Seismic Collapse Analysis of Reinforced Concrete Framed Structures Using The Nite Element Method

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2003; 32:2027–2046 (DOI: 10.1002/eqe.313)
Seismic collapse analysis of reinforced concrete framed

structures using the nite element method
Daigoro Isobe1; ∗; † and Michihiro Tsuda2

1 Institute of Engineering Mechanics and Systems; University of Tsukuba; 1-1-1 Tennodai Tsukuba-shi;
Ibaraki 305-8573; Japan
2 Hitachi Software Engineering Co. Ltd; Kanagawa; Japan
SUMMARY
A new nite element code using the Adaptively Shifted Integration (ASI) technique with a linear
Timoshenko beam element is applied to the seismic collapse analysis of reinforced concrete (RC)
framed structures. This technique can express member fracture as a plastic hinge located at either
end of an element with simultaneous release of the resultant forces in the element. Contact between
members is also considered in order to obtain results that agree more closely with actual behavior,
such as intermediate-layer failure. By using the proposed code, suciently reliable solutions have been
obtained, and the results reveal that this code can be used in the numerical estimation of the seismic
design of RC framed structures. Copyright ? 2003 John Wiley & Sons, Ltd.
KEY WORDS: adaptively shifted integration technique; seismic collapse analysis; reinforced concrete
structures; member fracture; nite element method
1. INTRODUCTION
In the conventional design of a building, in order to minimize calculation costs, only static
analysis in the horizontal and uniaxial directions is commonly carried out. This approach can
ensure the structural strength of the building if there is sucient strength to support the load in
the vertical direction. For a similar reason, the mass system model replaces the building layer
in dynamic analysis, and the complicated dynamic behavior of the structure at the member
level becomes dicult to examine suciently. Therefore, the development of a more precise
and more ecient dynamic analysis code is strongly desired.
Recently, signicant advances in the eld of computers have been eliminating the calcula-
tion cost restrictions, and various dynamic analysis codes are being developed. Among those
codes, there are some codes applicable to dynamic collapse problems which contain strong
non-linearities and discontinuities, such as the Distinct Element Method (DEM) [1] or the
∗ Correspondence to: Daigoro Isobe, Institute of Engineering Mechanics and Systems, University of Tsukuba, 1-1-1
Tennodai Tsukuba-shi, Ibaraki 305-8573, Japan.
† E-mail: isobe@kz.tsukuba.ac.jp
Received 9 August 2002

Revised 15 November 2002 and 24 January 2003
Copyright ? 2003 John Wiley & Sons, Ltd. Accepted 24 February 2003
2028 D. ISOBE AND M. TSUDA
Discontinuous Deformation Analysis (DDA) [2]. These codes have been applied to demo-
lition analyses and seismic collapse analyses [3–8]. The Applied Element Method (AEM),
which can predict crack initiation and propagation in the material [9], and can also follow the
total failure behavior from zero loading to complete collapse in a reasonable CPU time [10],
has also been developed and applied to detailed non-linear analysis of reinforced concrete
structures. In contrast, the widely used nite element codes can only be used after making
complicated modications to simulate fracture occurring in exural damage or shear damage
of reinforced concrete members. The main purpose of this study is to devise a new algorithm
applicable to these types of discontinuous problems by using the nite element method.
As shown in References [11–13], the highest computational eciency in nite element
analyses of framed structures including the geometrically non-linear, elasto-plastic analysis as
well as the geometrically linear, plastic collapse analysis and the dynamic collapse analy-
sis, has been achieved by using the Adaptively Shifted Integration (ASI) technique. In this
work, the ASI technique and the linear Timoshenko beam element, which can be easily im-
plemented into existing nite element codes, are extended and applied to seismic collapse
analyses.
In this technique, the numerical integration points in an elastically deformed beam element
are placed at the optimal points for linear analysis (midpoint in the linear Timoshenko beam
and Gaussian integration points in the cubic beam element). They are immediately shifted
after the occurrence of a fully plastic section in the element, in order to form a plastic hinge
exactly at the position of the fully plastic section. This shift is carried out according to
the previously established relations between the locations of numerical integration points and
those of plastic hinges [14]. In this manner, this technique results in a higher computational
accuracy with fewer elements than conventional nite element methods. A member fracture is
represented in this paper as the release of the resultant forces in an element immediately after
a plastic hinge is located at the section. With the proposed technique, problems associated with
modeling structural discontinuities, such as those mentioned above, can be easily tackled even
by conventional displacement-based nite element codes. More details about this technique
are given in subsequent sections of this paper.
Typically, the Total Lagrangian Formulation (hereafter abbreviated as TLF) and the Updated
Lagrangian Formulation (ULF) [15] are used in incremental non-linear structural analyses. As
the occurrence of extremely large rotations and strains is anticipated in dynamic collapse
analyses, the ULF is used in this paper. In order to maintain a high computational eciency,
an implicit time integration scheme using Newmark’s method is chosen for the analyses,
since this scheme is better suited for modeling low-frequency response. Also, a distributed
mass matrix is used in the implicit code to reduce response errors due to a rough nite
element subdivision.
In this study, the ASI technique is implemented into the nite element code in order to
develop a more precise and less calculation-time-consuming seismic collapse analytical tool.
The purpose of this study is to verify the validity of the ASI technique in seismic collapse
analysis and to construct a highly ecient structural design tool for reinforced concrete (RC)
structures. Simple numerical tests showing the validity of the modeling of an RC beam,
along with the validity of the implemented scheme, are carried out. Also, examples involving
structural discontinuities such as a member fracture are adopted to show the expandability
of the scheme. A contact algorithm is added to the code to reproduce phenomena such as
intermediate-layer failure.
Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:2027–2046
SEISMIC COLLAPSE ANALYSIS OF RC FRAMED STRUCTURES 2029
In Section 2, the implicit non-linear code based on the ULF is explained. In Section 3, the
general concept of the ASI technique is shown. Some simple numerical tests for RC columns
are shown in Section 4, along with the strength characteristics used in the analyses. The
results of applying this technique to seismic collapse analysis of an eight-story three-span RC
building are detailed in Section 5. Section 6 contains concluding remarks.
2. TIME INTEGRATION SCHEME BASED ON THE ULF
In this section, an implicit time integration scheme based on the ULF applied in seismic
collapse analyses is described.
Relations between generalized strain increments and nodal displacement increments
The relations between the generalized strain increments n ”1 –n ”6 and the nodal displacement
increments of a linear Timoshenko beam element at incremental step n, are expressed as
   

 
 (n xJ − n xI )= n l 



n 1
 
  


 
 
 − n 


 n
2 
 
 ( n yJ n yI )= l 


 
 
 

 n 3   (n w J − n w I )= ln 
= (1)

 n 4  
 (n zJ − n zI )= n l 


 
  

    


 n 5 



 (n u J − n u I )= n l − n yI · (1 − s)=2 − n yJ · (1 + s)=2 



 
   

n 6  
(n vJ − n vI )= l + n xI · (1 − s)=2 + n xJ · (1 + s)=2
n
where uI –zI are the six components of the nodal displacement at the rst node, and
uJ –zJ are the six components of the nodal displacement at the second node, respectively.
The term s (−16s61) denotes the location of the numerical integration point. Here, it should
be noted that the nodal displacement increments are based on the elemental coordinate at
incremental step n. The element length n l calculated at incremental step n is used to evaluate
the generalized strain increments; these strain increments are equal to the updated Green strain
increments.
The generalized strain increment vector and the nodal displacement increment vector are
expressed as
{n } = [nn B L ] · {n u } (2)

{n u } = [u T ] · [0 T ] · {u} (3)
where
{u}T = uI ; vI ; wI ; xI ; yI ; zI ; uJ ; vJ ; wJ ; xJ ; yJ ; zJ (4)
[nn B L ] denotes the displacement–strain matrix at incremental step n. [0 T ] and [u T ] in Equation
(3) are the transformation matrix from global coordinates to the initial elemental coordinates,
and the transformation matrix from the initial elemental coordinates to elemental coordinates
at step n, respectively. In this study, [u T ] is calculated by successive iteration and is expressed

as
[u T ] = [n T ] · [n−1 T ] · [n−2 T ] · · · · · · [3 T ] · [2 T ] · [1 T ] (5)
where [n T ] is the transformation matrix from elemental coordinates at step (n − 1) to elemental
coordinates at step n. The matrix [n T ] is calculated as
n ∗ 
T 0 0 0
n   0
n ∗
T 0 0 
T = n ∗  (6)
 0 0 T 0 
0 0 0 nT ∗
where
[n T ∗ ] = [n T ] · [n T ] · [n T ]
 
cos(n ) sin(n ) 0
 
= − sin( ) cos( ) 0 
n n 
0 0 1
 
cos(n ) 0 − sin(n )
 
× 0 1 0 
n n
sin( ) 0 cos( )
 
1 0 0
 
×  0 cos(n ) − sin(n )  (7)
0 sin(n ) cos(n )
By dening the nodal displacement increments between steps (n − 1) and n as n−1 u I , n−1 vI ,
n−1 w I ; : : : : : : ; n−1 xJ , n−1 yJ and n−1 zJ , cos(n ), cos(n ) and n in Equation (7) can
be calculated as
cos(n ) = {n−1 l + (n−1 w J − n−1 w I )}=

1
[{n−1 l + (n−1 w J − n−1 w I )}2 + (n−1 vJ − n−1 vI )2 ] 2 (8)
cos(n ) = {n−1 l + (n−1 w J − n−1 w I )}=
1
[{n−1 l + (n−1 w J − n−1 w I )}2 + (n−1 u J − n−1 u I )2 ] 2 (9)
n
= (n−1 zI + n−1 zJ )=2 (10)
where n−1
l is the element length at step (n − 1).
Relations between resultant force increments and generalized strain increments
The relation between the resultant force increment vector and the generalized strain increment
vector of a linear Timoshenko beam element is expressed as
{n R } = [D] · {n } (11)
where [D] is the stress–strain matrix. For an elastic element, [D] is given as
[D] = [De ]
 
EIx 0 0 0 0 0
 0 0 
 EIy 0 0 0 
 
 0 0 EA 0 0 0 
= 0
 (12)
 0 0 GK 0 0  
 
 0 0 0 0 x GA 0 
0 0 0 0 0 y GA
where E, G, A, K, Ix , Iy , x and y are Young’s modulus, shear modulus, cross-sectional area,

Saint-Venant’s torsional coecient, the moments of area inertia around the x and y axes, and
shear correction factors in the x and y axes, respectively. When the element becomes plastic
the stress–strain matrix [D] can be expressed as
[D] = [Dp ]
[De ] · {@f=@R} · @f=@R · [De ]
= [De ] − (13)
@f=@R · [De ] · {@f=@R }
where f is the plastic potential given as

2 2 2 2
R1 R2 R3 R4
f= + + + −1 (14)
Mx0 My0 N0 Mz0
where R1 , R2 , R3 and R4 are associated with the two components of bending moment, the
axial force and the torsional moment, respectively. The eect of shear forces on the yield
condition has been neglected in this study. The subscript ‘0’ indicates a fully plastic value
under the condition that each component of the resultant forces acts independently on the
cross-section of the member.
The strain vector and the stress vector at step (n + 1) can be obtained by transforming the
updated Kirchho stress increment vector to the Jaumann dierential form vector, which is
expressed as
{n+1
n } = {n } + {n }
n
(15)
{R J } = [n+1 A]{n R } (16)
{n+1
n R} = {n R} + {R }
n J (17)
The transformation matrix [n+1 A] is expressed as
 n+1 ∗ n+1 ∗ n+1 ∗ 

T11 T12 0 T13 0 0
 n+1 ∗ ∗ ∗ 
 T21 n+1
T22 0 n+1
T23 0 0 
 
 ∗ ∗ n+1 ∗ 
 0 0 n+1
T33 0 n+1
T31 T32 
[n+1 A] = 
 n+1 ∗ n+1 ∗ n+1 ∗

 (18)
 T31 T32 0 T33 0 0 
 
 0 n+1 ∗ n+1 ∗ n+1 ∗ 
 0 T13 0 T11 T12 
n+1 ∗ n+1 ∗ n+1 ∗
0 0 T23 0 T21 T22
where n+1
Tij∗ is the (i; j) term of the matrix [n+1 T ∗ ] in Equation (7).
Implicit time integration scheme using Newmark’s method
Since an explicit code uses a lumped mass matrix, which is simplied by lumping each el-
ement’s mass at each node of an element, the numerical error tends to increase with time
in the response analyses. Accordingly, in the ASI technique, which requires less element
subdivision, the numerical error caused by the use of lumped mass matrices is increased and
becomes apparent. Thus a distributed mass matrix should be used when numerical models
are controlled by low-frequency response, in cases such as seismic collapse analysis. In this
section, an implicit code implemented with a distributed mass matrix is
shown.
The dynamic equilibrium equation at t = tn can be formulated as
[M ]{n u } = {n E } − {nn F } (19)
where [M ], {n u }, {n E } and {nn F } are the mass matrix, acceleration vector at step n, nodal
external force vector at step n and internal force vector at step n, respectively.
The following equation is substituted into Equation (19) at t = tn+1 in the implicit code:
{n+1
n+1 F } = {n F } + [n K]{u}
n n
(20)
Then the following incremental stiness equation is evaluated:
[M ]{n+1 u } + [nn K]
{u} = {n+1 E } − {nn F } (21)
By neglecting residual forces, an implicit code is obtained by evaluating the following incre-
mental kinematic equation:
[M ]{u } + [nn K]
{u} = 0 (22)
is a stiness matrix at step n. The distributed mass matrix [M ] for the linear
where [nn K]
Timoshenko beam element used in the implicit code is given by
 
2A A
 
 2A A 0 
 
 2A A 
 
 
 2Ix Ix 
 
 2I 0 I 
 y y


l  Iz 
2Iz 
[M ] =   (23)
6  A 2A 
 
 
 A 0 2A 
 
 A 2A 
 
 
 Ix 2I x 
 
 0 Iy 2Iy 
Iz 2Iz
where , A, l, Ix , Iy and Iz are the density of the member, cross-sectional area, length of the
element, moments of area inertia around the x and y axes, and polar moment of area inertia,
respectively.
The incremental kinematic equation for a structure under excitation at xed points, which
is used in this paper, is as follows:
[M1 ]{u } + [M2 ]{u b } + [nn K 1 ]{u} + [nn K 2 ]{ub } = 0 (24)
The subscript ‘1’ indicates the coupled terms between free nodal points, subscript ‘2’ indicates
the coupled terms between free nodal and xed nodal points, and subscript ‘b’ indicates the
components at xed nodal points. Vectors {u } and {u} are the nodal acceleration increment
and the nodal displacement increment, respectively.
Under the assumption that the displacements at free nodal points are estimated by adding
quasi-static displacement increments {us } and dynamic displacement increments {ud }, the
displacements at free nodal points are given as
{u} = {us } + {ud } (25)
{us } is evaluated, by neglecting inertia force, as follows:
{us } = − [nn K 1 ]−1 [nn K 2 ]{u b } (26)
Substituting Equations (25) and (26) into Equation (24), the following equation is obtained:
[M1 ]{u d } + [nn K 1 ]{ud } = ([M1 ][nn K 1 ]−1 [nn K 2 ] − [M2 ]){u b } (27)
In this scheme, equivalent forces are calculated by substituting nodal acceleration increments at
xed points into the right side of the above equation, and the incremental kinematic equation
is solved by Newmark’s method. Damping matrices are not considered in this study.
3. ADAPTIVELY SHIFTED INTEGRATION TECHNIQUE
The relation between the location of a numerical integration point and that of the occurrence
of a plastic hinge in the linear Timoshenko beam element is obtained by considering the
equivalence conditions between the strain energy approximations of a linear Timoshenko beam
element and a physical model known as the rigid-bodies spring model (RBSM). Referring to
Figure 1, the relation is expressed by the following equation [14]:
s1 = − r1 (28)
where s1 (−16s1 61) and r1 (−16r1 61) are the position of the numerical integration point
and the position of the plastic hinge or member fracture, respectively.
When the entire element behaves elastically, the midpoint of the element (s1 = 0) is the
most appropriate integration point from considerations of accuracy and symmetry. The internal
force vector at step n based on the ULF, is expressed as

{nn F } = [0 T ]T · [u T ]T · [nn B L (0)]T · {nn R(0)
}dl (29)
nl
where the value in parentheses in the displacement–strain matrix [nn B L ] indicates the location
of the integration point, and that in the resultant force vector {nn R} indicates the point at which
stresses are evaluated, respectively. [0 T ] and [u T ] are the transformation matrices based upon
the ULF.
Using elementary beam theory, relations between bending moments R1 and R2 and shear
forces R5 and R6 can be expressed as
d R2 d R1
R5 = − ; R6 = − (30)
dz dz
Thus, the distributions of bending moment increments n R1 (s) and n R2 (s) along the element
length can be approximated by the following equations using the bending moment increments
n R1 (0) and n R2 (0) and the shear force increments n R5 (0) and n R6 (0) at the midpoint
Figure 1. Linear Timoshenko beam element and its physical equivalent.
of the element:
n R6 (0)n ls
n R1 (s) = n R1 (0) − (31a)
2
n R5 (0)n ls
n R2 (s) = n R2 (0) − (31b)
2
where n l is the element length at t = tn . Equations (31) show that bending moments are
subject to a linear change in an element and are likely to take the maximum value at either
end (s = ± 1). As other resultant forces have constant values in the element, a fully plastic
state can be determined with a yield function by comparing the distributions calculated from
Equations (31).
In dynamic collapse analyses using the ASI technique, member fracture is expressed by
shifting the numerical integration point according to Equation (28) immediately after the
occurrence of a fractured section on either end of the element, and simultaneously reducing
the resultant forces of the element. For instance, if a fully plastic section or a fractured section
rst occurs at the left end of an element (r = − 1), the numerical integration point is shifted
immediately to the right end of the element (s = 1) according to Equation (28). The released
force vector which operates on the element at the next step in the analysis is then expressed by

{nn F } = [0 T ]T · [u T ]T · [nn B L (1)]T · {nn R(
−1)}dl (32)
nl
Similarly, if a fully plastic section or a fractured section rst occurs at the right end of the ele-
ment (r = 1), the numerical integration point is shifted to the left end of the element (s = −1).
In the case of the implicit scheme, the incremental stiness matrices used in the algorithm,
when the entire element is elastic, are

[nn K L ] = [u T ]T · [0 T ]T [nn B L (0)]T [De (0)][nn B L (0)][0 T ] · [u T ]dl (33a)
nl

[nn K NL ] =
[u T ]T · [0 T ]T [nn G(0)]T n
[n S(0)][nn G(0)][0
T ] · [u T ]dl (33b)
nl
and [nn S]
where [De ], [nn G] are the elastic stress–strain, initial displacement and initial stress
matrices, respectively. In the case that the section is fully plastic or fractures rst at the left
end of the element (r = −1), the incremental stiness matrices are given by

[nn K L ] = [u T ]T · [0 T ]T [nn B L (1)]T [Dp (−1)][nn B L (1)][0 T ] · [u T ]dl (34a)
nl

[nn K NL ] =
[u T ]T · [0 T ]T [nn G(1)]T n
[n S(−1)][nn G(1)][0
T ] · [u T ]dl (34b)
nl
where [Dp ] is the plastic stress–strain matrix. It should be noted that when new hinges are
formed, the resultant force increments calculated at the new integration point are automatically
added to those originally existing at that point. As a result of using this procedure, a ‘non-
smoothness’ type of phenomenon does not appear in the calculation. More details of the
implicit ASI algorithm are explained in the author’s previous papers [11–13].
4. STATIC AND QUASI-STATIC ANALYSES OF RC COLUMNS
In this section, the validity of the algorithm when implementing an RC member model, is
conrmed by carrying out some analyses on simply supported columns subjected to static
and quasi-static loads. The crack strength and yield strength formulas [16] used for the RC
member are as follows.
· Flexural crack strength

NZe
Mc = 1:8 Fc · Ze + (35)
Ac
· Flexural yield strength
column
My = 0:5(ag y + N )g1 D (when Nmin 6N ¡0) (36a)

N
My = 0:5ag y g1 D + 0:5ND 1 − (when 06N 6Nb ) (36b)
bDFc
My = {0:5ag y g1 D + 0:24(1 + g1 )(3:6 − g1 )bD2 Fc }
× (Nmax − N )=(Nmax − Nb ) (when Nb ¡N 6Nmax ) (36c)
Nmin = −ag y (37a)

Nmax = bDFc + ag y (37b)
Nb = 0:22(1 + g1 )bDFc (37c)
beam
My = 0:9at y d (38)
· Shear crack strength

N 0:085
Qc = 1 + kc (500 + Fc ) bj (39)
150bD M=Qd + 1:7
· Shear ultimate strength

0:115ku · kp (180 + Fc ) N
Qy = + 2:7 pw ·s fwy + 0:1 bj (40)
M=Qd + 0:115 bD
Figure 2. Simply supported column subjected to shear force.
· Stiness reduction ratio at yield point

2
M d
y = (0:043 + 1:65n · pt + 0:043 + 0:330 ) (41)
Qd D
here, b is the column width, D is the column height, h0 is the inner measured height of the
column, d is the equivalent height = 0:9D, Ac is the cross-sectional area of the RC member = bD,
Fc is the compressive strength of concrete, s fy is the yield stress of tension reinforcement,
s fwy is the tension yield stress of shear reinforcement, Ze is the section modulus considering
2
reinforcement = 1:1Z = 1:1 bD6 , n is the ratio of Young’s moduli of concrete and reinforcement,
h0
N is the normal force, M=Qd is the shear span ratio = 2d , 0 is the normal force ratio = ANc Fc , j
is the distance between the center of stresses = 78 d, kc is the correction factor of the member
height = 0:7, ku is the correction factor of member dimension = 0:7, kp is the correction factor
at
of tension reinforcement ratio = 0:82pt0:23 , pt is the tension reinforcement ratio = bD , pw is the
at
shear reinforcement ratio = b , at is the cross-sectional area of tension reinforcement, aw is
the cross-sectional area of a pair of shear reinforcements and is the distance between each
shear reinforcement.
A simple numerical test is carried out for a simply supported column subjected to shear
force, as shown in Figure 2. Figure 3 shows the load–displacement curves obtained by the
conventional code and by the ASI technique. In this case, a trilinear-type model [17] is
used. The results show that the converged solution can be obtained with only two-element
(a) (b)
Figure 3. Static analysis of a simply supported column: (a) conventional code; (b) ASI technique.
Figure 4. Degrading trilinear model for reinforced concrete.
subdivision per member in the ASI technique, while more elements are required in order
to obtain the converged solution in the conventional nite element code. The CPU cost of
two-element subdivision per member in the ASI technique is about 1% of that for 32-element
subdivision per member in the conventional code, while the error against the exact solution
is 0.5% for the former case and 1.1% for the latter. This convergence of the ASI technique
is due to the instant shifting of the numerical integration point to the adequate position for
each member. It is conrmed that highly accurate solutions can be obtained in a minimum
calculation time. Also, the validity of the trilinear-type model used in the analysis is conrmed
by comparing the result with that of the experiment [18].
A Degrading trilinear (D-tri) model [17], as shown in Figure 4, is used to explain the
hysterisis loop of the bending stiness of the beam. The D-tri model is widely used for
the simplicity of its algorithm and its accuracy against the bending behavior of RC beams.
Formulas for shear crack strength and shear ultimate strength are only used for static analyses.
The RC beams are assumed to be of exural damage type, and shear and axial stinesses are
xed to be elastic in quasi-static and dynamic analyses. A change in the axial force inside
the beam is considered numerically in this study.
The D-tri model is applied to RC beams subjected to cyclic loads. Figure 5 shows load–
displacement curves obtained from the experiment on an RC column under repeated quasi-
static load [18], along with a numerical result obtained by the ASI technique with two-element
subdivision per member, which was also a well-converged solution compared to other results
subdivided into more elements. The authors judged the numerical model to be sucient for
expressing the characteristic of the RC beam when applied in such cases as above, under
seismic excitations.
5. SEISMIC COLLAPSE ANALYSIS CONSIDERING MEMBER FRACTURE

AND CONTACT
Phenomena with strong non-linearity and discontinuities, such as member fracture, are easily
analyzable using the proposed code. However, behavior such as the penetration of a member
through a oor could be observed in the analyses [13], since contact between members was
not considered. Actually, some structures in the Great Hanshin-Awaji earthquake collapsed at
intermediate oors and as a result the upper oors piled upon the lower oors. Thus, a contact
algorithm is added to the proposed nite element code to follow the intermediate-layer failure
phenomena as well as the elasto-plastic behavior.
Member fracture is assumed to occur when at least one of the following conditions has
been satised:

dfx dfy
− 1¿0; − 1¿ 0 (42a)
dfxexp dfyexp

x y
− 1¿0; − 1¿0 (42b)
xexp yexp
where dfxexp and dfyexp are the critical ductility factors for the x- and y-axes, and xexp and
yexp are the critical shearing strain for the x- and y-axes, respectively.
Figure 5. RC column under repeated quasi-static load: (a) experiment (Kanakubo et al.
[18]); (b) ASI technique (2 elem:=memb:).
Figure 6. Member fracture expressed by using ASI technique:

(a) Linear Timoshenko beam elements (b) RBSM.
When member fracture occurs according to the conditions given above, it is expressed by
shifting the numerical integration point of the fractured element with simultaneous release of
the resultant forces in the element, as shown in Figure 6. The distribution of element mass
at each node is controlled to split into half of the initial mass. It is also to be noted that as
elements and nodes are still treated as a continuous model after the total separation, new virtual
nodes for the fractured sections must be established at the post-processing stage. The elements
with virtual nodes, with stiness reduced to zero, are then visualized as rigid bars thereafter.
Contact of fractured elements is judged using post-processing data, which includes virtual
nodes for the fractured elements. Figure 7 shows the relation of coordinates between a frac-
tured element and other elements. A1 and A2 denote the two nodes of a fractured element Ef .
Likewise, Bi1 and Bi2 denote two nodes of other elements Ei (i = 1; 2; 3; : : : ; n). Using the coor-
dinates of each node, A1 (xf1 ; yf1 ; zf1 ), A2 (xf2 ; yf2 ; zf2 ) and Bi1 (xi1 ; yi1 ; zi1 ), Bi2 (xi2 ; yi2 ; zi2 ), the
condition of the four nodes of the two elements existing on a same plane can be dened as
f(x; y; z) ≡ {(yi1 − yf2 )(zi2 − zf2 ) − (yi2 − yf2 )(zi1 − zf2 )}(xf1 − xf2 )
+ {(xi2 − xf2 )(zi1 − zf2 ) − (xi1 − xf2 )(zi2 − zf2 )}(yf1 − yf2 )
+ {(xi1 − xf2 )(yi2 − yf2 ) − (xi2 − xf2 )(yi1 − yf2 )}(zf1 − zf2 ) = 0 (43)
Figure 7. Relation of coordinates between fractured element and other elements.
Also, the condition of elements existing at a specic distance from a fractured element is
assumed to be
|A1 Bi1 | + |A1 Bi2 | + |A2 Bi1 | + |A2 Bi2 |6Cl (Lf + Li ) (44)
where |A1 Bi1 |, |A1 Bi2 |, |A2 Bi1 | and |A2 Bi2 | are the distances between nodes, and Lf and
Li (i = 1; 2; 3; : : : ; n) are the lengths of fractured element Ef and other elements Ei (i = 1; 2; 3;
: : : ; n), respectively. Cl is a coecient related to contact length, and the value of 1.8 is used.
On the other hand, another condition is given to the elements that exist on the same plane
from the initial stage of the analysis. The condition of elements existing at a specic distance
from a fractured element on the same plane is given by
|A1 Bi1 | + |A1 Bi2 | 6 Li (45a)
|A2 Bi1 | + |A2 Bi2 | 6 Li (45b)
In this algorithm, the contact loop is automatically avoided if the distance between nodes is not
decreasing. For elements that exist on the same plane from the initial stage (Equation (43)),
they are judged to be in contact if Equations (45) are satised. For other elements which do
not exist on the same plane, and if only the distance between nodes satises Equation (44),
the elements are judged to be in contact according to the condition below, which refers to
the four nodes nearly forming a plane.
f(x; y; z)6Cf (46)
Cf is a coecient expressing the degree of planarity, which is xed by considering the sec-
tional properties of constituent members. Values of 2:0 × 102 m3 for columns and 5:0 × 104 m3
for beams are used in this paper.
Consequently, the contact between members is judged on the basis of two sets of conditions:
Equations (43), (45) and Equations (44), (46). Once two elements are judged to be in contact,
a total of four gap elements are xed between the nodes, as shown in Figure 8. Material
properties for the gap elements are assumed to be the same as those of other elements. In
the case of contact, impact force is delivered through those gap elements to the connecting
elements. The oor slabs are not modeled in the numerical examples, however, the contact
Figure 8. Binding condition of fractured element and gap elements.
Table I. Member list (column).
Floor Fc (N=mm2 ) Sectional size (mm) Bar arrangement Bar arrangement

(Outer columns) (Inner columns)
8 20.6 650 × 650 8-D25 8-D25
2-D10@100 2-D10@100
7 20.6 650 × 650 8-D25 8-D25
2-D10@100 2-D10@100
6 20.6 650 × 650 8-D25 8-D29
2-D10@100 2-D10@100
5 20.6 650 × 650 8-D25 8-D29
2-D10@100 2-D10@100
4 23.5 700 × 700 8-D29 8-D32
2-D10@100 2-D10@100
3 23.5 700 × 700 8-D29 8-D32
2-D10@100 2-D10@100
2 23.5 700 × 700 8-D29 8-D32
2-D10@100 2-D10@100
1 23.5 750 × 750 16-D29 16-D32
2-D10@100 2-D10@100
between members and the oor is considered by selecting proper coecients in the conditions
given above.
Seismic collapse analyses are carried out on an eight-story three-span RC building [19]
with individual span length of 6:0 m, and the height of each story is 3:5 m. Member lists
are shown in Tables I and II. The main reinforcement quantity of the columns on the 4th
oor is articially set to 1=10 that in Table I in order to induce member fracture. Values of
5.0 (column) and 20.0 (beam) for the critical ductility factors, and 4:0 × 10−3 for the critical
shearing strain are used in these analyses. These values were estimated from the experiments
on RC beams [16–18]. The total number of elements is 464, and that of nodes is 340.
Incremental time is 5 ms, and the total number of steps is 1000. Nearly 20 times the dead
load is initially applied to the structure to imitate actual conditions. The original El Centro
seismic wave (NS and EW components) is given precisely at the xed points on the ground
oor, according to Equations (24) to (27).
Figure 9(a) shows the case in which only member fracture is considered, and Figure 9(b)
shows the result obtained after the contact algorithm was added to the code. Failure initiates
Table II. Member list (beam).
Floor Fc (N=mm2 ) Sectional size (mm) Bar arrangement

R 20.6 600 × 800 4-D22
2-D10@200
8 20.6 600 × 800 4-D22
2-D10@200
7 20.6 600 × 800 4-D22
2-D10@200
6 20.6 600 × 800 4-D25
2-D10@200
5 23.5 600 × 800 4-D25
2-D10@200
4 23.5 600 × 800 4-D25
2-D10@200
3 23.5 600 × 800 4-D29
2-D10@200
2 23.5 600 × 800 4-D29
2-D10@200
Figure 9. Seismic collapse analysis of RC framed structure: (a) without using contact
algorithm; and (b) using contact algorithm.
on the 4th oor at 2:6 s, where the main reinforcement quantity was articially reduced
to 10%. At the next instant, bending moments between the lateral beams and the columns
become critical at upper oors, causing member fracture throughout the entire structure. The
mass of upper oors falls to the ground in Figure 9(a), while in Figure 9(b), where the
contact algorithm is used, the impact energy of the falling members is dispersed to the lower
structure, causing additional damage. The collapse procedure stops at 5:0 s in both cases but
indicates completely dierent collapse modes. By considering the contact between members,
the reproduction of seismic damage observed in actual earthquakes, such as intermediate-layer
failure, became possible. However, failure initiates at midspan of the beam in both cases,
which is dierent from the actual behavior where failure tends to start from beam–column
connections. The dierence may be due to the underestimation of the critical shearing strain
and disregard of wall and oor slabs, which may naturally lead to structural weakness against
excessive loads. The computing time using SUN Ultra 5 (CPU: 270 MHz, memory: 128 MB)
was approximately 50 minutes.
6. CONCLUDING REMARKS
In this paper, a non-linear nite element code using the ASI technique with an RC beam
model was developed in order to analyze seismic collapse problems including structural dis-
continuities. The fracture of a section was modeled by shifting the numerical integration point
with simultaneous release of the resultant forces. The proposed code was improved by consid-
ering the contact between members in order to obtain results that agree more closely with the
actual behavior. The results revealed that this code can be used in the numerical estimation of
the seismic design of RC framed structures. However, the damping matrices, as well as the
wall and oor slabs, should also be considered in the future studies to improve the accuracy.
REFERENCES
1. Cundall PA. A computer model for simulating progressive, large-scale movement in blocky rock system.
Proceedings of the International Symposium on Rock Mechanics 1971, II-8, 129 –136.
2. Shi GH, Goodman RE. Discontinuous deformation analysis. Proceedings of the 25th U.S. Symposium on Rock
Mechanics 1984, 269 – 277.
3. Meguro K, Hakuno M. Simulation of collapse process of structures due to earthquake. Proceedings of the
Symposium on Computational Methods in Structural Engineering and Related Fields 1991, Vol. 15, 325 – 330
(in Japanese).
4. Itoh M, Yoshida N, Utagawa N, Kondo I. Simulation of blast demolition of reinforced concrete buildings.
Proceedings of the Third World Congress on Computational Mechanics 1994, 1152–1153.
5. Ma MY, Barbeau P, Penumadu D. Evaluation of active thrust on retaining walls using DDA. Journal of
Computing in Civil Engineering (ASCE) 1995; 1:820 – 827.
6. Kondo I, Utagawa N, Ito M, Yoshida N. Numerical method to simulate collapse behavior in blasting demolition
of space framed structure. Proceedings of the 13th Symposium on Computer Technology of Information,
Systems and Applications 1990, 49 – 54 (in Japanese).
7. Yarimer E. Demolition by controlled explosion as a dynamical process. Structures under Shock and Impact:
Proceedings of the First International Conference, Cambridge, Massachusetts, USA, 1989, 411– 416.
8. Tosaka N, Kasai Y, Honma T. Computer simulation for felling patterns of building. Demolition Methods and
Practice: Proceedings of the Second International Symposium, Tokyo, Japan, 1988, 395 – 403.
9. Tagel-Din H, Meguro K. Nonlinear simulation of RC structures using applied element method. Structural
Engineering=Earthquake Engineering 2000; 17:137 –148.
10. Tagel-Din H, Meguro K. Analysis of small scale RC building subjected to shaking table tests using applied
element method. Proceedings of the 12th World Conference on Earthquake Engineering, Auckland, New
Zealand 2000.
11. Toi Y, Isobe D. Adaptively shifted integration technique for nite element collapse analysis of framed structures.
International Journal for Numerical Methods in Engineering 1993; 36:2323 –2339.
12. Toi Y, Isobe D. Finite element analysis of Quasi-static and dynamic collapse behaviors of framed structures by
the adaptively shifted integration technique. Computers and Structures 1996; 58:947– 955.
13. Isobe D, Toi Y. Analysis of structurally discontinuous reinforced concrete building frames using the ASI
technique. Computers and Structures 2000; 76:471– 481.
14. Toi Y. Shifted integration technique in one-dimensional plastic collapse analysis using linear and cubic nite
elements. International Journal for Numerical Methods in Engineering 1991; 31:1537 –1552.
15. Bathe KJ. Finite Element Procedures. Prentice-Hall, pp. 565 –566, 1995.
16. Architectural Institute of Japan, Design guidelines for earthquake resistant reinforced concrete buildings based
on inelastic displacement concept (draft), 1997 (in Japanese).
17. Umemura H. Dynamic Seismic Design for Reinforced Concrete Buildings, Giho-do, 1973 (in Japanese).
18. Kanakubo T, Sonobe Y. Study on seismic performance of reinforced concrete members using high-strength
light-weight concrete. Journal of Structural and Construction Engineering (Transactions of AIJ) 1992;
441:63 – 72 (in Japanese).
19. Kajima Urban Disaster Prevention Workshop, Seismic Damage of Buildings, Kajima-Shuppankai, 1996 (in
Japanese).

Seismic Collapse Analysis of Reinforced Concrete Framed Structures Using The Nite Element Method

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Seismic Collapse Analysis of Reinforced Concrete Framed Structures Using The Nite Element Method

Загружено:

Авторское право:

Доступные форматы

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2003; 32:2027–2046 (DOI: 10.1002/eqe.313)

Seismic collapse analysis of reinforced concrete framed

Daigoro Isobe1; ∗; † and Michihiro Tsuda2

Received 9 August 2002

2. TIME INTEGRATION SCHEME BASED ON THE ULF

Relations between generalized strain increments and nodal displacement increments

{n } = [nn B L ] · {n u } (2)

at step n, respectively. In this study, [u T ] is calculated by successive iteration and is expressed

cos(n ) = {n−1 l + (n−1 w J − n−1 w I )}=

Relations between resultant force increments and generalized strain increments

where E, G, A, K, Ix , Iy , x and y are Young’s modulus, shear modulus, cross-sectional area,

where f is the plastic potential given as

The transformation matrix [n+1 A] is expressed as

 n+1 ∗ n+1 ∗ n+1 ∗ 

Implicit time integration scheme using Newmark’s method

[M ]{n u } = {n E } − {nn F } (19)

Then the following incremental stiness equation is evaluated:

[M1 ]{u } + [M2 ]{u b } + [nn K 1 ]{u} + [nn K 2 ]{ub } = 0 (24)

{u} = {us } + {ud } (25)

{us } is evaluated, by neglecting inertia force, as follows:

{us } = − [nn K 1 ]−1 [nn K 2 ]{u b } (26)

3. ADAPTIVELY SHIFTED INTEGRATION TECHNIQUE

Figure 1. Linear Timoshenko beam element and its physical equivalent.

4. STATIC AND QUASI-STATIC ANALYSES OF RC COLUMNS

· Flexural crack strength

My = 0:5(ag y + N )g1 D (when Nmin 6N ¡0) (36a)

Nmin = −ag y (37a)

Figure 2. Simply supported column subjected to shear force.

· Stiness reduction ratio at yield point

Figure 4. Degrading trilinear model for reinforced concrete.

5. SEISMIC COLLAPSE ANALYSIS CONSIDERING MEMBER FRACTURE

Figure 6. Member fracture expressed by using ASI technique:

Figure 7. Relation of coordinates between fractured element and other elements.

|A1 Bi1 | + |A1 Bi2 | 6 Li (45a)

|A2 Bi1 | + |A2 Bi2 | 6 Li (45b)

Figure 8. Binding condition of fractured element and gap elements.

Table I. Member list (column).

Floor Fc (N=mm2 ) Sectional size (mm) Bar arrangement Bar arrangement

Table II. Member list (beam).

Floor Fc (N=mm2 ) Sectional size (mm) Bar arrangement

Вам также может понравиться

{n } = [nn B L ] · {n u } (2)

cos(n ) = {n−1 l + (n−1 w J − n−1 w I )}=

where E, G, A, K, Ix , Iy , x and y are Young’s modulus, shear modulus, cross-sectional area,

[M ]{n u } = {n E } − {nn F } (19)

Then the following incremental stiness equation is evaluated:

[M1 ]{u } + [M2 ]{u b } + [nn K 1 ]{u} + [nn K 2 ]{ub } = 0 (24)

{us } = − [nn K 1 ]−1 [nn K 2 ]{u b } (26)

My = 0:5(ag y + N )g1 D (when Nmin 6N ¡0) (36a)

Nmin = −ag y (37a)

· Stiness reduction ratio at yield point