Active Interaction Control of Civil Structures. Part 1: SDOF Systems

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2002; 31:161–178 (DOI: 10.1002/eqe.104)
Active interaction control of civil structures.

Part 1: SDOF systems
Yunfeng Zhang∗; † and W. D. Iwan

Department of Civil Engineering; ML 104-44; California Institute of Technology; 1201 E. California Blvd;
Pasadena; CA 91125; U.S.A.
SUMMARY
This paper presents a family of semi-active control algorithms termed as active interaction control (AIC)
used for seismic response control of civil structures. AIC control algorithms include active interface
damping (AID), optimal connection strategy (OCS) and tuned interaction damping (TID). A typical
SDOF AIC system consists of a primary structure, an auxiliary structure and an interaction element.
The auxiliary structure typically has sti=ness comparable to that of the primary structure while its
natural frequency is much higher than that of the primary structure. Interactions between the primary
and the auxiliary structures are de?ned by speci?c AIC control logic such that vibrational energy is
extracted from the primary structure into the auxiliary structure during a locking phase and dissipated
in the auxiliary structure in the subsequent unlocking phase. The stability of AIC control algorithms is
shown using the Lyapunov direct method. The eAcacy of AIC control algorithms is demonstrated by
the results of numerical simulations of SDOF systems subjected to seismic ground motions. Practical
issues such as sampling period and time delay are also investigated in this study. Copyright ? 2001
John Wiley & Sons, Ltd.
KEY WORDS: structural control; semi-active control; seismic response; time delay
1. INTRODUCTION
Since the concept of structural control was ?rst introduced into civil structures application [1],
a wide variety of active control strategies and hardware con?gurations have been developed
to protect structures against wind and seismic hazards [2; 3]. Although active control systems
can greatly reduce the vibrational level of structures subjected to wind or seismic excitation,
the reliance on external power and the reliability of fully active control devices does present
impediments to its wide acceptance and application in civil structures. Semi-active control
systems provide a promising alternative in this respect. Semi-active control devices potentially
o=er the reliability of passive devices, yet maintain the versatility and adaptability of fully
∗ Correspondence to: Yunfeng Zhang, Department of Civil and Environmental Engineering, Lehigh University, 13 E.
Packer Avenue, Bethlehem, PA 18015, U.S.A.
† E-mail: zyf@its.caltech.edu
Received 30 September 2000

Revised 8 April 2001
Copyright ? 2001 John Wiley & Sons, Ltd. Accepted 19 April 2001
162 Y. ZHANG AND W. D. IWAN
Figure 1. Schematic of: (a) AIC system model, (b) interaction elements,
(c) two con?gurations of AIC system.
active systems [4]. A semi-active system consists of a device whose mechanical properties can
be modi?ed in real time using an active component that operates using extremely low power.
Various semi-active control systems have been proposed and some have been implemented
in full-scale structures in Japan and the U.S. in recent years [5–7]. Extensive studies have
indicated that appropriately implemented semi-active control systems perform signi?cantly
better than passive devices and have the potential to achieve the level of performance of fully
active systems [8].
Active interaction control (AIC) algorithms have been proposed as a semi-active control
strategy [9–13]. AIC algorithms include the active interface damping (AID), optimal connec-
tion strategy (OCS), and newly developed tuned interaction damping (TID) algorithms. All
of the AIC algorithms are founded upon the same instantaneous optimal control strategy that
involves minimization of an energy-based performance index at every time instant. A typical
AIC system consists of a primary structure (PS), an auxiliary structure (AS) and an interac-
tion element (IE), as shown in Figure 1(a). The AS typically has sti=ness comparable to that
of PS but relatively much lower mass, so its frequency is much higher than PS’s. Interaction
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 31:161–178
ACTIVE INTERACTION CONTROL OF CIVIL STRUCTURES 163
between PS and AS is de?ned by a speci?c AIC control logic such that vibrational energy
is extracted from PS into AS with the help of an IE during a locked phase and dissipated in
AS in the subsequent unlocked phase.
2. FORMULATION
The following assumptions are made in this study:

• Only response in the horizontal direction is considered.
• All system parameters are known in advance and do not change during the excitation
duration. The PS and AS remain within the linear elastic range.
• The mass of IE is negligible compared to that of PS and AS and therefore the dynamics
of IE is neglected.
• Only present and past values of base acceleration and system state variables are available
to determine the control input.
Under these conditions, the equations of motion for the PS and AS subjected to base
excitation xOg are expressed as:
O + C ẋ(t) + Kx(t) = − MLxOg (t) + Su(t)
M x(t) (1)
where
x(t) = {x1 (t) x2 (t)}T is the relative displacement vector

m1 0 c1 0
M= is mass matrix; C = is damping matrix
0 m2 0 c2

k 0
K= 1 is sti=ness matrix
0 k2
L = {1 1}T ; S = {−1 1}T
and m1 ; m2 ; c1 ; c2 ; k1 ; k2 and x1 ; x2 represents the mass, damping factor, sti=ness, and

relative displacement of the PS and AS respectively, as shown in Figure 1(a). The subscripts
1 and 2 are used to denote the PS and AS respectively. xOg is the ground acceleration. u(t) is
the control force developed within the IE, the sign of u(t) is de?ned such that u(t) is positive
when the IE is in tension.
2.1. Interaction element (IE)
Two types of interaction elements are considered herein. As depicted in Figure 1(b), a Type
1 IE consists only of an On=O= locking device, which is capable of providing a rigid con-
nection between PS and AS when locked, and upon unlocking the force within the IE is
instantaneously reduced to zero. A Type 2 IE is comprised of a fuse device and an On=O=
locking device described above. Examples of the fuse device could be Coulomb frictional
damper or a hydraulic pressure-limiting device. It is noteworthy that the recently developed
magneto-rheological damper can also be used for the Type 2 IE.
Under the assumption that during a locking phase the velocities and accelerations of the
PS and AS are equal, the control force u(t) developed reactively within the IE is expressed
as
m1 k2 x2 (t) − m2 k1 x1 (t) m1 c2 − m2 c1
u(t) = + ẋ1 (t)
m1 + m 2 m1 + m 2
For Type 2 IE, u(t) is kept lower than or equal to a preset maximum control force level
umax .
2.2. Control strategy
The aim of the control strategy is to reduce the motion of the PS as much as possible. This
objective can be achieved by minimizing the time derivative of the relative vibrational energy
of the PS at every time instant. The relative vibrational energy of the PS is de?ned herein as
E1 ≡ 12 m1 ẋ12 + 12 k1 x12 (2)
Di=erentiating Equation (2) and substituting Equation (1) yield the time derivative of E1 as
follows:
Ė1 = (m1 xO1 + k1 x1 )ẋ1 = [−u(t) − m1 xOg − c1ẋ1 ]ẋ1 (3)
In practice, the control algorithm has to be implemented in discrete time. The duration [ta ; tb ]
of external disturbance is uniformly partitioned into a set of appropriate short time intervals,
each of duration Tt and referred to as a control-sampling period. Consider a representative
short time interval, the kth sampling period, de?ned as t ∈ [tk ; tk+1 ]. As noted above, e=ort is
directed towards minimizing the change in E1 for the sampling period, denoted as TE1k , or
causing this change to be as negative as possible.
tk+1
TE1k = [−u(t) − m1 xOg − c1 ẋ1 ]ẋ1 dt
tk
= −[u(t)ẋ1 ]|t=tk Tt − [m1 xOg ẋ1 + c1 ẋ12 ]|t=tk Tt + O(Tt 2 ) (4)
If the sampling period is small enough, the higher order term O(Tt 2 ) may be neglected. In
addition, the second term in Equation (4) is constant regardless of the operating state selected
for the IE. Hence, to the degree of approximation considered, the di=erence in values of TE1k
for di=erent IE operating states depends solely on the ?rst term of Equation (4).
The control strategy can now be described as follows. At the beginning of each sampling
period, the state variables of the PS and AS are measured; a control processor then uses
this information to determine an appropriate operating state for the IE: locked or unlocked.
In the locked state, interactions between the PS and AS are activated, while interactions are
deactivated in the unlocked state. The attachment criteria for the AIC control algorithms have
been summarized in Table I. It can be seen that one condition which will drive the ?rst term
at the right-hand side of Equation (4) as negative as possible are employed by all of the AIC
control algorithms. Finally, the control signal made at the beginning of each sampling period
is sent to the controller valve of the IE to switch or maintain the operating state of the IE.
This procedure is then repeated for each consecutive sampling period [10].
Table I. Characteristics of the AID, OCS, and TID control algorithms.

Control algorithms Attachment AS dynamics IE type and
conditions operating states
AID u(t)ẋ(t)¿0 Utilized Type 1, free=rigid
OCS u(t)ẋ1 (t)¿0 Utilized Type 1, free=rigid
ẋ1 (t) = ẋ2 (t)
TID u(t)ẋ1 (t)¿0 Utilized Type 2, free=rigid
ẋ1 (t) = ẋ2 (t) |u(t)|6umax
When the PS and AS are attached with dissimilar velocities, the induced impact may cause
damage to either or both the PS and AS. This detrimental impact e=ect can be eliminated
if an additional attachment condition is introduced. This new attachment condition requires
equal PS and AS velocities in order to initiate an interaction between the PS and AS.
However, other important reasons exist for the adoption of this additional attachment condi-
tion. In an AIC system, the function of the AS can be regarded as an actuator delivering a con-
trol force to the PS. The value of the control force delivered by the AS is approximately equal
to the elastic spring restoring force in the AS. Therefore, the larger the AS displacement, the
larger the control force will be. It is found that the PS and AS usually achieve equal velocities
at the time instant when the AS is close to its maximum displacement. A large control force
will then be generated and more energy will be dissipated from the PS if the attachment
between the PS and AS is delayed until the PS and AS have equal velocities.
Although the OCS control algorithm generally exhibits better control performance than the
AID control algorithm if time delay is ignored, the excessive number of attachment between
the PS and AS presents a new problem for the OCS system [13]. The cause for this annoying
e=ect is explained as follows: when a locked state is going to be initiated, the velocity of
the PS and AS is small as inherent in the additional equal-velocity attachment condition.
Therefore, the PS will not move far from its current position before it is pushed back by
the large AS restoring force. When the PS is pushed back and thus the velocity of the PS
changes direction, the PS and AS are then detached according to the attachment criteria. After
a half cycle of free oscillation of the AS, this short-duration attach=detach cycle is repeated
again. A great number of high-frequency small-amplitude oscillations are then generated in
the displacement time history of the PS.
If the AS motion can be attenuated before the next attachment starts, this excessive
attachment-number problem is expected to be at least partially eliminated. Furthermore, if
the attachment can be initiated at the time instant when the amplitude of the AS motion is
zero, the behaviour of this OCS system will be quite similar to that of an AID system, while
the undesirable impact e=ect due to velocity di=erence is eliminated.
The most straightforward way to reduce the amplitude of the AS motion is to provide
damping to the AS to damp out its free oscillation motion in the unlocked state. In the
present study, linear viscous damping is used. Numerical simulation results have veri?ed the
validity of this approach [14].
Careful study of the problem of excessive attachment motivates the concept of tuned in-
teraction damping (TID) control algorithm. A longer duration in the locked state is expected
by using the Type 2 IE discussed above. The maximum control force level (e.g., slip force
level for a coulomb damper) which can be developed in the IE can be tuned for the optimal
control performance of the TID system. Limiting the control force can also prevent damage
to the AS and PS caused by excessively large control forces.
2.3. Control algorithm stability
Clearly, the AIC control algorithms belong to an instantaneous optimal control category with
a performance index given by a quadratic function
J (t) = 12 z1T (t)Pz1 (t) (5)
where

k1 0
P= ; z 1 = { x1 ẋ1 }T (6)
0 m1
The PS’s equation of motion may be expressed in state space form as
ż1 (t) = Az1 (t) + Bu(t) (7)
where

0 1 0
A= ; B= (8)
−k1 =m1 −c1 =m1 −1=m1
in which the excitation term is dropped because it is not relevant to the stability of the
structure [15].
A possible Lyapunov function is given by
V (t) = z1T (t)Qz1 (t)¿0 (9)
from which
V̇ (t) = ż1T Qz1 + z1T Qż1 = z1T (QA + AT Q)z1 + 2z1T QBu (10)
Clearly, the PS without control is stable. This means that for any given symmetric positive
semi-de?nite matrix I0 , we can always obtain a symmetric positive semi-de?nite matrix Q by
solving
QA + AT Q = −I0 (11)
Therefore the ?rst term in Equation (10) becomes −z1T I0 z1 60
because I0 is positive semi-
de?nite. If it can also be shown that the second term in Equation (10) is less than or equal
to zero, it is clear that V̇ (t) is negative semi-de?nite and by Lyapunov direct method, the
primary system with control is stable.
We can choose the Q matrix in Equation (11) same as the P matrix in Equation (5). Then

0 0
I0 = −(QA + AT Q) = (12)
0 2c1
Obviously Q and I0 are symmetric, positive semi-de?nite. Now, the second term in Equa-
tion (10) becomes
z1T QBu(t) = −m1 u(t)ẋ1 (t) (13)
Figure 2. Response of the PS in free vibration.
Recall that for the AIC control strategy, the control force is determined at every time instant
such that u(t)ẋ1 (t)¿0. Therefore, it has been shown V̇ (t) is negative semi-de?nite for the
control algorithm under consideration and thus the AIC system is stable.
3. NUMERICAL SIMULATION
In this study, the natural period and damping ratio of the PS are set to 1 s, and 2 per cent,
respectively. The dynamics of the AS is fully determined by the following dimensionless
parameters:
k2 $2 2
= ; = ; = (14)
k1 $1 1
In the current numerical study, the values of ; U and are taken as 2, 20, and 1 unless
otherwise speci?ed. For the TID system, the value of is usually set to 7.5.
Two con?gurations, Type A and Type B, are considered for the AIC system, as illustrated
in Figure 1(c). A sampling period of 0:004 s is used except otherwise speci?ed in Section 3.4.
Figure 3. Control force–PS displacement curve in a free vibration, Type A.
3.1. Idealized case I: Free vibration
Consider a PS at rest, released from its initial position x0 = 1, and an AS at rest at its zero
position at the start. The damping coeAcient of the PS is set to zero to investigate the e=ect
of the AIC control algorithms. Figure 2 plots the displacement time history of the AIC system
in free vibration. It is seen that an AIC system with = 2 has a larger equivalent damping
coeAcient than one with = 0:5. Also observed is that Type A and B con?gurations yield
very similar result. In Figure 3, the relationship of control force and PS displacement (i.e.,
hysteresis loop) are plotted for each of the AIC control algorithms. For the OCS algorithm
with = 1, a large control force still persists even when the PS’s motion has been greatly
attenuated. This large control force is caused by the unattenuated motion of the AS. High-
frequency locked=unlocked cycles with the OCS system are observed from a great number
Figure 4. Displacement–time history response of the PS controlled by the AID, OCS and TID algorithms
under sinusoidal excitation, Type A.
of narrow stripes in its hysteresis diagram. For the TID algorithm with = 7:5 and umax = 15
(for = 0:5) or 20 (for = 2), the problem of high-frequency attach=detach cycles is clearly
lessened.
3.2. Idealized case II: Sinusoidal excitation
Next, consider an AIC system disturbed by a sinusoidal base excitation with a forcing period
of 3 s. The damping ratio of the PS is set as 2 per cent. Figures 4–6 plot the relative
displacement, absolute acceleration and number of attachment time history for the AID, OCS
and TID systems, respectively.
Suppression of peak displacement and absolute acceleration in seismic events is equally
important in structural control of building structures. In the TID system, because of its
bounded control force level, the acceleration of the PS is also bounded, as shown by
Figure 5.
The above-mentioned over-activation problem inherent in the OCS system is once again
identi?ed. As previously expected, the attachment duration is increased and the number of
locked=unlocked cycles is reduced in the TID algorithm. The TID algorithm also yields a much
Figure 5. Acceleration–time history response of the PS controlled by the AID, OCS and TID algorithm
under sinusoidal excitation, Type A.
more favourable continuous hysteresis loop shape in contrast to the narrow-striped hysteresis
loop shape for the OCS system, as shown in Figure 7.
3.3. Seismic excitation
Four historical records were used to evaluate the proposed control algorithms. They include:
(1) The N–S component of El Centro (ELC) record from the 1940 Imperial valley earthquake.
(2) The N–S component of Hachinohe (HAC) from the 1968 Tokachi-oki earthquake. (3) The
N–S component of Sylmar Co. Hospital (SCH) from the 1994 Northridge earthquake. (4) The
N–S component of Kobe (KOB) from the 1995 Hyogo-ken Nanbu earthquake. Due to space
limitations, only the results for the ELC ground motion are shown herein. Interested readers
are referred to Reference [14] for a detailed description of other results that are not provided
in this paper.
A typical displacement time history of the PS for the AID, OCS, and TID (Type A con-
?guration) are plotted in Figure 8 for the ELC ground motion. In Table II, the peak and
root-mean-square (RMS) values of the signi?cant response quantities are listed for both Type
A and B con?guration of the AIC systems.
Figure 6. Attachment time histories controlled by the AID, OCS and TID algorithms under sinusoidal
excitation (1: attachment, 0: detachment), Type A.
Table II. Acceleration, velocity, and displacement of the PS and AS under ELC ground motion.∗
Algorithms Displacement (cm) Velocity (cm s−1 ) Acceleration (cm s−2 )

Peak RMS Peak RMS Peak RMS
Uncontrolled A=U 16.8 4.0 117.7 25.6 664.2 157.0
A=L 8.8 3.6 98.9 39.1 1045.0 425.5
AID PS 2.2 0.5 20.6 3.7 378.5 60.1

AS 4.2 0.7 508.0 24.9 67.0G 3.3G
Type A OCS PS 1.4 0.3 12.1 2.2 308.1 62.4
AS 3.8 0.9 468.0 52.5 61.6G 7.1G
TID PS 2.3 0.5 25.1 3.6 259.9 56.7
AS 2.3 0.7 221.2 23.8 35.0G 3.4G
AID PS 2.2 0.5 20.2 3.7 375.6 63.0

Type B OCS PS 1.4 0.3 13.5 2.2 309.2 71.0
TID PS 2.2 0.5 24.3 3.6 260.7 60.0
∗
Note: A=U = Always unlocked; A=L = Always locked; 1G = 981:5 cm s−2 .
Figure 7. Control force vs PS displacement (hysteresis loop) under sinusoidal excitation, Type A.
It is seen that the peak and RMS displacements and velocities of the PS controlled by
the AID, OCS and TID algorithms are greatly reduced. In general, the performance of the
TID system is nearly equal to the AID system, while the performance of the OCS system is
slightly better than the TID system in terms of the peak displacement reduction.
However, the TID algorithm appears more attractive in regard to the AS response. It is
observed from Table II that all the response quantities of the AS in the TID system are
much lower than those of the AID and OCS systems. This favourable AS performance of
the TID algorithm may extend the lifetime of the AS and reduce the repair cost and prevent
disturbance to building functions in practice.
Table III lists the number of attachments between the PS and AS in the entire duration
(30 s) of the input ground motions. It is observed that the attachment number in the TID
system is much less than that of the OCS system. In practice, it is likely that the short-
duration attachment cycle may not be executed because of mechanical limitations. For the
OCS system, short-duration attachment cycles account for a large portion of narrow-striped
hysteresis loops. But for the TID system, narrow-striped hysteresis loops only account for a
limited portion of total dissipated energy because of the bounded control force. Therefore,
more robust performance is expected for the TID system in real practice.
3.4. Sampling period

The length of the sampling period directly a=ects the control performance of the AIC system.
Clearly, if the sampling period is set too long, the instantaneous optimal condition may not be
Figure 8. Displacement–time histories of the PS controlled by the AID, OCS and TID algorithms
excited by the ELC ground motion, Type A.
Table III. Number of attachments between the PS and AS in the entire duration (30 s) of input
ground motion ELC, HAC, SCH, and KOB for the AID, OCS, and TID control algorithms.
Algorithm Earthquake record
ELC HAC SCH KOB
AID 191 192 210 192
Type A OCS 283 261 285 270
TID 195 177 195 185
AID 188 193 204 194

Type B OCS 264 256 276 272
TID 190 180 195 184
optimal during the following sampling period and the performance of the control system may
thus deteriorate. On the other hand, an over-re?ned sampling period will impose infeasible
requirements on costly data acquisition systems and computer equipment. Hence, a proper
sampling period should be an appropriate balance between these two extremes. Figure 9 plots
Figure 9. E=ect of control–sampling period—displacement time histories of the PS controlled by the

AIC algorithms under ELC ground motion (dt = control–sampling period), Type A.
the displacement time histories of the PS controlled by the AID, OCS, and TID algorithms
under ELC ground motion. Figure 10 shows the response spectra of the AIC systems subjected
to ELC ground motion. It is seen that the TID system is much more robust to sampling period
changes than are the AID and OCS systems. Among the three AIC control algorithms, the
OCS system exhibits the highest sensitiveness to sampling period changes.
3.5. Time delay
In an AIC system, time delay can arise from (1) data acquisition, processing, and transmission,
and (2) the time taken for the actuator to fully achieve its speci?ed state.
In this study, time delay is assumed to be a constant time interval between the instant when
response variables are measured and the instant when the actuator achieves its full operating
state. It is further assumed that the length of the time delay is a multiple of the sampling
period.
The e=ect of time delay on the performance of the AIC system is examined by computing
the displacement response and the response spectra of the PS corresponding to various values
of time delay. The results for the ELC ground motion are shown in Figures 11 and 12. It is
Figure 10. E=ect of control–sampling period—response spectra of the PS controlled by the AIC
algorithms under ELC ground motion (dt = control–sampling period), Type A.
seen that the system performance deteriorates as the length of time delay increases. However,
it is worth noting that even the AIC system with a time delay of 0:16 s still has a better
performance than the uncontrolled case. The OCS algorithm is observed to be much more
sensitive to time delay than the TID algorithm.
Time delay in the AIC system will not pose stability problems because the input energy to
the AIC systems is bounded. Because of its high-frequency nature, the frequency of the AS
falls outside the predominant frequency range of seismic excitations and thus the response
of the AS is usually very small if no interaction with the PS is initiated. In an AIC system
with time delay, instead of being removed from the PS to the AS, the vibrational energy is
transferred from the AS to the PS at some delayed time point. But the motion of the PS is
still bounded because the total input energy is bounded. No instability of the AIC system due
to time delay is observed in the study.
4. CONCLUSIONS
In this paper, a family of highly e=ective semi-active control algorithms, termed AIC control
algorithms, is presented. An AIC system utilizes dynamic interactions between the primary
Figure 11. E=ect of time delay–displacement time histories of the PS controlled by the AIC algorithms
under ELC ground motion (TD = time delay), Type A.
structure and auxiliary structure to reduce the response of the primary structure under earth-
quake or wind excitation. These interactions are accomplished with the help of an actively
controlled interaction element linking the primary structure to the auxiliary structure. The hy-
draulic valve developed for the AVS system [5] can be used for the interaction element in
an AIC system with only slight modi?cation.
The response and characteristics of an AIC system in free vibration, harmonically forced
vibration and seismic excitation are investigated wherein the primary structure and auxiliary
structure are each modelled as an SDOF oscillator.
The following conclusions are drawn based on the analytical and numerical results presented
in the preceding sections and other results in Reference [14]:
1. An e=ective AIC system utilizes the dynamics of an auxiliary structure with large sti=ness
and high natural frequency that implies a light mass of the auxiliary structure.
2. The AIC control algorithms are inherently stable.
3. All of the three AIC control algorithms considered lead to signi?cant reduction in dis-
placement and velocity response of the primary structure under severe seismic
excitation.
Figure 12. E=ect of time delay—response spectra of the PS controlled by the AIC algorithms under
ELC ground motion (TD = time delay), Type A.
4. Among the three AIC control algorithms, the OCS control algorithm generally has the best
control performance in terms of displacement and velocity response reduction. However,
the OCS system also has a high number of locked=unlocked cycles. This problem can be
partly alleviated by adding proper damping to the auxiliary structure. An optimal value
for the damping coeAcient of the auxiliary structure is around 15%. By connecting a fuse
device with maximum force being optimally tuned in series with the On=O= locking de-
vice, the TID system partly solves the problems with the OCS system while maintaining
a displacement response comparable to the OCS system. Considering the impact problem
and generally low performance of the AID system compared with the other AIC systems,
the TID control algorithms is felt to have an advantage over the AID and OCS control
algorithms in the future applications.
5. Large sampling period and time delays usually result in poor performance of the AIC con-
trol system. However, numerical simulation results show that the TID control algorithm is
less sensitive to changes in sampling period and time delay.
ACKNOWLEDGEMENTS
The authors express their appreciation to Professor T. Kobori for initially conceiving and developing
the approach of active interaction control for structures, and for providing the inspiration to pursue
this research. The authors also express their gratitude to the Kobori Research Complex of the Kajima
Corporation that has provided partial support for this research.
REFERENCES
1. Yao JPT. Concept of structural control. Journal of Structural Division, ASCE 1972; 98:1567–1574.
2. Housner GW, et al. Structural control: Past, present and future. Journal of Engineering Mechanics, ASCE
1997; 123(9):897–929.
3. Spencer Jr. BF, et al. Controlling buildings: A new frontier in feedback. Control System, IEEE 1997; 17:19–35.
4. Fujino Y, et al. Structural control: Basic concepts and applications. Proceedings of the 1996 ASCE Structures
Congress, Chicago, IL, April 15–18, 1996.
5. Kobori T, Takahashi M, Nasu T, Niwa N, Ogasawara K. Seismic response controlled structure with active
variable sti=ness system. Earthquake Engineering and Structural Dynamics 1993; 22:925–941.
6. Patten W, et al. Controlled semiactive hydraulic vibration absorber for bridges. Journal of Structural
Engineering, ASCE 1996; 122:187–192.
7. Kobori T. Mission and perspective towards future structural control research. Proceedings of the Second World
Conference in structural control, Kyoto, Japan, 1998.
8. Spencer Jr. BF, et al. New applications and development of active, semi-active and hybrid control techniques for
seismic and non-seismic vibration in the USA. Proceedings of International Post-SMiRT Conference Seminar
on Seismic Isolation, Passive Energy Dissipation and Active Control of Vibration of Structures, Cheju, Korea,
August 1999.
9. Hayen JC, Iwan WD. Response control of structural systems using active interface damping. Proceedings of
the First World Conference on Structural Control, Los Angeles, USA, 1994.
10. Hayen JC. Response control of structural systems using semi-actively controlled interactions. EERL Report
95-03, Caltech, Pasadena, USA, 1995.
11. Iwan WD, Wang LJ. New developments in active interaction control. Proceedings of the Second International
Workshop on Structural Control, Hong Kong, China, 1996.
12. Iwan WD, Wang LJ. A comparison of control algorithms for active interaction control of civil structures.
Proceedings of the Second World Conference on Structural Control, Kyoto, Japan, 1998.
13. Wang LJ. Active interaction control for civil structures. EERL Report 97-04, Caltech, Pasadena, USA, 1997.
14. Zhang Y. Semi-active control of dynamically excited structures using active interaction control. Ph.D. Thesis,
EERL, California Institute of Technology, Pasadena, California, 2001.
15. Yang JN, et al. Stable controllers for instantaneous optimal control. Journal of Engineering Mechanics, ASCE
1992; 118(8):1612–1630.

Active Interaction Control of Civil Structures. Part 1: SDOF Systems

Загружено:

Сведения о документе

Исходное описание:

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

Active Interaction Control of Civil Structures. Part 1: SDOF Systems

Загружено:

Авторское право:

Доступные форматы

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2002; 31:161–178 (DOI: 10.1002/eqe.104)

Active interaction control of civil structures.

Yunfeng Zhang∗; † and W. D. Iwan

Received 30 September 2000

The following assumptions are made in this study:

x(t) = {x1 (t) x2 (t)}T is the relative displacement vector

and m1 ; m2 ; c1 ; c2 ; k1 ; k2 and x1 ; x2 represents the mass, damping factor, sti=ness, and

2.1. Interaction element (IE)

2.2. Control strategy

= −[u(t)ẋ1 ]|t=tk Tt − [m1 xOg ẋ1 + c1 ẋ12 ]|t=tk Tt + O(Tt 2 ) (4)

Table I. Characteristics of the AID, OCS, and TID control algorithms.

2.3. Control algorithm stability

Figure 2. Response of the PS in free vibration.

Figure 3. Control force–PS displacement curve in a free vibration, Type A.

3.1. Idealized case I: Free vibration

3.2. Idealized case II: Sinusoidal excitation

3.3. Seismic excitation

Algorithms Displacement (cm) Velocity (cm s−1 ) Acceleration (cm s−2 )

AID PS 2.2 0.5 20.6 3.7 378.5 60.1

AID PS 2.2 0.5 20.2 3.7 375.6 63.0

3.4. Sampling period

AID 188 193 204 194

Figure 9. E=ect of control–sampling period—displacement time histories of the PS controlled by the

3.5. Time delay

Вам также может понравиться