You are on page 1of 10

Structural Pounding via Contact Force Model

S. Naserkhaki & F.N.A. Abdul Aziz


Universiti Putra Malaysia, Serdang, Malaysia

H. Pourmohammad
Islamic Azad University, Karaj Branch, Karaj, Iran

SUMMARY:
Pounding between adjacent structures under dynamic loading alters dynamic responses of the structures which
may cause their malfunctioning or damage. Intensive force is produced between the pounded structures at the
contact point in a short time which can be modelled by a contact force model. In this paper, equations of motion
of the structures are presented while the pounding forces are included when the pounding is occurred. Three
cases; free falling of a steel ball, pounding between adjacent rods with point masses on free movement and
pounding between adjacent buildings due to earthquake are presented and the results are shown graphically. The
pounding negatively impacts the displacements and velocities of the pounded structures. The pounding is
particularly detrimental to the adjacent buildings since increases the developed story shears.

Keywords: Pounding, Contact force model, Adjacent structures

1. INTRODUCTION

Pounding between moving bodies is significant issue of dynamic systems where pounded bodies
impose intensive forces on each other for a relatively short time. Particularly, the pounding between
the adjacent buildings during earthquake excitations has been concern of civil engineers during last
couple of decades because several reports of building damages have been reported due to pounding.
The pounding between the adjacent buildings occurs when they are closely located and vibrate out of
phase due to different dynamic properties. Each building’s floor is simply considered as a moving
body with a concentrated mass during earthquake excitation which pounds to adjacent building’s
floor/column. Intensive forces, up to several times the forces induced by the earthquake, are produced
in the contact point due to the pounding.

Two models are available in order to obtain response of the pounded bodies; stereomechanical and
contact force models. The momentum conservation principle is used in the stereomechanical model
which considers coefficient of restitution to model the pounding (Muthukumar and DesRoches, 2006).
Velocities of the pounded bodies immediately after pounding are calculated based on the relative
velocities of the moving bodies immediately before pounding. Then, each pounded body continues its
movement according to new velocity imposed to it. Anyway, moving bodies are not supposed to have
mutual movements in this model. Besides, this model is unable to provide duration of the pounding
and intensity of the pounding force. This is while the mutual movement between the adjacent
buildings is very important since their movement is due to external forces imposed by the earthquake
excitation and moving of each floor affects moving of the other floors. Therefore, stereomechanical
model is not appropriate to model the pounding specially the pounding between the adjacent buildings.

On the other hand, contact force model provides the advantages of the pounding force and duration by
introducing a spring and a viscous damper in the contact point to develop pounding force. The
characteristics of the spring and viscous damper could be either linear or nonlinear. Four contact force
models are available depending on their linearity and damping (Muthukumar and DesRoches, 2006); i.
linear elastic (Fig. 1(a)), ii. linear visco-elastic (Fig. 1(b)), iii. nonlinear elastic (Fig. 1(c)) and iv.
nonlinear elastic with nonlinear damping (Fig. 1(d)). Eliminating the damping from the linear visco-
elastic and nonlinear elastic with nonlinear damping models reduces them to the linear elastic and
nonlinear elastic models, respectively.

Figure 1. Pounding force and displacement relationship; a) linear elastic, b) linear visco-elastic, c) nonlinear
elastic, d) nonlinear elastic with nonlinear damping (source: Muthukumar and Desroches, 2006)

Jankowski (2005) analyzed and compared these models and found that the least errors occurred when
he used the nonlinear elastic with nonlinear damping model and linear visco-elastic model to simulate
the pounding force. Mucthukumar and DesRoches (2006) analyzed and compared the
Stereomechanical, nonlinear elastic with nonlinear damping and linear visco-elastic models. Their
results demonstrated although the methodologies were different, the displacements from the three
models were almost similar and the differences between various model responses were relatively
small. It is general agreement that contact model do not significantly affect building responses (Abdel
Raheem, 2006; Filiatrault et al., 1995; Maison and Kasai, 1992; Anagnostopoulos, 1992). It has been
shown that effect of pounding on the building responses can be reasonably achieved by means of any
contact force model. The linear visco-elastic model is been used in this present study because it offers
the advantages of obtaining the pounding force considering energy dissipation during the pounding.
Furthermore it is simple, reliable and practical.

The linear visco-elastic contact force model consists of a linear spring representing the stiffness of the
pounding and a viscous damper representing the energy dissipation during pounding. It is inactive
when the adjacent bodies vibrate individually and freely; however, it is activated when the separation
gap is closed, causing the bodies pound together. The pounding force, F, developed during the
pounding is as follow:

F = k p δu + c p δu& (1)

where kp and cp are the pounding stiffness and damping, respectively. The pounding stiffness could be
correlated to either axial stiffness of the pounded bodies or their lateral stiffness. The pounding
damping is correlated to coefficient of restitution (the ratio of the contact and separation velocities). δu
and δu& are the relative displacement and relative velocity of the pounded bodies, respectively, given
by:
δu = u1 - u 2 − SeparationGap (2)
δu& = u&1 - u& 2 (3)

Where ui and u&i are the displacements and velocities of the pounded bodies, respectively.

Three cases of the pounding of structures are presented in this paper and the pounding forces as well
as responses of the pounded structures are discussed. Initially, free falling of a steel ball on a
stationary rigid beam is presented; its movement is shown graphically and discussed. Then, pounding
between two adjacent rods with point masses on free vibration is presented and the mutual movements
of the rods are shown graphically and discussed. Finally, the mutual movement of two adjacent
buildings due to earthquake induced pounding is presented. Seismic responses of the adjacent
buildings are shown graphically and discussed in terms of the buildings’ displacements and story
shears.

2. FREE FALLING OF A STEEL BALL (SDOF CASE)

A ball is falling down from an initial height due to gravity and pounds to the ground surface. Upon the
pounding, two equal forces in opposite direction are developed between the ball and surface. The ball
is instantaneously stopped and then is moved back upward due to pounding force induced by the
surface since the surface is rigid and without movement. Anyhow, some energy dissipates during the
pounding and the ball is not back to the initial situation. Besides, gravity is still acting on the ball and
returns back the ball to the surface. These actions and reactions are repeated until the ball is fully
stopped.

Several experiments of the free falling of the steel balls on the upper surface of a rigid beam with a
leaf spring were tested by Goland et al. in 1955. Jankowski (2005) numerically simulated one of these
experiments which is schematically shown in Fig.2. Jankowski (2005) calculated only pounding forces
developed during the first pounding between the steel ball and the rigid beam via different contact
force models and attempted to compare accuracy of these models. Anyhow, our concern is to present
all poundings between the steel ball and the rigid beam and follow the movement of the ball until the
fully stop.

Figure 2. Schematic model of the free falling ball on the stationary rigid surface

Equation of motion of the free falling ball pounded to the rigid beam is as follow:

mu&& + F = mg (4)

where m is the mass of the steel ball with diameter of 0.004 m which is falling from initial height of
h=0.051 m, u&& is the acceleration of the steel ball and g is acceleration of earth gravity. If u-h<0, the
steel ball is falling freely, the pounding force is zero (F=0). However, immediately upon the pounding
(u-h=0), the pounding force is developed between the steel ball and the rigid beam (F>0). During the
pounding (u-h>0, at this condition it is numerically assumed that the steel ball penetrates the rigid
beam), the pounding force exists continuously (F>0) though immediately vanishes when the steel ball
separates from the rigid beam. The characteristics of the linear visco-elastic model is assumed in a
way that the maximum pounding force is obtained equal to 80.7 N as it was obtained from
experimental test.

Time histories of the movement of the steel ball as well as the pounding forces are shown in Fig. 3 and
Fig. 4, respectively. As it is observed from Fig. 3, the steel ball is falling down freely and is pounded
to the rigid beam in the time of 0.102 sec with downward contact velocity equal to -0.998 m/s. During
this pounding, the maximum pounding force equal to 80.7 N is developed (Fig. 4). The pounding force
changes direction of the movement of the steel ball. This means that the rigid beam motivates the steel
ball vertically by exertion of motive pounding force (or motive velocity equal to separation velocity).
Separation velocity is upward and equal to 0.652 m/s which is less than the contact velocity since
some energy dissipates during pounding. The steel ball is motivated by the pounding force and reaches
to height of 0.025 at the summit and then returns back to the rigid beam because it is subjected to
gravity force all time. The steel ball is again pounded to the rigid beam, but, downward contact
velocity is less than the previous one and as well the pounding force is smaller than the previous one.
These sequences are periodically repeated while heights of summits, contact and separation velocities,
period between two consecutive poundings and pounding forces are reduced. Finally, the steel ball is
fully stopped at time of 0.472 sec and the pounding forces are vanishes.

Figure 3. Time history of the movement of the falling steel ball

Figure 4. Time history of the pounding forces between the steel ball and rigid beam
3. POUNDING BETWEEN ADJACENT RODS WITH POINT MASSES ON FREE
MOVEMENT (2DOF CASE)

A SDOF system moves freely if no obstacle interrupts its movement; however, adjacent SDOF system
interrupts its movement if separation gap between them is less than amplitude of the moving system.
In the latter case two SDOF systems pound together during the movement; and therefore, there is
interaction between them and each system affects movement of the other system which is called 2DOF
pounding.

Zhu et al. (2002) conducted several experimental tests to evaluate seismic performance of elevated
bridges with emphasis on pounding of girders and carried out 3D numerical modelling of these
experiments. Among their cases, there is a 2DOF case of pounding between adjacent rods with point
masses on free movement as shown in Fig. 5. Both rods had been supposed to have equal masses
(m1=m2=2 kg) and equal springs (k1=k2=210.125 N/m) in the original example. Zhu et al. (2002) also
considered stiffness of contact force model equal to 154740 N/m. In addition to their assumptions, we
added damping ratio of 5% to calculate viscous damping coefficient of either SDOF system (c1, c2)
and coefficient of restitution equal to 0.65 for the linear visco-elastic model in order to take into
account dissipation of energy during the movement and pounding, respectively. Equation of the
motion of the 2DOF pounding case of adjacent rods becomes as follow:

Figure 5. Schematic model of the pounding between the adjacent rods with point masses

⎡m1 0 ⎤ ⎧u&&1 ⎫ ⎡c1 0 ⎤ ⎧u&1 ⎫ ⎡k 1 0 ⎤ ⎧u1 ⎫ ⎧ F ⎫ (5)


⎢ 0 ⎨ ⎬+ ⎨ ⎬+ ⎨ ⎬+⎨ ⎬=0
⎣ m 2 ⎥⎦ ⎩u&&2 ⎭ ⎢⎣ 0 c 2 ⎥⎦ ⎩u& 2 ⎭ ⎢⎣ 0 k 2 ⎥⎦ ⎩u 2 ⎭ ⎩- F ⎭

where ( u&&1 , u&&2 ),( u&1 , u& 2 ) and ( u1 , u 2 ) are the acceleration, velocity and displacement of adjacent rods.
This equation consists of two uncoupled equation corresponding to each rod if u1-u2-separation
gap>0. However, both equations are coupled through the pounding force, F, if u1-u2-separation
gap≤0.

Initially, it is supposed that adjacent rods withstand 0.10 m from each other while first rod is
motivated due to an initial displacement of u0=-0.10 m. Time history of the movement of the adjacent
rods are shown in Fig. 6(a). It is observed from this figure that the first rod moves freely and its
amplitude is reduced gradually and finally stops after 10 seconds. The first rod does not pound to the
second rod since separation gap between them is larger than the maximum amplitude of the movement
(u1-u2-0.10>0). So, the second rod is not motivated and keeps stopped all time.

Now, it is supposed that the adjacent rods withstand immediately from each other with zero separation
gap. Again, the first rod is motivated due to an initial displacement of u0=-0.10 m. Time history of the
movement of the adjacent rods are shown in Fig. 6(b). The first rod which is pulled 0.10 m from its
origin, is triggered and returns back to the origin while intends to cross it due to stored energy. Since
the second rod stands at the origin, the first rod pounds to the second rod at the time about 0.160 sec.
Upon the pounding, the pounding forces are developed as shown in Fig. 7. Immediately when the
pounding occurs, the contact velocities of the first and second rods are 0.940 m/s and 0, respectively.
During the pounding, the pounding force (maximum of 316 N) is developed which pushes the first rod
back and the second rod away. Therefore, the first rod is almost stopped and the second rod is
motivated. Immediately after the pounding when the rods are separated, the separation velocities of the
first and second rods are 0.149 m/s and 0.785 m/s, respectively. The second rod moves and returns
back to the origin and both rods pound together again (Fig. 6(b)). The second pounding is less
intensive than the first pounding since some energy has already dissipated during the movement and
the first pounding. This procedure is repeated while each pounding is less intensive than the previous
one and rate of reduction of the pounding forces is higher for next pounding (Fig. 7). After 2 seconds,
both rods accompany each other and move almost in a same order (Fig. 6(b)). Anyhow, they still
pound together during the movement and the last pounding happens at the time 6.540 sec but very
small pounding forces are developed at last poundings which are not visible in the Fig. 7. The
amplitudes of the movement of both rods in the pounding case are less than no-pounding case
(compare Fig. 6(a) and Fig. 6(b)). Finally, both rods are stopped before 10 seconds.

Figure 6. Time histories of the movements of the adjacent rods

Figure 7. Time history of the pounding force between the adjacent rods
4. POUNDING BETWEEN ADJACENT BUILDINGS DUE TO EARTHQUAKE (MDOF
CASE)

The pounding between the adjacent buildings occurs due to earthquake excitation if they are closely
located and have different dynamic properties and vibrate out of phase. This phenomenon has caused
building damage during almost every earthquake. The pounding of the adjacent buildings can be
simplified as the pounding of different masses corresponding to each building at the same floor level.

Filiatrault et al. (1995) tested a series of the pounding cases of two adjacent 1/8 scale single bay
moment resisting steel frame models (8-story and 3-story frames) on the uni-axial earthquake
simulator shaking table. A rigid diaphragm and its associated mass were simulated by a thick steel
plate and three attached stacks of concrete blocks at each floor. A fully welded base plate of each
column was screwed to the shaking table to provide a rigid and fixed base. Based on the free vibration
test, fundamental period of the 8-story and 3-story buildings were obtained equal to 0.605 sec and
0.341 sec, respectively. In addition, three special pounding elements were designed to measure the
pounding force time histories. Based on these pounding elements, the stiffness of the contact model
was detected equal to 12800 MN/m.

Numerical simulation of one of the tests due to the scaled 1940 El Centro earthquake (I-ELC180,
PGA=0.15g) is presented in this section. Figure 8 shows the schematic model of the adjacent buildings
connected by linear visco-elastic contact force model. The equation of the motion of these MDOF
adjacent buildings is as follow:

Figure 8. Schematic model of pounding between adjacent buildings

⎡[m]1 0 ⎤ ⎧ {u&&}1 ⎫ ⎡[c ]1 0 ⎤ ⎧ {u&}1 ⎫ ⎡[k ]1 0 ⎤ ⎧ {u}1 ⎫ ⎧ {F } ⎫ ⎡[m]1 0 ⎤ ⎧{1}⎫


+ + ⎬+⎨ ⎬ = −⎢ u&& (6)

⎣ 0 [m]2 ⎥⎦ ⎨⎩{u&&}2 ⎬⎭ ⎢⎣ 0 [c]2 ⎥⎦ ⎨⎩{u&}2 ⎬⎭ ⎢⎣ 0 ⎥ ⎨
[k ]2 ⎦ ⎩{u}2 ⎭ ⎩- {F }⎭ ⎣ 0 [m]2 ⎥⎦ ⎨⎩{1}⎬⎭ g
When the adjacent buildings are separated about 0.02 m from each other, they vibrate individually and
freely due to earthquake. Time histories of the movement of the adjacent buildings at the 3rd floor are
shown in Figure 9(a) for no-pounding case. Different phases and amplitudes of the displacements due
to different fundamental periods of the adjacent building are observed in this figure. These adjacent
buildings tend to pound together if the separation gap gets narrower because the phase of the
displacement is different. Time histories of the movement of the adjacent buildings at the 3rd floor are
shown in Figure 9(b) while they are adjusted very close and with no (virtually zero) separation gap.
The adjacent buildings with no separation gap pound together at several times; hence, vibration of
each building affects the other. The interaction between the adjacent buildings is through the pounding
force which is equal in both buildings but opposite direction ({F}). Time history of the pounding force
developed between the adjacent buildings at the 3rd floor is shown in Figure 10. The adjacent
buildings pound together from the initial seconds of the vibration, but the first significant and intense
pounding between the adjacent buildings occurs at the time of 1.61 sec with the intensity of 998 N
(Figure 10). From this time onward, the amplitudes of the displacements of the adjacent buildings are
increased (Figure 9). The most intense pounding occurs between the adjacent buildings at the time of
4.49 sec with the intensity of 6058 N (Figure 10). Each building prevents movement of the other
building in pounding side while pushes it to no-pounding side (Figure 9). The 8-story building is more
influential during the pounding because it possesses larger total mass than the 3-story building. It is
apparent from the Figure 9(b) that the 8-story building pushes the 3-story building away and changes
its maximum displacement (particularly, at the time around 4 sec).

Figure 9. Time histories of the movement of the adjacent buildings at the 3rd floor

Figure 10. Time history of the pounding forces between the adjacent buildings at the 3rd floor
The pounding between the adjacent buildings impacts not only their displacement but also story shear
produced at each floor due to earthquake excitation. Time histories of the story shears of the adjacent
buildings at the 3rd floor are shown in Figure 11. When the adjacent buildings vibrate individually and
freely due to earthquake, the maximum story shear produced at the 3rd floor of the 3-story and 8-story
buildings are 284 N Figure 10(a) and 509 N Figure 10(b), respectively. The pounding causes the
adjacent buildings experience larger story shears. The maximum story shear produced at the 3rd floor
of the 3-story and 8-story buildings are 348 N Figure 10(a) and 853 N Figure 10(b), respectively. It is
observed that the maximum story shears of both buildings are increased significantly. It is concluded
that for the closely located adjacent buildings with different dynamic properties where potential of the
pounding between the adjacent buildings is high, the pounding effect must be considered for the
building design. The seismic design of the building is performed based on the earthquake induced
story shears and displacements. The structural elements are designed to resist the story shear from
earthquake excitation and their displacements are checked to remain in allowable limits. If the
pounding effect is not considered for the seismic design of the building, it may suffer detrimentally
from the pounding because of the exceeding of the seismic responses, particularly the story shear,
from the safety limit.

Figure 11. Time histories of the story shear of the adjacent buildings at the 3rd floor

4. CONCLUSION

Three different cases of the pounding between the adjacent structures are presented in this paper. The
pounding force is simulated by a linear visco-elastic contact force model which is activated during the
pounding only. The equations of the motion of these cases are presented and solved numerically.
Finally, the results are presented graphically and discussed.

The first case is the free falling of the steel ball on the upper surface of the rigid beam. It is
numerically and graphically shown that the steel ball falls down and after pounding to the rigid
surface, is motivated by pounding force. The movements of the steel ball are repeated periodically
until it is stopped because its energy is dissipated during the poundings. The second case is the
pounding between the adjacent rods with point masses on free movement. The first rod is moved as a
result of an initial displacement but pounds to the second rod. Therefore, the rods interact together
while due to this interaction, the movement of each rod is changed and the energy of movement is
dissipated until the both rods are stopped. The last case is the pounding between the adjacent buildings
due to earthquake. The adjacent buildings are pound together because they are closely located and
vibrate out of phase. Due to the pounding, the seismic responses of both buildings are changed.
Particularly, the story shear of the buildings are increased which could lead to detrimental
consequences for the buildings. As a conclusion, the effect of pounding between the adjacent
structures should be considered in their seismic design while linear visco-elastic contact force model is
an appropriate model for this purpose.

REFERENCES

Anagnostopoulos, S.A. and Spiliopoulos, K.V. (1992). An Investigation of Earthquake Induced Pounding
between Adjacent Buildings. Earthquake Engineering and structural Dynamics 21,289-302.
Filiatrault, A., Wagner, P. and Cherry, S. (1995). Analytical Prediction of Experimental Building Pounding.
Earthquake Engineering and structural Dynamics 24,1131-1154.
Goland, M., Wickersham, P.D. and Dengler, M.A. (1955). Propagation of elastic impact in beams in bending.
Journal of Applied Mechanics 22,1–7.
Jankowski, R. (2005). Non-Linear Viscoelastic Modelling of Earthquake-Induced Structural Pounding.
Earthquake Engineering and structural Dynamics 34,595-611.
Maison, B.F. and Kasai, K. (1992). Dynamics of Pounding When Two Buildings Collide. Earthquake
Engineering and structural Dynamics 21,771-786.
Muthukumar, S. and DesRoches, R. (2006). A Hertz Contact Model with Non-Linear Damping for Pounding
Simulation. Earthquake Engineering and structural Dynamics 35,811-828.
Abdel Raheem, S.E. (2006). Seismic Pounding between Adjacent Building Structures. Electronic Journal of
Structural Engineering 6,66-74.
Zhu, P., Abe, M. and Fujino, Y. (2002). Modelling Three-Dimensional Non-Linear Seismic Performance of
Elevated Bridges with Emphasis on Pounding of Girders. Earthquake Engineering and structural Dynamics
31,1891-1913.

You might also like