Hysteretic Shell Finite Element

A. N. Moysidis 1 and V. K. Koumousis, M.ASCE 2

Abstract: A hysteretic shell finite element for the nonlinear, static, and dynamic analysis of structures is presented, formulated on the basis
of classical theory of plasticity and finite deformation. The generalized smooth, rate-independent three-dimensional (3D) Bouc-Wen model
is expressed in tensorial form incorporating the von Mises yield criterion and different types of nonlinear hardening laws. Based on this
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

approach, a hysteretic shell finite element is derived in which the shell is considered as a number of fully bonded layers along the thickness.
The elastic mixed interpolation of tensorial components with nine nodes (MITC9) element is extended by considering as additional hysteretic
degrees of freedom the plastic strains, backstresses, and the variable yield stress. These are considered at the Gauss points of two faces and
all interlaminar interfaces, the evolution of which is described by Bouc-Wen-type equations. Using this formulation, the effect of the non-
linear hardening on the response of a shell structure and in particular the phenomenon of ratcheting is investigated. The developed hysteretic
shell element accounts for geometric nonlinear analysis and incorporates two constituent functionally graded materials. Numerical results are
presented, demonstrating the efficacy, accuracy, and generality of the proposed approach. DOI: 10.1061/(ASCE)EM.1943-7889.0001589.
© 2019 American Society of Civil Engineers.
Author keywords: Inelastic analysis; Hysteretic shell finite elements; Bouc-Wen model; Nonlinear kinematic hardening; Ratcheting;
Geometric nonlinearity; Functionally graded materials.

Introduction new MITC4+ shell finite element that performs reliably even when
the mesh undergoes large displacements and significant distortions
Systematic work has been undertaken to develop reliable and com- during the response. Ko et al. (2017b) also presented a comparison
putationally effective shell finite elements suitable for the analysis of the performance of MITC shell elements in several benchmark
of all different types of shell structures. The seminal isoparametric problems. Additional applications involving the MITC9 shell
shell element of Ahmad et al. (1970) constitutes the basis of most element can be found in the works of Cinefra and Valvano (2016),
modern shell finite-element formulations. The displacement field Carrera et al. (2016), Cinefra et al. (2015, 2016), and Wisniewski
inside this shell element is interpolated from three displacement and Panasz (2013).
degrees of freedom (DOFs) and two rotation DOFs at each node, In all aforementioned works, the kinematic assumption is
incorporating the Reissner and Mindlin theory, determining the made that the thickness remains invariable during deformation,
so-called degenerated solid shell elements. Subsequently, the linear which is not valid when finite strains are considered. Hughes and
formulation was extended to material and geometric nonlinear
Carnoy (1983), Simo et al. (1990), Bischoff and Ramm (1997), and
analysis (Bathe and Bolourchi 1980; Hughes and Liu 1981).
Toscano and Dvorkin (2007) made significant contributions in the
Because shells elements are susceptible to the detrimental effects
field of shell analysis accounting for finite strains. Moreover, other
of shear and membrane locking, remarkable effort has been under-
approaches addressing the inelastic behavior of the shell structures
taken to alleviate these phenomena. Various techniques have been
have been presented by Pinsky and Jang (1988) and Schimmels and
suggested, such as (selective) reduced integration (Pawsey and
Clough 1971; Zienkiewicz et al. 1971), the assumed strain stabi- Palazotto (1994). Another method, which is based on the multiple-
lization procedure (Belytschko et al. 1989), and use of an enhanced quadrature eight-node hexahedral finite element for large deforma-
interpolation of the transverse shear strains and membrane strains in tion analysis of Liu et al. (1998), was presented by Masud et al.
the natural coordinate system (Huang and Hinton 1986). (2000), where a continuum-based shear-deformable finite-element
The mixed interpolation of tensorial components (MITC) ap- formulation for geometrically nonlinear analysis of thick layered
proach proposed by Bathe and Dvorkin (1986) and Bucalem and composite shells was developed, and the corotational kinematic
Bathe (1993) for plate and shell elements constitutes an alternative framework for large rotation analysis was employed. This method
efficient method for treating shear and membrane locking. The de- was extended to incorporate elastoplastic models (Masud and
velopment of enhanced MITC elements presents a continuing in- Tham 2000) and damage models (Tham et al. 2005).
terest (Bathe et al. 2003). Recently, Ko et al. (2017a) proposed a Furthermore, shell finite elements are of major importance in
composites, and great interest has been expressed in analyzing the
1
Ph.D. Candidate, Institute of Structural Analysis and Aseismic behavior of functionally graded material (FGM) shell structures in
Research, National Technical Univ. of Athens, Zografou Campus, Athens different applications (Arciniega and Reddy 2007; Chakravorty et al.
15780, Greece. Email: [email protected] 1998; Masud and Panahandeh 1999; Nayak and Bandyopadhyay
2
Professor, Institute of Structural Analysis and Aseismic Research, 2005; Woo and Meguid 2001).
National Technical Univ. of Athens, Zografou Campus, Athens 15780, In plasticity problems, when a material is loaded beyond its elas-
Greece (corresponding author). Email: [email protected]
tic limit, evolution of plastic strains is initiated. Smooth hysteretic
Note. This manuscript was submitted on February 5, 2018; approved on
October 3, 2018; published online on February 25, 2019. Discussion period models, such as the Bouc-Wen model (Sivaselvan and Reinhorn
open until July 25, 2019; separate discussions must be submitted for indi- 2000; Wen 1976), are able to encapsulate all material elastoplastic
vidual papers. This paper is part of the Journal of Engineering Mechanics, phases under cyclic loading in rate tensorial form. Compared with
© ASCE, ISSN 0733-9399. classical plasticity, smooth hysteretic modeling is very efficient

© ASCE 04019027-1 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 with respect to modeling various hysteretic behaviors, including the where the dot indicates the derivative with respect to time λ̇ is the
effects of damage-induced phenomena such as stiffness degradation, rate of plastic multiplier; Qðσij Þ is the plastic potential function,
strength deterioration, and pinching (Charalampakis and Koumousis which in associative plasticity coincides with the yield function
2009; Kottari et al. 2014). Qðσij Þ ¼ Φðσij Þ (Crisfield 1991)
In this work, the MITC9 shell element is extended to handle
mainly the inelastic response of shell structures following a more ∂Φðσij Þ ∂ΦðσÞ
physical formulation by incorporating a smooth hysteresis model ε̇pl
ij ¼ λ̇ or ε̇pl ¼ λ̇ ð3Þ
∂σij ∂σ
modified to also incorporate geometrical nonlinear analysis. More
specifically, additional hysteretic degrees of freedom are introduced
that control the plastic strains, backstress components, and variable A complementarity condition holds, i.e., λ̇ · Φ ¼ 0, which re-
yield stress. These are defined at every Gauss point of the faces and quires either the yield function or λ̇ to be zero, and there is no plas-
interfaces of the layers into which the thickness of the shell is sub- tic flow. Apart from the evolution of the plastic strain, an evolution
divided, and Bouc-Wen-type evolution equations describe their of the yield surface may be present in the form of hardening, and
thus the yield surface undergoes expansion, i.e., isotropic and/or
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

behavior. The Bouc-Wen evolution equations are derived based

on the classical theory of plasticity in a consistent way, constituting translation, i.e., kinematic hardening
an additional set of first-order nonlinear differential equations that,
together with the equations of dynamic or static equilibrium, de- Φ ¼ Φðσij ; αij ; kÞ ¼ fðσij − αij Þ − k ¼ 0 ð4Þ
termine the system of equations that fully describes the inelastic
response. The main advantage of the proposed method relies on the where k = variable yield stress; and α = tensorial backstress kin-
fact that the elastic predictor/return mapping algorithm is embodied ematic hardening parameter that represents the evolution of the
into the first-order nonlinear differential evolution equations of center of the yield surface in the stress space. The backstress
Bouc-Wen type, which directly account for inelasticity in a natural evolves as a function of the plastic multiplier λ̇ and the kinematic
way, constraining the solution to stay always on the yield surface hardening function G
and satisfy the flow rule by definition per se.
The remaining work is organized as follows. In the “Inelastic α̇ij ¼ λ̇Gij or α̇ ¼ λ̇G or fα̇g ¼ λ̇fGg ð5Þ
Behavior and Bouc-Wen Hysteretic Model” section, the basic ðtensor components notationÞ ðtensor notationÞ ðmatrix-vector notationÞ
notions and relations of associative plasticity are presented, and
the hysteretic Bouc-Wen model is derived accounting for Prager’s
The isotropic hardening is defined as follows (Dafalias et al.
kinematic hardening or Armstrong-Frederick (AF) nonlinear hard-
2008b):
ening, Chaboche additive decomposition scheme, and the multipli-
cative AF kinematic hardening law, which are briefly presented in
compatible form that fits the hysteretic formulation. In the “MITC9 k̇ ¼ ck ðks − kÞε̇ps ð6Þ
Shell Element” section, the main features and steps of MITC9 shell
element formulation are presented. In the “Material Nonlinearity: where ks = saturation limit of k; ck = parameter controlling the rate
Elastic and Hysteretic Matrix” and “Material and Geometric Non- of evolution of k toward ks ; and the equivalent plastic strain is given
linearity: Tangent Matrix” sections, the elastic stiffness, hysteretic, Z Z
and tangent matrices are derived. In the “Solution Procedure” sec- X
εps ¼ δεps ¼ dεps ¼ ε̇ps dt ð7Þ
tion, the system of equations is presented for dynamic analysis.
Finally, numerical results are presented that justify the validity and
accuracy of the proposed formulation, underlining the generality This is accumulated from the equivalent plastic strain rates
of the proposed approach and its applicability to a wide range of
engineering problems. Ratcheting and wrinkling of tubes and the pffiffiffiffiffiffiffiffi
2=3ðε̇pl ∶ε̇pl Þ1=2 ¼ λ̇BðσÞ;
ε̇ps ¼
analysis of functionally graded shells are treated in particular.  
pffiffiffiffiffiffiffiffi ∂ΦðσÞ ∂ΦðσÞ 1=2
where BðσÞ ¼ 2=3 ∶ ð8Þ
∂σ ∂σ
Inelastic Behavior and Bouc-Wen Hysteretic Model
In the elastic range, materials exhibit no internal damage, thus By substituting Eq. (8) in Eq. (6), the following relation is
returning to zero stress-strain when unloaded. In stress space, the derived:
elastic domain is delimited by an external boundary i.e., the yield
surface that is defined by a yield function of the form k̇ ¼ ck ðks − kÞλ̇BðσÞ ¼ λ̇F; F ¼ ck ðks − kÞBðσÞ ð9Þ
Φðσij ; σy Þ ¼ fðσij Þ − σy ¼ 0 ð1Þ
where isotropic hardening evolves as a function of the plastic multi-
where σy = yield stress, whereas any admissible stress state must plier λ̇ and isotropic hardening function F. For the case of von
satisfy the condition Φðσij ; σy Þ ≤ 0. Mises yield criterion, BðσÞ ¼ 1 (Crisfield 1991).
Loading further the material, plastic yielding (or plastic flow), In addition, the total strain tensor is considered as the sum of an
pl
i.e., evolution of plastic strains, is initiated, manifested as perma- elastic component εel ij and a plastic component εij (assumption of
nent strains at unloading (de Souza Neto et al. 2008). This is additive decomposition), and thus
described by the plastic flow rule
pl
εij ¼ εel
ij þ εij or ε ¼ εel þ εpl ð10Þ
∂Qðσij Þ ∂QðσÞ ∂QðfσgÞ
ε̇pl
ij ¼ λ̇ or ε̇ ¼ λ̇
pl
or fε̇pl g ¼ λ̇
∂σij ∂σ ∂fσg Furthermore, the stress increment is linearly related to the elastic
ðtensor components notationÞ ðtensor notationÞ ðmatrix-vector notationÞ
strain increment in the plastic region and can be expressed by the
ð2Þ following constitutive relation (Khan and Huang 1995):

© ASCE 04019027-2 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 pl  T
σ̇ij ¼ Cijkl ε̇el
kl ¼ Cijkl ðε̇kl − ε̇kl Þ or σ̇ ¼ C∶ε̇ ¼ C∶ðε̇ − ε̇ Þ or
el pl
∂Φ
λ̇ ¼ H1 H2 λ1 ½Cfε̇g ð16Þ
fσ̇g ¼ ½Cfε̇el g ¼ ½Cðfε̇g − fε̇pl gÞ ð11Þ ∂fσg

To derive the Bouc-Wen relations, the two Heaviside functions

where C ¼ Cijkl ei ⊗ ej ⊗ ek ⊗ el is the fourth-order isotropic
are smoothed using the following expressions (Fig. 1):
elasticity tensor; and ½C = two-dimensional array in which the
 
components Cijkl are stored.  fðσij − αij ÞN
For plastic flow to occur, the stresses must remain on the yield H 1 ¼   ;
 N≥2 ð17Þ
k
surface (consistency condition) and hence (Simo and Hughes 1998)
    and because there is no plastic deformation during unloading,
∂Φ T ∂Φ T ∂Φ i.e., λ̇ ¼ 0, then
Φ̇ ¼ dfσg þ dfαg þ dk ¼ 0 or
∂fσg ∂fαg ∂k     ∂Φ T
∂Φ ∂Φ ∂Φ ∂Φ T 1 þ sgn ∂fσg fσ̇g
Φ̇ ¼ ∶dσ þ ∶dα þ dk ¼ 0 ð12Þ H2 ¼ H fσ̇g ¼
∂fσg 2
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

∂σ ∂α ∂k
 T 
∂Φ
In order to find λ̇, Eq. (11) is premultiplied by the flow vector ≈ β þ γsgn ½Cðfε̇g − fε̇pl gÞ
∂fσg
f∂Φ=∂fσggT , and using Eqs. (3), (5), (9), and (12), one obtains    
∂Φ T ∂Φ
  ¼ β þ γsgn ½C fε̇g − λ̇
∂Φ T ∂fσg ∂fσg
λ̇ ¼ λ1 · ½Cfε̇g ð13Þ  T 
∂fσg
∂Φ
¼ β þ γsgn ½Cfε̇g ð18Þ
where ∂fσg

      −1 Usually, the values β ¼ 1=2 and γ ¼ 1=2 are considered,

∂Φ T ∂Φ ∂Φ T ∂Φ
λ1 ¼ − fGg − Fþ ½C but in general, the parameters should comply with the follow-
∂fαg ∂k ∂fσg ∂fσg ing restriction to yield a thermodynamically admissible model
ð14Þ (Charalampakis and Koumousis 2009; Erlicher and Point 2004):

Relation Eq. (13) holds only when yielding has occurred. Thus, −β ≤ γ ≤ β ð19Þ
by introducing the following Heaviside type functions
Finally, using Eqs. (3) and (16), the following Bouc-Wen model
  for the evolution of plastic strains is derived:
1; Φ¼0 1; ð∂Φ=∂σÞ∶dσ ≥ 0
H 1 ðΦÞ ¼ ; H 2 ðΦÞ ¼      
0; Φ<0 0; ð∂Φ=∂σÞ∶dσ < 0  fðσij − αij ÞN ∂Φ T
fε̇pl g ¼   β þ γsgn
 ½Cfε̇g ½Rfε̇g
ð15Þ k ∂fσg
|fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
H1 H2
a single relation is established for the plastic multiplier in the whole ð20Þ
stress space (Fig. 1), which constitutes the main intervention of
smooth hysteretic models, and more specifically the Bouc-Wen where the interaction matrix ½R is expressed
model (Bouc 1967; Wen 1976)   
∂Φ ∂Φ T
½R ¼ λ1 ½C ð21Þ
∂fσg ∂fσg

and determines the necessary interrelations between the plastic

strain components to ensure that the stresses remain on the yield
surface, accounting also for the hardening law (consistency condi-
tion). By substituting Eq. (16) in Eq. (5) the following Bouc-Wen
model for the evolution of backstress is derived:
     
 fðσij − αij ÞN ∂Φ T
fα̇g ¼   β þ γsign
 ½Cfε̇g
k ∂fσg
|fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
H1 H2
   
∂Φ T
× λ1 fGg ½C fε̇g ð22Þ
∂fσg

Similarly, by substituting Eq. (16) in Eq. (6), the evolutionary

equation of Bouc-Wen type for the isotropic hardening is obtained
     
 fðσij − αij ÞN ∂Φ T

k̇ ¼   β þ γsign ½Cfε̇g
k  ∂fσg
|fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
H1 H2
 T
∂Φ
Fig. 1. Heaviside-type functions controlling the inelastic behavior. × λ1 F ½Cfε̇g ð23Þ
∂fσg

© ASCE 04019027-3 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 In this rate form, the Bouc-Wen model for the plastic strains and nonlinear kinematic and isotropic hardening can incorporate any yield
criterion and hardening law. In the case of von Mises yield criterion, the following expression is used:
1
fðσij Þ ¼ pffiffiffi ½ðσxx − σyy Þ2 þ ðσyy − σzz Þ2 þ ðσxx − σzz Þ2 þ 6ðτ 2xy þ τ 2yz þ τ 2xz Þ1=2 ð24Þ
2

8 9 8 9
Kinematic hardening is accounted for by incorporating various >
< >
x
= X
9 >
< i>
x
= X9
t ~i
hardening laws: y ¼ N i ðξ; ηÞ yi þ N i ðξ; ηÞζ i V ð30Þ
• linear Prager’s law (Crisfield 1997) > >
: ; i¼1 >
: ; > i¼1
2 3
z zi mid
 
2 ∂Φ
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

α̇ ¼ λ̇ hP ¼ λ̇G ð25Þ ~ 3 is obtained by dividing the vector v̂i connecting the upper
i
3 ∂σ where V 3
and lower points by its scalar length ti . The Jacobian operator
• Armstrong-Frederick model (Frederick and Armstrong 1966) relating the natural coordinate derivatives to the local coordinate
  derivatives is defined appropriately (Bathe 1996).
2 ∂Φ The strains in the direction normal to the midsurface are as-
α̇ ¼ λ̇ hAF − cAF BðσÞα ¼ λ̇G ð26Þ
3 ∂σ sumed to be negligible; thus, the displacement field throughout the
element will be defined by the three Cartesian components of the
• Chaboche additive decomposition scheme (Chaboche 1989, ith midsurface node displacement and the increment in the director
1991) vector V~ i3. Furthermore, assuming that the director vectors remain
! straight during the deformation process and the thickness remains
XM
2 i ∂Φ
α̇ ¼ λ̇ h − cCh BðσÞαi
i
¼ λ̇G ð27Þ constant during the deformation process, the displacement field is
i¼1
3 Ch ∂σ given by
8 9 8 9
• multiplicative AF kinematic hardening law (Dafalias et al. >
< >
u
= X
9 >
< i>
u
= X
9
2008a, b) t ~i
v ¼ N i ðξ; ηÞ vi þ N i ðξ; ηÞζ i ΔV 3 ð31Þ
> >
: ; i¼1 > >
: ; i¼1 2
X
4 w wi
α̇ ¼ λ̇ Gi ¼ λ̇G; α̇4 ¼ λ̇G4
i¼1
where u, v, and w = displacements in the directions of the global
2 ∂Φ i x-, y-, and z-axes. In addition, for linear kinematic description,
Gi ¼ hiD − cD BðσÞαi ði ¼ 1;2;3Þ
3 ∂σ i.e., infinitesimal displacements, rotations, and strains, the incre-
 
2 h4D ∂Φ 4 4 4  ∂Φ ment in the director vector V~ i3 is defined by two rotations of the
G4 ¼ W 1 − BðσÞα4 ; W 1 ¼ cD þ hD − cD α4 ∶
3 c4D ∂σ ∂σ vector, αi and β i , about orthogonal directions normal to it, V ~ i1
2 ∂Φ 4 and V ~ 2 , as follows:
i
G4 ¼ h4 − cD BðσÞα4 ð28Þ
3 D ∂σ
ΔV
i ~ 2 þ βiV
~ 3 ¼ −αi V i ~1 i
ð32Þ

MITC9 Shell Element and with this, Eq. (31) can be expressed (Ahmad et al. 1970)
Consider the shell element in Fig. 2 and let ξ, η be two curvilinear
coordinates in the middle surface of the shell and ζ be a linear co-
ordinate in the thickness direction. Assume further that ξ, η, and
ζ vary between −1 and þ1 on the respective faces of the element.
Thus, one can write a relationship between the Cartesian coordi-
nates of any point of the shell and the curvilinear coordinates in
the form (Ahmad et al. 1970)
8 9 8 9
>
< >
x
= X > xi >
9
ð1 þ ζÞ < =
y ¼ N i ðξ; ηÞ yi
>
: > ; i¼1 2 > : > ;
z zi top
8 9
> xi >
X 9
ð1 − ζÞ < =
þ N i ðξ; ηÞ yi ð29Þ
i¼1
2 > : > ;
zi bottom

where N i ðξ; ηÞ = standard interpolation functions (Bathe 1996)

used in isoparametric formulation. Eq. (29) can be rewritten in a
~ i3, the midsurface co-
form specified by the unit director vector V
Fig. 2. Nine-node MITC shell element.
ordinates, and ti as follows:

© ASCE 04019027-4 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 ~ i1 ð0Þ ¼ t V
in interval t ∈ ½0; 1 with initial conditions at t ¼ 0 V ~ i1 ,
V~ 2 ð0Þ ¼ V
i t ~ 2 , and V
i ~ 3 ð0Þ ¼ V
i t ~ 3.
i

If θ1 ¼ αi t and θ2 ¼ β i t, the ODE system in Eq. (38) is refor-

mulated herein as follows:
d ~i ~ i3 g
fV g ¼ −β i fV
dt 1
d ~i ~ i3 g
fV g ¼ α i fV
dt 2
d ~i ~ i2 g þ β i fV
~ i1 g
fV g ¼ −αi fV ð39Þ
dt 3
The analytical solution of Eq. (39) results in
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

tþΔt ~ i i ~ 1 g þ A12 f V
~ 1 ð1Þg ¼ A11 f V t i ~ 2 g þ A13 f V
t ~ 3g
i t i
f V1 g ¼ fV
tþΔt ~ i ~ i1 g þ A22 ft V
~ i2 ð1Þg ¼ A21 ft V ~ i2 g þ A23 ft V
~ i3 g
f V2 g ¼ fV
tþΔt ~ i i ~ 1 g þ A32 f V
~ 3 ð1Þg ¼ A31 f V t i ~ 2 g þ A33 f V
t ~ 3g
i t i
f V3 g ¼ fV ð40Þ

Fig. 3. Finite displacements and rotations of a shell element. where

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
α2i þ β 2i cos α2i þ β 2i
A11 ¼
8 9 8 9 α2i þ β 2i
>
< >
u
= X
9 >
< i>
u
= X
9   pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ti αi β 2i þ α2i cos α2i þ β 2i
v ¼ N i ðξ; ηÞ vi þ ~ ~
N i ðξ; ηÞζ ½ −V 2 ; V 1 
i i
A22 ¼
> >
: ; i¼1 > >
: ; i¼1 2 βi α2i þ β 2i
w wi  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
αi β i 1 − cos α2i þ β 2i
ð33Þ A12 ¼ A21 ¼
α2i þ β 2i
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
In the case of finite displacements and rotations, within the step αi sin α2i þ β 2i
t to t þ Δt, at any node i (Fig. 3), the following relation holds: A23 ¼ −A32 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
α2i þ β 2i
tþΔt ~ i t ~i pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V3 ¼ ½tþΔtt Ri  V 3 ð34Þ β i sin α2i þ β 2i
A13 ¼ −A31 ¼ − pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where ½tþΔtt Ri  is a rotation tensor. The rotation defined in Eq. (34) α2i þ β 2i
t ~ i tþΔt ~ i
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
is a rotation around an axis normal to the plane ð V 3; V 3 Þ; A33 ¼ cos α2i þ β 2i ð41Þ
t ~i t ~i
therefore, the rotation axis is in the plane ð V 1 ; V 2 Þ and is
given by The second-order approximation of the increment in the director
vector is given by the second-order Taylor series expansion of
θ~ ¼ αi V
t ~i t ~i
1 þ βi V2 ¼ f φ χ ψg ð35Þ either Eq. (34) or (40)
The rotation tensor is obtained by Argyris (1982) tþΔt ~ i t ~i t ~i t ~i 1 2 2 t ~i
V3 − V 3 ¼ −αi V 2 þ β i V 1 − ðαi þ β i Þ V 3 þ : : : ð42Þ
2 3 2
1 0 0  
6 7 sinðθÞ 1 sinðθ=2Þ 2 2 Hence, the second-order approximation of the increment in the
½ t Ri  ¼ 4 0 1 0 5 þ
tþΔt
½S þ ½S ð36Þ
θ 2 θ=2 ~ i3 is defined
director vector V
0 0 1
~3 ¼ i tþΔt ~ i 0 ~i
ΔV V3 − V 3
where
tþΔt ~ i t ~i t ~i 0 ~i
2 3 ¼ V3 − V 3 þ V3 − V3
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0 −ψ χ
6 7 t ~i 0 ~i t ~i t ~i 1 2 2 t ~i
θ¼ φ2 þ χ2 þ ψ2 and ½S ¼ 4 ψ 0 −φ 5 ð37Þ ¼ð V 3 − V 3 Þ − αi V 2 þ β i V 1 − ðαi þ β i Þ V 3 þ :: : ð43Þ
2
−χ φ 0 0
~ 3 is the director vector in the initial configuration.
i
where V
t ~i The important feature of the MITC9 shell element formu-
Alternatively, the increment in the director vector V 3 can be
evaluated solving the following system of ordinary differential lation is the use of a mixed interpolation scheme to render the re-
equations (ODE): sulting element relatively distortion-insensitive (Bathe 1996; Bathe
and Dvorkin 1986). In the natural coordinate system of the shell
d ~i dθ ~ i element, the covariant base vectors are defined
fV g ¼ − 2 fV 3g
dt 1 dt
∂fxg ∂fxg ∂fxg
d ~i dθ ~ i fgξ g ¼ ; fgη g ¼ ; fgz g ¼ ð44Þ
fV g ¼ 1 fV 3g ∂ξ ∂η ∂ζ
dt 2 dt
d ~i dθ ~ i
fV g ¼ − 1 fV
dθ2 ~ i where fxg is the vector of coordinates of coordinates f x y z gT
2g þ fV 1 g ð38Þ
dt 3 dt dt [Eq. (30)]. In the natural system, the strain tensor at any point inside

© ASCE 04019027-5 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 the element is expressed in terms of covariant tensor components
and contravariant base vectors
ε ¼ ε~ ij gi gj ; i; j ¼ ξ; η; ζ ð45Þ

where the tilde indicates that the tensor components are in the con-
vected coordinate system. The covariant and contravariant base
vectors are in general not of unit length and satisfy the following
relationship:
gi gj ¼ δ ij ; i; j ¼ ξ; η; ζ ð46Þ

where δij = Kronecker delta (δ ij ¼ 1 for i ¼ j and δij ¼ 0 for

i ≠ j). Eq. (45) can also be expressed
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

ε ¼ ε~ ξξ gξ gξ þ ε~ ηη gη gη þ ε~ ξη ðgξ gη þ gη gξ Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
in-layer strains

þ ε~ ξζ ðgξ gζ þ gζ gξ Þ þ ε~ ηζ ðgη gζ þ gζ gη Þ ð47Þ

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
transverse strains

The Cartesian components of the strain tensor εkl fek g ⊗ fel g

are given by
εkl ¼ ε~ mn ðfgm gT fek gÞðfgn gT fel gÞ ð48Þ
Fig. 4. Tying points used for the strain components of the MITC9 shell
The evaluation of these covariant components is achieved by element.
using the following relation for the strain components in terms of
the base vectors:
   T  
1 ∂fxg ∂fug ∂fxg The interpolation scheme used for the evaluation of the covariant
ε~ ij ¼ þ strain component ε~ ij ðξ; η; ζÞ at any point inside the element is
2 ∂ξ i ∂ξ i ∂ξ j
   T  
∂fug ∂fxg ∂fxg X
n
þ − ε~ ij ðξ; η; ζÞ ¼ hk ðξ; ηÞ~εkij ðξ; η; ζÞ ð53Þ
∂ξ j ∂ξ i ∂ξ j
k¼1
ξ i ¼ ξ; η; ζ; i; j ¼ ξ; η; ζ ð49Þ
where hk ðξ; ηÞ = appropriate interpolation functions; and
Eq. (49) can be rewritten in the following form: ε~ kij ðξ; η; ζÞ = covariant strain component evaluated using Eq. (50)
        at tying point k by direct interpolation using the finite-element
1 ∂fug T ∂fxg ∂fug T ∂fxg
ε~ ij ¼ þ displacement/rotation assumptions (Fig. 4).
2 ∂ξ i ∂ξ j ∂ξ j ∂ξ i The end result is a strain-displacement relation in the following
   
1 ∂fug T ∂fug form:
þ
2 ∂ξ i ∂ξ j fεg ¼ ½BL  fdg ð54Þ
|{z} |{z} |{z}
ξ i ¼ ξ; η; ζ; i; j ¼ ξ; η; ζ ð50Þ ð6×1Þ ð6×45Þ ð45×1Þ

Therefore, the linear components of the strains in the convected where

coordinate system are fdg ¼ f fd1 g : : : fd9 g g; fdi g ¼ f ui vi wi αi β i g ð55Þ
   
1 ∂fug T ∂fug T
ε~ ij jlinear ¼ fgj g þ fgi g The stress-strain law should satisfy the shell assumption that the
2 ∂ξ i ∂ξ j
stress normal to the shell surface is negligible; thus, the stress in the
ξ i ¼ ξ; η; ζ; i; j ¼ ξ; η; ζ ð51Þ direction of vector V ~ i3 should be set to zero, with the elastic con-
stitutive matrix given
and the nonlinear components are
0 2 31
    1 v 0 0 0 0
1 ∂fug T ∂fug B 6 7C
ε~ ij jnonlinear ¼ ξ i ¼ ξ; η; ζ; i; j ¼ ξ; η; ζ B 6ν 7C
2 ∂ξ i ∂ξ j B 6 1 0 0 0 0 7C
B 6 7C
ð52Þ B 60 0 0 0 0 0 7C
B 6 7C
B E 6 1−ν 7C
½C ¼ ½QT B 60
B1 − ν 2 6
0 0 0 0 7C½Q
7C
To avoid membrane and shear locking, the in-layer strains
B 6 2 7C
and transverse shear strains are interpolated independently with B 6 1−ν 7C
B 60 0 0 0 cs 0 7 C
appropriately selected interpolations. Then, these interpolations are B 6 2 7C
B 6 7C
tied to the usual displacement interpolations (Bathe et al. 2003; @ 4 1 − ν 5A
Bucalem and Bathe 1993). The in-layer strain interpolation yields 0 0 0 0 0 cs
2
the membrane and bending response of the element, and the trans-
verse shear strain interpolation gives the transverse shear one. ð56Þ

© ASCE 04019027-6 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 where cs = shear correction factor of 5=6; and ½Q is the transfor-
mation matrix (Bathe 1996) that transforms the stress-strain law
from a ξ, η, ζ Cartesian shell-aligned system to the global Cartesian
coordinate system.

Material Nonlinearity: Elastic and Hysteretic Matrix Gauss-Point

layer n-1
The elastic deformation field is extended by introducing an addi- layer 2
tional vector of hysteretic degrees of freedom, which herein are the layer 1
plastic strains at Gauss points of all layer interfaces (Owen and interface n
Hinton 1980) interface 2
interface 1
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

( 
interface 1 interface 2 interface n Fig. 5. Layered MITC9 shell element.
fzgT ¼ zfflfflfflffl}|fflfflfflffl{ zfflfflfflffl}|fflfflfflffl{ zfflfflfflffl}|fflfflfflffl{ ; : : : ;
fεpl gT1,1 ; fεpl gT1,2 ; · · · fεpl gT1;n
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Gauss point 1
( 
 ) interface 1 interface 2 interface n
interface 1 interface 2
zfflfflfflffl}|fflfflfflffl{ zfflfflfflffl}|fflfflfflffl{
interface n
zfflfflfflffl}|fflfflfflffl{ fzα g ¼
T zfflfflffl}|fflfflffl{ zfflffl}|fflffl{ zfflffl}|fflffl{ ; : : : ;
ð57Þ fαgT1,1 ; fαgT1;2 ; · · · fαg1;n
fεpl gT9,1 ; fεpl gT9,2 ; · · · fεpl gT9;n |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Gauss point 1
Gauss point 9
 interface 1 interface 2 interface n
)
:::; zfflfflffl }|fflfflffl { zfflffl}|fflffl{ zfflffl}|fflffl{ ð60Þ
fαgT9,1 ; fαgT9;2 ; · · · fαgT9;n
where fεpl gi;j = plastic component of the strain vector at the ith |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Gauss point of the jth interface (Fig. 5). At this point, a hysteretic Gauss point 9
linear interpolation field ½N pl  can be considered utilizing appropri-
ate shape functions so that and
( 
interface 1 interface 2 interface n
fzk gT ¼ z}|{ z}|{ z}|{ ; : : : ;
fεpl g ¼ ½N pl  fzg ð58Þ k1 , 1 ; k1,2 ; · · · k1;n
6×1 6×ð54nÞ ð54nÞ×1 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Gauss point 1
 )
interface 1 interface 2 interface n
where n − 1 = total number of layers. :::; z}|{ z}|{ z}|{ ð61Þ
k9,1 ; k9 , 2 ; · · · k9;n
For simplicity reasons, this interpolation can be reduced further, |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
and the plastic component of the strain vector [Eq. (57)] at the ith Gauss point 9
Gauss point of the jth interface may be expressed
The expression for the principle of virtual work in this case is
written
½α ½β½γ ½δ½ε ½ζ Z 1Z 1Z 1
2 3 fε̄gT fσg det½Jdξdηdζ ¼ fd̄gT fPex g ð62Þ
0 0 1 0 0 0 0 0 0 0
−1 −1 −1
6 7
60 0 0 1 0 0 0 0 0 07
6 7 where det½J = determinant of the Jacobian operator; fd̄g = virtual
6 7
fεpl gi;j ¼6 0 0 0 0 1 0 0 0 0 0 7 fzg ¼ ½N pl i;j fzg displacements; fε̄g = corresponding virtual strains; and fPex g is the
6 7
60 ··· 0 0 0 0 1 0 0 0 ··· 07 external nodal loading vector. The principle of virtual work can be
6 7
6 7 alternatively formulated as follows:
60 0 0 0 0 0 1 0 0 07
4 5 Z 1Z 1Z 1
0 0 0 0 0 0 0 1 0 0 fε¯~ gT fσg
~ det½Jdξdηdζ ¼ fd̄gT fPex g ð63Þ
6×ð54nÞ −1 −1 −1

α ¼ 1; β ¼ ð6nÞði − 1Þ þ 6ðj − 1Þ where fε¯~ g = covariant components of the virtual strain; and fσg
~ =
contravariant components of the stress, given by
γ ¼ ð6nÞði − 1Þ þ 6ðj − 1Þ þ 1;
δ ¼ ð6nÞði − 1Þ þ 6ðj − 1Þ þ 6 σ~ ij ¼ σmn ðfem gT fgi gÞðfen gT fgj gÞ ð64Þ
ε ¼ ð6nÞði − 1Þ þ 6ðj − 1Þ þ 7; ζ ¼ 54n ð59Þ where σmn = Cartesian components of the stress tensor
σmn fem g ⊗ fen g. By means of Eq. (11), the principle of virtual
work [Eq. (62)] is expressed
which extracts identically the corresponding plastic strains for the Z 1Z 1Z 1
vector fzg and can be directly used in the numerical evaluations of fε̄gT ½Cðfεg − fεpl gÞ det½Jdξdηdζ ¼ fd̄gT fPex g
the matrix integrals in the finite-element calculations at the Gauss −1 −1 −1
points. In a similar way, additional vectors fza g and fzk g contain- ð65Þ
ing the backstresses fαg [Eq. (5)] and the variable yield stress k
[Eq. (6)] at Gauss points of all layer interfaces, respectively, can be Substituting relation Eqs. (57) and (58) into relation Eq. (65),
defined the following expression is obtained:

© ASCE 04019027-7 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Z Z Z Z Z
1 1 1
fd̄g T
½BL  ½Cð½BL fdg − ½N pl fzgÞ det½Jdξdηdζ
T Sij δEij dV ¼ δW or S~ ij δ E~ ij dV ¼ δW ð71Þ
−1 −1 −1 V V

¼ fd̄gT fPex g ð66Þ Furthermore

   ~ T
or ∂Eij T ∂ Eij
δEij ¼ fδdg or δ E~ ij ¼ fδdg ð72Þ
Z 1Z 1
Z 1 ∂fdg ∂fdg
½BL T ½C½BL fdg det½Jdξdηdζ
−1 −1 −1 and assuming that the loading is deformation-independent, i.e., con-
Z 1Z 1 Z 1
centrated nodal loading that does not change direction as a function
− ½BL T ½C½N pl fzg det½Jdξdηdζ ¼ fPex g ð67Þ of the deformations, then
−1 −1 −1

δW ¼ fδdgT fPex g ð73Þ

Finally, the following constitutive equation that incorporates
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

plasticity is obtained at the element level: Substituting Eq. (72) into Eq. (71) the following relations are
( ) derived:
h i fdg
½k  −½k
½ke fdg − ½kh fzg ¼ ð45×45Þ ð45×ð54nÞÞ
e h  ¼ fPex g ð68Þ Z  
fzg ∂Eij
fδdgT Sij dV ¼ fδdgT fPex g or
V ∂fdg
where n − 1 = total number of layers into which the thickness of the Z  ~ 
~ ij ∂ Eij
shell is subdivided; ½ke  = MITC9 elastic stiffness matrix; and ½kh  is fδdg T S dV ¼ fδdgT fPex g ð74Þ
V ∂fdg
the herein introduced hysteretic stiffness matrix. In the elastic stiff-
ness matrix ½ke , the shear correction factor of 5/6 has been taken Hence, equilibrium is established when
into account in the integration along the thickness. The number of
columns of the hysteretic matrix ½kh  corresponds to six compo- fRg ¼ fPin g − fPex g ¼ f0g ð75Þ
nents of strain at each of the nine Gauss points for the n interfaces
of the thickness of the shell. where fRg is the residual force vector; and fPin g is the internal
The additional unknown vector fzg, containing all plastic strains force vector given by
at all Gauss points of all interfaces, follows an evolutionary equa- Z   Z  ~ 
∂Eij ∂ Eij
tion of Bouc-Wen type given in relation Eq. (20) independently for fPin g ¼ Sij dV or fPin g ¼ S~ ij dV
every six-component plastic strain vector at every particular Gauss V ∂fdg V ∂fdg
point. From the aforementioned, it becomes evident that the pro- ð76Þ
posed formulation can be used also for other types of elements that
directly incorporate hysteretic behavior (Moysidis and Koumousis Applying a truncated Taylor series expansion to Eq. (75), a new
2015; Sofianos and Koumousis 2017; Triantafyllou and Koumousis residual force vector fRn g is determined
2014).
∂fRg
fRn g ¼ fRo g þ fδdg
∂fdg
Material and Geometric Nonlinearity: Tangent Matrix ∂fPin g
¼ fRo g þ fδdg ¼ fRo g þ ½kt fδdg ¼ f0g ð77Þ
∂fdg
In the case of finite deformations (Fig. 3), the principle of virtual
work is written (Bonet and Wood 2008) where fRo g is the old residual force vector; and ½kt  is the tangent
Z stiffness matrix of the shell element. By means of Eq. (76) the
S∶δEdV ¼ δW ð69Þ tangent stiffness matrix is evaluated
V Z   
∂fPin g ∂ ∂Eij
½kt  ¼ ¼ Sij dV or
where S is the second Piola-Kirchhoff stress tensor; E is the Green- ∂fdg V ∂fdg ∂fdg
Lagrange strain tensor [Eq. (45)]; V = volume of the element in the Z   ~ 
initial un-deformed configuration; and δW = virtual work. The in- ∂ ∂ Eij
½kt  ¼ S~ ij dV ð78Þ
ner product S∶δE is defined as Sij δEij . The principle of virtual V ∂fdg ∂fdg
work can be equivalently defined in terms of the covariant compo-
nents of the virtual Green-Lagrange strain tensor and the contravar- The components of the matrix ½kt  are
iant components of the second Piola-Kirchhoff stress tensor as Z   
∂ ∂Eij
follows (Dvorkin 1995): ½kt kl ¼ Sij dV or
V ∂dl ∂dk
Z Z   ~ 
~ EdV
S∶δ ~ ¼ δW ð70Þ ∂ ∂ Eij
½kt kl ¼ S~ ij dV ð79Þ
V V ∂d l ∂dk
~ E
The inner product S∶δ ~ is defined as S~ ij δ E~ ij . Because the shell The integrands of Eq. (79) can be chain-differentiated to obtain
can undergo large displacements and large rotations and the con-   
stitutive relations are nonlinear, Eq. (69) or (70) cannot be solved ∂ ∂Eij ∂Sij ∂Eij ∂ 2 Eij
Sij ¼ þ Sij or
directly. Thus, an approximate solution is obtained by referring all ∂dl ∂dk ∂dl ∂dk ∂dk ∂dl
variables to a previously calculated known equilibrium configura-   ~ 
∂ ∂ Eij ∂ S~ ij ∂ E~ ij ~ ij ∂ 2 E~ ij
tion and using the linearized version of the virtual work equation, S~ ij ¼ þS ð80Þ
which is rewritten ∂dl ∂dk ∂dl ∂dk ∂dk ∂dl

© ASCE 04019027-8 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 where Eq. (80) utilizes the Einstein (summation) convention on the ½M S fÜg þ ½CS fU̇g þ ½K S fUg − ½HS fZg ¼ fPS ðtÞg
i and j indices. The term ∂Sij =∂dl of Eq. (80) needs to be evalu-
ated, which can be chain-differentiated to obtain fŻg ¼ ff 1 ðfUg; fU̇g; fZg; fAg; fKgÞg

∂Sij ∂Sij ∂Ers ∂E fȦg ¼ ff2 ðfUg; fU̇g; fZg; fAg; fKgÞg
¼ ¼ ðCt Þijrs rs ð81Þ
∂dl ∂Ers ∂dl ∂dl fK̇g ¼ ff3 ðfUg; fU̇g; fZg; fAg; fKgÞg ð88Þ

where Ct is the tangential modular tensor. In the case of elastoplas- subject to the initial conditions
tic material, under the assumption of large displacements and large
rotations but small strains, the plasticity material model described fUð0Þg ¼ fUg0 ; fU̇ð0Þg ¼ fU̇g0 ; fZð0Þg ¼ fZg0 ;
in the previous paragraphs can be directly employed by simply sub-
fAð0Þg ¼ fAg0 ; fKð0Þg ¼ fKg0 ð89Þ
stituting the second Piola-Kirchhoff stresses and Green-Lagrange
strains for the small displacement engineering stress and strain where fZg is the ðnz × 1Þ vector of hysteretic degrees of freedom,
measures, respectively (Bathe 1996). Within the framework of which contains the ð54nint × 1Þ vectors fzg of all elements of the
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Bouc-Wen hysteresis model, the tangent module Ct is obtained in structure; fAg is the ðnα × 1Þ vector that contains the ð54nint × 1Þ
matrix-vector notation by combining Eqs. (11) and (20) vectors fzα g of all elements of the structure; fKg is the ðnk × 1Þ
vector that contains the ð9nint × 1Þ vectors fzk g of all elements of
fṠg ¼ ½CðfĖg − H 1 H 2 ½RfĖgÞ ¼ ð½C − H1 H2 ½C½RÞfĖg ð82Þ
the structure; and the functions ff1 g, ff2 g, and ff3 g contain the
evolutionary equations of Bouc-Wen type given in Eqs. (20), (22),
Hence
and (23), respectively, governing the evolution of all the hysteretic
½Ct  ¼ ½C − H1 H2 ½C½R ð83Þ variables at the Gauss points of the two faces and all interlaminar
interfaces.
Combining Eqs. (79)–(81), the tangent matrix in the case of The evolution equations are structured in uncoupled sets for
material and geometric nonlinearity is evaluated each Gauss point at every layer for all the elements of structure
Z   (Argyris et al. 1979). This is reflected in the inner structure of the
∂E ∂Eij ~ ij ∂ 2 E~ ij hysteretic matrix. Proceeding from Eq. (68) at the element level, the
½kt kl ¼ ðCt Þijrs rs þS dV ð84Þ
V ∂dl ∂dk ∂dk ∂dl hysteretic forces along the direction of the element’s degrees of
freedom are determined (Moysidis and Koumousis 2015)
where summation on the i and j indices is assumed.
ffih g ¼ ½kih  fzi g ð90Þ
45×54nint

Solution Procedure The element contributions can be raised to the structural level by
augmenting the preceding relation as follows:
The elemental stiffness ½ke  and hysteretic matrices ½kh  derived us-
ing Eq. (68) and the tangent matrices ½kt  derived using Eq. (84) are fFih g ¼ ½Hi  fzi g ð91Þ
assembled to form the structural stiffness matrix ½K S , structural nf ×54nint
hysteretic matrix ½H S , and structural tangent matrix ½K t . For a
shell structure with nf degrees of freedom, ðnint − 1Þ is the number where only 54 of the nf rows of the matrix ½Hi  are nonzero. Thus,
of layers into which the shell is subdivided, nelem is the number the total hysteretic force results in
of the shell elements of the structure, nz ¼ nelem × ð54 × nint Þ,
X
nel
na ¼ nelem × ð54 × nint Þ, and nk ¼ nelem × ð9 × nint Þ, the equation fFh g ¼ ½H S fZg ¼ ½H i fzi g ð92Þ
of motion is expressed i¼1

½M S fÜg þ ½CS fU̇g þ fPin g ¼ fPS ðtÞg ð85Þ where

where ½M S  and ½CS  are the mass and viscous damping square sym- ½H S  ¼ ½ ½H1  ½H 2  ::: ½H nelem   ð93Þ
metric ðnf × nf Þ matrices of the structure, respectively; fPin g is the
This reveals the uncoupled contribution of the elements at the
ðnf × 1Þ internal force vector, which depends on the deformation of
structural level acting in hysteretic manner within the system indi-
the shell structure; and fPS ðtÞg is the ðnf × 1Þ vector of external
vidually, underlining the inherent parallelism of the formulation.
forces. For the inelastic case, when the nonlinear effects of large
Therefore the hysteretic matrix of the structure ½HS  is formed
displacements and rotations are not taken into account, the expres-
in a concatenated manner from the individual element contributions
sion for the internal force is simplified [Eq. (68)]
participating in a constant form in the linear equilibrium Eq. (88).
fPin g ¼ ½K S fUg − ½HS fZg ð86Þ Time-stepping methods are needed to determine the solution in
increments at time instances ti following Newmark’s method
and the equation of motion is expressed (Chopra 2011)
½M S fÜg þ ½CS fU̇g þ ½K S fUg − ½HS fZg ¼ fPS ðtÞg ð87Þ ½M S fΔÜgi þ ½CS fΔU̇gi þ ½K S fΔUgi ¼ fΔPS gi þ ½H S fΔZgi
where ½HS  is the global hysteretic ðnf × nz Þ matrix of the structure. fΔÜgi ¼ fÜgiþ1 − fÜgi
The linear equations of motion [Eq. (87)] depend on hysteretic de- fΔU̇gi ¼ fU̇giþ1 − fU̇gi
grees of freedom, which follow the nonlinear evolution equations
[Eqs. (20), (22), and (23)]. Thus, the set of Eq. (87), together with fΔUgi ¼ fUgiþ1 − fUgi
the evolution equations of plastic strains [Eq. (20)] and kinematic fΔPS gi ¼ fPS ðtiþ1 Þg − fPS ðti Þg
[Eq. (22)] and isotropic [Eq. (23)] hardening, are needed to estab-
lish the solution fΔZgi ¼ fZgiþ1 − fZgi ð94Þ

© ASCE 04019027-9 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 These are further set into the following form: correction is strictly numerical to achieve the desired accuracy,
whereas in a radial return scheme, the predictor step is linear and
½K̂fΔUgi ¼ fΔP̂gi all inelastic considerations are in the corrector step.
fŻg ¼ ff 1 ðfUg; fU̇g; fZg; fAg; fKgÞg The proposed formulation provides the nodal displacements,
elastic and plastic strains, stresses, backstresses, and the variable
fȦg ¼ ff2 ðfUg; fU̇g; fZg; fAg; fKgÞg yield stress at all Gauss points of the two faces and all interlaminar
fK̇g ¼ ff3 ðfUg; fU̇g; fZg; fAg; fKgÞg ð95Þ interfaces that satisfy the inelastic constitutive relations and equi-
librium without any additional iterative process. Therefore, by solv-
where ing in steps the system of equations of motion and the evolution
equations, the scheme always stays on the yield function and sat-
γ 1 isfies the flow rule by default. Hence, the local iterations of radial
½K̂ ¼ ½K S  þ ½CS  þ ½M S  ð96Þ
βΔt βðΔtÞ2 return or backward Euler method are avoided in solving for the
equilibrium of the system.
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

fΔP̂gi ¼ ðfΔPS gi þ ½H S fΔZgi Þ þ ½afU̇gi þ ½bfÜgi ð97Þ

and Numerical Examples

 
1 γ 1 γ A FORTRAN code was developed to implement the proposed
½a ¼ ½M  þ ½CS ; ½b ¼ ½M  þ Δt − 1 ½CS 
βΔt S β 2β S 2β formulation and comparisons were made using Abaqus version
ð98Þ 2017 code.

All matrices, ½M S , ½CS , ½K S , and ½HS , are evaluated only Example 1: Axially Loaded Tube
once at the beginning of the analysis procedure and are stored
in skyline or sparse form accompanied by appropriate renumbering In this example, a pressurized tube thick enough to deform plasti-
of the degrees of freedom [Cuthill-McKee (CM) algorithm or re- cally is subjected to axial compression and buckles into an axisym-
verse Cuthill-McKee (RCM) algorithm] (George and Liu 1981), if metric wrinkling mode. The wrinkle amplitude is initially small,
needed. Thus, less computer memory for the storage of the matrices but gradually the axial rigidity is reduced and eventually a limit
is required and the number of arithmetic operations is minimized load instability appears, which is followed by collapse (Hassan and
when the LDLT decomposition is carried out. The solution pro- Kyriakides 1992; Jiao and Kyriakides 2009, 2011a, b). The problem
cedure for the inelastic case is presented in the Appendix. consists of a cylindrical shell with midsurface radius R ¼ 0.0276 m
This solution procedure has certain advantages compared with and length 2L ¼ 0.2151 m (Fig. 6). The axially loaded cylindrical
the standard schemes and results into a significant reduction in com- shells are extremely imperfection-sensitive structures; thus the
puting time. The main benefit comes from Steps 2.6.1 and 2.6.4 following two different types of imperfections are introduced:
in the Appendix, which directly determine solutions that satisfy the • initial axisymmetric imperfection
equilibrium. This avoids the laborious evaluation of the nodal inter-    
nal forces of standard methods, which correspond to the element πz πz
ΔR ¼ t ω0 þ ω1 cos cos ð99Þ
stresses, employing numerical integration over the element volume Nλ λ
in every inner step. In addition, there is no need to calculate the
consistent tangent stiffness matrix of the structure, ½K t , as in stan- where ω0 and ω1 = axisymmetric imperfection amplitudes
dard Newton-Raphson schemes, which is a computationally expen- with value 0.13%; 2λ = wavelength of the wrinkling buckling
sive procedure. A drawback of the proposed scheme is that it does mode under monotonic compression with value λ ¼ 0.5565R ¼
not attain the quadratic convergence of Newton-Raphson method, 0.0154 m; and N ¼ 7 (2L ¼ 2Nλ).
requiring more but much faster iterations for the same accuracy.
Nevertheless, dynamic problems usually require small time steps
and thus few iterations per step.
Y
The Bouc-Wen evolution equations in relations Eq. (88) can be
solved in every step of Newmark method using any type of inte-
grator scheme for systems of first-order nonlinear differential equa-
tions such as the Livermore family of solvers (Radhakrishnan and Z
Hindmarsh 1993), allowing for robust and unconditionally stable
solutions. Plastic flow rules are incremental in nature, and standard
finite-element method (FEM) solution procedures have to follow
small equilibrium steps to trace their path. Accuracy is needed to
0.02m 2 L = 0.2151m 0.02m
integrate the flow rule within an increment and to keep the solution
on the yield surface. The solution is finally equilibrated only at the Y
end of each increment after a number of equilibrium iterations. R = 0.0276m
Therefore, classical elastoplastic solution procedures are based t0 = 0.0020m
on incremental predictor/corrector schemes such as the backward X
Euler scheme (Simo and Hughes 1998; de Souza Neto et al. 2008),
which eventually accumulate some error. This is attributed to errors
in the integration of the flow rule and their relation to the complete
Fig. 6. Geometry and imperfections of the axially loaded cylindrical
incremental/iterative solution procedure. A basic advantage of the
shell. Thickness ¼ ×4, axisymmetric imperfection ¼ ×400, and thick-
proposed method is that it accounts for the inherent inelasticity,
ness imperfection ¼ ×300.
i.e., the flow rule and consistency condition after yielding and the

© ASCE 04019027-10 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Axial Force Table 1. Multiplicative AF kinematic hardening rule parameters
Parameter Value
h1 (MPa) 413,685
c1 20,000
Node A
h2 (MPa) 22,256
c2 400
h3 (MPa) 3,137
c3 10
h4 (MPa) 99,284
c4 1,800
h4 800
c4 5,000
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Node B

120

100

axial force (kN)

80

60 Bouc-Wen - p=5500.0kPa
40 Abaqus - p=5500.0kPa
uz = 0 Bouc-Wen - p=0.0kPa
20
Abaqus - p=0.0kPa
Fig. 7. Finite-element mesh and boundary conditions.
0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
displacement of point A (m)
• linearly varying thickness imperfection Fig. 8. Chaboche additive decomposition scheme with four
  components.
Δt 2jzj
tðzÞ ¼ t0 1 − 1− ; −L≤z≤L ð100Þ
t0 L

where t0 ¼ 0.0020 m = average thickness; and Δt ¼ 1.778 × 100

10−5 m = amplitude of the variation. 80
The tube is discretized into 28 × 46 nine-node MITC9 shell 60
40
axial force (kN)

elements, and the computational model consists of eight layers

and 5,208 nodes (Fig. 7). The material parameters are E ¼ 20
181.332 GPa, ν ¼ 0.3, and σy ¼ 129.621 MPa, where E is the 0
Young’s modulus, ν is the Poisson’s ratio, and σy is the yield stress. -20
In addition, the material follows the von Mises yield criterion. -40
Two cases are examined. In the first case, the internal pressure -60
is zero (p ¼ 0.0 kN=m2 ), whereas in the second case, the internal -80 Bouc-Wen - p=5500.0kPa
Bouc-Wen - p=0.0kPa
Abaqus - p=5500.0kPa
Abaqus - p=0.0kPa
pressure is considered as p ¼ 5,500.0 kN=m2 ½≈ 0.5855ðσy t=RÞ. -100
0 0.001 0.002 0.003 0.004 0.005
The proposed hysteretic model is verified by comparing with displacement of point A (m)
Abaqus for a material following the Chaboche additive decompo-
sition scheme with four components and the parameters h1 , c1 , h2 , Fig. 9. Chaboche additive decomposition scheme with three
c2 , h3 , c3 , h4 , and c4 given in Table 1. In Fig. 8, the response of components.
the tube is plotted in terms of vertical displacement at Node A
versus the applied axial force for the two cases of internal pressure
(throughout this example, compressive load and deformation are
positive). The postbuckling response is obtained following the arc- 140
1
length method, and the plastic strains and backstress components 120 2 3
are computed by evolving the Bouc-Wen equations. It is evident 4
axial force (kN)

100
that the internal pressure leads to reduction of the limit load.
Moreover, the proposed hysteretic model is compared with 80
Abaqus for a material following the Chaboche additive decompo- 60 Bouc-Wen - p=0.0kPa
sition scheme with three components and the parameters h1 , c1 , h2 ,
40 Bouc-Wen - p=5500.0kPa
c2 , h3 , and c3 given in Table 1. In Fig. 9, the vertical displacement
of Node A versus the applied axial force is plotted. 20
In the Fig. 10, the response of the tube is plotted in terms of 0
vertical displacement at Node A versus the applied axial force 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
for the two cases of internal pressure when the material follows displacement of point A (m)
the multiplicative AF kinematic hardening rule with the parameters
Fig. 10. Postbuckling response.
from Table 1.

© ASCE 04019027-11 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 140
p=0.0Kpa p=5500.0kPa
120

axial force (kN)

100
80
60
40
20
0
-20
0 0.002 0.004 0.006 0.008 0.01 0.012
vertical displacement of point A (m)

Fig. 13. Vertical displacement (m) at Node A of the tube versus the
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

applied axial force (kN).

0.001

evolution of wrinkles during the cycling (m)

0.0008

Fig. 11. Distribution of the von Mises stress (second Piola-Kirchhoff)

(kN=m2 ) for Points 1 and 2 (Fig. 10) on the outer face of tube with 0.0006

deformation scale factor ¼ ×2.

0.0004

0.0002

0
Initial length of the Initial length of the
tube – p = 0.0 kPa tube – p = 5500.0 kPa

Fig. 14. Evolution of wrinkling versus initial tube length.

0.0010
Peak radial displacements at

0.0009
0.0008
0.0007
node B (m)

0.0006 p=0.0kPa
0.0005 p=5500.0kPa
0.0004
0.0003
0.0002
0.0001
0.0000
1 8 15 22 29 36 43 50
Fig. 12. Distribution of the von Mises stress (second Piola-Kirchhoff) number of loading cycle
(kN=m2 ) for Points 3 and 4 (Fig. 10) on the outer face of tube with
deformation scale factor ¼ ×2. Fig. 15. Evolution of wrinkling versus loading cycle.

In Fig. 15, the magnitude of the radial displacement of Node B

In Figs. 11 and 12, the distribution of the von Mises stress is plotted against the number of the loading cycles for an internal
(second Piola-Kirchhoff) on the outer face of the tube is shown for pressure of 0.0 or 5,500.0 kPa. Apparently, the internal pressure
Points 1, 2, 3, and 4 of Fig. 10. gives rise to larger radial displacements.
Subsequently, the cylindrical shell is cyclically loaded under In Fig. 16, the distribution of the von Mises stress on the outer
load control. The applied load has an amplitude of 53.761 kN about face of the shell is illustrated at the end of the loading for the
a compressive mean load of 37.084 kN, and the number of load two cases.
cycles is 50. In Fig. 13, the vertical displacement of Node A versus
the applied axial force is plotted, where the ratcheting is mani-
Example 2: Dynamic Analysis of Scordelis-Lo Roof
fested. This is also illustrated in Fig. 14, where it can be noticed
that the axial scans of the radial displacements gradually shift In this example, the Scordelis-Lo roof is subjected to a time-varying
toward larger values as the number of load cycles increases, indi- load (Ricker pulse) with maximum intensity qz ¼ −15 kN=m2,
cating that the diameter of the tube is gradually growing. uniformly distributed on the surface of the shell in the Z-direction

© ASCE 04019027-12 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Z
Y

Node
ode A

50m
50
25m

40°
40°
40
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 18. Geometry and boundary conditions of Scordelis-Lo roof.

Fig. 16. Distribution of the von Mises stress (second Piola-Kirchhoff)

(kN=m2 ) at the end of the cyclic loading on the outer face of the tube
with deformation scale factor ¼ ×10.

10

5
load (kN/m2)

-5

-10

-15
Fig. 19. Computational model consisting of 1,750 MITC9 shell
-20 elements.
0 2 4 6 8
time (s)

Fig. 17. Load (kN=m2 ) versus time (s). 0.05

Z-displacement of node A (m)

0.04 Bouc-Wen
t = 3.292s Abaqus
−π2 0.362 ðt−3Þ2
qz ðtÞ ¼ −15ð1 − 2π2 0.362 ðt − 3Þ2 Þe ð101Þ 0.03

The load value versus time is plotted in Fig. 17. 0.02

The structure consists of a cylindrical shell roof supported by
0.01
rigid diaphragms at both curved sides. The length of the cylinder
is 50 m, the radius is 25 m, and the thickness is 0.25 m. The geom- 0
etry and boundary conditions of the shell are shown in Fig. 18.
Due to symmetry, only a quarter of the shell is modeled. Thus, -0.01
the roof is discretized into 35 × 50 nine-node MITC9 shell ele- 0 1 2 3 4 5 6 7 8
time (s)
ments, and the computational model consists of 12 layers and 7,171
nodes. The computational model is shown in Fig. 19. Fig. 20. Vertical Z-displacement of Node A (m).
The material parameters are E ¼ 432 GPa, Et ¼ 43.2 GPa,
ν ¼ 0.0, and σy ¼ 15 MPa, where E is the Young’s modulus, Et
is the tangent modulus, ν is the Poisson’s ratio, and σy is the yield
stress. The density of the material is 7.85 Mg=m3 . In addition, the derived displacement history curve is compared with the one ob-
material follows the von Mises yield criterion, and linear kinematic tained using the Abaqus code.
hardening is considered. The numerical solution is obtained apply- In Figs. 21 and 22, the vertical Z-component of velocity and
ing the Newmark’s method and choosing a time step of 0.002 s. acceleration history of Node A are plotted, respectively.
The roof is also modeled in Abaqus (Hibbitt et al. 2001) using From all previous figures and comparisons, it is apparent that
7,000 S4R5 shell elements, and the number of resulting nodes is the solution obtained based on the proposed formulation agrees
also 7,171 for comparison reasons. well with the solution obtained using Abaqus. The total comput-
In Fig. 20, the response of the shell is plotted in terms of vertical ing time needed using a FORTRAN code implementing the pro-
Z-displacement at Node A versus time. In the same figure, the posed formulation was 9.688 × 103 s as opposed to 38.555 × 103 s

© ASCE 04019027-13 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 0.1

Z-velocity of node A (m/s)

Bouc-Wen
Abaqus
0.05

-0.05

-0.1
0 1 2 3 4 5 6 7 8
time (s)

Fig. 21. Vertical Z-velocity of Node A (m=s).

Fig. 24. Distribution of the von Mises stress (engineering stress)
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

(kN=m2 ) at the upper face of shell with deformation scale factor ¼ ×20.
1.5
Z-acceleration of node A (m/s2)

Bouc-Wen
1 Abaqus
0.5

-0.5

-1

-1.5
0 1 2 3 4 5 6 7 8
time (s)

Fig. 22. Vertical Z-acceleration of Node A (m=s2 ).

Fig. 25. Magnitude of plastic strain at the upper face of shell with
deformation scale factor ¼ ×20.
30 proposed method
standard method - full N-R
25
In Fig. 24, the distribution of the von Mises stress at the upper
20 face of the shell is illustrated at time t ¼ 3.292p
s. ffiffiffiffiffiffiffiffi
time (s)

In Fig. 25, the permanent plastic strain [εps ¼ 2=3ðεpl ∶εpl Þ1=2 ,
15
Eq. (8)] at the upper interface of the shell is shown at time
10 t ¼ 8.0 s.
5
Example 3: Functionally Graded Material Shells
0
0 1000 2000 3000 4000 In the third example, a cylindrical shell with free edges is subjected
time step to two opposite loads, as shown in Fig. 26 (Masud and Tham 2000).
The length of the cylinder, L, is 10.35 m, the radius, R, is 4.953 m,
Fig. 23. Comparison of computational efficiency. and the thickness, t, is 0.094 m. Due to symmetry of geometry and
loading, only the part ABCD of the shell can be modeled (in Fig. 27,
with 20 × 30 nine-node MITC9 shell elements, eight layers, and
needed by a FORTRAN code implementing the standard Newmark 2,501 nodes).
algorithm combined with Newton-Raphson iterations within each The material properties are E ¼ 10,500 kPa, Et ¼ 1,050 kPa,
time step. This corresponds to a decrease of 74.84%. Both analy- ν ¼ 0.3125, and σy ¼ 105 kPa. The material follows the von
ses were performed in a personal computer (PC) fitted with an Intel Mises yield criterion, and linear kinematic hardening is considered.
Core i7 central processing unit (CPU) at 1.70 GHz and 4.00 GB of The nonlinear effects of large rotations and displacements are taken
random access memory (RAM). In Fig. 23, the required time in into account.
seconds for each time step is plotted against the time step for 4,000 In Fig. 28, the response of the cylindrical shell is plotted in terms
steps, where the baseline corresponds to elastic behavior and the of radial displacement at Node A versus the applied load. The re-
increased time to inelastic behavior. sults obtained using the proposed formulation are also compared
In this comparison, the same tolerance was set for convergence with Abaqus and the results presented by Masud and Tham (2000).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Subsequently, the efficiency of the hysteretic MITC9 shell
ðjþ1Þ T ðjþ1Þ ðjÞ T ðjÞ element is investigated in the case of the nonlinear analysis of func-
ðfΔUgi Þ fΔUgi − ðfΔUgi Þ fΔUgi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < 10−5 tionally graded shells. FGMs are advanced engineering materials
ðjÞ T ðjÞ used in, for example, the aircraft and aerospace industry, computer
ðfΔUgi Þ fΔUgi
circuit industry, bioengineering, among others. These materials
ð102Þ were developed to improve the behavior of laminated composite

© ASCE 04019027-14 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 F
L
2
L Node B
2

kN
Node A

F
Z

Y Node C
Node D m
R
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

F
Fig. 28. Radial displacement (m) of Node A versus applied force (kN).
Fig. 26. Cylindrical shell with free edges.

of the two phases are given as a function of distance z from the

middle surface of the shell
V 1 ðzÞ ¼ 1 − V 2 ðzÞ
 
2z þ t n
V 2 ðzÞ ¼
2t
t t
− ≤z≤ ð103Þ
2 2
where t = thickness of the shell; and n = power-law exponent,
0 ≤ n ≤ ∞. The value n ¼ 0 represents a shell from the material
Phase 1, whereas as n tends to infinity the volume fraction of
Phase 1 tends to zero. Accordingly, the variation of the Young’s
modulus, E, and Poisson’s ratio, ν, across the thickness t of the
shell can be expressed (Arciniega and Reddy 2007; Woo and
Meguid 2001)
 
2z þ t n
EðzÞ ¼ V 1 ðzÞE1 þ V 2 ðzÞE2 ¼ ðE2 − E1 Þ þ E1
2t
 n
2z þ t
νðzÞ ¼ V 1 ðzÞν 1 þ V 2 ðzÞν 2 ¼ ðν 2 − ν 1 Þ þ ν1 ð104Þ
2t

Alternatively, in the presence of an elastic constituent material

Fig. 27. Computational model of the cylindrical shell.
(Ceramic phase 1) and elastoplastic constituent material (Metal
phase 2), the variation in Young’s modulus, E, Poisson’s ratio, ν,
yield stress, σy , and elastoplastic tangent modulus, Et , across the
thickness t of the shell can be expressed (Giannakopoulos et al.
1995)
shells, the main deficiency of which is the material mismatch be-
tween different materials, such as a metal and a ceramic, which νðzÞ ¼ V 1 ðzÞν 1 þ V 2 ðzÞν 2 ¼ ðν 2 − ν 1 ÞV 2 ðzÞ þ ν 1
often leads to interlaminar stress concentrations and subsequently
V 2 ðzÞE2 qþE
qþE2 þ ð1 − V 2 ðzÞÞE1
1
to delamination failures. Within the thickness of a functionally EðzÞ ¼
graded shell, the volume fraction of the constituent materials, mi- V 2 ðzÞ qþE
qþE2 þ ð1 − V 2 ðzÞÞ
1

crostructure, and properties vary gradually and smoothly from the q þ E2 E1

one surface to the other. Consequently, the gradation of the material σy ðzÞ ¼ σy2 V 2 ðzÞ þ ð1 − V 2 ðzÞÞ
q þ E1 E2
properties through the thickness eliminates jumps in the stress and
displacement distributions and results in a material that combines V 2 ðzÞEt2 qþE
qþE2 þ ð1 − V 2 ðzÞÞE1
1

the desirable properties of the constituent materials, such as thermal Et ðzÞ ¼

V 2 ðzÞ qþE
qþE1
þ ð1 − V 2 ðzÞÞ
and corrosive resistance, toughness, strength, and hardness. t2

For a functionally graded shell with two elastic constituent ma- 2z þ t

V 2 ðzÞ ¼ ð105Þ
terials (Phase 1 and Phase 2), the variation of the volume fractions 2t

© ASCE 04019027-15 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Stress
ceramic
composite
A
B H
E1
C metal
H2 q
E

E2

Strain
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 29. Stress-strain curve representing the constitutive response of

the composite as a function of the volume fractions (Giannakopoulos
et al. 1995).

Fig. 31. Radial displacement (m) of Node D versus applied force (kN).

Fig. 30. Radial displacement (m) of Node A versus applied force (kN).

where q = ratio of stress to strain transfer (Fig. 29)

σ1 − σ2
q¼ ; 0 < q < þ∞ ð106Þ
ε1 − ε2
Fig. 32. Radial displacement (m) of Node C versus applied force (kN).
Two Elastic Materials
In the first case, a functionally graded cylindrical shell (L ¼
10.35 m, R ¼ 4.953 m, and t ¼ 0.094 m) with free edges is sub- Finally, in Fig. 33, the distribution of the von Mises stress on the
jected to two opposite loads, as shown in Fig. 26 (Arciniega and outer face of the shell is illustrated.
Reddy 2007). The material properties for the constituents are
E1 ¼ 15.1 × 108 kPa, E2 ¼ 7.0 × 108 kPa, and ν 1 ¼ ν 2 ¼ 0.3, and Ceramic (Elastic) and Metal (Elastoplastic)
the variation of these parameters follows the rule of Eq. (104). The
nonlinear effects of large rotations and displacements are taken into In this case, an annular functionally graded plate is subjected to a
account. distributed transverse shear force qz , as shown in Fig. 34 (Arciniega
In Figs. 30–32, the response of the functionally graded cylin- and Reddy 2007). The internal radius of the plate, Ri , is 6.0 m, the
drical shell is plotted in terms of radial displacement at Nodes external radius, Re , is 10.0 m, and the thickness, t, is 0.03 m. The
A, D, and C, respectively, versus the applied load for different val- computational model is presented in Fig. 35, and consists of 8 × 60
ues of the power-law exponent n of Eq. (104). The results obtained nine-node MITC9 shell elements, eight layers, and 2,057 nodes.
using the proposed formulation are also compared with the results The material properties for the ceramic constituent (Al2 O3 )
presented by Arciniega and Reddy (2007). are E1 ¼ 3.80 × 108 kPa and ν 1 ¼ 0.25, whereas for the metal

© ASCE 04019027-16 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 kN/m
F
m
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 36. Vertical Z-displacement (m) of Node A versus applied force

(kN=m).
Fig. 33. Distribution of the von Mises stress (second Piola-Kirchhoff)
(kN=m2 ) on the outer face for n ¼ 1 with deformation scale
factor ¼ ×1.

qz

Node A
Z Y

Re Ri X

Fig. 34. Functionally graded annular plate.

Fig. 37. Distribution of the von Mises stress (second Piola-Kirchhoff)
(kN=m2 ) at the upper face of the functionally graded plate with defor-
mation scale factor ¼ ×1.

1995). The nonlinear effects of large rotations and displacements

are taken into account.
In Fig. 36, the response of the functionally graded plate is
plotted in terms of vertical displacement at Node A versus the
applied load.
Finally, in Fig. 37, the distribution of the von Mises stress on the
outer face of the plate is illustrated.

Concluding Remarks

A hysteretic MITC9 shell finite element was developed that is suit-

able for the inelastic and geometric nonlinear static and dynamic
Fig. 35. Computational model of the functionally graded annular plate. analysis of shell structures. A smooth rate-independent hysteretic
model of Bouc-Wen type was incorporated in the constitutive rela-
tions of the standard finite-element formulation yielding the elastic,
hysteretic, and tangent element matrices. Plastic strains, backstress
phase (Ni), the properties are E2 ¼ 2.14 × 108 kPa, ν 2 ¼ 0.31, components, and the variable yield stress at all shell layer interfaces
sy ¼ 1.48 × 105 kPa, and Et ¼ 0.003E2 . The variation of these were introduced as additional unknowns together with correspond-
parameters follows the rule of Eq. (105), and the parameter q of ing evolution equations. The system of governing equations was
Eq. (106) is considered as 4.5 × 106 kPa (Giannakopoulos et al. solved numerically, employing efficient solution schemes.

© ASCE 04019027-17 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 The basic advantage of the proposed method relies on the fact the inelastic constitutive relation in the principle of virtual work in
that the physical inelastic response is handled through a system of a separable form, distinguishing the elastic and hysteretic part.
first-order differential equations. This provides the nodal displace- This results in structural matrices that are evaluated only once,
ments, elastic and plastic strains, backstress components, the var- at the beginning of the analysis procedure. Thus, it becomes evi-
iable yield stress, and stresses at all Gauss points of every layer dent that the proposed formulation directly accounts for inelastic-
that satisfy the inelastic constitutive relations and equilibrium ity in a natural way by solving in coupled form the equilibrium
without any additional iterative process. Therefore, by solving the equations together with the nonlinear evolution equations. This
system of differential equations numerically, the scheme always avoids the inconsiderate elastic predictions, which are belatedly
stays on the yield function and satisfies the flow rule ab initio. followed by plastic corrections. For these reasons, the proposed
Consequently, the local iterations of Newton-Raphson method are formulation turns out to be computationally more efficient for
avoided at the expense of the numerical solution of first-order evo- the same accuracy compared with standard methods, which was
lution equations for the introduced additional hysteretic unknowns. verified with a set of examples and comparisons with existing
Especially, for the inelastic case, the proposed formulation utilizes results.
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

Appendix. Solution Procedure for the Inelastic Case

The solution procedure for the inelastic case is performed in steps as follows.
Step Action
1.1 Select Δt
1.2 Solve fÜg0 ¼ ½M S −1 ½fPS g0 − ½CS fU̇g0 − ½K S fUg0 
 
1 γ 1 γ γ 1
1.3 Evaluate ½a ¼ ½M s  þ ½CS , ½b ¼ ½M S  þ Δt − 1 ½CS , ½K̂ ¼ ½K S  þ ½C  þ ½MS 
βΔt β 2β 2β βΔt S βðΔtÞ2
2.0 For every time step i ¼ 1; 2; : : : ; N
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
ð0Þ ðfΔPgi ÞT fΔPgi
2.1 Start with fΔZgi ¼ fΔZgi−1 , fΔPS g0 ¼ fΔZg0 ¼ f0g
ðfΔPgi−1 ÞT fΔPgi−1
ð0Þ ð0Þ ð0Þ ð0Þ
2.2 Evaluate fΔP̂a gi ¼ fΔPS gi þ ½afU̇gi þ ½bfÜgi , fΔP̂b gi ¼ ½H S fΔZgi , fΔP̂gi ¼ fΔP̂a gi þ fΔP̂b gi
ð1Þ ð0Þ ð1Þ ð1Þ
2.3 Solve fΔUgi ¼ ½K̂−1 fΔP̂gi ⇒ fUgi ¼ fUgi−1 þ fΔUgi
ð1Þ ð1Þ ð1Þ
2.4 Compute fΔZgi , fΔAgi , fΔKgi by evolving the Bouc-Wen Eqs. (20), (22), and (23)
ð1Þ ð1Þ ð1Þ ð1Þ
2.5 fΔP̂b gi ¼ ½H S fΔZgi ⇒ fΔP̂gi ¼ fΔP̂a gi þ fΔP̂b gi
2.6.0 For every inner iteration j ¼ 1; 2; : : : ,
ðjþ1Þ ðjÞ
2.6.1 Solve fΔUgi ¼ ½K̂−1 fΔP̂gi
ðjþ1Þ ðjþ1Þ
2.6.2 fUgi ¼ fUgi−1 þ fΔUgi
ðjþ1Þ ðjþ1Þ ðjþ1Þ
2.6.3 Compute fΔZgi , fΔAgi , and fΔKgi by evolving the Bouc-Wen Eqs. (20), (22), and (23)
ðjþ1Þ ðjþ1Þ ðjþ1Þ ðjþ1Þ
2.6.4 fΔP̂b gi ¼½H S fΔZgi ⇒
fΔP̂gi ¼ fΔP̂ g þ fΔP̂b gi
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia i
u ðjþ1Þ ðjþ1Þ ðjÞ ðjÞ
uðfΔP̂gi T
Þ fΔUgi − ðfΔP̂gi Þ fΔUgi
T
2.6.5 If t ðjÞ T ðjÞ
< ε, then go to Step 2.7 Else j ¼ j þ 1 and go to Step 2.6.1
ðfΔP̂gi Þ fΔUgi  
γ γ γ
2.7 fΔU̇gi ¼ fΔUgi − fU̇gi þ Δt 1 − fÜgi
βΔt β 2β
1 1 1
2.8 fΔÜgi ¼ fΔUgi − fU̇gi − fÜgi
βðΔtÞ2 βΔt 2β
2.9 fUgiþ1 ¼ fUgi þ fΔUgi , fU̇giþ1 ¼ fU̇gi þ fΔU̇gi , fÜgiþ1 ¼ fÜgi þ fΔÜgi , fZgiþ1 ¼ fZgi þ fΔZgi , fAgiþ1 ¼ fAgi þ fΔAgi ,
fKgiþ1 ¼ fKgi þ fΔKgi

References Argyris, J. H., J. St. Doltsinis, and K. J. Willam. 1979. “New developments
in the inelastic analysis of quasistatic and dynamic problems.” Int. J.
Ahmad, S., B. M. Irons, and O. C. Zienkiewicz. 1970. “Analysis of thick Numer. Methods Eng. 14 (12): 1813–1850. https://doi.org/10.1002/nme
and thin shell structures by curved finite elements.” Int. J. Numer. Meth- .1620141206.
ods Eng. 2 (3): 419–451. https://doi.org/10.1002/nme.1620020310. Bathe, K. J. 1996. Finite element procedures. Upper Saddle River, NJ:
Arciniega, R. A., and J. N. Reddy. 2007. “Large deformation analysis of Prentice Hall.
functionally graded shells.” Int. J. Solids Struct. 44 (6): 2036–2052. Bathe, K. J., and S. Bolourchi. 1980. “A geometric and material nonlinear
https://doi.org/10.1016/j.ijsolstr.2006.08.035. plate and shell element.” Comput. Struct. 11 (1–2): 23–48. https://doi
Argyris, J. 1982. “An excursion into large rotations.” Comput. Methods .org/10.1016/0045-7949(80)90144-3.
Appl. Mech. Eng. 32 (1–3): 85–155. https://doi.org/10.1016/0045 Bathe, K. J., and E. N. Dvorkin. 1986. “A formulation of general shell
-7825(82)90069-X. elements—The use of mixed interpolation of tensorial components.”

© ASCE 04019027-18 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Int. J. Numer. Methods Eng. 22 (3): 697–722. https://doi.org/10.1002 Erlicher, S., and N. Point. 2004. “Thermodynamic admissibility of Bouc-
/nme.1620220312. Wen type hysteresis models.” C. R. Mec. 332 (1): 51–57. https://doi.org
Bathe, K. J., P. S. Lee, and J. F. Hiller. 2003. “Towards improving the /10.1016/j.crme.2003.10.009.
MITC9 shell element.” Comput. Struct. 81 (8–11): 477–489. https://doi Frederick, C. O., and P. J. Armstrong. 1966. A mathematical representation
.org/10.1016/S0045-7949(02)00483-2. of the multiaxial Bauschinger effect. C.E.G.B. Rep. No. RD/B/N660.
Belytschko, T., B. L. Wong, and H. Stolarski. 1989. “Assumed strain George, A., and J. W. H. Liu. 1981. Computer solution of large sparse
stabilization procedure for the 9-node Lagrange shell element.” Int. positive definite systems. Englewood Cliffs, NJ: Prentice-Hall.
J. Numer. Methods Eng. 28 (2): 385–414. https://doi.org/10.1002/nme Giannakopoulos, A. E., S. Suresh, M. Finot, and M. Olsson. 1995. “Elasto-
.1620280210. plastic analysis of thermal cycling: Layered materials with composi-
Bischoff, M., and E. Ramm. 1997. “Shear deformable shell elements tional gradients.” Acta Metall. Mater. 43 (4): 1335–1354. https://doi
for large strains and rotations.” Int. J. Numer. Methods Eng. 40 (23): .org/10.1016/0956-7151(94)00360-T.
4427–4449. https://doi.org/10.1002/(SICI)1097-0207(19971215)40: Hassan, T., and S. Kyriakides. 1992. “Ratcheting in cyclic plasticity. Part I:
23<4427::AID-NME268>3.0.CO;2-9. Uniaxial behavior.” Int. J. Plast. 8 (1): 91–116. https://doi.org/10.1016
Bonet, J., and R. D. Wood. 2008. Nonlinear continuum mechanics for finite /0749-6419(92)90040-J.
element analysis. Communications in numerical methods in engineer- Hibbitt, H. D., B. Karlsson, and E. P. Sorensen. 2001. Abaqus/standard
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

ing. Cambridge, UK: Cambridge University Press. user’s manual. Providence, RI: Hibbitt, Karlsson, & Sorensen.
Bouc, R. 1967. “Forced vibration of mechanical system with hysteresis.” Huang, H. C., and E. Hinton. 1986. “A new nine node degenerated shell
In Proc., 4th Conf. on Nonlinear Oscillations. Prague, Czech Republic: element with enhanced membrane and shear interpolation.” Int. J.
Academia. Numer. Methods Eng. 22 (1): 73–92. https://doi.org/10.1002/nme
Bucalem, M. L., and K. J. Bathe. 1993. “Higher-order MITC general shell .1620220107.
elements.” Int. J. Numer. Methods Eng. 36 (21): 3729–3754. https://doi Hughes, T. J. R., and E. Carnoy. 1983. “Nonlinear finite element shell
.org/10.1002/nme.1620362109. formulation accounting for large membrane strains.” Comput. Methods
Carrera, E., M. Cinefra, G. Li, and G. M. Kulikov. 2016. “MITC9 shell Appl. Mech. Eng. 39 (1): 69–82. https://doi.org/10.1016/0045-7825(83)
finite elements with miscellaneous through-the-thickness functions 90074-9.
for the analysis of laminated structures.” Compos. Struct. 154: 360–373. Hughes, T. J. R., and W. K. Liu. 1981. “Nonlinear finite element analysis of
https://doi.org/10.1016/j.compstruct.2016.07.032. shells. Part I: Three-dimensional shells.” Comput. Methods Appl. Mech.
Chaboche, J. L. 1989. “Constitutive equations for cyclic plasticity and Eng. 26 (3): 331–362. https://doi.org/10.1016/0045-7825(81)90121-3.
cyclic viscoplasticity.” Int. J. Plast. 5 (3): 247–302. https://doi.org/10 Jiao, R., and S. Kyriakides. 2009. “Ratcheting, wrinkling and collapse of
.1016/0749-6419(89)90015-6. tubes under axial cycling.” Int. J. Solids Struct. 46 (14–15): 2856–2870.
Chaboche, J. L. 1991. “On some modifications of kinematic hardening https://doi.org/10.1016/j.ijsolstr.2009.03.018.
to improve the description of ratchetting effects.” Int. J. Plast. 7 (7): Jiao, R., and S. Kyriakides. 2011a. “Ratcheting and wrinkling of tubes due
661–678. https://doi.org/10.1016/0749-6419(91)90050-9. to axial cycling under internal pressure. Part I: Experiments.” Int. J.
Chakravorty, D., P. K. Sinha, and J. N. Bandyopadhyay. 1998. “Applica- Solids Struct. 48 (20): 2814–2826. https://doi.org/10.1016/j.ijsolstr
tions of FEM on free and forced vibration of laminated shells.” J. Eng. .2011.05.027.
Mech. 124 (1): 1–8. https://doi.org/10.1061/(ASCE)0733-9399(1998) Jiao, R., and S. Kyriakides. 2011b. “Ratcheting and wrinkling of tubes
124:1(1). due to axial cycling under internal pressure. Part II: Analysis.” Int.
Charalampakis, A. E., and V. K. Koumousis. 2009. “A Bouc-Wen model J. Solids Struct. 48 (20): 2827–2836. https://doi.org/10.1016/j.ijsolstr
compatible with plasticity postulates.” J. Sound Vib. 322 (4–5): .2011.05.026.
954–968. https://doi.org/10.1016/j.jsv.2008.11.017. Khan, A. S., and S. Huang. 1995. Continuum theory of plasticity, 440.
Chopra, A. K. 2011. Dynamics of structures: Theory and applications to New York: Wiley.
earthquake engineering: Pearson education. Upper Saddle River, NJ: Ko, Y., P. S. Lee, and K. J. Bathe. 2017a. “The MITC4+ shell element
Pearson Prentice Hall. in geometric nonlinear analysis.” Comput. Struct. 185: 1–14. https://doi
Cinefra, M., and S. Valvano. 2016. “A variable kinematic doubly-curved .org/10.1016/j.compstruc.2017.01.015.
MITC9 shell element for the analysis of laminated composites.” Ko, Y., Y. Lee, P. S. Lee, and K. J. Bathe. 2017b. “Performance of
Mech. Adv. Mater. Struct. 23 (11): 1312–1325. https://doi.org/10.1080 the MITC3+ and MITC4+ shell elements in widely-used benchmark
/15376494.2015.1070304. problems.” Comput. Struct. 193: 187–206. https://doi.org/10.1016/j
Cinefra, M., S. Valvano, and E. Carrera. 2015. “A layer-wise MITC9 finite .compstruc.2017.08.003.
element for the free-vibration analysis of plates with Piezo-patches.” Int. Kottari, A. K., A. E. Charalampakis, and V. K. Koumousis. 2014. “A con-
J. Smart Nano Mater. 6 (2): 85–104. https://doi.org/10.1080/19475411 sistent degrading Bouc-Wen model.” Eng. Struct. 60: 235–240. https://
.2015.1037377. doi.org/10.1016/j.engstruct.2013.12.025.
Cinefra, M., S. Valvano, and E. Carrera. 2016. “Thermal stress analysis Liu, W. K., Y. Guo, S. Tang, and T. Belytschko. 1998. “A multiple-
of laminated structures by a variable kinematic MITC9 shell element.” quadrature eight-node hexahedral finite element for large deformation
J. Therm. Stresses 39 (2): 121–141. https://doi.org/10.1080/01495739 elastoplastic analysis.” Comput. Methods Appl. Mech. Eng. 154 (1–2):
.2015.1123591. 69–132. https://doi.org/10.1016/S0045-7825(97)00106-0.
Crisfield, M. A. 1991. Vol. 1 of Non-linear finite element analysis of solids Masud, A., and M. Panahandeh. 1999. “Finite-element formulation for
and structures: Essentials. New York: Wiley. analysis of laminated composites.” J. Eng. Mech. 125 (10): 1115–1124.
Crisfield, M. A. 1997. Vol. 2 of Non-linear finite element analysis of solids https://doi.org/10.1061/(ASCE)0733-9399(1999)125:10(1115).
and structures: Advanced topics. New York: Wiley. Masud, A., and C. L. Tham. 2000. “Three-dimensional corotational frame-
Dafalias, Y. F., K. I. Kourousis, and G. J. Saridis. 2008a. “Corrigendum to work for elasto-plastic analysis of multilayered composite shells.”
‘Multiplicative AF kinematic hardening in plasticity’.” Int. J. Solids AIAA J. 38 (12): 2320–2327. https://doi.org/10.2514/2.901.
Struct. 45 (10): 2861–2880. https://doi.org/10.1016/j.ijsolstr.2008.01.001. Masud, A., C. L. Tham, and W. K. Liu. 2000. “Stabilized 3-D co-rotational
Dafalias, Y. F., K. I. Kourousis, and G. J. Saridis. 2008b. “Multiplicative formulation for geometrically nonlinear analysis of multi-layered
AF kinematic hardening in plasticity.” Int. J. Solids Struct. 45 (10): composite shells.” Comput. Mech. 26 (1): 1–12. https://doi.org/10.1007
2861–2880. https://doi.org/10.1016/j.ijsolstr.2008.01.001. /s004660000144.
de Souza Neto, E. A., D. Peri, and D. R. J. Owen. 2008. Computational Moysidis, A. N., and V. K. Koumousis. 2015. “Hysteretic plate finite
methods for plasticity. Computational methods for plasticity—Theory element.” J. Eng. Mech. 141 (10): 4015039. https://doi.org/10.1061
and applications. Chichester, UK: Wiley. /(ASCE)EM.1943-7889.0000918.
Dvorkin, E. N. 1995. “Nonlinear analysis of shells using the MITC formu- Nayak, A. N., and J. N. Bandyopadhyay. 2005. “Free vibration analysis of
lation.” Arch. Comput. Methods Eng. 2 (2): 1–50. https://doi.org/10 laminated stiffened shells.” J. Eng. Mech. 131 (1): 100–105. https://doi
.1007/BF02904994. .org/10.1061/(ASCE)0733-9399(2005)131:1(100).

© ASCE 04019027-19 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027

 Owen, D. R. J., and E. Hinton. 1980. Finite elements in plasticity: Theory Sofianos, C. D., and V. K. Koumousis. 2017. “Hysteretic beam element
and practice. Swansea, UK: Pineridge Press. with degrading smooth models.” Arch. Appl. Mech. 88 (1–2): 253–269.
Pawsey, S. F., and R. W. Clough. 1971. “Improved numerical integration of https://doi.org/10.1007/s00419-017-1263-8.
thick shell finite elements.” Int. J. Numer. Methods Eng. 3 (4): 575–586. Tham, C. L., Z. Zhang, and A. Masud. 2005. “An elasto-plastic damage
https://doi.org/10.1002/nme.1620030411. model cast in a co-rotational kinematic framework for large deformation
Pinsky, P. M., and J. Jang. 1988. “Elastoplastic shell element based on analysis of laminated composite shells.” Comput. Methods Appl. Mech.
assumed covariant strain interpolations.” J. Eng. Mech. 114 (6): 1045– Eng. 194 (21–24): 2641–2660. https://doi.org/10.1016/j.cma.2004.07.050.
1062. https://doi.org/10.1061/(ASCE)0733-9399(1988)114:6(1045). Toscano, R. G., and E. N. Dvorkin. 2007. “A shell element for finite
Radhakrishnan, K., and A. C. Hindmarsh. 1993. Description and use strain analyses: Hyperelastic material models.” Eng. Comput. 24 (5):
of LSODE, the Livermore solver for ordinary differential equations. 514–535. https://doi.org/10.1108/02644400710755898.
Lawrence Livermore National Laboratory Rep. No. UCRL-ID-113855. Triantafyllou, S. P., and V. K. Koumousis. 2014. “Hysteretic finite elements
Washington, DC: National Aeronautics and Space Administration. for the nonlinear static and dynamic analysis of structures.” J. Eng.
Schimmels, S. A., and A. N. Palazotto. 1994. “Nonlinear geometric and Mech. 140 (6): 04014025. https://doi.org/10.1061/(ASCE)EM.1943
material behavior of shell structures with large strains.” J. Eng. Mech. -7889.0000699.
120 (2): 320–345. https://doi.org/10.1061/(ASCE)0733-9399(1994) Wen, Y.-K. 1976. “Method for random vibration of hysteretic systems.”
Downloaded from ascelibrary.org by East Carolina University on 03/05/19. Copyright ASCE. For personal use only; all rights reserved.

120:2(320). J. Eng. Mech. Div. 102 (2): 249–263.

Simo, J. C., and T. J. R. Hughes. 1998. Computational inelasticity. Wisniewski, K., and P. Panasz. 2013. “Two improvements in formulation
New York: Springer. of nine-node element MITC9.” Int. J. Numer. Methods Eng. 93 (6):
Simo, J. C., M. S. Rifai, and D. D. Fox. 1990. “On a stress resultant 612–634. https://doi.org/10.1002/nme.4399.
geometrically exact shell model. Part IV: Variable thickness shells with Woo, J., and S. A. Meguid. 2001. “Nonlinear analysis of functionally
through-the-thickness stretching.” Comput. Methods Appl. Mech. Eng. graded plates and shallow shells.” Int. J. Solids Struct. 38 (42–43):
81 (1): 91–126. https://doi.org/10.1016/0045-7825(90)90143-A. 7409–7421. https://doi.org/10.1016/S0020-7683(01)00048-8.
Sivaselvan, M. V., and A. M. Reinhorn. 2000. “Hysteretic models for Zienkiewicz, O. C., R. L. Taylor, and J. M. Too. 1971. “Reduced integration
deteriorating inelastic structures.” J. Eng. Mech. 126 (6): 633–640. technique in general analysis of plates and shells.” Int. J. Numer. Meth-
https://doi.org/10.1061/(ASCE)0733-9399(2000)126:6(633). ods Eng. 3 (2): 275–290. https://doi.org/10.1002/nme.1620030211.

© ASCE 04019027-20 J. Eng. Mech.

J. Eng. Mech., 2019, 145(5): 04019027