This action might not be possible to undo. Are you sure you want to continue?

# Dynamic analysis of steel frames with ﬂexible connections

Miodrag Sekulovic

*

, Ratko Salatic, Marija Nefovska

Faculty of Civil Engineering, University of Belgrade, Bulevar kralja Aleksandra 73, 11000 Belgrade, Yugoslavia

Received 1 June 2001; accepted 14 February 2002

Abstract

This paper deals with the eﬀects of ﬂexibility and damping in the nodal connections on the dynamic behavior of

plane steel frames. A ﬂexible eccentric connection is idealized by nonlinear rotational spring and dashpot in parallel.

Thus, the eﬀects of viscous and hysteretic damping on dynamic response of frame structures are taken into consid-

eration. A numerical model that includes both nonlinear connection behavior and geometric nonlinearity of the

structure is developed. The complex dynamic stiﬀness matrix for the beam with ﬂexible connections and linear viscous

dampers at its ends is obtained. Several examples are included to illustrate the eﬃciency and accuracy of the present

model. Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Steel frame; Semi-rigid connection; Nonlinear dynamic analysis

1. Introduction

The conventional methods of analysis and design of

frame structures are based on the assumption that the

joint connections are either fully rigid or ideally pinned.

The models with ideal connections simplify analysis

procedure, but often cannot represent real structural

behavior. Therefore, this idealization is not adequate as

all types of connections are more or less, ﬂexible or semi-

rigid. It is proved by numerous experimental investiga-

tions that have been carried out in the past [1–4].

Based on experimental study due to static monotonic

loading tests for various types of connections, many

models have been done to approximate the connection

behavior. The simplest one is the linear model that has

been widely used for its simplicity [5–7]. However, this

model is good only for the low level loads, when the

connection moment is quite small. In each other case,

when the connection rigidity may rapidly decrease

compared with its initial value, a nonlinear model is

necessary. Several mathematical models to describe the

nonlinear behavior of connections have been formulated

and broadly used in research practice [8–11].

So far, most experimental and theoretical work is

limited to static analysis of steel frames with ﬂexible

connections. Very few papers have been devoted to the

dynamic analysis although the ﬂexibility of connections

with energy dissipation has a great inﬂuence on dy-

namic behavior of these types of structures. Under cy-

clic loads, the connection hysteresis loop increases the

energy absorption capacity and hysteretic damping may

signiﬁcantly reduce dynamic response of real structures.

Therefore, modelling of the nodal connection is impor-

tant for the design and accuracy in the dynamic frame

structure analysis. However, as the experimental data

for the connection behavior under cyclic loading are

rather poor, it is diﬃcult to make corresponding math-

ematical model. The experiments carried out by Popov

and coworkers [12–16] and also in Refs. [17,18] show

that the hysteresis loops under repeated and reversed

loading are very stable, so the moment–rotation func-

tions obtained by static tests can be extended to the

dynamic analysis.

The dynamic analysis of frames with ﬂexible con-

nections using linear moment–rotation relationship has

been studied in several papers [19–21]. Kawashima and

Fujimoto [20] obtained the complex dynamic stiﬀness

Computers and Structures 80 (2002) 935–955

www.elsevier.com/locate/compstruc

*

Corresponding author. Tel.: +381-11-3218552; fax: +381-

11-3370223.

E-mail address: sekulovic@grf.bg.ac.yu (M. Sekulovic).

0045-7949/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.

PII: S0045- 7949( 02) 00058- 5

matrix for a uniform beam element with linear rota-

tional springs and dashpots at its ends. The inﬂuence of

the ﬂexibility and eccentricity in the connections on

dynamic behavior of plane frames within the linear

theory was investigated by Suarez et al. [21]. Although

the linear constitutive model of nodal connections is

very easy to use, it is inadequate in the term that it is

applicable only to a small range of the initial rotations

and because it cannot represent hysteretic damping to be

a primary source of passive damping in the frame

structures. A bilinear moment–rotation function, which

is also accurate only for a small rotation range, was used

by Sivekumaran [22] and Yousef-Agha et al. [23]. The

eﬀect of hysteretic damping resulting from the nonlinear

ﬂexible connection on the dynamic response of the frame

was studied by Shi and Altury [24]. They developed a

numerical model based on the complementary energy

approach using the Ang and Morise [6] function for the

moment–rotation relations at connections. Chan and

Ho [25] proposed a numerical method for linear and

nonlinear vibration analysis of frame with semi-rigid

connections. They adapted the conventional cubic

Hermitian shape functions for a uniform beam with end

springs and derived the element matrices using the

principle of total potential energy. The inﬂuence of both

hysteretic and viscous damping at connections on seis-

mic response of the steel frames was considered by

Sekulovic et al. [26]. The combined eﬀects of material

yielding and connection ﬂexibility in static and dynamic

problems have been discussed in detail by Chan and

Chui [36]. Several nonlinear ﬂexible connection models

under cyclic loading were established in the past decade

[27–30].

The present study is an extension of the author’s

previous work [31], regarding static analysis of ﬂexibly

joint frames, on a more general case of the dynamic

analysis. Two types of nonlinearities are considered:

geometric nonlinearity of the structure and material

(constitutive) nonlinearity of the connections. These

nonlinearities are interactive. The eccentricity of the

connections is also considered. To describe the nonlinear

behavior of the connection under cyclic loading, the

independent hardening model is used. So, the eﬀect of

hysteretic damping on dynamic behavior of the structure

is directly included through the connection constitutive

relation. Moreover, the inﬂuence of viscous damping at

connections on dynamic response of frame structures is

considered. For a uniform beam with rotational springs

and dashpots attached at its ends the complex dynamic

stiﬀness matrix is obtained. The stiﬀness matrix has been

obtained based on analytical solutions of governing

diﬀerential equations second order analysis, so that each

beam represents one element. Nodal displacements and

rotations are chosen as the primary unknowns, while

displacements and rotations of the element ends are

eliminated. Thus, the number of degrees of freedom

are the same as for the system with rigid connections.

Besides, the consistent mass and damping matrices

are derived. These matrices are based on the physical

properties of the member and given in an explicit

form. The present matrices are more general than the

corresponding matrices previously obtained by other

authors.

Energy dissipation exists in frame structures under

dynamic loads. The primary sources of energy dissipa-

tion, is known, may be hysteretic behavior of connec-

tions and the friction between elements forming the

beam–column assemblage. In addition, diﬀerent types of

energy dissipation devices can be installed into connec-

tions in order to increase the structural energy absorp-

tion capacity. For this reason, in the present model, the

total energy dissipation is conﬁded to the joint connec-

tions. Two types of energy dissipation are assumed.

They are: hysteretic damping due to nonlinear behavior

of connections and viscous damping at the connections.

In general, the eﬀects of these dampings are coupled.

Also, they can be considered separately using either

linear constitutive relation for the connections or zero

value for the viscous damping coeﬃcients at the con-

nections. As it is assumed that all structural elements,

except the connections, remain elastic through the whole

loading range, the energy dissipation at plastic hinges

cannot be observed. Also, energy dissipation due to ra-

diation damping at the supports is not included in this

consideration. The other types of energy dissipation that

may exist in real frame structures can be included in the

present model in the usual way, by mass and stiﬀness

proportional damping matrix.

Based on theoretical problem formulation, a com-

puter program was developed in order to increase dy-

namic analysis eﬃciency and design of steel frames. A

parametric study has been performed in order to esti-

mate the inﬂuence of nonlinear connection ﬂexibility

and viscous damping at connections on the frame dy-

namic (seismic) response. The present numerical model

is restricted to 2D frame systems. It can be extended on

a more general case of 3D analysis without diﬃculty.

Besides, the proposed beam element can be easily

incorporated into existing commercial programs for

structural analysis.

2. Formulation of structural element

A beam element with ﬂexible, eccentric and viscous

damping connections is shown in Fig. 1. The ﬂexible

connections are represented by nonlinear rotational

springs at beam ends. Thus, only the inﬂuence of

bending moment on the connection deformation is

considered, while the inﬂuences of axial and shear forces

are neglected. The connection spring element is assumed

massless and dimensionless in size. The eccentricity is

936 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

modelled by short inﬁnitely stiﬀ elements whose lengths

are e

1

and e

2

. The linear viscous damping at nodal

connections are modelled by dashpots acting at beam

ends.

2.1. Stiﬀness matrix and nodal force vector

The stiﬀness matrix and the nodal force vector for the

ﬂexible eccentric beam have been represented in a pre-

vious work by the authors [31]. It will be brieﬂy sum-

marized herein. Joint displacements and rotations are

the primary unknowns, while displacements and rota-

tions of the beam ends can be eliminated as has been

shown in [31]. Thus, the number of degrees of freedom is

the same as for the beam element with fully rigid con-

nections. Consequently, the function describing lateral

displacement vðxÞ for the element with ﬂexible eccentric

connections can be written in the usual way by inter-

polation function matrix and nodal displacements vector

as:

vðxÞ ¼ NðxÞðI þ GÞq ¼

¯

NNðxÞq; ð1Þ

where

NðxÞ ¼ N

1

ðxÞ N

2

ðxÞ N

3

ðxÞ N

4

ðxÞ ½ ; ð2aÞ

q

T

¼ v

1

u

1

v

2

u

2

f g; ð2bÞ

G ¼

1

D

0 De

1

0 0

g

21

g

22

g

23

g

24

0 0 0 ÀDe

2

g

41

g

42

g

43

g

44

_

¸

¸

_

_

¸

¸

_

; ð2cÞ

denoting the matrix of interpolation functions obtained

based on the analytical solutions of the second order

analysis equations [32], the nodal displacement vector

and the correction matrix, respectively. Interpolation

functions N

i

ðxÞ, i ¼ 1; . . . ; 4 and elements of correction

matrix G are given in Appendix A.

Stiﬀness matrix for the beam element with ﬂexible

eccentric connection is obtained through the total po-

tential energy, that can be written as

U ¼ U

a

þ U

f

þ U

s

; ð3Þ

where

U

a

¼

EI

2

2A

k

4

l; ð4aÞ

U

f

¼

EI

2

_

l

0

v

2

;xx

dx; ð4bÞ

U

s

¼

2

i¼1

c

i

a

2

i

; ð4cÞ

denoting strain energy of the beam, axial (U

a

) and

ﬂexural (U

f

) and strain energy of the springs (U

s

). Strain

energy due to axial deformation and bending are cou-

pled since parameter k

2

includes derivatives of both axial

and lateral displacements [31]. With the assumption that

k

2

¼ const., these two part of the strain energy can be

expressed independently. Thus, after substituting Eq. (1)

into Eq. (4b) the following can be obtained:

U

f

¼

1

2

q

T

ðk

II

þ k

ef

Þq; ð5Þ

where matrices k

II

and k

ef

are deﬁned as

k

II

¼ EI

_

l

0

ðN

00

ðxÞÞ

T

N

00

ðxÞ

_ _

dx; ð6aÞ

k

ef

¼ G

T

k

II

þ k

II

G þ G

T

k

II

G; ð6bÞ

denoting beam stiﬀness matrix with the rigid connec-

tions according to the second order analysis and cor-

rection matrix that accounts for the eﬀects of ﬂexibility

and eccentricity, respectively. Analytical expression for

the elements of matrix k

II

and the appropriate expan-

sions in the power series form, convenient for the nu-

merical analysis, are given by Goto and Chen [33].

The simpliﬁed form of this matrix, corresponding to

the linearized second order analysis, can also be used. In

that case, the stiﬀness matrix has the form

k

II

¼ k

0

þ k

g

; ð7Þ

Fig. 1. A beam with ﬂexible, eccentric and viscous damping

connections.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 937

where k

0

is the conventional stiﬀness matrix and k

g

is the

geometric stiﬀness matrix of the uniform beam. In this

case, the simpliﬁed form of the matrix G, with functions

/

i

, i ¼ 1; . . . ; 4, are replaced by 1.0, can be used.

The strain energy of the springs can be expressed in

the form

U

s

¼

1

2

q

T

k

s

q; ð8Þ

where

k

s

¼ G

T

CG: ð9Þ

The explicit form of matrices G and C can be found in

[31]. From Eqs. (5) and (8) the total strain energy due to

the bending for the beam with ﬂexible and eccentric

connections can now be written as:

U ¼ U

f

þ U

s

¼

1

2

q

T

ðk

II

þ k

ef

þ k

s

Þq: ð10Þ

The equivalent generalized end force vector due to

distributed loads along the beam pðxÞ is obtained in the

usual maner:

Q ¼

_

l

0

pðxÞ

¯

NN

T

ðxÞ dx ¼ ðI þ GÞ

T

_

l

0

pðxÞN

T

ðxÞ dx: ð11Þ

Components of the vector Q, for some simple load

distributions and temperature change are given in the

closed form in [32]. In general case, elements of the

vector Q are computed numerically.

2.2. Element mass matrix

Assuming that the mass density q is constant, the

element consistent mass matrix m can be formulated as:

m ¼

_

V

q

¯

NN

T

ðxÞ

¯

NNðxÞ dx; ð12Þ

where

¯

NNðxÞ is the matrix of modiﬁed shape functions

deﬁned by Eq. (1). After substitution of Eq. (1) into Eq.

(12), the consistent element mass matrix, for the uniform

beam with ﬂexible eccentric connections, can be written

as:

m ¼ m

0

þ m

ef

; ð13Þ

where

m

0

¼

_

V

qN

T

ðxÞNðxÞ dx; ð14aÞ

m

ef

¼ G

T

m

0

þ m

0

G þ G

T

m

0

G: ð14bÞ

In the above relations, m

0

denotes conventional mass

matrix for the beam element and m

ef

denotes the mass

correction matrix.

2.3. Complex dynamic stiﬀness matrix

Apart from nonlinear rotational springs, rotational

viscous dashpots are attached at beam ends, as shown in

Fig. 1. The total moment at each nodal connection

(i ¼ 1, 2) can be given in terms of relative rotation h

between beam end and column face and relative angular

velocity

_

hhðtÞ as:

M

i

ðtÞ ¼ k

i

h

i

ðtÞ þ c

i

_

hh

i

ðtÞ; i ¼ 1

0

; 2

0

; ð15Þ

where k

i

and c

i

are rotational spring stiﬀness and rota-

tional viscous damping coeﬃcients, while dot over the

symbols denotes diﬀerentiation with respect to time. The

tangent or secant form of the above relation may be

written if nonlinear springs and/or dashpots are con-

sidered. In the case of periodic response with circular

frequency x the following relation between the ampli-

tudes may be derived:

M

ið0Þ

¼ k

Ã

i

_

hh

ið0Þ

; i ¼ 1

0

; 2

0

; ð16Þ

where complex ﬂexural stiﬀness k

Ã

i

of the connection is

deﬁned as the ratio between moment and relative rota-

tion amplitudes:

k

Ã

i

¼

M

i

ðtÞ

h

i

ðtÞ

¼ k

i

þ jxc

i

; j ¼

ﬃﬃﬃﬃﬃﬃﬃ

À1

p

; ð17aÞ

h

i

ðtÞ ¼ h

ið0Þ

e

jxt

: ð17bÞ

The beam end force vector RðtÞ can be expressed in

terms of the end displacement vector qqðtÞ and relative

end rotation vector hðtÞ as:

RðtÞ ¼ kfqqðtÞ À hðtÞg; ð17aÞ

where

RðtÞ ¼ f T

1

ðtÞ M

1

ðtÞ T

2

ðtÞ M

2

ðtÞ g; ð18Þ

qqðtÞ ¼ fvv

1

ðtÞ uu

1

ðtÞ vv

2

ðtÞ uu

2

ðtÞ g; ð19Þ

h

T

ðtÞ ¼ f 0 h

1

ðtÞ 0 h

2

ðtÞ g; ð20Þ

are end force vector, end displacement vector and rela-

tive end rotation vector of the member respectively (Fig.

1), while k is the classical or the second order ﬂexural

stiﬀness matrix of a uniform beam, that depends on the

type of analysis.

After the elimination of relative end rotation vector

hðtÞ, Eq. (17a) transforms to:

RðtÞ ¼

kk

Ã

qqðtÞ; ð20Þ

where

kk

Ã

¼ ðI À SÞ

T

kðI À SÞ þ S

T

kk

s

S ¼ k þ k

1

þ k

2

; ð21aÞ

k

1

¼ ÀkS À S

T

k; ð21bÞ

938 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

k

2

¼ S

T

ðk þ K

*

ÞS; ð21cÞ

qqðtÞ ¼ qq

ð0Þ

e

jxt

: ð21dÞ

The explicit forms of matrices k

1

and k

2

are given in

Appendix A. The matrix S can be obtained from matrix

G putting e

1

¼ e

2

¼ 0, and it can be found in Appendix

A. The matrix k

Ã

is a complex ﬂexural stiﬀness matrix of

uniform beam with ﬂexible connection according to the

linear or second order analysis, including both ﬂexible

and viscous phenomena. The elements of this matrix

corresponding to the linear analysis are given in Ap-

pendix A.

Expanding the elements of the dynamic stiﬀness ma-

trix in series with respect to the circular frequency x and

neglecting higher terms than the third order, the fol-

lowing expansion is obtained in the decomposed form:

k

Ã

¼ k þ jxc À x

2

m; ð22Þ

where k is the static stiﬀness matrix; c, the damping

matrix and m, the mass matrix for the uniform beam

with ﬂexible springs and dashpots at its ends.

It should be noted that if neither eccentricities nor

springs and dashpots are present, the matrix k trans-

forms to the classical beam element ﬂexural stiﬀness

matrix k

0

. The matrix c is consistent element damping

matrix, which is based on physical properties of the

member. The elements of the matrix c and matrix m for

a uniform beam according to the linear analysis are

provided in Appendix A.

The proposed viscous damping at beam ends causes

that viscously damped system does not satisfy Caughey

and O’Kelly’s condition [34]. The response of a multi-

degree-of-freedom system cannot be expressed as a lin-

ear combination of its corresponding modal responses.

So, the system is nonclassically damped and it generally

has complex valued natural modes. It is necessary to

elucidate physical interpretation of solutions represented

by complex conjugate pairs of characteristic values. In

order to establish the relationship between coeﬃcient c

i

of viscous damping in joints and modal relative damping

factor f

k

for k mode shape, a speciﬁc procedure has been

established.

Provided that the amount of damping in the system is

not very high, the characteristic values occur in complex

conjugate pairs with either negative or zero real parts.

Let k

i

and

kk

i

be a pair of characteristic values deﬁned by:

k

i

¼ Àe

i

þ jx

di

; ð23aÞ

kk

i

¼ Àe

i

À jx

di

: ð23bÞ

Further, let xx

di

be the modulus of k

i

, i.e.:

xx

di

¼

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

e

2

i

þ x

2

di

_

: ð24Þ

The corresponding pseudo-damping factor 11

i

is:

11

i

¼

e

i

xx

di

: ð25Þ

Based on the parametric study, the relationship between

coeﬃcient c

i

of viscous damping in joints and modal

pseudo-relative damping factor 11

i

for i mode shape can

be obtained, and the corresponding curve is presented in

Fig. 2.

3. Semi-rigid connection modelling

Numerous experimental results have shown that the

connection moment–rotation relationships are nonlinear

over the entire range of loading for almost all types of

connections. To describe connection behavior, diﬀerent

mathematical models have been proposed. In this study,

the three parameter power model proposed by Richard

and Abbott [8] and Kishi et al. [27] is used to represent

moment–rotation behavior of the connection under

monotonic loading. This model can be formulated as:

M ¼

k

0

h

½1 þ ðh=h

0

Þ

n

1=n

; ð26Þ

where k

0

initial connection stiﬀness; n, the shape pa-

rameter; h

0

¼ M

u

=k

0

, the reference plastic rotation and

M

u

, the ultimate moment capacity. Accordingly to Eq.

(26), M–h functions for the two types of connection

(double web angle (DWA), top and seat angle with

double web angle (TSDWA)) are shown in Fig. 3a. The

ﬁrst of these connections are rather weak and the second

is relatively stiﬀ. The details of these connections can be

found in Ref. [35].

The independent hardening model was adopted to

simulate the inelastic connection behavior under cyclic

loading. In this model, the characteristics of connec-

tions are assumed to be unchanged through the loading

Fig. 2. Relationship between coeﬃcient c

i

and modal pseudo-

relative damping factor 11

i

.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 939

cycles. The moment–rotation curve under the ﬁrst cycle

of loading unloading and reverse loading remain un-

changed under the repetititon of loading cycles. The

skeleton curve used in the model was obtained from

three parameter power model. The cyclic moment–

rotation curve based on this model is schematically

shown in Fig. 3b. The independent hardening model is

simple and easily applicable to all types of steel frames

connection models. More information about this model

can be found in Refs. [35,36].

4. Numerical procedures

The equations of motion of a frame subjected to

dynamic loading can be written in the following form:

´

MM

€

UU þ

´

CC

_

UU þ

´

KKU ¼ F; ð27Þ

or

´

MM

€

UU þ

´

CC

_

UU þ

´

KKU ¼ À

´

MM

€

UU

g

; ð28Þ

in which

´

MM is the mass matrix,

´

CC is viscous damping

matrix and

´

KK is static stiﬀness matrix for the system of

structural elements. The time dependent vectors

€

UU,

_

UU

and U are the relative node accelerations, velocities and

displacements respectively, while the vectors F and

€

UU

g

are externally applied loads and ground accelerations.

The equations of motions are integrated using step-by-

step integration, with a constant acceleration assump-

tion within any step.

To solve the nonlinear equations, that are nonlinear

in terms of the displacements as well as the axial force,

secant stiﬀness method is used. This method is very

simple in computer implementation and also gives con-

vergent solutions for design loadings. In each time step,

the load increment DF or D

€

UU

g

is divided into a few

smaller subincrements (Fig. 4) and iterative procedures

are employed. The iterative algorithm is based on eval-

uating secant stiﬀness matrix, which depends on the

stiﬀness of connections, represented by slope of its mo-

ment–rotation curve at any particular moment value.

The convergence is obtained when the diﬀerences be-

Fig. 3. (a) Three parameter power model and (b) independent

hardening model.

Fig. 4. Secant stiﬀness method in a case of nonlinear connec-

tion behavior.

940 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

tween two consecutive cycles displacements at all joints

reach the prescribed tolerance. The current connection

stiﬀness becomes the starting connection stiﬀness for the

next load subincrement. The convergent solutions for all

load subincrements are accumulated to obtain the total

nonlinear response within time step.

5. Numerical examples

Based on the above theoretical considerations, a

computer program has been developed and dynamic

analysis of plane frames with diﬀerent number of stories

and bays, as well as diﬀerent types of connections and

loads, has been performed. For illustration, only some

typical results are presented herein.

5.1. Ten-storey single bay frame

Ten-storey single bay plane steel frame 40.00 m high

and 8.00 m wide has been analyzed. The geometrical and

material properties of this frame are shown in Fig. 5.

Two types of semi-rigid beam-to-column connections

(TSDWA and DWA) with both linear and nonlinear

moment–rotation relations were considered. For com-

parison the same frame with rigid joints was analyzed.

Linear (ﬁrst order) and a geometrically nonlinear (sec-

ond order) analyses of the frame were carried out for all

aforementioned connection types. The results of linear

and nonlinear analyses of the frame with fully rigid

joints obtained by this study have been compared with

the corresponding results obtained by the known soft-

ware package SAP 2000 [37]. These two sets of results

are quite close to each other. For the semi-rigid type

connections database developed by Chen and Kishi [38]

was used. The following examples include vibration and

transient analysis of steel frame shown in Fig. 5, if

subjected to uniform and seismic ground excitations.

5.1.1. Natural frequencies

The natural frequencies and the corresponding peri-

ods for the ﬁrst three modes are determined for the cases

of fully rigid and linear semi-rigid connection (DWA,

TSDWA) and shown in Table 1. The change in natural

frequencies due to variation of joint stiﬀness (ﬁxity fac-

tor) is shown in Fig. 6. The natural frequencies are

normalized by dividing their values by the frequencies

obtained for the frame with rigid joints and ﬁxity factors

are deﬁned as:

k

i0

¼

3EI

l

c

i

1 À c

i

_ _

; ð29Þ

where k

i0

is initial connection stiﬀness that varies from 0

in the case of pinned connection to 1 for the case of

fully rigid connection and c

i

is ﬁxity factor whose values

are normalized form 0 to 1.

As seen in Fig. 6, the connection ﬂexibility has a sig-

niﬁcant inﬂuence on variation of the natural frequencies

particularly on the lower frequencies. This fact can be

Fig. 5. Layout and properties of single-bay ten-storey frame

investigated.

Table 1

Natural frequencies of the frame investigated

Type of

connection

Natural frequencies (rad/s) Periods (s)

First mode Second mode Third mode First mode Second mode Third mode

Rigid 6.328 17.523 31.116 0.993 0.359 0.202

TSDWA 5.727 16.088 28.611 1.097 0.391 0.220

DWA 4.647 13.519 24.247 1.352 0.465 0.259

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 941

very important for seismic analysis of frame structures,

as the lower modes may generally have the principal in-

ﬂuence on seismic response of buildings.

Fig. 7 shows the inﬂuence of eccentricity beam-

to-column connection on the variation of natural

frequencies. It can be seen that the eccentricity of con-

nections may have a practical inﬂuence depending on

the type and size of the connection.

5.1.2. Transient analysis

The transient displacement analysis of the frame

shown in Fig. 5 has been performed for the two cases of

ground motions: two steps sudden acceleration and an

earthquake excitation. Gravitational loads are also in-

cluded and they are considered as additional lumped

masses at the beam nodes.

5.1.3. Two-steps ground acceleration

The frame is assumed to be subjected to the sudden

discontinuous two-steps uniform ground acceleration

shown in Fig. 10. The transient response analysis of the

frame with various connection types according to the

ﬁrst order and second order analyses has been carried

out. Characteristic results of the lateral displacements

and accelerations at the top of frame as well as bending

moments and shear forces at the base of the frame for

the various types of connections are presented in Table

2. The eﬀects of viscous damping at joint connections on

the deﬂection and internal forces of the frame are also

included. The envelopes of lateral displacement and

shear force of the frame with various connection types

according to the ﬁrst order and second order analyses

are plotted in Figs. 8 and 9.

It can be seen from Table 2 and Figs. 8 and 9 that the

frame with semi-rigid connections has a larger lateral

displacements, but smaller shear forces when compared

with the fully rigid connection. These diﬀerences in-

crease with decrease in the connection stiﬀness. Conse-

quently, the diﬀerence in maximum displacement at the

top of the frame with rigid joints and semi-rigid type of

joints are (in percent): 15.1 and 48.6 for linear or 19.1

and 64.0 for nonlinear types of TSDWA and DWA

connections, respectively. The diﬀerences (in percent) in

the shear forces at the base of the frame are: 3.9 and 13.2

for linear or 5.5 and 20.9 for nonlinear types of TSDWA

and DWA connections, respectively.

It is obvious that there is a signiﬁcant diﬀerence be-

tween the results obtained for the frame with rigid joints

and the frames with semi-rigid (DWA and TSDWA)

connections especially in the case of the weak connec-

tions type (DWA).

The time histories of the lateral displacements at the

left top node of the frame with various connection types

according to the ﬁrst order and second order analyses

are plotted in Fig. 10. It can be seen from Fig. 10 that

the frame with nonlinear connections has longer am-

plitudes and period when compared with the rigid joint

case. The displacement amplitudes and period increase

with a decrease in joint stiﬀness. They are longer in

DWA connection case than in the TSDWA connection

case. Besides, the nonlinear connections dampen and

produced nonrecoverable deﬂection due to the presence

of permanent deformations at connections. On the

contrary, the frame with either fully rigid and linear

Fig. 6. The inﬂuence on connection ﬂexibility on the natural

frequencies.

Fig. 7. The inﬂuence of connection eccentricity on the natural

frequencies.

942 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

connection types produces no hysteretic damping. Fig.

10 shows also that the second order analysis further

magniﬁes the aforementioned nonlinear eﬀects on the

frame deﬂection response.

Fig. 8. Lateral displacement envelopes (a) and shear force envelopes (b) according to the ﬁrst order analysis.

Table 2

Maximum displacements and internal forces of frame investigated

Type of

connec-

tions

Maximum displacement

of node A (cm)

Maximum bending mo-

ment of node B (kNm)

Maximum shear force of

node B (kN)

Maximum acceleration

of node A (m/s

2

)

First order Second order First order Second order First order Second order First order Second order

Rigid

Present

study

2.84 3.10 161.16 166.59 58.00 55.86 0.863 0.736

SAP2000 2.85 3.10 161.50 166.60 58.13 55.81 0.905 0.953

TSDWA

Linear

c ¼ 0

3.27 3.58 164.33 174.39 53.39 51.98 1.23 1.31

Linear

c ¼ 50; 000

2.96 3.24 154.49 162.27 50.26 48.47 0.696 0.767

Nonlinear

c ¼ 0

3.38 3.66 166.39 176.23 52.43 50.39 1.05 1.13

Nonlinear

c ¼ 50; 000

3.00 3.30 153.66 161.98 48.10 45.95 0.673 0.709

DWA

Linear

c ¼ 0

4.22 4.60 178.02 182.67 47.59 42.61 1.38 0.712

Linear

c ¼ 50; 000

2.96 3.46 137.93 153.12 37.64 36.34 0.475 0.479

Nonlinear

c ¼ 0

4.66 5.66 177.51 203.79 40.03 34.95 0.705 0.765

Nonlinear

c ¼ 50; 000

3.68 5.36 150.98 192.86 35.58 34.45 0.460 0.465

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 943

Fig. 9. Lateral displacement envelopes (a) and shear force envelopes (b) according to the second order analysis.

Fig. 10. Time history displacement at the left top node of the frame with various connection types. According to the (a) ﬁrst order

analysis and (b) second order analysis.

944 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

The inﬂuence of viscous damping at connections on

displacement response of frame with some types of

connections is shown in Figs. 11 and 12. These ﬁgures

show that viscous damping alters the deﬂection response

of the frame signiﬁcantly, particularly in case a weak

connection type (DWA). It is obvious that the dis-

placement response of the frame decays with time for

both linear and nonlinear types of connections. In the

linear type connection case plotted in Figs. 11a and 12a

there is only viscous damping, so the frame oscillates

about its initial position. In the nonlinear type connec-

tion case plotted in Figs. 11b and 12b, there are both

viscous and hysteretic damping, so the frame oscillates

about its permanent drift position which exists as a re-

sult of the permanent nonrecoverable rotations of con-

nections.

5.1.4. Earthquake excitation

The frame is assumed to be subjected to the ﬁrst four

seconds of Montenegro earthquake (1979) NS compo-

nent motion shown in Fig. 13b. The peak ground ac-

celeration was 0.4 g at about third second. The

displacement response at the top of the frame with two

types of nonlinear connections and rigid jointed frame

according linear and second order analyses plotted in

Fig. 14. This ﬁgure shows considerable diﬀerence be-

tween the responses of rigid jointed frames and frames

with nonlinear connection types. The main reason for it

is hysteretic damping which exists only in nonlinear

connection case. It reduced transient deﬂection response

gradually decreasing its amplitude with time.

Fig. 14a shows that in both TSDWA and DWA

nonlinear types of connection there are permanent de-

ﬂection drift (due to large connection rotation) to the

positive side (forward permanent deﬂection) in the ﬁrst

case, and to the negative side (backward permanent

deﬂection) in the second case. Fig. 14b also shows that

the frame with DWA connection demonstrates re-

markably diﬀerent response from the others. After

about 3 s in this case, the frame deforms suddenly to a

Fig. 11. The inﬂuence of viscous damping at connections on displacement response of the frame with TSDWA connections. (a) Linear

type of connection and (b) nonlinear type of connection.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 945

Fig. 12. The inﬂuence of viscous damping at connections on displacement response of the frame with DWA connections. (a) Linear

type of connection and (b) nonlinear type of connection.

Fig. 13. Lateral load history. Accelerogram (a) and spectrum (b) for Montenegro earthquake (1979), Petrovac NS component.

946 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

peak value that reaches over 60 cm, and oscillates about

this permanent deﬂection. The main reason for this is

the appearance of large rotational deformations at the

joint connections.

The lateral displacement and shear force envelopes of

the frame with the previous connection types obtained

according linear and second order analyses are shown in

Figs. 15 and 16. The frames with ﬂexible nonlinear con-

nections under applied earthquake motion have smaller

lateral displacements and shear forces when compared

with the rigid jointed frame. It is necessary bear in mind

that any earthquake is an excitation with a wide range of

frequencies. The predominant frequencies of the applied

earthquake are within the range from 2 to 10 Hz (peri-

ods 0.1–0.5 s). The lowest natural frequencies of the

investigated frames (rigid, TSDWA, DWA) are much

higher than the predominant earthquake frequencies,

while the second and the third natural frequencies are

within the range of predominant frequencies of the ap-

plied earthquake (Fig. 13b). It obviously has a great

inﬂuence on displacement response of these frames.

Time history acceleration responses of the frame with

rigid and two types nonlinear (TSDWA, DWA) connec-

tions according to linear and second order analyses are

shown in Fig. 17. It is obvious that there is a substantial

hysteretic damping eﬀect on the acceleration response of

the frame with nonlinear connections. On the contrary,

in the case of rigid jointed frame, the acceleration re-

sponse is not dampened, so the large ampliﬁcation of

the acceleration response exists. For the applied

ground motion, the acceleration is ampliﬁed from the

base to top of the frame by factors 6.5, 3.9 and 1.7 for

rigid, TSDWA and DWA case of connections, respec-

tively.

The hysteresis M–h loops at joint C of the frame with

TSDWA and DWA type of connections are shown in

Fig. 14. Time history displacement with various connection types according to the (a) ﬁrst order analysis and (b) second order analysis.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 947

Fig. 18. It can be seen that the connections undergo

strong rotational deformations during the applied earth-

quake motion.

5.2. Single-bay two-storey frame

For the purpose of comparison of the analysis in this

paper with existing computational methods the single-

bay two-storey frame shown in Fig. 19 has been ana-

lyzed. Vibration and transient response analysis of the

frame were investigated by Chan and Ho [25] and Chan

and Chui [36]. They applied the numerical model based

on the linearized second order theory assuming the lin-

ear and nonlinear types of connections. The stiﬀness and

geometric matrices of the uniform beam with end

springs were obtained using conventional cubic Hermi-

tian shape functions. Two elements per beam and one

element per column were applied. The ﬂush end plate

ﬂexible connection type was assumed and modelled by

Chen–Lui exponential model [39] shown in Fig. 19b.

Fig. 15. Lateral displacement envelopes (a) and shear force envelopes (b) according to the ﬁrst order analysis.

Fig. 16. Lateral displacement envelopes (a) and shear force envelopes (b) according to the second order analysis.

948 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

The transient response of the frame was performed for

the two cases of lateral loads (cyclic and impact) with

and without the presence of gravitational loads.

The same frame has been analyzed by the present

numerical model and the results compared with those

previously obtained by Chan and coworkers [25,36].

Fig. 17. Time history acceleration with various connection types according to the (a) ﬁrst order analysis and (b) second order analysis.

Fig. 18. Hysteresis M–h loops at joint C of the frame with TSDWA (a) and DWA (b) type of connections.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 949

The natural frequencies of the frame for the fully

rigid and linear semi-rigid (ﬂush end plate and TSDWA)

connections have been determined and shown in Table

3. It can be seen that they are agree well with the results

by Chan and Ho.

The time histories of the displacement at node B and

the hysteretic loops at node A of the frame under the

lateral cyclic loads obtained by Chan and Chui and in

the current study are shown in Fig. 20. It can be seen

that the response curves are very close.

Displacement response of the frame subjected to the

impact loads obtained by Chan and Chui and by the

present study is shown in Fig. 21.

It can be seen that the response curves have the same

character, but there are some diﬀerences between their

amplitudes and periods. These diﬀerences are small at

the beginning and they gradually increase with time. As

expected, they are larger in the case the presence than in

the case the absence of gravitational loads. The frame

analyzed by the present study has smaller amplitudes

and nonrecoverable deﬂections, as well as shorter peri-

ods when compared with the same frame analyzed by

Chan and Chui.

6. Conclusion

An eﬃcient method to perform dynamic analyses of

steel frame structures with ﬂexible connections has been

presented in this paper. A numerical model that includes

both nonlinear connection behavior and geometric

nonlinearity of the structure has been developed. The

complex dynamic stiﬀness matrix for a prismatic beam

with rotational springs and dashpots attached at its ends

was obtained in an explicit form. The stiﬀness matrix

was based on the analytical solutions of the second order

equations, so that each beam corresponds to one ﬁnite

element.

On the bases of the above theoretical considerations

and the results of the applied numerical analysis, it is

evident that the ﬂexible joint connections greatly inﬂu-

ence the dynamic behavior of steel frames. The connec-

tion ﬂexibility may signiﬁcantly alter both vibration

and the response of frames. An increase in the connec-

tion ﬂexibility reduces the frame stiﬀness, and thus the

eigenfrequencies, particularly the lower ones, which may

have a primary inﬂuence on dynamic response of the

structure.

From the results of numerical examples it can be

concluded that the structural responses of the frames

with nonlinear connections and the frames with con-

ventional type of connections (rigid or linear) are con-

siderably diﬀerent. It shows that the eﬀect of hysteretic

damping on structural response is signiﬁcant. Therefore,

the nonlinear constitutive model for connections should

be used in design and response analysis of real frame

structures. The linear model is inadequate as it cannot

represent a hysteretic behavior of connection under cy-

clic loads.

From the results, it can also be concluded that the

viscous damping at connections may considerably re-

duce the displacement response and internal forces of

the frame, particularly in the case of weak connection

types. The inﬂuence of the geometric nonlinearity in-

creases with the gravitational loads and the lateral frame

deﬂections. It is higher in the frame with ﬂexible con-

nections than with rigid joints.

The connections are vital structural components that

are very often responsible for the behavior and safety of

frame structures subjected to strong dynamic (seismic)

loads. Therefore, connection design and modelling have

a great practical importance.

Fig. 19. Single-bay two-storey frame. (a) Layout and (b) moment–rotation curves of ﬂush end plate connection.

Table 3

Natural frequencies of the single-bay two-storey frame

Type of

connection

Natural frequencies (rad/s)

Present work Chan and Ho

Rigid 10.50 10.35

Flush end plate 7.28 7.30

TSDWA 7.11 7.13

950 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

Appendix A.

The interpolation functions N

i

ðxÞ, i ¼ 1; . . . ; 4, for the

compressive member ðN < 0Þ, are:

N

1

ðxÞ ¼ D

À1

½1 À cos x À xsinx þ xn sinx

À sin xsinxn þ ð1 À cos xÞ cos xn;

N

2

ðxÞ ¼ lðDxÞ

À1

½xcos x À sin x þ xnð1 À cos xÞ

þ ð1 À cos x À xsinxÞ sinxn

þ ðsinx À xcos xÞ cos xn;

N

3

ðxÞ ¼ D

À1

½1 À cos x À xn sin x þ sin xsinxn

À ð1 À cos xÞ cos xn;

N

4

ðxÞ ¼ lðDxÞ

À1

½sinx À x þ xnð1 À cos xÞ

À ð1 À cos xÞ sinxn þ ðx À sinxÞ cos xn;

where

D ¼ 2ð1 À cos xÞ À xsin x; n ¼

x

l

; x ¼ l

ﬃﬃﬃﬃﬃﬃﬃ

N j j

EI

_

:

Fig. 20. Dynamic behavior of two-storey frame under cyclic loads. (a) Displacement response at node B and (b) hysteretic loops at

node A.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 951

The shape functions for the tensile member ðN > 0Þ can

be obtained from the foregoing expressions replacing

x ¼ jx, and using the relations shx ¼ Àj sin jx and

chx ¼ cos jx.

The elements of the correction matrix G are:

g

21

¼ Àg

23

¼

À6

lD

½g

1

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

;

g

22

¼

À6e

1

lD

½g

1

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

À 4½g

1

/

3

þ g

1

g

2

ð4/

2

3

À /

2

4

Þ;

g

24

¼

À6e

2

lD

½g

1

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

À 2g

1

/

4

;

g

41

¼ Àg

43

¼

À6

lD

½g

2

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

;

g

42

¼

À6e

1

lD

½g

2

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

À 2g

2

/

4

;

g

44

¼

À6e

2

lD

½g

2

þ 2g

1

g

2

ð2/

3

À /

4

Þ/

2

À 4½g

2

/

3

þ g

1

g

2

ð4/

2

3

À /

2

4

Þ;

D ¼ ð1 þ 4g

1

/

3

Þð1 þ 4g

2

/

3

Þ À 4g

1

g

2

/

2

4

; g

i

¼

EI

lk

i

i ¼ 1; 2:

Fig. 21. Displacement response of two-storey frame under impact loads. (a) Without gravity and (b) with gravity loads.

952 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

The elements of matrix k

1

according to the second

order theory are:

k

1 11

¼ Àk

1 13

¼ k

1 33

¼ À

EI

l

3

D

12l/

2

ða þ bÞ;

k

1 22

¼ À

EI

l

3

D

4l

2

ð2/

3

c þ /

4

f Þ;

k

1 12

¼ Àk

1 23

¼ À

EI

l

3

D

½6l/

2

ðc þ f Þ þ 2l

2

ð2/

3

a þ /

4

bÞ;

k

1 44

¼ À

EI

l

3

D

4l

2

ð2/

3

d þ /

4

eÞ;

k

1 14

¼ Àk

1 34

¼ À

EI

l

3

D

½6l/

2

ðd þ eÞ þ 2l

2

ð2/

3

b þ /

4

aÞ;

k

1 24

¼ À

EI

l

3

D

2l

2

½2/

3

ðe þ f Þ þ /

43

ðd þ cÞ;

k

1 jk

¼ k

1 kj

:

The elements of matrix k

2

according to the second

order theory are:

k

2 11

¼ Àk

2 13

¼ k

2 33

¼

EI

lD

2

4/

3

ða

2

_

þ b

2

Þ þ 4/

4

ab þ

a

2

g

1

þ

b

2

g

2

_

;

k

2 12

¼ Àk

2 23

¼

EI

lD

2

4/

3

ðac

_

þ bf Þ þ 2/

4

ðaf þ bcÞ þ

ac

g

1

þ

bf

g

2

_

;

k

2 14

¼ Àk

2 34

¼

EI

lD

2

4/

3

ðae

_

þ bdÞ þ 2/

4

ðad þ beÞ þ

ae

g

1

þ

bd

g

2

_

;

k

2 22

¼

EI

lD

2

4/

3

ðc

2

_

þ f

2

Þ þ 4/

4

cf þ

c

2

g

1

þ

f

2

g

2

_

;

k

2 44

¼

EI

lD

2

4/

3

ðd

2

_

þ e

2

Þ þ 4/

4

ed þ

e

2

g

1

þ

d

2

g

2

_

;

k

2 24

¼

EI

lD

2

4/

3

ðce

_

þ df Þ þ 2/

4

ðcd þ ef Þ þ

ce

g

1

þ

df

g

2

_

;

where a ¼ g

21

¼ Àg

23

, b ¼ g

41

¼ Àg

43

, c ¼ g

22

, d ¼ g

44

,

e ¼ g

24

, f ¼ g

42

and e

1

¼ e

2

¼ 0.

Analytical expressions for the functions /

i

,

i ¼ 1; . . . ; 4 can be found in Ref. [35].

The elements of complex dynamic stiﬀness matrix kÃ

according to the ﬁrst order analysis are:

k

Ã

11

¼ Àk

Ã

13

¼ k

Ã

33

¼

12EI

l

3

D

Ã

ð1 þ g

Ã

1

þ g

Ã

2

Þ; k

Ã

14

¼ Àk

Ã

34

¼

6EI

l

2

D

Ã

ð1 þ 2g

Ã

1

Þ;

k

Ã

12

¼ Àk

Ã

23

¼

6EI

l

2

D

Ã

ð1 þ 2g

Ã

2

Þ; k

Ã

44

¼

4EI

lD

Ã

ð1 þ 3g

Ã

1

Þ;

k

Ã

22

¼

4EI

lD

Ã

ð1 þ 3g

Ã

2

Þ; k

Ã

24

¼

4EI

lD

Ã

; k

Ã

jk

¼ k

Ã

kj

g

Ã

i

¼

EI

l

1

k

Ã

i

; k

Ã

i

¼ k

i

þ jxc

i

; j ¼

ﬃﬃﬃﬃﬃﬃﬃ

À1

p

;

D

Ã

¼ 1 þ 4g

Ã

1

þ 4g

Ã

2

þ 12g

Ã

1

g

Ã

2

:

The elements of damping matrix c are:

c

11

¼ Àc

13

¼ c

33

¼

36EI

l

3

D

2

ðh

1

þ h

2

þ 4g

1

h

2

þ 4g

2

1

h

2

þ 4g

2

h

1

þ 4g

2

2

h

1

Þ;

c

12

¼ Àc

23

¼

12EI

l

2

D

2

ð2h

1

þ h

2

þ 2g

1

h

2

þ 10g

2

h

1

þ 12g

2

2

h

1

Þ;

c

14

¼ Àc

34

¼

12EI

l

2

D

2

ðh

1

þ 2h

2

þ 10g

1

h

2

þ 12g

2

1

h

2

þ 4g

2

h

1

Þ;

c

22

¼

4EI

lD

2

ð4h

1

þ h

2

þ 24g

2

h

1

þ 36g

2

2

h

1

Þ;

c

24

¼

8EI

lD

2

ðh

1

þ h

2

þ 3g

1

h

2

þ 3g

2

h

1

Þ;

c

44

¼

4EI

lD

2

ðh

1

þ 4h

2

þ 24g

1

h

2

þ 36g

2

1

h

2

Þ; c

jk

¼ c

kj

;

where h

i

¼ ðc

i

EIÞ=ðlk

2

i

Þ, i ¼ 1; 2.

The elements of mass matrix m are:

m

11

¼ Àm

13

¼ m

33

¼

36EI

l

3

D

3

g

1

g

2

ðg

2

h

2

1

þ 8g

2

2

h

2

1

þ 20g

3

2

h

2

1

þ 16g

4

2

h

2

1

À 4g

1

g

2

h

1

h

2

À 8g

2

1

g

2

h

1

h

2

À 8g

1

g

2

2

h

1

h

2

À 16g

2

1

g

2

2

h

1

h

2

þ g

1

h

2

2

þ 8g

2

1

h

2

2

þ 20g

3

1

h

2

2

þ 16g

4

1

h

2

2

Þ;

m

12

¼ Àm

23

¼

12EI

l

2

D

3

g

1

g

2

ð2g

2

h

2

1

þ 18g

2

2

h

2

1

þ 52g

3

2

h

2

1

þ 48g

4

2

h

2

1

À 6g

1

g

2

h

1

h

2

À 8g

2

1

g

2

h

1

h

2

À 16g

1

g

2

2

h

1

h

2

À 24g

2

1

g

2

2

h

1

h

2

þ g

1

h

2

2

þ 6g

2

1

h

2

2

þ 8g

3

1

h

2

2

Þ;

m

14

¼ Àm

34

¼

12EI

l

2

D

3

g

1

g

2

ðg

2

h

2

1

þ 6g

2

2

h

2

1

þ 8g

3

2

h

2

1

À 6g

1

g

2

h

1

h

2

À 16g

2

1

g

2

h

1

h

2

À 8g

1

g

2

2

h

1

h

2

À 8g

1

g

2

2

h

1

h

2

À 24g

2

1

g

2

2

h

1

h

2

þ 2g

1

h

2

2

þ 18g

2

1

h

2

2

þ 52g

3

1

h

2

2

þ 48g

4

1

h

2

2

Þ;

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 953

m

22

¼

4EI

lD

3

g

1

g

2

ð4g

2

h

2

1

þ 40g

2

2

h

2

1

þ 132g

3

2

h

2

1

þ 144g

4

2

h

2

1

À 8g

1

g

2

h

1

h

2

À 24g

1

g

2

2

h

1

h

2

þ g

1

h

2

2

þ 4g

2

1

h

2

2

Þ

m

24

¼

8EI

lD

3

g

1

g

2

ðg

2

h

2

1

þ 7g

2

2

h

2

1

þ 12g

3

2

h

2

1

À 5g

1

g

2

h

1

h

2

À 12g

2

1

g

2

h

1

h

2

À 12g

1

g

2

2

h

1

h

2

À 36g

2

1

g

2

2

h

1

h

2

þ g

1

h

2

2

þ 7g

2

1

h

2

2

þ 12g

3

1

h

2

2

Þ;

m

44

¼

4EI

lD

3

g

1

g

2

ðg

2

h

2

1

þ 4g

2

2

h

2

1

À 8g

1

g

2

h

1

h

2

À 24g

2

1

g

2

h

1

h

2

þ 4g

1

h

2

2

þ 40g

2

1

h

2

2

þ 132g

3

1

h

2

2

þ 144g

4

1

h

2

2

Þ

m

jk

¼ m

kj

:

References

[1] Jones SW, Kirby PA, Nethercot DA. The analysis of

frames with semi-rigid connections—a state-of-the-art

report. J Construct Steel Res 1983;3(2):2–13.

[2] Nethercot DA. Steel beam-to-column connections—a re-

view of test data. London: CIRIA; 1985.

[3] Davisson JB, Kirby PA, Nethercot DA. Rotational stiﬀ-

ness characteristics of steel beam-to-column connections. J

Construct Steel Res 1987;8:17–54.

[4] Moree DB, Nethercot DA, Kirby PA. Testing steel frames

at full scale: appraisal of results and implications for

design. Struct Eng 1993;71:428–35.

[5] Monforton GR, Wu TS. Matrix analysis of semi-rigidly

connected frames. J Struct Div, ASCE 1963;89(ST6):13–

42.

[6] Ang KM, Morris GA. Analysis of three-dimensional frame

with ﬂexible beam-column connection. Can J Civil Eng

1984;11:241–54.

[7] Chan SL, Zhau ZH. A pointwise equilibrating polynomial

(PEP) element for nonlinear analysis frames. J Struct Eng,

ASCE 1994;120(6):1703–17.

[8] Richard RM, Abbott BJ. Versatile elastic-plastic stress-

strain formula. J Eng Mech Div, ASCE 1975;101(EM4):

511–5.

[9] Lui EM, Chen WF. Analysis and behaviour of ﬂexibly-

jointed frames. Eng Struct 1986;8(2):107–18.

[10] Kishi N, Chen WF. Moment–rotation relations of semi-

rigid connections with angles. J Struct Eng, ASCE

1990;116(ST7):1813–34.

[11] Wu FS, Chen WF. A design model for semi-rigid connec-

tions. Eng Struct 1990;12(2):88–97.

[12] Popov EP, Pinkey RB. Cyclic yield reversal in steel

building connections. J Struct Div, ASCE 1969;95(3):

327–53.

[13] Vedero VT, Popov EP. Beam-column sub assemblages

under repeated loading. J Struct Div, ASCE 1972;98(5):

1137–59.

[14] Tsai KC, Popov EP. Cyclic behaviour of end-plate

moment connections. J Struct Div, ASCE 1990;116(11):

2917–30.

[15] Popov EP, Bertero VV. Cyclic loading of steel beams

and connections. J Struct Div, ASCE 1973;99(6):1189–

204.

[16] Krawinkler H, Bertero VV, Popov EP. Inelastic behavior

of steel beam-to-column subassemblages. Report No.

UBC/EERC-71/07 Earthquake Engineering Research Cen-

ter, University of California, Berkeley, CA, 1971.

[17] Stelmack TW, Marley MJ, Gerstle KH. Analysis and tests

of ﬂexibly connected steel frames. J Struct Eng, ASCE

1986;112(7):1573–88.

[18] Korol RM, Ghobarah A, Osman A. Extended end-plate

connections under cyclic hading: behavior and design. J

Construct Steel Res 1990;16:253–80.

[19] Lionberger SR, Weaver JW. Dynamic response of frames

with non-rigid connections. J Eng Mech Div, ASCE

1969;95(EM1):95–114.

[20] Kawashima S, Fujimoto T. Vibration analysis of frames

with semi-rigid connections. Comput Struct 1984;19:85–

92.

[21] Suarez LE, Singh MP, Matheu EE. Seismic response of

structural frameworks with ﬂexible connections. Comput

Struct 1996;58(1):27–41.

[22] Sivakumaran KS. Seismic response of multistorey steel

buildings with ﬂexible connections. Eng Struct 1988;10:

239–48.

[23] Yousef-Agha W, Aktan HM, Olowokere OD. Seismic

response of low-rise steel frames. J Struct Div, ASCE

1989;115(3):594–607.

[24] Shi G, Atluri SN. Static and dynamic analysis of space

frames with non-linear ﬂexible connections. Int J Numer

Meth Eng 1989;28:2635–50.

[25] Chan SL, Ho GWM. Nonlinear vibration analysis of steel

frames with semi-rigid connections. J Struct Eng, ASCE

1994;120:1075–87.

[26] Sekulovic M, Salatic R, Mandic R. Seismic analysis of

frames with semi-rigid eccentric connections. Twelfth

Conference on Earthquake Engineering. New Zealand

2000; paper no. 273.

[27] Kishi N, Chen WF, Goto Y, Matsuoka KG. Design aid of

semi-rigid connections for frame analysis. Eng J, AISC

1993;30(3):90–107.

[28] Goto Y, Suzuki S, Chen WF. Analysis of critical

behaviour of semi-rigid frames with or without load his-

tory in connections. Int J Solids Struct 1991;27(4):467–

83.

[29] Goto Y, Suzuki S, Chen WF. Stability behaviour of semi-

rigid sway frames. Eng Struct 1993;15(3):209–19.

[30] Huang J, Morris G. Analysis of ﬂexibly connected frames

under nonproportional loading. In: Bjorhovode R et al.,

editors. Proc of Second Int Workshop on connections in

steel structures. Pitsburg, Chicago: AISC; 1991.

[31] Sekulovic M, Salatic R. Nonlinear analysis of frames with

ﬂexible connections. Comput Struct 2001;79(11):1097–

107.

[32] Sekulovic M, Malcevic I. Second-order analysis of frame

structures. Theor Appl Mech 1994;20:209–34.

[33] Goto Y, Chen WF. Second-order elastic analysis for frame

design. J Struct Div, ASCE 1987;113(7):1501–19.

954 M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

[34] Caughey TH, O’Kelly MEJ. Classical normal modes in

damped linear dynamic systems. J Appl Mech, ASME

1965;32:583–8.

[35] Chen WF, Goto Y, Liew JYR. Stability design of semi-

rigid frames. New York: John Willey & Sons; 1996.

[36] Chan SL, Chui PPT. Nonlinear static and cyclic analysis of

steel frames with semi-rigid connections. Amsterdam:

Elsevier; 2000.

[37] SAP 2000. Integrated ﬁnite element analysis and design of

structures. Berkeley, California: Computers and Struc-

tures, Inc.; 1996.

[38] Chen WF, Kishi NN. Semirigid steel beam-to-column

connections data base and modeling. J Struct Eng, ASCE

1989;120(6):1703–17.

[39] Lui EM, Chen WF. Behavior of braced and unbraced

semi-rigid frames. Int J Solids Struct 1988;24(9):893–913.

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955 955

936

M. Sekulovic et al. / Computers and Structures 80 (2002) 935–955

matrix for a uniform beam element with linear rotational springs and dashpots at its ends. The inﬂuence of the ﬂexibility and eccentricity in the connections on dynamic behavior of plane frames within the linear theory was investigated by Suarez et al. [21]. Although the linear constitutive model of nodal connections is very easy to use, it is inadequate in the term that it is applicable only to a small range of the initial rotations and because it cannot represent hysteretic damping to be a primary source of passive damping in the frame structures. A bilinear moment–rotation function, which is also accurate only for a small rotation range, was used by Sivekumaran [22] and Yousef-Agha et al. [23]. The eﬀect of hysteretic damping resulting from the nonlinear ﬂexible connection on the dynamic response of the frame was studied by Shi and Altury [24]. They developed a numerical model based on the complementary energy approach using the Ang and Morise [6] function for the moment–rotation relations at connections. Chan and Ho [25] proposed a numerical method for linear and nonlinear vibration analysis of frame with semi-rigid connections. They adapted the conventional cubic Hermitian shape functions for a uniform beam with end springs and derived the element matrices using the principle of total potential energy. The inﬂuence of both hysteretic and viscous damping at connections on seismic response of the steel frames was considered by Sekulovic et al. [26]. The combined eﬀects of material yielding and connection ﬂexibility in static and dynamic problems have been discussed in detail by Chan and Chui [36]. Several nonlinear ﬂexible connection models under cyclic loading were established in the past decade [27–30]. The present study is an extension of the author’s previous work [31], regarding static analysis of ﬂexibly joint frames, on a more general case of the dynamic analysis. Two types of nonlinearities are considered: geometric nonlinearity of the structure and material (constitutive) nonlinearity of the connections. These nonlinearities are interactive. The eccentricity of the connections is also considered. To describe the nonlinear behavior of the connection under cyclic loading, the independent hardening model is used. So, the eﬀect of hysteretic damping on dynamic behavior of the structure is directly included through the connection constitutive relation. Moreover, the inﬂuence of viscous damping at connections on dynamic response of frame structures is considered. For a uniform beam with rotational springs and dashpots attached at its ends the complex dynamic stiﬀness matrix is obtained. The stiﬀness matrix has been obtained based on analytical solutions of governing diﬀerential equations second order analysis, so that each beam represents one element. Nodal displacements and rotations are chosen as the primary unknowns, while displacements and rotations of the element ends are eliminated. Thus, the number of degrees of freedom

are the same as for the system with rigid connections. Besides, the consistent mass and damping matrices are derived. These matrices are based on the physical properties of the member and given in an explicit form. The present matrices are more general than the corresponding matrices previously obtained by other authors. Energy dissipation exists in frame structures under dynamic loads. The primary sources of energy dissipation, is known, may be hysteretic behavior of connections and the friction between elements forming the beam–column assemblage. In addition, diﬀerent types of energy dissipation devices can be installed into connections in order to increase the structural energy absorption capacity. For this reason, in the present model, the total energy dissipation is conﬁded to the joint connections. Two types of energy dissipation are assumed. They are: hysteretic damping due to nonlinear behavior of connections and viscous damping at the connections. In general, the eﬀects of these dampings are coupled. Also, they can be considered separately using either linear constitutive relation for the connections or zero value for the viscous damping coeﬃcients at the connections. As it is assumed that all structural elements, except the connections, remain elastic through the whole loading range, the energy dissipation at plastic hinges cannot be observed. Also, energy dissipation due to radiation damping at the supports is not included in this consideration. The other types of energy dissipation that may exist in real frame structures can be included in the present model in the usual way, by mass and stiﬀness proportional damping matrix. Based on theoretical problem formulation, a computer program was developed in order to increase dynamic analysis eﬃciency and design of steel frames. A parametric study has been performed in order to estimate the inﬂuence of nonlinear connection ﬂexibility and viscous damping at connections on the frame dynamic (seismic) response. The present numerical model is restricted to 2D frame systems. It can be extended on a more general case of 3D analysis without diﬃculty. Besides, the proposed beam element can be easily incorporated into existing commercial programs for structural analysis.

2. Formulation of structural element A beam element with ﬂexible, eccentric and viscous damping connections is shown in Fig. 1. The ﬂexible connections are represented by nonlinear rotational springs at beam ends. Thus, only the inﬂuence of bending moment on the connection deformation is considered, while the inﬂuences of axial and shear forces are neglected. The connection spring element is assumed massless and dimensionless in size. The eccentricity is

. ð2aÞ ð2bÞ ð1Þ 2 X ci a2 . i i¼1 ð4cÞ denoting strain energy of the beam. eccentric and viscous damping connections. . With the assumption that k 2 ¼ const. kII ¼ EI 0 ð5Þ ð6aÞ ð6bÞ kef ¼ GT kII þ kII G þ GT kII G.. (1) into Eq. ð7Þ . Analytical expression for the elements of matrix kII and the appropriate expansions in the power series form. 2A EI 2 Z 0 l ð3Þ ð4aÞ Uf ¼ Fig.1.M. . i ¼ 1. the number of degrees of freedom is the same as for the beam element with fully rigid connections. the function describing lateral displacement vðxÞ for the element with ﬂexible eccentric connections can be written in the usual way by interpolation function matrix and nodal displacements vector as: e vðxÞ ¼ NðxÞðI þ GÞq ¼ N ðxÞq. / Computers and Structures 80 (2002) 935–955 937 0 1 6 g21 G¼ 6 D4 0 g41 2 De1 g22 0 g42 0 g23 0 g43 3 0 g24 7 7. The linear viscous damping at nodal connections are modelled by dashpots acting at beam ends. Consequently. these two part of the strain energy can be expressed independently. after substituting Eq. the nodal displacement vector and the correction matrix. 4 and elements of correction matrix G are given in Appendix A. In that case. convenient for the numerical analysis. denoting beam stiﬀness matrix with the rigid connections according to the second order analysis and correction matrix that accounts for the eﬀects of ﬂexibility and eccentricity. A beam with ﬂexible. 2. ÀDe2 5 g44 ð2cÞ denoting the matrix of interpolation functions obtained based on the analytical solutions of the second order analysis equations [32]. Interpolation functions Ni ðxÞ. where NðxÞ ¼ ½ N1 ðxÞ N2 ðxÞ N3 ðxÞ N4 ðxÞ . corresponding to the linearized second order analysis. Stiﬀness matrix for the beam element with ﬂexible eccentric connection is obtained through the total potential energy. are given by Goto and Chen [33]. while displacements and rotations of the beam ends can be eliminated as has been shown in [31].xx ð4bÞ Us ¼ modelled by short inﬁnitely stiﬀ elements whose lengths are e1 and e2 . v2 dx. where Ua ¼ EI 2 4 k l. . that can be written as U ¼ Ua þ Uf þ Us . Sekulovic et al. Thus. Thus. Joint displacements and rotations are the primary unknowns. qT ¼ f v1 u1 v2 u2 g. (4b) the following can be obtained: Uf ¼ 1qT ðkII þ kef Þq. axial (Ua ) and ﬂexural (Uf ) and strain energy of the springs (Us ). Strain energy due to axial deformation and bending are coupled since parameter k 2 includes derivatives of both axial and lateral displacements [31]. 1. Stiﬀness matrix and nodal force vector The stiﬀness matrix and the nodal force vector for the ﬂexible eccentric beam have been represented in a previous work by the authors [31]. It will be brieﬂy summarized herein. . The simpliﬁed form of this matrix. the stiﬀness matrix has the form kII ¼ k0 þ kg . respectively. can also be used. respectively. 2 where matrices kII and kef are deﬁned as Z lh i ðN00 ðxÞÞT N00 ðxÞ dx.

ð16Þ The equivalent generalized end force vector due to distributed loads along the beam pðxÞ is obtained in the usual maner: Z l Z l e Q¼ pðxÞ N T ðxÞ dx ¼ ðI þ GÞT pðxÞNT ðxÞ dx: ð11Þ 0 0 where complex ﬂexural stiﬀness kiÃ of the connection is deﬁned as the ratio between moment and relative rotation amplitudes: kiÃ ¼ Mi ðtÞ ¼ ki þ jxci . q where RðtÞ ¼ f T 1 ðtÞ M 1 ðtÞ T 2 ðtÞ M 2 ðtÞ g. v v hT ðtÞ ¼ f 0 h1 ðtÞ 0 h2 ðtÞ g. . . From Eqs. . with functions /i . are replaced by 1.938 M. (1). . (1) into Eq. rotational viscous dashpots are attached at beam ends. i ¼ 10 . 2 where ks ¼ G CG: T 2. the simpliﬁed form of the matrix G. m0 denotes conventional mass matrix for the beam element and mef denotes the mass correction matrix. can be used. ð17aÞ ð17bÞ Components distributions closed form vector Q are of the vector Q. i ¼ 1.3. After substitution of Eq. The strain energy of the springs can be expressed in the form Us ¼ 1qT ks q. the element consistent mass matrix m can be formulated as: Z e e ð12Þ m ¼ q N T ðxÞ N ðxÞ dx. hi ðtÞ j¼ pﬃﬃﬃﬃﬃﬃﬃ À1. while k is the classical or the second order ﬂexural stiﬀness matrix of a uniform beam. elements of the computed numerically. V qðtÞ ¼ f 1 ðtÞ u1 ðtÞ 2 ðtÞ u2 ðtÞ g. 1. ð21aÞ ð21bÞ In the above relations. as shown in Fig. that depends on the type of analysis. Element mass matrix Assuming that the mass density q is constant. The tangent or secant form of the above relation may be written if nonlinear springs and/or dashpots are considered. Complex dynamic stiﬀness matrix Apart from nonlinear rotational springs. . i ¼ 10 . In the case of periodic response with circular frequency x the following relation between the amplitudes may be derived: _ Mið0Þ ¼ kiÃ hið0Þ . The total moment at each nodal connection (i ¼ 1. while dot over the symbols denotes diﬀerentiation with respect to time. for some simple load and temperature change are given in the in [32]. can be written as: m ¼ m0 þ mef .0. where Z m0 ¼ qNT ðxÞNðxÞ dx. for the uniform beam with ﬂexible eccentric connections. end displacement vector and relative end rotation vector of the member respectively (Fig. 1). ð15Þ ð8Þ ð9Þ The explicit form of matrices G and C can be found in [31]. the consistent element mass matrix. (12). Eq. 20 . / Computers and Structures 80 (2002) 935–955 where k0 is the conventional stiﬀness matrix and kg is the geometric stiﬀness matrix of the uniform beam. V The beam end force vector RðtÞ can be expressed in terms of the end displacement vector ðtÞ and relative q end rotation vector hðtÞ as: RðtÞ ¼ kfðtÞ À hðtÞg. 4. ð18Þ ð19Þ ð20Þ ð17aÞ e where N ðxÞ is the matrix of modiﬁed shape functions deﬁned by Eq. Sekulovic et al. 2) can be given in terms of relative rotation h between beam end and column face and relative angular _ velocity hðtÞ as: _ Mi ðtÞ ¼ ki hi ðtÞ þ ci hi ðtÞ. ð20Þ mef ¼ GT m0 þ m0 G þ GT m0 G: ð14bÞ where kÃ ¼ ðI À SÞT kðI À SÞ þ ST ks S ¼ k þ k1 þ k2 .2. In general case. After the elimination of relative end rotation vector hðtÞ. (17a) transforms to: q RðtÞ ¼ kÃ ðtÞ. In this case. k1 ¼ ÀkS À ST k. (5) and (8) the total strain energy due to the bending for the beam with ﬂexible and eccentric connections can now be written as: U ¼ Uf þ Us ¼ 1qT ðkII þ kef þ ks Þq: 2 ð10Þ where ki and ci are rotational spring stiﬀness and rotational viscous damping coeﬃcients. ð13Þ ð14aÞ are end force vector. hi ðtÞ ¼ hið0Þ ejxt : 2. 20 .

h0 ¼ Mu =k0 . (26). a speciﬁc procedure has been established. The elements of the matrix c and matrix m for a uniform beam according to the linear analysis are provided in Appendix A. 2. c. The corresponding pseudo-damping factor 1i is: 1i ¼ ei : xdi ð25Þ ð22Þ where k is the static stiﬀness matrix. i. the characteristic values occur in complex conjugate pairs with either negative or zero real parts. the system is nonclassically damped and it generally has complex valued natural modes. [35]. It should be noted that if neither eccentricities nor springs and dashpots are present. n. 3a. let xdi be the modulus of ki . the mass matrix for the uniform beam with ﬂexible springs and dashpots at its ends. Expanding the elements of the dynamic stiﬀness matrix in series with respect to the circular frequency x and neglecting higher terms than the third order. M–h functions for the two types of connection (double web angle (DWA). the three parameter power model proposed by Richard and Abbott [8] and Kishi et al. The elements of this matrix corresponding to the linear analysis are given in Appendix A. To describe connection behavior. which is based on physical properties of the member. and it can be found in Appendix A. Accordingly to Eq. top and seat angle with double web angle (TSDWA)) are shown in Fig. In this model. the damping matrix and m. the relationship between coeﬃcient ci of viscous damping in joints and modal pseudo-relative damping factor 1i for i mode shape can be obtained. [27] is used to represent moment–rotation behavior of the connection under monotonic loading. The matrix c is consistent element damping matrix. the matrix k transforms to the classical beam element ﬂexural stiﬀness matrix k0 . the characteristics of connections are assumed to be unchanged through the loading . In order to establish the relationship between coeﬃcient ci of viscous damping in joints and modal relative damping factor fk for k mode shape. The independent hardening model was adopted to simulate the inelastic connection behavior under cyclic loading.M.: qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ xdi ¼ e2 þ x2 : i di ð23aÞ ð23bÞ Based on the parametric study. 3. The matrix S can be obtained from matrix G putting e1 ¼ e2 ¼ 0. the following expansion is obtained in the decomposed form: k ¼ k þ jxc À x m. Semi-rigid connection modelling Numerous experimental results have shown that the connection moment–rotation relationships are nonlinear over the entire range of loading for almost all types of connections. In this study. and the corresponding curve is presented in Fig. It is necessary to elucidate physical interpretation of solutions represented by complex conjugate pairs of characteristic values. ð26Þ ð24Þ where k0 initial connection stiﬀness. Relationship between coeﬃcient ci and modal pseudo relative damping factor 1i . Let ki and ki be a pair of characteristic values deﬁned by: ki ¼ Àei þ jxdi . the shape parameter. 2. So. The response of a multidegree-of-freedom system cannot be expressed as a linear combination of its corresponding modal responses. the ultimate moment capacity. the reference plastic rotation and Mu . qðtÞ ¼ ð0Þ ejxt : q * ð21cÞ ð21dÞ The explicit forms of matrices k1 and k2 are given in Appendix A. diﬀerent mathematical models have been proposed. / Computers and Structures 80 (2002) 935–955 939 k2 ¼ ST ðk þ KÞS. Sekulovic et al. The details of these connections can be found in Ref. The proposed viscous damping at beam ends causes that viscously damped system does not satisfy Caughey and O’Kelly’s condition [34]. The matrix kÃ is a complex ﬂexural stiﬀness matrix of uniform beam with ﬂexible connection according to the linear or second order analysis. Ã 2 Fig. ki ¼ Àei À jxdi : Further. The ﬁrst of these connections are rather weak and the second is relatively stiﬀ.e. Provided that the amount of damping in the system is not very high. This model can be formulated as: M¼ k0 h ½1 þ ðh=h0 Þn 1=n . including both ﬂexible and viscous phenomena.

U and U are the relative node accelerations. Numerical procedures The equations of motion of a frame subjected to dynamic loading can be written in the following form: c U þ C U þ K U ¼ F. with a constant acceleration assumption within any step. € the load increment DF or DUg is divided into a few smaller subincrements (Fig. 3b. secant stiﬀness method is used. More information about this model can be found in Refs. which depends on the stiﬀness of connections. The moment–rotation curve under the ﬁrst cycle of loading unloading and reverse loading remain unchanged under the repetititon of loading cycles. The skeleton curve used in the model was obtained from three parameter power model. 3. [35. The time dependent vectors U. represented by slope of its moment–rotation curve at any particular moment value. The cyclic moment– rotation curve based on this model is schematically shown in Fig. / Computers and Structures 80 (2002) 935–955 b in which c is the mass matrix. Sekulovic et al. while the vectors F and Ug are externally applied loads and ground accelerations. The iterative algorithm is based on evaluating secant stiﬀness matrix. ð27Þ . In each time step. velocities and € displacements respectively. cycles. Secant stiﬀness method in a case of nonlinear connection behavior.940 M. To solve the nonlinear equations. 4. 4) and iterative procedures are employed.36]. b M€ b _ M€ ð28Þ Fig. This method is very simple in computer implementation and also gives convergent solutions for design loadings. The convergence is obtained when the diﬀerences be- Fig. The equations of motions are integrated using step-bystep integration. The independent hardening model is simple and easily applicable to all types of steel frames connection models. C is viscous damping M b matrix and K is static stiﬀness matrix for the system of € _ structural elements. (a) Three parameter power model and (b) independent hardening model. that are nonlinear in terms of the displacements as well as the axial force. b M€ b _ or c U þ C U þ K U ¼ À c Ug . 4.

5. 5. a computer program has been developed and dynamic analysis of plane frames with diﬀerent number of stories and bays. The change in natural frequencies due to variation of joint stiﬀness (ﬁxity fac- Fig. For comparison the same frame with rigid joints was analyzed.359 0.00 m high and 8. Linear (ﬁrst order) and a geometrically nonlinear (second order) analyses of the frame were carried out for all aforementioned connection types. The convergent solutions for all load subincrements are accumulated to obtain the total nonlinear response within time step. The geometrical and material properties of this frame are shown in Fig. This fact can be Table 1 Natural frequencies of the frame investigated Type of connection Rigid TSDWA DWA Natural frequencies (rad/s) First mode 6. The results of linear and nonlinear analyses of the frame with fully rigid joints obtained by this study have been compared with the corresponding results obtained by the known software package SAP 2000 [37].247 Periods (s) First mode 0. TSDWA) and shown in Table 1. has been performed.352 Second mode 0. Layout and properties of single-bay ten-storey frame investigated. the connection ﬂexibility has a signiﬁcant inﬂuence on variation of the natural frequencies particularly on the lower frequencies.647 Second mode 17. Ten-storey single bay frame Ten-storey single bay plane steel frame 40. / Computers and Structures 80 (2002) 935–955 941 tween two consecutive cycles displacements at all joints reach the prescribed tolerance.523 16.M. The current connection stiﬀness becomes the starting connection stiﬀness for the next load subincrement.220 0. 5.116 28. The natural frequencies are normalized by dividing their values by the frequencies obtained for the frame with rigid joints and ﬁxity factors are deﬁned as: 3EI ci ki0 ¼ . if subjected to uniform and seismic ground excitations.519 Third mode 31. 6.097 1.391 0. Sekulovic et al.259 .611 24. 5. Natural frequencies The natural frequencies and the corresponding periods for the ﬁrst three modes are determined for the cases of fully rigid and linear semi-rigid connection (DWA. only some typical results are presented herein.993 1. For the semi-rigid type connections database developed by Chen and Kishi [38] was used.465 Third mode 0. Numerical examples Based on the above theoretical considerations. 5.088 13.727 4.1. tor) is shown in Fig.1. These two sets of results are quite close to each other. As seen in Fig. Two types of semi-rigid beam-to-column connections (TSDWA and DWA) with both linear and nonlinear moment–rotation relations were considered.00 m wide has been analyzed.1. ð29Þ l 1 À ci where ki0 is initial connection stiﬀness that varies from 0 in the case of pinned connection to 1 for the case of fully rigid connection and ci is ﬁxity factor whose values are normalized form 0 to 1.202 0.328 5. The following examples include vibration and transient analysis of steel frame shown in Fig. For illustration. 5. 6. as well as diﬀerent types of connections and loads.

The diﬀerences (in percent) in the shear forces at the base of the frame are: 3.0 for nonlinear types of TSDWA and DWA connections. Besides. as the lower modes may generally have the principal inﬂuence on seismic response of buildings.3. 5. 7 shows the inﬂuence of eccentricity beamto-column connection on the variation of natural frequencies. Fig.1. Characteristic results of the lateral displacements and accelerations at the top of frame as well as bending moments and shear forces at the base of the frame for the various types of connections are presented in Table 2. On the contrary. The inﬂuence of connection eccentricity on the natural frequencies. The transient response analysis of the frame with various connection types according to the ﬁrst order and second order analyses has been carried out. 10. Gravitational loads are also included and they are considered as additional lumped masses at the beam nodes.1 and 48. Fig. It can be seen from Fig. 7. Transient analysis The transient displacement analysis of the frame shown in Fig. the nonlinear connections dampen and produced nonrecoverable deﬂection due to the presence of permanent deformations at connections. very important for seismic analysis of frame structures. 8 and 9 that the frame with semi-rigid connections has a larger lateral displacements. They are longer in DWA connection case than in the TSDWA connection case. The displacement amplitudes and period increase with a decrease in joint stiﬀness. 6. the frame with either fully rigid and linear . 8 and 9. The eﬀects of viscous damping at joint connections on the deﬂection and internal forces of the frame are also included.2. but smaller shear forces when compared with the fully rigid connection. Sekulovic et al. 10. 5 has been performed for the two cases of ground motions: two steps sudden acceleration and an earthquake excitation.1 and 64. The time histories of the lateral displacements at the left top node of the frame with various connection types according to the ﬁrst order and second order analyses are plotted in Fig.5 and 20.1. It can be seen that the eccentricity of connections may have a practical inﬂuence depending on the type and size of the connection.9 and 13. The inﬂuence on connection ﬂexibility on the natural frequencies.942 M. It can be seen from Table 2 and Figs. Fig. the diﬀerence in maximum displacement at the top of the frame with rigid joints and semi-rigid type of joints are (in percent): 15.9 for nonlinear types of TSDWA and DWA connections. respectively. / Computers and Structures 80 (2002) 935–955 5. The envelopes of lateral displacement and shear force of the frame with various connection types according to the ﬁrst order and second order analyses are plotted in Figs. respectively.2 for linear or 5. Consequently.6 for linear or 19. 10 that the frame with nonlinear connections has longer amplitudes and period when compared with the rigid joint case. It is obvious that there is a signiﬁcant diﬀerence between the results obtained for the frame with rigid joints and the frames with semi-rigid (DWA and TSDWA) connections especially in the case of the weak connections type (DWA). These diﬀerences increase with decrease in the connection stiﬀness. Two-steps ground acceleration The frame is assumed to be subjected to the sudden discontinuous two-steps uniform ground acceleration shown in Fig.

27 176.39 50.673 943 Second order 0.13 53. Sekulovic et al.50 164.33 154. connection types produces no hysteretic damping.39 45.953 1.98 48.23 161.96 3.68 4.64 40. 000 Nonlinear c¼0 Nonlinear c ¼ 50. Lateral displacement envelopes (a) and shear force envelopes (b) according to the ﬁrst order analysis.66 3.736 0.95 34.13 0.03 35.95 Maximum acceleration of node A (m/s2 ) First order 0.31 0.46 5.39 162.23 0.85 3. Fig.81 51.M.39 153.765 0.38 0.10 3. / Computers and Structures 80 (2002) 935–955 Table 2 Maximum displacements and internal forces of frame investigated Type of connections Rigid Present study SAP2000 TSDWA Linear c¼0 Linear c ¼ 50.66 3.36 178.66 Second order 166.84 2.30 Maximum bending moment of node B (kN m) First order 161.12 203.45 1.05 0.10 Second order 55.66 5.49 166.98 Maximum shear force of node B (kN) First order 58. 8.59 37.26 52.38 3.460 0. .696 1.61 36.47 50. 10 shows also that the second order analysis further magniﬁes the aforementioned nonlinear eﬀects on the frame deﬂection response.43 48.00 Second order 3.22 2.00 58.58 3.27 2.16 161.98 182.767 1.60 174.86 55.24 3.51 150.709 4.60 3.712 0.465 Fig. 000 Nonlinear c¼0 Nonlinear c ¼ 50. 000 Maximum displacement of node A (cm) First order 2.96 4.705 0.67 153.93 177.475 0.905 1.479 0.86 47.59 166.863 0.34 34.10 3.02 137.58 42.79 192. 000 DWA Linear c¼0 Linear c ¼ 50.

10. / Computers and Structures 80 (2002) 935–955 Fig. Sekulovic et al. Time history displacement at the left top node of the frame with various connection types. Fig. 9. Lateral displacement envelopes (a) and shear force envelopes (b) according to the second order analysis. According to the (a) ﬁrst order analysis and (b) second order analysis. .944 M.

the frame deforms suddenly to a Fig. In the linear type connection case plotted in Figs. The peak ground acceleration was 0. It reduced transient deﬂection response gradually decreasing its amplitude with time. These ﬁgures show that viscous damping alters the deﬂection response of the frame signiﬁcantly. This ﬁgure shows considerable diﬀerence between the responses of rigid jointed frames and frames with nonlinear connection types. 14. 14b also shows that the frame with DWA connection demonstrates remarkably diﬀerent response from the others. Fig. 11a and 12a there is only viscous damping. there are both viscous and hysteretic damping. and to the negative side (backward permanent deﬂection) in the second case. The inﬂuence of viscous damping at connections on displacement response of the frame with TSDWA connections. particularly in case a weak connection type (DWA). 13b.M. 5. . 11b and 12b.4. It is obvious that the displacement response of the frame decays with time for both linear and nonlinear types of connections. so the frame oscillates about its initial position. The main reason for it is hysteretic damping which exists only in nonlinear connection case. After about 3 s in this case.1. In the nonlinear type connection case plotted in Figs. Fig. so the frame oscillates about its permanent drift position which exists as a result of the permanent nonrecoverable rotations of connections. 11 and 12.4 g at about third second. The displacement response at the top of the frame with two types of nonlinear connections and rigid jointed frame according linear and second order analyses plotted in Fig. (a) Linear type of connection and (b) nonlinear type of connection. 11. / Computers and Structures 80 (2002) 935–955 945 The inﬂuence of viscous damping at connections on displacement response of frame with some types of connections is shown in Figs. Sekulovic et al. 14a shows that in both TSDWA and DWA nonlinear types of connection there are permanent deﬂection drift (due to large connection rotation) to the positive side (forward permanent deﬂection) in the ﬁrst case. Earthquake excitation The frame is assumed to be subjected to the ﬁrst four seconds of Montenegro earthquake (1979) NS compo- nent motion shown in Fig.

Accelerogram (a) and spectrum (b) for Montenegro earthquake (1979). (a) Linear type of connection and (b) nonlinear type of connection.946 M. The inﬂuence of viscous damping at connections on displacement response of the frame with DWA connections. 13. Lateral load history. / Computers and Structures 80 (2002) 935–955 Fig. Fig. Petrovac NS component. 12. Sekulovic et al. .

5. and oscillates about this permanent deﬂection. the acceleration is ampliﬁed from the base to top of the frame by factors 6.M. It obviously has a great inﬂuence on displacement response of these frames. On the contrary. Sekulovic et al. respectively. 17. 14. For the applied ground motion. DWA) connections according to linear and second order analyses are shown in Fig. The main reason for this is the appearance of large rotational deformations at the joint connections. The predominant frequencies of the applied earthquake are within the range from 2 to 10 Hz (periods 0. the acceleration response is not dampened.1–0.5 s). TSDWA. 13b).9 and 1. 15 and 16. peak value that reaches over 60 cm. The hysteresis M–h loops at joint C of the frame with TSDWA and DWA type of connections are shown in . It is necessary bear in mind that any earthquake is an excitation with a wide range of frequencies. 3. in the case of rigid jointed frame. so the large ampliﬁcation of the acceleration response exists. Time history acceleration responses of the frame with rigid and two types nonlinear (TSDWA. The frames with ﬂexible nonlinear connections under applied earthquake motion have smaller lateral displacements and shear forces when compared with the rigid jointed frame. The lateral displacement and shear force envelopes of the frame with the previous connection types obtained according linear and second order analyses are shown in Figs. Time history displacement with various connection types according to the (a) ﬁrst order analysis and (b) second order analysis. It is obvious that there is a substantial hysteretic damping eﬀect on the acceleration response of the frame with nonlinear connections. while the second and the third natural frequencies are within the range of predominant frequencies of the applied earthquake (Fig. / Computers and Structures 80 (2002) 935–955 947 Fig. The lowest natural frequencies of the investigated frames (rigid. DWA) are much higher than the predominant earthquake frequencies.7 for rigid. TSDWA and DWA case of connections.

948 M. 15. 16. Vibration and transient response analysis of the frame were investigated by Chan and Ho [25] and Chan and Chui [36]. It can be seen that the connections undergo strong rotational deformations during the applied earthquake motion.2. 19 has been analyzed. Lateral displacement envelopes (a) and shear force envelopes (b) according to the second order analysis. 18. . The stiﬀness and geometric matrices of the uniform beam with end springs were obtained using conventional cubic Hermitian shape functions. Lateral displacement envelopes (a) and shear force envelopes (b) according to the ﬁrst order analysis. Single-bay two-storey frame For the purpose of comparison of the analysis in this paper with existing computational methods the singlebay two-storey frame shown in Fig. / Computers and Structures 80 (2002) 935–955 Fig. They applied the numerical model based on the linearized second order theory assuming the linear and nonlinear types of connections. 5. Fig. 19b. Sekulovic et al. Two elements per beam and one element per column were applied. Fig. The ﬂush end plate ﬂexible connection type was assumed and modelled by Chen–Lui exponential model [39] shown in Fig.

The same frame has been analyzed by the present numerical model and the results compared with those previously obtained by Chan and coworkers [25. 18. The transient response of the frame was performed for the two cases of lateral loads (cyclic and impact) with and without the presence of gravitational loads. Fig. 17. / Computers and Structures 80 (2002) 935–955 949 Fig. Hysteresis M–h loops at joint C of the frame with TSDWA (a) and DWA (b) type of connections. . Time history acceleration with various connection types according to the (a) ﬁrst order analysis and (b) second order analysis.M.36]. Sekulovic et al.

A numerical model that includes both nonlinear connection behavior and geometric nonlinearity of the structure has been developed.28 7. Sekulovic et al. Single-bay two-storey frame.13 The natural frequencies of the frame for the fully rigid and linear semi-rigid (ﬂush end plate and TSDWA) connections have been determined and shown in Table 3. The stiﬀness matrix was based on the analytical solutions of the second order equations. It shows that the eﬀect of hysteretic damping on structural response is signiﬁcant. Conclusion An eﬃcient method to perform dynamic analyses of steel frame structures with ﬂexible connections has been presented in this paper. From the results of numerical examples it can be concluded that the structural responses of the frames with nonlinear connections and the frames with conventional type of connections (rigid or linear) are considerably diﬀerent. connection design and modelling have a great practical importance. . As expected. It is higher in the frame with ﬂexible connections than with rigid joints. 19. (a) Layout and (b) moment–rotation curves of ﬂush end plate connection. The frame analyzed by the present study has smaller amplitudes and nonrecoverable deﬂections.50 7. Table 3 Natural frequencies of the single-bay two-storey frame Type of connection Rigid Flush end plate TSDWA Natural frequencies (rad/s) Present work 10. it is evident that the ﬂexible joint connections greatly inﬂuence the dynamic behavior of steel frames.950 M.30 7. The complex dynamic stiﬀness matrix for a prismatic beam with rotational springs and dashpots attached at its ends was obtained in an explicit form. The time histories of the displacement at node B and the hysteretic loops at node A of the frame under the lateral cyclic loads obtained by Chan and Chui and in the current study are shown in Fig. as well as shorter periods when compared with the same frame analyzed by Chan and Chui. but there are some diﬀerences between their amplitudes and periods. Displacement response of the frame subjected to the impact loads obtained by Chan and Chui and by the present study is shown in Fig. From the results. it can also be concluded that the viscous damping at connections may considerably reduce the displacement response and internal forces of the frame. and thus the eigenfrequencies. Therefore. particularly the lower ones. Therefore. / Computers and Structures 80 (2002) 935–955 Fig. It can be seen that the response curves have the same character. which may have a primary inﬂuence on dynamic response of the structure. These diﬀerences are small at the beginning and they gradually increase with time. so that each beam corresponds to one ﬁnite element. particularly in the case of weak connection types. It can be seen that the response curves are very close. 21. 20. The linear model is inadequate as it cannot represent a hysteretic behavior of connection under cyclic loads. The inﬂuence of the geometric nonlinearity increases with the gravitational loads and the lateral frame deﬂections. The connection ﬂexibility may signiﬁcantly alter both vibration and the response of frames. The connections are vital structural components that are very often responsible for the behavior and safety of frame structures subjected to strong dynamic (seismic) loads. An increase in the connection ﬂexibility reduces the frame stiﬀness. they are larger in the case the presence than in the case the absence of gravitational loads.11 Chan and Ho 10. the nonlinear constitutive model for connections should be used in design and response analysis of real frame structures. 6. It can be seen that they are agree well with the results by Chan and Ho.35 7. On the bases of the above theoretical considerations and the results of the applied numerical analysis.

Dynamic behavior of two-storey frame under cyclic loads. N3 ðxÞ ¼ DÀ1 ½1 À cos x À xn sin x þ sin x sin xn À ð1 À cos xÞ cos xn. for the compressive member ðN < 0Þ. . . .M. (a) Displacement response at node B and (b) hysteretic loops at node A. . 20. l rﬃﬃﬃﬃﬃﬃﬃ jN j : x¼l EI D ¼ 2ð1 À cos xÞ À x sin x. The interpolation functions Ni ðxÞ. N4 ðxÞ ¼ lðDxÞÀ1 ½sin x À x þ xnð1 À cos xÞ À ð1 À cos xÞ sin xn þ ðx À sin xÞ cos xn. N2 ðxÞ ¼ lðDxÞÀ1 ½x cos x À sin x þ xnð1 À cos xÞ þ ð1 À cos x À x sin xÞ sin xn þ ðsin x À x cos xÞ cos xn. Sekulovic et al. 4. are: N1 ðxÞ ¼ DÀ1 ½1 À cos x À x sin x þ xn sin x À sin x sin xn þ ð1 À cos xÞ cos xn. i ¼ 1. Appendix A. where x n¼ . . / Computers and Structures 80 (2002) 935–955 951 Fig.

and using the relations shx ¼ Àj sin jx and chx ¼ cos jx. The elements of the correction matrix G are: g21 ¼ Àg23 ¼ À6 ½g1 þ 2g1 g2 ð2/3 À /4 Þ/2 .952 M. 3 4 EI lki g22 ¼ À6e1 ½g1 þ 2g1 g2 ð2/3 À /4 Þ/2 lD À 4½g1 /3 þ g1 g2 ð4/2 À /2 Þ. Displacement response of two-storey frame under impact loads. 2: gi ¼ . (a) Without gravity and (b) with gravity loads. g24 ¼ lD D ¼ ð1 þ 4g1 /3 Þð1 þ 4g2 /3 Þ À 4g1 g2 /2 . lD À6e2 ½g2 þ 2g1 g2 ð2/3 À /4 Þ/2 À 4½g2 /3 lD þ g1 g2 ð4/2 À /2 Þ. / Computers and Structures 80 (2002) 935–955 Fig. Sekulovic et al. 21. 4 i ¼ 1. lD g41 ¼ Àg43 ¼ À6 ½g2 þ 2g1 g2 ð2/3 À /4 Þ/2 . The shape functions for the tensile member ðN > 0Þ can be obtained from the foregoing expressions replacing x ¼ jx. lD g42 ¼ À6e1 ½g2 þ 2g1 g2 ð2/3 À /4 Þ/2 À 2g2 /4 . 3 4 g44 ¼ À6e2 ½g1 þ 2g1 g2 ð2/3 À /4 Þ/2 À 2g1 /4 .

lDÃ Ã Ã kjk ¼ kkj EI 2 4l ð2/3 c þ /4 f Þ. l3 D giÃ ¼ k1 12 ¼ Àk1 23 ¼ À k1 44 ¼ À pﬃﬃﬃﬃﬃﬃﬃ EI 1 . . . i ¼ 1. lD k1 jk ¼ k1 kj : The elements of matrix k2 according to the second order theory are: k2 11 ¼ Àk2 13 ¼ k2 33 EI a2 b2 . d ¼ g44 . 2 ¼ 6EI Ã ð1 þ 2g1 Þ. lDÃ 4EI . l2 DÃ Ã k24 ¼ 4EI Ã ð1 þ 3g1 Þ. lD2 8EI ðh1 þ h2 þ 3g1 h2 þ 3g2 h1 Þ. l3 D Ã Ã k12 ¼ Àk23 ¼ 6EI Ã ð1 þ 2g2 Þ. ¼ 2 4/3 ðae þ bdÞ þ 2/4 ðad þ beÞ þ þ g1 g2 lD k2 22 k2 44 EI c2 f 2 . f ¼ g42 and e1 ¼ e2 ¼ 0. l3 DÃ Ã Ã k14 ¼ Àk34 m14 ¼ Àm34 12EI 2 3 ¼ 2 3 ðg2 h2 þ 6g2 h2 þ 8g2 h2 1 1 1 l D g1 g2 2 2 2 À 6g1 g2 h1 h2 À 16g1 g2 h1 h2 À 8g1 g2 h1 h2 À 8g1 g2 h1 h2 2 2 2 3 À 24g1 g2 h1 h2 þ 2g1 h2 þ 18g1 h2 þ 52g1 h2 2 2 2 4 þ 48g1 h2 Þ. 4/3 ðce þ df Þ þ 2/4 ðcd þ ef Þ þ þ g1 g2 lD2 c12 ¼ Àc23 ¼ c14 ¼ Àc34 ¼ c22 ¼ 4EI 2 ð4h1 þ h2 þ 24g2 h1 þ 36g2 h1 Þ. lD2 4EI 2 ðh1 þ 4h2 þ 24g1 h2 þ 36g1 h2 Þ. The elements of mass matrix m are: m11 ¼ Àm13 ¼ m33 36EI 2 3 4 ¼ 3 3 ðg2 h2 þ 8g2 h2 þ 20g2 h2 þ 16g2 h2 1 1 1 1 l D g1 g2 2 2 À 4g1 g2 h1 h2 À 8g1 g2 h1 h2 À 8g1 g2 h1 h2 2 2 2 3 4 À 16g1 g2 h1 h2 þ g1 h2 þ 8g1 h2 þ 20g1 h2 þ 16g1 h2 Þ. ¼ 2 4/3 ða2 þ b2 Þ þ 4/4 ab þ þ g1 g2 lD k2 12 ¼ Àk2 23 EI ac bf . lD2 cjk ¼ ckj . l3 D EI ½6l/2 ðd þ eÞ þ 2l2 ð2/3 b þ /4 aÞ. j ¼ À1.M. c ¼ g22 . l2 D2 k1 14 ¼ Àk1 34 ¼ À k1 24 EI ¼ À 3 2l2 ½2/3 ðe þ f Þ þ /43 ðd þ cÞ. . . Sekulovic et al. ¼ 2 4/3 ðc2 þ f 2 Þ þ 4/4 cf þ þ g1 g2 lD EI e2 d 2 . 2 2 2 2 k2 24 ¼ where a ¼ g21 ¼ Àg23 . lDÃ Ã k22 ¼ 4EI Ã ð1 þ 3g2 Þ. l2 DÃ . 2 2 2 12EI Ã Ã ð1 þ g1 þ g2 Þ. i ¼ 1. kiÃ ¼ ki þ jxci . l kiÃ Ã Ã Ã Ã DÃ ¼ 1 þ 4g1 þ 4g2 þ 12g1 g2 : EI 2 4l ð2/3 d þ /4 eÞ. 2. 4 can be found in Ref. l3 D EI ½6l/2 ðc þ f Þ þ 2l2 ð2/3 a þ /4 bÞ. [35]. Analytical expressions for the functions /i . l3 D2 12EI 2 ð2h1 þ h2 þ 2g1 h2 þ 10g2 h1 þ 12g2 h1 Þ. l2 D2 12EI 2 ðh1 þ 2h2 þ 10g1 h2 þ 12g1 h2 þ 4g2 h1 Þ. ¼ 2 4/3 ðac þ bf Þ þ 2/4 ðaf þ bcÞ þ þ g1 g2 lD k2 14 ¼ Àk2 34 EI ae bd . / Computers and Structures 80 (2002) 935–955 953 Ã k44 ¼ The elements of matrix k1 according to the second order theory are: k1 11 ¼ Àk1 13 ¼ k1 33 ¼ À k1 22 ¼ À EI 12l/2 ða þ bÞ. b ¼ g41 ¼ Àg43 . c24 ¼ c44 ¼ where hi ¼ ðci EIÞ=ðlki2 Þ. l3 D The elements of damping matrix c are: c11 ¼ Àc13 ¼ c33 ¼ 36EI 2 2 ðh1 þ h2 þ 4g1 h2 þ 4g1 h2 þ 4g2 h1 þ 4g2 h1 Þ. e ¼ g24 . ¼ 2 4/3 ðd 2 þ e2 Þ þ 4/4 ed þ þ g1 g2 lD EI ce df . The elements of complex dynamic stiﬀness matrix kÃ according to the ﬁrst order analysis are: Ã Ã Ã k11 ¼ Àk13 ¼ k33 ¼ m12 ¼ Àm23 12EI 2 3 4 ¼ 2 3 ð2g2 h2 þ 18g2 h2 þ 52g2 h2 þ 48g2 h2 1 1 1 1 l D g1 g2 2 2 À 6g1 g2 h1 h2 À 8g1 g2 h1 h2 À 16g1 g2 h1 h2 2 2 2 3 À 24g1 g2 h1 h2 þ g1 h2 þ 6g1 h2 þ 8g1 h2 Þ.

J Struct Div. CA. Analysis and behaviour of ﬂexiblyjointed frames.120:1075–87. Eng Struct 1990. [20] Kawashima S. [3] Davisson JB. London: CIRIA. Comput Struct 1984. 1991. Aktan HM. Ho GWM.120(6):1703–17. Chicago: AISC. Struct Eng 1993. Chen WF. Analysis and tests of ﬂexibly connected steel frames.101(EM4): 511–5. J Struct Div. Wu TS. ASCE 1973.58(1):27–41.20:209–34. J Struct Eng. Analysis of three-dimensional frame with ﬂexible beam-column connection. Fujimoto T. J Struct Eng. J Construct Steel Res 1983.95(3): 327–53.113(7):1501–19. Report No. Salatic R. Nethercot DA. Extended end-plate connections under cyclic hading: behavior and design. Chen WF. ASCE 1990.8(2):107–18. [12] Popov EP. J Construct Steel Res 1987. [30] Huang J. [16] Krawinkler H. Nonlinear analysis of frames with ﬂexible connections. Comput Struct 2001. [2] Nethercot DA. ASCE 1994. 1985. [21] Suarez LE. A pointwise equilibrating polynomial (PEP) element for nonlinear analysis frames. Berkeley. [32] Sekulovic M. Matheu EE. Zhau ZH. Int J Solids Struct 1991. Cyclic yield reversal in steel building connections. Analysis of ﬂexibly connected frames under nonproportional loading. [7] Chan SL. Olowokere OD. Popov EP. Goto Y. Mandic R. [24] Shi G.89(ST6):13– 42. ASCE 1963. Analysis of critical behaviour of semi-rigid frames with or without load history in connections. ASCE 1975. [31] Sekulovic M.79(11):1097– 107. Static and dynamic analysis of space frames with non-linear ﬂexible connections.30(3):90–107. Kirby PA. [18] Korol RM.116(11): 2917–30.. Cyclic loading of steel beams and connections.954 M. 273. J Eng Mech Div. J Struct Div. Salatic R. Beam-column sub assemblages under repeated loading. AISC 1993. Cyclic behaviour of end-plate moment connections. Chen WF. Kirby PA. [19] Lionberger SR. ASCE 1994. Morris GA.16:253–80. Bertero VV. J Construct Steel Res 1990. Eng Struct 1988. [13] Vedero VT. Seismic response of structural frameworks with ﬂexible connections. [22] Sivakumaran KS. [17] Stelmack TW. Seismic response of low-rise steel frames. Proc of Second Int Workshop on connections in steel structures. [27] Kishi N. Bertero VV. J Struct Eng. 2 2 2 m24 ¼ m44 ¼ 4EI 2 2 ðg2 h2 þ 4g2 h2 À 8g1 g2 h1 h2 À 24g1 g2 h1 h2 1 1 lD3 g1 g2 2 3 4 þ 4g1 h2 þ 40g1 h2 þ 132g1 h2 þ 144g1 h2 Þ 2 2 2 2 mjk ¼ mkj : References [1] Jones SW. Moment–rotation relations of semirigid connections with angles. In: Bjorhovode R et al. J Struct Div. Seismic analysis of frames with semi-rigid eccentric connections. ASCE 1989. Int J Numer Meth Eng 1989. [23] Yousef-Agha W. . ASCE 1987. ASCE 1986. The analysis of frames with semi-rigid connections—a state-of-the-art report. [4] Moree DB.112(7):1573–88. J Eng Mech Div. Popov EP. Suzuki S. Pinkey RB. Gerstle KH. Theor Appl Mech 1994. Rotational stiﬀness characteristics of steel beam-to-column connections. Atluri SN.11:241–54.71:428–35. Nonlinear vibration analysis of steel frames with semi-rigid connections. [5] Monforton GR. Suzuki S. Design aid of semi-rigid connections for frame analysis. Eng Struct 1986. Matrix analysis of semi-rigidly connected frames. Sekulovic et al. Malcevic I. editors.115(3):594–607.116(ST7):1813–34. Seismic response of multistorey steel buildings with ﬂexible connections. Testing steel frames at full scale: appraisal of results and implications for design.12(2):88–97. UBC/EERC-71/07 Earthquake Engineering Research Center. Eng J. Dynamic response of frames with non-rigid connections. paper no.19:85– 92.27(4):467– 83. 1971.3(2):2–13. Chen WF. Can J Civil Eng 1984. / Computers and Structures 80 (2002) 935–955 m22 ¼ 4EI 2 3 4 ð4g2 h2 þ 40g2 h2 þ 132g2 h2 þ 144g2 h2 1 1 1 1 lD3 g1 g2 2 2 À 8g1 g2 h1 h2 À 24g1 g2 h1 h2 þ g1 h2 þ 4g1 h2 Þ 2 2 8EI 2 3 ðg2 h2 þ 7g2 h2 þ 12g2 h2 À 5g1 g2 h1 h2 1 1 1 lD3 g1 g2 2 2 2 2 À 12g1 g2 h1 h2 À 12g1 g2 h1 h2 À 36g1 g2 h1 h2 2 3 þ g1 h2 þ 7g1 h2 þ 12g1 h2 Þ. Comput Struct 1996. Nethercot DA. Weaver JW. ASCE 1972.28:2635–50. [14] Tsai KC. Ghobarah A. Second-order analysis of frame structures. Singh MP. Chen WF.8:17–54. [33] Goto Y. J Struct Div. [25] Chan SL. Stability behaviour of semirigid sway frames. [26] Sekulovic M. Versatile elastic-plastic stressstrain formula.99(6):1189– 204. J Struct Div. [29] Goto Y. Vibration analysis of frames with semi-rigid connections. Nethercot DA. Chen WF.95(EM1):95–114. Pitsburg. [9] Lui EM. ASCE 1969.15(3):209–19. [6] Ang KM. [10] Kishi N. Kirby PA. Inelastic behavior of steel beam-to-column subassemblages. J Struct Eng. [8] Richard RM. Second-order elastic analysis for frame design. Marley MJ. Osman A. New Zealand 2000. ASCE 1969. Steel beam-to-column connections—a review of test data. Twelfth Conference on Earthquake Engineering. A design model for semi-rigid connections. [15] Popov EP. [11] Wu FS. University of California. Chen WF.98(5): 1137–59. Popov EP. ASCE 1990.10: 239–48. Morris G. Eng Struct 1993. Abbott BJ. Matsuoka KG. [28] Goto Y. J Struct Div.

955 [37] SAP 2000. Inc. J Struct Eng. Chui PPT. . O’Kelly MEJ.24(9):893–913. ASME 1965. ASCE 1989. Integrated ﬁnite element analysis and design of structures. Stability design of semirigid frames. Int J Solids Struct 1988. J Appl Mech.M. 2000. Kishi NN. 1996. / Computers and Structures 80 (2002) 935–955 [34] Caughey TH. Liew JYR. 1996. [39] Lui EM. Nonlinear static and cyclic analysis of steel frames with semi-rigid connections. Goto Y.32:583–8. Chen WF. [36] Chan SL. Behavior of braced and unbraced semi-rigid frames.. Amsterdam: Elsevier. Berkeley. Classical normal modes in damped linear dynamic systems. New York: John Willey & Sons. California: Computers and Structures. Semirigid steel beam-to-column connections data base and modeling. [35] Chen WF. [38] Chen WF. Sekulovic et al.120(6):1703–17.