NASA

Technical

Memorandum

106526

A Coupled/Uncoupled
Deformation
and Fatigue
Damage Algorithm Utilizing the Finite
Element Method

Thomas

E. Wilt

University of Toledo
Toledo, Ohio
and
Steven
Lewis

M. Arnold
Research

Cleveland,

March

Center

Ohio

1994

N94-36561

(NASA-TM-106525)
A
COUPLEO/UNCOUPLED
DEFORMATION
AND
FATIGUE
DAMAGE
ALGORITHM
UTILIZING
THE
FINITE
ELEMENT
METHOD
(NASA.
Lewis

Research

Center)

26

Unclas

p
G3/39

National Aeronautics and

Space Administration

001053a

A Coupled/Uncoupled

Deformation

Utilizing

and Fatigue

The Finite

Element

Damage

Algorithm

Method

T. E. Wilt*
University
Toledo,

of Toledo
Ohio 43606

S. M. Amold
NASA Lewis Research
Cleveland

Center

Ohio

Abstract
A fatigue

damage

computational

based fatigue damage

model
mented
into the commercial
Damage

is introduced

into

algorithm

utilizing

a multiaxial,

isothermal,

for unidirectional
metal matrix composites
finite element
code MARC using MARC
the finite

element

solution

through

the

continuum

has been impleuser subroutines.

concept

of effective

stress which fully couples

the fatigue damage
calculations
with the finite element
deformation solution.
An axisymmetric
stress analysis was performed
on a circumferentially
reinforced
behave

ring, wherein
elastic-perfectly

anisotropic
resented

continuum

1.0

based

by an isotropic

and damage

distribution

both the matrix

plastic.

cladding

The composite

plasticity

plasticity

model,

model.

and the composite

core behavior
and similarly,

Results

core were

was represented
the matrix

are presented

assumed
using

cladding

to

Hill's

was rep-

in the form of S-N curves

plots.

Introduction
In advanced

durability

engine

in addition

designs,

materials

to decreased

which

weight

allow

higher

are desirable.

ites (MMCs)
may provide
these benefits. For example,
(TMCs) rotors are projected
to have significant
benefits
and lower
ever,

weight,

as compared

to fully realize

life prediction
Analysis
histories

methods
of typical

requires

the benefits

to the nickel
offered

speeds

and longer

matrix

compos-

titanium
metal matrix composite
in terms of increased
rotor speeds

and titanium

by MMCs,

operating

The use of metal

rotors

currently

in service.

computationally

efficient

to complex

thermomechanical

How-

design

and

must be developed.
aerospace

structures

the use of computational

subjected
approaches

*Resident Research Associate at Lewis Research Center

such as the finite

element

load
method.

In this regard,

it is to desirable

to develope

a life prediction

algorithm

that can be used in

conjunction
with the finite element method. Historically,
two basic approaches
have been
used in predicting
the life of structures;
uncoupled
or fully-coupled
deformation-damage
methods.
A typical

analysis

a finite

element

model,

the number

analysis,

and then,

of cycles

"local"
fracture mechanics
new finite element mesh
crack

maybe

controlled
scheme.

modeled
by

and damage

method.

fatigue

damage

energy

the su'ess

initiation

rate

properties,

uses

taken

MARC,

is coupled

In Section

in the context

2, the requisite

computationally-coupled
of a reinforced
MMC
results

2.0

The

continuum
posites

in terms

Damage

fatigue

damage

study.

Specifically,

element

release

deformation

code

the com-

in which

the use of provided

through

user

the

subrou-

the degradation

of the material

method.

equations

of the evolution

a node

the effects of damage are accounted

for
concept allows the current damage state

element

damage

of the crack is then

with

is a fully-coupled

finite

damage

Subsequently,

are presented.

of damage

In Section

3 the

Finally, an example
will be presented
and

in the ring cross

section.

Formulation

damage

[1]. The

method

through

response

of the finite
fatigue

The propagation

in the present

stress,
stress

to a fatigue

from

the crack. This requires

a
tip zone. For example,
the

fatigue damage
algorithm
will be outlined.
ring, representing
a typical engine component,

will be presented

Fatigue

as input

are predicted.

in conjunction

a nonlinear

to MARC

into the deformation

which,

nodes.
criteria

to the uncoupled

tines. By utilizing the concept of effective

in the finite element
solution. The effective
to be incorporated

state data

of a crack

of double

release

This is the approach

developed

algorithm

the stress state for each element

approach is then used to propagate

be constructed
to model the crack

the alternative

scheme

of obtaining

using

to the

using a series

a strain

As mentioned,
putational

consists

calculations

mechanics
model

utilize

model

a recently

for the fatigue

is phenomenological,

stress

developed

multiaxial,

of unidirectional
based,

and

isothermal,

metal

assumes

matrix

a single

comscalar

internal
damage variable,
the evolution
of which is anisotropic.
The present
multiaxial,
isothermal,
continuum
damage model for unidirectional
metal matrix composites
may be
expressed

as, [1]

dD

Dr_ 1

j" [1-(1-D)

1_+1] %

is a function

(EQ 1)

where D k and D k_ 1 is the amount of damage

respectively,
and AN k is the number of cycles
(which

aN

of the current

stress

at the current
at the current

state) is defined

as,

and previous
increments,
stress state (o k) , and a k

<%>
(xk = 1-a(--_--

(.)

where

( ) are the Macauley

the current

and previous

brackets.

increments,

Note,

(EQ 2)

the subscript/superscript

respectively.

The fatigue

k and

limit surface,

k - 1 denote
Off, is defined

as,

1 max

The static fracture

surface,

t_ u, is defined

_u
Lastly,

the normalized

max Fifo

stress

= 1-

(o/_ (t) - o/_(t0))

as,

max
t

amplitude,

(u)

(o_:(t))

F,n, is defined

1 max
to rn_F(,n)(o-ij(t)
In the above
some

time

expressed

equations,
during

load

cycle.

The

general

as,

12' and 13 are physically

of the current

form

for

invariants

1 (didjSij)
"4

12= a ais ksk - (aAs p 2

13 =

are a function

d i denoting

F(:0,

4co_ ) - 1
9
}
rl_ )
12 + 313

meaningfull

1
ll = _SijSi
j _didiSjkS_+

which

(EQ 5)

load

cycle,

(u), or (,7,)

and

t, is

may

be

as,

) - 1) I l +
in which/1,

(EQ4)

--Ok(t0))

to is the time at the beginning

the

(EQ 3)

- 1

of the current

the materials'

(didjSij)

deviatoric

(EQ 6)

[1], i.e.,

2
(EQ 7)

stress

state,S/_

= o.k.
1
q - "_Offll
i/' and a vector

fiber orientation.

Note, that in the expression

for o_k, (O,,) = 0 indicates
static fracture
and, as will be
discussed
in the following
section, the finite element
is considered
to have failed completely. On the other hand, whenever
(Off) = 0 indicates
that the current
stress state is
below the fatigue limit and thus ct t is set equal to 1. This presents
a special case when
integrating
the fatigue damage expression,
EQ. 1, and will be considered
later in this section.

First,
Thus,

consider

a current

integrating

EQ

state of stress,

o k, which

is above

the fatigue

limit, i.e.

at #: 1.

1.,

[I-

(I-D)13+I]

l-%Dk

^13

(EQ 8)

f) Io,_,= FmANk
which

results

in an expression

for the number

( [I - (I -Dr) I_+I]I-%

of cycles,

AN k, at the current

_ [I - (I-Dk_

stress,

o k, i.e.,

i) 13+I] 1-ctk )
(EQ 9)

AN k
P_m(l -0c k) (13+ 1)
Note

that D k_ 1 is the total amount

is the total amount

To calculate

of damage

the cycles

of damage

at the beginning

of the load block

and D k

at the end of this load block.

to failure,

let D k = 1, which

(1-

results

[1-(1-Dk_l)l_+l]

in the following,

1-%)

(EQ 10)

aNp,
In the present

P_( 1- %) ( 13+ 1)

computational

and D k_ 1 are known.

scheme,

That

since

the damage

increment

is controlled,

both D k

is,
(EQll)

D k = Dk_I+AD
where

AD

increment

is the user

specified

in the number

As will be shown
the damage

increment

of cycles

in the following

in damage.

for each element,

section,

Thus

EQ.

9 is used

to predict

AN_, due to the increment

it is also necessary

to re-write

EQ. 9 in terms

I-%

1-

Now

of

D k, i.e.,
1
[_+1

D k =

the

in damage.

/1-

consider

ct k = 1. Thus,

{ [1 -

(1-Dk_l)

the case
EQ.

_._mANk} 1-ct,_

13+1]

in which

+ (1-o_

the current

stress

jO_Q12)

k) (13+ 1)

state is below

the fatigue

limit,

i.e.

1 takes the form,

(EQ 13)

Dr_ 1 1

Upon
Dk-

integrating

the above

1, may be expressed

equation,

J FmdN

+I

the increment

as,

in cycles,

AN k, with initial

damage,

(EQ 14)

AN k =

For the number

log [1-

of cycles

(1-Dk)

13+1] -log[l-

to failure,

ANFd

-log

[1 -

(1-D__1)_+1]

1)

let D k =

"_

1, i.e.,

(1 -Dk_

1) I_+1]
(EQ 15)

NF' =
Alternatively,

the following

1)

expression

for the damage

D k, may

be expressed

as,

p+l
D k = 1The

effect

{1-

[1-

of damage

(1-Dk_l)[_+llexp((]3+

is included

in the finite

(EQ16)

1)P_AN

k) }

stress

analysis

element

by utilizing

the

concept
of effective
stress [2]. Based on the hypothesis
of strain-equivalence
[3,4], the
effect of damage
may be accounted
for by simply degrading
the elastic and plastic material properties.

The degraded

elastic
[C]

and similarly,

the plastic

material

Comnutational

using

implementation

a few select

has been

coupled

The MARC

by,

k) [C]

(EQ 17)

e.g. yield

(1-Dk)

stress

t)y, are degraded,

(EQ 18)

t_y

of the fatigue damage calculations

were developed
in the
code MARC.
MARC provides
various
user subroutines
[5]
of constitutive

MARC

user

user

subroutines

HOOKLW

properties,

respectively.

ment

global

required

element
were,

solution

ELEVAR,

has been

criteria,

new

ELEVAR
attained,

is called

elements,

fatigue

the elastic

of "applied"

load cycle

will be discussed

at the end

and is intended

later

of each

to be used

all integration

points

subsequent
damage calculations
written in a sufficiently
general

are averaged

The

material

load

to output

increele-

is used to store
load cycle. The

in this section.

The present version

of the fatigue damage
algorithm
utilizes _
damage calculations.
For example,
the stresses
for each integration
and then

model

and ANPLAS.
and plastic

ment quantifies
at the end of a given increment.
In this algorithm
ELEVAR
the current converged
stress state for each element
during the "applied"
meaning

etc. By

damage

scheme.
HOOKLW,

are used to degrade

subroutine

convergence

failure

the continuum-based

finite

and ANPLAS
The

models,

subroutines,

with the nonlinear

subroutines
once

is calculated

Scheme

The present implementation

context
of the finite element
that allow

matrix

(1 -D

properties,
fly

3,0

constitutive

to give one stress

point

quantities
in the
are determined

state for each

element.

All

use these average quantifies.

However,
the program
was
form so that all of the damage
calculations
may be per-

formed at each integration point

required

is increased

A flowchart

dimensions

with

minimal

for various

of the developed

modifications.

storage

life prediction

Specifically,

all that

is

arrays.

scheme

is shown

in Figure

1. The deforma-

tion analysis is the actual finite element run. The fatigue and failure calculations
are contained in the MARC subroutine
ELEVAR.
First, note there are two levels of failure criteria
checks;

element

level and structural

level.

The element

face check which is part of the fatigue damage

included
such as a check on total mechanical
these

failure

criteria,

that element

equal

to the maximum

amount

is considered

of damage

level includes

a static

model, (EQ. 4). Additional

strains, etc. If an element
to have

allowed.

"failed"

As shown

fracture

criteria
violates

and its damage,

sur-

may be
one of
D, is set

in Fig. 1, for a coupled

anal-

ysis the damage calculations

are terminated
and a deformation
analysis is re-run in order
to account for the stress redistribution
due to that element's
failure. For an uncoupled
analysis,

when

element

deformation
calculations.

failure

occurs

the analysis

analysis,
instead,
it continues
The structural
level criteria

This could

take

the form

of a check

does

not loop

back

to the next element

monitors
the global

upon

selected

nodal

and

perform

another

and performs
the damage
response
of the structure.

displacements

which

if they vio-

late a specified
displacement
criteria the structure
is considered
to have failed. For exampie, the tip displacement
of a turbine
blade may be required
to stay within
a given
tolerance.

Again,

performed

since

jected

note

that for an uncoupled

the present

to the initial

"applied"

the algorithm

may include

the beginning

of the analysis.

The

above

fatigue

mentioned

load cycle
the option

"applied"

data file, and are used to achieve

rence

of damage,

"applied"

cycle

internally

monitor

damage

i.e. material
must

calculation

phase.

The

initial

the stress

rediswibution

state.
states

in the structure
the number

number

load

cycle

has been

of load increments

used

depends
on the nonlinearity
of the smactural response
and requires
of the user. Note, since the algorithm,
in its present form, requires
number

of increments

AUTOLOAD
The

subroutine

and stores
increment
from

cannot

ELEVAR

automatic

load

ELEVAR,

DAMAGE

is

sub-

In the future,

the MARC

at

input

due to the occur-

of load increments

per

so that the program

can

completed

and begin

in the load cycle

the

usually

experience
on the part
the user to specify the

incrementing/stepping,

i.e.

MARC's

be used.
is called

at each

increment

the average
stress (strain is optional)
of the current
"applied"
load cycle,

within

When
m_,

option,

in a cycle,

check

when

for each element

through

by the user. This is necessary

"applied"

criteria

the structure

_
damage

are user defined

In addition,

failure

assumes

is in a completely

to include

degradation.

a given

no structural

algorithm

load cycles

be specified

when

analysis

damage

see Figure
is entered

during

the "applied"

load

cycle

state for each element.

At the last load
the subroutine
DAMAGE
is now called

2
for each element,

various

element

quantities,

such

as, at,,

Ou' F,n are calculated

and stored. When DAMAGE
has been called for the last elein the mesh, the fatigue damage calculations
are performed.
Figure 2 shows the gen-

eral algorithm
for the fatigue damage
FORTRAN
source code for the fatigue

calculations.
In addition,
Appendix
I contains
the
damage algorithm.
The code is presented
in a form

that allows the fatigue damage

or any other

deformation

Presently,
user

the damage

specifies

calculations

analysis
calculations

the allowable

to be used

as a subroutine

in a finite

are controlled

increment

by the increment

in damage,

for example

in damage,
AD

subroutine,

minimum

number

SORTN,
of cycles

is called

to failure

and is chosen

NF,,," = minANeF
Once

the controlling

amount
formed

number

Figure

by a specified
3 shows

is the actual
quent
which

which

as the "controlling

has been

In

and the given eleand stored. Next a

element

has the

element",

i.e.,
(EQ 19)

e = 1 --->numel

of cycles

the cycle

cycles, shown
is determined

the next "applied"

redistribution
due

determined,

the corresponding,

actual

load cycle,

to merely

indicate

in dashed
in SORTN.

files.

damage

number

Some

These

of fatigue

attains

experience
reduced

load

element

The subse-

analysis.

cycle

load

cycle,

was also written

damage

would

NFmin

the element

cycles

and remaining
evolution

comments

on the fatigue

the user

specified

considered

material

with difficulty

PATRAN

at specified

increments

cycles

damage

value

be used

properties

convergence
of

need

damage,

damage

of 95% was

global

for each

[6] element
during

table

element,

the

of the curand a sum-

are also generated.

algorithm

allowable

investigation

a summary

to failure

in each element

in any subsequent

in achieving

Further

generates

files containing

maximum

a cutoff

which

distributions

output

the damage

5%.

in Fig. 3. The resulting

cycles is repeated
sequence,
i.e. one

17 and 18.

In addition,

in the next section,

below

that the "applied"

initiation.

"applied"

PATSTR,

files contain

to fail and is no longer

presented

as shown

using EQS.

analysis.

general

known.

to the newly calculated

element
damage D_. This is done through
HOOKLW,
for the elastic constants,
and ANPLAS,
for the aniso-

ratios

table showing

an element

Recall

is already

are the predicted

cycles corresponding
to NFmin
it is assumed
that the stress state in each element

of damage

for the subsequent

a subroutine,

fatigue

in the code.

and used in the finite

lines,
Here

is performed

the location

yield stress

results

used

damage

the predicted
NFmlncycles
and at the end of NFrai n, each element
of damage as calculated
in CALCD.
Note for a coupled
analysis,

are degraded
according
the MARC subroutines

Finally,

scheme

that is applied

element's

This sequence
of"applied"
load cycle and predicted
has failed. For an uncoupled
analysis,
only one

"applied"

In preparation

the controlling

load cycle is run in the finite element analysis to account for the stress
to the new damage
state in each element
(i.e. D_) and again a new

NFmin is predicted.
until the structure

tropic

amount,

load history

remains
constant during
has incurred
an amount

mary

(15%).

of damage,
D_, in all of the remaining
elements must be re-calculated.
This is perin subroutine
CALCD using EQ. 10 or EQ. 13. Note that since the damage was

incremented

rent

(see Fig. 2) to determine

AD. The

= 0.15,

CALCN,
using EQ. 9 or EQ. 12, based upon the new value of damage
ment's stress state, the number of cycles to failure, A/_ F, is calculated
"sorting"

element,

program.

First,

the element

is assumed

calculations.

specified
when

convergence

to be made.

based

In the example
upon

the element
difficulties

once

preliminary
stiffness
needs

was
to

be

addressed.

Second,

plasticity_ model
ening,

in the present

and the corresponding

specific

block of fatigue

tion/update

coupled

was used. This idealization

update

cycles.

of the internal

damage-deformation

eliminates

of the internal

variables,

In order to accurately

account

variables

analysis,

the need to account

which

for cyclic

may occur

during a

for the hardening,

through the load block would

be required.

hard-

a projecIn addition,

it is usually assumed that the fatigue damage calculations

are applied to a "stabilized"
stress redistribution.
Thus, when hardening
is present, more than one "applied"
load cycle
may be necessary

4,0

Examnle

in order to achieve

Aoolication:

As stated previously,
computationally

the stabilized

A Cladded

MMC

one of the primary

efficient

method

redistribution.

Ring Insert

motivations

for predicting

of this research

the fatigue

is to establish

aerospace

com-

life of typical

ponents.
This includes the ability to predict the location(s)
of damage initiation
and to be
able track the propagation
of damage throughout
the smacture. With this in mind, the
fatigue

damage

this specific
consideration

algorithm

was applied

to a cladded

MMC

structure
are two-fold.
First, it represents
in advanced
engine designs. Secondly,

try and load conditions,

maximum
circumferential
the cladding,

ring.

The reasons

for choosing

a MMC rotor insert currently

under
because of its axisymmetric
geome-

qualitative
stress distributions
are known a priori. For example,
stress in the core occurs at its inner diameter
and likewise for

thus providing

some

intuitive

feel for where

damage

initiation

will occur

as

well as how it may propagate.

The composite
core was described
by Hill's
anisotropic
model available
in MARC [5], while the matrix cladding was
elastic

perfectly-plastic.

deformation
fatigue

The

analysis

damage

elastic

are given

model

and

in Table

are given

inelastic

1, while

in Table

elastic-plastic
constitutive
assumed to be isotropic
and

material

parameters

required

material

parameters

for the

utilizes

the iso-

the associated

2. Note that the matrix

cladding

for

the

tropic form of the fatigue damage

model, i.e. co= = 0a = e0= = rl= = rl/z = rim = 1,
whereas
the composite
core is represented
by the transversely
isotropic
form of the model.
The finite element
model, representing
225 nodes and 64 8-node axisymmetric
pressure

load was applied

With regards
ously

made

observed

the inner

to the deformation

and compared

thus providing

Two types

along

of fatigue

the cross-section
of the ring, Figure 4, consisted
of
elements
(MARC element number 28). A uniform
diameter

analysis,

to limited

burst

experimental

a level of confidence
life analyses

of the ring.
pressure
data

in the finite

were performed,

predictions

[7,8].

Very

element

namely,

have

good

been

correlation

modeling

an uncoupled

previwas

of the ring.
and a coupled

analysis.
The uncoupled
life prediction
each of the four elements
in the radial

results were obtained

by taking the stress state in
direction
of the composite
core of the ring, Figure

4b. Here

state was relatively

it was assumed

one element
uncoupled

would
analysis,

all of the elements

no fatigue

with the matrix

cladding.

causing

fatigue

infinite

that the stress

represent

calculations

were

This is due to the initially

lives

to be calculated.

constant

in a column
performed
low stress

core.

on the elements
levels

As will be shown,

in the z-direction,

of the composite

in the matrix

it is only

thus
In the

associated
cladding

in the coupled

analysisthatfinite

lives and damage

are predicted

for the cladding,

due to stress redistribu-

tion effects.
Figure 5 shows the results of the uncout_led fatigue damage analysis.
As expected,
element 1 has the shortest fatigue life, thus, damage is predicted
to initiate along the composite core inner
analysis
burst

diameter.

results.

fatigue

the fatigue

Figs.

as predicted
analysis.

damage

Finally,

coupled

deformation

5 and 6, one observes

life predicted

by the coupled

by the coupled

This

difference

damage in the ring cross-section.

effects which are automatically
fatigue

the _

and fatigue

that- at pressures
analysis

damage

close

is close

to the

to that of the

analysis,
since at high stress levels, once the damage
initiates
in the core,
failure of the ring occurred
rapidly. On the other hand, at low stress levels, the

life

uncoupled

6 shows

By comparing

pressure,

uncoupled
"structural"

Figure

analysis

may

is longer

be viewed

than

that predicted

as the effect

from

of propagation

This propagation
is caused by the stress redistribution
captured
by performing
a fully coupled
deformation
and

analysis.

Fig.

7 shows

two

selected

damage

distribution

plots

in the ring

cross-section

produced
by the coupled
fatigue damage analysis.
Note that in Fig. 7a, the damage
tiates along the inner diameter
of the composite
core. Conversely,
in Fig. 7b structural
ure of the ring is depicted
redistribution,

the

of the

(i.e. the composite

the matrix

cladding

core has completely

has accumulated

significant

failed)
amounts

inifail-

and due to stress

of damage.

&0_Smnlna
A coupled/uncoupled
The algorithm

utilizes

deformation

and fatigue

a multiaxial,

isothermal,

damage

algorithm

stress-based,

has

transversely

been

presented.

isotropic

contin-

uum fatigue damage model in which the fatigue damage calculations

are coupled
with the
nonlinear
finite element
solution using the concept of effective
stress. Incorporated
in the
life prediction

scheme

The

has been

space
age

algorithm
component
distribution

presented
However,

are failure

criteria

applied

to a cladded

and results

have been

plots

the

over

ring

at both the element

MMC

presented

ring insert
in terms

cross-section.

All

and structural

representing

of S-N curves
of the

a typical
along

fatigue

MMC rings under contract

become available,
a similar

the present

fatigue

damage

Acknowledgment
The first author would
NCC3-248,

algorithm

with Textron Lycoming,

finite element
analysis
and continuum

like to acknowledge

at NASA

Lewis

Research

fatigue

aeroresults

available.
on simi-

NAS3-27027.
Once these
will be conducted
to verify

damage

model.

support for this work under

Center.

level.

with dam-

damage

are qualitative
in nature since no experimental
results
are currently
full scale burst pressure and fatigue tests are currently
being performed

lar cladded
test results

grant,

checks

a cooperative

6.0 References
1. Arnold,

S. M., and Kruch,

Creep and Fatigue

2. ].,emaitre,
Press,

Matrix

J. L., Mechanics

Damage

Composites,

J. L.: Continuum

J. L.: Continuum

and Crack

Growth,

Revision

Palo Alto,

6. PATRAN
7. Arnold,

NASA

of Solid Materials,

Models

for

TM-105213,

Cambridge

1991.

University

Damage

Mechanics:

Part I - General

Concepts,

Damage

Mechanics:

Part II - Damage

Growth,

User

J. Appl.

K.5, Volume

Mech.,

J. Appl.

Initi-

D: User

Subroutines,

MARC

Analysis

Research

Corpo-

CA.
Manual

S. M., and Wilt, T. E.: A Deformation

8. Wilt, T. E., and Arnold,

Element

Crack

Vol. 55, 1988, pp. 65-72.

and Life

tially Reinforced
SiCfH 15-3 Ring, DE-Vol. 55, Reliablity,
Prevention,
Ed. R. J. Schaller, 1993, pp. 231-238.

Finite
15

Mechanics

Vol. 55, 1988, pp. 59-64.

5. MARC,
ration,

Metal

Continuum

1990.

4. Chaboche,
ation

of Unidirectional

J. and Chaboche,

3. Chaboche,
Mech.,

S.: Differential

Methodology,

S. M.: A Computationally-Coupled
HITEMP

Review

10

Prediction
Stress

OfA

Circumferen-

Analysis

and Failure

Deformation

1993, Vol. II, NASA

and Damage

CP19117,

pp. 35:1-

TABLE

1. Material

Elastic:

Properties

For Deformation
(MPa)
SiC/Ti 15-3 Composite
Material

(1 denotes

fiber

E2=E3

= 114457.

Inelastic:
o = 276.

Yl

5.

Oy

Matrix

v12 = 0.28

Y2 -

Oy

Ref. [7]

direction)

E 1 = 183959.

Model,

Material

y3

V13

V23

= 0.32

1.

Oy

(Ti 15-3)

Elastic:
E = 74466.

v = 0.32

Inelastic:
c

= 514.
Y

TABLE 2. Material

Properties For Fatigue Model,

(MPa)
SiC/Ti 15-3 Composite
Material

= 10694.

co

Ref. [7]

= 5.5

off = 1972.

o)17 = 12.482

13 = 1.842

com = 11.8

a =0.012

11,, = rift = rim = 1.0

M=

22371.
Matrix

Material

(Ti 15-3) - Isotropic

= 6081.

Simplification
co

= 1.0
U

o/7 = 965.

_/'t = 1.0

13 = 2.27

tom = 1.0

a =0.0365

riu = _fl

M=

6205.

11

= rim =

1.0

_1

Deformation

] Structural

Level

Analysis

Failure

Criteria

I_

X____

Check

no Sai/.re
Ll_, t F,Aglllml/,.]._A

Failure

Criteria

Check

I.....fai/.ure

no failure

[ Fatigue

Damage

................................................

Calculations

,',,_%%

failure

no failure

Life Prediction
where

Completed

of arrow

.....................

Coupled

_11_.-

Uncoupled

._"_ """"'_""'_

Box indicates
K

Figure

Analysis

Path

Analysis

basic modules

Path

of life prediction

algorithm

1: Coupled/Uncoupled

Life Prediction

12

Scheme

SUBROUTINE ELEVAR
Enter

Structural Level Failure Criteria ]

I Element

Level

F lilure

Criteria

_]

(calculate cycles to failure

[--cALcN
b_ed
I current damage)
,,,,

DA AO
i
T

II1

fatigue

damage

.......

....

SORTN

(find minimum

(control

cycles

to failure

CALCD
actual amount

based on Nf min)

Exit
I

, _ Boxes containcode found in Appendix I

Figure

2: ELEVAR

based

on Nf min)

(calculate

i'

on

Subroutine

13

Calculations

of damage

2 nd "applied"
load cycle

1st "applied"
load cycle

cycles

A
continue
to failure

/ \/......'/\
'\

,/

nncnuplr.d.analy_i_
to failure
ennnl_'_

Figure

3: Cycle

analv_i._

Scheme

For Uncoupled

14

and Coupled

Analysis

84.78

= I

81.56
7n 7

='_1
r
I

67.99

3.88

I
20.77

13.716

1.60

J_
T

r
f

3.175

4.22

Center
Line
a) Actual Ring Geometry
(dimensions are in ram.)
Z

ID

1 234

OD

r
r

b) Idealized

Ring Geometry

Figure 4: Cladded MMC Ring Geometry

15

and Finite Element Model

1.00

--

El.#1

0.90
0.80

-.......
-

0.70
e',l

El.#4

El.#3

0.6O
ogo
"-.X_,,,
030

_"

E1 .#2

0.20
0.10
0.00

.0

, ,, ....J ....... J ...... .J ....... J ........ I ....... J ........ I ...... .J ...... .I

lOt
102
10 3
10'
10 _
106
10_
10'
10_

Cycles To Failure (N F)

Figure 5: Life Prediction For Uncoupled Analysis

16

1.00

........
0-0010o

I ........ I ........ I ........ I ........ I ........ I

10 2
lOs
10 _
lOs
106
107

Cycles To

Failure

Figure 6: Life Prediction

(NF)

For Coupled

17

Analysis

0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1.0
a) N/NF

= 0.2

b) N/NF

Cycles to failure,

Figure 7: Fatigue

Damage

N F = 155

Distribution

18

= 1.0

in Ring Cross

Section

APPENDIX
Fatigue

Damage

C_" ,_,_" "__"P_ "_" _ _ ??_"_?_ _"/__ "_?__" _;'.._..#_#######

SUBROUTINE

PURPOSE:
INPUT:

-#_--_- -_.._###################

PERFORM

FATIGUE

DAMAGE

OUTPUT:

IMPLICIT

REAL*8

(A-H,O-Z)

LOGICAL

IDAM,FALT

DIMENSION

AMP(6),WAA2(126)

....................

CONSTANTS

...................

C
ANGDEG = 0.0
ANGRAD = (3.141592654/180.)*ANGDEG
OMU = 1.
OMFL= 1.
OMM= 1.
ETAU = 1.
ETAFL = 1.
ETAM = 1.
BETA = 2.27
DENM = 1.
A = 0.2302
SIGFL = 20.3
XML = 900.
SIGU -- 128.
C
C LOOP OVER
C TOTAL NUMBER

OF INCREMENTS,

ITOT

C
CALL

CALCULATIONS

CURRENT
NUMBER OF CYCLES TO FAILURE,
BASED UPON DAMAGE INCREMENT,
DELD

C MATERIAL

WAA2,

STRESS HISTORY FOR EACH ELEMENT (PASSED

THROUGH
ARRAY WAA2)
DAMAGE INCREMENT,
DELD
NUMBER OF INCREMENTS
IN "APPLIED"
LOAD CYCLE,
PREVIOUS
AMOUNT OF DAMAGE, DO

Subroutine

DAM AGE(FUMAX,FFLMAX
,FMMAX,XNF,
D0,DELD,DDN,rIDT)

&
C
C
C
C
C
C
C
C
C
C
C
C
C
C

I:

ZEROR(AMP,6)

19

DDN,

ITOT

DO 50 J= 1,trOT
DO 55 K=l,6
INDX = ((J-1)*6)+K

aMP(K) = W_(n_X)
55
C
C

CONTINUE

................

C CALCULATE F_U
C

..............

C
CALL FFUNC(AMP, ANGR AD,OMU,ETAU,S IGU,FU)
C
IF(FU.GT_JMAX')
THEN
FUMAX = FU
FFU = 1.0 - FUMAX
ENDIF
CONTINUE

50
C
C LOOP OVER
C TOTAL NUMBER OF INCREMENTS, ITOT
C
DO 100 I=l,rroT-1
JS=I+l
DO 110 J=JS3TOT

CALL ZEROR(AME6)
DO 120 K=1,6
INDX1 = ((I-1)*6)+K
INDX2 = ((J-1)*6)+K

AMP(K) = WAA20NDX _)- WAA2gNDX2)

120
C
C

CONTINUE

................

C CALCULATE F FL
C

................

C
CALL FFUNC(AMPj_GR

AD,OMFL _I'AFL,S IGFL,FFL)

C
IF(FFL.GT.FFLMAX) THEN
FFLMAX = FFL
FFFL = 0.5*FFLMAX - 1.0
ENDIF
C
C

...............

C CALCULATE F_M
C

...............

C
CALL FFUNC(AMP, ANGRAD,OMM JETAM,XML,FM)
C

n::ff'M.GT.FMMhX)
FMMAX = FM
FFM = 0.5*FMMAX
ENDIF
C
110

CONTINUE

20

I00
C
C

CONTINUE

..............

C < > FUNCTION

C

..............

C
IF(FFFL.LT.0.0) THEN
F"F_LH = 0.0
ELSEIF(FF_.GE.0.0)
THEN
FFFLH = FFFL
ENDIF
IF(FFU.LT.0.0) THEN
FFUH = 0.0
ELSEIF(FFU.GE.0.0) THEN
FFUH = FFU
ENDIF
C
C

.................

C CALCULATE ALPHA
C

.................

C
NFAIL = 0
IF(FFUH.NE.0.00) THEN
ALPHA = A*(FFFLH/FFUH)
C

.......................

C IF FFUH=0 FAILURE CASE

C

........................

ELSE
NFAIL = 1
XNF = -999.
ENDIF
C
C

.....................................

C IDAM=.FALSE. IF NO PRIOR DAMAGE EXISTS

C IDAM=.TRUE. IF PRIOR DAMAGE EXISTS
C

......................................

C
IDAM = .FALSE.
IF(D0.NE.0.0) THEN
IDAM = .TRUE.
ENDIF
FALT = .FALSE.
IF(IDAM .AND. ALPHA.EQ.0.) THEN
FALT = .TRUE.
ELSEIF(.NOT.IDAM
XNF= 1.E9
ENDIF

.AND. ALPHA.EQ.0.) THEN

C
DSTAR = DO + DELD
IF(DSTAR.GT.0.95) DSTAR = 0.95
C
IF(XNF .NE.-999.) THEN
CALL C ALCN(BETA,DS TAR ,D0,ALPH A,FFM ,FALT,XNF, DDN)
ENDIF

21

C
C
STOP
END
SUBROUTINE
C
C
C
C
C

FFUNC(STRS.ANG,OMEGA .ETA,DENM,F)

PURPOSE: CALCULATE; F_U F FL F_M

CALLED FROM: DAMAGE
IMPLICIT REAL*8 (A-H,O-Z)

C
DIMENSION STRS(6)
C
C

......................

C CALCULATE INVARIANTS
C

......................

C
CALL INVAR(STRS,ANG,XI l ,XI2,XI3)
C
A = 1.0/(DENM*DENM)
B = 4.0*OMEGA*OMEGA- 1.0
C = B/(ETA*ETA)
C
F = DSQRT(A*(B*XI I+C*XI2+(9J4.)*XI3))
C
RETURN
END
C###_,_'_?"
C
C
C
C
C

? ?_';';//4;#############################################
SUBROUTINE INVAR(S IG,ANG ,XI 1,XI2,XI3)
PURPOSE: CALCULATE INVARIANTS; I 1 I 2 1 3
CALLED FROM: FFUNC
IMPLICIT REAL*8 (A-H,O-Z)

C
DIMENSION SIG(6)
C
DI = DCOS(ANG)
D2 = DSIN(ANG)
D3 = 0.0
Dll = Dl*Dl
D22 = D2*D2
D33 = D3*D3
D12 = DI*D2
D23 = D2*D3
Dl3 = DI*D3
C
PRESS = (SIG(1)+SIG(2)+SIG(3))/3.
Sll = SIG(1) - PRESS
$22 = SIG(2) - PRESS
$33 = SIG(3) - PRESS

22

S12 = SIG(4)
$23 = SIG(5)
S13 = SIG(6)
C
C

...............

C [S_ij]* [S_ij]
C

...............

C
&

XJ2 = 0.5*(Sll*Sl
I + $22"$22
+2.*$23"$23+
2.*S13"S13)

+ $33"$33

+ 2.*S12"S12

C
C

................

C [D_ij]* [S_ij]
C

.................

C
&

XI = DII*Sll
+ D22"$22
D23*S23+D13*D13)

+ D33"$33

+ 2.*(D12"S12

C
C

......................

C [D_ij]* [S_j_]* [S_ki]

C

.....................

C
&
&
&
&
&

XIH = DII*(SII*S11+S12"S12+S13"S13)
D22"(S12"S12+$22"$22+$23"$23)
D33"(S13"S13+$23"$23+$33"$33)
2.*D12"(Sl1"S12+$22"S12+S13"$23)
2.*D13"(Sl1"S13+S12"$23+$33"S13)
2.*D23"(S 12"S 13+$22"$23+$23"$33)

+
+
+
+
+

C
XI3 = XI**2
XI1 = XJ2 - XIH + 0.25"XI3
XI2 = XIH - XI3
C
RETURN
END
SUBROUTINE
C
C
C

ZEROR(AIDIM)

PURPOSE:

ZERO

IMPLICIT

REAL*8

AN ARRAY
(A-H,O-Z)

C
DIMENSION
A(IDIM)
DO 100 I= 1 jDIM
A(D = 0.0
CONTINUE
RETURN
END
C# _;'__;_;_;;;_;;L__;? _"_" ? #_#_
SLrBROUTINE
100

_;__FY;_L_.__._._ _#_#_#_#_###_##__
CALCN(BETA ,D,D0,ALPH A ,FFM,FALT, XNF, DDN)

C
IMPLICIT
C
C

PURPOSE:

REAL*8

(A-H,O-Z)

CALCULATE

CYCLES

TO FAILURE

23

C
C
C

CALLED FROM: DAMAGE

LOGICAL FALT

C
C
XFACI -- I. - (l. - D)**(BETA+I.)
XFAC2 -- I. - (I. - D0)**(BETA+I.)
IF(XNF .LT. 1.E9) THEN
IF(.NOT. FALT) THEN
DENOM = (ALPHA*(BETA+ 1.)*(FFM**BETA))
DDN = ((XFACI**ALPHA) - (XFAC2**ALPHA))/DENOM
IF(DDN.LT. 1) THEN
XNF= -999.
ENDIF
ELSEIF(FALT) THEN
DDN = (LOG(XFAC 1)- LOG(XFAC2))/((BETA+ I)*(FFM**BETA))
IF(DDN.LT. 1) THEN
XNF = -999.
ENDIF
ENDIF
ELSEIF(XNF .EQ. 1.E9) THEN
DDN -- 1.E9
ENDIF
C

24

REPORT

DOCUMENTATION

Form Approved
OMB NO. 0704-0188

PAGE

Publicreporting
burdenfor thiscollection
ofinformation
is estimated
to average1 hourper response,
including
the timefor rewewing
instructions,searching
existingdatasources,
gathering
andmaintaining
thedataneeded,andcompleting
andreviewing
thecollection
of information.Sendcommentsregarding
thisburdenestimateor any otheraspectofthis
colkmct_n
ofinformation,includingsuggestions
for reducing
thisburden,toWashington
Heack:luariers
Services,Directorate
for Information
Operations
andReports,1215Jefferson
DavisHighway,
Suite1204,Arlington,
VA 222024302,andto theOfficeof Management
andBudget,Paperwork
Reduction
Project(0704-0188),Washington.
DC 20503.
1. AGENCY USE ONLY (Leave b/ank)

2. REPORT DATE

3. REPORT TYPE AND DA/t=._ COVERED

March 1994

Technical Memorandum

4. TITLE AND SUBTITLE

5. FUNDING NUMBERS

A Coupled/Uncoupled Deformation and Fatigue Damage Algorithm

Utilizing the Finite Element Method
WU-505-63-12

6. AUTHOR(S)

Thomas E. Wilt and Steven M. Arnold

8. PERFORMING ORGANIZATION
REPORT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)

National
Lewis

Aeronautics
Research

Cleveland,

and Space

Ohio

E-8652

44135-3191

9. SPONSORING/MONITORING

National

Aeronautics

Washington,

D.C.

funded

Center.

and

Space

Administration
NASA

20546-0001

by NASA

Responsible

Lewis

Research

Cooperative

person,

1211.DISTRIBUTION/AVAILABILITY

Unclassified

10. SPONSORING/MONITORING
AGENCY REPORT NUMBER

AGENCY NAME(S) AND ADDRESS(ES)

11. SUPPLEMENTARY NOTES

Steven
M. Arnold,
NASA
(work

Administration

Center

Steven

Center

and Thomas

Agreement

NCC3-248)

M. Arnold,

organization

E. Wilt,

University

and Resident
code

5220,

of Toledo,

Research
(216)

Associate

TM-106526

Toledo,

Ohio

43606

at the Lewis

Research

433-3334.
12b. DISTRIBUTION CODE

STATEMENT

- Unlimited

Subject Category 39

13. ABSTRACT (Maximum 200 words)

A fatigue damage computational algorithm utilizing a multiaxial, isothermal, continuum based fatigue damage model
for unidirectional metal matrix composites has been implemented into the commercial finite element code MARC
using MARC user subroutines. Damage is introduced into the finite element solution through the concept of effective
stress which fully couples the fatigue damage calculations with the finite element deformation solution. An
axisymmetric stress analysis was performed on a circumferentially reinforced ring, wherein both the matrix cladding
and the composite core were assumed to behave elastic-perfectly plastic. The composite core behavior was represented
using Hill's anisotropic continuum based plasticity model, and similarly, the matrix cladding was represented by an
isotropic plasticity model. Results are presented in the form of S-N curves and damage distribution plots.

15. NUMBER OF PAGES

14. SUBJECT TERMS

26

Fatigue; Continuum damage; Isothermal; Anisotropic; Inelastic;

Finite element method
17. SECURITY CLASSIRCATION
OF REPORT

Unclassified
NSN 7540-01-280-5500

18. SECURITY CLASSIFICATION

OF THIS PAGE

Unclassified

16. PRICE CODE

A03

19. SECURITYCLASSIRCATION
OF ABSTRACT

20. LIMITATION OF ABSTRACT

Unclassified
Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Stcl.Z39-18
298-102