Вы находитесь на странице: 1из 18

Engineering Structures 212 (2020) 110526

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Nonlinear dynamic analysis of frames with isotropic and kinematic T


hardening/softening
S. Mohammadzadeh, H. Moharrami

Faculty of Civil and Environmental Engineering, Tarbiat Modares University, Tehran, Iran

ARTICLE INFO ABSTRACT

Keywords: Static nonlinear analysis of frames and plane stress/strain structures via mathematical programming algorithms
Nonlinear dynamic analysis has been addressed by many researchers. It has been shown that the method of Dissipated Energy Maximization
Duhamel's integral (DEM) is an efficient algorithm for nonlinear static analysis. This study extends the application of DEM method
Lumped plasticity to the nonlinear dynamic analysis of frames considering bending moment-axial force interaction. The nonlinear
Mathematical programming
static analysis algorithm that is the basis of nonlinear dynamic analysis, and corresponding assumptions in-
Isotropic
Kinematic
cluding linear-kinematics, lumped-plasticity, piecewise-linear yield function, and the associated flow rule are
Softening briefly explained. The dynamic analysis that is carried out by Duhamel integral method is fully formulated. The
Hardening proposed method traces the nonlinear equilibrium path through a linear mathematical programming process and
makes modifications on response of Duhamel's integral to yield the nonlinear response. In addition, the
Bauschinger effect (kinematic hardening) is included in the formulations to get more realistic responses. Several
examples illustrate the efficiency and accuracy of the proposed method. It has been shown that the method is the
most accurate and fastest algorithm compared to conventional methods of nonlinear dynamic analyses.

1. Introduction in every iteration. Since this procedure is very time consuming, espe-
cially for a structure with considerable number of degrees-of-freedom,
Most of current building structures are moment resisting frames in 1969, Oden [4] proposed the Modified-Newton-Raphson (MNR)
with/without lateral load resisting systems. To capture the real re- method; this method divides the load intensity into several load in-
sponse of structures to lateral loads it is necessary to carry out a non- crements and updates the stiffness matrix at the beginning of every load
linear dynamic analysis. However, current nonlinear analysis algo- increment and uses the same stiffness matrix for the next iterations of
rithms are time consuming and most of engineers prefer not to do such that increment. Similarly, many methods based on force control, dis-
an analysis. Therefore, developing user friendly algorithms for non- placement control, arc-length control, work control, generalized dis-
linear dynamic analysis of frames that are fast enough to attract design placement control, orthogonal residual procedure, and energy mini-
engineers, may be a very important issue for researchers in the field. mization schemes have been developed to overcome a specific
There are many methods to determine the response of a frame shortcoming(s) of other methods [5]. Conventionally, a new method
structure with elastoplastic behavior. Two important branches of overcomes the shortcoming(s) of other methods, but it suffers from
methods in this field are incremental-iterative methods and mathema- different shortcoming(s). Being very time consuming, having low ac-
tical programming-based methods. curacy, incapability of capturing the bifurcation in equilibrium path,
Incremental iterative methods along with a direct time integration singularity of stiffness matrix in certain circumstances, and increasing
algorithm are the most popular methods in commercial programs of number of iterations in some situations are some of shortcomings of
nonlinear dynamic analysis [1–3]. Contrary to pure incremental these methods [5–11].
methods in which errors become accumulated and considerably high, Application of plasticity on incremental-iterative methods especially
the incremental-iterative methods try to eliminate the accumulation of in frames may be followed by two basic approaches: (1) lumped plas-
errors. For instance, Newton-Raphson (NR) method is one of basic ticity and (2) distributed plasticity. The former one considers the plastic
methods of this type that checks the equilibrium in every increment and behavior of material concentrated in some predefined sections of the
tries to diminish the unbalance forces at the end of each iteration. The structure and treats the rest of structure as elastic, while the latter one
original Newton-Raphson method, updates the tangent stiffness matrix distributes the plastic behavior through use of numerous fiber elements


Corresponding author.
E-mail addresses: mohammadzadeh.saeed@modares.ac.ir (S. Mohammadzadeh), hamid@modares.ac.ir (H. Moharrami).

https://doi.org/10.1016/j.engstruct.2020.110526
Received 14 May 2019; Received in revised form 10 March 2020; Accepted 10 March 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

in the structure. Distributed plasticity may result in high accuracy MU¨ + CU + KU = P (1)
provided that the difficulties associated with numerical integration, in which U is the displacement vector, U is the velocity vector, Ü is
required for the computation of element forces and stiffness matrix, are the acceleration vector, M is the mass matrix, C is the damping matrix,
overcome and the problem of required memory associated with nu- K is the stiffness matrix and P is the external exciting force vector
merous fiber elements especially in large structures, are solved (see during the time step t .
[12–14]). Nevertheless, there are studies in the literature that prove the Assume that the solution is known up to the time t and the solution
usefulness of lumped plasticity approach [15–18] due to the simple of time t + t is to be calculated. The Duhamel’s integral is used to
application and extremely less computational cost which still provides solve the problem linearly. Since Duhamel’s integral is not a vector but
promising accuracy approved by many regulations. is a scalar solution, a modal decomposition should be performed. Eq.
Another approach in nonlinear analysis is based on mathematical (2) presents the decoupled modal form of Eq. (1) with a scaled unit
programming. Started by the limit state analysis of structures, mathe- modal mass:
matical programming methods have experienced many modifications
through the time. Maier, proposed a method which used mathematical M U¨ + C U + K U = P (2)
programming in static analysis of structures with perfect-plastic beha- where
vior [19,20]. Later on, he linearized the yield surface to be able to
consider the interaction of internal forces in frame structures [21]. In 10
M = T ·M· = I = 01
his method, the product of plastic multipliers in yield conditions, which
(3)
were nonlinear terms, was adopted as object function in optimization
problem; then he used Quadratic Programing (QP), to find plastic 2 1 0
1
multipliers. Due to the large QP solution effort and being unable to C = T ·C · = 0 2 2 2
consider local unloading, and in order to extend the application of the (4)
method, the QP method was substituted with Linear Complementarity
Programming (LCP) and Restricted Basis Linear Programming (RBLP) 2
1
0
methods [22]. In these methods the nonlinear complementarity con- K = T ·K · = 0
2
2
ditions are implicitly satisfied during the solution procedure. This (5)
modification results in a more efficient and speedy method and enables
capturing the local unloading (even in the presence of softening). Maier U= ·U ; U = ·U ; U¨ = · U¨ ; P = T ·P
(6)
et al. [22] briefly described the fully-holonomic procedure, but it was in which is the mass scaled mode shape matrix. Consider a scalar
fully discussed by Ardito et al. [23,24] for continuum problems. They equation related to one mode:
generalized the classical limit state analysis for a combination of limit
and deformation analysis and maximized the load factor as an objective u¨ + 2 u + 2u =p (7)
function under linear and complementarity conditions to assess the The following equation shows numerical solution of Duhamel’s in-
structural safety with respect to diverse alternative limit states in a fully tegral for Eq. (7):
holonomic manner.
e ti + 1
Tangaramvong, and Tin-Loi [25] utilized, the penalty function u (ti + 1) = [AD (ti + 1)sin( D ti + 1) BD (ti + 1)cos( D ti + 1)]
method to solve the mathematical programming problem. In this m D (8)
method the penalty parameter has to be chosen carefully to avoid nu- in which:
merical instabilities.
Recently, Mahini et al. formulated the Dissipated Energy p (ti + 1 ) p (ti + 1 )
AD (ti + 1) = AD (ti) + p (ti) ti I1 + I4
Maximization (DEM) [26] approach to determine non-holonomic so- ti + 1 ti + 1 (9)
lution of frame structures with material nonlinearity. They overcame
the aforementioned shortcomings associated with other mathematical BD (ti + 1) = BD (ti ) + p (ti ) ti
p (ti + 1 )
I2 +
p (t i + 1 )
I3
programming techniques. The method applies the complementarity ti + 1 ti + 1 (10)
conditions implicitly during the pivot finding by introducing some re-
where
strictions in solution procedure that makes this method capable of
finding the non-holonomic response as well as capturing the local un- ti e ti
I1 = e cos( )d = ( cos( )+ D sin( D ))
loading. This present study employs the well-known Duhamel's integral
D D
ti 1 ( )2 + D
2 ti 1
to extend the application of DEM method in nonlinear dynamic analysis (11)
of frame structures. In order to find the nonlinear responses in each
time step, some modifications related to the nonlinear responses are I2 =
ti
e sin( )d =
e
( sin( )
ti
D D Dcos( D ))
made on the response of the Duhamel's integral. There is also another ti 1 ( )2 + D
2 ti 1
modification made on the influence matrix (to be explained later), to (12)
make it applicable on dynamic problems. In addition, the Prager's rule ti
has been employed to introduce the Bauschinger effect (kinematic I3 =
ti 1
e sin( D )d
hardening) to the formulations of the proposed method.
D ti
= I '2 + I '1
( )2 + D
2 ( )2 + D
2 ti 1 (13)
2. The proposed method
ti
I4 = e cos( D )d
Two stages are considered in the proposed method. ti 1

D ti
= I '1 I '2
( )2 + 2 ( )2 + 2 ti 1 (14)
2.1. Stage 1: linear solution
D D

where D = 1 + , I'1 and I'2 refer to I1 and I2 without substitution of


The first stage involves linear solution of the equation of motion limit points (ti 1, ti )and:
using Duhamel's integral. Consider the equation of motion, which is
p (ti + 1) = p (ti + 1) p (ti ) (15)
written in the following form:

2
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

It should be noted that, due to the fact that values of I1 to I4 are


constant (for constant Δt) during the whole analysis process, they can
be computed in the pre-analysis phase and stored for future use.

2.2. Stage 2: modification of linear response

Assume that in a number of sections, due to linear analysis the in-


ternal forces exceed their limit capacities. In this situation, the ex-
ceeded forces should be released and redistributed to the whole struc-
ture. This is done by adding the required deformations at these sections.
Suppose that the force vector in excess of plastic capacity of a section at
time t + t is Ftp+ t . One necessary component in determining the
value of Ftp+ t is the influence matrix ( fv ) which is used to find the
intensity of released forces and brings the linear responses back on the
equilibrium track.

2.2.1. Influence matrix


Fig. 1. Hexagonal yield surface.
The Influence Matrix fv , is composed of internal forces of the whole
structure due to impose of a unit displacement in certain direction at
the critical sections one at a time. It is calculated statically for a static deformation vector is perpendicular to the corresponding yield surface.
analysis or dynamically for a dynamic analysis in a pre-analysis phase To utilize the Simplex method for nonlinear analysis, the yield surface
and is constant during the whole analysis process. is linearized. For example, a hexagonal yield surface as presented in
Generally there are three possible types of hinges in a section cor- Fig. 1 is defined at each node of I or WF sections.
responding to a 2D frame, namely axial, shear and bending deforma- Fig. 1 contains M̄s and N̄s which are equal to M / Mp and N /Np , re-
tions. Ignoring plasticity in shear, for each critical section s with the spectively; in which Mp and Np are the plastic moment and plastic axial
potential of being plastic hinge, two influence vectors say fvA, s and fvB, s , force, respectively. According to Fig. 1, matrix s for any section (s) is
corresponding to Axial and Bending deformations, are calculated and defined as follows:
located at the corresponding columns of the influence matrix fv . In fact 1 1 1 1
fvA, s and fvB, s are respectively, the vector of internal forces in all pre- Np 0 Np Np 0 Np
defined sections due to axial and bending deformation in section s. For = 0 0
s 0 1
0 0 1
0
compatibility of matrix calculations, the influence matrix fv will con- (1
1
N¯ 0) M Mp (1
1
N¯ 0) M (1
1
N¯ 0) M Mp (1 N¯ 0 ) M
1

tain zeros corresponding to shear deformation. Its general shape is as p p p p


3× 6 (19)
follows: Following Eq. (19), the overall matrix is defined; as follows:

fv = f vA,1 0 f vB,1 fvA,2 0 f vB,2 fvA, s 0 fvB, s f vA, m 0 fvB, m 1 [0]3 × 6 : [0]3 × 6
[0]3 × 6 2 : [0]3 × 6
(16) = :
: : :
[0]3 × 6 [0]3 × 6 :
in which m is the number of critical sections in the structure. m 3m × 6m (20)
In order to find the influence vector in a dynamic analysis, one
The increment of plastic deformation vector is calculated as an in-
needs to impose a unit local deformation at the corresponding section
ternal product of unit plastic deformation matrix by corresponding in-
and excite the whole system with the internal forces arisen from that
cremental plastic multiplier vector. i.e.:
unit deformation for one time step.
As was stated before, the fvA, s , which is the influence vector of axial p
t+ t = Xt + t (21)
deformation on sth section, reflects the internal forces at all critical
sections and it is comprised of m sub-vectors as follows: in which, Xt + t is the vector (yet unknown) of plastic multipliers
and is defined; as follows:
fvA, s , 1
X1t + t x s1
f vA, s , 2 x s2
fvA, s = Xt + = X 2t + t ; X st + = and Xt + = Xt + Xt
: t t t t
: :
f vA, s , m X mt+ x s6
(17) t 6m × 1 6× 1 (22)

Noting that there are three internal force components at each sec- Using the influence matrix, the force vector in excess of plastic ca-
tion, the sub-vector fvA, s , i has the following form: pacity can now be formulated; as follows:
fvA, s , i = [ni , vi , mi ]T (18) Ftp+ t = fv . . Xt + t (23)

In Eq. (18) ni, vi, and mi stand for axial, shear, and moment at It should be noted that the increments of plastic multipliers are to be
section i due to unit axial deformation at section s. Therefore each calculated via linear programming. One important equation that should
column of influence matrix is a vector of order 3 m and the influence be met in all sections of all elements is the yield function equations:
matrix is a matrix of order 3m × 3m . Y= T (F
intt + Fintt + t )+ T. fv . Xt + t {0} (24)
Similar to Chen et al. [27] who suggested sectional constitutive
model to account for a combined axial force and bending moment in a in which and are the variable parameter related to the hard-
section hinge, in this paper, a sectional yield surface for the combina- ening/softening (evolution) rule governing the material behavior. In
tion of internal forces was considered. Based on this assumption, plastic this paper, four different evolution rules have been formulated through
hinge in a section may not occur unless the combination of axial force the proposed method: Perfect Plasticity, Isotropic evolution, Kinematic
and bending moment reach to the plastic yield surface . As an analogy evolution, and Combined evolution.
to the flow rule hypothesis, it is assumed that the increment of plastic A complementarity condition that must also be met in all sections is

3
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

that the product of yield function and the increment of the plastic and X̄t + t is a vector containing the evolution variables of each section
multiplier is zero. i.e. either the force combination at a section is such and is arranged in the following form:
that it satisfies the plastic hinge condition, (Y = 0) and there is an
increment on plastic deformation ( Xt + t 0) or, the section is not a X¯ 1t + t
X¯ 2t + x¯ s1
plastic hinge (Y<0) and there is no increment on plastic deformation X¯ t + t = t ; X¯ s t + t =
: x¯ s2
( Xt + t = 0) . This complementarity condition can be written as follows: 2× 1
X¯ mt + t 2m × 1 (30)
Xt + t
TY =0 (25)
x̄ s1 and x̄ s2 are defined in Fig. 2b. According to this figure, the three-
Another complementarity condition that also should be met is that phase evolution rule can be formulated using the definition of
every time the problem is solved, only one yield surface can be acti- equivalent plastic strain. The equivalent plastic strain in section (s) at
vated. This can be written as follows: time t + t , is defined as follows:
x sp x sq = 0; p q in which: p , q = 1, 2, 3, , 6; s = 1, 2, 3, ,m s
p
= Q¯s X st + t (31)
(26)
in which: Q̄s = [ ns1, ns2, ns3, , ns6 ]1 × 6 and nsj is the magnitude of normal
where x sp refers to the increment of plastic multiplier related to the p- vector to the jth yield surface in critical section s. According to Fig. 2b,
th yield surface in s-th critical section. the isotropic evolution variables (x̄ s0 , x̄ s1, x̄ s2 ) are associated with three
more conditions, that should be met on each section in the analysis
Case 1: (Perfect Plasticity). In case of perfect plasticity, is a unit vector
process.
({1}3m × 1) and is a zero vector ({0}3m × 1). To find the increments of
plastic multipliers that make the yield function (18) satisfy in all x¯ s0 p
s1
sections, an optimization problem is established. According to a recent x¯ s1 p
s2
p
s1
study by Mahini et al. [26] the objective function of such an p
x¯s0 + x¯ s1 + x¯ s2 = s (32)
optimization problem can be taken as the total dissipated energy:

max({1}T Xt + t )
subjected to: Solving the last equation for the evolution variable x̄ s0 and replacing
T (F the result into the first inequality, the constraints will be reduced into
Y= intt + Fintt + t ) + T . fv . Xt + t {1} {0}
only two inequalities ( x¯ s1, x¯ s2 ) at each section. The compact matrix form
( Xt + t )T Y = 0 of these additional conditions can be displayed as follows:
x sp x sq = 0; p q in which: p , q = 1, 2, 3, , 6, s = 1, 2, 3, ,m
Xt + {0} Zt + t = Zt + Q Xt + t + W X¯ t + t L {0} (33)
t

(27) in which Q is a 2m × 6m matrix that contains the magnitude of


normal vectors in each critical section and is defined using Qs matrices:
Case 2: (Isotropic evolution). If a three-phase isotropic evolution, such as
the one in Fig. 2, is adopted as the softening model, for an isotropic Q1 O
:
O
evolution, and in Eq. (18) can be defined as follows: O Q2 : O
O Q¯ s1 × 6
Q= O O : ; Qs =
= HX¯ t + t + {1} = HX¯ t + H ¯Xt + t + {1} O O O [0]1× 6
: O
2×6
= {0} (28) O O :
O O Qm 2m × 6m (34)
in which, H is a 3m × 2m matrix containing the softening/hardening
slopes in Fig. 2-b, as follows: and W is a 2m × 2m matrix comprised of Ws matrices:

H1 O O W1 O O
: :
O H2 : O O W2 : O
hs 0 O 1 1
H= O O O
Hs = hs 0 W= O O : ; Ws =
O O O 1 0
O O : O : 2×2
: O hs 0 O O : O
O O : 3×2
Hm O O Wm (35)
O O 3m × 2m (29) 2m × 2m

Fig. 2. Constitutive model in terms of normalized internal actions: (a) typical 6-mode piecewise linear yield surface and its evolution with isotropic softening (b)
three phase softening model [28].

4
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 3. Constitutive model in terms of normalized internal forces: (a) typical 6-mode piecewise linear yield surface and its evolution with kinematic softening, (b)
three phase softening model.

Fig. 4. Constitutive model with typical 6-mode piecewise linear yield surface: (a) isotropic softening, (b) kinematic softening (c) combined isotropic-kinematic
softening.

Beside these linear inequality constraints, to ensure logical sequence


of activation of hardening/ softening phase, the following nonlinear
constraints must be considered:

p
x¯s1 ( x¯ s1 x¯ s2 + sp s1) = 0
p p
x¯ s2 (x¯ s1 s 2 + s1) = 0 (37)

or in compact form for the whole structure:

( X¯ t + t )T (Z ) = {0} (38)

In summary, the mathematical programming problem that is needed


to find the increment of plastic multiplier can now be established as
Fig. 5. The portal one-story frame structure with perfect plastic behavior.
follows:

and L contains the plastic limits that is defined according to Eq.(32):

L1 p
L 2 s1
L= ; Ls = p p
: s2 s1 2 × 1
Lm 2m × 1 (36)

Table 1
Properties of the elements used in portal one-story frame structure.
Element No. Type A (cm2) I (cm4) Plastic Moment (kN.cm) Plastic Axial Force (kN) Modulus of Elasticity (kN/cm2) Density (kg/m3)

2 IPE200 28.5 1943 5300 680 20,000 7850


1, 3 IPE240 39.1 3892 8800 940 20,000 7850

5
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 6. The horizontal displacement responses at node 2 as a function of time.

Fig. 7. The Yield surface evolutions in node 4 obtained (a) The Proposed Method (b) Newmark method.

max{1}T Xt + t
subjected to:
Y= T (F
intt + Fintt + t ) + T. fv . Xt + t H (X¯ t + X¯ t + t ) {1}
{0}
Zt + t = Zt + Q Xt + t + W X¯ t + t L {0}
( Xt + t )T Y = 0
( X¯ t + t )T Zt + t = 0
x sp x sq = 0; p q in which: p , q = 1, 2, 3, , 6, s = 1, 2, 3, ,m
Xt + t {0}
X¯ t + t {0}
Fig. 8. 2D frame structure with isotropic hardening.
(39)

Case 3: (Kinematic evolution (The Bauschinger Effect)). In case of three


phase kinematic evolution, according to Fig. 3 and Prager's rule for the

6
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Table 2
Properties of the elements used in example 2 and 3.
Element No. Type A (cm2) I (cm4) Plastic Moment (kN.cm) Plastic Axial Force (kN) Modulus of Elasticity (kN/cm2) Density (kg/m3)

1 IPE200 28.5 1943 5300 680 20,000 7850


2 IPE240 39.1 3892 8800 940 20,000 7850

ns1 0 0 0 0 0
0 ns 2 0 0 0 0
p 0 0 ns 3 0 0 0
= Q X s t+ t ; Q =
s
_ ks _ ks 0 0 0 ns 4 0 0
0 0 0 0 ns5 0
0 0 0 0 0 ns 6 6× 6 (41)

As before, nsj is the magnitude of normal vector to the jth yield


surface in critical section s. According to Eq. (40), the definition of
plastic deformation increments in Eq. (21), and relation between evo-
lution variables and equivalent plastic deformations (Eq. (31) and
(32)), the Prager's back stress could be written in the following form:
Fig. 9. The three-phase isotropic hardening.
= Hk s s X s t + t = Hk s s (Q
¯s
1) T sX
kt + t
_ ks (42)
Bauschinger effect, and are defined; as follows:
in which, Hk s is a matrix containing the softening/hardening slope
= {1} in Fig. 3-b and is defined as follows:
= c ps (40)
10
Hk s = (hs )
in which ps is the plastic strains in section s and c is the Prager's constant 01 (43)
parameter. In this case, the equivalent plastic strain ( sp ) in section s,
and T s is the transformation matrix that is defined as follows:
should be calculated separately for each yield surface:

1.1 Fig. 10. Vertical Displacement at node 2 as a


function of time ( t = 0.05 s).

0.9

0.7

0.5
Vertical Displacement of node 2 (cm)

0.3

0.1

-0.1 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-0.3

-0.5

-0.7

-0.9

-1.1
Time (sec.)
Newmark-ArcLength ( =1) Proposed Method ( =1) Newmark-ArcLength ( =0)

Proposed Method ( =0) Proposed Method ( =0.6) Linear Solution

7
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 11. Vertical Displacement at node 2 as


1 a function of time ( t = 0.2 s).

0.8

0.6
Vertical Displacement of node 2 (cm)

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-0.2

-0.4

-0.6

-0.8

-1
Time (sec.)
Newmark-ArcLength ( =1; t=0.05) Proposed Method ( =1; t=0.2)

Newmark-ArcLength ( =1; t=0.2)

Fig. 12. The Yield surface evolutions in node 2 of element 1 with full isotropic Fig. 13. The Yield surface evolutions in node 2 of element 1 with full kinematic
hardening behavior. hardening.

8
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Zkt + t = Zkt + Qk Xt + t + Wk X¯ kt + t Lkt + t {0} (48)

in which:

Q1 k O :
O
O Q2 k : O
O
Qk = O O : O : Qsk
O O
: O
O O : m
O O Q k 12m × 6m
Ns1 O O O O O

=
O Ns2 O O O O
O O Ns3 O O O
Nsj = {n0 }sj

O O O Ns4 O O
O O O O Ns5 O
O= {00}
O O O O O Ns6 12 × 6 (49)

W 1k O :
O
O W 2k : O
O
Wk = O O : O : W sk
Fig. 14. The Yield surface evolutions in node 2 of element 1 with 60% isotropic O O
: O
and 40% kinematic hardening. O O : m
O O W k 12m × 12m
Ws1 O O O O O
Table 3
CPU solution time for the responses of example 2 t = 0.05 s.
O Ws2 O O O O Wsj = 1 1
W O O O 1 0
= O O s3
Method CPU time (s) O O O Ws 4 O O 0 0
O=
O O O O Ws5 O 0 0
Elastic Solution – Duhamel's Integral 0.37 O O O O O Ws6 12 × 12 (50)
Elastic Solution – Newmark method 0.14
Newmark-AL – Full Isotropic 89.53
Proposed Method – Full Isotropic 0.92
Ls1
Newmark-AL – Full Kinematic 55.51
Proposed Method – Full Kinematic 0.62 L1k Ls 2
p
Newmark-AL – Combined 91.64 L2k Ls 3 s1
Lk = ; Ls k = ; Lsj =
Proposed Method – Combined 0.87 : Ls 4 p
s2
p
s1
Lm k 12m × 1
Ls5
Ls6 12 × 1 (51)
100000000000
001000000000
000010000000 Additionally, the logical sequence of hardening/softening phase
Ts = activation demands the following nonlinear equation to be met in all
000000100000
000000001000 sections:
000000000010 6 × 12 (44)
( X¯ kt + t )T Zkt + t =0 (52)
Additionally, kt + t is a vector containing the kinematic evolution
X̄ s
variables of section s and is arranged in the form of X s t + t matrix, as The mathematical programming problem that is to find the incre-
follows: ment of plastic multiplier can now be established as follows:
X¯ s1
max{1}T Xt + t
X¯ 1kt + t X¯ s2
subjected to:
X¯ 2 kt + X¯ s3 x¯s1j
X¯ kt + t = t ; X¯ s kt + t = ; X¯ sj = Y= T (F + Fintt + Hk ( Q 1) T X¯ kt + t )
: X¯ s 4 x¯s2j
intt t kt
_ k
2×1
X¯ m kt + t 12m × 1 X¯ s5 T.
+ fv . Xt + t {1} {0}
X¯ s6 (45)
12 × 1 Zkt + = Zkt + Qk Xt + t + Wk X¯ kt + t Lk {0}
t

In this case, at each section s every yield surface j has its own three ( Xt + t )T Y = 0
evolution variables ( x¯ s0j , x¯s1j , x¯s2j ) which according to Fig. 3-b should ( X¯ kt + t )T Zkt + t = 0
admit the following conditions: x sp x sq = 0; p q; p , q = 1, 2, 3, , 6, s = 1, 2, 3, , m
Xt + {0}
x¯ s0j p
s1
t
X¯ t + t {0} (53)
x¯s1j p
s2
p
s1
x¯s0j + x¯ s1j + x¯s2j = s
pj
(46)
Case 4: (Combined Isotropic-Kinematic evolution). The combined
After applying the same simplifications made in the isotropic case, Isotropic-Kinematic evolution is followed by adopting a share for
the compact matrix form of these additional conditions can be dis- Isotropic and Kinematic rules in final deformation. Considering that
played as follows: ( ) represents the share of isotropic evolution (0 1), (1- ) will
represent the share of kinematic deformation and according to Fig. 4,
Z s kt + t = Z s kt + Q s k Xst + t + W s k X¯ s kt + t Ls kt + t {0} (47)
the mathematical programming problem could be written in the
If Eq. (47) is written for the whole structure: following form:

9
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

0.8 Fig. 15. Vertical Displacement at node 2 as


a function of time ( t = 0.05 s).

0.6

0.4
Vertical Displacement of node 2 (cm)

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-0.2

-0.4

-0.6

-0.8

-1
Time (sec.)
Newmark-ArcLength ( =1) Proposed Method ( =1) Newmark-ArcLength ( =0)

Proposed Method ( =0) Proposed Method ( =0.6) Linear Solution

max{1}T Xt + t Modification of P (ti + 1) leads to the modification of AD (ti + 1) ,


subjected to: BD (ti + 1) , and accordingly modification of linear responses that results in
precise non-linear responses.
Y= T (F
intt + Fintt + t kt (1 ) Hk ( Q 1) T X¯ kt + t )
_ k
3. Steps of the proposed algorithm
+ T. fv . Xt + H (¯Xt + X¯ t + t ) {1}
t
Zt + t = Zt + Q Xt + t + W X¯ t + t L {0}
Computational steps of the proposed algorithm are presented as
Zkt + t = Zkt + Qk Xt + t + Wk X¯ kt + t Lk {0} follows:
( Xt + t )T Y = 0
( X¯ t + t )T Zt + t = 0
1- Pre-analysis Initial Calculations
( X¯ kt + t )T Zkt + t = 0
Phase. I. Perform modal decomposition and calculate the shape modes
x sp x sq = 0; p q in which: p , q = 1, 2, 3, , 6, s = 1, 2, 3, ,m and frequencies
Xt + t {0} II. Compute I1to I4 for all time steps
III. Calculate the influence matrix for all critical sections
X¯ t + t {0}
(54) 2. Analysis pha- For each time step:2.1 Stage one:
se I. Calculate AD (ti + 1) and BD (ti + 1) using Eqs. (9) and (10)
II. Calculate u (ti + 1) using Eq. (8)
III. Calculate the internal forces: Fintt + t = Fintt + K Ut+ t
The mathematical programming problem shown in Eqs. (27), (39), IV. Check for : T . fv . Xt + T (Fintt + t ) {0}
(53), and (54) have nonlinear constraints (marked with asterisk) which If Yes, start another Time Step 2.1. If Not Goto 2.2.: 2.2 Stage Two:
are satisfied implicitly during pivot finding in simplex table. For more I. Establish the Simplex Table and solve it to obtain the increment
information and detailed steps of solving these mathematical pro- of plastic multipliers.
II. Modify Pi + 1:
gramming problems, refer to Mahini et al. [28]. P (ti + 1) = Pi + 1 Pi + fv . Xt + t
After finding the increment of plastic multipliers, the elastoplastic III. Calculate AD (ti + 1) and BD (ti + 1) using Eqs. (9) and (10)
responses can be evaluated through modification of P (ti + 1) as follows: IV. Calculate u (ti + 1) using Eq. (8)
V. Calculate the internal forces:
P (ti + 1) = P (ti + 1) P (ti) + Fip+ 1 = Pi + 1 Pi + fv . Xt + t (55) Fintt + t = Fintt + K Ut+ t + fv . Xt + t
Goto Next Time Step 2.1
Leading to modified modal forces:
P (ti + 1) = T· P (ti + 1) (56)

10
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

0.8 Fig. 16. Vertical Displacement at node 2 as


a function of time ( t = 0.2 s).

0.6

0.4
Vertical Displacement of node 2 (cm)

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-0.2

-0.4

-0.6

-0.8

-1
Time (sec.)
Newmark-ArcLength ( =1; t=0.05) Proposed Method ( =1; t=0.2)
Newmark-ArcLength ( =1; t=0.2)

Fig. 17. The Yield surface evolutions in node 2 of element 1 with full isotropic Fig. 18. The Yield surface evolutions in node 2 of element 1 with full kinematic
softening. softening.

4. Numerical examples hardening and softening behavior; the fourth example is a portal four-
story frame structure with kinematic hardening behavior under El-
In this section in order to demonstrate the performance of the Centro earthquake record. The last example investigates the same
proposed method, several problems are solved. The first example is a structure with elastic-perfectly plastic behavior for all sections in which
2D Frame structure with perfect plastic behavior; The second and third the structure has been driven to collapse.
examples are frame structures with isotropic, kinematic, and combined

11
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

4.1. Portal one-story frame structure

A portal one-story frame as presented in Fig. 5 is considered. In


order to see the effect of axial forces on the moment capacity of col-
umns in the context of lumped plasticity, two axial forces are applied at
the top of columns. A similar example has been solved by Nguyen and
Kim [12], in which they have considered distributed plasticity, P-Δ and
P-δ effect using fiber element method.
The dimensions are: L = 400 cm and h = 300 cm . The mechanical
properties of the elements in portal frame are presented in Table 1.
The material is assumed to have perfect plastic behavior on element
3 (both ends selected as critical sections in the proposed method). The
exciting force, F(t) in Fig. 5, is provided as:
F(t) = 150sin(1.2t ) kNt = 0to5 s. (57)
To investigate the accuracy and robustness of the proposed method
this example is solved and compared to Newmark method with lumped
plasticity using MATLAB, and Newmark method using Beams With
Hinges (BWH) element in OpenSEES software. The responses of hor-
izontal displacement at node 2 are presented in Fig. 6.
Fig. 19. The Yield surface evolutions in node 2 of element 1 with 60% isotropic Fig. 6, clearly shows that the results of the proposed method is al-
and 40% kinematic. most the same as the Newmark method with lumped plasticity and it
has infinitesimal difference with OpenSEES. This small difference is due
to the definition of the two yield surfaces. Time wise, the responses of
Table 4
the proposed method (MATLAB code) has been obtained in 0.35 s,
CPU solution time for the responses of example 3 t = 0.05 s.
while that of Newmark with lumped plasticity (MATLAB code) is ob-
Method CPU time (s) tained in 134.65 s and Newmark with BHW (OpenSEES) is obtained in
Elastic Solution – Duhamel's Integral 0.50
1.35 s i.e. the proposed method is more than 4 times faster than
Elastic Solution – Newmark 0.13 OpenSEES and is 380 times faster than Newmark method using MA-
Newmark-AL – Full Isotropic 33.00 TLAB.
Proposed Method – Full Isotropic 0.85 Fig. 7(a) and (b) present the change in yield conditions at node 4
Newmark-AL – Full Kinematic 24.62
during the analysis process by the proposed and Newmark-lumped
Proposed Method – Full Kinematic 0.64
Newmark-AL – Combined 31.61 plasticity methods, respectively. These figures show that (a) the re-
Proposed Method – Combined 0.73 sponses are very similar to each other, (b) unlike some of mathematical
programming algorithms for nonlinear analysis, the proposed method
has no tendency to get stuck at the corner points of the piecewise linear
yield surface.

4.2. 2D frame structure with hardening evolutions

This example considers a 2D-Frame structure with an inclined


member to show how the interactive yield surfaces in element 1 could
be activated. Fig. 8 presents the geometry of this structure.
In this figure, L1 = 250 cm , L2 = 200 cm , and h = 150 cm . The
properties of elements are presented in Table 2.
In this example, N̄0 in Fig. 1 is considered to be 0.15 and the three-
phase isotropic hardening, presented in Fig. 9, is assumed to be the
governing behavior of material. In this figure, sp1and sp2 are assumed to
be 0.05 and 0.25, respectively and js_m is assumed to be 1.5. The ex-
citing force at node 2 is provided as:
F(t) = 340sin(0.5t 2) kN ; 0 t 3.5sec. (58)
The responses of vertical displacement at node 2 obtained by the
proposed method and Newmark-Arclength method using 0.05 s as time
step duration are presented in Fig. 10.
According to this figure, the responses of each case (isotropic, ki-
nematic, and combined) are very similar to each other and the only
tangible difference is the difference between the responses of various
cases in hardening. This is reasonable because in the case of full kine-
matic hardening the locus of yield surfaces move and while in one
surface it results in hardening in the opposite surface it behaves as if it
is softening. This results in early activation of opposite yield surface and
Fig. 20. The four-story portal 2D frame structure under horizontal earthquake
leads to larger displacements especially at the opposite yield surface.
excitation.
To see the effect of time step on the performance of proposed
method and compare it to Newmark method, this example was solved
using t = 0.2 s. In Fig. 11, the results have been compared to the case

12
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Table 5
Section Properties utilized in the four story building.
d (mm) bf mm) tw (mm) t f (mm) A (cm2) I (cm4) Plastic Moment (kN.cm) Plastic Axial Force (kN)

WH 400 × 400 400 400 8 14 142 45,179 57,302 3331


WH 350 × 350 350 350 8 12 110 26,322 38,356 2587
WH 300 × 300 300 300 8 12 94.1 16,350 27,945 2211
WH 500 × 300 500 300 8 16 133 63,080 64,890 3136
WH 400 × 250 400 250 8 12 90.1 26,140 34,000 2117

300 than Newmark’s method. However, the proposed method although uses
200 the Duhamel's Integral, is much faster (about 100 times faster) than Arc
Length algorithm in all types of problems including full isotropic be-
Ground Acc. (cm/s2 )

100
havior, full kinematic behavior and the combination of the two.
0
0 5 10 15 20 25 30
4.3. 2D frame structure with softening evolutions
-100

-200 Consider the previous example, presented in Fig. 8 with the same
-300
element types and properties, but with a softening behavior, presented
in Fig. 2b with adopted sp1and sp2 equal to 0.05 and 0.25, respectively
-400
Time (s)
and js_min = 0.8. The exciting force is provided; as follows:

Fig. 21. The El-Centro earthquake acceleration record with peak ground ac- F(t) = 290sin(0.5t 2) kNt = 0to3.5s. (59)
celeration of 310 cm/s2.
Other properties of this example are the same as the previous ex-
ample. Assuming Δt = 0.05, the nonlinear dynamic responses of node 2
in vertical direction that is obtained by the proposed method and
Newmark-Arclength method is depicted in Fig. 15. It is observed that
the responses are in close agreement to each other in all three cases.
Another important fact about Fig. 15 is that unlike the previous ex-
ample in which the maximum displacement belonged to the full kine-
matic case, in this example the maximum displacement belongs to full
isotropic case as a result of softening behavior. This is because in iso-
tropic case softening makes the whole yield surface become smaller.
Similar to the case of hardening, the effect of time step on the results
of softening was investigated. Fig. 16 compares the responses of these
two methods using t = 0.2 s with the response of Newmark-Arclength
method using t = 0.05 s as a reference method. It is seen that longer
time steps in the proposed method results in considerably less amount
of errors than the Newmark-Arclength method.
Figs. 17, 18, and 19 present the yield surface evolutions in node 2 of
element 1 with full isotropic, full kinematic, and combined 60% iso-
tropic and 40% kinematic softening, respectively. Again, the close
agreement in responses of both methods made us to present only the
yield surface evolutions obtained by the proposed method; however
according to Table 4 the consumed CPU time for the proposed method
Fig. 22. The horizontal displacement at the rooftop. is much less than the Newmark-Arclength method.

of t = 0.05 s as the reference response. When the time step increases, 4.4. Four story frame structure under El-Centro earthquake [27]
the results of Newton-Raphson algorithm undergo some errors while,
the proposed method keeps on following the reference response. It is This example has been reported in [27]. As displayed in Fig. 20, a
concluded that the proposed method is more precise than Newmark four-story frame structure is subjected to earthquake. The data on di-
nonlinear dynamic analysis algorithms. mensions, section types, member numbering, and vertical load num-
Figs. 12, 13, and 14 present the yield surface evolutions in node 2 of bering of the frame are shown in this figure. All sections are welded
element 1 with full isotropic, full kinematic, and combined 60% iso- steel H-sections, the properties of which are displayed in Table 5.
tropic and 40% kinematic hardening, respectively. Note that since the The lumped masses at the rooftop are 5 t for exterior nodes and 10 t
responses of Newmark-Arclength and the proposed method were very for interior one. They are 10 t and 20 t, for the other floors. The vertical
similar to each other, only the responses of the proposed method are loads on the beam-column joints are assumed to be P1,2,3 ex
= 200 kN and
presented. Having said that, one may notice the difference of the two P4ex = 100 kN for exterior columns, and P1,2,3 in
= 400 kN and
algorithms in their CPU time of nonlinear dynamic analysis. P4in = 200 kN for interior columns. A damping ratio of 0.02 is con-
Table 3 compares the CPU time of solution for each algorithm. It is sidered for the first two modes. In this example, the plasticity is as-
seen that the Duhamel's Integral takes more time for elastic analysis sumed only for columns. To consider the kinematic coefficient equal to
3/97, (as was assumed in the paper), sp1, sp2 and js_m in Fig. 9 are

13
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 23. The horizontal displacement at the rooftop including main events through all time history.

Fig. 24. The Yield surface evolutions at the bottom section of column 111. (a) the proposed method (b) The time integration based methods.

assumed 0.0, 1.00, and 0.031 respectively. Other properties such as According to this figure, only 5 sections experience plastic deformation
modulus of elasticity are the same as previous examples. The El-Centro and all these plastic hinges disappear after about 6 s; however, the
(NS 1940) earthquake, as displayed in Fig. 21, with peak ground ac- structure has an infinitesimal (about −0.5 cm) remaining plastic de-
celeration of 310 cm/s2 is used as the horizontal ground excitation. formation which is hard to distinguish. It is also notable that since the
Chen et al. [27], has solved this example using a time integration amount of sp1 is zero, all these 5 sections, which experienced plastic
method based on two element models namely Section Method and Fiber deformation, undergo hardening; however, the highest amount of
Method. Their horizontal displacement responses at the rooftop are hardening, according to the results, belongs to the bottom section of
compared to the proposed method in Fig. 22. It can be seen that there column 112 with a final back-force equal to about −1000 kN in axial
are only infinitesimal differences in the results of their method and the direction.
proposed algorithm. One important point that is noticeable in this Figs. 24 and 25 illustrate the yield surface evolutions at the bottom
figure is that when compared to the responses obtained by the proposed sections of columns 111 and 112, respectively. According to these fig-
method, there is some amplitude decay errors associated with the time ures, even though the proposed method utilizes piecewise linear yield
integration-based methods after formation of considerable plastic surface that is a bit different than assumptions of the reference paper,
hinges. the obtained responses are very similar. Despite the ignorable differ-
To show an overall picture on the history of plastic hinge formation ences, according to Table 6 there are considerable savings in solution
all over the building structure, the rooftop horizontal displacement times of proposed method. It should be noted that all three methods are
including main events through time history is presented in Fig. 23. utilized in computers with the same specifications (Pentium IV

14
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 25. The Yield surface evolutions at the bottom section of column 112. (a) The proposed method (b) The time integration based methods.

Table 6 multiplied by 1.2 in order to drive the structure to the collapse. Fig. 26
CPU solution time for the responses of example 4. presents the rooftop horizontal displacement including the main events
Method CPU Time (s) from plastic hinge formation until collapse which occurs at about 9 s.
According to this figure the first wave of the El-Centro earthquake,
Sectional 20.76 at about time 3 to 4 s, activates a few plastic hinges in the columns and
Fiber element 65.48
some more hinges on the beams. The second wave started from 4 to 5 s
Proposed 5.08
causes remarkable plastic deformation, as a result of formation of many
hinges. The plastic deformations are accumulated after time 6 s and the
Processor, CPU 2.8 GHz, RAM 2.0 G.). According to this table, the structure is pushed towards a mechanism and this leads to the collapse
proposed method consumes much less computational time than the of the structure.
other two integration-based methods. Another interesting result that is noticeable in Fig. 26 is that the
time integration-based methods result in period elongation and ampli-
4.5. Collapse investigation of the Four-story structure tude decay errors.
Figs. 27 and 28 present the yield surface events, obtained by the
In order to investigate the performance of the proposed method in a proposed method, at the bottom section of column 111 and at the right
collapse situation, the same structure of the previous example, but with section of beam 241, respectively.
different conditions, is considered. It is assumed that all member sec- The solution time (measured until one time-step before collapse) for
tions are characterized as critical with an elastic-perfectly plastic be- this example is 3.22 s for the proposed method and 43.56 s for the time
havior. Additionally, the earthquake loading in the previous example is integration-based method. This means the growing number of plastic

Fig. 26. The horizontal displacement at the rooftop including main events through time history until collapse.

15
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

Fig. 27. The Yield surface events at the bottom section of column 111.

Fig. 28. The Yield surface events at the right section of beam 241.

16
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

hinges affects negatively the solution time of time integration-based CRediT authorship contribution statement
methods while it has ignorable effect on the solution time of the pro-
posed method. S. Mohammadzadeh: Methodology, Investigation, Formal analysis,
Software, Validation, Writing - original draft. H. Moharrami:
5. Summary and conclusion Supervision, Conceptualization, Writing - review & editing.

A Dissipated Energy Maximization (DEM) based approach has been Declaration of Competing Interest
proposed to solve the frame structures with material nonlinearity under
dynamic loads. The approach is composed of two stages. In stage one, The authors declare that they have no known competing financial
the proposed method utilizes the well-known Duhamel's Integral to interests or personal relationships that could have appeared to influ-
calculate the linear dynamic responses. In stage two, a Linear ence the work reported in this paper.
Programming (LP) problem is established to find the increment of
plastic multipliers associated with the linearized yield surfaces. The LP Appendix A. Supplementary material
problem comprises the Dissipated Energy Maximization (DEM) as ob-
jective function and yield conditions as the constraints. There are also Supplementary data to this article can be found online at https://
several nonlinear complementarity conditions that are implicitly sa- doi.org/10.1016/j.engstruct.2020.110526.
tisfied along with the solution process. Then the Restricted Basis Linear
Programming (RBLP) algorithm is used to solve the problem. The References
plastic multipliers are then used to modify the values of internal forces
and the responses of the Duhamel's Integral. [1] Chang S-Y. Performances of non-dissipative structure-dependent integration
Five numerical examples were solved using the proposed method methods. Struct Eng Mech 2018;65:91–8.
and the well-known time integration-based methods with either [2] Mohammadzadeh S, Ghassemieh M, Park Y. Extended implicit integration process
by utilizing nonlinear dynamics in finite element. Struct Eng Mech
Newton-Raphson or Arc-Length control iterations. The first Example 2017;64:495–504.
proved that the proposed method does not stick to corners of yield [3] Mohammadzadeh S, Ghassemieh M, Park Y. Structure-dependent improved Wilson-
surfaces and has an approved accuracy. The second and third examples θ method with higher order of accuracy and controllable amplitude decay. Appl
Math Model 2017;52:417–36.
demonstrated the potency of the proposed method in problems invol- [4] Oden JT. A general theory of finite elements II. Applications. Int J Numer Meth Eng
ving isotropic, kinematic, and combined behavior. The fourth example 1969;1:247–59.
compared the proposed method with two other time integration-based [5] Leon SE, Paulino GH, Pereira A, Menezes IF, Lages EN. A unified library of nonlinear
solution schemes. Appl Mech Rev 2011;64:040803.
methods under horizontal excitation of a real earthquake record and [6] Caballero A, Willam K, Carol I. Consistent tangent formulation for 3D interface
proved the merits of the proposed method from the viewpoint of ac- modeling of cracking/fracture in quasi-brittle materials. Comput Methods Appl
curacy and computational efficiency compared to other methods; fi- Mech Eng 2008;197:2804–22.
[7] Rezaiee-Pajand M, Salehi-Ahmadabad M, Ghalishooyan M. Structural geometrical
nally, the last example, where perfect plasticity was adopted as material
nonlinear analysis by displacement increment. Asian J. Civil Eng. (BHRC)
behavior, assessed the robustness of the proposed method close to 2014;15:633–53.
collapse condition and demonstrated its performance in collapse si- [8] Kala Z. Computation of equilibrium paths in nonlinear finite element models.
tuation. MATEC Web Conf: EDP Sci 2016:04026.
[9] Rezaiee-Pajand M, Estiri H. Finding equilibrium paths by minimizing external work
As a conclusion to the proposed method, the following remarks are in dynamic relaxation method. Appl Math Model 2016;40:10300–22.
highlighted: [10] Andreaus U, Dell’Isola F, Giorgio I, Placidi L, Lekszycki T, Rizzi NL. Numerical si-
mulations of classical problems in two-dimensional (non) linear second gradient
elasticity. Int J Eng Sci 2016;108:34–50.
(1) It was shown that the proposed algorithm yields the nonlinear re- [11] Yang Y-B, Shieh M-S. Solution method for nonlinear problems with multiple critical
sponse with a marginal improvement in accuracy compared to the points. AIAA J 1990;28:2110–6.
time integration-based methods. [12] Nguyen P-C, Kim S-E. Second-order spread-of-plasticity approach for nonlinear
time-history analysis of space semi-rigid steel frames. Finite Elem Anal Des
(2) Since the proposed method does not need to update stiffness matrix, 2015;105:1–15.
mode shape analysis and the I1 ~ I4 integrals in the whole solution [13] Nguyen P-C, Kim S-E. Distributed plasticity approach for time-history analysis of
process, it is much faster than other classical methods. steel frames including nonlinear connections. J Constr Steel Res 2014;100:36–49.
[14] Nguyen P-C, Doan NTN, Ngo-Huu C, Kim S-E. Nonlinear inelastic response history
(3) The proposed method traces the formation of plastic hinges during
analysis of steel frame structures using plastic-zone method. Thin-Wall Struct
time steps. Therefore, it is not necessary to use variable time step. 2014;85:220–33.
This characteristic provides more accurate results than the other [15] Nguyen P-C, Kim S-E. Nonlinear elastic dynamic analysis of space steel frames with
semi-rigid connections. J Constr Steel Res 2013;84:72–81.
methods when the time step is not small enough.
[16] Nguyen P-C, Kim S-E. Nonlinear inelastic time-history analysis of three-dimensional
(4) It was shown that the proposed method is able to model the hard- semi-rigid steel frames. J Constr Steel Res 2014;101:192–206.
ening and softening properties of materials. [17] Rahai A, Nafari SF. A comparison between lumped and distributed plasticity ap-
(5) The only approximation in the proposed method refers to the pie- proaches in the pushover analysis results of a pc frame bridge. Int J Civil Eng
2013;11:217–25.
cewise linearization of yield surface, constant time step and nu- [18] Shing PB, Tanabe T-a. Modeling of inelastic behavior of RC structures under seismic
merical solution of Duhamel's Integrals. Otherwise it yields exact loads: ASCE Publications; 2001.
solutions. [19] Maier G. A quadratic programming approach for certain classes of non linear
structural problems. Meccanica 1968;3:121–30.
(6) The Dissipated Energy Maximization (DEM) is a very good ap- [20] Maier G. Quadratic programming and theory of elastic-perfectly plastic structures.
proach for this type of formulation of nonlinear dynamic analysis Meccanica 1968;3:265–73.
procedures. [21] Maier G. A matrix structural theory of piecewise linear elastoplasticity with inter-
acting yield planes. Meccanica 1970;5:54–66.
(7) The proposed method unlike other classical methods, is not itera- [22] Maier G, Giacomini S, Paterlini F. Combined elastoplastic and limit analysis via
tive and does not have convergence criterion to meet. It yields the restricted basis linear programming. Comput Methods Appl Mech Eng
response of the structure as an exact solution of a mathematical 1979;19:21–48.
[23] Ardito R, Cocchetti G, Maier G. On structural safety assessment by load factor
problem.

17
S. Mohammadzadeh and H. Moharrami Engineering Structures 212 (2020) 110526

maximization in piecewise linear plasticity. Eur J Mech-A/Solids 2008;27:859–81. to elastic-perfectly plastic analysis of planar frames. Arch Mech 2013;65:171–94.
[24] Ardito R, Cocchetti G, Maier G. Generalized limit analysis in poroplasticity by [27] Yongsheng C, Bin W, Tianlin P. A novel sectional constitutive model for beam-
mathematical programming. Arch Appl Mech 2010;80:57. column element. Eng Rev 2016;36:41–52.
[25] Tangaramvong S, Tin-Loi F. Simultaneous ultimate load and deformation analysis of [28] Mahini M, Moharrami H, Cocchetti G. Elastoplastic analysis of frames composed of
strain softening frames under combined stresses. Eng Struct 2008;30:664–74. softening materials by restricted basis linear programming. Comput Struct
[26] Mahini M, Moharrami H, Cocchetti G. A dissipated energy maximization approach 2014;131:98–108.

18

Вам также может понравиться