Materials and Design 189 (2020) 108456

Contents lists available at ScienceDirect

Materials and Design

journal homepage: www.elsevier.com/locate/matdes

Shape optimization of SMA structures with respect to fatigue

Xiaojun Gu a, b , Yinfeng Cao b , Jihong Zhu b, *, Jun Wang a, b , Weihong Zhang b , Ziad Moumni c, b
a
Unmanned System Research Institute, Northwestern Polytechnical University, 710072 Xi’an, China
b
State IJR Center of Aerospace Design and Additive Manufacturing, Northwestern Polytechnical University, Xi’an 710072, China
c
IMSIA, ENSTA-Paris, CNRS, CEA, EDF, Institut Polytechnique de Paris, 91120 Palaiseau, France

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• A non-parametric shape optimization

approach based on optimality crite-
ria is applied to change the structural
configuration.
• Both low- and high-cycle fatigue cri-
teria are used to compute the fatigue
factor.
• The design optimization framework is
accomplished by relying on commer-
cially available software and tools.

A R T I C L E I N F O A B S T R A C T

Article history: This paper presents an efficient structural optimization approach for shape memory alloys (SMAs) with
Received 8 October 2019 respect to fatigue. In the proposed method, a nonparametric shape optimization approach based on opti-
Received in revised form 14 December 2019 mality criteria is applied to change the structural configuration. The design optimization framework is
Accepted 24 December 2019 accomplished by relying on commercially available software and tools. In addition, both low- and high-cycle
Available online 8 January 2020
fatigue criteria are used to compute the fatigue factor at each material point, indicating its degree of safeness
with respect to fatigue. Meanwhile, a 3D constitutive model is utilized to predict, with good accuracy, the
Keywords: stabilized thermomechanical stress state of a SMA structure subjected to multiaxial nonproportional cyclic
Shape memory alloys
loading. Finally, numerical examples are tested to demonstrate the effectiveness of the proposed method.
Fatigue
© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
Structural optimization
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction strains (pseudoelasticity). The unique abilities of SMAs to accommo-

Shape memory alloys (SMAs) exhibit special properties as a result
of solid-solid martensitic phase transformation, such as recover-
ing their shape by simple heating after being inelastically strained
(shape memory effect: SME), or accommodating large recoverable
* Corresponding author.
E-mail addresses: gu.xiaojun@nwpu.edu.cn (X. Gu), jh.zhu@nwpu.edu.cn (J. Zhu),
zhangwh@nwpu.edu.cn (W. Zhang), ziad.moumni@ensta-paristech.fr (Z. Moumni).
date recoverable deformation has motivated increased usage of these
materials in engineering applications ranging from biomedicine [6]
to automotive engineering [18] and aerospace industry [16,17]. In
many of these applications, SMAs are subjected to cyclic loading
and their fatigue behavior becomes highly important. A good under-
standing of fatigue in SMAs is necessary to improve the safety
of SMA structures in industrial applications. Moreover, Designing
and optimizing SMA structures against fatigue require an efficient
optimization approach as well as a reliable fatigue criterion.

https://doi.org/10.1016/j.matdes.2019.108456
0264-1275/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
2 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456
The fatigue of a SMA structure is usually named as low- or high-
cycle fatigue according to its number of cycles to failure. In low-cycle
fatigue (e.g. SMA damping control devices), the SMA undergoes a
large range of phase transformation, resulting in significant inelas-
tic strains and hysteresis energy. High-cycle fatigue (e.g. SMA stents)
of SMA has received little attention and the investigations reported
in the literature particularly focus on a shakedown-based analysis
[12,9,27,2]. To predict the fatigue lifetime, some macro fatigue cri-
teria have been proposed in both 1D and 3D. The 1D criteria are
usually proposed through rotating-bending fatigue tests on wires,
while the 3D ones are always established by using energy-based
approaches. For 1D criteria, Tobushi et al. [32] proposed a Manson-
Coffin law to model their rotating-bending fatigue results. It links
the strain amplitude to the number of cycles to failure. They also
showed that this criterion parameter values are different between
air and water and this is explained by temperature effect on phase
transformation. Song et al. [29] proposed a damage model, where
a damage variable is defined as the ratio of accumulated dissipated
energy after a prescribed number of cycles N to that obtained at
fatigue failure Nf ; the fatigue criterion is established by setting the
critical damage value to 1. The criterion can reasonably predict the
fatigue lifetimes of pseudoelastic SMAs in the load cases with various
mean stresses, peak stresses and stress amplitudes. Zheng et al. [36]
performed a series of experimental works on the fatigue behavior
of pseudoelastic NiTi plates and their results showed that fatigue
cracks initiate in the so-called active zone where SMAs undergoes
locally cyclic phase transformation. For 3D criteria, The Manson-
Coffin criterion is extended by [28] where the alternating equivalent
transformation strain is related to the number of cycles to failure.
The improved criterion can predict the multiaxial fatigue life of SMAs
and provide a satisfactory normalization for different testing modes,
including torsion, tension/tension and bending. Moumni et al. [25]
established an energy based fatigue model that predicts the fatigue
lifetime using hysteresis energy at stabilized cycles. Since the SMA
spends most of its lifetime at the stabilized state, this criterion links
the stabilized hysteresis energy to the number of cycles to failure.
Kan et al. [19] modified it by replacing the power law equation by
a logarithmic one. Based on the new definitions of equivalent stress
and non-proportionality for the non-proportional multiaxial cyclic
deformation of super-elastic NiTi SMAs, Song et al. [30] extended
their previous damage fatigue model [29] to a 3D one so that it
can be used for the prediction of multiaxial fatigue with various
non-proportional multiaxial paths.
Comparing to the study of fatigue performance of SMA materi-
als, the research and development related to the optimized design
of SMA devices have not been the subject of sufficient investiga-
tion. Using SMA wire actuators, Strelec et al. [31] designed an airfoil
considering the coupled fluid–structure interaction problem and uti-
lized a genetic algorithm (GA) based optimization tool with local
search steps. Wind tunnel tests indicated an increase in lift for a
given flow velocity and angle of attack by activating the optimized
SMA wire actuators. Langelaar et al. [21-23] proposed detailed SMA
structural design methods incorporating gradient-based optimiza-
tion, both topology and shape optimization procedures for the design
of SMA devices are described. The most influential work recently
is given by [15,26], who proposed a general design optimization
framework for SMA components and assemblies thereof actuation
applications. A structural response model with respect to concerned
design variables is proposed using design of experiments (DOE) tech-
niques, and then been used to search the desired configuration using
gradient or non-gradient based optimizer. By relying on commer-
cially available software and utilizing tools already well established
in design optimization, the proposed approach is shown to be
very efficient for analyzing morphing aerostructure. Despite these
recent developments, optimization of SMAs with respect of fatigue
was poorly investigate and virtually unaccounted for by current
optimization methods, which are not efficient on dealing with fatigue
problems.
In this paper, we focus our analyses on improving the fatigue
lifetime of pseudoelastic SMAs structures under multiaxial cyclic
loading. Particularly, How to enhance the fatigue performance by
changing the structural configuration of SMAs. An efficient approach
for fatigue design and optimization of SMA structures is proposed.
This approach include several component: the constitutive model
and the fatigue criteria and the optimizer. The constitutive model
used here in based on the 3D rate-dependent model developed in
our previous work [13], which can simulate pseudoelastic SMAs
subjected to multiaxial and cyclic loading. For low-cycle fatigue,
energy based fatigue criterion [25] is used; for high-cycle fatigue,
the shakedown-based fatigue criterion [12] for SMA is utilized here
to calculate the fatigue factor, which indicating the degree of safe-
ness with respect to high cycle fatigue. The fatigue factors at each
material point are evaluated by the optimizer, which is based on the
Optimality Criteria (OC) algorithm.
The manuscript is organized as follows: after detailed description
of the optimization method in Section 2, a brief review of the 3D
constitutive model and fatigue criteria of SMAs is given in Section 3.
Optimization cases considered low- and high-cycle fatigue are pre-
sented in Section 4 and Section 5, respectively. Conclusion is finally
provided in Section 6.
2. Optimization methods
2.1. Boundary value problem
The boundary value problem is written here for the case of
quasi-static evolution of a SMA structure occupying a volume Y∈R3
subjected to prescribed loading in the time internal [0, T]. Body forces
f(x, t) act in the domain Y. The boundary C is defined by the follow-
ing mechanical boundary components: The contact forces Td (x, t) are
applied on a static boundary Ct , displacement ud (x, t) is prescribed on
a kinematic boundary Cu , the design variable for shape optimization
is applied on a design boundary Cd . The condition Ct ∩ Cu ∩ Cd = ∅
makes Td (x, t) and ud (x, t) design-independent, meaning that they
remain unchanged with the variation of the design configuration.
Using Fourier’s law of heat conduction q(x, t) = −k∇h(x, t), the
thermal boundary conditions are defined as follows: The inner heat
generation Q(x, t) acts over Y. Heat convection conditions act on the
portion Ch of the boundary C, and the specified temperature hd (x, t) is
constrained over the remaining portion Ch such that Ch + Ch = C.
Starting from an initial configuration at time t = 0 in which the state
variables are known, the problem consists in solving the following
set of equations over Y for t ∈ [0, T]:
Static and kinematic admissibility
s ∈ K s , u ∈ Ku , (1)
where the set Ks , Ku are defined by
 
Ks = s | divs + f = 0 in Y, s • n = T d on CT ,
 
Ku = u | u = ud on Cu , (2)
and n is an outward unit vector normal to the boundary CT .
Heat equations
q ∈ Kq , h ∈ Kh , (3)
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 3

where the set Kq , Kh are defined by where n is the iteration number, D(n) is the step direction and a is
the distance along this direction. The search steps improve the sat-
 
Kq = q | −divq + Q̇ = qCp ḣ in Y, q • n = h(h − hr ) on Ch , isfaction of the objective function, or the degree of compliance with
  the constraints, or both. The iterative search procedure is terminated
Kh = h | h = hd on Ch , (4) when the solution X(n+1) satisfies the Kuhn-Tucker conditions (7).
To compute the step direction D(n) and the distance a, several
strategies are used in shape optimization. Here, a nonparametric
where hr is the ambient temperature of the surrounding environ-
shape optimization approach based on optimality criteria is applied.
ment and h is the heat transfer coefficient. n is an outward unit vector
Physical intuitive based optimality criteria postulate criteria that
normal to the boundary Ch .
describe the optimum. The redesign rule is derived from the optimal-
ity criteria and the response behavior of the system for modifications
2.2. Shape optimization problem of the design variables. The design space for so called nonparametric
shape optimization is build up by implicit parameter from a chosen
Structural optimization problems are commonly formulated as
set of surface nodes from the finite element model. As shown in Fig. 1,
constrained minimization problems. The typical shape optimization
the outer normal on the node is taken as the optimization displace-
problem is formulated as:
ment vector, while for nodes on the perimeter of the set of design
nodes the one sided normal from the adjoining element surface is
find: Cd (X), X = (x1 , x2 , . . . , xn ) taken. Once the design response, which is fatigue factor in this case,
min: f (X), is bigger the reference value, the node grow along the outer normal
on the node, otherwise, the node shrink along the normal. After the
s.t.: gj (X) ≥ 0 ( j = 1, 2, . . . , J),
design nodes relocation, mesh smoothing will be executed for all the
hk (X) = 0 (k = 1, 2, . . . , K), model.
Cd (X) ∈ Yd , (5)
2.3. Description of the optimization process and tools

where X are design variables describing the design boundary, Cd (X)

The objective of this work is the optimization design of SMA
is the optimal boundary of Cd . f(X) is the objective function, gj (X) are
structure with respect of low- and high-cycle fatigue. In this section,
inequality constraints for j = 1toJ, hk (X) are equality constraints
the detailed description of optimization process and tools utilized
for k = 1toK. Yd is the design space that constrained by the design
is presented. The framework of the optimization process can be
boundary Cd .
described at three levels:
The solution of a constrained optimization problem in Eq. (5)
can often be found by using the so-called Lagrangian method. The
• Optimization iteration, which integrates the structural
Lagrangian is defined as
responses to change the structural configuration to improve
fatigue performance.

p

m
L (X, k, g) = f (X) − kj gj (X) − gk hk (X), (6)
j=1 k=1

where kj and gk are Lagrange multipliers. In general, the Lagrangian

L (X, k, g) is the sum of the original objective function and terms that
involve Lagrange multipliers and the constraints. The Kuhn-Tucker
conditions, which are necessary condition for objective function f(X)
at solution point X to be a minimum are:

∂ f (X)  ∂ gj (X) 
p m
∂L ∂ hk (X)
= − kj − gk = 0 (i = 1, 2, . . . , n),
∂ xi ∂ xi j=1
∂ xi
k=1
∂ xi
∂L
= kj gj (X) = 0, kj ≥ 0 ( j = 1, 2, . . . , J),
∂ kj
∂L
= hk (X) = 0 (k = 1, 2, . . . , K). (7)
∂ hk

The Kuhn-Tucker conditions above only guarantee that the solution

point X is a local optimum. A sufficient condition for optimality in
the whole solution space is that the Hessian matrix of the Lagrangian
function is positive definite:

∂ 2L
sT s > 0, ∀s = 0. (8)
∂ X2

In practice, to solve the shape optimization problem in Eq. (5),

a numerical iterative search procedure is applied. It starts from an
initial configuration X(0) and proceeds in small steps:

X (n+1) = X (n) + aD(n) , (9) Fig. 1. Design nodes update strategy.

4 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

Fig. 2. Optimization process.

• Calculation of the solution of the global boundary value expansion tensor a are functions of phase composition:
problem and fatigue life for a given structural configuration.
 −1
• Calculation of the local constitutive responses at a given mate-
K = (1 − z)K −1 −1
A + zK M , (11)
rial point.

As shown in Fig. 2, the optimization tool used in this work is a = (1 − z)aA + zaM . (12)
Tosca Structure1 from Simulia. It can run any command-based anal-
ysis tools during the optimization iteration automatically. The finite The volume fraction of martensite is constrained to the interval
element analysis tool used here is Abaqus from Simulia. Combin- [0, 1], and the recoverable inelastic deformation of martensite is lim-
ing Abaqus’ user material subroutine (UMAT) interface, the struc- ited by a maximum e0 that depends on the material used. The further
tural responses of a given configuration is calculated using Abaqus. assumption of complete orientation of stress-induced martensite
The quantitative structural responses is read by the fatigue life once it forms reduces the constraint on the orientation strain to an
solver, which is written in Python. The Python script is executed by equality. These constraints can be expressed as
Abaqus/CAE without GUI, and read the structural responses from the 
2
Abaqus result ODB file. fatigue factors for every material points are z ≥ 0, 1 − z ≥ 0, and e0 − eori : eori = 0, (13)
written in ASCII files, which will be read by optimization tool Tosca 3
to change the structural configuration according to the redesign rule. The loading functions for forward and reverse phase transforma-
The new structural configuration will be the new input of the finite tion Fz1 and Fz2 and for martensite reorientation Fori can be defined
element analysis tool to start the next iteration. as
1   2

Fz1 = ElMA s : s + PMA (trs)2 + s + B : (eori + aMA Dh)
2 3
3. 3D constitutive model and fatigue criteria for SMAs

b 2
− C(h) − (G + b)z − a(1 − z) − (a − b)z + e , (14)
3.1. Constitutive model 2 0
1   2

The prediction of the fatigue lifetime requires the computation Fz2 = − ElMA s : s + PMA (trs)2 − s + B : (eori + aMA Dh)
2 3
of the asymptotic thermomechanical state (stress, strain, and dissi-

b 2
pated energy). To this end, an efficient constitutive model is needed. + C(h) + (G − b)z − a(1 − z) + (a − b)z + e , (15)
2 0
In the present global approach, the constitutive model is developed 
  
 2 2 2 
within a thermomechanically consistent framework based on the  
Fori =  s + B − 2 s + B : eori eori  − zY, (16)
principle of virtual power (PVP) in compliance with the first and
second laws of thermodynamics [33-35].
The derivation of the model gives the following stress-strain
relation:
s = K : [e − zeori − er − aDh],
where s and e are the Cauchy stress tensor and total strain tensor. er
is the residual strain tensor. The elastic stiffness tensor K and thermal
 3 3e 3 
0 VM
where ElMA , PMA , a, b, a, b, Y and aMA are material constants. C(h)
is a function of temperature. B is a measure of internal stress that
develops during cyclic training and allows unassisted transformation
of austenite into oriented martensite. The evolution of the marten-
(10)
site volume fraction z (forward and reverse phase transformation)
and martensite orientation strain tensor eori are governed by the
consistency conditions associated with the loading functions. They
can be expressed in terms of the Kuhn-Tucker conditions as

ż ≥ 0, Fz1 (s, e ori , z, h) ≤ 0, żFz1 = 0,

ż ≤ 0, Fz2 (s, e ori , z, h) ≤ 0, −żFz2 = 0,
1
Tosca Structure is a platform for structural optimization based on industry stan-
dard FEA packages (Abaqus, ANSYS, MSC Nastran, etc.). g ≥ 0, Fori (s, eori , z) ≤ 0, gFori = 0, (17)
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456
Fig. 3. Evolution of the hysteresis area with the number of cycles.
where g is positive scalar, so that ėori = Ng, and N is a vector indi-
cating the direction of the orientation strain rate. For the detailed
derivation of the model, see authors’ previous work [14].
3.2. Fatigue criteria for SMAs
3.2.1. Energy based low-cycle fatigue criterion for SMAs
Under cyclic loading, the mechanical response of pseudoelastic
SMAs stabilizes and settles into a closed stress-strain loop, after sev-
eral transient cycles, as show in Fig. 3. For low-cycle fatigue, it is
shown that the dissipated energy of the stabilized cycle is a relevant
parameter for the estimation of fatigue life [25].
To calculate low-cycle fatigue life of SMAs, [24] introduced the
effect of hydrostatic pressure based on the work of [25]. The fatigue
criterion proposed by [24] is used here:
Wd (x) + aPmax (x) = m[Nf (x)]p ,
where x is a material point in a SMA structure, Wd (x) is the hystere-
sis area of the stabilized cycle, Pmax (x) is the maximum hydrostatic
pressure of the stabilized cycle, Nf (x) is the number of cycles to fail-
ure, a, m and p are material parameters. For the tensile-compressive
results from [24], a = 0.0025, m = 5.19mJ/m3 and p = −0.2196.
The results are reported in Fig. 4.
To apply the low-cycle fatigue criteria in Eq. (18) on shape opti-
mization, a fatigue factor flow (x), which indicates the degree of
safeness with respect to low-cycle fatigue, is defined as
flow (x) = Wd (x) + aPmax (x).
3.2.2. Shakedown based high-cycle fatigue criterion for SMAs
In order to achieve high-cycle fatigue performance, the asymp-
totic macroscopic behavior of a SMA structure is expected to be
purely elastic. In this case, defects appear and are then confined
in certain grains because of stress concentration due to anisotropy
and heterogeneity of the material at this scale. The shakedown the-
ory provides a necessary condition for fatigue safety in structures
subjected to time-varying load. The elastic shakedown of a struc-
ture corresponds to a situation where energy dissipation becomes
bounded and the response elastic after a certain number of load-
ing cycles. This path-independent state, which is also independent of
the initial state of the structure, corresponds to a high-cycle fatigue
5
Fig. 4. Energy based low-cycle fatigue criteria from [24].
regime. In the authors’ previous work [12], a shakedown based
high-cycle fatigue criterion for SMAs is presented.
G(x) + āPmax (x) ≤ 1, (20)
where x is a given material point, G(x) is a fatigue factor calculated
based on the state of point x, Pmax (x) is the maximum hydrostatic
pressure at the stabilized cycle and ā is a material constants. The
proposed criterion permits to compute a fatigue factor at each mate-
rial point indicating its degree of safeness with respect to high-cycle
fatigue.
G(x) + āPmax (x) ≤ 1 ⇒ the material point at x is safe,
(21)
G(x) + āPmax (x) > 1 ⇒ the material point at x is not safe.
Detailed definition of the high-cycle fatigue criteria defined in Eq.
(18)
(20) can be found in [12]. The high cycle fatigue factor fhigh (x) defined
as
fhigh (x) = G(x) + āPmax (x), (22)
which indicates the degree of safeness with respect to high-cycle
fatigue at material point x.
4. Optimization case 1: SMA plate with a hole in the center
In this section, a center-holed SMA plate case is presented to
demonstrate the optimization process. The shape of the center hole
(19)
is optimized to enhance the low-cycle fatigue life of the plate under
uniaxial tensile-unloading condition. The geometry of the model is
shown in Fig. 5. Due to symmetry, only one quarter of the model is
used in the simulations. The plate is 100 mm × 120 mm × 15 mm
in size and contains a hole 25 mm in radius at the center. The mesh
consists of 128 hexahedral elements with linear integration (C3D8)
for a total of 747 degrees of freedom. uniaxial loading is considered,
involving vertical traction of 50 MPa. The material parameters used
for this simulation is shown in Table 1. It’s notable that the material
parameters used in this case keep constant under cyclic loading.
In this optimization case, the objective is to minimize the max-
imum low-cycle fatigue factor of the SMA plate under the vertical
traction. The design variable is the shape of the center hole. The con-
straint is the total volume of the new design cannot be larger than
6 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

0.4 40

39
0.3

)
38
0.2

low

( )
Max(
37

0.1
36

0 35
1 3 5 7 9 11 13 15
Iteration number

Fig. 6. The convergence history of the objective function (blue) and the constraint
(red). (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)

fatigue factor flow (X) decrease from 0.328 to 0.073, the distributions
of fatigue factor flow (X) before and after optimization are shown in
Fig. 7. Meanwhile, the stress also decreases, which means that less
phase transformation occurs during the loading-unloading process.
Fig. 8 shows the maximum stress at time step 0.5 decreases from
141.9 MPa to 99.4 MPa, the corresponding maximum martensite
volume fraction decreases from 0.96 to 0.16, as shown in Fig. 9.
Fig. 5. The geometry of center-holed SMA plate. After the optimization, fatigue test was carried out to validate
the optimization results. Two samples were prepared: the original
design and the optimized design. The samples was cut using pseu-
doelastic NiTi alloy sheet with the thickness of 0.6 mm. The samples
the one of the initial configuration. The shape optimization problem
size are described in Fig. 10. The experiments were carried out using
is formulated as:
a Bose Electroforce 3510-AT fatigue test machine. A Flir infrared
camera X6540 was used to scan temperature on the surface of the
find: Cd (X), X = (x1 , x2 , . . . , xn ) samples, which were coated with a matte spray paint to maximize
the emissivity. All samples were tested under stress controlled cyclic
min : max(flow (X)),
loading with a maximum force of 3000 N at 0.25 Hz. Each design was
s.t.: V(X) ≤ 37.7, repeated twice. Fig. 11a shows the samples before and after fatigue
Cd (X) ∈ Yd , (23) test. Fig. 11b shows the temperature distribution of the samples at
the 10th cycle with the maximum force. The temperature of orig-
inal design is higher than the optimized design because the phase
where flow (X) is the low-cycle fatigue factor defined in Eq. (19), V(X)
transformation zone of original design is bigger. The fatigue life of
is the total volume of the SMA plate, which is constrained to be less
the original design were 458 and 521. The fatigue life of the opti-
than the initial configuration value of 37.7 cm3 .
mized design were 3245 and 3688. The fatigue test results agree
The simulation process converged after 15 iterations. A desk-
with simulated results in terms of fatigue life comparison of the two
top equipped with Intel hexa-core Xeon processors @1.6 GHz and
design.
16 GB of RAM is used to run the simulations. The total solution
time is 6 min. The iterative curve is shown in Fig. 6. As we can see,
during the iterative process, the volume of the center hole almost 5. Optimization case 2: self-expanding SMA stent
remains constant, the shape of the hole in the center changes its
shape gradually from a circle to a “rectangle”. The maximum of the In this section, a self-expanding stent is used as the optimization
case study. Stent implantation is one of the most effective minimally
invasive treatments for the obstruction of coronary arteries [11].
Based on different deployment approaches, stents can be classified
Table 1
as two types: balloon-mounted and self-expanding stents. Balloon-
Material parameters obtained using data form [4].
mounted stents are manufactured in a crimped state and mounted
EA 45,200 MPa EM 26,400 MPa
onto a tightly-folded balloon catheter and expanded plastically and
a 0.24 MPa b 0.096 MPa
m 0.3 e0 0.08 deployed by inflating the balloon. Self-expanding stents are man-
Y 30 MPa G 0.92 MPa ufactured at a diameter slightly above the artery and are crimped
a 2500 MPa b 6250 MPa and constrained by a delivery catheter to the smaller diameter
n 0.0145 MPa K−1 j 0.038 MPa until the intended deployment position is reached, where the con-
A0f 320 K h 340 K
straint is removed and the stent deployed [8]. Nickel titanium, also
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 7

+3.28e−01
+3.12e−01
+2.95e−01
+2.79e−01
+2.62e−01
+2.45e−01
+2.29e−01
+2.12e−01
+1.96e−01
+1.79e−01
+1.63e−01
+1.46e−01
Max: 3.28e−01 +1.30e−01
+1.13e−01
+9.68e−02
+8.02e−02
+6.37e−02
+4.72e−02 Max: 7.34e−02
+3.07e−02
+1.41e−02
−2.38e−03

Fig. 7. Comparison of the low-cycle fatigue factor distribution before and after optimization.

141.886
136.561
131.235
125.910
120.585
115.260
109.934
104.609
99.284
93.959
88.634
Max: 141.9 83.308
77.983
72.658
67.333
62.008
Max: 99.43
56.682
51.357
46.032
40.707
35.382

Fig. 8. Comparison of the stress distribution at step time 0.5 before and after optimization.

known as Nitinol, exhibits pseudoelastic and biocompatible prop- minimal influence on the geometry and properties of the anatomy
erties that make it particularly suitable for self-expanding stents. [7,8,10].
Compared to a balloon-mounted stent, a Nitinol self-expanding stent Experimental evaluation of the fatigue behavior involves long
can be designed to provide constant gentle outward pressure with lasting fatigue test. Even with an accelerated fatigue test, a 400

0.962
0.913
0.865
0.816
0.767
0.719
0.670
0.622
0.573
0.524
0.476
Max: 0.962
0.427
0.379
0.330
0.281
0.233 Max: 0.162
0.184
0.136
0.087
0.038
−0.010

Fig. 9. Comparison of the martensite volume fraction distribution at step time 0.5 before and after optimization.
8 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

Fig. 10. The samples dimensions in mm: (a) the original design; (b) the optimized design.

Fig. 11. (a) the samples before and after fatigue test; (b) temperature distribution.
0.898 mm

J− K+
0.615 mm
0.08 mm

K− J+

0.589 mm

Stent
Stent
m
m
1.5 mm

Artery
08
0.
m

m
m
5m

25

Artery
1.2

0.

45 degree

Fig. 12. Stent and artery modeling: (a) dimensions of the unit cell; (b) mesh and boundary conditions of the unit cell; (c) position and mesh of stent and artery; (d) dimensions of
the stent and artery.
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 9

Fig. 13. Material parameters of Nitinol used for manufacturing stent: (a) DSC results of phase transformation temperatures; (b) stabilization of pseudoelastic loop.

million cycles test can last months. Finite element analysis (FEA) is
an extremely useful complement and has proven its effectiveness
and capacity to provide a better and more detailed understanding for
fatigue analysis and design. The aim of this section is to provide a fast
and cost-effective method to evaluate fatigue resistance of Nitinol
stent components, which is then applied to a specific stent geometry
to illustrate the effectiveness of the method.
5.1. FEM implementation
elements through the thickness direction and the total number of
elements used to mesh the unit cell is 5000. The coronary artery is
modeled as a cylinder with an outer diameter of 2.5 mm and thick-
ness of 0.5 mm. Reduced integration 8-node elements (C3D8R) are
used to mesh the artery with 5 element layers through the thickness
direction. The total number of elements used to mesh the artery is
6000. As shown in Fig. 12c and d, the expanded stent has a bigger
diameter compared to the artery, which can expand the artery after
deployment.

5.1.1. Geometric model and boundary conditions 5.1.2. Material properties

Numerous Nitinol stent designs are used in clinical practice, most
of which having two fundamental constituents: expandable ring ele-
ments and connecting elements [3]. The geometry of the stent in
this study is based on the design of self-expanding Nitinol stent
manufactured by Covidien™, shown in Fig. 12b, which can also be
found in [3]. The stent is cut from a Nitinol tube into an open lat-
tice design with “peak-to-peak” connectors. Since the stent pattern
repeats itself, only a unit cell of the stent with appropriate boundary
conditions is considered (Fig. 12b). For the boundary planes K+ and
K−, the periodic boundary conditions can be written as
K+ K−
uK+ K− K+
q = uq , u0 = u0 , and uz = uK−
z ,
where uq , u0 and uz are the displacements along q, 0 and z directions
in cylindrical coordinates, respectively. The constraints in Eq. (24)
means that a pair of corresponding nodes on plane K+ and K− have
the same displacements. The same periodic boundary conditions is
also enforced on the boundary planes J+ and J−. It is notable that the
three translational movement of one node on the stent is constrained
to avoid rigid motion. The unit cell stent model is meshed with 3
The finite element simulation of the stent is performed with the
following loading conditions: (i) Stent crimping and deployment,
whereby an initially expanded nitinol stent is compressed by pre-
scribing inward radial displacement conditions on the outer surface
at a low temperature of 263 K (− 10 ◦ C), the displacement is then
removed at the human body temperature of 310 K (37 ◦ C), which
makes the stent self-expand and come into contact with the artery
wall. (ii) Stent operation under cyclic diastolic-systolic loading.
The stent is assumed to be made of pseudoelastic Nitinol (52 Ni
%wt). The transformation temperatures of the material are deter-
mined by differential scanning calorimetry (DSC), which is shown in
Fig. 13a. Fig. 13b shows the stress-strain response under the stress-
(24) controlled cyclic tensile loading, performed at room temperature
298 K. The hysteresis loop evolves and reaches a stabilized state
when the number of cycles increases.
The material parameters used for stent simulation, which are
characterized from the uniaxial experimental data in Fig. 13b, are
listed in Table 2. Fig. 14a and b show the uniaxial stress-strain
response of the Nitinol SMA at temperatures 263 K and 310 K respec-
tively, which are the temperature conditions for stent crimping and
stent operation. It’s notable from the figures that the phase trans-
formation stress decreases when the temperature decreases. This is

Table 2
Material parameters used for finite element simulations.

EA 40,000 MPa EM 30,000 MPa

m 0.3 q 6400 kg m−3
Cp 470 J kg−1 K−1 B(sat) 100 MPa
aA 1.1 × 10−5 K−1 aM 1.1 × 10−5 K−1
k 18 W m−1 K−1 t 10
s rs 30 MPa s rf 80 MPa
(0) (sat)
s ms 455 MPa s ms 260 MPa
(0) (sat)
s mf 605 MPa s mf 470 MPa
(0) (sat)
s as 340 MPa s as 235 MPa
(0) (sat)
s af 80 MPa s af 90 MPa
c(0) 0.058 c(sat) 0.054
0 0 K−1 u 0.016
A0f 258.5 K hr 310 K
10 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

Fig. 14. Stress-strain curve at different temperatures: (a) 263 K; (b) 310 K.

helpful to crimping the stent in lower temperature, which makes the
material “softer”.
The modeling of arterial mechanical behavior is based on the
work of [5, 20], who performed arterial stiffness identification of the
human carotid artery and proposed the characterization of the aortic
wall stress-strain relationship. In particular, an isotropic “bilinear”
constitutive model consistent with the mean values of their exper-
imental measurement in the circumferential direction is used. The
relation between Young’s modulus Eartery and strain e are
160 kPa, e ≤ 0.025,
Eartery =
900 kPa, e > 0.025,
which is also implemented in Abaqus using user material subroutine
UMAT.
5.1.3. Stent crimping and deployment
First, the expanded stent is compressed by a delivery catheter
at 263 K (− 10 ◦ C), which makes the stent deform easier. A dis-
placement of 0.85 mm in the radial direction is applied on the outer
surface of the stent, resulting in a folded diameter of 1.3 mm of
the folded stent. To simplify the simulation, self-contact of the stent
during crimping is not considered here. The artery material is not
considered in analyzing stent crimping. Then, the stent is released
from the catheter inside the artery. During the deployment stage,
contact condition is applied between the outer surface of crimped
stent and the inner surface of the artery. The displacement applied
during the crimping process is then removed and the stent is allowed
to expand progressively by itself. When the self-expanding stent
reaches the inner surface of the artery, the contact is activated and
the stent continues to expand outward against the artery wall until
reaching a position in which the outward forces due to expansion
of the stent are balanced by the compressive forces exerted by the
artery wall on the stent.
(25)
5.1.4. Pulsating loading
After deployment of the stent, the stent-artery system is sub-
jected to a continuous cyclic diastolic-systolic loading at the human
body temperature of 310 K (37 ◦ C). This cyclic loading condition is
a result of physiologic blood pressure loading acting on the internal
surface of the stent and artery. The pulsatile loading ranges between
80 mmHg for diastolic pressure and 160 mmHg for systolic pres-
sure, which followed the US Food and Drug Administration (FDA)
Guidance and based on the ASTM fatigue testing standard for vas-
cular stents [1]. The cyclic loading makes the stent reach elastic
shakedown state after a certain number of cycles, In this case, ten
cycles are needed in this case.

(a) (b) 101.2E−03

(c) 101.2E−03
1.0E+00
937.5E−03 94.8E−03 94.8E−03
875.0E−03 88.5E−03 88.5E−03
812.5E−03 82.2E−03 82.2E−03
750.0E−03 75.9E−03 75.9E−03
687.5E−03 69.5E−03 69.5E−03
625.0E−03 63.2E−03 63.2E−03
562.5E−03 56.9E−03 56.9E−03
500.0E−03 50.6E−03 50.6E−03
437.5E−03 44.3E−03 44.3E−03
375.0E−03 37.9E−03 37.9E−03
312.5E−03 31.6E−03 31.6E−03
250.0E−03 25.3E−03 25.3E−03
187.5E−03 19.0E−03 19.0E−03
125.0E−03 12.6E−03 12.6E−03
62.5E−03 6.3E−03 6.3E−03
0.0E+00 0.0E+00 0.0E+00

Fig. 15. Simulated martensite volume fraction for the original stent design under: (a) fully crimped loading; (b) systolic pressure; (c) diastolic pressure.
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 11

(a) 860.37 (b) 130.02

806.60 121.90
752.82 113.79
699.05 105.68
645.28 97.56
591.50 89.45
537.73 81.33
483.96 73.22
430.18 65.11
376.41 56.99
322.64 48.88
268.87 40.77
215.09 32.65
161.32 24.54
107.55 16.42
53.77 8.31
0.00 0.20

(c) 156.74 (d) 0.92

146.95 0.85
137.17 0.78
127.38 0.72
117.60 0.65
107.81 0.58
98.03 0.52
88.24 0.45
78.46 0.38
68.68 0.31
58.89 0.25
49.11 0.18
39.32 0.11
29.54 0.04 danger zones
19.75 −0.02
9.97 −0.09
0.18 −0.16

Fig. 16. Simulated results for the original stent design: (a) stress distribution fully crimped; (b) stress distribution under systolic pressure; (c) stress distribution under diastolic
pressure; (d) high-cycle fatigue factor distribution.

5.1.5. High-cycle fatigue analysis (22), the maximum fatigue factor is 0.92. The danger zones in terms
Since the stent is found to achieve elastic shakedown, failure by of fatigue, which are the locations having large fatigue factor, are
fatigue may only take place after a large number of cycles. In order pointed out in Fig. 16d. These danger zones are the locations sub-
to evaluate the fatigue endurance of the stent, the high-cycle fatigue jected to the highest stress, also under hydrostatic tension, which
criterion defined in Eq. (22) is used. It is notable that only phase are bad for fatigue due to tensile loads assisting the opening of
transformation is considered in this case: cracks.
Shape optimization is carried out to improve high-cycle fatigue

w̄∗ performance under cyclic diastolic-systolic blood pressure loading
, 0 < z < 1,
G(x) = b̄ (26)
v̄∗ conditions. The reduced geometric model shown in Fig. 17 is used to
, z = 0.
b̄ simplify the optimization process and reduce computation time. The
objective of the optimization is to minimize the maximum high-cycle
where w̄∗ and v̄∗ are two values calculated by the fatigue criterion fatigue factor of the SMA stent under cyclic diastolic-systolic blood
at every material point. The material parameters b̄ and b̄ in the pressure loading. The constraint is the total volume of the new design
criterion are used the 30% of tensile stress values, under which the
material remains elastic in mixed-phase and pure austenite phase,
respectively. The material parameter ā is obtained from [2].
⎧
⎪
⎪
⎨b̄ = 27.6 MPa,
b̄ = 76.8 MPa, (27)
⎪
⎪
⎩ā = 0.011 MPa−1 .

The material parameters used here are for reference only. They Artery
Stent
are used to figure out which can figure out danger zones, before
proceeding to optimizing the geometry with respect to fatigue.

5.1.6. Simulation results and optimization process

The stent release, followed by operation under diastolic-systolic
blood pressure cycles is simulated using the original stent design
shown in Fig. 12. Fig. 15 shows the martensite volume fraction in the
stent at fully crimped, systolic and diastolic blood pressure, respec-
tively. As shown in Fig. 15b and c, the martensite volume fractions
are the same, which means there is no phase transformation during
the cyclic diastolic-systolic blood pressure loading. This is a neces-
sary condition for the stent to have high-cycle fatigue. Fig. 16a is
the stress distribution after being completely crimped, and the max-
imum stress is 860.37 MPa. Fig. 16b and c are the stress distributions
in the stent under systolic and diastolic blood pressure, respectively,
and the maximum stresses are 130.02 MPa and 156.74 MPa, respec- Fig. 17. The single strut reduced stent model considered for the optimization
tively. Fig. 16d is the high-cycle fatigue factor calculated using Eq. problem.
12 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

1 1.8 where fhigh (X) is the high-cycle fatigue factor defined in Eq. (22), V(X)
is the total volume of the SMA stent, which is constrained to be less
than the original design V̄.
The optimization process converged after 8 iterations. The total
solution time is 3 h 37 min. The convergence curves are shown in
0.8 1.4 Fig. 18. As we can see, during the iterative optimization process,
)

the strut geometry evolves gradually from a rectangular block to

a dogbone-like shape, meanwhile, the volume of the stent is kept
h ig h

( )
almost constant. The maximum fatigue factor fhigh (X) decreases from
Max(

0.899 to 0.571, the distributions of the fatigue factor fhigh (X) before
0.6 1 and after optimization are shown in Fig. 19, with a more uniform
distribution reached after 8 iterations. Meanwhile, the stress also
decreases, as shown in Figs. 20 and 21, the maximum stress in the
stent under systolic and diastolic blood pressure decreases from
151.3 MPa to 139.8 MPa, and from 126.3 MPa to 121.3 MPa, respec-
0.4 0.6 tively. The martensite volume fraction decreases from 0.081 to 0.016,
1 2 3 4 5 6 7 8 as shown in Fig. 22.
Iteration number A new CAD model of the stent is built according to the optimized
stent configuration illustrated in Fig. 18. The original design and new
Fig. 18. The convergence history of the objective function (blue) and the constraint design of the stent are shown in Fig. 23, It’s notable that these two
(red). designs have the same volume.
The optimized stent model is simulated under crimp-release,
cyclic diastolic-systolic blood pressure loading. Fig. 24 shows the
cannot be larger than the one of the original design. The design vari-
martensite volume fraction in the stent at fully crimped, systolic
able is the free edge of the stent in Fig. 17. The shape optimization
and diastolic blood pressure, respectively. Compare to the original
problem is formulated as:
design shown in Fig. 15, The maximum stabilized martensite volume
fraction is reduced from 0.101 to 0.025. Fig. 25a is the stress dis-
find: Cd (X), X = (x1 , x2 , . . . , xn ) tribution after being completely crimped, and the maximum stress
min : max(fhigh (X)), is 665.10 MPa Compared to the original design (860.37 MPa), it’s
reduced 22.7 %. Fig. 25b and c are the stress distribution of the
s.t.: V(X)/V̄ ≤ 1,
stent under systolic and diastolic blood pressure, respectively, and
Cd (X) ∈ Yd , (28) the maximum stress are 122.67 MPa and 139.70 MPa, respectively.

(a) 0.899 (b) 0.762

0.772 0.653
0.645 0.543
0.519 0.434
0.392 0.325
0.265 0.215
0.138 0.106
0.011 −0.004
−0.116 −0.113

(c) 0.673 (d) 0.623

0.576 0.530
0.479 0.437
0.382 0.345
0.285 0.252
0.188 0.159
0.091 0.067
−0.006 −0.026
−0.103 −0.118

(e) 0.598 (f) 0.586

0.475 0.463
0.351 0.340
0.227 0.217
0.104 0.094
−0.020 −0.029
−0.144 −0.153
−0.268 −0.276
−0.391 −0.399

(g) 0.576 (h) 0.571

0.470 0.439
0.365 0.307
0.260 0.175
0.155 0.043
0.050 −0.088
−0.055 −0.220
−0.161 −0.352
−0.266 −0.484

Fig. 19. Evolution of the high-cycle fatigue factor in iterations 1 to 8, corresponding to Figs. (a) to (h).
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 13

(a) 151.3 (b) 138.4

132.4 121.1
113.5 103.9
94.7 86.6
75.8 69.3
56.9 52.0
38.0 34.7
19.1 17.5
0.2 0.2

(c) 137.3 (d) 140.7

120.2 123.1
103.0 105.5
85.9 88.0
68.7 70.4
51.6 52.9
34.5 35.3
17.3 17.8
0.2 0.2

(e) 138.9 (f) 136.8

121.5 119.7
104.2 102.6
86.9 85.5
69.5 68.4
52.2 51.4
34.9 34.3
17.5 17.2
0.2 0.1

(g) 140.2 (h) 139.8

122.7 122.3
105.2 104.9
87.7 87.4
70.2 70.0
52.7 52.5
35.2 35.1
17.7 17.6
0.2 0.2

Fig. 20. Evolution of the von Mises stress in the stent under systolic pressure in iterations 1 to 8, corresponding to Figs. (a) to (h).

(a) 126.3 (b) 116.9

110.5 102.3
94.8 87.7
79.0 73.1
63.2 58.5
47.5 44.0
31.7 29.4
16.0 14.8
0.2 0.2

(c) 117.2 (d) 120.7

102.6 105.6
87.9 90.5
73.3 75.5
58.7 60.4
44.1 45.4
29.5 30.3
14.8 15.3
0.2 0.2

(e) 119.4 (f) 117.6

104.5 102.9
89.6 88.2
74.7 73.6
59.8 58.9
44.9 44.2
30.0 29.6
15.1 14.9
0.2 0.2

(g) 121.4 (h) 121.3

106.2 106.2
91.1 91.1
75.9 75.9
60.8 60.8
45.7 45.6
30.5 30.5
15.4 15.4
0.2 0.2

Fig. 21. Evolution of the von Mises stress in the stent under diastolic pressure in iterations 1 to 8, corresponding to Figs. (a) to (h).
14 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456

(a) 0.081 (b) 0.044

0.071 0.038
0.061 0.033
0.051 0.027
0.041 0.022
0.031 0.016
0.020 0.011
0.010 0.005
0.000 0.000

(c) 0.031 (d) 0.031

0.027 0.027
0.023 0.023
0.019 0.019
0.015 0.016
0.011 0.012
0.008 0.008
0.004 0.004
0.000 0.000

(e) 0.028 (f) 0.024

0.024 0.021
0.021 0.018
0.017 0.015
0.014 0.012
0.010 0.009
0.007 0.006
0.003 0.003
0.000 0.000

(g) 0.018 (h) 0.016

0.016 0.014
0.013 0.012
0.011 0.010
0.009 0.008
0.007 0.006
0.004 0.004
0.002 0.002
0.000 0.000

Fig. 22. Evolution of the maximum martensite volume fraction in the stent in iterations 1 to 8, corresponding to Figs. (a) to (h).

Fig. 23. Geometric comparison of the original and new stent designs.
 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456 15

(a) (b) (c)

1.0E+00 25.0E−03 25.0E−03
933.3E−03 23.4E−03 23.4E−03
866.7E−03 21.7E−03 21.7E−03
800.0E−03 20.0E−03 20.0E−03
733.3E−03 18.4E−03 18.4E−03
666.7E−03 16.7E−03 16.7E−03
600.0E−03 15.0E−03 15.0E−03
533.3E−03 13.4E−03 13.4E−03
466.7E−03 11.7E−03 11.7E−03
400.0E−03 10.0E−03 10.0E−03
333.3E−03 8.3E−03 8.3E−03
266.7E−03 6.7E−03 6.7E−03
200.0E−03 5.0E−03 5.0E−03
133.3E−03 3.3E−03 3.3E−03
66.7E−03 1.7E−03 1.7E−03
0.0E+00 0.0E+00 0.0E+00

Fig. 24. Simulated martensite volume fraction of the new stent design under: (a) fully crimped loading; (b) systolic pressure; (c) diastolic pressure.

Compared to the original design (130.02 MPa and 156.74 MPa), it is
reduced 5.7 % and 10.9 %, respectively. Fig. 25d shows the high-cycle
fatigue factor calculated using Eq. (22), the maximum fatigue factor
is 0.69, a reduction of 25 % compared to the original design.
6. Conclusion
In this paper, engineering analysis tools are utilized to develop
an efficient methodology for enhancing the fatigue lifetime of SMA
structures. Specifically, an energy based low-cycle fatigue model
and a shakedown based high-cycle fatigue model have been imple-
mented in a flexible nonparametric shape optimization framework.
Two optimization case studies are provided, one is a SMA plate with
a hole in the center, the other is a self-expanding stent. For each
case, the optimization problem is formulated. The design objective in
each case is the minimization of the maximum fatigue factor, and the
constraint is the total volume of the structure. The optimized results
for each case indicated that the new configuration improved the
fatigue behavior dramatically, compared to the original one. Further
work is needed in order to experimentally validate the optimization

(a) 665.10 (b) 122.67

620.76 114.51
576.42 106.34
532.08 98.18
487.74 90.02
443.40 81.85
399.06 73.69
354.72 65.53
310.38 57.36
266.04 49.20
221.70 41.03
177.36 32.87
133.02 24.71
88.68 16.54
44.34 8.38
0.00 0.22

(c) 139.70 (d) 0.69

130.40 0.64
121.10 0.58
111.80 0.52
102.50 0.47
93.20 0.41
83.90 0.36
74.61 0.30
65.31 0.24
56.01 0.19
46.71 0.13
37.41 0.07
28.11 0.02
18.81 −0.04
9.51 −0.09
0.21 −0.15

Fig. 25. Simulated results of the new stent design: (a) stress distribution fully crimped; (b) stress distribution under systolic pressure; (c) stress distribution under diastolic
pressure; (d) high-cycle fatigue factor distribution.
16 X. Gu, Y. Cao, J. Zhu, et al. / Materials and Design 189 (2020) 108456
procedure, which can then be utilized in designing SMA structures of
arbitrary geometry with respect to fatigue.
CRediT authorship contribution statement
Xiaojun Gu: Writing - original draft. Yinfeng Cao: Investiga-
tion. Jihong Zhu: Writing - review & editing. Jun Wang: Writing -
review & editing. Weihong Zhang: Writing - review & editing. Ziad
Moumni: Writing - review & editing.
Declaration of competing interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgments
This work is supported by National Natural Science Founda-
tion of China (11802243, 11802241), Natural Science Foundation
of Shaanxi Province (2019JQ-176), Key Project of NSFC (51790171,
51761145111, 51735005) and NSFC for Excellent Young Scholars
(11722219).
References
[1] ASTM F2477-07, Standard Test Methods for In Vitro Pulsatile Durability Testing
of Vascular Stents, ASTM International, West Conshohocken, PA, 2013.
[2] F. Auricchio, A. Constantinescu, C. Menna, G. Scalet, A shakedown analysis of
high cycle fatigue of shape memory alloys, Int. J. Fatigue 87 (2016) 112–123.
[3] C. Bonsignore, A decade of evolution in stent design, Proceedings of the Interna-
tional Conference on Shape Memory and Superelastic Technologies, 2003, pp.
519–528.
[4] C. Bouvet, S. Calloch, C. Lexcellent, A phenomenological model for pseudoelas-
ticity of shape memory alloys under multiaxial proportional and nonpropor-
tional loadings, Eur. J. Mech. A. Solids 23 (1) (2004) 37–61.
[5] A. Danpinid, J. Luo, J. Vappou, P. Terdtoon, E. Konofagou, In vivo characterization
of the aortic wall stress-strain relationship, Ultrasonics 50 (7) (2010) 654–665.
[6] T. Duerig, A. Pelton, D. Stöckel, An overview of nitinol medical applications,
Mater. Sci. Eng. A 273-275 (1999) 149–160.
[7] T. Duerig, D. Tolomeo, M. Wholey, An overview of superelastic stent design,
Minim. Invasive Ther. Allied Technol. 9 (3-4) (2000) 235–246.
[8] T. Duerig, M. Wholey, A comparison of balloon- and self-expanding stents,
Minim. Invasive Ther. Allied Technol. 11 (4) (2002) 173–178.
[9] X. Feng, Q. Sun, Shakedown analysis of shape memory alloy structures, Int. J.
Plast. 23 (2) (2007) 183–206.
[10] X. Gong, A. Pelton, T. Duerig, N. Rebelo, K. Perry, Finite element analysis
and experimental evaluation of superelastic nitinol stent, SMST-2003 Pro-
ceedings of the International Conference on Shape Memory and Superelastic
Technologies, 2003, pp. 453–462. Pacific Grove, CA.
[11] L. Gu, S. Zhao, S. Froemming, Arterial wall mechanics and clinical implications
after coronary stenting: comparisons of three stent designs, Int. J. Appl. Mech.
04 (02) (2012) 1250013.
[12] X. Gu, Z. Moumni, W. Zaki, W. Zhang, Shakedown based model for high-cycle
fatigue of shape memory alloys, Smart Mater. Struct. 25 (11) (2016) 115012.
[13] X. Gu, W. Zaki, Z. Moumni, W. Zhang, Implementation of the large time
increment method for the simulation of pseudoelastic shape memory alloys,
ASME 2015 Conference on Smart Materials, Adaptive Structures and Intel-
ligent Systems, American Society of Mechanical Engineers. 2015, pages
V001T03A015–V001T03A015.
[14] X. Gu, W. Zhang, W. Zaki, Z. Moumni, An extended thermomechanically cou-
pled 3d rate-dependent model for pseudoelastic smas under cyclic loading,
Smart Mater. Struct. (2017).
[15] D. Hartl, D. Lagoudas, F. Calkins, Advanced methods for the analysis design and
optimization of sma-based aerostructures, Smart Mater. Struct. 20 (9) (2011)
094006.
[16] D. Hartl, D. Lagoudas, F. Calkins, J. Mabe, Use of a ni60ti shape memory alloy for
active jet engine chevron application: I. Thermomechanical characterization,
Smart Mater. Struct. 19 (1) (2009a) 015020.
[17] D. Hartl, J. Mooney, D. Lagoudas, F. Calkins, J. Mabe, Use of a ni60ti shape mem-
ory alloy for active jet engine chevron application: II. Experimentally validated
numerical analysis, Smart Mater. Struct. 19 (1) (2009b) 015021.
[18] J. Jani, M. Leary, A. Subic, M. Gibson, A review of shape memory alloy research
applications and opportunities, Mater. Des. 56 (2014) 1078–1113.
[19] Q. Kan, G. Kang, W. Yan, Y. Dong, C. Yu, An energy-based fatigue failure model
for super-elastic niti alloys under pure mechanical cyclic loading, Third Inter-
national Conference on Smart Materials and Nanotechnology in Engineering,
8409, 2012, pp. 84090F.
[20] T. Khamdaeng, J. Luo, J. Vappou, P. Terdtoon, E. Konofagou, Arterial stiffness
identification of the human carotid artery using the stress-strain relationship
in vivo, Ultrasonics 52 (3) (2012) 402–411.
[21] M. Langelaar, F. van Keulen, Design optimization of shape memory alloy active
structures using the r-phase transformation, The 14th International Sympo-
sium on: Smart Structures and Materials & Nondestructive Evaluation and
Health Monitoring, International Society for Optics and Photonics. 2007, pages
65250W–65250W.
[22] M. Langelaar, F. van Keulen, Modeling of shape memory alloy shells for design
optimization, Comput. Struct. 86 (9) (2008a) 955–963.
[23] M. Langelaar, F. van Keulen, Sensitivity analysis of shape memory alloy shells,
Comput. Struct. 86 (9) (2008b) 964–976.
[24] C. Morin, A comprehensive approach for fatigue analysis of shape memory
alloys,PhD thesis. Ecole Polytechnique. 2011.
[25] Z. Moumni, A. Van Herpen, P. Riberty, Fatigue analysis of shape memory alloys:
energy approach, Smart Mater. Struct. 14 (5) (2005) S287.
[26] S. Oehler, D. Hartl, R. Lopez, R. Malak, D. Lagoudas, Design optimization and
uncertainty analysis of sma morphing structures, Smart Mater. Struct. 21 (9)
(2012) 094016.
[27] M. Peigney, On shakedown of shape memory alloys structures, Ann. Solid
Struct. Mech. (2014) 17–28.
[28] A. Runciman, D. Xu, A. Pelton, R. Ritchie, An equivalent strain/coffin-manson
approach to multiaxial fatigue and life prediction in superelastic nitinol medi-
cal devices, Biomaterials 32 (22) (2011) 4987–4993.
[29] D. Song, G. Kang, Q. Kan, C. Yu, C. Zhang, Damage-based life prediction model
for uniaxial low-cycle stress fatigue of super-elastic niti shape memory alloy
microtubes, Smart Mater. Struct. 24 (8) (2015) 085007.
[30] D. Song, G. Kang, C. Yu, Q. Kan, C. Zhang, Non-proportional multiaxial fatigue of
super-elastic niti shape memory alloy micro-tubes: damage evolution law and
life-prediction model, Int. J. Mech. Sci. 131-132 (2017) 325–333.
[31] J. Strelec, D. Lagoudas, M. Khan, J. Yen, Design and implementation of a shape
memory alloy actuated reconfigurable airfoil, J. Intell. Mater. Syst. Struct. 14
(4-5) (2003) 257–273.
[32] H. Tobushi, T. Hachisuka, T. Hashimoto, S. Yamada, Cyclic deformation and
fatigue of a tini shape-memory alloy wire subjected to rotating bending, J. Eng.
Mater. Technol. 120 (1) (1998) 64–70.
[33] J. Wang, Z. Moumni, W. Zhang, A thermomechanically coupled finite-strain
constitutive model for cyclic pseudoelasticity of polycrystalline shape memory
alloys, Int. J. Plast. 97 (2017) 194–221.
[34] J. Wang, Z. Moumni, W. Zhang, Y. Xu, W. Zaki, A 3d finite-strain-based constitu-
tive model for shape memory alloys accounting for thermomechanical coupling
and martensite reorientation, Smart Mater. Struct. 26 (6) (2017) 065006.
[35] J. Wang, Z. Moumni, W. Zhang, W. Zaki, A thermomechanically coupled finite
deformation constitutive model for shape memory alloys based on hencky
strain, Int. J. Eng. Sci. 117 (2017) 51–77.
[36] L. Zheng, Y. He, Z. Moumni, Investigation on fatigue behaviors of niti polycrys-
talline strips under stress-controlled tension via in-situ macro-band observa-
tion, Int. J. Plast. 90 (2017) 116–145.