Vous êtes sur la page 1sur 38

-31*

Dchngir nJ Pll-Or1-t Proresses in


Vbtt uvc,~ omnositos

Ctrbric~.Univ. (Ea~iantd)

Aprol

ultzl

WTC,

IITu

EEg

11;
I- L-HM!C
ALUA-to~tlurr
ARACA-

OER

4 j

7t

-wrtsn'

DISCLAIMER NOTICE

THIS DOCUMENT IS BEST

QUALITY AVAILABLE. THE


COPY FURNISHED TO DTIC
CONTAINED A SIGNIFICANT
NUMBER OF PAGES WHICH DO
NOT REPRODUCE LEGIBLY.

BIBLIOGRAPHIC INFORMATION

PB85-163343
Debonding and Pull-Out Processes in

Fibrous Composites,

1984
by J.

K. Wells,

PERFORMER:

and P.

W. R.

Cambridge Univ.

Beaumont.
(England).

Dept.

of

Engineering.
CUED/C-MATS/TR-l 10

The processes of debonding and pull-out in fibrous


composites are described. Models predicting the debond
length and the probability distribution
of pull-out lengths
of fibres and bundles are derived. These lengths are
functions of the fibre, matrix and interface properties.
Prediction is then compared with experiment and a simple
relationship between pall-out
and debond lengths is found.
An understanding of the debonding and pull-out processes is
important because they affect
the fracture toughness of
fibre compcsites.
KEYWORDS:

*Fiber composites,

*Failure,

*Foreign technology.

Available from the National Technical Information


SPRINGFIELD, VA. 22161
PRICE CODE:

/i

PC E04/MF E04

Service,

.9'

..

S-,

..

.. .

.i ~i!

i.!,

P885- 163343

i.

-..

S.....-

'.. .

.. ...

. .

..

,... ...

- .....9.9.F......... ... ...

i!

"I

'I,9

.::.~k!i::.'-
i.
' -,

!-

(/v

.
,9

9.r

""

- - -p9

-...

.-...

""

",

"

...,.

: ,:,:-.'

.. } '

,..

.4.

:. ,

-' 2..

'

'-

.: :, ", - . ,--" "X : " -"-. : "'. ; . ' :. -;- ., . ' . -. .-, :,.
:
..

.9 .

"

:'~ :N"

T O

: " 9 ' .9,


; :
9

F O
9.

E H
/-.
T O

I
E

9..,.... ,

A
V C

.:

.... .

..

'4!

"k
..

: ..,-

.. ...
;..

--

.. J `;.

. .. ,. ., :...

`:``,! :
/

; '1:'-.:

. . ", '. ' (`~ U ::

.94,

'7i
DEBONDING AND PUL!,-OUT PROCESSES IN
FIBROUS COMPOSITES

J K Wells and P W R Beaumont

CUED/C/MATS/TR11O

(1984)

......... .'S..

ijLL-OUiT PROCLSSES 114


DE;ZD3OIi;G h't, !-U
iFIBRiOUS C(IPOSITES

S/

Weells
11

d P W R Benucont

Cambridge univetita"y !7.n'


iering
Department
T'rumpington Street, Cambridge CB2 !PZ

ASTIRACT
The
described.

processes

of

debonding and

pull-out

in

fibrous composites

Models predicting the debond length and the prcbability distri-

buticn of pull-out lengths of fibres and bundles are derived.


are
is

functions
then

out and

of

compared

debond

the
with

lengths

pull-out processes

is

fibre,

matrix

experiment

is

found.

and

and

An

interface
simple

*..

These lengths

properties.

relationship

understanding

of

Prediction
between

and

important because they affect the fracture toughness

of fibre composites.

/
/

'II

*},;ow~ at L- iRes

pull-

the debonding

'

are

.roh Coiltr2,

C.h.e.,.tsev Rued,

Sunbur', on Thames,

Middlesex

I.

1NTRUDUCTION

The
higher
have

impact

than

toughne:;s

that

of

been propose~d

of

either
[i1

its

3]

fibre

reinforced

constituent

based

polymer

phases.

composite

A number of

is

reasons

of fibre pull-out and

on the processes

debonding which absorb energy in the composite.

This paper analyses


out ard dcbonding

the fracture of a composite,

lengths

to be calculated.

not only on single fibres,

These

enabling the pull-

processes

can operate

but also "bundles" of fibres.

DESCRIPTION OF THE FAILURE PROCESSES


Consider

applied

load

(Fig.

Under

monotonic

small

crack

transferred
forces
fails.

a composite

loading

in
by

1).

eventually

A cylindrical

fibres

to the

the

at

material

forms.

to the

become

continuous

Perpendicular

the matrix
shear

of

so

crack

Load,

fibres
large
at

the

notch-tip

once

tarried
still

is

by

and

the matrix,

intact.

a
is

These shear

fibre and matrix

propagates

from

crack surface along the fibre as the applied load increases.


is

to an

a notch.

fractures

the bond between

the interface

parallel

fibre direction

which are

that

aligned

the matrix
This process

called de'oneino.

Some

interfacial
ture.
fibre.
the

This

load

between

fibre

and

matrix

is

still

possible

by

forces due to matrix shrinkage onto the fibre during manufacfriction

Because

fibre is

transfer

of

produces

the

variable

non-uniform
strength

stress

of the

along

fibre

the

along

debonded

its

able to break some distance from the matrix crack-plane

length
where

the stress is
highest.
After fracture,
the composite typically shjw:This process is
ea
from it
with fibres orotruding
matrix crack-pian

called ?u~ -Out.

.....

...

..

I'
THE PROCESS OF DEBONDING

The process cf debonding is


debond

stress

and

the

rate

of

controlled
increase

by two parameters7

the fibre

stress

length

of

along

the

of

debonded fibre due to friction.


After the matrix cracks,
crack

front

is

the fibre stress at the intirfacial

the debond stress,

cd

and

the applied

fiLre

(debond)

stress when

a length x has debonded is:

*.

where

o(x)

f(x)

is

When o(x)

+ f(x)

(1)

the stzess -in

reaches

the fibre

the ultimate

on its

caused by friction

st,.ength of

the fibre,

and the debond length may be determined provided

oaf

surface.

failure

occurs

d and f(x) are known.

/8

3.1
The

Calculation of the Debond Stress


debond

"Thc shear
load,
T ,
o

stress

can

be predicted

exceeded.

example,

occurs

shear-lag

analysis

[4

6].

the

shear strength

similar

results

of the

interface

are obtained,

for

Takaku and Arridge [6] predict:


r

(Zn)

2O (2
-

is

the

Young's

modulus

Other derivations

on T0O , E and G
o f

(2)
f

U'

modulus.

when

In references

Ef

where

stress at the interface can be calculated as a fu.iction of applied

assuming debonding
is

using

of

also agree

The function

the

fibre

and

the

in

on the dependence

involving fibre radius,

tive radius of its surrounding matrix cylinder,

matrix

of debond
r,

fn

denends

shear
stress

and the effecon the

rzc~ise

fibre goem.etry considered.


Such stress-based analyses nave the disadvantage of taking no account
of any

stress concentration

at t,

crack

front

.nl

may..

therefore,

over-

estimate the debond stress.


An

alternative

approach

based

on

the

energetics

of

failure, is

for

interfacial

described by Outwater and Murphy [7] where:


4EfG
=(13f f 2c~

ad

G 2c

is

the

cracking.
since
load

Outwater

they do not
condition.

Clearly,

mode

critical

anU Murphy's
state whether

Wells

only if

strain

release

rate

derivation of this equation


the model

rederived

[8]

there is

energy

is

their

is

confusing

in the fixed grips or fixed


model

to

clarify

this

point.

both sufficient stress to nucleate an interfacial

crack and a favourable energy balance can the crack prbpagate.


The
and

(3)

most

is

to

convenient

experimentally

on fibre radius.

Wells

wires

diameter

of

varying

shown in Fig.

way

2.

[8]

of

distinguishing

investigate

therefore

the

dependence

measured

embedded

in

between

equations

of

debond

(2)

stress

the debond stress of steel

epoxy

resin,

We see that the debond stress is

and

the

results

are

more accurately predicted

by the energy condition.

3.2

Interfacial Frictional Str,'ss Transfer


The

frictional

to compressive
during

cure

stress

transfer

radial stres:.-,-

and

thermal
estimated

stresses

can

be

although

the

model

does

fibre

and

matrix

is

due

produced both by the shrinkage of the resin


-- match

by

not

between

effects

simple

take

during

analysis

into

account

cooling.

developed
the

The
by

radial

Harris

constraining

[9]

effects

of-surrounding fibres.

3.2.1

Si!.ile Linear

MAodel

If

assumed

a fi,,re is

infi:i

-Jy stiff,
4

//,

there is

no Poisson contrac-

tion transverse to the direction of applied load,


facial

str;s

is

uniform.

The

stress

in

and the frictional inter-

fibre embedded

to

a distance

x is therefore:

ff
r
f

a(x) =

A2

where Tf = uP and 4 is
i

[and

is

the

approximation

3.2.2

the coefficient of friction between fibre and matrix

average

radial

compressive

stress

around

the

fibre.

This

has been used extensively in the literature.

Non-linear Model

Although
estimates
be

fibres are

generally

stiff,

the simple

linear model over-

the frictional stress transfer because Poisson contraction

ignored.

Full

allowance

for

this

effect

is

made

in

the

cannot

derivaticn

of Wells [8] which shows that:


a(x)=a

(a - a

(4)

where

where v
matrix

""

o f

and

E are

(subscripts

the
f

Poisson's

and

between

fibre and matrix,

between

fibre

build-up will

2 fEm
fm
Er(l+v
f ff

B =

Vf

ratio and

m respectively),
rf the fibre

and matrix.
fall as

Equation

the

axial

stress that can be produced

(4)

fibre

Young's modulus
p

radius

the

coefficient

and c0

predicts

of

the

of

by frictional loading is

friction

'misfit

that the rate

load increases.

fibre and

strain'

of stress

The maximum shear


o

when

the Poisson

contraction of the fibre is equal to the residual strain of the matrix.


If

the

front at ..hicl

oricin

of

the

the fibre stress i Jd

ld

c,(x)
/

co-ordinates

(op

d)-

is

taken

then ecuction (4)

as

the

debond

becomes:

crack

3.3

Calculation of the Debond Length


Failure

that

portion

strength of

of

fibre of

fibre,

of
the

fibre,

debonded length,

dZ is

uniform

between
cr.

the

strength
matrix

when

occurs

the

surfaces,

crac,.

stress

reaches

to propagate

The debond crack ceases

in
the

and

the

wC

therefore given by the condition:

dFz
f=

- (a

22

-)

f"p

"-nn
hence
2

(-0

d =

an Cp
f
p

is

where Z
(Fig. 3).

the

If

final debonded

length

on both

a d" > 7E no debonding occurs:

if

sides

of

the matrix crack

> a p and (ff" >

CT

debonding

extends along the entire length of fibze.

3.4

Effect of Reinforcement on the Nature of Debon~ing


Equation

will

have a

03)

predicts

low debond

called 'bundle

that

stress.

low

Consequently,

or

large

fibres

radius

which may

phenomenon

be

This process may operate even though

can occur.

debonding'

stiffness

9the

fibre debond stress exceeds the fibre strength and no individual


debonding would be predicted.
by resin.
stiffness

""

The

Bundles

consist of a group of fibres bonded


Such a bundle has a lower

They behave as large single fibres.


and

interfacial

strength

than

parameter

G2c,

single
will

fibre,
also

matrix and interface properties.


.6

fibre

be

but
a

with

larger

combination

of

radius.
the

pure

a bundle will have a corrugated surface,

After debonding,
fore

any

cause

interlocking

residual
the

small movement

single

bund'ie

fibre

may

be

these corrugations.

of

compressive

the bundlc with

of

strain

case.

on

approximated

of

the relevant

an

air-gap

mater`al

properties.

accounts

which

for

the

of

(3)

and

'socket'

produce

fibres,

distribution

equations

by

to its

This will

the bundle

stress

The

respect

and there-

an effective

similar

and

debond

(5)

after

will

to that
stress

of
a

of

substitution

These corrugations may also provile


whitening

observed

in

debonded

glass

A debonded single fibre would not produce a sufficiently

fibre composites.

large airgap to create such whitening effects.

3.4.1

Calculation of Burdle Prroerties for Calculation of Bundle

Debond Lenath
The bundle stiff.ness
of-mixturcs

where

the

and strength

fracture

of

mav be calculated using the rule-

weak

fibres

the bundle to 80% of the ideal rule-of-mixtures


ratio

of

bundle

volume fraction (v
The

is
=

that

of

typical

the

reduces

strength

prediction (10].

composite

of

of

Poisson's

appropriate

fibre

0.32).

interface

parameter,

G2c'

is

assumed

to be

linear

furction

of the two constituent material properties:

cG
2
where

G
l

and

interface and
around

G are
2

,a

the

rf)

G2 1

critical

(7)

energy

pure resin respectivel-i.

the edge of the bundle,

a,

release

rates

given by the square packing approxima-

is

If

,olumf-:

frac-ion

c~

of

The spacing between fibres centres

tion:

w h ore V,,

for" fracture

[Irs
7-

in the bndle

(8]

Wells

measured

approximately

carbon

fibres

50,

500

2 and 60 Jm-'

epoxy

in

and matrix was

Jm-'

for

an

for E-glass,

respectively.

found to be about 5%,

The

epoxy

rcsin

matrix

and

Kevlar and high strength

misfit

strain

between

bundle

and the radius of a bundle,

rb,

Was

typically 500 wm.

3.5

Comnarison of Observed an". Predicted Debond Le: qths in Mcdel Co)moosites


Experimentz

have

been

carried

out

on

model

composite

specimens

where debond lengtLs may be accurately measured and compared to theoretical


predictions.
a

specimen

The specimens
consistcd

of

have
single

been described
layer

of

[3 1 .

previously

lass

inforcing

Brieflv,
to..s

clo Z

to the tensile side of a small epoxy beam loaded in three point bending.
TABLE 1:

Variazion in Debond Lengths with Specimen Type

~PE

DE5OND LENGTH (ram)

SPECIMEN

No. OF

NOTE

30

"OBSERVFD

PREDICTED

2.7 0.2

3.9

Vf = 23%
rb = 0.5 m

90

5.8 + 0.6

8.7

7.1 1.0

10

7.9

Vf

40%

0.45 mn

As 1 with
G1=

The observed
resu't

eing

rer s~ecJ;en,.
(i)

the

lenoths of bundle

a,, !rage

R';ut.

of

a:

le..st

dobonding
100

are shown

measurements

in Table
(10

measurome'*'s

for sooomens of three types are orasc;ted:

Type A had no weft

fibrts

holding

th.e main) to"w.

1 each

in n .1.

bundle spreading to occur during manufacture.

(ii)

Type

B had

the

bundle

small

number

together

and

of

light weft

reducing

the

fibres

present,

inter-fibre

holding

spacing.

This

corresponds to a composite with a locally higher volume fraction.


(iii)

Type C were specimens prepared as type B, but the fibres were sprayed
with mould relcase fluid, producing d weak fibre/matrix bond.
Table

(3),

(6),

(7)

also
and

shows
(8),

the

with

predicted

debond

substitution

of

lengths

material

using

equations

properties

listed

in Table 2.
TABLE 2:

Properties of Fibres and Epoxy Resin

FIBRE
RESIN
CARBON

GLASS

Strength

(MPa)

1650

2480

80

Young's Modulus

(GPa)

70

230

(11m)

0.2

0.2

Radius
Poisson's Ratio

Values

of G 1 , G and misfit strain,


2o

previous section,
that
case.

no

co' are those reported

except in the case of type C specimens where it

fibre-matrix
Agreement

0.35

bonding

between

occurred

observed

and

and

consequently

predicted

values

G
is

is

the

assumed
in

this

generally

good

in

although the predictions are typically 30% higher than the observed value.

Tl

The

PROCESS OF PULL-OUT

fundamcntal

the reinforciicT

Icc,-euln

fibre.

origin

of

In the

11.sence of

pull-cut

is

the

variable

strenoth

flaws

lw

of

a fibre

ir

would break in the region of mnximum stress (i.e. between the faces of
However, when a brittle
a matrix crack),
and no pull-out would result.
fibre

carries

fracture

at a

non-uniform

large

load

flaw in

at a point of higher stress.


this

case

fibre

fracture

along

region

of

This is

will

its

length

low

the

stress

fibre

at

point

'A'

either

or at a minor

shown schematically

occur

may

in

away

Fig.

from

flaw

4.

the

In

region

of maximum stress and produce pull-out during crark propagation.

4.1

A Statistical Model for Pull-Out


After

the matrix

tip increases

c'racks,

the

load

on

Friction

causing debonding.

a fibre

between

close

fibre

to

the

crack

and matrix gives

rise to a non-uniform stress distribution along the fibre length.


The variable strength of a brittle fibre is controlled by the distribution

of

flaws

along

of such material

is

well

the material up to a
in its

its

Experiments

show

that

the

described by a Weibull distribution.

stress d,

fra-tion

of

the fibrres

strength

On loading

P(a)

will

where ao is

a characteristic strength and m the Weibull modulus.

Consider

debonded

fibre

to be non-uniformly loaded,

Fig.

with

sections of length 6x
th
The stress in the i
section increases

5.

series

zero to a. as the debond crack propagates

secti6n

is

the matrix

at. the

point

crack.

The

of

maximum fibre

probability

of

of

stress,

failure

The n th

along the fibre.

in

i.e.
loading

in

the plane
section

zero load tc a.1 is given by the cumulative probability of failure P(.

is

"ail;

simplest form:
P(a) = 1

from

length.

not

Howe'er,

the

simply

P(o. ).

highly stressed

probability
It

section has

alsQ

of

faElre

depe-

not brc..
-1.0 -

upon

occurring
the

before the

in

the

probability

th

that

flaw in the i

th

i
i.

of

from

).

segment
A more

section

causes failure of the fibre.


j > i,

This is

given by the sum of 1

P(cr.)

for all

i.e. all sections more hiohly loaded than section i*.


The

relative

probability

of

fracture

occurring

in

section

is

therefore given by:


n
P(a,)
j 1 j=i+Il (I-

f, 1

Equation

P(o.)) 6x

(10) may be re-expressed

(10)

to present an integral

form of

the cumula-

tive probability function:

{ifkx
d/2 dx

P(o(x")) dxx

0P(a(xlf d1
F(x)

(11)

d/2

d/2

P(U(.X-)){r

f
0

(C(x')) dx"} dx'

1
xt

The pull-out length is

given by:

d
p

and F(x')
quently

is
the

the cumulative probability


cumulative

probability

of x being less than

distribution

of

the

x'.

pull-out

Conselength

being less than or equal to Z is:


P

2
The

length of

This

is

model

p
assumes

fibre which

juqtlfied

since

that

is
the

the

flaw

small by comparison
average

slowly with increasing gauge length.


S.

entire

strength
It

spectrum

is

repeated

in

with the pull-out length.


of

the fibres

changes

only

implies that the full range of

flaws must be present in short lengths of fibre.

*An allowance for the probability of a severe flaw causing fracture in a


"section under lower stress, before fibre failure, occurs in section i,
made by the use of the cumnnlntive Weibull distribution, P(o).

10-

is

The Effect of the Reinforcement on the Nature of Pull-Out

4.2

For

certain

composites

the

occur because

the debond stress

behaviour

anticipated

or high

is

fibre stiffness,

in

debonding

is

greater than

materials

as found

in

a bundle of fibres
behaves
,/

as

with

fibre.

fibres

Although

fibres

is

does

not

the fibre strernth.

This

strong

bond

fibre/matrix

fibre reinforced

(which debonds more easily

large

single

carbon

such cases the pull-out of individual

--7

of

systems.

not possible,

but instead

than the individual

composite,

the

In

bundle

fibres)

still

has

a variable

strength which can be described by a Weibull distribution with

a Weibull

modulus

about

3.5

times

larger

strength of about 80% of the rule-cf-mixute


10].

The

stress

distribution

in

the

than

the

single

fibre

znd

prediction [Harlow and Pheonix,

bundle

is

similar

to

that

in

the

single fibre (Section 3).


The nature

of the dull-out process

debonding process.
Glass

and

debonds

If

Kevlar*

then

the

should
bundle

and pull-out as a
predicted

fibres can debond,


normally
will

The

of

process

the appropriate substitutions

this.

However,

away

intact

may

therefore

from

are made.

if

only

the matrix

composite;

analysed

be

controlled

by the

they will pull-out individually.

fracture

small piece

to do.

do

is

this

using

In general,

is

the bundle
crack

plane

what CFRP is

equation

(11)

when

a composite may show

a comLination of the two types of pullout.

4.2.1

Fractographic observations of Pull-Out


Fractured

*Although Kevlar

samples of epoxy containing E-glass,

is

polymeric

fibre,

it

appears

Kevlar

49 and carbon

to behave as a brittle

material.
In the absence of any detailed information, the teibull distriis
bution (with a Weibull modulus equal to that of glass and carbon)
used to characterise the fibre stLength.

/
-

l1

(Grafil
in

EX-AS)

order

surface

to

verify

of a

individual

have been examined

pull-out

the carbon

a scanning

above predictions.

unidirectional

fibre

By contrast;

the

in

or

(0/90]

with

Figure

or

fibre reinforced

no

'I/

are

in

agreement

Section 4.2, although

variations

with
in

microscope

shows

Glass

matrix

and

between

epoxy shows a

has fractured and pulled-out with intact ratrix


observations

laminate.

little

electron

the

fracture

Zcvlar
the

show

fibres.

solid bundle which

binding the fibres.

the debonding

(SE!4)

behaviour

These

predicted

the properties of the fibre/matrix

in

bond

could allow the other mode of pull-out to occur.

4.3

Comparison of Predicted and Observed Distributions in Comoosites


For a typical composite with brittle fibre reinforcement the function

P(a)

is

of the Weibull form

(eqn.

9)

has

been

and for a debonded

fibre the stress

distribution is:

OWx)

The

integral

aP

(aP

(equation

ad) e

11)

evaluated

numerically

for

various

materials and the predictions compared with experiment.

4.3.1

Results for GFRP

Beaumont and Anstice [111 measured a large number of pull-out lengths


%n GFRP in an effort to ascertain their distribution.
were well
than

for

described by a Weibull distribution but of a much


the

fibre

strength

distribution.

The

average

of

lower modulus
the

Weibull

parameters from over 2000 measurements of Dull-out lengths are:

9
0

= 0.24

I'2

0.08 mm

2.2 0.42

This distribution
7?

The pull-out lengths

is

shown

in

Fig.

7 and is

-12 -

comparcd

with the prediction

eqn.

using

(11),

Good agreement is

SIof
typical

E-glass/epoxy

which is

in

Table 2.

The model predicts that the most probable

(the point at which the slope is

intuitively correct

By comparison,

listed

found between the predicted and experimental distributions

at small pull-out
Sexcept
lengths.
pull-out length

properties

since the

stress

the Weibull distribution

maximised)

is

predicts

is

at Z
p

0,

maximised at that point.


zero probability

of zero

length pull-outs.

4.3.2

Results for CFRP


The procedure described in

Section 4.3.1 has been applied to bundle

pull-out of high strength carbon fibres using data from WelJs

[8].

Charac-

teristic Weibull parameters for 500 measurements were:

t=
0.47
0
M

1.9

0.03 m
o.i

This distribution is

from eqn.

(11)

in Fig.

8,

and may be compared with the result

using typical values for carbon fibre in

and a bundle misfit


di.stributions,

shown

strain

noted

in

of 3%.

Section

epoxy (see Table 2)

The main differences


4.3.1,

are

evident

in

shape

although

of the

agreement

between average values is good.

4.4

A Method for the Rapid Calculation of Pull-Out Lengths


So far the theory has successfully predicted the shape and position

of
and

the

bundle

bility

cumulative
pull-out

distribution

probability
cases.
is

However,

composite

has

beeor

properties

on

and

required.
the

and

of
an

position

between

-13-

\I

often

both

individual

the complete
average

value

fibre
probaof

the

Consequently the effect of chang-

shape

investigated and correlations

for

calculations

time-consuming

pull-out length iz all that is


ing

distributions

of

the

distribution

pull-out and debond lengths

have been sought as a means of convenient prediction.

1.
4.4.1

Effect of Varying 14aterial Properties on Pull-Out Distributions


From

any
in

parameters
the

Changes

length.

In

pull-out

values

in

distribution

the flaw

in most of these

is

affected

or stress

parameters

by

distributions

will also affect

the

the effect of varying fibre misfit strain,

fibre strength and radius on the probability distribution,

which

particular,

length

changes

Figure 9 shows

Weibull modulus,
using

the

(11),

which-reflect

fibre.

debond

eqn.

are

it

otherwise

should

be

typical of

noted

that

an

the

E-glass/epoxy

distribution

composite.

is

relatively

insensitive to the Weibull modulus.

S4.4.2

Correlation between Fibre Pull-out and Fibre De& ond Length

Figure 10 shows the relationship between the average pull-out length,.


and

/ i
S/

*1

several

Kevlar

the

fibre

parameters

debonn
vary.

length,

kd

The average

(as

calculated

pull-out length

eqn.

6)

when

for both glass

and

reinforced material may be estimated by:

6.8

This approximation is accurate to within 10% for


k

< 2.1 mm).

stress,

4.4.3

is

lengths

The behaviour due

to changring

Gl,

< 0.3 mm (or


and therefore

the debo~id

least well predicted.

Correlation between Bundle Pull-Out and Bundle Debond Length


Figure
for

11

shows

typical

high

similar

fibre misfit

strain.

variation

strength

correlation between 2. and


However,

for

carbon
changes

as

noted

Li

using

14

in

between
fibre

of

ccr%,posite.

fibre

the

pull-out

strength,

previous

and

debond

There
radius

section,

is

a
and

changes

due to variation:s in dehonc.


less,

an

approximate

2p < 0.32 mm (7.5 mn

stress do not follow the same trend.

relationship
< Id < 12 mn)

may

be

found

in

the

Neverthe-

region

to an accuracy of 20%,

0.22

mm <

namely:

pd
p

35

/
5

SSU.MMARY
The

energy

of debonding

absorption

and

pull-out.

in

composites

These processes

the lengths to be calculated from fibre,


The
single

study

fibre

shows

and

that

bundle

two

have

been

on

the

studied

length

to enable

resin and interface properties.

types

debonding.

dependent

is

of

deboriding

Debond

are

lengths

possible,

have

been

namely
measured

and found to be in agreement with the predictions of the theory.


A model has been
non-uniform
a

result

of

otress,
the

proposed

predicting

two

types

of

for the fracture of brittle fibres


the

probability

debonding

of

(single

fracture

fibre

and

under

sites.

As

bundle),

two

corresponding modes of pull-out have been proposed and observed in practice.

Acknowledgements
The authors
M F Ashby.
Engineering

wish

One of us
Research

to acknowledge
(JKW)

acknowledges

Council in

the

15 -

discussions

the support

form of a research

the course of this work.

-.

helpful

with Professor

of the Science

and

studentship during

f
References
1

Kelly,

Kaelbe, D Uf,J.

Kirk, J N, Munro, M and Beaumont,

Greszczuk,

Lawrence,

Takaku,

Outwater, J D and Murphy, M C, Paper lc,


24th Annual Technical
Conference of Composites Section SPI (1969)

Wells,

Harris,

B, J. Mater.

10

Harlow,

0 G and Pheonix,

11

Beaumont,

A, Proc.

Roy.

Soc.

Adhesion,

P J,

J.

Mater.

A and Arridge,

J K, PhD dissertation,

P W R, J.

I
it

Ii.

S~~pwrb/r;not"2
S~13

Mater.

Sci.,

Phys.

D:

Appl.

Phys.,

6, (1973),

September-1982

173

Comp. Mater.,

P D, J.

(2978)

2197

Cambridge University,

S L, J.

13,

42

7 (1972),

13 (1978),

P W R and Anstice,

'"I

Sei.,

Sci.,

63

5 (1K73), 245

R G C, J.

03 84

A282 (1964)

L B, ASTM STP 452 (1969),

\~
/ *

Lond.

Mater.

12 (1978), 195

Sci.,

15 (1980),

2619

2038

l/

Figure Captions
1

Schematic of composite fracture ahead of a notch.

Variation of debond stress with wire radius.


Solid line; debond
stress calculated from equation 3 with Ef

210 GPa,

and G 2 .

200 Jm-2

Dashed line; debond stress from equation 2 with Gm = 1.1 GPa,


To = 20 MPa,
rm = 10 mm.
3

Calculation of the debond length

Schematic showing origin of pull-out

Schematic of debonded fibre

Pull-out in
plastic

Predicted and best-fit Weibull distributions for fibre pull-out


lengths in GFRP.
c o = 0.75%.

Predicted and best-fit Weibull distributions for bundle pull-out


lengths in CFRP.
b = 3%.

Effect of changing material parameters on fibre pull-out distribution


in GFRP.
Parameter varying:
(a) misfit strain, (b) Weibull mcdulus,
(c) fibre strength, and (d) fibre radius.

10

Correlations between fibre pull-out ane debond lengths Fo:r (a)


and 'b) Kevlar reinforced material.

(a)

glass,

(Parameter varied

o,

(b)

of;

Kevlar and (c)

+,

co; x, rf;

carbon fibre reinforced

and A,

glass,

G])

Predicted correlation between bundle pull-out and debond lengths


for high strength 7arbon fibre composite (see figure 10 for key)

11

LI

I!
/

Pre-existing

notch

Fibres

)c
d/

"f

i -.

ill

Pull-Out
Fibre fracture
-II

'U-

Fibres debonding

2000
Outwater and
8

Murphy

0l

."

rI

Takaku and
Arridge

1000

,I

500
WIRE RADIUS 9jm)

I
1000

,Ii

a x)=O.- (CrP-adO

MATRIX CRACK PLANE


i

c\
UP

.,

,
T

POSITION

XI

\Li

S.

Region Of maXimum stress


*

Debonded [ength

Frction

he e,

FIBRE STRESS

FIBRE STRENGTH
FIBRE STRESS

.1A

Tar

El

<
2

cOP
-1

.4'..

I,

/1
-

,/,1

i.A

I
'h

!!j\/IK;

I.,

...

.4..
A

'S
.'*'

f
I

'1

L.

.4

I-

-t
-

.......
....

GLASS

a05

Thec.ry-=.

.,.Avrc,

best-fit Weibui[

distribution, to data

0.5
PULL-OUT LENGTH (mm)

I'-

f,-J

i\ICARBON

.ca

Theory
0
CL0

Average. best-fit Weibull


distribution to ddta

-5

Hw
D

0
,1

1-5
PULL-OUT LENGTH (mm)

-0

PULL-OUT LENGTH (m0.5

1~

m101

05

PULL-OUT LENGTH (mm)

0.5

11-75

2-0

2-25 Gpa

iii
S10

LL

PULL- OUT LENGTIH (mm)

2-0

,/
Io

I<

rf6

S~

1*o
Icr_

K
H

--.

,! !

.i

I
I:i

PULL-OUT LENGTH (mm)

2.0

...........

IGLASS
I4

0-70

p
-J

0-50-4-

-#

tI

*i

DEBONO

LENGTH (mmr)

~KEVLAR 49
-0.7E
x

I-.

~01
CL

0.2

0.1-

2j0312
DEODLNGH(m

-.

CARBONI

E 0-4-20

p =Ld/35
-20
A

D
A

0
0-2-

~O.1
-,

I1

fBUNDLE
01

10

DEBOND LENGTH (mm)

12

14

Vous aimerez peut-être aussi