Sunteți pe pagina 1din 24

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL.

24, 81 1-834 (1995)

EARTHQUAKE SIMULATOR TESTING OF POUNDING BETWEEN


ADJACENT BUILDINGS
MANOLIS PAPADRAKAKIS
Institute of Struciural Analysis and Aseismic Research, National Technical University, Zografou Campus, Athens 15773. Greece
AND
HARRIS P. MOUZAKIS
Laboratory for Earthquake Engineering, National Technical University, Zografou Campus, Athens 15700, Greece

SUMMARY
Two-storey-building frames have been designed and built in an effort to investigate the pounding phenomenon of
adjacent buildings during earthquakes. Static tests have been first performed to determine the static characteristics and
the support conditions of the test structures. Sinusoidal and random acceleration signal tests have been subsequently
performed to determine the dynamic characteristics of the test structures. Two series of tests were carried out using the
shaking table simulator. In the first case, the dynamic response has been recorded without pounding, while in the second
case, the test structures have been placed with zero-gap separation and pounding was induced. Input energy time
histories were correlated and compared with and without pounding. The experimental results were compared with
analytical ones based on a formulation of the contact impact problem by the Lagrange multiplier method. Good
agreement between the experimental and the analytical results was achieved.

INTRODUCTION
Adjacent structures with inadequate clear spacing between them have suffered considerable structural and
non-structural damage as a result of their collision during earthquakes. The different dynamic characteristics
of the buildings make them vibrate out of phase and pounding occurs if the at-rest separation is insufficient to
accommodate their relative motions.
Pounding of adjacent buildings during earthquakes has received considerable attention in the last few
years. This is due to the fact that many incidents of seismic pounding have been recorded in many parts of the
world. The research work which has been published with regard to this problem is focused on observations of
the pounding damage due to actual earthquakes,',2 on analytical studies of the dynamics of pounding with
different types of idealizations and simulations of the impact p h e n ~ m e n o n ~and
- ~ on suggestions for
mitigating the dynamic implication of pounding.6 The analytical models presented in the past are based on
a number of assumptions, while no verification of their validity has ever been performed by correlating the
analytical results with data obtained from experimental tests.
In this work experimental tests were performed in an effort to investigate the pounding phenomenon and
correlate the experimental with analytical results. An attempt was made firstly to identify the static
characteristics and the support conditions, and secondly the dynamic characteristics of the test structures.
Two series of tests were performed subsequently. In the first case the dynamic response was recorded without
pounding, while in the second case the test structures were placed with zero-gap separation and pounding
was induced. Input energy time histories were recorded and compared with and without pounding. The
experimental results were compared subsequently with analytical results based on a formulation of the
impact problem by the Lagrange multiplier methodY4

CCC 0098-8847/95/0608 1 1-24


0 1995 by John Wiley & Sons, Ltd.

Received 29 April 1994


Revised 1 December 1994

812

M.PAPADRAKAKIS A N D H. P. MOUZAKIS

DESIGN AND CONSTRUCTION O F TEST STRUCTURES


The experimental and analytical studies presented in this work were performed using the test structures of
Figure 1 and their analytical model. The two-storey-building frames are made out of reinforced concrete with
the following material characteristic strengths: concrete 22 Mpa, and reinforcement 400 Mpa. The crosssectional properties of the structural members and the masses of the test structures are shown in Table I. The
dimensions of the test structures were selected on the basis of the capabilities and the characteristics of the
shaking table at the Laboratory for Earthquake Engineering of the National Technical University of
Athens.
Test structure ( I ) is designed to be flexible with respect to test structure (11) which is stiffer. This is achieved
by reducing the columns cross-sectional area and increasing the thickness of the slab at the first level of the
flexible test structure. Special care was taken in order to avoid local damage at the areas of the slabs where
pounding is going to take place. Figure 2 depicts the layout of the contact area and the detail of the
reinforcement to prevent local damage.
The test structures are fixed on the shaking table and the support conditions are determined by static
testing. The test structures are designed to remain elastic under an excitation with an acceleration design
spectrum of 1.Og. Rigid slab response is assumed at each floor. The structures are cast in two phases with
a time interval of 10 days.
EXPERIMENTAL SET-UP AND INSTRUMENTATION
The test structures are placed on the shaking table with aligned floor levels. The reference frames are based
outside the shaking table and thus the recorded displacements are the absolute ones. The overall view of the

T E S T STRUCTURE (11)

TEST STRUCTURE

(1)

/
C 4 1 40/20

CB1 4 0 / 2 0 C C 1 2 0 / 2 0

t-- 1.70
CA2 40/20

+-

\\

CD1 2 0 / 2 0

1.60 __i

CB2 40/20CC2 20/20

CD2 2 0 / 2 0

B1 20130

S H A K I N G

C A I 40/20

C B I 4 0 1 2 0 CC1 20/20

T A B L E

CD120/20

Figure 1. General arrangements of the test structures: (I) flexible; (11) stiff (dimensions in metres)

813

TESTING OF POUNDING

Table I. Second moments of area of beams and masses of


the test structures

Test
structure I

6.75

2.74 x lo3

6.75

2.10 x lo3

Test
structure I1

7.10

2.25 x lo3

6.75

2.25 x lo3

Figure 2. Reinforcement detail of the contact area (dimensions in metres)

two test structures with the reference frames is shown in Figure 3 and the schematic representation of the
instruments is depicted in Figure 4.
The displacements and accelerations along the direction of the excitation of each level of the test structures
are recorded at two different measurement points as shown in Figure 4. In order to determine the transverse
response of the test structures during testing, the accelerations at each level are also recorded in this direction.
The sampling rate is 1000 samples per second for each channel measuring the test structures and the shaking
table response. The overall response of the slabs is determined by the mean value of the two records while any
induced torsional response is also calculated. The time history of the gap is determined by the mean value of
the two displacement transducers as depicted in Figure 4.
IDENTIFICATION OF STRUCTURAL CHARACTERISTICS
Static testing is performed in order to determine the support conditions on the shaking table and the stiffness
parameter (EI) of the members.' These tests are realized by imposing static forces with two actuators between

814

M. PAPADRAKAKIS A N D H. P. MOUZAKIS

Figure 3. Overall view of the test structures with reference frames

Q
4

180

0-

32

30
35 0 1 7
29

38

LEVEL 1

41

19

31

LEVEL 2

Figure 4. Schematic representation of the instrumentation

the two test structures at each slab separately and measuring the total displacements and rotations of each
floor and the vertical movement of the supports. The parameters to be determined are the moduli of elasticity
E l , E 2 and for a two-dimensional (2D) idealization of the test structures the translational k 3 , k4 and
rotational k 5 , k6 stiffness coefficients of the support conditions as shown in Figure 5. The computed values of
the static characteristics of the structures are shown in Table 11. The stiffness coefficients were calculated by
considering the mean values of the corresponding measured displacements of the support conditions.

815

TESTING OF POUNDING
u9

u1 0

f-

u2

u3

S H A K I N G

T A B L E

Figure 5. Analytical model of the test structures

Table 11. Calculated static characteristics of the test structures


~

El
(kN/m2)

EZ

hl

(kN/m2)

(m)

k3

= k4

(kN/m)

k 5 = k,j
(kNm/rad)

Test
structure I

1.839 x lo7 1.280 x lo7

2.60

721 938

17 776

Test
structure I1

1.625 x lo7 1.150 x lo7

2.60

1351012

71 136

The natural frequencies, the mode shapes and the modal damping of the test structures are identified by
analysing the recorded signals in the frequency domain as well as in the time domain. In the frequency
domain, two approaches were used and compared. -In the first approach the test structures are excited by
a random (broad band white noise) acceleration signal and in the second approach ambient vibration
measurements are performed. In the time domain the test structures are excited by a sinusoidal acceleration
signal at the resonance of their first modes in order to evaluate the modal damping. The dynamic
characteristics of the aforementioned approaches are depicted in Table 111.

816

M. PAPADRAKAKIS AND H. P. MOUZAKIS

Table 111. Calculated dynamic characteristics of the test structures


Ambient excitation

Random excitation
Flexible

Flexible

Stiff

Stiff
-

1st

mode
Natural
frequency

4.1 1

2nd
mode
11.03

1st
mode

2nd
mode

1.22

26.38

1.00
0.511

- 1.00

1st
mode

2nd
mode

1st
mode

2nd
mode

5.20

12.40

8.80

29.60

(H4
Mode
shapes
Modal
damping

1a0
050
4.8%

0.632
- 1.00

6.0%

0.533

5.5%

Natural frequencies and mode shapes measurements


The accelerations at each level of the test structures are recorded and the natural frequencies are computed,
using the corresponding transfer functions,' and summarized in Table 111. The two approaches produced
similar transfer functions and the two natural frequencies of vibration of the test structures are identified by
comparing the peaks and the associated frequencies of the Fourier amplitude spectrum of the acceleration
recorded and those observed at each level from the transfer function curves. The ambient excitation
measurements correspond to stiffer structures and can be explained by the smaller amplitude of the
excitation of the latter case.
Damping measurements
The damping parameters are also evaluated both in frequency and in time domains. The half power
bandwidth method" is used to evaluate the modal damping in the frequency domain. The damping ratios
11,t2are depicted in Table 111. The logarithmic decrement method is used in order to identify the damping of
the first mode in the time domain. The calculated value of the damping ratio
with this method is also
depicted in Table 111. It can be seen that the results obtained by the two approaches for the damping ratio of
the first mode are comparable. The values t1 = l 2= 5 per cent are taken for the analytical studies performed
in a subsequent section.
Earthquake test using shaking table simulator
Two series of tests are performed. In the first case the test structures are placed with an adequate gap
between them so that the dynamic response is recorded without pounding. The same tests are repeated in the
second case but the structures are placed with zero-gap separation and pounding is induced. In all. tests
a horizontal translational motion of the shaking table has been created only by controlling its translational
and rotational degrees of freedom. It should be mentioned that no differences in the achieved acceleration
and displacement responses of the shaking table are observed for the two series of tests.
The desired input motion is a ramped sinusoidal displacement signal with peak displacement u = 0 1 3 cm
and a frequency f = 4.105 Hz. This frequency corresponds to the resonance of the first mode of the flexible
structure. The achieved accelerations and displacements of the shaking table are shown in Figure 6. 'These
signals were filtered by a 30 Hz digital low-pass filter. The corresponding time histories of accelerations and
displacements with pounding are not presented because they are identical with those recorded without
pounding. Three tests were performed for the case of pounding. The achieved maximum shaking table
accelerations and displacements are shown in Table IV. These series of tests are performed in order to induce
controlled non-linear response during the experiments and observe the development of the cracking patterns

TESTING OF POUNDING

817

0.10

;;; 0 0 5
v

ow

-0.10

,, , , , , , , , ,
0.00

, , , , , I , ,

, , I , , , , , ,I

,, , ,

( , , ( ,

,I

, (

1.03

200

J.00

4.w
TIME (JK)

1.w

2.m

3.00

1.00
TIME (nec)

,, , , , I ,

, ( , (

5.00

,, , , ,, , , , , , , , , , , , , , ,,,
6.00

7.w

8 0

a)
0.20

z 0 10

zi

b)

ow

, , , , , , , , I I

5.00

, , I I I I , , I ( 1 1 , , , 1 I I ,

8.00

7.w

1.00

Figure 6. Records obtained on shaking table: (a) acceleration; (b) displacement

Table IV. Shaking table response


No.
test

Acceleration
(m/sec2)

Displacement
(cm)

0.80
1.24
2.40

0.122
0.189
0 366

~~

2
3

of the elements of the structure as well as the effect of local permanent deformations at the area of contact.
The results presented in subsequent sections correspond to Test 1 except for the energy time histories studies
where the results from the three tests are presented.
DISCUSSION OF THE EXPERIMENTAL RESULTS
Displacement and acceleration response
From the recorded displacements, it was observed that torsional effects were not induced in the structures.
Figure 7 depicts the displacement response of the two levels of the flexible test structure with and without
pounding. It can be seen that its response without pounding suffers large displacements due to the resonance
induced by the input motion. When pounding is occurring the amplitude of the second level is reduced by
a factor of 2-5 and that of the first level by a factor of 2-0.A non-symmetric response along the equilibrium
state is produced while the response remains periodic.
Figure 8 depicts the displacement response of the two levels of the stiff test structure with and without
pounding. In this structure the symmetric response along the equilibrium state is observed both with and

818

M. PAPADRAKAKIS AND H. P. MOUZAKIS


-_.-.
POUNDING

LML 1
-

NO POUNDING

1 50

0.75

I Ii h n

-1

A .

- - POUNDING
NO POUNDING

3 00

LML 2

30

I5

-0.00

v,

- 1 50

-3.00

, - , , , , , , , , , , , , , , I , ,, , , , , , , , , , , , , , , , , I,,, , , , , , , , , , , , , , , , , j , , , , , , , , , , , , ~, , , , ,
Figure 7. Flexible test structure. Displacement response

without pounding. The amplitude, in the case of pounding is increased by a factor of 3 for the first level and
by a factor of 2 for the second level.
The results presented in Figures 9 and 10 show a dramatic increase of the acceleration peaks when
pounding is occurring. These peaks have reached up to a sixfold increase approximately in both structures
but they take place within a very short-time duration. The instantaneous phenomenon of the impact is :-hown
very clearly. The acceleration peaks which correspond to the occurrence of impact are taking place at the
maximum negative gap distance. From these figures, it can be observed that when a peak acceleration occurs
in one level during pounding the corresponding acceleration of the other level is not affected either at or after
impact has taken place.
Gap response
Figure 11 depicts the recorded gap opening during pounding. It can be observed that a consistent
penetration is recorded at both diaphragm levels. This is attributed to local deformations in the contact area
in connection with the instrumentation set-up as shown in Figure 4. These local deformation effects were
taken into account in the analytical studies as they affect the real activation of contact of the two structures.

819

TESTING OF POUNDING

TIME (sec)

__-

080

-6

- POUNDING
NO POUNDING

LML 2

0.40

0.w

2
vl

-0.10

-0.80
0.00

1.00

2.00

3.00

4.00

5.00

7.00

6.00

8.00

TIME (sec)

Figure 8. Stiff test structure. Displacement response

The detailed windows of the gap response shown in Figure 12, for the first and second level, reveal that
pounding is occurring periodically in two phases. In the first phase the pounding takes place at the second
level (point 1) followed by the first level (point 2), while in the second phase pounding occurs at first level
(point 3) followed by the second level (point 4).
Energies

In addition to the displacement and acceleration time histories, energy time histories are calculated. The
absolute energy equation of an N-storey building subjected to an earthquake ground motion is given by"

-21{ 4 " m l { 4

+ j{tiITCcl{duj +

{fTWul =

I( il

miiti)

dug

(1)

where [ m ] , [c] and { u } are the diagonal mass matrix, viscous damping matrix and relative displacement
vector, respectively, mi is the lumped mass associated with the ith floor, ii,, is the absolute acceleration at the
ith floor as is measured on the test structures during the experiments and (fs} is the restoring force vector.

820

M. PAPADRAKAKIS A N D H. P. MOUZAKIS
NO POUNDING

LEVEL 1

NO POUNDING

l.W

1.00
A

rn

5
w

L M L 2

2.W

2 W

0.W

0.00

9
-1.00

-l.M)

-2.w

-2.00
2.00

4.00

6.00

0 00

8.00

6.00

TIME (sec)

POUNDlNG LEVEL 1

4m

1.00

2.00

TIME (sec)

POUNMNG

LML 2
I

8.00

6.W
0-

4.00

2.W

0.00

-4.00 0 OD

TIME (sec)

6.W

8.00

Figure 9. Flexible test structure. Acceleration response

The right-hand side term of equation (1) is the absolute input energy of the systems which can also be
written as

in which all the terms are measured quantities. The absolute input energy represents the work done by the
total base shear at the foundation on the foundation displacement.
Figure 13(a) depicts the absolute input energy demand of the system considered in this study with and
without pounding. In the case of pounding the results of the three tests shown in Table IV are depicted in
Figure 13(b). The input energy of the system is the sum of the input energies of each of the two structures.
From Figure 13(a) where the total input energy demand is considerably lower in the case of pounding, the
beneficial effect of pounding on the overall response of the system can be observed. It has to be mentioned
that in the case of pounding Eindoes not correspond to the input energy induced at each test structure, since
an impact force is also present during contact. However, as the duration of the impact is almost negligible, the
work done by the impact force is not significant and therefore Ein can be considered a fairly accurate
approximation of the input energy and the impact term could not be taken into account. Figure 13(b) shows
the effect of local permanent deformations at the contact area which were observed during Test 3. The
measured permanent deformation after Test 3 was 4 mm. This allowed the flexible structure to vibrate with
longer amplitude leading to much higher input energy demand of the system.

821

TESTING OF POUNDING
NO POUNDING

LEML 1

NO POUNDING

LEVEL 2

TIME (sec)

TIME (sec)

POUNMNG LML 1

'
POUNMNG LEML 2

4.w

0 00

2.w

4.00

6.00

8.00

TIME (sec)

Figure 10. Stiff test structure. Acceleration response

The same observation is valid for the flexible test structure where the input energy demand with pounding
is only a fraction of the corresponding energy without pounding as shown in Figure 14. This picture is
however totally reversed for the stiff building where the input energy demand with pounding appears to be
four times more than the corresponding energy without pounding. It can also be seen that the maximum
input energy demand of the stiff test structure subjected to a sinusoidal acceleration at its resonant exciting
frequency, with the same duration and amplitude as the previous one, is less than the corresponding energy
during pounding.
Force-displacement interaction diagrams
Figure 15 depicts the shear force-displacement interaction diagrams for both floors of the flexible test
structure corresponding to the case without pounding, while Figure 16 depicts the corresponding inertial
force-displacement interaction diagrams for the case with pounding. The results show in both figures the
linear response of the test structures under the excitation of Test 1 since no significant hysteretic behaviour is
revealed.
Cracking patterns
During the third test, it was observed that beam B4 of Level 1 of the flexible test structure, having the
thickest slab opposite the contact area, exhibited cracking with a pattern shown in Figure 17. In all three tests
an increase in the accelerations recorded at measurement points 33 and 34 was observed, compared to those
recorded at measurement point 35 (Figure 4).

822

M. PAPADRAKAKIS A N D H. P. MOUZAKIS

-1.00

-pp7
I I I I I I,
I I I I
,,

TIME (eec)

Figure 1 1 . Gap time history

ANALYTICAL MODEL AND NUMERICAL IMPLEMENTATION


Formulation of the dynamic problem
The experimental results are compared with the analytical ones based on the formulation presented by
Papadrakakis et a1.4 This formulation of building pounding is viewed as a contact impact problem and is
treated by the Lagrange multiplier method.
The Lagrange multiplier method is used to enforce the compatibility constraints of the corresponding
displacements due to contact. Then the typical equilibrium equation of motion

[MI@} + CCI{C) + CKl{u} = { W >


is transformed, for linear response, as follows:

[,"

:[+I]:

*:[+I

KO:l[l]=[p:t)]

(3)

(4)

TESTING OF POUNDING
GAP OPENING

1,00

, , ~W OPENING

LEVEL 1

-2

om

b-

823

0.00

v)

E -0.50

LEVEL 2

1W

om
0.0

-050

- 1 .oo
-1.00

3 80

3 70

3.00

-1 50
3

3 90

TIME (sec)

FLEXIBLE TEST STRUCTURE

^a

TIME (sec)

LEVEL 1

;
;

5w

FLEXIBLE TEST STRUCTURE

lorn

LEVEL 2

500

0.00

0.00

6
<

-5.m

-5.W

-1000

-10 00

3 7a

3 00

0
TIME (sec)

^a

^a

5.w

!$

0.00

-5.03

G*

390

I 00

L M L 2

5w
5.w

STIFF TEST STRLCTURE

l0.W

l0.W

3 80
TIME (sac)

0.00

-5.00

-1000

-1000

3 7a

3 00

TIME (sec)

3eu
TIME (sec)

3 90

4 00

Figure 12. Windows of the gap time history and the acceleration response of the test structures

in which [ K J is the (m x n) contact matrix of the form

r... 1

... - 1
... 1 ... - 1
...
1 ... - 1

I...

...I

..I
...

["'1

= (6)

"I1

(A} is the vector of nodal contact forces; [MI,[Cj, [K]are the combined mass, damping and stiffness
matrices for the two buildings. The dimension rn corresponds to the total number of constraints or the total
number of nodal pairs of contact, while n is the total number of degrees of freedom of the system and (6) is
the vector of initial nodal gaps (Figure 18).

824

M. PAPADRAKAKIS A N D H. P. MOUZAKIS

/INPUT
/INPUT
/INPUT

.....

s m 00

ENERGY WITH POUNDlNG


ENERGY WITH POUNDlNG
ENERGY WITH POUNMNG

TEST 3

1 0 w 00

.--m.

73wo 00
0
>
w

20w.00

z
1000 00

0 00

0.00

1 .DO

2 00

3.00

4.00

5 00

8.00

7.00

8 0

TIME (Set)

b)
Figure 13. Input energy time histories of the system: (a) Test 1; (b) Tests 1-3

Solution of dynamic equations


The time integration of dynamic response is performed with the Newmark method in which the
displacement and velocity vectors at time t At are given by the finite difference formulae

+ %At + [(i- p ) { % } + p { r + A ' i i } ] A t 2


{f+Arlij
= {'G) + [(I
y){'ii) + ~ f ' + ~ ~ i i ] ] A t

{'+A'u} = {'u}

The implicit Newmark scheme requires the solution of the following system in each time step for the linear
problem:

825

TESTING OF POUNDING

___

STIFF TEST STAUCTURE/INPUT ENERGY WTHOUT POUNDING


TEST STRUCTURE/INPUT ENERGY WTHOUT POUNDING RESONANCE

- STIFF

.-...STIFF TEST STRUCTURE/INPUT ENERGY WTH POUNDING

1200 w

9oo.w
A

-ol3

v
-3

66

w.00

300.00

0 00
2 00

0.00

_....
FLEXfBLE

TEST

3.00

4.00
TIME (sec)

5 00

6.00

8.00

7.00

TRUCTURE INPUT ENERGY WITH POUNDING


ENERGY WITHOUT POUND

- FLEXIBLE TEST TRUCTUR&lNPUT


7

4wo.00

5000.00

:I
zw0.00

a5

1000.00

0.00
0

0 OD

Figure 14. Input energy time histories of each test structure

LEVEL 1

1.25Et005

1 25EtOOS

6.25EtDO4
.-..
z

6.25Et004
v

0
w

!?
3

O.OOf+OW

O.WE+OW

i5

-e.zs~too4

-&ZbE+004

- 1.25E+005
-0.02

!?
2

,L M L Z

-0 01

0.W

DISPLACEMENT (cm)

o.ai

0.02

m2

-0.02

-0 01

DISPLACEMENT (cm)

Figure 15 Flexible test structure. Shear force-displacement interaction diagram (without pounding)

826

M. PAPADRAKAKIS A N D H. P. MOUZAKIS
m

l.ZSE+OOS

1.2SE+005

2El

ZX)E+OOS

2.X)E+005

O.WE+OW

O.D3E+OM)

Pz

Gz

--I.mE+w

-1.z3E+ooa

-2.6OE+005

-2.K+o05
-0.E5

-010

0.00
0.w
DISPLACEMENT (cm)

DISPLACEMENT (cm)

Figure 16. Flexible test structure. Inertia force-displacement interaction diagram (with pounding)

Figure 17. Cracking pattern of the level 1 of flexible test structure

TESTING OF POUNDING
RESULTS
__ _ ANALYnCAL
EXPERIMENTAL RESULTS

3 00

s
z

827
FLEXIBLE

1.80

0,W

3
IT
a
vr

-1.m

.
1,, ,,

-3.60
0.00

,,, , , , , , j , , , , ,

I ,

2.M

1.00

,,I ,I

'I'

, , , , , , , I , , , , ,, ,, , , ,, , I , , , , I , , ,

3.00

4.00

5.m

8.W

I
, , I I I , I I I

7.00

8.W

TIME (nec)

-ANALYncAL

RESULTS
EXPERIMENTAL RESULTS

where

[It] =

!" 1, {-;=["I
It,,

(9)

in which [Zj],
{ k j )( j = I,II, ...) are the effective stiffness matrix and load vector of the Newmark method
respectively for building j.
This increase in the size system of equations is solved by using a condensation procedure for all the
displacement degrees of freedom of the system. The size of the condensed coefficient matrix depends on the
number of nodes in contact which, in general, is much smaller than the total degrees of freedom of the system.
The condensed matrix is symmetric and positive definite. In the solution technique adopted in this study the
factorization of each modified stiffness matrix is performed once, and only forward and backward substitutions are needed for each contact iteration and each time step. A skyline storage technique is used for storing
and factorizing the modified stiffness matrix.

828

M. PAPADRAKAKIS AND H. P. MOUZAKIS

COMPARISON OF EXPERIMENTAL AND ANALYTICAL RESULTS


Two mathematical models were considered for the analytical tests of this study. In the first model the
columns were discretized with three beam elements so that more regular distribution of masses can be
obtained. For the second model the masses were concentrated on the floor levels only. Both models use
a two-dimensional idealization for the test structures and rigid slab response at each floor level.
In order to eliminate the inability of the shaking table to reproduce precisely a given earthquake signal and
avoid errors in the comparison between analytical and experimental results, the input for the analytical study
is taken to be the recorded acceleration response of the shaking table.
Figure 19 depicts the comparison between the experimental and the analytical time histories of the
displacement response of the second level of the flexible and stiff test structures without pounding. These
results correspond to a distributed mass idealization. Similar analytical tests performed for lumped masses
produced less accurate results although the differences were not significant.
Figures 20 and 21 show the time histories of recorded and calculated displacement responses of the flexible
and stiff test structures respectively when pounding is occurring. The values p = 1/2 and y = 1/2 are selected

._._.
EXPERIMENTAL RESULTS

-ANALYTICAL

RESULTS

LEVEL 1

TIME (sec)

..._.EXPERIMENTAL RESULTS

0 00

- ANALYTICAL

LEVEL 2

RESULTS

0 40

-$

000

t
z
W

5 -0.40

II)

6 -0.80

-1.20

-1.60

TIME (see)

Figure 20. Flexible test structure. Comparison between experimental and analytical displacement response

829

, d

-0.40

-0.80

080

-5

, , , ! , , , , ,I , , , , , , , , , I, , , I , , , , , I / , , , , , , , , , , , , , , , , , , , , , ( , , , , , , , , , , , , I , I,,,,,,,,,

_ _ _ _ _ EXPERIMENTALRESULTS
RESULTS

-ANALynCAL

LEVEL2

0.40

g om
2E

-0.40

-0.80

TIME (see)

Figure 21. Stiff test structure. Comparison between experimental and analytical displacement response

__

0.50

z
W

0.00

0.W

- - _ EXPERIMENTAL RESULTS
RESULTS

- ANALYTICAL

LEVEL 1

0.00

i
2
w

I
V
w

5
a

LEVEL 2

E,

6-

- EXPERIMENTAL RESULTS
ANALMICAL RESULTS

-0.50

$3 -0.50

k?
-1.00

140

3.65

3.90

4 15

TIME (sac)

Figure 22. Flexible test structure. Comparison between experimental and analytical displacement response in selected time interval

830

M. PAPADRAKAKIS AND H. P. MOUZAKIS

for the Newmark parameter^,^ while the time integration step is At = 0001 sec. Figures 20 and 21 show that
the correlation between analytical and experimental results is satisfactory taking into consideration the effect
of the non-linearities involved in the polyparametric (stiffness, damping, gap, mass) system identification of
the test structures. The discrepancy of the amplitude of the response observed in the time inte:rval
3.40-4.40 sec is attributed to the uncertainty with which the dynamic characteristics of the system can be
identified.
In order to investigate in more detail the predicted response by the analytical model in this time segment,
the initial conditions obtained from the experimental results at time t = 3.40 sec are introduced in the
numerical model. The comparative results shown in Figures 22 and 23 reveal that good agreement can be
achieved when the uncertainty concerning the dynamic characteristics of the test structures is restricted to
a short time period.

_ _ - EXPERIMENTAL RESULTS
ANALYTICAL RESULTS

0 75

___

LEVEL 1

E
$ 0.25
0.25
$

0.75

b-

LEVEL 2

$fl

EXPERIMENTAL RESULTS
ANALYTICAL RESULTS

125
1.25

0.23

3
-0.25
-025

n
fl
-0 26
-0.26

-0 75
3 40

3 65

3 90
TIME (set)
(sec)

4 15

4.10
4 10

-0 75
-0.75

14

3 10

3 90

3 65

4 15

110

TIME (sec)

Figure 23. Stiff


experimenttal and analytical displacement response in selected time interval
Stiff test structure. Comparison between experimental

30

30

30

Figure 24. Adjacent 5-storey frames (dimensions in metres)

83 1

TESTING OF POUNDING

ANALYTICAL TESTS
In order to investigate the mitigation of pounding between actual buildings a series of tests was performed on
two 5-storey buildings, with zero gap between them, as shown in Figure 24. Three test cases were considered.
In the first test case the stiff building has fundamental period T , = 0.52 sec and the flexible building
TI = 1.25 sec. In the second test case the dynamic characteristics of the stiff building remain the same, while
the flexible building is made stiffer with T I = 0 9 7 sec. In the third test case the flexible building has the same
characteristics as in the first case, while the stiff building becomes more flexible with T , = 0-69 sec. Dynamic
analyses were carried out for two earthquake motions depicted in Figure 25. Figures 26 and 27 demonstrate
the input energy demand for each building when pounding and no pounding is taking place. Table V gives
the percentage change of the maximum input energy demand when the flexible structure of the first case
becomes stiffer (test case 2) and the corresponding stiff structure becomes more flexible (test case 3). The
percentage variation shown in Table V is calculated as the ratio of the difference of the input energy demand
with and without pounding to the input energy demand without pounding.
4.0
h

2.00
\

-E.,

0.w

!I

-2.W
-4.00

IO.00

5.00

0.W

15.00

25.00

20.00

TIME (aec)

2w

0.w

!i

8 -2.00

2
-4.00

0.W

5.00

1o.W

re.'^

m.m

2s.h

m.00

~5.m

TIME (sec)

b)
Figure 25. Earthquake acceleration records: (a) Kalamata, (b) El-Centro

Table V. Percentage variation of the total input energy


demand for the three test cases
El-Centro earthquake Kalamata earthquake
Test case
1
2
3

Flexible
1 1.76

- 4.32
40.33

Stiff

Flexible

Stiff

9492
76.27
- 39.78

- 2.7
- 13.61

43.75

2544

- 7.79

4.16

40.00

832

M. PAPADRAKAKIS A N D H. P. MOUZAKIS

12S.M

1m.w

ism

k5
3-

w.m

25.M

0.M

10.b
i(i.bd
TIME (Ssc)

0.64

TIME (Ssc)

1m.m

S7RLClURE/lNpUl
_ - FLEXIBLE
FLEXIBLE STAKNRE/INFUT

E N W WITH PWNMNC
E N W W M U T FWMlNG

2a.b

u.w

7s.m

= W.,.Oo
3

2 2s.w
5.

0.M

1O.b

TIME

STRVCTVRE/INFUT
_ _ FLEXIBLE
FLEXIBLE STRVCNRE/INWT
I

ENERW W I I H PWNMNG
ENERGY WrmOUT PCUNINC
I

lw.w

TIFF
, - - STIFF

1S.b

20.b

(Sec)

STRUCNRE/INPW ENERGY WITH PWNDlNC


STRUCNRE/INPVI ENERGY WHOUT POUNMNG

c)
Figure 26. Kalamata earthquake. Input energy demand: (a) Test case 1; (b) Test case 2; (c) Test case 3

25.00

833

TESTING OF POUNDING

1W.W

1D.W

1u.m

,-.
2 1W.W

g
2

75.00

ti

75.04

z
w

3
z

lW.00

50.00
50.00

25.00

21.00

0.00

0.00

0.00

500

10.DO

TIME (Scc)

20.00

15.00

2500

30.00

35.03

W.00

TIME (Sec)

lam

2w.m

STIFF STRUCTURE/INPUl
- - STIFF STRUCTURE/INFVT

ENERGY WITH P W N M N C
ENERGY WITHOUT POUNDING

1W.W

75.M

B3

g-

M.m
25.00

0.W

0.00

TIME (Sec)

_SlRKlURE/I
- - FLEXIBLE
FLEXIBLE STRLCTJRE/I

J l ENERGY W m l P W N M N G
JT ENERGY WjliOUT WUWI

STIFF SlRUCTLJRE/INPVI
- - S7FF STRUCTURE/INFVT
I

500

10W

ENERGY WITH PWNDlNC


ENERGY WITHOUT WUNDIM:

15.04

20.00
WOO
TIME (Sec)

30.00

3500

Figure 27. El-Centro earthquake. Input energy demand: (a) Test case 1; (b) Test case 2; (c) Test case 3

(404

834

M. PAPADRAKAKIS AND H. P. MOUZAKIS

CONCLUSIONS
On the basis of the experimental and analytical results presented, the following conclusions can be drawn:
(1) The experimental results demonstrate not only the amplification effect of poucding on the stiff
structure adjacent to a flexible one when the excitation is near the resonance of the flexible, but they also
reveal that the induced input energy of the stiff structure with pounding is larger than the corresponding
energy during its resonance. On the other hand, the beneficial effect of pounding for the flexible structure and
for the input energy of the whole system was also shown. However, when local damage is induced at the area
of contact, leading to an increase of the gap, then the beneficial effect of pounding for the flexible structure is
reduced.
(2) The acceleration diagrams recorded reveal a dramatic increase of acceleration peaks, up to six limes,
with respect to those recorded without pounding in both test structures. These peaks occur only at the level
where the pounding takes place without affecting the other level. Thus, the induced displacements in the
structure, due to pounding and, consequently, to the stresses produced in the columns, are not directly
affected by these acceleration peaks because the impulse is of a very short duration and is resisted by the
inertia of the slabs. These peak accelerations, however, are believed to affect the equipment and the secondary
elements of short natural period at the level where pounding takes place.
(3) Earthquake responses obtained from the analytical models without pounding are in good agreement
with those recorded from experimental testing.
(4)In the case of pounding the comparison between analytical and experimental results is satisfactory
taking into account that the analytical results are more sensitive to the identification of the dynamic
characteristics of the structures and to the accurate representation of the contact conditions. It may,
therefore, be concluded that the analytical approach, based on the Lagrange multiplier method, is a reliable
tool for studying the pounding phenomenon of adjacent buildings during earthquakes.
(5) The analytical results of the two 5-storey buildings confirm the earlier observation that the induced
input energy of the stiff structure with pounding is larger than the corresponding energy without pounding,
but the magnitude of the amplification is dependent on the characteristics of the earthquake motion. The
results also reveal that it is generally more beneficial in the case of pounding to make the stiff structure more
flexible than making the flexible structure stiffer.
ACKNOWLEDGEMENTS

The work has been supported by a grant from the General Secretary of Research and Technology of Greece.
The authors are grateful to Professor P.G. Carydis for his valuable discussions and for providing the sha.king
table facilities of the Laboratory for Earthquake Engineering and to Mr. S. Bitzarakis for his assistance in the
preparation of this study.
REFERENCES
1. The Central Greece earthquakes of February-March 1981. Reconnaissance and Engineerina Report, EERI, 1982.
2. K. Kasai, V. Jeng, P. C. Patel, J. A. Munshi and B. F. Maison, Seismic pounding effects-Survey and analysis, Proc 10th work! con$
earthquake eng., Madrid, Spain, 1992, pp. 3893-3898.
3. S . A. Anagnostopoulos and K. V. Spiliopoulos, An investigation of earthquake induced pounding between adjacent buildings,
Earthquake eng. struct. dyn. 21, 289-302 (1992).
4. M. Papadrakakis, H. Mouzakis, N. Plevris and S. Bitzarakis, A Lagrange multiplier solution method for pounding of buildings
during earthquakes, Earthquake eng. struct. dyn. 20, 981-998 (1991).
5. B. F. Maison and K. Kasai, Dynamics of pounding when two buildings collide, Earthquake eng. struct. dyn. 21, 771-786 (1992).
6 . B. D. Westenno, The dynamics of interstructural connection to prevent pounding, Earthquake eng. struct. dyn. 18,687-699 ( 1 989).
7. P. G . Carydis, J. D. Aristizabal and R. J. Pulkrabek, Six dof earthquake simulator at the National Technical University of Athens,
Proc. 7th Europian con$ earthquake eng. Athens, Greece, Vol. 3, 1982, pp. 4 7 1 4 7 8 .
8. A. E. Aktan, V. V. Bertero, A. A. Chowdhury and T. Nagashima, Experimental and analytical predictions of the mechanical
characteristics of a l j 5 scale model of a 7-storey R/C frame-wall building structure, Report No UCB/EERC-83/13, Earthquake
Engineering Research Center, University of California, Berkeley CA, 1983.
9. J.-H. Shen and A. Astaneh, Seismic response evaluation of an instrumented six story steel building, Report No. LrCB/EERC-PO/ZO,
Earthquake Engineering Research Center, University of California, Berkeley, CA, 1990.
10. R. W. Clough and J. Penzien, Dynamics ofStructures, Mc Craw-Hill, New York, 1975.
11. C. M. Uang and V. V. Bertero, Evaluation of seismic energy in structures, Earthquake eng. struct. dyn. 19, 77-90 (1990).

S-ar putea să vă placă și