You are on page 1of 17

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2005; 63:14611477


Published online 17 March 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nme.1315
Simulation of strength difference in elasto-plasticity for
adhesive materials
Rolf Mahnken
1, ,
and Michael Schlimmer
2,
1
Chair of Engineering Mechanics (LTM), University of Paderborn, Warburger Str. 100,
D-33100 Paderborn, Germany
2
Institute for Materials, University of Kassel, Mnchebergstr. 3, D-34109 Kassel, Germany
SUMMARY
Experimental evidence of certain adhesive materials reveals elastic strains, plastic strains and hardening.
Furthermore, a pronounced strength difference effect between tension, torsion or combined loading
is observed. For simulation of these phenomena, a yield function dependent on the rst and second
basic invariants of the related stress tensor in the framework of elasto-plasticity is used in this work.
A plastic potential with the same mathematical structure is introduced to formulate the evolution
equation for the inelastic strains. Furthermore, thermodynamic consistency of the model equations is
considered, thus rendering some restrictions on the material parameters. For evolution of the strain like
internal variable, two cases are considered, and the consequences on the thermodynamic consistency
and the numerical implementation are extensively discussed. The resulting evolution equations are
integrated with an implicit Euler scheme. In particular, the reduction of the resulting local problem is
performed, and for the nite-element equilibrium iteration, the algorithmic tangent operator is derived.
Two examples are presented. The rst example demonstrates the capability of the model equations to
simulate the yield strength difference between tension and torsion for the adhesive material Betamate
1496. A second example investigates the deformation evolution of a compact tension specimen with
an adhesive zone. Copyright 2005 John Wiley & Sons, Ltd.
KEY WORDS: adhesive materials; strength difference; elasto-plasticity; thermodynamic consistency; nite
elements; parameter identication
1. INTRODUCTION
In recent years, adhesion has gained growing attention for the bonding process of industrial
products. Examples of particular application are the industrial branches of automotive industry,
the aeroplane construction and the building industry. Compared to standard bonding processes,
such as blind riveting, ow drill screwing, self-piercing riveting and clinching, adhesive

Correspondence to: R. Mahnken, Chair of Engineering Mechanics (LTM), University of Paderborn, Warburger
Str. 100, D-33100 Paderborn, Germany.

E-mail: rolf.mahnken@ltm.uni-paderborn.de

E-mail: schlimmer@uni-kassel.de
Received 10 May 2004
Revised 12 October 2004
Copyright

2005 John Wiley & Sons, Ltd. Accepted 5 January 2005
1462 R. MAHNKEN AND M. SCHLIMMER
processing has several advantages: rstly, in particular due to automation, adhesive bonding is
economical. Furthermore, as a technical aspect, adhesive is able to make seams tight where
otherwise an additional sealing would be necessary. In this way, structural adhesive bonds in
vehicle construction permit a stiffness increase of the car body as well as an improved crash
performance at a nearly constant weight. Moreover, the adhesion technique offers ideal features
for mixed construction.
Despite the increasing range for application of structural adhesives in vehicle construction,
the demand of the vehicle industry for reliable and efcient numerical design methods to
consider the adhesive bonds in the construction process has not yet been satised. However, in
view of reducing development times for new model ranges, calculation and simulation becomes
absolutely necessary. Therefore, in this publication, a simulation procedure related to the group
of so-called crash optimized adhesives is presented.
In particular, we are concerned with the adhesive material Betamate 1496. Experimental
observations of this material have been published by Schlimmer in Reference [1] for combined
tension torsion tests. The main conclusions of the results are as follows:
1. There exist an almost linear relation in the initial regime of the stress strain diagram up
to a certain threshold (Phase I).
2. Above the threshold, the stress strain relation increases non-linearly with a lower slope
as in Phase I up to a maximum point (Phase II).
3. Increasing loading above the maximum point renders a decreasing stress strain relation
(Phase III).
4. The stress strain relation for unloading in the Phase II regime follows approximately the
slope of the initial stress strain relation of Phase I.
5. The stress strain relations in Phase II and Phase III are dependent on the tension torsion
loading relation. In particular, the yield strength is largest for pure tension and smallest
for pure torsion (strength difference effect).
In this way, the macroscopic stress strain relations for adhesive materials exhibit similar charac-
teristics as observed in metals. In particular, the effects of elastic strain, remaining plastic strain
and hardening can be interpreted in the observations. However, contrary to the micromechanical
development of dislocations in metals, the underlying microscopic phenomena for adhesive
materials are not fully understood so far and deserve further investigations.
The purpose of this work is to concentrate on the observed macroscopic effects of adhesive
materials. This encompasses the formulation and the efcient numerical treatment of a consti-
tutive model, thus enabling to simulate the above-mentioned macroscopic stress strain relations
with a general nite element programme.
The above macroscopic phenomenon of strength difference is, from a more general point
of view, an example of so-called asymmetric effects or as denoted by Altenbach et al. [2]
non-classical effects. These are dened by the observation that a certain type of experiment,
such as a tension test, is not sufcient in order to characterize the material for different loading
scenarios. Instead, additional independent types of experiments, such as compression, shear and
hydrostatic tests, are necessary in order to get a more comprehensive (although in general still
not complete) characterization of the material. It should be emphasized, that the strength
difference effect examined in this paper is restricted to isotropic materials, and therefore
carefully has to be distinguished from the effects of initial (material) anisotropy and induced
anisotropy (such as a Bauschinger effect or damaged induced anisotropy).
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1463
Several publications can be found in the literature for simulation of inelastic material
behaviour with strength difference effects. Many of the approaches are based on a stress
potential dependent on the stress tensor and further state variables, e.g. which describe the
state of hardening, softening or damage, respectively. Typically, the potential is formulated
with polynomials in terms of invariants of the related stress tensor. One of the earlier formu-
lations of this type is given by Schlimmer [1] in 1974. Along this line, further constitutive
equations have been formulated, e.g. in Spitzig et al. [3], Zolochevskii [4], Altenbach et al. [2],
Ehlers [5], Betten et al. [6] and Mahnken [7, 8] among others.
A yield criterion for an isotropic elastic/plastic continuum can be written in terms of the
rst invariant of the Cauchy stress tensor and the second and third basic invariants of the
deviatoric Cauchy stress tensor, respectively. As discussed extensively in Ehlers [5], a general
yield function of this type allows simultaneously for the two effects of pressure stress state
and deviatoric stress state dependence, where the latter can be different, e.g. in tension and
compression.
In order to simulate the experimental data currently available from Schlimmer [1] for adhesive
materials, the third invariant is disregarded in the approach of this paper. Instead a yield function
dependent on the rst invariant of the stress tensor and the second invariant of the deviatoric
stress tensor according to Schlimmer [1, 9] is considered. Additionally, a plastic potential with
the same mathematical structure is introduced to formulate the evolution equation for the
inelastic strains. Two different approaches (subsequently referred as Case 1 and Case 2)
are postulated for evolution of a strain like internal variable. Upon considering thermodynamic
consistency of the model equations, for both Cases 1 and 2 some restrictions on the material
parameters are derived. Furthermore, Case 2 gives a restriction on evolution of the strain like
internal variable.
A further part of the paper is devoted to numerical aspects. An implicit Euler backward
scheme is used for integration of the evolution equations, thus resulting in a (local) non-linear
problem of dimension 7. It is shown that this problem can be reduced to an equivalent
one-dimensional problem for Case 2, whereas the dimension for Case 1 is two. The result-
ing non-linear equation is solved with a Newton algorithm. Additionally, the corresponding
algorithmic tangent operator for the (global) nite element equilibrium iteration of the nite
element structure scheme is derived.
An outline of this work is as follows: Section 2 introduces the yield function and the plastic
potential to formulate the evolution equation for the inelastic strains. Furthermore, consequences
of thermodynamic consistency are investigated. Section 3 describes the integration scheme and
the implementation into a nite element programme. For illustrative purposes, two examples
are presented in Section 4: rstly, the model is applied to simulate the yield strength difference
observed experimentally by Schlimmer [1] for adhesive materials. A second example investigates
the deformation evolution of a sheartension specimen.
1.1. Notations
Square brackets [] are used throughout the paper to denote function of in order to dis-
tinguish them from mathematical groupings with parenthesis (). The three basic invariants
I
1
[A], I
2
[A], I
3
[A] of an arbitrary second-order tensor A are dened as
I
1
[A] = 1 : A, I
2
[A] =
1
2
1 : A
2
, I
3
[A] =
1
3
1 : A
3
(1)
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1464 R. MAHNKEN AND M. SCHLIMMER
Here, 1 is a second-order unit tensor, dened as 1 u = u for arbitrary vector (rst-order tensor)
u. Consequently, the deviatoric/volumetric additive decomposition of A can be written as
A = A
dev
+ A
vol
where A
dev
= I
dev
: A, A
vol
=
1
3
I
1
[A]1, I
dev
= I
1
3
1 1 (2)
where I is a fourth-order unit tensor, dened as I : A = A.
2. CONSTITUTIVE EQUATIONS
2.1. Strain decomposition and Hookes law
Upon using standard notation within a geometrically linear theory, the following relations are
valid:
1. =
el
+
pl
2.
el
= C
1
:
3. C = 2GI
dev
+ K1 1
(3)
Equation (3.1) expresses the additive decomposition of the total strain tensor into an elastic
part
el
and a plastic part
pl
, and in Equation (3.2), the stress tensor is related to the
elastic strain tensor
el
by the fourth-order elasticity tensor C. For the case of isotropic linear
elasticity, C is given in Equation (3.3), where G and K denote the shear and bulk modu-
lus, respectively, and the fourth-order projection tensor I
dev
is dened in Equation (2). The
plastic strain tensor
pl
of Equation (3.1) is obtained from evolution equations as discussed
next.
2.2. Yield function and plastic potential
A yield criterion for an elastic/plastic continuum in terms of the symmetric second-order stress
tensor can be written in the form [] = 0. For an isotropic material, the yield function
reduces as [I
1
, I
2
, I
3
] = 0, where the three basic invariants I
1
, I
2
, I
3
of the Cauchy stress
tensor are determined according to Equation (1). Alternatively, the yield function can be recast
into the form [I
1
, I

2
, I

3
] = 0, where the prime refers to invariants of the deviatoric stress
tensor
dev
, i.e.
I
1
= 1 : , I

2
=
1
2
1 :
2
dev
, I

3
=
1
3
1 :
3
dev
(4)
As discussed extensively in Schlimmer [9] and Ehlers [10], a general yield function of this
type allows simultaneously for the two effects of pressure stress state and deviatoric stress
state dependence, where the latter can be different, e.g. in tension and compression. However,
the experimental observations of these effects require sophisticated and careful measurement
techniques. This has been carried out, e.g. Spitzig et al. [3] for martensite materials, and the
resulting experimental observations are simulated in Mahnken [7].
In order to simulate the experimental data currently available for adhesive materials, the
third invariant I

3
is disregarded, thus restricting our attention to the following yield
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1465
function:
1. = I

2

1
3

2. = Y
2
a
1
Y
0
I
1
a
2
I
2
1
3. Y = Y
0
+ R[e
v
]
4. R[e
v
] = q(1 exp[be
v
]) + He
v
(5)
Remarks
1. The above yield function basically has the mathematical structure formulated by Schlimmer
[9] in 1974. Let us recall from References [5, 9] that it reduces to certain well-known
criteria. Then for the case a
1
= a
2
= 0, the von Mises cylinder is obtained. For Y = Y
0
,
= a
1
/2 = 0, a
2
= a
2
1
/4 the right circular cone
_
3I

2
= Y
0
I
1
of Drucker and
Prager is retrieved. Furthermore, a
1
= 0, a
2
= 0 establishes Greens ellipsoid [10].
2. The stress like internal variable R[e
v
] of Equation (5.4) plays the role of a hardening
function dependent on the strain like internal e
v
dened in the next section.
3. The related material parameters characterizing are Y
0
, a
1
, a
2
, q, b, H.
In order to account for a non-associated ow rule, additionally a plastic potential

is intro-
duced with the same mathematical structure as the yield function dened in Equation (5)
1.

= I

2

1
3

2.

= Y
2
a

1
Y
0
I
1
a

2
I
2
1
(6)
Remark
Compared to the yield function additional material parameters related to the plastic potential

are a

1
and a

2
.
2.3. Rate equations
For evolution of the plastic strain tensor, the following rate equation is derived from the plastic
potential as:
1.
pl
=

*
=

dev
+

t 1
2. t =
1
3
(a

1
Y
0
+ 2a

2
I
1
)
(7)
and where the plastic multiplier is obtained from the loading and unloading conditions

0,

= 0, 0 (8)
In order to formulate a rate equation for the strain like internal variable e
v
two different
approaches are considered: for both cases, the following equivalence of dissipated power
expressions are postulated:
Case 1. e
v
Y = :
pl
e
v
=
1
Y
:
pl
=

Y
(t I
1
+ 2I

2
)
Case 2. e
v
Y
0
= :
pl
e
v
=
1
Y
0
:
pl
=

Y
0
(t I
1
+ 2I

2
)
(9)
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1466 R. MAHNKEN AND M. SCHLIMMER
Remark
For the case a
1
= a

1
and a
2
= a

2
the ow rule (7) is associative; otherwise it is non-associative.
2.4. Thermodynamic consistency
A thermodynamic argument requires the dissipation for arbitrary stress and strain states to be
non-negative, see e.g. Truesdell and Noll [11]. Without going into details of the thermodynamic
framework, the (reduced) dissipation as a consequence of the inelastic strain rate (7) is calculated
according to D = :
pl
R e
v
0. Upon considering Cases 1 and 2 in Equation (9), the
following relations are obtained:
Case 1. D =(Y R) e
v
= Y
0
e
v
0
Case 2. D =(Y
0
R) e
v
0
(10)
Since Y
0
is always non-negative, the requirement for Case 1 in Equation (10) is equivalent to
the following inequality:
:
pl
=

: (
dev
+ t 1)
=

:
dev
+

3
: 1a

1
Y
0
+

3
: 12a

2
I
1
=2

2
+

3
I
1
a

1
Y
0
+
2

3
a

2
I
2
1
0
In the above derivations, additionally Equation (4) for the rst stress invariant I
1
and the
second deviatoric stress invariant I

2
has been used. Several conclusions can be envisaged
in order to satisfy the above inequality. Since, due to Equation (8), the plastic multiplier

and, furthermore, I

2
and I
2
1
are always non-negative, we choose the restrictions a

1
= 0,
a

2
0 on the material parameters, thus satisfying inequality (10) for all stress and strain
states.
Concerning Case 2 in Equation (10), additionally the term (Y
0
R) must be non-negative.
For the specic hardening function (5.4) with Y
0
, H, q 0, this is satised by the restriction
e
v
(Y
0
q)/H. To summarize the following restrictions on the material parameters and the
strain like internal variable e
v
are sufcient for thermodynamic consistency of the constitutive
equations:
Case 1. Y
0
0, a

1
= 0, a

2
0
Case 2. Y
0
0, a

1
= 0, a

2
0, H 0 q 0 e
v

Y
0
q
H
(11)
Remark
As will be seen in Section 4.1, simulation of the experimental data for adhesive materials with
strength difference requires a
1
>0. From the remark in Section 2.3, it follows that the resulting
ow rule is non-associative.
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1467
3. NUMERICAL IMPLEMENTATION
3.1. Integration scheme for the evolution equations
Following standard integration procedures in nite element techniques a strain-driven algorithm
is considered, where the total strain tensor
(n+1)
and initial values
(n)

pl
,
(n)
e
v
are given at a
representative time step
(n)
t . Then, it is the object to nd the corresponding quantities
(n+1)

pl
,
(n+1)
e
v
at time
(n+1)
t =
(n)
t + t consistent with the constitutive equations of the previous
sections; see e.g. Reference [12].
For numerical integration of the rate equations (7) and (9), an Euler backward rule renders
the following update scheme for the plastic strain tensor:
1.
(n+1)

pl
=
(n)

pl
+
pl
2.
pl
=
(n+1)

dev
+
(n+1)
t 1
3.
(n+1)
t =
1
3
(a

1
Y
0
+ 2a

2
(n+1)
I
1
)
(12)
and the strain like internal variable.
1.
(n+1)
e
v
=
(n)
e
v
+e
v
2. e
v
=
_

Y
(
(n+1)
t
(n+1)
I
1
+ 2
(n+1)
I

2
) for Case 1

Y
0
(
(n+1)
t
(n+1)
I
1
+ 2
(n+1)
I

2
) for Case 2
(13)
respectively, and where Case 1 and Case 2 refer to the distinction introduced in Equation (9).
In order to simplify the notation, the index n + 1 referring to the actual time step will be
omitted subsequently.
3.2. Volumetric/deviatoric split of the integration scheme
The integration scheme (12) can be written in terms of the volumetric and deviatoric stress
tensor as follows: upon inserting Equation (12.1) into Equation (3.1), solving for
el
and
inserting into Equation (3.2) renders
=
tr
C :
pl
, where
tr
= C : (
(n)

pl
) (14)
Furthermore, using the relation
C :
pl
= 2G
dev
+3t K1 (15)
the deviatoric/volumetric split of Equation (14) is obtained as
1.
vol
=
tr
vol
K3t 1,
tr
vol
=
1
3
I
tr
1
1, I
tr
1
= 3K tr[
(n)

pl
]
2.
dev
=
tr
dev
2G
dev
,
tr
dev
= 2G(
dev

(n)

pl
dev
)
(16)
where tr[
(n)

pl
] = 1 : (
(n)

pl
). Note that the trial stress tensors
tr
dev
and
tr
vol
are
determined directly from the given quantities
(n+1)
and
(n)

pl
; however, the nal stress tensors

dev
and
vol
are obtained iteratively as discussed next.
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1468 R. MAHNKEN AND M. SCHLIMMER
3.3. Reduction of the integration scheme to a single scalar equation for Case 2
In the following, it will be shown that Equation (16) can be reduced to a single scalar equation
analogously to the radial-return scheme of elasto-plasticity, e.g. described in Reference [12].
For reasons explained below, we will concentrate on Case 2 introduced in Equation (9).
Inserting Equation (12) into Equation (16.1) implies the scalar equation
1
3
I
1
=
1
3
I
tr
1
K3t =
1
3
I
tr
1
K(a

1
Y
0
+ 2a

2
I
1
)
which can be solved for I
1
as
I
1
=
I
tr
1
3Ka

1
Y
0
1 + 3K2a

2
(17)
Upon rewriting, Equation (16.2) yields

dev
=
tr
dev
, where = (1 + 2G) (18)
which means that
dev
is coaxial to
tr
dev
. Taking the inner product of both sides in Equation (18)
renders

dev
:
dev
=
tr
dev
:
tr
dev

2
I

2
= I

tr
2
(19)
where analogously to Equation (4) the second invariant of the deviatoric trial stress tensor
I

tr
2
=
1
2

tr
dev
:
tr
dev
has been introduced.
Using Equation (12.2) and Equation (19), the strain like internal variable of Equation (13.1)
can now explicitly be written as
e
v
=
(n)
e
v
+

Y
0
_
t I
1
+
2

2
I

tr
2
_
(20)
Note that this explicit formulation is not possible for Case 1 introduced in Equation (9), since,
according to Equation (5) the hardening variable Y = Y[e
v
] is dependent on the strain like
variable e
v
.
Next, using the loading and unloading conditions (8) and furthermore Equation (5.1), it
follows that = I

2

1
3
H = 0 for the case of loading, which allows the following non-linear
residual of Equation (19) in terms of the plastic multiplier
1. r[] = 3I

tr
2

2
where
2. = Y
2
a
1
Y
0
I
1
a
2
I
2
1
3. I
1
=
I
tr
1
3Ka

1
Y
0
1 + 3K2a

2
4. = 1 + 2G
5. Y = Y
0
+ R
6. R = q(1 exp[be
v
]) + He
v
7. e
v
=
(n)
e
v
+

Y
0
_
t I
1
+
2

2
I

tr
2
_
(21)
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1469
The above non-linear equation is solved iteratively with a Newton method

(k+1)
=
(k)

1
J
(k)
r[
(k)
], k = 0, 1, 2, . . . , where J =
*r
*
(22)
From the relations (21), the following result is obtained for J:
1. J =
_

2
*
*
+ 22G
_
where
2.
*
*
= 2YY

*e
*
(a
1
Y
0
+ 2a
2
I
1
)
*I
1
*
3.
*I
1
*
= 3K
I
1
2a

2
+ a

1
Y
0
1 + 3K2a

2
4.
*e
*
=
1
Y
0
_
t I
1
+
2

2
I

tr
2
_
+

Y
0
_
2a

2
3
*I
1
*
I
1
+ t
*I
1
*

8G

3
I

tr
2
_
(23)
Note that additionally the notation Y

= dY/de
v
has been used.
3.4. Final stress and strain determination
After having determined the plastic multiplier , the nal stress tensor is obtained as
=
dev
+
vol
where
dev
=
1


tr
dev
,
vol
=
1
3
I
1
1 (24)
and the plastic strains are updated according to relations (11) and (13), respectively.
3.5. Algorithmic tangent modulus
The algorithmic tangent modulus necessary for applying a Newton method for iterative solution
of the global equilibrium problem requires the derivative of the stress tensor dened in
Equation (24) with respect to the total strain tensor , i.e.
C =
d
d
(25)
Straightforward differentiation renders the following result:
C =
2G

I
dev
+
_
1
3
*I
1
*
1
2G

2

tr
dev
_

*
*
+
B
1 + 3K2a

2
1 1 (26)
where *I
1
/* is given in Equation (23.3). The term */* is obtained as follows: we
consider the local problem (21) as an implicit function and conclude
r[, []] = 0
dr
d
=
*r
*
+
*r
*
*
*

*
*
= J
1
*r
*
(27)
where J is already dened in Equation (22). The result for *r/* is as follows:
*r
*
=
_
a
1
Y
0
+ 2a
2
I
1
2YY

t
1
Y
0
_

2
3K
1 + 3K2a

2
1 +
_
3
4YY

Y
0

2
_
2G
tr
dev
(28)
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1470 R. MAHNKEN AND M. SCHLIMMER
The above results for the stress tensor Equation (24), the plastic strains Equations (12) and (13)
and the algorithmic tangent operator can be implemented into a general nite element pro-
gramme. The example shown below has been obtained with the programme ABAQUS [13].
4. REPRESENTATIVE EXAMPLES
4.1. Simulation of experimental data
Experimental measurements for macroscopic characterization of strength difference of adhesive
materials are difcult to obtain and require advanced testing procedures. In the following, we
describe briey the procedure used by Schlimmer in Reference [1].
For preparation of the experiments, cylindrical specimen were glued as shown in Figure 1(a).
The materials are aluminium for the cylindrical specimen and Betamate 1496 for the adhesive
material, respectively. Then the specimens are subjected to combined loading in tension and
shear as shown Figure 1(a). An experimental measurement technique according to Figure 1(b)
is used in order to obtain the axial average deformation v = (v
1
+v
2
)/2 and the circumferential
deformations u and u
F
for the upper and lower part, respectively. Based on these quantities,
axial strains and shear strains

x
= ln
_
1 +
v
d
_
,
xy
= arctan
_
u u
F
d + v
_
(29)
are determined, where d = 2 mm denotes the thickness of the adhesive zone.
It is obvious, that the above quantities are not fully consistent with the geometric linear
theory assumed in this paper. Therefore, the extension of the constitutive equations to a geo-
metrically non-linear theory is of major importance in the ongoing project.
Figure 1. (a) Combined loading in tension and torsion for cylindrical specimen glued with Betamate
1496; and (b) experimental measurement of axial and circumferential displacements.
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1471
Table I. Shear/axial strain ratios in four different experiments.
Shear/axial strain ratio
No. Designation =
1
2

xy
/
xx
1 Tension 0
2 Torsion
3 G3E6 0.5
4 G6E3 2.0
0
0 0.05 0.1 0.15
5
10
15
20
25
30
35
40
Axial strain [-]
Axial stress vs Axial strain
A
x
i
a
l

s
t
r
e
s
s

[
M
P
a
]Tension, Exp.
G3E6, Exp.
G6E3, Exp.
Tension
G3E6
G6E3
Figure 2. Axial stress versus axial strain for three different shear/axial strain ratios (cf. Table I).
Dashed lines with symbols represent experiment, whereas solid lines represent simulation.
0
5
10
15
20
25
30
35
0 0.1 0.2 0.3 0.4
Shear strain [-]
Shear stress vs shear strain
S
h
e
a
r

s
t
r
e
s
s

[
M
P
a
]Torsion, Exp
G3E6, Exp.
G6E3, Exp.
Torsion
G3E6
G6E3
Figure 3. Shear stress versus shear strain for three different shear/axial strain ratios (cf. Table I).
Dashed lines with symbols represent experiment, whereas solid lines represent simulation.
Concerning further strain variables, additional assumptions are needed for the combined
aluminium/adhesive specimen. In this way, it is assumed that due to the higher stiffness of the
aluminium cylinder the additional axial strains are zero, i.e.
yy
=
zz
= 0. Consequently, in
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1472 R. MAHNKEN AND M. SCHLIMMER
Table II. Resulting material parameters for Betamate 1496 obtained from least-squares minimization.
b a
1
a
2
a

1
a

2
E [dimen- Y
0
q [dimen- H [dimen- [dimen- [dimen- [dimen-
[MPa] sion less] [MPa] [MPa] sion less] [MPa] sion less] sion less] sion less] sion less]
Value 1588.7 0.34 36.11 8.08 101.1 33.96 0.643 0.122 0.0 0.279
full detail, the coefcients of the strain tensor for axial and torsional loading are as follows:

Axial
=
_
_
_

xx
0 0
0 0 0
0 0 0
_

_
,
Torsion
=
_
_
_
0
1
2

xy
0
1
2

xy
0
0 0 0
_

_
(30)
Experiments have been performed with four different ratios of
xy
and
xx
as summarized in
Table I. Note that the rst and second experiments refer to the special cases of pure tension
and pure torsion, respectively.
The nal results of the experiments are summarized in Figures 2 and 3, respectively. In
particular, the results demonstrate the dependence of the threshold value on the loading ratio
in Table I.
For simulation of the experimental effects, material parameters Y
0
, q, b, a
1
, a
2
, a

1
, a

2
introduced in Section 2 have to be determined by taking into account the constraints for
the strains of Equation (30). Furthermore, the restriction a

1
= 0 in Equation (11) is taken into
account. Several strategies can be envisaged for the task of parameter identication; see e.g.
Reference [14]. For the problem at hand, a non-linear least-squares functional is formulated
considering the experimental data of Reference [1]. Then a Simplex Nelder algorithm is
used for minimization; see e.g. Reference [15]. The resulting parameter set is summarized in
Table II. From relation (13), we conclude that thermodynamic consistency of the constitutive
equations is guaranteed as long as the strain like internal variable satises
e
v

Y
0
q
H
= 0.82 (= 82%)
which is very much beyond a small strain theory. Therefore, for the purpose of this work,
Case 1 is justied.
In addition to the experimental data, the result of the simulation is also shown in Figures 2
and 3 for the axial stress and the shear stress, respectively. The comparison demonstrates the
capability of the model to simulate the strength difference including the hardening behaviour
for the different loading cases with accuracy sufcient for practical purposes.
4.2. Compact tension specimen with adhesive zone
In this example, the constitutive equations including the material parameters of Table I are
used in order to simulate a compact tension specimen with an adhesive zone. The geometry of
the specimen is shown in Figure 4. The specimen consists of two parts made out of steel with
Youngs modulus E = 210 GPa and Poisson ratio = 0.3. These are assumed to be glued with
the adhesive material Betamate 1496 with a thickness of d = 5 mm. A total load of 2800 N is
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1473
Figure 4. Compact tension specimen: geometry of the sample.
applied to the holes in vertical direction, thus resulting in a Mode I scenario for the crack tip.
The nite element mesh is shown in the top left part of Figure 5, and the top right part of
Figure 5 shows the resulting deformed mesh with a scale deformation factor of 3. The bottom
part of Figure 5 denes the location of the Center Node for the results of Figure 7. Figure
6 depicts the von Mises stress, the pressure (which is the negative rst stress invariant divided
by three) and the strain like internal variable dened in Equation (9) in the adhesive zone.
Figure 7 exhibits some load dependent quantities in the Center Node of the adhesive zone
near the crack tip. Figure 7 (top) shows the von Mises stress and the pressure and Figure 7
(bottom) the development of the strain like internal variable.
5. SUMMARY AND OUTLOOK
This contribution has concentrated on the numerical implementation of a constitutive model pro-
posed by Schlimmer [1, 9], thus enabling to simulate the strength difference in elasto-plasticity
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1474 R. MAHNKEN AND M. SCHLIMMER
Figure 5. Compact tension specimen. Top left: undeformed mesh; top right: deformed mesh of adhesive
zone with deformation scale factor equal to 3; bottom: location of Center Node.
for adhesive materials. Two different approaches (referred as Case 1 and Case 2) are pos-
tulated for evolution of a strain like internal variable. Upon considering thermodynamic con-
sistency of the model equations, for both Cases 1 and 2, some restrictions on the material
parameters are derived. Furthermore, Case 2 gives a restriction for evolution of the strain like
internal variable. For the specic material Betamate 1496, this limit becomes 82%, which,
however, is not signicant for small strain problems.
An implicit Euler backward scheme is used for integration of the evolution equations, thus
resulting in a non-linear local problem of dimension 7. It is shown that for Case 2, this problem
can be reduced to a one-dimensional problem, which is solved with a Newton algorithm. The
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1475
Figure 6. Compact tension specimen. Top: von Mises stress; middle: pressure; bottom: strain like
internal variable in the adhesive zone. The deformation scale factor is equal to 3.
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
1476 R. MAHNKEN AND M. SCHLIMMER
-40
-30
-20
-10
0
10
20
30
40
50
60
0 0.2 0.4 0.6 0.8 1
Load factor
s
t
r
e
s
s
e
s

[
M
P
a
]
von Mises
Pressure
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.4 0.6 0.8 1
Load factor
i
n
t
e
r
n
a
l

v
a
r
i
a
b
l
e

[
-
]
Int. Variable
0.2
Figure 7. Compact tension specimen: Development of load dependent quantities ver-
sus load factor in the Center Node of Figure 5. Top: von Mises stress and pressure;
bottom: strain like internal variable.
dimension of the non-linear problem for Case 1 is two. Furthermore, the associated algorithmic
tangent operator for the global FE-equilibrium iteration scheme is derived.
The comparison of calculated and experimental results demonstrates the capability of the
model to simulate the related characteristic for a Betamate 1496 of strength difference and
hardening effects.
The resulting algorithm has been obtained in the beginning of a project on simulation of
adhesive materials. In this respect, it has to be noted that the experimental quantities are not
fully consistent with the geometric linear theory assumed in the model formulation. Therefore,
the extension of the constitutive equations to a geometrically non-linear theory is of major
importance in the ongoing work.
Furthermore, it has to be remarked that principally a different behaviour in tension and
compression can be simulated by incorporation of odd power terms for the rst invariant of the
stress tensor into the yield function. However, as observed, e.g. by Spitzig [3] for martensite
materials, the relation between the rst and second invariant may not be identical in tension and
compression tests, if these are superposed by a hydrostatic stress state. Similar effects might
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477
SIMULATION OF STRENGTH DIFFERENCE 1477
occur for adhesive materials, which would require also the third invariant of the deviatoric stress
tensor in future work; see e.g. Reference [7]. Further envisaging aspects are rate-dependence
and damage effects.
ACKNOWLEDGEMENTS
For providing the nite element topology of the compact tension specimen, Dr. Jendrny from the
Laboratorium fr Werkstoff- und Fgetechnik, University of Paderborn is gratefully acknowledged.
REFERENCES
1. Schlimmer M. Grundlagen zur Berechnung des mechanischen Verhaltens von strukturellen Klebverbindungen
des Fahrzeugbaus. In Proceedings of Mechanisches Fgen und Kleben, Hahn O (ed.), 2003.
2. Altenbach H, Altenbach J, Zolochevsky A. Erweiterte Deformationsmodelle und Versagenskriterien der
Werkstoffmechanik. Deutscher Verlag fr Grundstofndustrie: Stuttgart, 1995.
3. Spitzig WA, Sober RJ, Richmond O. Pressure dependence of yielding and associate volume expansion in
tempered martensite. Acta Metallurgica 1975; 23:885893.
4. Zolochevskii AA. Modication of the theory of plasticity of materials differently resistant to tension and
compression for simple loading processes. Soviet Applied Mechanics (transl. Prikladnaya Mekhanika) 1989;
24(12):12121217.
5. Ehlers W. A single-surface yield function for geomaterials. Archive of Applied Mechanics 1995; 65:246259.
6. Betten J, Sklepus S, Zolochevsky A. A creep damage model for initially isotropic materials with different
properties in tension and compression. Engineering Fracture Mechanics 1998; 59(5):623641.
7. Mahnken R. Strength difference in compression and tension and pressure dependence of yielding in elasto-
plasticity. Computer Methods in Applied Mechanics and Engineering 2001; 90:50575080.
8. Mahnken R. Theoretical, numerical and identication aspects of a new model class for ductile damage.
International Journal of Plasticity 2002; 18:801831.
9. Schlimmer M. Flieverhalten plastisch kompressibler Werkstoffe. Dissertation, RWTH Aachen, 1974.
10. Green RJ. A plasticity theory for porous solids. International Journal of Mechanical Sciences 1972; 14:
215224.
11. Truesdell C, Noll W. The Nonlinear Field Theories of Mechanics (2nd edn). Springer: Berlin, 1965.
12. Simo JC, Hughes TJR. Computational inelasticity, interdisciplinary applied mathematics. Mechanics and
Materials, vol. 7. Springer: Berlin, 1998.
13. ABAQUS Finite Element Program, Version 6.3. Hibbit, Karlsson, Sorensen, Inc.: Providence, RI, USA, 2002.
14. Mahnken R. Identication of material parameters for constitutive equations. In Encyclopedia of Computational
Mechanics, Stein E, de Borst, Hughes (eds), vol. 2. Wiley: Chichester, 2004.
15. Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes in Fortran. Cambridge University
Press: Cambridge, 1992.
Copyright 2005 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2005; 63:14611477

You might also like