1=0RE1
END IF
NN=.5*(II0)+SQRT(RE1)
IF (NN.LE.0) THEN
NN=0.0
73
END IF
IF (Q1.LE.0) THEN
Q1=J0**2J
END IF
S=SQRT(Q1)
DELTA=(NN+0.35*S)/(E0)
STATEV(11)=(1DELTA**V*TIME(2)/T0)**(1/(1+M))
RDASTRAN11=3.0*PHI**(N1.0)*TAO11/(2.0*E0*SIG0*100)
* /(STATEV(11)**N)
RDASTRAN22=3.0*PHI**(N1.0)*TAO22/(2.0*E0*SIG0*100)
* /(STATEV(11)**N)
RDASTRAN12=3.0*PHI**(N1.0)*TAO12/(2.0*E0*SIG0*100)
* /(STATEV(11)**N)
DDAMSTRAN11=RDASTRAN11*DTIME
DDAMSTRAN22=RDASTRAN22*DTIME
DDAMSTRAN12=RDASTRAN12*DTIME
STATEV(12)=DDAMSTRAN11+STATEV(12)
STATEV(13)=DDAMSTRAN22+STATEV(13)
STATEV(14)=DDAMSTRAN12+STATEV(14)
DS1=C1*(DSTRAN(1)DCRSTRAN11)
74
DSTRESS11=C1*(DSTRAN(1)DCRSTRAN11)+C2*(DSTRAN(2)
DCRSTRAN22)+
* C3*(DSTRAN(NTENS)DCRSTRAN12)
DSTRESS22=C4*(DSTRAN(1)DCRSTRAN11)+C5*(DSTRAN(2)
DCRSTRAN22)+
* C6*(DSTRAN(NTENS)DCRSTRAN12)
DSTRESS12=C7*(DSTRAN(1)DCRSTRAN11)+C8*(DSTRAN(2)
DCRSTRAN22)+
* C9*(DSTRAN(NTENS)DCRSTRAN12)
STRESS(1)=DSTRESS11+STRESS(1)
STRESS(2)=DSTRESS22+STRESS(2)
STRESS(NTENS)=DSTRESS12+STRESS(NTENS)
DO K1=1,NTENS
DO K2=1,NTENS
DDSDDE(K2,K1)=0
END DO
ENDDO
Q11=EMOD1*EBULK
Q12=ENU*EMOD2*EBULK
Q21=ENU*EMOD2*EBULK
Q22=EMOD2*EBULK
75
Q66=EG
DDSDDE(1,1)=Q11*COS(THETA)**4.0
* +2.0*(Q12+2*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q22*SIN(THETA)**4.0
DDSDDE(1,2)=(Q11+Q224*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q12*(SIN(THETA)**4.0+COS(THETA)**4.0)
DDSDDE(2,1)=DDSDDE(1,2)
DDSDDE(2,2)=Q11*SIN(THETA)**4.0
* +2.0*(Q12+2*Q66)*SIN(THETA)**2.0*COS(THETA)**2.0
* +Q22*COS(THETA)**4.0
DDSDDE(NTENS,1)=(Q11Q122*Q66)*SIN(THETA)*COS(THETA)**3.0+
* (Q12Q22+2*Q66)*SIN(THETA)**3.0*COS(THETA)
DDSDDE(1,NTENS)=DDSDDE(NTENS,1)
DDSDDE(NTENS,2)=(Q11Q122*Q66)*SIN(THETA)**3.0*COS(THETA)
* (Q12Q22+2*Q66)*SIN(THETA)*COS(THETA)**3.0
DDSDDE(2,NTENS)=DDSDDE(NTENS,2)
DDSDDE(NTENS,NTENS)=(Q11+Q222*Q122*Q66)*SIN(THETA)**2
* *COS(THETA)**2.0+
* Q66*(SIN(THETA)**4+COS(THETA)**4.0)
RETURN
END
76
APPENDIX B
*Heading
*Node
1, 0., 20., 0.
……
*Element, type=S8R,ELSET=CREEP
……
*Nset, nset=B1
creep, SNEG
*shell SECTION,ELSET=CREEP,MATERIAL=com1
0.4
77
2230, 2230, 0
** MATERIALS1
*MATERIAL,NAME=COM1
*USER MATERIAL,CONSTANTS=16,TYPE=MECHANICAL
*DEPVAR
14
*INITIAL CONDITIONS,TYPE=SOLUTION
CREEP,0.0,0.0,0.0,0.0,0.0,0.0,0.0
0.0,0.0,0.0,0.0,0.0,0.0,0.0
*Boundary
B1, 2, 2
*STEP,INC=30
*STATIC
1.E7,1.E6
*Dsload
vessel, P, 0.5
*EL PRINT,FREQ=1
S,
SDV1,
SDV4
78
*END STEP
*STEP,INC=100000
CREEP STEP
*VISCO,CETOL=1E4
1.E6,10
** OUTPUT REQUESTS
*PRINT,FREQ=1,RESIDUAL=YES
*EL PRINT,FREQ=1
SDV
*OUTPUT,FIELD,FREQ =1
*ELEMENT OUTPUT
S,E,CE,
SDV
*NODE OUTPUT
*END STEP
79