Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/compstruc
a
Department of Structural Mechanics, University of Pavia, Via Ferrata 1, 27100 Pavia, Italy
b
Department of Mechanics, Structures and Environment, University of Cassino, Via G. Di Biasio, 43, 03043 Cassino, Italy
Received 25 November 2002; accepted 3 July 2003
Abstract
The present work addresses a simple and effective one-dimensional model able to reproduce the superelastic be-
haviour as well as the shape memory effect. In particular, it considers the transformations from austenite to single
variant martensite and from single variant martensite to austenite, taking into account the influence of the temperature.
Moreover, the training and the two way memory effects are modelled considering the evolutions of the phase-trans-
formation stress thresholds as well as of the residual ‘‘permanent’’ deformation. The time integration of the evolutive
equations is performed adopting a backward Euler scheme and the finite time step is solved through a modified return-
map algorithm. The proposed SMA constitutive law and numerical procedure are adopted to develop numerical ap-
plications. Finally, the ability of the model to reproduce experimental data is assessed.
2003 Elsevier Ltd. All rights reserved.
Keywords: Shape-memory alloys; Pseudoelastic behaviour; Shape memory effects; Training effects; Two way memory effects;
Numerical procedure
0045-7949/$ - see front matter 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0045-7949(03)00319-5
2302 F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317
describe micro-scale level effects, such as, nucleation, (TWME). This effect can be used to realize new potential
interface motion, twin growth [1,15]; the micro–macro- applications, as reversible fasteners, temperature-sensi-
level models which combine micro-mechanical elements, tive actuators, retrievable medical implants, toys and
such as habit planes, martensitic variants, with macro- novelty items.
scale thermodynamics and adopt proper homogeniza- SMA models which take into account the training
tion techniques to derive macro/global parameters from effects and the two way memory effect have been pre-
the micro/local ones [11,19,17,26]; the macro-level sented in literature. Among the others, Tanaka et al. [28]
models which are able to describe some of the major developed a model for the cyclic uniaxial response of
SMA macroscopic features [18,4,24,27,7,25,8]. The shape-memory alloys subjected to thermal and/or me-
macro-models are particularly suitable for the study of chanical loads based on three internal variables, the lo-
SMA structural elements as they can be implemented cal residual stress and strain and the residual volume
more easily in numerical codes. fraction of martensite accumulated during cycles. Lex-
One of the most interesting feature is the SMA re- cellent and Bourbon [20] proposed a one-dimensional
sponse under cyclic loading conditions. In fact the ma- pseudoelastic model for the analysis of the tensile cyclic
terial response progressively changes during the cycles, response of SMA, introducing a new internal state pa-
reaching a limit and stable path after a certain number rameter representing the instantaneous residual mar-
of stress/temperature loading–unloading cycles, i.e. after tensite volume fraction. Abeyaratne and Sang-Joo Kim
the so-called material training. From a phenomenolog- [2] developed a one-dimensional model based on an
ical point of view the training effects on the stress–strain energetic approach, assuming that the critical value of
curve result in the two following aspects: the driving forces required for nucleating and propa-
gating the phase transformations is affected by the defect
• a progressive decrement of the initial and final stress density which depends on the number of loading cycles.
thresholds of the phase transformations, Bo and Lagoudas [10] described a one-dimensional
• a progressive increment of a residual ‘‘permanent’’ model which accounts for the training effects introduc-
deformation. ing irreversible plastic strain. Lexcellent et al. [21] de-
veloped a model in the framework of thermodynamics
Both the described phenomena have a micro-me- introducing a training term in the expression of the free
chanical interpretation which is still under investigation. energy of self-accommodated martensite.
In particular, two micromechanical interpretations are The present work addresses a one-dimensional (1D)
discussed in literature, one based on residual stresses due model able to reproduce the pseudoelastic and shape
to the development of dislocations induced in the ma- memory behaviour of SMA under cyclic loading con-
terial during the cycles, the other on residual permanent ditions, including training and two way memory effects.
martensite accumulated during training [16]. Referring The aim is to develop a one-dimensional model that is
to the first interpretation, the progressively change in the able to take into account many of the special features of
SMA response under cyclic loading is attributed to the SMA behaviour, such as the different response in tension
oriented residual stresses occurring during the disloca- and in compression, the different elastic properties of
tions arrangement. The nucleation and growth of pref- austenite and martensite, the reorientation process, the
erential martensite variants are favored by these residual training effects and the TWME, and at the same time
stresses that tend to be relaxed by shape change. that is suitable for the implementation in a numerical
Moreover, since the generation of dislocations is strictly code.
linked to the development of plasticity, it can be de- The choice to develop a one-dimensional model is
duced that the training effects are connected with a re- motivated by the fact that most of structural elements in
sidual plastic strain. The second interpretation is based SMA applications can be designed using 1D constitutive
on the progressive increment of residual permanent relations, involving the bending and axial response of
oriented martensite variants that occur because of the beams. Furthermore, in a 1D frame, it is possible to
dislocations arrangement. The presence of the disloca- accurately account for the above mentioned several
tions does not allow the complete martensite–austenite special features of SMA behaviour.
transformation during the unloading and heating pro- In the model the transformations from austenite to
cess. Moreover, the small residual martensite plates single variant martensite and from single variant mar-
grow during subsequent cycles. tensite to austenite are considered, taking into account
The presence of permanent deformation, due to the influence of the temperature. Moreover, the micro-
plastic strains or to the residual martensite variants, mechanical interpretation that assumes the training
occurring during the material training allows to obtain effects due to the residual permanent martensite fraction
reversible spontaneous shape change during cooling and is adopted. Hence, the material training under cyclic
heating processes without application of any external loading is modeled adopting exponential evolutions of
stress, which is known as the two way memory effect the austenite–martensite transformation stress thresh-
F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317 2303
olds and of the residual ‘‘permanent’’ deformation. Two evolutionary processes which may produce
Constitutive equations are written in evolutive form. variations of nS are considered:
The time integration of the evolutive equations is per-
formed adopting a backward Euler scheme and the finite • the conversion of austenite into single variant mar-
time step is solved through a modified return-map al- tensite (A fi S),
gorithm. The proposed SMA constitutive law and nu- • the conversion of single variant martensite into aust-
merical procedure are adopted to develop numerical enite (S fi A).
applications. Finally, the ability of the model to repro-
duce experimental data and the TWME is assessed. Hence the analysis is restricted to the case that the
The paper is organized as follows. First a one- temperature is greater than TsAM , i.e. only austenite––
dimensional model is presented, then the time-discrete single variant martensite phase transformations are
integration algorithm for the model, together with a considered. The austenite volume fraction is obtained as:
solution algorithm belonging to the return map family,
nA ¼ 1 nS ð1Þ
is described. Finally, some numerical results, which
show the model ability to simulate the axial and bending In the present model, the micromechanical interpreta-
behaviour of SMA under cyclic loading, are presented. tion that considers the progressive increment of the
permanent deformation during cyclic loading due to an
increment of the irreversible residual martensite fraction
is adopted. Thus, the martensite fraction is set as:
2. Time-continuous model
nS ¼ nR þ n ð2Þ
In the following an uniaxial model able to describe
the pseudoelastic and shape memory effects of SMA where n represent the reversible martensite fraction and
under cyclic loading conditions is presented. In Fig. 1, nR is the residual irreversible martensite fraction.
all the possible phase transformations for the shape-
memory alloys are reported. 2.1. Strain decomposition and elastic relation
The pseudoelastic effect occurs when a loading–un-
loading cycle is performed at a temperature higher then Limiting the discussion to a small deformation re-
TfSA while the shape memory effect is obtained by a gime, the additive decomposition of the total strain e is
combination of stress and temperature path. considered:
A scalar internal variable nS , which represents the e ¼ ee þ nS b nR ðb jÞ þ aðT T0 Þ ð3Þ
single variant martensite fraction, is introduced.
where ee is the elastic strain, nS b is the inelastic strain
due to the phase transition; in particular, b is an internal
variable describing the change of martensite reorienta-
tion [7,22], j is a training parameter whose value is set in
the next subsection, a is the thermal coefficient and T0 is
the reference temperature.
The elastic strain is assumed to be linearly related to
the stress:
r ¼ Eee ¼ Eðe nS b þ nR ðb jÞ aðT T0 ÞÞ ð4Þ
500
1st Cycle
400
50th Cycle
300
σ [MP a ]
200
100
0
0.00 0.01 0.02 0.03 0.04 0.05 0.06
ε
Fig. 2. Stress–strain response of a OMRCO wire of a circular section of Ni–Ti shape memory alloy under cyclic loading: experimental
results.
test on a Ni–Ti wire of circular section, producted by increase of the temperature TsSA and TfSA between two
OMRCO (Glendora, CA, USA), is described. The ex- limit values denoted as:
perimental stress–strain curve, reported in Fig. 2, rep-
For no-trained material ði:e: virginÞ
resents the typical pseudoelastic mechanical response
under cyclic loading of Ni–Ti shape memory alloy. This TsSA ¼ Ts1SA
and several other tests on the same material were per- TfSA ¼ Tf1SA
formed in the Laboratory of Biological Structure Me- ð18Þ
For fully-trained material
chanics (Labs) at the Politecnico of Milano (Italy). In
Fig. 2, only the 1st and the 50th of the 100 performed TsSA ¼ Ts2SA
cycles, are represented. The other cycles are not reported TfSA ¼ Tf2SA
in order to obtain a more clear representation of the
training effects. From the whole cyclic test, it results that with Ts1SA 6 Ts2SA and Tf1SA 6 Tf2SA .
in correspondence of the 50th cycle the material re- Thus in the proposed model, it is assumed that rAS
s ;
sponse has already reached a limit and stable path. It rf ; TsSA and TfSA evolve between the two limiting
AS
can be noted that the initial and final stress thresholds of values, reported in (17) and (18), following an expo-
the phase transformations austenite–martensite and nential law in terms of accumulative measures of the
martensite–austenite decrease during the cyclic loading phase transformation occurrence. In particular, it is
of different quantities. Thus each value of the stress is set:
bounded between two values. In particular, for the
rAS AS AS AS AS SA
s ¼ rs1 ðrs1 rs2 Þ½1 expðbs c Þ ð19Þ
phase transformation A fi S the values of rAS s and rASf
are limited between the following values:
rAS
f ¼ rAS AS AS AS SA
f1 ðrf1 rf2 Þ½1 expðbf c Þ ð20Þ
For no-trained material ði:e: virginÞ
rAS AS
s ¼ rs1 TsSA ¼ Ts1SA þ ðTs2SA Ts1SA Þ½1 expðbSA AS
s c Þ ð21Þ
rAS
f ¼ rAS
f1
ð17Þ
For fully-trained material TfSA ¼ Tf1SA þ ðTf2SA Tf1SA Þ½1 expðbSA AS
s c Þ ð22Þ
rAS
s ¼ rAS
s2 where
rAS ¼ rAS
f f2
c_ AS ¼ n_AS
S c_ SA ¼ n_SA
S ð23Þ
with rAS AS AS AS
s1 P rs2 and rf1 P rf2 .
In the phase transformation S fi A the initial and and bAS AS SA
s , bf , bs and bSA
s are material constants mea-
final stress thresholds decrease as a consequence of the suring the material ability to be trained. The evolution
2306 F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317
600
AS
560 bs =2
AS
bs =1
AS
520 bs =0.5
AS
bs =0.25
[MPa]
480
s
AS
σ
440
400
360
0 1 2 3 4 5 6 7 8 9 10
SA
(a) γ
0.012
0.010
0.008 bR=2
bR=1
bR=0.5
0.006
bR=0.25
ξR
0.004
0.002
0.000
0 1 2 3 4 5 6 7 8 9 10
AS
(b) γ
tion. Eq. (24) is plotted in Fig. 3(b) for different values By linearization, Eqs. (26) and (27) become:
of bR .
kþ1 k dR1 AS
0 ¼ R1 ¼ R1 þ AS dkS ð30Þ
dkS k
3. Time-discrete model and solution algorithm
The evolutive equations governing the shape-memory dR2 SA
0¼ Rkþ1
2 ¼ Rk2 þ SA dkS ð31Þ
alloys phase transformations are integrated developing a dkS k
step by step time algorithm. In particular, once the so-
lution at time tn is known, a backward-Euler implicit For HAS ¼ 1, substituting formula (5), (7) and (29) in
integration procedure is adopted to evaluate the solution Eq. (26), it results:
at time tnþ1 ¼ tn þ Dt. In the following the subscript Ôn’ AS
indicates a quantity evaluated at the time tn while no Rk1 ¼ kAS;k
S ðLAS;k
1 Þa pAS ð1 nS;n kAS;k
S ÞLAS;k
2 ð32Þ
subscript indicates a quantity evaluated at the current
time tnþ1 . The symbol D indicates the variable increments !
at the time step Dt. dR1 AS;k aAS kAS;k
S aAS C AS T ðEA ES Þ
¼ ðL1 Þ 1 þ
The discretization form of the evolutive equations dkAS
S
EA ES LAS;k
1
k
(23) is: !
ð1 nS;n kAS;k ÞC AS T ðEA ES Þ
þp AS
LAS;k
2 þ S
cAS ¼ cAS AS AS
cSA ¼ cSA SA SA EA ES
n þ ðnS nS;n Þ n ðnS nS;n Þ
ð25Þ ð33Þ
In Eq. (25) it is assumed that the conditions (14) and (15) where
for the phase transformation activation are satisfied.
C AS T ðES þ ðnS;n þ kAS;k ÞðEA ES ÞÞ
The time-discrete form of the evolutionary equations (6) LAS;k
1
AS
¼ Sf;n gþ S
The time integration of the evolutive equation (16) of the • Evaluate the phase transformation limit parameters:
reorientation parameter b when it occurs, i.e. when rAS AS AM
AS f C Ts
absðrÞ > rSS , gives: Sf;n ¼ þ RAS
f;n
ES
Dt c½eL b½r rSS when r > rSS
b ¼ bn þ rAS AS AM
Dt c½eL þ b½r þ rSS when r < rSS AS s;n C Ts AS
Ss;n ¼ þ Ss;n
ð42Þ En
rAS
f ¼ rAS AS AS SA
f1 ðrf1 rf2 Þ½1 expðbr c Þ ð44Þ ntr1
s ¼ nS;n
Etr2 ¼ ES else
solve S fi A evolutionary equations:
C AS compute kSA
S from the Newton–Raphson tech-
GAS;tr2 ¼ g T nique
Etr2
compute the training parameters rAS
s and rAS
f
• Check trial states:
endif
if GAS;tr1 < Ss;n
AS
then
ns ¼ ns;n
solution found 3.2. Algorithmic tangent modulus
else if GAS;tr2 > Sf;n
AS
then
nS ¼ 1 In the following, the construction of the tangent
solution found modulus consistent with the time-discrete model is dis-
else cussed.
solve A fi S evolutionary equations: The differentiation of the constitutive equation (4)
compute kAS S from the Newton–Raphson tech- gives:
nique
compute the training parameters TsSA , TfSA and nR
endif dr ¼ E A½e nS b þ nR ðb jÞ aðT T0 Þ
on
Solution scheme for the S fi A phase transformation. þ E 1 A b R ðb jÞ BðnS nR Þ de
onS
• Compute trial states: ð48Þ
First trial state: no phase transformation
ntr1
S ¼ nS;n
where
ntr2
S ¼ nR oE EA ES
E ¼ ¼ E2 ð50Þ
onS EA ES
EA ES
Etr2 ¼
ES þ ntr2
S ðEA ES Þ For HAS ¼ 1, it results:
Table 1
Material data for the three different SMA adopted for the numerical investigations
Material M1
EA ¼ 70 000 MPa ES ¼ 10 000 MPa eþ
L ¼ 0:08
TsAM ¼ 10 C TfAM ¼ 5 C Ts1SA ¼ 30 C Tf1SA ¼ 31 C
Ts2SA ¼ 30 C Tf2SA ¼ 31 C
AS;þ
rs1 ¼ 600 MPa rAS;þ
s2 ¼ 380 MPa rAS;þ
f1 ¼ 700 MPa rSS;þ ¼ 30 MPa
rAS;þ
f2 ¼ 700 MPa
C AS;þ ¼ 6 MPa/C C SA;þ ¼ 8 MPa/C a ¼ 0:00002 C1 nL ¼ 0:012
aAS ¼ 1:1 pAS ¼ 1:0 aSA ¼ 1:0 pSA ¼ 0:97
bAS
s ¼ bf
AS
¼ 0:8 bSA
s ¼ bf
SA
¼ 0:8 bR ¼ 0:6
Material M2
EA ¼ 33 000 MPa ES ¼ 18 300 MPa eþ
L ¼ 0:05 nL ¼ 0:125
TsAM ¼ 4 C TfAM ¼ 3 C Ts1SA ¼ 8 C Ts2SA ¼ 11 C
Tf1SA ¼ 10 C Tf2SA ¼ 14 C
AS;þ
rs1 ¼ 200 MPa rAS;þ
s2 ¼ 90 MPa rAS;þ
f1 ¼ 200 MPa AS;þ
rf2 ¼ 140 MPa
C AS;þ ¼ 12 MPa/C C SA;þ ¼ 15 MPa/C rSS;þ ¼ 30 MPa a ¼ 0:00002 C1
aAS ¼ 1:1 pAS ¼ 1:0 aSA ¼ 1:0 pSA ¼ 0:97
bAS
s ¼ bf
AS
¼ 0:1 bSA
s ¼ bf
SA
¼ 0:1 bR ¼ 0:5
Material M3
EA ¼ 33 000 MPa ES ¼ 18 300 MPa eþ
L ¼ 0:08 nL ¼ 0:3
TsAM ¼ 10 C TfAM ¼ 5 C Ts1SA ¼ 28 C Tf1SA ¼ 32 C
Ts2SA ¼ 28 C Tf2SA ¼ 32 C
AS;þ
rs1 ¼ 40 MPa rAS;þ
f1 ¼ 80 MPa rSS;þ ¼ 30 MPa a ¼ 0:00002 C1
AS;þ
rs2 ¼ 40 MPa rAS;þ
f2 ¼ 80 MPa
C AS;þ ¼ 6 MPa/C C SA;þ ¼ 8 MPa/C aAS ¼ aSA ¼ 1:0 pAS ¼ pSA ¼ 1:0
bAS
s ¼ bf
AS
¼ 0:2 bSA
s ¼ bf
SA
¼ 0:2 bR ¼ 0:5 T ¼ 60 C
F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317 2311
800.00
stress[M Pa]
400.00
0.00
800
600
stress [MPA]
400
200
0
0 .0 0 .5 1 .0 1 .5 2 .0 2 .5 3 .0 3 .5 4 .0 4 .5 5 .0 5 .5 6 .0 6 .5 7 .0
(b) s tra in [% ]
Fig. 4. Pseudoelastic response of a NiTi alloy subjected to equal loading cycles with 7% maximum elongation: experimental results (a);
numerical results (b).
that the numerical response reported in Fig. 4(b) is in stress threshold value rASS and the increment of the re-
good accordance with the experimental behaviour re- sidual martensite fraction nR tend to stabilize after the
ported in Fig. 4(a). The stress value rAS
S and the residual first 20 cycles.
martensite fraction nR vary more significantly in the first In Fig. 6(a) the experimental stress–strain response of
cycle then after 20 cycles the values of rAS
S and nR tend to the material subjected to different loading cycles char-
stabilize. acterized by an increasing value of the maximum elon-
In Fig. 5(a) the evolution of the stress value rAS S is gation that varies from 2% to 8% is plotted.
plotted in terms of the accumulated strain during the In Fig. 6(b) the mechanical pseudoelastic response
S fi A phase transformation, defined as gSA ¼ cSA eL , obtained by the proposed model is represented. It can
and in Fig. 5(b) the residual martensite fraction nR be pointed out that the numerical results (see Fig. 6(b))
versus the accumulated strain during the A fi S phase are in good accordance with the experimental results
transformation, set as gAS ¼ cAS eL ; is represented. The (see Fig. 6(a)). In Fig. 7(a) the stress value rAS
S is plot-
experimental data and numerical results are represented ted in terms of the accumulated strain gSA and in
with diamonds and with a continuous line, respectively. Fig. 7(b) the residual martensite fraction nR versus the
It can be noted that the experimental and numerical data accumulated strain gAS is represented. It can be pointed
are in a very good accordance, thus the choice of the out that the variation of rASS and of nR ; occurring dur-
model parameters is satisfactory. In particular, the ing the S fi A and the A fi S phase transformations,
model is able to simulate the evolution of the stress value respectively, are influenced by two opposite effects;
rAS
S and of the residual martensite fraction nR during the the variations of the parameters rAS S and nR tend to
cyclic loading. Furthermore both the decrement of the decrease as these parameters are modelled adopting
2312 F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317
600
Experimental
560
Numerical
520
[MP a ]
480
s
AS
σ
440
400
360
0 20 40 60 80 100 120 140 160
SA
(a) g [%]
0.14
0.12
0.10
Experimental
0.08 Numerical
ξR
0.06
0.04
0.02
0.00
0 20 40 60 80 100 120 140 160
AS
(b) g [%]
exponential functions (19) and (24), on the other elongation equal to 6%. The parameters of the material
hand they tend to increase as a consequence of the in- M2 are set on the basis of the experimental data. In the
creasing of the maximum deformation of the loading following the tensile behaviour of the wire is investi-
cycles. gated, thus only the parameters characterizing the tensile
Also in this case, the experimental and numerical response are reported in Table 1.
data are in a very good accordance, thus the choice of In Fig. 8 the stress–strain response of the OMRCO
the model parameters appears satisfactory. wire is represented and in particular the 1st and the
Then, the experimental data related to the OMRCO 50th cycle are plotted. It can be pointed out that ex-
wire of circular section of Ni–Ti shape memory alloy, perimental and numerical results are in a very good
represented in Fig. 2, are taken in consideration. The accordance. The proposed model is able to catch the
experimental stress–strain curve represents the pseudo- variation of the initial and final stress thresholds of the
elastic mechanical response of the wire subjected to phase transformations and the increment of permanent
loading cycles characterized by the same maximum deformation.
F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317 2313
800.00
stress[M Pa]
400.00
0.00
800
600
stress [MPa]
400
200
0
0 1 2 3 4 5 6 7 8
(b) s tra in [% ]
Fig. 6. Pseudoelastic response of a NiTi alloy subjected to loading cycles with an increasing maximum elongation from 2% to 8%:
experimental results (a); numerical results (b).
4.2. Two way shape memory effect In order to obtain a fully trained shape memory
material the beam is subjected to a initial history of 10
A cantilever beam characterized by an elastic core cycles of bending loading–unloading, with a maximum
and two SMA layers one on the top and the other on the bending moment equal to 200 Nmm at a constant
bottom is analyzed. It is assumed that the shape memory temperature T ¼ 60 C. The whole training takes 20 s. In
alloy has the same mechanical behaviour in tension and Fig. 9 the loading–unloading bending moment is plotted
in compression so only the data characterizing the ten- versus the transversal displacement of the free end of the
sile behaviour of material M3 are reported in Table 1. beam for the three examined cases. It can be pointed out
Three analyses are performed using different values that after 10 cycles the mechanical response of the beam
for the Young modulus of the elastic core: is stable in all the three cases as the SMA material is
completely trained. Then, the beam is subjected to 2
1st case E ¼ 10 000 MPa cycles of change of temperature, ranging between 11 and
2nd case E ¼ 20 000 MPa 100 C.
3rd case E ¼ 30 000 MPa In Fig. 10 the transversal displacement of the beam
free end is plotted versus time for the three different
The beam length is equal to 10.0 mm, the rectangular values of the elastic core Young modulus. It can be
cross section is 1.0 mm1.0 mm and each SMA layer noted that at the end of the bending training, i.e. at the
has thickness equal to 0.1 mm so the elastic core thick- time t ¼ 20 s, the beam remains in a deformed config-
ness is of 0.8 mm. uration because of the permanent strain accumulated
The beam finite element developed in [22] is adopted during the cycles. After that time, i.e. for t > 20 s, two
to perform the numerical analyses. thermal cycles are performed, thus the beam swings
2314 F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317
600
Experimental
Numerical
550
[MPa]
500
s
AS
σ
450
400
0 10 20 30 40 50
SA
(a) g [%]
0.12
Experimental
0.10 Numerical
0.08
0.06
ξR
0.04
0.02
0.00
0 10 20 30 40 50
AS
(b) g [%]
between two configurations: the beam is able to recover memory effect for shape-memory alloys under cyclic
a greater part of the permanent strain during cooling, loading conditions is presented.
while it goes back to the deformed shape during heating. The model is based on scalar internal variables:
Moreover the influence of the Young modulus of the the martensite fraction and the variables represent-
elastic core on the described two way shape memory ing the material training effects. In particular, the
effect is small. training effects characterizing the pseudoelastic and
shape memory behaviour under cyclic loading condi-
tions, such as the variation of the initial and final
stress thresholds of the phase transformations and of
5. Conclusions the residual martensite fraction, are modelled. The
relatively large set of parameters is necessary for
In the present work a one-dimensional (1D) model modelling such complex phenomena. Since the model
able to reproduce the pseudoelastic and the shape parameters have a clear physical meaning they can be
F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317 2315
500
1st cycle
400
stress [MPa] 50th cycle
300
200
numerical results
100
experimental results
0
0.00 0.01 0.02 0.03 0.04 0.05 0.06
strain
Fig. 8. Stress–strain response of a OMRCO wire under cyclic loading: comparison between experimental and numerical results.
200
100
50
0 x 100
0.00 0.03 0.06 0.09 0.12 0.15
w [mm]
Fig. 9. Bending moment versus transversal displacement for three different values of the Young modulus of the elastic core.
set by performing suitable tensile and compressive tational tools to be used during the design process of
experimental tests. SMA-based devices.
Attention is devoted to the integration of the model The capacity of the model to reproduce experimental
and to the formulation of a solution algorithm. The data and the two way shape memory effect has also been
latter is obtained from a well-known family of elastic assessed. Its ability to describe the material response for
predictor–inelastic corrector algorithms, properly mod- complex loading history (partial loading–unloading
ified to take into account the type of evolution processes pattern) has been numerically tested.
occurring in shape-memory materials. This aspect is of The presented model results simple and able to
non-negligible interest for the development of compu- properly reproduce the pseudoelastic and shape-memory
2316 F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317
x 100
0.15
E=10000 MPa
0.12
0.03
0.00
0 3 6 9 12 15 18 21 24
t [s]
Fig. 10. Transversal displacement versus time for three different values of the Young modulus of the elastic core.
behaviour in tension and bending under cyclic loading [7] Auricchio F, Sacco E. A temperature-dependent beam for
conditions. shape-memory alloys: constitutive modelling, finite-ele-
ment implementation and numerical simulations. Comput
Meth Appl Mech Eng 1999;174:171–90.
[8] Auricchio F, Sacco E. Thermo-mechanical modelling of
Acknowledgements
a superelastic shape-memory wire under cyclic stretch-
ing-bending loadings. Int J Solids Struct 2001;38:6123–
The present work has been partially developed within 45.
the joint French-Italian ‘‘Lagrangian laboratory’’ pro- [9] Auricchio F, Faravelli L, Magonette G, Torra V, editors.
ject as well as partially supported by the Italian National Shape Memory Alloys. Advances in Modelling and
Center of Research (C.N.R.) and the University of Applications. CIMNE 2001, Barcelona.
Cassino. [10] Zhonghe Bo, Lagoudas DC. Thermomechanical modeling
of polycrystalline SMAs under cyclic loading, Part III:
evolution of plastic strains and two-way shape memory
References effect. Int J Eng Sci 1999;37:1175–203.
[11] Boyd J, Lagoudas DC. A thermodynamical constitutive
[1] Abeyaratne R, Knowles J. On the driving traction acting model for shape-memory materials. Part I. The monolithic
on a surface of strain discontinuity in a continuum. J Mech shape-memory alloy. Int J Plast 1996;12:805–42.
Phys Solids 1990;38:345–60. [12] Buehler WJ, Wiley RC. Nickel-based alloys. Technical
[2] Abeyaratne R, Sang-Joo Kim. Cyclic effects in shape Report, 1965. US-Patent 3174851.
memory alloys:a one-dimensional continuum model. Int J [13] Castellano MG, editor. Proceeding of the Final Workshop
Solids Struct 1997;34:3273–89. of ISTECH Project: Shape Memory Alloy Devices for
[3] Auricchio F. Shape memory alloys: applications, microm- Seismic Protection of cultural heritage Structures. June 23,
echanics, macromodelling and numerical simulations. PhD 2000, Ispra, Italy.
dissertation, University of California at Berkeley, Depart- [14] Duerig TW, Melton KN, St€ okel D, Wayman CM, editors.
ment of Civil Engineering, 1995. Engineering aspects of shape-memory alloys. Butterworth-
[4] Auricchio F, Taylor RL, Lubliner J. Shape-memory alloys: Heinemann; 1990.
macromodelling and numerical simulations of superelastic [15] Falk F. Pseudo-elastic stress–strain curves of polycrystal-
behavior. Comput Meth Appl Mech Eng 1997;146:281– line shape memory alloys calculated from single crystal
312. data. Int J Eng Sci 1989;27:277–84.
[5] Auricchio F, Sacco E. A 1D model for superelastic shape- [16] Filip P, Mazanec K. The two way memory effect in TiNi
memory alloys with different elastic properties between Alloys. Scripta Mater 1996;35:349–54.
austenite and martensite. J Non-linear Mech 1997:1101–14. [17] Huang M, Brinson L. A multivariant model for single
[6] Auricchio F, Sacco E. A superelastic shape-memory alloy crystal shape memory alloy behavior. J Mech Phys Solids
beam model. J Intel Mater Syst Struct 1997:1–13. 1998;46:1379–409.
F. Auricchio et al. / Computers and Structures 81 (2003) 2301–2317 2317
[18] Leclercq S, Lexcellent C. A general macroscopic descrip- [24] Raniecki B, Lexcellent C. Thermodynamics of isotropic
tion of the thermomechanical behavior of shape memory pseudoelasticity in shape-memory alloys. Eur J Mech A:
alloys. J Mech Phys Solids 1996;44:953–80. Solids 1998;17:185–205.
[19] Lexcellent C, Goo B, Sun Q, Bernardini J. Characteriza- [25] Qidwai MA, Lagoudas DC. On thermomechanics and
tion, thermomechanical behavior and micro-mechanical- transformation surfaces of polycrystalline NiTi shape
based constitutive model of shape memory Cu–Zn–Al memory alloy material. Int J Plast 2000;16:1309–
single crystal. Acta Materialia 1996;44:3772–80. 43.
[20] Lexcellent C, Bourbon G. Thermodynamical model of [26] Siderey N, Patoor E, Berveiller M, Eberhardt A. Constit-
cyclic behavior of Ti–Ni and Cu–Zn–Al shape memory utive equations for polycrystalline thermoelastic shape
alloys under isothermal undulated tensile tests. Mech memory alloys. Part I. Intergranular interactions and
Mater 1996;24:59–73. behavior of grains. Int J Solids Struct 1999;36:4289–
[21] Lexcellent C, Leclercq S, Gabry B, Bourbon G. The two 315.
way shape memory effect of shape memory alloys: an [27] Souza A, Mamiya E, Zouain N. Three-dimensional model
experimental study and phenomenological model. Int J for solids undergoing stress-induced phase transforma-
Plast 2000;16:1155–68. tions. Eur J Mech A: Solids 1998;17:789–806.
[22] Marfia S, Reddy JN, Sacco E. Superelastic and shape [28] Tanaka K, Nishimura F, Hayashi T, Tobushi H, Lexcel-
memory effects for laminated beams. AIAA Journal, lent C. Phenomenological analysis on subloops and cyclic
accepted for publication. behavior in shape memory alloy under mechanical. Mech
[23] Pelton AR, Hodgson D, Duerig T, editors. Proceedings of Mater 1995;19:281–92.
the First International Conference on Shape Memory and [29] Wayman CM. Introduction to the crystallography of
Superelastic Technologies, 1994. martensitic transformations. MacMillan; 1964.