Professional Documents
Culture Documents
art ic l e i nf o a b s t r a c t
Article history: In this study we apply the total Lagrangian formulation based on positions to develop a strategy for the
Received 13 November 2013 physical and geometrical nonlinear analysis of plane structures and mechanisms. The developed frame
Received in revised form elements have cross sections constituted by laminas of different materials and widths. The Reissner–Mindlin
15 May 2014
kinematics is adopted to include shear effects into the formulation. These elements are connected among
Accepted 1 July 2014
Available online 24 July 2014
each other by semi-rigid connections. Both elements and connections have multi-linear elastoplastic
behavior with isotropic hardening. An alternative flow direction rule, which allows the reproduction of
Keywords: any stress–strain curve and the determination of closed solution for the plastic multiplier, is shown. All steps
Physical and geometrical nonlinear to reproduce the formulation are provided. Various examples are used to demonstrate the good behavior of
dynamics
the formulation, including a progressive collapse analysis of a tall building subjected to a real seismic load.
Frames
& 2014 Elsevier B.V. All rights reserved.
Machines
Finite element methods
Semi-rigid connections
Progressive collapse
http://dx.doi.org/10.1016/j.finel.2014.07.001
0168-874X/& 2014 Elsevier B.V. All rights reserved.
2 H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15
works that follow the same localized plastic hinge strategy to We start the formulation description by defining the Reissner
model the behavior of frame elements, like [14–22]. kinematic for laminate frame elements passing through the
Some authors [23–25] use distributed plasticity to model two establishment of the total potential energy including plastic
and three dimensional frames. One advantage of distributed dissipation for both frame elements and semi-rigid connections.
plasticity, when compared to plastic hinges, is the better repre- After that, the principle of stationary energy is employed to write
sentation of the plastic evolution over cross sections, without the the dynamic equilibrium equations (regarding positions) and the
necessity of a previous knowledge of moment–curvature curves. elastoplastic evolution is defined. The Newmark time integration
However, both plastic hinges and the existent distributed plasticity algorithm and the Newton–Raphson procedure are combined to
frame models do not consider the shear stress influence like the solve the resulting time step nonlinear system of equations.
finite element proposed in this work. Examples are presented to validate the proposed formulation
In the context of three dimensional analyses advanced works and to show its possibilities.
such as [26–28] should be mentioned. In these works cable, truss
and beam-column elements are specifically designed to guarantee
efficiency for practical applications. Authors include second order 2. Frame element Reissner kinematics
formulations for the beam-column element and apply lumped
mass matrix to proceed with time integration when all elements The FE formulation presented here is called positional as it is
are coupled together. Moreover, fiber concept plasticity is intro- based on positions, not displacements [46,47]. The main advan-
duced to consider distributed plasticity along elements. tages of this total Lagrangian strategy are the establishment of
Concerning problems in which semi-rigid connections are gradient deformation without the explicit use of the chain rule and
present one can cite the works of [29–36]. All these works use the achievement of quasi-constant mass matrix [49–52]. The chain
second order geometrical description, which limits the range of rule operation appears as a simple numerical (2 2) matrix
applications, i.e., displacements and rotations should not be large inversion, which allows an easy generalization of this procedure
as required, for example, in machine and progressive collapse for any class of nonlinear mechanical problems.
analysis including the post-critical behavior of ruined structures. In this section we present the complete development of the
In order to implement semi-rigid connections in any computa- alternative 2D positional laminate Reissner kinematics to be used
tional code it is necessary to use results from works that study in the proposed finite element formulation.
the experimental behavior of these connections, as, for example,
[37,38]. Some works, using experimental results, to establish 2.1. Initial configuration
empirical mathematical models for connections behavior are also
present in literature; see, for example, the works of [29,39–44]. The positional formulation is based on two mappings, one
Based on these studies and in second order instability theory some related to the initial configuration and another related to the
authors [45] adopted elastoplastic power laws to model the non- current configuration. To describe the initial configuration map-
linear behavior of semi-rigid connection for the analysis of 2D ping one starts with the reference line approximation, see Fig. 1a,
structures subjected to seismic loads. These authors also adopted by the following expression:
localized plasticity at extremity of elements (as the elastic
f oi ðξ1 Þ ¼ xm
i ðξ1 Þ ¼ ϕℓ X iℓ
m m
ð1Þ
instability approach has been adopted along bars) to model the
physical nonlinear behavior of 2D frames. in which i is the coordinate direction (1 or 2), m represents the
In the present work an alternative position based on finite reference line and ℓ the element node (or shape function). In
element (FE) formulation [46,47] is developed to comprise geo- expression (1) the repetition of index ℓ indicates summation
metrical and physical nonlinearity of both frame members and (Einstein notation). Fig. 1a shows 4 nodes (cubic approximation);
connections for dynamic analysis of machines and structures. The however any number of nodes or approximation order may be
formulation, originally developed and presented in this work, is chosen.
!m
geometrically exact and, considering the Reissner kinematic In Fig. 1a f 0 ðξ1 Þ is the mapping from the non-dimensional
hypothesis, includes shear stress contribution for both displace- variable ξ1 to the initial reference line. To define the cross sections
initial position one calculates the nodal angle θk that the normal
0
ment and failure criterion. Based on [48] the developed elasto-
plastic algorithm is multi-linear with an alternative flow direction vector vik makes with the horizontal direction (x1 axis), see Fig. 1b, as
rule, which allows the reproduction of any stress–strain curve and
the determination of closed solution for the plastic multiplier.
θ0k ¼ arctgðv2ðkÞ =v1ðkÞ Þ ð2Þ
Moreover, semi-rigid elastoplastic connections develop large rota- in which
tions allowing realistic analysis of unload situations for which ! !
connections and/or frame elements suffer plastic deformations. v1k ¼ T 2k =jj T jj and v2k ¼ T 1k =jj T jj ð3Þ
Fig. 1. Reference line parameterization for initial configuration (cubic approximation). (a) Reference line parametrization and (b) cross section initial positions.
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 3
and
dϕℓ ðξ1 Þ m
T ik ¼ X ð4Þ
dξ k iℓ
1 ξ1
tion
In Fig. 2 one can find any point inside the continuum (laminate
frame) using the following expressions:
!
lam
h
f 01 ðξ1 ; ξ2 Þ ¼ xlam ξ ξ ϕ ξ ξ cos ðϕℓ ðξ1 Þθℓ Þ
1ℓ lam 0
ð 1 ; 2 Þ ¼ ð
ℓ 1 ÞX þ d þ
1
2 2
Fig. 3. Current configuration mapping – detaching angles.
ð6Þ
!
lam
h
f 02 ðξ1 ; ξ2 Þ ¼ xlam
2 ðξ1 ; ξ2 Þ ¼ ϕℓ ðξ1 ÞX 2ℓ þ d ξ2 sin ðϕℓ ðξ1 Þθ0ℓ Þ
lam
þ
2
ð7Þ
in which ξ2 is the second non-dimensional variable, d
lam
is the
lam
distance of the lamina mid line from the reference line, h is
the height of a lamina and the pair ð cos ðϕℓ ðξ1 Þθℓ Þ; sin ðϕℓ ðξ1 Þθℓ ÞÞ
0 0
is the parameterized unit orthogonal vector. In Eqs. (6) and (7) the
symbol m is suppressed, and from now on X iℓ is simply the initial
nodal coordinate.
lam
The width of each lamina ðb Þ is introduced at the integration
procedure as an additional weighting value. Fig. 4. Change of configuration – positional mapping.
2.2. Current configuration One can see from Fig. 3 and from Eqs. (8) and (9) that cross
sections remain straight, but not orthogonal to the reference line.
The necessary information to build the initial configuration This characterizes the general Reissner kinematics. Moreover we
comprises the reference line nodal coordinates and the distance of kept h
lam
and b
lam
unchanged, limiting our constitutive relation to
each lamina from the reference line (assumed constant) as well as accept any shear elastic modulus (G), but null Poisson ratio
the height and width of each lamina (assumed constant). The regarding the longitudinal strain relations, see Eq. (16).
current configuration is achieved by a nonlinear process that uses
a trial position to start the solution procedure. So, at this stage, we
do not worry about the solution process and write the current 2.3. Change of configuration function (deformation) and its gradient
configuration similarly to the initial one, as
! As the mappings from the non-dimensional space to initial and
lam
h
f 11 ðξ1 ; ξ2 Þ ¼ ylam ξ ξ ϕ ξ ξ cos ðϕℓ ðξ1 Þθℓ Þ current configurations had been defined in previous section, we
lam
ð 1 ; 2 Þ ¼ ð
ℓ 1 ÞY 1ℓ þ d þ
1
2 2 start the description of the deformation of the analyzed body
ð8Þ (laminate frame). This is done by joining the two mappings of
! Figs. 2 and 3 in a single representation, see Fig. 4, for which the
h
lam !
f 12 ðξ1 ; ξ2 Þ ¼ ylam function f describes the deformation from initial configuration
2 ðξ1 ; ξ2 Þ ¼ ϕℓ ðξ1 ÞY 2ℓ þ d ξ2 ð sin ðϕℓ ðξ1 Þθℓ
lam
þ
2 !
ðB0 Þ to the current one ðBÞ. As is well known [46] f is written as a
ð9Þ ! !
composition of mappings f 0 and f 1 as
Here ylam
are the current coordinates of a general point inside the
i ! ! ! 1
frame element and lamina, Y iℓ are current nodal coordinates and f ¼ f 1ð f 0Þ ð10Þ
θℓ are the current angles of cross sections, see Fig. 3. !
and the gradient of f , called here A (a 2 2 tensor), is written
! !
from the gradient of f 0 and f 1 as
Fig. 2. Point P at a general cross section of the initial element configuration. ð12Þ
4 H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15
!_ !
element. No summation is implied. in which y δt ¼ δ y has been used.
By the energy conjugate principle [53], the first term of the first
integral of Eq. (19) is the second Piola–Kirchhoff stress and the
3.2. Frame element strain energy
derivative of the strain energy stored in elastic connections is the
!int
As mentioned before the Saint-Venant-Kirchhoff specific strain internal moment M . Therefore, in order to simplify the under-
energy is adopted here to relate the Green strain (E) and the standing of the physical nonlinear procedure shown in the next
section, one rewrites Eq. (19) as In this case, instead of using the elastic strain energy, for both the
Z frame element and connections, one adopts the Helmholtz free
∂E ! !int ! ! !
δΠ ¼ S : !dV 0 U δ Y þ M U δ Y F U δ Y energy potential [54].
Z
V0 ∂Y Z Following this reasoning the total potential energy for isother-
! ! !€ ! ! ! !
q U ϕdS0 U δ Y þ ρ0 Y U ϕ ϕ dV 0 U δ Y ¼ 0 ð20Þ mal problems is written as
S0 V0 Z
! ! !
!m ! ! Π ð Y Þ ¼ Ψ ðE; αÞdV 0 þ ΘSR ððY η3α Y η3β Þ; αÞ F U Y
To obtain Eq. (20) the property ∂ y =∂ Y ¼ ϕ and the nodal V0
! Z Z
! !m ! _ _
ρ0 ! !
characteristic of δ Y have been used. Moreover, the same approx- 1
q U y ð Y ÞdS0 þ y U y dV 0 ð24Þ
imation adopted to position is used for acceleration. S0 2 V0
The understanding of Eq. (20) can be further improved as where Ψ and Θ are, respectively, the frame elements and the
SR
!int ! !int
! ! ! ! ! connections free energy potentials. These potentials are written as
δΠ ¼ F Uδ Y þ M Uδ Y F Uδ Y L U Q Uδ Y
a function of the Green strain tensor, the relative angle positions at
!€ ! !_ ! !
þMU Y Uδ Y þDU Y Uδ Y ¼ 0 ð21Þ connections and the thermodynamic parameters α and α.
The variation of the total potential energy is zero at the
in which L is the matrix that transforms the distributed forces into equilibrium position, i.e.
nodal equivalent ones, M is the constant mass matrix and D is the
Z Z
mass proportional damping [52,54,55]. Due to the arbitrariness of ∂Ψ ∂E ! ∂ΘSR ! ! ! ! !
! δΠ ¼ : !dV 0 U δ Y þ ! U δ Y F U δ Y q U ϕdS0 U δ Y
δ Y Eq. (21) represents the geometrical nonlinear dynamic equili- V 0 ∂E ∂ Y ∂Y S 0
brium equation, i.e. Z
! € ! ! ! !
!int !int ! ! !€ !_ ! þ ρ0 Y U ϕ ϕ dV 0 U δ Y ¼ 0 ð25Þ
F þ M F L U Q þ M U Y þD U Y ¼ 0 ð22Þ V0
It is important to note that this work deals with relatively in which parameters α and α are not present due to their intrinsic
slender bars whose cross section rotational inertia is neglected; !int
relation with E and S and with the relative rotation and M to be
see [47] for more information regarding this subject.
shown in Section 4.2.
To close this item we show the pair of internal moments, for an
Even for inelastic problems, the derivative of the free energy
elastic semi-rigid connection, associated to a global node η related
potential with reference to Green strain is the second Piola–Kirchh-
to the initial (or final) point of a frame element α and to the final
off stress tensor. For the same reason, the energy conjugate of the
(or initial) point of a frame element β, as
free energy at elastoplastic connections with repect to relative
η ðηαβÞ η η η η ðηαβÞ η η
F 3α ¼ M ηα ¼ k ðY 3α Y 3β Þ and F 3β ¼ M β ¼ k ðY 3α Y 3β Þ angles are internal moments. In this way Eq. (25) can be written
ð23Þ exactly as Eq. (21) or (22). However, with a physical nonlinear
! meaning; i.e., while the passage from Eq. (19) to Eq. (20) is done by a
In Appendix A one finds explicit expressions to calculate ∂E=∂ Y simple differentiation of the quadratic potential shown by
present in Eq. (20). In Section 4 the nonlinear dynamic equilibrium Eq. (16), now it is necessary to define an inelastic (plastic for
is extended to include plasticity. In Section 5 the Newmark time instance) relation to describe the material behavior and its
integration and the Newton–Raphson procedure are combined to evolution rule.
solve the resulting nonlinear dynamic system of equations.
4.2. Elastoplastic constitutive relation
4. Inelastic Procedure
In this item, we summarize the constitutive elastoplastic relation
The previous section is used to introduce the geometrical developed by [48] and [56]. We follow a general 3D description in
nonlinearity concept into the total Lagrangian dynamic formula- order to adequate the constitutive relation to any finite element
tion based on positions. This section takes advantage of the kinematic, avoiding volumetric locking. A brief description of the
previous equations to introduce elastoplasticity into the nonlinear main equations is given for both frame and connections.
dynamic problem and briefly presents the multi-linear elastoplas-
tic evolution model, adopted for both frame end semi-rigid 4.2.1. Frame element
connections. Although the developed displacements are high, the strain level
present in our applications is relatively small. Therefore, the Green
4.1. Equilibrium equation strain approximates the linear strain and the second Piola–Kirchhoff
stress can be used in place of the Cauchy stress. Following this
The difference between the elastic and inelastic procedures is reasoning we adopt the additive strain decomposition as
the way the potential energy of external forces is transferred to the
deformed body and dissipated during the dynamic process. E ¼ Ee þ Ep or Ee ¼ E Ep ð26Þ
6 H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15
in which E e and E p are, respectively, the elastic and plastic parts of Using these variables we calculate the stress level considering
E. Therefore, the free energy potential can be written as ΔEp ðinstead of dEp Þ as the main unknown, as follows:
1 1 S ðn þ 1Þ ¼ ℭ:ðt E ðn þ 1Þ t E pn ΔE p Þ ¼ t S ðn þ 1Þ ℭ:ΔE p ð36Þ
Ψ ðE Ep ; αÞ ¼ Ee : ℭ : Ee þ hα2 ð27Þ
2 2 t
where the trial stress S ðn þ 1Þ is known. Using this value we
where ℭ is the elastic constitutive tensor and h is the isotropic calculate the trial yielding expression
hardening parameter. qffiffiffiffiffiffi
f ðn þ 1Þ ¼ t J 2 t χ Sy
The second Piola–Kirchhoff stress and the thermodynamic t
ð37Þ
force χ are written as
t
If f ðn þ 1Þ r 0 the step is elastic and the trial variables are the
∂Ψ t
S¼ ¼ ℭ : E e ¼ ℭ : ðE E p Þ ð28Þ correct ones. However, if f ðn þ 1Þ 4 0 then the step is elastoplastic
∂E and f ðn þ 1Þ should be zero, resulting in
∂Ψ Δλ ¼ t f ðn þ 1Þ =ðG þHÞ ð38Þ
χ¼ ¼ h α ð29Þ
∂α
Finally, the correct stress tensor to be introduced in Eq. (25) results
As mentioned after Eq. (25), in order to eliminate α from
by substituting (38) into (32) and using (32) and (36), i.e.
equilibrium equation it is necessary to relate E p and α. In classical
0 1
formulations this is done introducing a plastic potential F S; χ
[57] for which the plastic flow is given by ∂Ψ t f ðn þ 1Þ Bt S ðn þ 1Þ C
¼ S ¼ tS ℭ:B
@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi C
A ð39Þ
∂E ðG þ HÞ G tr ð t
S Þ
∂F ∂F 2 J 2ðn þ 1Þ þ P pffiffiffiffiffiffiffiffiffiffiffiffi
t
E_ ¼ λ_ α_ ¼ λ_
p t
J 2ðn þ 1Þ I
and ð30Þ 9k
∂S ∂χ
Moreover, the constitutive elastoplasticity tensor is written as [56]
or in infinitesimal notation !
∂S ∂2 Ψ t
S :ℭ
∂F ∂F ℭep ¼ ¼ ¼ ℭ E : η t pffiffiffiffi ¼ ℭ ℭp ð40Þ
dE p ¼ dλ and dα ¼ dλ ð31Þ ∂E ∂E∂E S : ℭ : η þ 2 J2 H
∂S ∂χ
where λ is the plastic multiplier. In the adopted formulation
[48,56] Eqs. (29) and (31) are replaced by α ¼ λ and 4.2.2. Elastoplastic model for semi-rigid connections
dE ¼ ηdλp
and dχ ¼ hdλ ð32Þ In this item the previous general elastoplastic procedure
(multi-linear) is simplified to accomplish the semi-rigid connec-
for which tion. Firstly, the notation of Eqs. (15) and (23) is changed to
!
S trðSÞ η η η
G G R3αβ ¼ ðY 3α Y 3β Þ ð41Þ
η ¼ Dp :S ¼ pffiffiffiffi þ I and Dp ¼ pffiffiffiffiðℭp Þ 1 ð33Þ
J2 2G 9kp J2 η
in which R3αβ is the relative rotation between frame bars α and β
t
In Eq. (33) J 2 ¼ 1=2½S : S is the second invariant of the deviatory at a connection node η. From now on, to simplify the develop-
ments, this relative rotation will be called simply R. This variable is
stress S and ℭp is an elastic constitutive tensor, similar to ℭ
clearly one-dimensional and is separated into elastic and plastic
changing the ν by νp . Moreover k ¼ E=½3ð1 2νp Þ is the elasto-
p
parts, as
plastic bulk modulus. Following this strategy, when νp ffi 0:5
isochoric plasticity takes place and when ν ¼ νp the plastic flow R ¼ Re þ Rp or Re ¼ R Rp ð42Þ
occurs in the same direction of elastic flow. In this work we adopt The free Helmholtz energy potential, implicit in Eq. (24), is written
ℭp ¼ ℭ, defined by the second derivative of expression (16) with as
reference to strain. This choice releases any possible locking
1 1
related to Reissner kinematic. ΘSR ðR Rp ; αÞ ¼ kðRe Þ2 þ hα2 ð43Þ
To complete the elastoplasticity procedure one defines the 2 2
yielding failure expression as where k is the elastic stiffness of the connection and h its isotropic
pffiffiffiffi hardening. Following the previous item one writes the internal
f ¼ J 2 χ Sy ð34Þ moment and the thermodynamic force related to the internal
pffiffiffi variable α as
in which Sy ¼ σ y = 3 is the initial size of the Von-Mises surface
∂Θ
SR
(f ¼0) with σ y being the yielding uniaxial stress. M¼ ¼ kRe ¼ kðR Rp Þ ð44Þ
It is important to know that the plastic multiplier should satisfy ∂R
the Khun–Tucker conditions, related to the yielding surface (34), i.e.
∂θ
dλ Z 0; f r 0; dλ f ¼ 0
χ¼ ¼ hα ð45Þ
ð35Þ ∂α
which means that if f r 0 then dλ ¼ 0 and no plastic evolution occurs In the case of semi-rigid connections, the elastic limiting
and if, for some situation, one finds f 40 then dλ 4 0 should be moment M y 4 0 is used to write the failure expression as
achieved in order to guarantee the equality f ¼ 0. f ¼ jMj M y χ r 0 and f ¼ jMj M y þ hα r0 ð46Þ
In terms of incremental solution we assume isotropic hard-
ening (h), constant by parts, and a typical interval t n ; t n þ 1 for Following the steps described in the previous item, one achieves
which n is a previous (solved) step and n þ 1 is the current step. t
f
Therefore, one writes ΔRp ¼ signðMÞ: ð47Þ
kþh
t
E ðn þ 1Þ -elastic trial of the total strain in which sign represents signal. Moreover the tangent modulus is
t
E pðn þ 1Þ ¼ E pn -accumulated plastic strain ðassumed as trialÞ obtained as
t
λðn þ 1Þ ¼ λn -trial internal variable ∂2 Θ
SR
k:h
kt ¼ ¼ ð48Þ
t
χ ðn þ 1Þ ¼ χ n -trial thermodynamic internal force ∂R2 k þh
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 7
5. Newmark and Newton–Raphson procedures From Eq. (54) the Newton–Raphson procedure to solve the non-
linear equation (51) gives
In this section we combine the Newmark time integration and ! ! !0 ! !0
the Newton–Raphson algorithm to solve the nonlinear dynamic H U Δ Y s þ 1 ¼ g ð Y s þ 1Þ and Δ Y s þ 1 ¼ H1 U !
g ð Y s þ 1Þ ð55Þ
equilibrium described by Eqs. (25) or (22), rewritten here joining
! !0 !
the external conservative loads into one single vector F : where Y s þ 1 is a trial position (usually Y s as the first choice) for
! !int !int ! !€ !_ ! ! !0 !
g ¼ F þ M F þ M U Y þ DU Y ¼ 0 ð49Þ Y s þ 1 used in Eq. (51) to calculate g ð Y s þ 1 Þ. Solving Δ Y s þ 1 one
!
! ! calculates a new trial for Y s þ 1 as
in which g is a vector that is null if Y is the right position trial
!int !int ! !0 !
used to calculate internal forces F and M , if not it turns into Y Sþ1 ¼ Y sþ1 þΔ Y sþ1 ð56Þ
the unbalanced mechanical force. Expression (49) represents the Acceleration and velocity are corrected, for each iteration, by,
dynamic equilibrium equation at any time and has to be solved. In !
order to do so the first step is to write this equilibrium for a ! Y Sþ1 ! !_ _
! !€ !€
Y€ S þ 1 ¼ T S and Y S þ 1 ¼ Y S þ Δtð1 γ Þ Y S þ γΔt Y S þ 1
specific instant t s þ 1 , as follows: βΔt 2
ð57Þ
∂Π ∂Ψ ∂Θ
SR
!
g jS þ 1 ¼ ! ¼ ! þ !
∂Y ∂Y ∂Y
The stop criterion when a chosen tolerance (TOL) is satisfied, is
Sþ1 Sþ1 Sþ1
! !€ !_ ! !! !
F Sþ1 þMU Y Sþ1 þDU Y Sþ1 ¼ 0 ð50Þ jj g ð Y s þ 1 Þjj r TOL or jjΔ Y s þ 1 jj r TOL ð58Þ
in which internal forces are assumed as the derivatives of In order to conclude the description of the formulation the
Helmholtz free energy potentials regarding positions. Substituting second derivatives of the Helmholtz free energy potentials of
the Newmark approximations [52,55] into Eq. (50) gives frame elements and semi-rigid connections regarding positions
are presented. From Eqs. (25), (40), (48) and (53) one writes
∂Π ∂Ψ ∂Θ
SR
!
g jS þ 1 ¼ ! ¼ ! þ ! ∂2 Ψ ∂E ∂2 Ψ ∂E ∂Ψ ∂2 E ∂E ∂E
∂Y ∂Y ∂Y
ep
Sþ1 Sþ1 Sþ1 ! ! ¼ ! : ∂E∂E : ! þ ∂E : ! ! ¼ ! : ℭ : !
∂Y ∂Y ∂Y ∂Y ∂Y ∂Y ∂Y ∂Y
! M ! ! !
F Sþ1 þ U Y S þ 1 M U T S þD U R S ∂2 E
βΔt 2 þS : ! ! ð59Þ
∂Y ∂Y
γD ! !
þ U Y S þ 1 γΔtD U T S ¼ 0 ð51Þ
βΔt
∂2 Θ ∂2 Θ
SR SR
! ! ! ! ¼ ∂R∂R ¼ kt ð60Þ
where vectors T s and R s represent the dynamic contribution of ∂Y ∂Y
the past as
! _
!
! YS YS 1 €
! ! _
! !€
TS¼ þ þ 1 YS and R S ¼ Y S þ Δtð1 γ Þ_ Y S
βΔt 2 βΔt 2β
ð52Þ
and β and γ are the usual Newmark parameters [11].
!!
Eq. (51) can be understood simply by g ð Y S þ 1 Þ ¼ 0 and is
!
clearly nonlinear in Y S þ 1 . A Taylor expansion to solve this non-
linear equation is developed. The second derivative of the total
energy potential is then given by
∂2 Π ! ∂2 Ψ ∂2 Θ M γD
! ! ¼ H ¼ ∇ g ¼ ! ! þ ! ! þ þ ð53Þ
∂ Y ∂ Y S þ 1 ∂ Y ∂ Y S þ 1 ∂ Y ∂ Y S þ 1 βΔt
2 βΔt
The necessary expressions to calculate the last term of Eq. (59) is rotation, elastoplastic connections, elastoplastic behavior of mate-
presented in Appendix B. rials and all aspects working together.
Fig. 12. Connection moment rotation graphics – (a) Monotonic and (b) cyclic.
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 9
k ¼ 12 kN cm with elastic limit M y ¼ 3 kN cm. Fig. 12b shows the a degradation spreads over the beam; moreover the load stabilizes
connection behavior when subjected to a cyclic loading. after the displacement of 7 cm. After the imposed displacement of
As one can see in Fig. 12 the multi-linear strategy is capable of 2.2 cm a reduction in the load level occurs for the softening case
accurately reproducing any elastoplastic curve. and, consequently, in the bending moment. Moreover, no bending
Fig. 13 shows partial positions for 15 equally spaced loading moment spreads along the beam.
levels, demonstrating that the nonlinear geometric description is
exact. This level of rotation is not accomplished by formulations
6.4. Elastoplastic column
that adopt second order geometrical descriptions.
Fig. 15. Curves: stress strain and force displacement (under the load). (a) Stress–strain relation and (b) load versus displacement.
10 H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15
Fig. 16. Bending moments diagrams for various load levels. (a) Perfect elastoplastic and (b) softening.
Fig. 17. Eccentric column and results. (a) Geometry and (b) load versus central displacement.
Fig. 18. Analized frames, adapted from [57]. (a) Clamped and (b) simple supported.
The load P (Fig. 18) grows monotonically until the critical load the difference among the exact geometrical description and
shows up. Figs. 20 and 21 show that the results presented by our second order approximations.
formulation agree with the ones given by references. For semi- For connections A and B the displacement levels are very small
rigid connections the differences in results are less than 2% for all and failure occurs in an abrupt way. In order to show up the
cases. For rigid connection the difference is about 6%, explained by horizontal plateau (for A and B), Figs. 20 and 21, we used lateral
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 11
loads of 0:0015P and 0:003P instead of 0:001P and 0:002P. loading. The real event chosen is the Superstation Hill in 1987. The
However, this procedure does not change the critical load value. original unscaled time series record has been extracted from [61].
Both vertical (UP) and horizontal (360) accelerations are trans-
formed into displacements (cm) and applied as base movements
6.6. Seismic loading imposed on a tall frame
in the time domain analysis. The structure is a 25 storeys building,
Fig. 22, modeled with (a) free connections (articulated joints),
This example demonstrates the capability of the proposed
(b) rigid connections, (c) semi-rigid connections, (d) semi-rigid
formulation to model the behavior of structures to real seismic
connections with bracing and (e) semi-rigid connections with
bracing and pendulum.
All beams are made of wide flange W200 46. The columns are
divided as follows: (i) from the 1st to the 5th floor W610 155,
(ii) from the 6th to the 10th floor W610 140, (iii) from the 11th to
the 15th W610 113, (iv) from the 16th to the 20th floor
W460 74 and (v) from the 21st to the 25th W200 46. Bracings
are 3:606 m long and pose square cross section of 5:83 cm side.
The adopted material is structural steel with simplified proper-
ties depicted in Fig. 23. The adopted density of columns is
ρst ¼ 7000 kg/m3. For beams we adopt an equivalent density
ρeq ¼ 15; 333 kg/m3, in order to include the mass influence of a
reinforced concrete slab with 10 cm of height and 4 m of width.
The considered vertical load q ¼ 4088 kN/m is uniformly distrib-
uted over beams. The frame is considered simple supported and
the soil moves horizontally and vertically following the real
Fig. 19. Connections bending moment rotation diagrams. earthquake data.
Fig. 25. Progressive collapse of case (a). (a) 2.25 s, 2.5 s 2.75 s and 3 s – enlarged
Fig. 24. Moment–rotation curve for connections. displac. (b) 3 s, 3.25 s, 3.5 s, 3.75 s and 4 s – not enlarged.
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 13
7. Conclusions Acknowledgments
In this study an alternative position based on finite element The authors would like to thank FAPESP (São Paulo Research
formulation is presented and successfully implemented. The for- Foundation) and CNPq (National Council for Scientific and Tech-
mulation is originally presented here, providing all steps for its nological Development – Brazil) for the financial support of this
reproduction. It is capable of accurately solving geometrical and research (FAPESP: 11/15731-5; CNPq: 306152/2013-4).
physical nonlinear dynamic problems of both machines and
structures.
Examples show good behavior of the formulation for various !
Appendix A. Calculating ∂E=∂ Y
applications, covering progressive collapse of structures under
seismic loading. Future works include the implementation of
From expression (14), reproduces as,
damage mechanics, also allowing the analysis of reinforced con-
crete structures developing large displacements and under severe 1 1
E ¼ ðC IÞ ¼ ðAt UA IÞ ðA1Þ
dynamic loading. 2 2
14 H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15
one writes [3] I. Romero, F. Armero, An objective finite element approximation of the
∂E 1 ∂C 1 ∂ n t o 1 ∂ n 0 t 1 t 1 0 1
o kinematics of geometrically exact rods and its use in the formulation of an
! ¼ 2 ! ¼ 2 ! A U A ¼ 2 ! ½ðA Þ U ðA Þ U ½ðA Þ U ðA Þ energy-momentum conserving scheme in dynamics, Int. J. Numer. Methods
∂Y ∂Y ∂Y ∂Y Eng. 54 (2002) 1683–1716.
(" !# [4] M. Greco, H.B. Coda, Positional FEM formulation for flexible multi-body
1 t
1 ∂A
¼ ðA0 Þ t U ! U ½ðA1 Þ U ðA0 Þ 1 dynamic analysis, J. Sound Vib. 290 (2006) 1141–1174.
2 ∂Y [5] G. Jelenic, M.A. Crisfield, Dynamic analysis of 3D beams with joints in presence
" ! #) of large rotations, Comput. Methods Appl. Mech. Eng. 190 (2001) 4195–4230.
∂A1 [6] W.F. Chen, E.M. Lui, Structural Stability: Theory and Implementation, Elsevier;
þ ½ðA0 Þ t U ðA1 Þt U ! UðA 0 1
Þ ðA2Þ
∂Y New York, 1987.
[7] S.E. Kim, H.T. Thai, Nonlinear inelastic dynamic analysis of suspension bridges,
!
In which all terms are known except ∂A1 =∂ Y , detailed as follows: Eng. Struct. 32 (2010) 3845–3856.
" #
∂A1 ϕϖ;1 0
¼ ðA3Þ
∂Y 1ϖ 0 0
" #
∂A1 0 0
¼ ðA4Þ
∂Y 2ϖ ϕ ϖ ;1 0
2( lam
! ) (
lam
)3
h h
6 d þ ξ ½ cos ðϕk θk Þðϕz;1 θz Þϕϖ þ sin ðϕk θk Þϕϖ ;1
lam
sin ðϕℓ θℓ Þϕϖ 7
∂A1 ∂A1 6 2 2 2 7
6 7
¼ ¼6 ( ! ) ( ) 7 ðA5Þ
∂Y 3ϖ ∂θϖ 6 h
lam
h
lam 7
4 lam
d þ ξ ½ cos ðϕk θk Þϕϖ ;1 sin ðϕk θk Þðϕz;1 θz Þϕϖ cos ðϕℓ θℓ Þϕϖ 5
2 2 2
[8] E.M. Lui, A. Lopes, Dynamic analysis and response of semirigid frames, Eng.
in which Y iϖ is the degree of freedom i of node ϖ , component of Struct. 19 (1997) 644–654.
! [9] J.C. Awkar, E.M. Lui, Seismic analysis and response of multistory semirigid
vector Y .
frames, Eng. Struct. 21 (1999) 425–441.
[10] A. Ibrahimbegovic, R.L. Taylor, On the role of frame-invariance in structural
! ! mechanics models at finite rotations, Comput. Methods Appl. Mech. Eng. 191
Appendix B. Calculating ∂2 E=ð∂ Y ∂ Y Þ (2002) 5159–5176.
[11] J. Argyris, H.P. Mlejnek, Dynamics of structures, Texts on Computational
From expression (A2) one writes: Mechanics, 5, North-Holland, Amsterdam, 1991.
8 "
t #
>
>
>
> 0 t
2 U ðA Þ U ∂A
1
U ∂A1
UðA0 Þ 1 þ
> ! !
∂2 E 1< ∂Y ∂Y
! ! ¼ 2> "
t #
ðB1Þ
∂Y ∂Y >
>
> 0 t 2 1
∂ A
U ½ðA1 Þ UðA0 Þ 1 þ ½ðA0 Þ t U ðA1 Þt U ∂2 A 1 0 1
: þ ðA Þ U ! !
> ! ! U ðA Þ
∂Y ∂Y ∂Y ∂Y
∂2 A 1
∂Y 3ϖ ∂Y 3μ
2( !" #) ( )3
h
lam cos ðϕk θk Þϕμ;1 ϕϖ þ cos ðϕk θk Þϕϖ ;1 ϕμ þ lam
h
6 dlam þ ξ cos ðϕℓ θℓ Þϕϖ ϕμ 7
6
6 2 2 sin ðϕk θk Þðϕz;1 θz Þϕϖ ϕμ 2 7
7
ðB2Þ
¼6
6 ( !" #) ( )7
7
6 h
lam sin ðϕk θk Þϕϖ ;1 ϕμ þ sin ðϕk θk Þϕμ ϕμ;1 þ lam
h 7
4 dlam þ ξ sin ðϕℓ θℓ Þϕϖ ϕμ 5
2 2 þ cos ðϕk θk Þðϕz;1 θz Þϕϖ ϕm 2
References [14] Z.H. Zhou, S.L. Chan, Elastoplastic and large deflection analysis of steel frames
by one element per member I: one hinge along member, J. Struct. Eng.-ASCE
[1] G.G. Deierlein, A.M. Reinhorn, M.R. Willford, Nonlinear structural analysis for 130 (4) (2004) 538–544.
seismic design: a guide for practicing engineers, NEHRP Seismic Design [15] W.X. Ren, X.G. Tan, Z.C. Zheng, Nonlinear analysis of plane frames using rigid
Technical Brief No. 4, NIST, USA, 2010. body-spring discrete element method, Comput. Struct. 71 (1) (1999) 105–119.
[2] J.C. Simo, L. Vu-Quoc, On the dynamics of flexible beams under large overall [16] F. Armero, D. Ehrlich, Numerical modeling of softening hinges in thin Euler–
motions – the plane case: part 1, J. Appl. Mech., ASME, 53, 1986849–854. Bernoulli beams, Comput. Struct. 84 (10–11) (2006) 641–656.
H.B. Coda, R.R. Paccola / Finite Elements in Analysis and Design 91 (2014) 1–15 15
[17] D. Ehrlich, F. Armero, Finite element methods for the analysis of softening [41] K.M. Ang, G.A. Morris, Analysis of three-dimensional frames with flexible
plastic hinges in beams and frames, Comput. Mech. 35 (4) (2005) 237–264. beam-column connections, Can. J. Civ. Eng. 11 (1984) 245–254.
[18] D.W. White, Plastic hinge methods for advanced analysis of steel frames, J. [42] N. Kishi, W.F. Chen, Steel connection data bank program, Structural Engineer-
Constr. Steel Res. 24 (2) (1993) 121–152. ing Report no. CE-STR-86-18, School of Civil Engineering, Purdue Univ., West
[19] W.F. Chen, Y. Goto, J.Y.R. Liew, Stability Design of Semi-rigid Frames, John Lafayette, IN, 1986.
Wiley & Sons, USA, 1996. [43] N. Kishi, W.F. Chen, Data base of steel beam-to-column connections, Structural
[20] A. Landesmann, E.M. Batista, Advanced analysis of steel framed buildings to Engineering Report no. CE-STR-93-15, School of Civil Engineering, Purdue
brazilian standard and Eurocode-3, J. Constr. Steel Res. 61 (2005) 1051–1074. Univ., West Lafayette, IN, 1986.
[21] S.L. Chan, Z.H. Zhou, Elastoplastic and large deflection analysis of steel frames [44] K. Zhu, F.G.A. Al-Bermani, S. Kitipornchai, B. Li, Dynamic response of flexibility
by one element per member. II: three hinges along member, J. Struct. Eng.- jointed frames, J. Struct. Div. ASCE, 17, 1995575–580.
ASCE 130 (2004) 545–553. [45] M. Sekulovic, M. Nefovska-Danilovic, Contribution to transient analysis of
[22] C. Ngo-Hu, S.E. Kim, J.R. Oh, Nonlinear analysis of space steel frames using inelastic steel frames with semi-rigid connections, Eng. Struct. 30 (2008)
fiber plastic hinge concept, Eng. Struct. 29 (2007) 649–657. 976–989.
[23] A.R. Alvarenga, R.A.M. Silveira, Second-order plastic-zone analysis of steel [46] J. Bonet, R.D. Wood, J. Mahaney, P. Heywood, Finite element analysis of air
frames Part I: numerical formulation and examples of validation, Lat. Am. supported membrane structures, Comput. Methods Appl. Mech. Eng. 190 (5–
J. Solids Struct. 6 (2009) 323–342. 7) (2000) 579–595.
[24] P. Avery, M. Mahendran, Distributed plasticity analysis of steel frame struc- [47] H.B. Coda, M. Greco, A simple FEM formulation for large deflection 2D frame
tures comprising non-compact sections, Eng. Struct. 22 (2000) 901–919. analysis based on position description, Comput. Methods Appl. Mech. Eng. 193
[25] F. Gruttmann, R. Sauer, W. Wagner, Theory and numerics of three-dimensional (33–35) (2004) 3541–3557.
beams with elastoplastic material behavior, Int. J. Numer. Methods Eng. 48 [48] A.S. Botta, R.R. Paccola, W.S. Venturini, H.B. Coda, A discussion on volume
(2000) 1675–1702. change in the plastic phase, Commun. Numer. Methods Eng. 24 (2008)
[26] H.T. Thai, S.E. Kim, Practical advanced analysis software for nonlinear inelastic 1149–1162.
dynamic analysis of steel structures, J. Constr. Steel Res. 67 (3) (2011) 453–461. [49] H.B. Coda, R.R. Paccola, A positional FEM formulation for geometrical non-
[27] H.T. Thai, S.E. Kim, Nonlinear inelastic analysis of space frames, J. Constr. Steel linear analysis of shells, Lat. Am. J. Solids Struct. 5 (3) (2008) 205–223.
Res. 67 (4) (2011) 585–592. [50] H.B. Coda, A solid-like FEM for geometrically nonlinear 3D frames, Comp.
[28] H.T. Thai, S.E. Kim, Second-order inelastic dynamic analysis of steel frames Methods Appl. Mech. Eng. 198 (47–48) (2009) 3712–3722.
using hinge method, J. Constr. Steel Res. 67 (10) (2011) 1485–1494. [51] H.B. Coda, R.R. Paccola, Improved finite element for 3D laminate frame
[29] E.M. Lui, W.F. Chen, Behavior of braced and unbraced semi-rigid frames, Int. J. analysis including warping for any cross-section, Appl. Math. Model. 34 (4)
Solids Struct. 24 (9) (1988) 893–913. (2010) 1107–1137.
[30] W.S. King, The limit loads of steel semi-rigid frames analyzed with different [52] H.B. Coda, R.R. Paccola, A FEM procedure based on positions and uncon-
methods, Comput. Struct. 51 (5) (1994) 475–487. strained vectors applied to nonlinear dynamic of 3D frames, Finite Elem. Anal.
[31] L.M.C. Simões, Optimization of frames with semi-rigid connections, Comput. Des. 42 (4) (2011) 319–333.
Struct. 60 (4) (1996) 531–539. [53] R.W. Ogden, Nonlinear Elastic Deformation, Ellis Horwood, England, 1984.
[32] P.P.T. Chui, S.L. Chan, Vibration and deflection characteristics of semi-rigid [54] C. Lanczos, The Variational Principles of Mechanics, fourth ed, 1970, Dover
jointed frames, Eng. Struct. 19 (12) (1997) 1001–1010. Publications; New York.
[33] L. Xu, Second-order analysis for semi-rigid steel frame design, Can. J. Civ. Eng. [55] H.B. Coda, R.R. Paccola, Unconstrained finite element for geometrical non-
28 (2001) 59–76. linear dynamics of shells, Math. Probl. Eng. (2009) (Art. 575131).
[34] M. Sekulovic, R. Salatic, Nonlinear analysis of frames with flexible connections, [56] Ronaldo Rigobello, Humberto Breves Coda, Jorge Munaiar Neto, Inelastic
Comput. Struct. 79 (11) (2001) 1097–1107. analysis of steel frames with a solid-like finite element, J. Constr. Steel Res.
[35] T.S. Kruger, B.W.J. van Rensburg, G.M. du Plessis, Nonlinear analysis of 86 (2013) 140–152.
structural steel frames, J. Construct. Steel Res. 34 (1995) 285–306. [57] J.C. Simo, T.J.R. Hughes, Computational Inelasticity, Springer; New York, 1998.
[36] S.L. Chan, P.P.T. Chui, Nonlinear Static and Cyclic Analysis of Steel Frames with [58] K.M. Hsiao, R.T. Yang, A.C. Lee, A consistent finite element formulation for
Semi-Rigid Connections, Elsevier, Oxford, 2000. nonlinear dynamic analysis of planar beam, Int. J. Numer. Methods Eng. 37
[37] W.F. Chen, N. Kishi, Semi-rigid steel beam-to-column connections: Data base (1994) 75–89.
and modeling, J. Struct. Div. ASCE 115 (1) (1998) 105–119 (1989). [59] H.A. Elkaranshawy, M.A. Dokainish, Corotational finite element analysis of
[38] K.M. Abdalla, W.F. Chen, Expanded database of semi-rigid steel connections, planar flexible multibody systems, Comput. Struct. 54 (1995) 881–890.
Comput. Struct. 56 (4) (1995) 553–564. [60] L. Pinheiro, R.A.M. Silveira, Computational procedures for nonlinear analysis of
[39] R.M. Richard, B.J. Abbott, Versatile elastic–plastic stress–strain formula, J. Eng. frames with semi-rigid connections, Lat. Am. J. Solids Struct. 2 (2005)
Mech., Div. ASCE 101 (4) (1975) 511–515. 339–367.
[40] M.J. Frye, G.A. Morris, Analysis of flexibly connected steel frames, Can. J. Civ. [61] 〈http://peer.berkeley.edu/peer_ground_motion_database/spectras/21713/
Eng. 2 (3) (1975) 280–291. unscaled_searches/84357/edit〉.