Scientia Iranica B (2012) 19 (2), 249–257

Sharif University of Technology

Scientia Iranica
Transactions B: Mechanical Engineering
www.sciencedirect.com

A comparative analysis of some one-dimensional shape memory

alloy constitutive models based on experimental tests
H. Sayyaadi a,∗ , M.R. Zakerzadeh a , H. Salehi b
a
School of Mechanical Engineering, Sharif University of Technology, Tehran, 11155-9567, Iran
b
Institute of Aeronautical Research Center, Isfahan, Iran

Received 26 September 2010; revised 12 December 2011; accepted 17 January 2012

KEYWORDS Abstract Recently, Shape Memory Alloys (SMAs) have been receiving more attention and further study,
Shape memory alloys; due to their ability to develop extremely large, recoverable strains and great forces. In this paper, three
1-D constitutive models; major models of SMA behavior, used in the literature, for studying the static performance of SMA
Experimental data. components attributed to Tanaka, Liang and Rogers, and Brinson, have been analyzed and compared. The
major differences and similarities between these models have also been emphasized and presented in this
paper, based on the experimental data of the shape memory and superelastic behavior of an SMA wire. It
is shown that these models all agree well in their prediction of the superelastic behavior of SMAs at higher
temperatures, but the models developed by Tanaka, and Liang and Rogers cannot be used for predicting
the shape memory effect behavior of SMAs. It is also shown analytically that the original evolution kinetics,
proposed by Brinson, in a specified region, are inadmissible for some thermomechanical loading and initial
conditions. Furthermore, corrected evolution kinetics is addressed here in detail, that is; admissible and
valid in this region. According to this research, regarding the validation assessment of three major 1-D
constitutive models with experimental data, it will be shown that the Brinson model with the corrected
evolution kinetics developed by Chung et al. can be applied for the modeling of SMA smart structures,
such as flexible SMA beam structures.
© 2012 Sharif University of Technology. Production and hosting by Elsevier B.V. All rights reserved.

1. Introduction wide range of fields, such as aerospace, medical, civil and me-
chanical engineering [1].
One well known example of the real-world application of
Smart materials have been extensively used in recent years,
SMAs is hydraulic tube coupling used on the F-14 in 1971 [2].
due to their great potential in revolutionizing engineering ap-
Since then, engineers in various industries have continued
plications and design, particularly for active and passive con-
to use the unique properties of SMAs in solving engineering
trol of structures. Among these materials, Shape Memory Al- problems. One problem in the aerospace industry is shape
loys (SMA) have been receiving more attention and further control in morphing structures.
study, due to their ability to develop extremely large recov- Another application of SMA in the aerospace industry is
erable strains and great force. Their applications encompass a the development of a variable geometry airfoil. Through SMA
actuation, the airfoil effectively changes its configuration from
∗ Corresponding author.
symmetric to cambered [3]. Pairing SMAs and Micro Electro-
E-mail address: [email protected] (H. Sayyaadi). Mechanical Systems (MEMS) to decrease the turbulence of
an aerodynamic surface is another use in the aerospace
1026-3098 © 2012 Sharif University of Technology. Production and hosting by industry [4].
Elsevier B.V. All rights reserved. Peer review under responsibility of Sharif
With regard to their unique mechanical properties, the
University of Technology.
Ni–Ti SMAs have appealed to engineers much more than other
doi:10.1016/j.scient.2012.01.005 SMA types. The Ni–Ti SMAs are characterized by a very high
ultimate tensile strength of up to 1000 MPa, elongation to
failure reaching 50%, a high recovery stress up to 800 MPa or
recovery strain up to 8% and an excellent damping capacity [5].
To use these SMA wires in industrial applications, such as
shape control applications, it is necessary to have a meticulous
250 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257

understanding of the mechanical behavior of SMAs, in order to In Brinson evolution kinetics, under certain conditions, the
predict and control their potential. Although the characteriza- local evolution law for the formation of the martensitic fraction
tion of SMAs is based on the same principle used to test other leads to an inadmissible phase fraction (ξ > 1). This pertains to
materials, special consideration should be paid when monitor- a case where both stress and temperature induced martensite
ing the response of shape memory alloys in experimental tests, evolve simultaneously. Modified evolution laws presented by
due to their complex behavior [6]. Chung et al. to overcome this anomaly are analyzed [15].
Shape memory behavior in SMAs is due to a reversible To show how each of these models match with a given set
thermoelastic crystalline phase transformation between a high of experimental data of a SMA sample, the thermomechanical
symmetry parent phase (austenite) and a low symmetry properties of that sample are first obtained. Furthermore,
product phase (martensite) [7]. The phase transformation the phase diagram is constructed by determining the phase
occurs as a function of both temperature and stress. At zero boundaries using experimental data, wherein tests are carried
stress point, phase transformation is triggered at temperatures out under simple loading conditions like a constant stress
addressed by As , Af , Ms and Mf , which, respectively, are thermal cycle or a constant temperature mechanical (stress)
representative of austenite start and finish, and martensite cycle. A linear fit to the test data is used to obtain idealized
start and finish. At temperatures lower than Ms , applying stress phase boundaries.
causes a transformation from the ‘twinned’ martensite phase In the majority of work performed to obtain thermome-
to the stress-preferred or ‘detwinned’ martensite. This results chanical properties, the SMA were heated using Joule resistive
in large strains under relatively constant stress conditions. heating and the cooling was done by free convection. Further-
This strain can be wholly recovered by heating the material more, wire temperature was measured using a thermocouple
above Af and then cooling it without applying any stress. mounted on the wire, while the wire was heated by passing an
This phenomenon is called the Shape Memory Effect (SME). electric current through it. Since passing an electrical current
However, at temperatures above As , applying stress causes through wires results in local non-uniformities in wire temper-
a transformation from the austenite phase to the stress- ature [8], the test on the mentioned SMA wires in this paper
preferred martensitic phase. This process, which is reversible is carried out using a test machine with a controllable thermal
in a hysteresis loop by removal of the stress, is called chamber. By using a thermal chamber for temperature control,
pseudoelasticity [8]. electrical current passing through the wire is eliminated and
The physical behavior of SMA is a function of three relevant uniform wire temperature along the wire is ensured.
variables: stress, strain and temperature, and their related time This paper is organized as follows. Section 2 discusses the
rates. The SMA constitutive models try to describe the SMA Tanaka, Liang and Rogers, and Brinson constitutive models.
behavior as a function of these three variables. Most of these Section 3 analyzes the testing methods that are used to obtain
constitutive models are based on specific material parameters material parameters of a SMA wire (Ni–Ti, one way shape
and are determined by experimental tests [8]. memory, 0.25 mm diameter). In order to obtain the required
Several three-dimensional constitutive models have been thermomechanical properties and parameters of this wire,
developed for SMAs, such as the Boyd and Lagoudas model [9], careful experimental measurements have been carried out.
Ivshin and Pence [10], Graesser and Cozzarelli [11]. However, Section 4 discusses the validity of each of the mentioned
none of these models is widely used in engineering applications, models with experimental data. Finally, in Section 5, it is
because the developed models are too complicated and require shown analytically that the evolution kinetics proposed by
many parameter definitions. These models can be simplified for Brinson [14] is inadmissible for certain thermomechanical
the 1-D applications, but it is not reasonable to use 3-D models loading. Furthermore, a corrected evolution kinetics presented
for 1-D applications, such as SMA wires. Another advantage by Chung et al. [15] to overcome this anomaly is addressed.
of 1-D models is that the parameters are engineering-based
and simply determined by typical mechanical engineering 2. SMA constitutive models
experiments.
The Tanaka model is one of the first constitutive models for In this section, three of the 1-D constitutive models that
SMAs [12]. This formulation assumes that strain, temperature have been proposed to describe the SMA behavior are discussed.
and martensite volume fraction are the only state variables The reason for only these models being discussed is due to
for this model and the stress is determined based on these their applicability to the entire range of thermomechanical
variables. Also, phase transformation kinetics is expressed in an conditions. Furthermore, they are simple and widely used in
exponential form and is a function of stress and temperature. many engineering applications.
Liang and Rogers [13] proposed a model based on the rate form
of the Tanaka constitutive formulation. Nevertheless, a cosine 2.1. Tanaka model
function is used to model the martensite volume fraction [8].
A major and important shortcoming for both Tanaka Tanaka proposed a unified one-dimensional martensitic
and Liang and Rogers models is that they only explain phase transformation model in 1986 [12]. This formulation
phase transformations from martensite to austenite and was actually limited to the stress-induced martensite phase
austenite to martensite. Since the Shape Memory Effect transformation only. The basic assumption he made was that
(SME) at lower temperatures is caused by the conversion the thermomechanical process of the SMA material is fully
between stress-induced martensite and temperature-induced expressed by three major state variables: strain, temperature
martensite, these models cannot be implemented in the and martensite volume fraction [16].
detwinning of martensite, which is responsible for the SME [8]. The critical stress-temperature profile used in this constitu-
This problem was solved by the Brinson model [14]. In this tive model is shown in Figure 1. The two material constants,
model, the martensite volume fraction is separated into two CA and CM , called stress-influence coefficients, which indicate
parts, stress-induced and temperature-induced martensitic the influence of stress on the transition transformation, are ob-
volume fractions. tained from experimental tests.
 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257 251

Figure 2: Critical stress-temperature profiles used in Brinson model.

Figure 1: Critical stress-temperature profiles used in Tanaka model.

where aA , aM , bA and bM are four material constants:
The constitutive equation in this model relating to the state π −a A
variables stress (σ ), strain (ε ) and temperature (T ) in the terms aA = , bA = ,
(Af − As ) CA
of the martensite volume fraction (ξ ), is:
π −a M
σ − σ0 = E (ξ )(ε − ε0 ) + θ (T − T0 ) + Ω (ξ )(ξ − ξ0 ), (1) aM = , bM = .
(Ms − Mf ) CM
where (σ0 , ε0 , ξ0 , T0 ) represent the initial state or original
condition of the material. In this equation, E is the module of Furthermore, ξM and ξA are the initial martensite volume
elasticity and assumed to be a linear function of the martensite fractions, prior to the current transformation.
volume fraction:
E (ξ ) = EA + ξ (EM − EA ). (2) 2.3. Brinson model

Ω is called the phase transformation coefficient and is defined As stated before, the major shortcoming of both Tanaka,
as:
and Liang and Rogers models is that they can only explain
Ω (ξ ) = −εL E (ξ ), (3) the phase transformation from martensite to austenite and
its reverse transformation. Since the Shape Memory Effect
where εL is the maximum recoverable strain. The kinetics
(SME) at lower temperatures is caused by the conversion
equations describing the martensite volume fraction as an
between stress-induced martensite and temperature-induced
exponential function of stress and temperature are:
martensite, these models cannot be implemented to the
ξA→M = 1 − exp(aM (Ms − T ) + bM σ ), detwinning of martensite, which is responsible for the SME [8].
for T > Mf This problem was solved by the Brinson model [14]. In this
model, the martensite volume fraction (ξ ) is separated into
and CM (T − Ms ) < σ < CM (T − Mf ), (4)
stress-induced (ξs ) and temperature-induced components (ξT ):
ξM →A = exp(aA (As − T ) + bA σ ),
for T > As ξ = ξs + ξT . (8)

and CA (T − Af ) < σ < CA (T − As ), (5) The original form of the constitutive equation in this model is,
as follows [14]:
where aA , aM , bA and bM are material constants, in terms of
transition temperatures As , Af , Ms and Mf . σ − σ0 = E (ξ )ε − E (ξ0 )ε0 + Ω (ξ )ξs − Ω (ξ0 )ξs0
+ Θ (T − T0 ). (9)
2.2. Liang and Rogers model
Brinson and Huang, in their research, applied some different
This model has almost the same form of constitutive loading and unloading operations with special initial conditions
equation as proposed in the Tanaka model. However, for and also performed mathematical techniques, while using new
phase kinetics, a cosine function to describe the martensite parameter definitions to violate initial conditions. A simplified
volume fraction as a function of stress and temperature is governing equation is reduced to the following form [17]:
supposed, respectively [13]. The kinetics equations describe the
martensite volume fraction as a cosine function of stress and σ = E (ξ )(ε − εL ξs ) + Θ (T − T0 ). (10)
temperature, as follows:
To allow for the shape memory effect at temperatures below Ms ,
1 − ξA 1 + ξA the transformation phase equations of Liang and Roger Eqs. (6)
ξA→M = cos[aM (T − Mf ) + bM σ ] + ,
2 2 and (7) with critical stresses, as defined in Figure 1, are modified
for T > Mf , to describe the definition of ξT and ξs [14]. Variation of critical
stresses with temperature for transformation consistent with
and CM (T − Ms ) < σ < CM (T − Mf ), (6)
separation of ξ into two components is shown schematically in
ξM Figure 2.
ξM →A = cos [aA (T − As ) + bA σ ) + 1] ,
2 The evolution equations for calculation of the martensite
for T > As , fractions as a function of temperature and stress can now be
and CA (T − Af ) < σ < CA (T − As ), (7) represented, according to Figure 2, as follows:
252 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257

– Conversion to detwinned martensite:

For T > Ms & σscr + CM (T − Ms ) < σ < σfcr + CM (T − Ms )
 
1 − ξs0 π
ξs = cos [σ − σfcr − CM (T − Ms )]
2 σscr − σfcr
1 + ξs0
+ ,
2
ξT 0
ξT = ξT 0 − (ξs − ξs0 ). (11)
1 − ξs0
For T < Ms & σscr < σ < σfcr
  Figure 3: DSC test for the SMA wire and its transformation temperatures.
1 − ξs0 π 1 + ξs0
ξs = cos (σ − σf ) +
cr
,
2 σscr − σfcr 2 Table 1: The transformation temperatures obtained by the DSC test.

ξT 0 Martensite finish (Mf ) 46.8(°C)

ξT = ξT 0 − (ξs − ξs0 ) + ∆T ε , (12) Martensite start (Ms ) 56.7(°C)
1 − ξs0 Austenite start (As ) 69.4(°C)
76.5(°C)
where, if Mf < T < Ms and T < T0 Austenite finish (Af )

1 − ξT 0
∆T ε = {cos[aM (T − Mf )] + 1}.
2 heat to advance the endothermic transition. So, a characteristic
else: dip is seen in the heating curve. On the contrary, during cooling,
due to the exothermic transition from the austenite phase to
∆T ε = 0.
martensite, the SMA sample gives off heat and, therefore, a
- Conversion to austenite: characteristic peak is seen in the heat flow curve [18].
For T > As and CA (T − Af ) < σ < CA (T − As ) Figure 3 illustrates the result of a DSC test for the sample
SMA wire. The required power to maintain a constant heating
ξ0 σ
    
ξ= cos aA T − As − +1 , or cooling rate for the SMA specimen is shown on the ordinate
2 CA axis, and the temperature of the SMA wire is represented on the
ξs0 abscissa axis. The transformation temperatures are generally
ξs = ξs0 − (ξ0 − ξ ),
ξ0 determined by drawing tangent lines to the beginning and end
regions of the transformation peak/dip, and the baseline of the
ξT 0
ξT = ξT 0 − (ξ0 − ξ ). (13) heating and cooling curves. The transformation temperatures
ξ0 at zero-stress for the available SMA wire are obtained and
As discussed before, in the last section, it was shown that this tabulated in Table 1 [6].
evolution kinetics is incorrect for certain thermomechanical
loading and leads to an inadmissible martensite fraction 3.2. Loading test method
(ξ > 1).
Since the characteristic phase transformation temperatures
3. Experimental tests are stress-dependent, their value must be obtained at different
stress levels. To obtain the transformation temperatures at
In order to obtain the required thermomechanical properties different stress levels, isostress tests were performed on the
of a Ni–Ti alloy, careful experimental measurements have been SMA wire.
carried out on a FlexinolTM actuator wire manufactured by Since performing the constant stress test, using a tensile
Dynalloy, Inc. For the experiment, a one way shape memory, test machine, is very difficult, this test is done by hanging
0.25 mm in diameter with low temperature (70 °C) and a Ni–Ti the dead weights to the wire in the controllable temperature
SMA actuator wire, has been selected. chamber, as shown in Figure 4. During each measurement,
Several methods have been reported for determining changes in the position of the dead weight (expressing the
SMA transformation temperatures of which the Differential length of the wire), being a function of its temperature over
Scanning Calorimeter (DSC) test, the electrical resistance the transformations, were precisely observed and measured.
method and the applied loading method are the most A proximity inductive position sensor was used to measure
commonly used. Among these methods, in this paper, data from the position of the dead weights. A thermal chamber with a
the DSC test and loading methods are used. controller was used to control the temperature of the specimen.
One thermocouple mounted directly on the wire was used
3.1. DSC test method to monitor the temperature of the wire. Also, a HBM data
acquisition system was used to record the temperature and
The DSC test is the most popular and convenient way strain values.
for determining SMA phase transformation temperatures [18]. Furthermore, experimental stress–strain tests were carried
In this test, the austenite start (As ), austenite finish (Af ), out at different temperatures using an Instron Tensile Test
martensite start (Ms ) and martensite finish (Mf ) temperatures Machine with a temperature controllable chamber. This setup
at zero-stress can be obtained. When the SMA starts its was used to characterize quasi-static stress–strain properties at
transformation from martensite phase to austenite, it absorbs different temperatures (Figure 5).
 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257 253

Figure 4: Test set up used for constant stress tests.

3.2.2. Isothermal tests

Two sets of isothermal test (monotonic loading) are
performed on the specimen: first, monotonic loading below
the martensite finish temperature (Mf ) and second, monotonic
loading above the austenite finish temperature (Af ). It should
be mentioned here that similar to the isostress tests, in these
isothermal tests, the SMA wires were first cycled for 35 loading
times to establish stable behavior, and to remove the residual
deformation during testing; then, these isothermal tests were
performed completely.
Monotonic loading below martensite finish temperature (Mf )
Testing consists of putting stress on the material until
detwinning completes and to some maximum stress. The
maximum stress value is not known before starting the test.
Rather, it is selected during the test course by noting that
detwinning has completed. The sample is then unloaded.
The temperature of the specimen is finally homogeneously
Figure 5: Instron tensile test machine with controllable temperature chamber. increased until above Af , and any strain recovery is detected [6].
From this test, the detwinning start and finish stresses (σs ,
3.2.1. Isostress tests σf ), as well as the elastic modulus for martensite (E M ) and the
The basics of this test is that the transformation to maximum recoverable strain (εl ), can be approximated. Here,
the martensite phase and the reverse transformation are these loading/unloading cycles are applied to the specimen at
accompanied by a large deformation, while an external force T = 20, 25, 30 and 35 °C.
is applied to the specimen during cooling and shape recovery Monotonic loading above austenite finish temperature (Af )
during heating. Thus, as a primary effect of the phase The second monotonic test assesses material behavior at
transformation, the temperatures at which there is a large temperatures greater than Af (i.e., possible pseudoelasticity).
change in geometry can be considered as a straightforward Here, these loading/unloading cycles are applied to the
determination of the transformation temperatures [18]. It specimen at T = 80, 85, 90, 95 and 100 °C, and nearly full
should be mentioned here that before performing these tests, pseudoelasticity is observed. From strain–stress characteristics
the SMA wire should first be cycled to ensure consistent of the FlexinolTM actuator wire, the Young’s module of the
behavior. The available SMA wire is cycled for 50 times before austenite phase, E A , and the maximum recoverable strain, εl ,
testing, and it was seen that the behavior of the wire in could be established.
the isostress tests was consistent after this cycling process. The results derived by isothermal and isostress tests are
These isostress tests are performed at 30, 60, 90, 130, 170 and summarized in Table 2. The Tanaka and Brinson phase diagram,
190 MPa. based on experimental parameters, are shown in Figures 6 and
For each load case, the value of four characteristic transfor- 7, respectively.
mation temperatures of the wires can be easily measured by It should be mentioned that in the Brinson model, the
drawing tangents lines to the start and end regions of the trans- parameters, Ms and Mf , are addressed as the temperature
formation and the baseline of the heating and cooling curves. above which the martensite transformation stresses are a linear
Determination of the temperatures at which the phase trans- function of temperature, as depicted in Figure 7. However, in
formations begin and end, for different constant stresses, al- Tanaka, and Liang and Rogers models, these parameters are
lows construction of the final phase diagram. Then, by a simple defined at zero stress point, and these are the temperatures for
extrapolation, the values of the characteristic transformation the martensite start and finish, derived by cooling the austenite
temperatures Ms , Mf , As and Af of the SMA wire at ‘‘zero-stress phase without applying any kind of stress [8]. Thus, when
conditions’’ are obtained. calculating these constants from experimental critical points,
254 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257

Table 2: Experimentally derived material parameters.

Material Value
parameter

Mf 38.5(°C) (Tanaka) 43.9(°C)(Brinson)

Ms 43(°C)(Tanaka) 48.4(°C)(Brinson)
As 68(°C)
Af 73.75(°C)
CA 6.73(MPa/°C)
CM 6.32(MPa/°C)
εL 4.1 (%)
EA 31.5 (GPa)
EM 20 (GPa)
σs 25 (MPa)
σf 78 (MPa)

Figure 8: The experimental shape memory effect at T = 27 °C and comparison

with the mentioned models (T < Mf ).

Figure 6: Tanaka experimentally derived phase diagram.

Figure 9: The experimental pseudoelastic effect at T = 110 °C and comparison

with the mentioned models (T > Af ).

4. Comparison of 1-D constitutive model results with

experimental data

In order to demonstrate the effectiveness of the models

developed by Tanaka, Liang and Rogers and Brinson, a
comparative study has been accomplished here. In this study,
the shape memory and superelastic behavior, as well as
isostress test results, were investigated for the case of the
mentioned Ni–Ti SMA wire.
The experimental isothermal stress–strain plot at T = 27 °C
(that is below As ) is compared by the prediction of these
models in Figure 8. Since the whole loading and unloading
Figure 7: Brinson experimentally derived phase diagram. process of the specimen stays in the full martensite phase, from
these three mentioned constitutive models, only the Brinson
model can be applied for predictions at temperatures below
the numerical values used for Tanaka, and Liang and Rogers
As [8]. As seen from this figure, the Brinson model matches the
models for Ms and Mf differ from those used for the Brinson
experimentally measured characteristics of the shape memory
model. These different values for both models must be used in
effect quite well, and there is only some differences at the end
order to accomplish a fair comparison between them.
of the loading section.
The key feature of these results is the incompatibility However, at temperatures over As , all models are applicable,
between the Tanaka martensite transformation temperatures and in Figure 9, they are compared with experimental test data
(Ms , Mf ) and the DSC results. It should be mentioned here at a representative temperature, T = 110°C, starting from
that although the DSC is an indication of transformation pure austenite. From this isothermal test, it can be seen that
temperatures under no stress conditions, the values of Ms and all models match the experimentally measured characteristics
Mf that have been determined, correspond more closely to of pseudoelasticity fairly closely, and there are only small
those defined in the Brinson model [8]. differences in their transformation path [8]. However, the
 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257 255

Figure 11: The stress-temperature phase diagram.

Figure 10: The experimental isostress test at σ = 100 MPa and comparison
with the mentioned models (σ ≥ σf ).
(b) Temperature decreases while stress simultaneously in-
creases.
Brinson model has a more accurate prediction. This is a simple
consequence of the fact that in all models, the changes in To use transformation kinetics in this region, a modified model
stress detected during the loading and unloading processes must satisfy the following statements:
are driven by the same equations (constitutive Eqs. (1) and 1. In all conditions: ξ ≤ 1.
(9)). Initial changes in stress within the SMA are due to the 2. If σ = σfcr then ξs = 1.
elastic component of stress, until the value of stress reaches the 3. If T = Mf then ξ = 1.
value of critical stress for direct transformation from austenite
Although after Brinson’s original model, Bekker and Brin-
to martensite. After this, the transformation from austenite to
son [19] developed a kinetics that was robust and did not per-
martensite begins and the residual strain is gradually recovered.
mit the martensite volume fraction to exceed unity, Brinson’s
Since the Tanaka and the Liang and Rogers models recognize
original model is still the most widely used formulae describing
only one type of martensite, this residual strain recovery for
the behavior of SMAs. To meet the above conditions, a modifi-
these models is associated with the martensite volume fraction,
cation of Brinson’s martensite kinetics by Chung et al. is devel-
while for the Brinson model, it is associated with an increase
oped [15]. In this formulation, Eq. (12) is revised as:
only in the stress-induced martensite. The transformation is
For T < Ms and σscr < σ < σfcr :
completed when the full residual strain is recovered and
the sample fully becomes martensite. During the unloading
 
1 − ξs0 π 1 + ξs0
process, the inverse behavior is observed. ξs = cos (σ − σf ) +
cr
,
Since it will be shown in the next section that there exists 2 σscr − σfcr 2
a region between σs and σf , which under special conditions ∆T ε
(the original kinetics proposed by Brinson [14]) gives rise to ξT = ∆T ε − (ξs − ξs0 ), (14)
1 − ξs0
a physically untenable volume fraction (ξ > 1) [15], the
experimental constant stress test data at stress above σf is where, if Mf < T < Ms and T < T0
selected for comparison with the constitutive models.
1 − ξs0 − ξT 0
The experimental isostress strain-temperature plot at σ = ∆T ε = cos[aM (T − Mf )]
100 MPa (that is, above σf ) is compared by prediction of the 2
Tanaka, Liang and Roger, and Brinson models in Figure 10. As 1 − ξs0 + ξT 0
+ ,
seen from this figure, all models match the experimentally 2
measured characteristics of the isostress test quite closely, and else,
there are only minor differences in their phase transformation
paths. ∆T ε = ξ T 0 .
As can be seen from these equations, with respect to the
5. Correction to the Brinson evolution Brinson model, only the equation of ξT has been changed,
while the equation for ξs remained unchanged. Recently, by
Figure 11 illustrates a typical phase diagram with different combination of the Brinson thermomechanical constitutive
active transformation zones. There is a region (PQRS, shown in equation of SMA wires (with the corrected evolution kinetics
Figure 11) in which simultaneous evolution of the martensite developed by Chung et al. [15]) and the nonlinear bending
twinned and detwinned volume fractions could occur. equation of a flexible beam, modeling of a large deflection
The original Brinson model [14], in the case when Mf < flexible beam actuated by two active SMA wires is performed,
T < Ms and σscr < σ < σfcr , can be used only with and the results of the proposed model were also verified,
specific initial conditions, otherwise it gives rise to a physically with respect to PC-based experimental set-up values [20]. The
inadmissible volume fraction (ξ > 1). In this region, for some experimental results showed that the proposed model could
initial conditions, this model leads to an incorrect martensite predict the behavior of the smart structure with externally
volume fraction under two conditions: attached SMA wires with moderate accuracy. It means that the
Brinson thermomechanical constitutive equation of SMA wires
(a) Temperature decreases while stress is fixed. could be used to model the behavior of structures actuated
256 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257

Table 3: Changes in the martensite volume fractions of the Brinson and the Chung et al. models for different initial conditions.

Initial Process Brinson model Chung et al. suggested Initial Process Brinson model Chung et al. suggested
condition model condition model

ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-1 ξ =1 ξ =1 Case-1 ξ =1 ξ =1
ξT = 0 ξT = 0 ξT = 0  ξT = 0 
ξs = 0 ξs = 0 ξs = 0.2 ξs = 0.2
ξs0 = 0 Case-2 ξ =1 ξ =1 ξs0 = 0.2 Case-2 ξ = 1.2 ξ =1
ξT = 1 ξT = 1 ξT = 1 ξT = 0.8
ξT 0 = 1 ξT 0 = 0.5
(inadmissible)
ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-3 ξ =1 ξ =1 Case-3 ξ = 1.5 ξ =1
ξT = 0 ξT = 0 ξT = 0.5 ξT = 0
(inadmissible)

ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-1 ξ =1 ξ =1 Case-1 ξ =1 ξ =1
ξT = 0  ξT = 0  ξT = 0 ξT = 0
ξs = 0.5 ξs = 0.5 ξs = 1 ξs = 1
ξs0 = 0.5 Case-2 ξ= ξ =1 ξs0 = 1 Case-2 ξ =2 ξ =1
ξT = 1 ξT = 0.5 ξT = 1 ξT = 0
ξT 0 = 0.2 ξT 0 = 0
1.5 (inadmissible) (inadmissible)
ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-3 ξ= ξ =1 Case-3 ξ =2 ξ =1
ξT = 0.8 ξT = 0 ξT = 1 ξT = 0
1.8 (inadmissible) (inadmissible)

ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-1 ξ =1 ξ =1 Case-1 ξ =1 ξ =1
ξT = 0  ξT = 0  ξT = 0 ξT = 0
ξs = 0.5 ξs = 0.5 ξs = 0 ξs = 0
ξs0 = 0.5 Case-2 ξ= ξ =1 ξs0 = 0 Case-2 ξ =1 ξ =1
ξT = 1 ξT = 0.5 ξT = 1 ξT = 1
ξT 0 = 0.5 ξT 0 = 0
1.5 (inadmissible)
ξs = 1 ξs = 1 ξs = 1 ξs = 1
   
Case-3 ξ= ξ =1 Case-3 ξ =2 ξ =1
ξT = 0.5 ξT = 0 ξT = 1 ξT = 0
1.5 (inadmissible) (inadmissible)

by SMA wires with great accuracy. The major benefit of using 6. Conclusions
the Brinson thermomechanical constitutive equation is in its
simplicity, as well as its proper predictions. As a result of this research work, three major models of
It is proved in [15] that all the mentioned preconditions are
SMA behavior, used in the literature for studying the static
satisfied. Table 3 summarizes the changes in the martensite
performance of SMA components, attributed to Tanaka, Liang
volume fraction of the Brinson model, and the suggested model
by Chung et al. after stress or/and temperature changes for 3 and Rogers, and Brinson, have been analyzed and compared.
cases: The major differences and similarities between these models
have also been elaborated and presented based on experimental
Case-1.
data of the shape memory and superelastic behavior, as well as
T = Ms (fixed),
the isostress test of a Ni–Ti SMA wire.
σ = σscr → σfcr . Based on the results accomplished in this paper, the
Case-2. following conclusions can be drawn:
σ = σscr (fixed),
• It has been found that the models developed by Tanaka,
T = Ms → Mf . and Liang and Rogers cannot be used for predicting the
Case-3. shape memory effect behavior of SMAs. This is due to the
T = Ms → Mf , fact that in the constitutive equations used in those models,
σ = σscr → σfcr . the transformational component is proportional to the total
martensite volume fraction. However, the model developed
As seen from the above results, the following statement can be
concluded for the region PQRS shown in Figure 11: by Brinson introduces two state variables: stress-induced
and temperature-induced martensite volume fractions. This
1. When stress increases, while the temperature is constant
(Case-1), the Brinson model, like the Chung et al. model, enables prediction of the shape memory effect in SMAs.
shows proper results for any initial conditions. • The models developed by Tanaka, Liang and Rogers, and
2. When temperature decreases, while stress is constant (Case- Brinson all agree well in their predictions of the superelastic
2), the Brinson model shows proper results, only if the behavior of SMAs at high temperatures, when the alloys stay
initial condition is temperature-induced martensite (ξs0 = in the fully austenite phase. However, the Brinson model
0, ξT 0 = 1) or pure austenite (ξs0 = 0, ξT 0 = 0). For other has more accurate predictions of the superelastic behavior
initial conditions, the Chung et al. model must be used. of SMAs.
3. When temperature decreases, while stress simultaneously
increases (Case-3), the Brinson model shows proper results,
• It was shown analytically that the original evolution kinetics
only if the initial condition is temperature-induced marten- proposed by Brinson [14] in a specified region (Mf <
site (ξs0 = 0, ξT 0 = 1). For other initial conditions, the T < Ms and σscr < σ < σfcr ) is inadmissible for some
Chung et al. model should be used. certain thermomechanical loading and initial conditions.
4. Regardless of initial conditions, the Chung et al. model Furthermore, a corrected evolution kinetics developed by
produces results of ξ ≤ 1, if σ = σfcr , then ξs = 1, and, Chung et al. [15] is reviewed that is both admissible and valid
if T = Mf , then ξ = 1. in this region.
 H. Sayyaadi et al. / Scientia Iranica, Transactions B: Mechanical Engineering 19 (2012) 249–257 257

In general, it can be recommended that for analysis of the [15] Chung, J., Heo, J. and Lee, J. ‘‘Implementation strategy for the dual
shape memory effect and superelastic behavior of SMA wires, transformation region in the Brinson SMA constitutive model’’, Smart
Materials and Structure, 16, pp. N1–N5 (2007).
the Brinson constitutive model with the corrected evolution [16] Elahinia, M.H. and Ahmadian, M. ‘‘An enhanced SMA phenomenological
kinetics developed by Chung et al. [15] should be applied. model: I. The shortcomings of the existing models’’, Journal of Smart
Material and Structure, 14, pp. 1297–1308 (2005).
[17] Brinson, L.C. and Huang, M.S. ‘‘Simplifications and comparisons of shape
References memory alloy constitutive models’’, Journal of Intelligent Material Systems
and Structures, 7, pp. 108–114 (1996).
[1] Duerig, T., Melton, K., Stockel, D. and Wayman, C., Engineering Aspects of [18] Abel, E., Luo, H., Pridham, M. and Slade, A. ‘‘Issues concerning the
Shape Memory Alloys, Butterworth–Heinemann Publishers, London (1990). measurement of transformation temperatures of NiTi alloys’’, Journal of
[2] Melton, K.R. ‘‘General applications of shape memory alloys and smart Smart Material and Structure, 13, pp. 1110–1117 (2004).
materials’’, K. Otsuka and C.M. Wayman, Eds., In Shape Memory Materials, [19] Bekker, A. and Brinson, L.C. ‘‘Phase diagram based description of the
10, pp. 220–239, Cambridge University Press, Cambridge (1999). hysteresis behavior of shape memory alloys’’, Acta Materialia., 46,
[3] Strelec, J.K., Lagoudas, D.C., Khan, M.A. and Yen, J. ‘‘Design and pp. 3649–3655 (1998).
implementation of a shape memory alloy actuated reconfigurable wing’’, [20] Zakerzadeh, M.R., Salehi, H. and Sayyaadi, H. ‘‘Modeling of a nonlinear
Journal of Intelligent Material Systems and Structures, 14, pp. 257–273 Euler–Bernoulli flexible beam actuated by two active shape memory
(2003). alloy actuators’’, Journal of Intelligent Material Systems and Structures, 22,
[4] Mani, R., Lagoudas, D. and Rediniotis, O. ‘‘MEMS based active skin pp. 1249–1268 (2011).
for turbulent drag reduction’’, Proceedings of SPIE, Smart Structures and
Materials, 5056, San Diego, CA, pp. 9–20 (2003).
[5] Otsuka, K. and Wayman, C.M., Shape Memory Materials, Cambridge
Hassan Sayyaadi received his B.S. Degree from Amirkabir University of
University Press, Cambridge (1999).
Technology, Tehran, Iran, in 1987, his M.S. Degree from Sharif University of
[6] Lagoudas, D.C., Shape Memory Alloys: Modeling and Engineering Applica-
Technology, Tehran, Iran, in 1990, and his Ph.D. Degree from the University
tions, Springer (2007).
of Tokyo, Japan, in 2001, all in Mechanical Engineering. From 1990 to 1996,
[7] Wayman, C.M. and Duerig, T.W., Engineering Aspects of Shape Memory
and since 2001, he has been with the Mechanical Engineering Department of
Alloys, Butterworth–Heinemann, Boston (1990).
Sharif University of Technology. He is now Associate Professor and his research
[8] Prahlad, H. and Chopra, I. ‘‘Comparative evaluation of shape memory alloy
interests include dynamics and control, robotics and mechanisms, artificial
constitutive models with experimental data’’, Journal of Intelligent Material
intelligence and neural networks.
Systems and Structures, 12, pp. 383–395 (2001).
[9] Boyd, J.G. and Lagoudas, D.C. ‘‘A thermodynamic constitutive model for
the shape memory materials part I. The monolithic shape memory alloys’’,
International Journal of Plasticity, 6, pp. 805–842 (1998). Mohammad Reza Zakerzadeh received his M.S. and B.S. Degrees in Mechanical
[10] Ivshin, I. and Pence, T.J. ‘‘A thermo mechanical model for one variant shape Engineering from Sharif University of Technology, Tehran, Iran, in 2005 and
memory material’’, Journal of Intelligent Material Systems and Structures, 5, 2007, respectively, where he is now a Ph.D. Degree student. He is currently
pp. 455–473 (1994). working on the modeling and control of a flexible large deflection beam by
[11] Graesser, E.J. and Cozzareli, F.A. ‘‘A proposed three dimensional constitu- Shape Memory Alloy (SMA) wire actuations for use in intelligent structures,
tive model for shape memory alloys’’, Journal of Intelligent Material Systems under the supervision of Professor Hassan Sayyaadi. His major focus is the
and Structures, 5, pp. 78–89 (1994). modeling and compensating hysteresis behavior of SMA actuators. He has
[12] Tanaka, K. ‘‘A thermomechanical sketch of shape memory effect: one- published about ten papers in this area.
dimensional tensile behavior’’, Research Mechanicanical, 18, pp. 251–263
(1986).
[13] Liang, C. and Rogers, C.A. ‘‘One-dimensional thermomechanical constitu-
tive relations for shape memory material’’, Journal of Intelligent Material Hamid Salehi is an expert in the field of aircraft structures. He received his
Systems and Structures, 1, pp. 207–234 (1990). B.S. Degree in Mechanical Engineering from Iran University of Science and
[14] Brinson, L.C. ‘‘One-dimensional constitutive behavior of shape memory Technology, Tehran, Iran, in 2005, and graduated from Sharif University of
alloys: thermomechanical derivation with non-constant material func- Technology, Tehran, Iran, with an M.S. Degree in Aerospace Engineering, in
tions’’, Journal of Intelligent Material Systems and Structures, 4, pp. 229–242 2008. He has published some papers in the fields of structural impact, composite
(1993). structures and smart structures.