You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228733852

Benchmark problem for control of base isolated buildings

Article · January 2002

CITATIONS READS
14 1,008

5 authors, including:

Sriram Narasimhan Satish Nagarajaiah


University of Waterloo Rice University
124 PUBLICATIONS   2,560 CITATIONS    291 PUBLICATIONS   11,094 CITATIONS   

SEE PROFILE SEE PROFILE

Henri Gavin
Duke University
109 PUBLICATIONS   2,544 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Data-driven Structural Dynamics & System Identification with Sparse Representation & Low-rank Modeling View project

Seismic and windstorm protection of structures through passive control View project

All content following this page was uploaded by Satish Nagarajaiah on 26 May 2014.

The user has requested enhancement of the downloaded file.


BENCHMARK PROBLEM FOR CONTROL OF BASE ISOLATED
BUILDINGS

Sriram Narasimhan 1 ,
Satish Nagarajaiah 1 , Member, ASCE,
Henri Gavin2 , Member, ASCE,
and Erik A. Johnson 3 , Assoc. Member, ASCE

ABSTRACT
This paper presents the benchmark problem definition for seismically excited base-isolated build-
ings. The objective of this benchmark study is to provide a well defined base isolated building with
a broad set of carefully chosen parameter sets, performance measures and guidelines to the partici-
pants, so that they can evaluate their control algorithms. The control algorithms may be passive, active
or semi-active. The benchmark structure considered is an eight story base isolated building similar
to existing buildings in Los Angeles, California. The base isolation system includes both linear and
non-linear bearings and devices. The superstructure is considered to be a linear elastic system with
lateral-torsional behavior. A new nonlinear dynamic analysis program has been developed and made
available to facilitate direct comparison of results of different control algorithms.
Keywords: Benchmark, Nonlinear, Base Isolated Building, Control Algorithms.

INTRODUCTION
Base isolation systems, such as sliding and elastomeric bearing systems, reduce the super-
structure response, but with increased base displacements. In the case of near-fault ground
motions, which are characterized by strong impulsive motion, excessive displacements can
occur resulting in the failure of the isolation bearings. Current practice is to provide non-linear
passive dampers to limit the bearing displacements, however, this increases the forces in the
superstructure and also at the isolation level. In order to reduce these excessive motions of the
base, active or semi-active devices are adopted.
Some of the commonly used isolation systems are elastomeric isolation systems and sliding
isolation systems. In buildings, sliding isolation systems support and decouple the superstruc-
ture from ground. Sliding bearings consist of Teflon - stainless steel, flat or spherical, interface.
Sliding bearings dissipate energy due to frictional behavior. Elastomeric isolation systems con-
sist of laminated rubber bearings reinforced with steel plates. Energy dissipation capacity is
provided by a lead plug within the rubber unit, as in lead-rubber bearings, or by inherent damp-
ing capacity of the rubber.
1
Graduate Student and Associate Professor, Dept. of Civ. and Env. Engrg., Rice Univ., 6100 S. Main Street,
Houston, TX 77005. E-mail: nsriram@rice.edu, nagaraja@rice.edu
2
Assistant Professor, Dept. of Civ. and Env. Engrg., Duke Univ., Durham, NC 27708. E-mail: hp-
gavin@duke.edu
3
Assistant Professor, Dept. of Civ. and Env. Engrg., Univ. of Southern California, Los Angeles, CA 90089.
E-mail: johnsonE@usc.edu.

1
TABLE 1. Floor Masses and Eccentricities
Floor 1 2 3 4 5 6 7 8
Mass 176 154 141 141 139 105 98 73
(kips−sec2 /f t)
ex -7.8 -7.8 -7.9 -6.6 -6.6 1.4 11.3 9.1
Eccent.
(f t) ey 8.7 6.4 -6.6 -11.3 -26.5 -32.8 -27.6 -30.8

TABLE 2. Modal Periods


N-S E-W T
Mode 1 2 3 1 2 3 1 2 3
Period (secs) 0.76 0.31 0.17 0.89 0.37 0.22 0.67 0.25 0.17

Extensive research has been performed both analytically and experimentally for active,
semi-active and hybrid control systems in civil engineering applications. Active control has
been implemented in many structures recently. Semi-active control of linear and non-linear
structures using novel devices such as Magneto-Rheological (MR) dampers, Electro-Rheological
dampers and variable stiffness systems has gained significant attention in the recent years. The
effectiveness of structural control strategies and different control algorithms has been demon-
strated, by many researchers, experimentally and analytically.
In order to provide a common basis for comparing different controllers a series of bench-
mark problems for fixed base building structures have been developed and studied widely
(Ohtori et al. 2001 and Yang et al. 2000). Participants of this benchmark study can propose
control strategies for the base isolated building model and shall define, evaluate, and report the
results for the proposed strategy. A set of evaluation criteria have also been developed for the
sake of comparison of various control strategies.

BENCHMARK BASE ISOLATED STRUCTURE


The structure is a base-isolated eight-story, steel-braced framed building, 270.4-ft long and
178-ft wide. The floor plan is L-shaped as shown in Fig.1. The benchmark structure considered
is an eight story base isolated building similar to existing buildings in Los Angeles, California.
A detailed 3D model of the fixed-base superstructure is developed with a rigid floor slab as-
sumption. The building has setbacks after the fifth floor. The superstructure bracing is located
at the building perimeter. Metal decking and a grid of steel beams support all concrete floors.
The steel superstructure is supported on a reinforced concrete slab, which is integral with con-
crete beams below, and drop panels below each column location. The isolators are connected
between these drop panels and the footings below. The seismic isolation system consists of 61
spherical sliding bearings and 31 linear elastomeric bearings as shown (Fig. 1). The superstruc-
ture is modeled as a linear elastic system. The superstructure members, such as beam, column,
bracing, and floor slab are modeled in detail (Fig.2). Three master DOF at the center of mass
of each floor are used in the condensed model. Hence, only 24 DOF are retained in modeling
the superstructure using fixed base properties. The model including the superstructure and iso-
lation system consists of 27 DOF. The mass of each floor and eccentricities between the center
of mass and center of stiffness at each floor, are shown in Table 1. The computed periods Tn
for the first nine modes in the fixed-base condition are shown in Table 2. The damping ratio is
assumed to be 5% in all modes.

2
FIG. 1. Building Plan, Bearing Locations and Elevation (All units in feet: 1 m =
3.28 ft)

THREE DIMENSIONAL NONLINEAR DYNAMIC ANALYSIS


Base isolated buildings are designed such that the superstructure remains elastic. Hence,
in this study the superstructure is modeled by a condensed linear elastic system. Also, the
localized nonlinearities at the isolation level allow condensation of the linear superstructure. In
addition, the equations of motion are developed in such a way that the fixed-base properties are
used for modeling the linear superstructure. The isolation system is modeled using linear or
nonlinear elements representing bearings, dampers or hydraulic devices, described later. The
base and the floors are assumed to be infinitely rigid in plane. The superstructure and the base
are modeled using 3 master degrees of freedom (DOF) per floor at the center of mass. Each
nonlinear isolation bearing or device is modeled explicitly using discrete biaxial Bouc-Wen
model, and the forces in the bearings or devices are transformed to the center of mass of the
base using a rigid base slab assumption. The equations of motion for the super-structure and
the base are as follows:
³ ´
MÜ + CU̇ + KU = −MR Üg + Üb (1)
h ³ ´i ³ ´
RT M Ü + R Üg + Üb + Mb Üg + Üb + Cb U̇b + Kb Ub + f + f c = 0 (2)
where M,C and K = superstructure mass, damping, and stiffness matrices, respectively, in
the fixed-base condition; R = influence matrix; Ü, U̇, and U represent the floor acceleration,
velocity and displacement vectors, respectively, relative to the base; Üb = vector of base ac-
celeration relative to the ground; Üg = vector of absolute ground acceleration; Mb = diagonal

3
FIG. 2. Detailed Three Dimensional Superstructure Model.

mass matrix of the rigid base; Cb = resultant damping matrix of viscous isolation elements;
Kb = resultant stiffness matrix of linear elastic isolation elements and f = vector containing
the forces mobilized in the isolation bearings and devices, and fc = control forces. Combining
Eq.(1) and (2) and using modal transformation U = φU0 (φ = modal matrix of the fixed-
base superstructure, normalized with respect to the mass matrix; and U0 = modal displacement
vector relative to the base), the following equation is obtained

Mc Üc +Cc U̇c +Kc Uc +Fc = BE Üg = Pc (3)


½ ¾ 0 · T ¸ · ¸
U φ Mφ φT MR φT Cφ 0
where Uc = ; Mc = ; Cc = ;
Ub RT Mφ RT MR + Mb 0 Cb
· T ¸ ½ ¾ · ¸
φ Kφ 0 0 φT MR
Kc = ; Fc = ; BE = − .
0 Kb f + fc RT MR + Mb
in which the following diagonal matrices are obtained because the modal matrix φ is nor-
malized with respect to the mass: φT M φ = I, φT Kφ = ω 2 , and φT Cφ = 2ζω, with ω =
diagonal matrix of natural frequencies of the fixed base structure and ζ = diagonal matrix of
damping ratios of the fixed base structure. Eq.(3) can also be written in incremental form, at
time t + ∆t (∆t = time step), as

M∆Üc +Cc ∆U̇c +Kc ∆Uc +∆Fc = ∆Pc (4)


½ ¾
Uc
Using X = , and Eq.(3), the state space equations can be formulated as
U̇c
³ ´
Ẋ (t) = AX (t) +Bc Fc (t) +EÜg (t) = g X, Fc , Üg (5)
· ¸ · ¸ · ¸
0 I 0 0
A= , Bc = ,E = .
−M−1
c Kc −M−1
c Cc −M−1
c −M−1
c BE

4
The Eq.(5) is solved using unconditionally stable Newmark’s constant-average acceleration
method, which can also be derived from trapezoidal rule given by

∆t
Xk+1 = Xk + (gk + gk+1 ) (6)
2
¡ ¢
where gk+1 = g Xk+1 , Fck+1 , ugk+1 ; hence, the method is implicit, needing iteration. In
incremental formulation iterations can be avoided.

The forces, f, mobilized in the elastomeric isolation bearings or devices can be modeled
by a elasticviscoplastic model with strain hardening

fx = kp Ux + cv U̇x + (ke − kp ) U y zx (7)

fy = kp Uy + cv U̇y + (ke − kp ) U y zy (8)


where ke = pre-yield stiffness, kp = post-yield stiffness, cv =viscous damping coeffi-
cient of the elastomeric bearing or device, U y is the yield displacement, and zx and zy are
dimensionless hysteritic variables (Park et al.1986).

 ³ ³ ´ ´ ³ ³ ´ ´ 
½ ¾ ½ ¾ 2 γ sgn U̇ z ½ ¾
żx U̇x zx x x + β z z
x y γ sgn U̇ z
y y + β U̇x
U y
=a −  ³ ³ ´ ´ ³ ³ ´ ´ 
ży U̇y zx zy γ sgn U̇x zx + β zy2 γ sgn U̇y zy + β U̇y
(9)
in which parameters a, β and γ are dimensionless quantities, Ux , Uy and U̇x , U̇y , represent
the displacements and velocities respectively, that occur at the isolation bearing or device.

Eq. (7), (8) and (9) can also be used to model sliding bearings with flat or spherical sliding
surface, by means of a small yield displacement U y (because of rigid plastic behavior and large
pre-yield stiffness) setting cv = 0 and (ke − kp ) U y = µN

fx = kp Ux + µN zx (10)

fy = kp Uy + µN zy (11)
where µ is the coefficient of friction and N is the average normal force at the bearing
(normal force variation is neglected). In a similar manner other devices such as nonlinear
fluid dampers can also be modeled using Eq. (7), (8) and (9). Eq.(9) is solved using the
unconditionally stable semi-implicit Runge-Kutta method (Rosenbrook 1964) suitable for the
solutions of stiff differential equations. Eq’s. (5), (7), (8) and (9) are solved using the efficient
predictor-corrector algorithm and verified using 3D-BASIS program (Nagarajaiah et al. 1991
a, b).

MATLAB IMPLEMENTATION
The analytical model is implemented using MATLAB and SIMULINK as shown in Fig.(3)-
(4). The analysis program comprises of a data file, a preliminary file to assemble the data
into system matrices for input into the non-linear analysis block 3D-BASIS-SAIC, which is a
SIMULINK based S function program as shown in Fig. 3. The full implementation procedure

5
input
Structure Data

input Input Data Form System Matrices Data


Output Data
Nonlinear Dynamic Analysis Simulink Block

MATLAB Program Prelim.m


input
Control Data

FIG. 3. Schematics for MATLAB Implementation

is also shown in Fig.4. The inputs to the nonlinear analysis block are the seismic excitation
and the control forces provided by the control devices. The non-linear response is calculated
using a predictor-corrector algorithm. All the sensor and control devices can be modeled in this
program as SIMULINK blocks and the outputs of these models fed into the analysis S-function
block.
Control Design
The proposed nominal isolation system has a fundamental isolation period of TI =2.5 sec,
with a frictional damping of µ = 10%. The benchmark participants can vary the isolation pe-
riod, TI , between 1.0 and 5.0 sec; however, to have a common basis for comparison of different
control strategies the participants need to perform one set of simulations for the nominal case
TI =2.5 sec in addition to their own control simulations. Participants must strongly justify and
defend significant deviations from the nominal isolation system.
The benchmark study participants, designers/researchers, are to design an appropriate pas-
sive, active, or semi-active control strategy, or a combination of them. It is left to the de-
signers/researchers to define the type, appropriate model and location of the sensor device(s)/
sensor(s) and to develop control algorithms. The S function program 3D-BASIS-SAIC will
remain invariant to the various control strategies developed and implemented by the partici-
pants. The various control strategies can be compared to one another by having the model and
evaluation criteria common to all controllers.
The analysis model from Eq. (5) can be written as
³ ´
Ẋ = g X, Fc , Üg (12)
³ ´
Y0 = h X, Fc , Üg (13)

where, X is the state vector, Y0 is the vector of outputs, which are to be specified in the
input data file. The control algorithm is required to take the discrete form
¡ ¢
Xk+1 = gc Xk , Yk , Fck , uk , Ugk (14)

uk = hc (Xk , Yk ) (15)
where, Xk is the discrete state vector at time t = kT , u is the control command. The
sensors and other measurement devices can be modeled similarly.

6
yo

Responses
[t', X]
Matrix
THREE-D-BASIS-SAIC J1toJ8
Earthquake Selector

Selected Outputs Performance Indices


S-Function Matrix
1 Selector
z Demux
Selected Output signal 1
Unit Delay signal out
signal 2
Mux
Sensors for Structure
yD Noise1
signal 1
Demux Device Response signal out
signal 2
J9 Sensors for Devices
Noise2
Performance Index
Matrix
Selector
Demux Device Outputs
signal signal in
Matrix
device output
Selector
Discrete Controller
selected responses
f and Device Output

control devices

FIG. 4. SIMULINK Block Diagram for Simulation of Control Algorithms

Earthquake Ground Motions


Candidate earthquakes that are to be used for the analysis are Imperial Valley-El Centro,
Northridge-Rinaldi, Northridge-Sylmar, Northridge-Newhall, Kobe, and Ji Ji-TCU. The bi-
directional ground motion for each record will be provided. The vertical ground motion will
not be considered.

EVALUATION CRITERIA
The following evaluation criteria are defined for the benchmark problem. These criteria
should be used as a common evaluation tool to benchmark the performance of the control
algorithm proposed.
1. Peak Base Shear (at isolation level)/Peak Base Shear for Uncontrolled Structure, J1 (q) =
maxkVo (t,q)k
t
maxkV̂o (t,q)k
.
t
2. Peak Structural Shear (at First Story Level)/Peak Structure Shear for Uncontrolled Struc-
maxkV1 (t,q)k
ture, J2 (q) = t
maxkV̂1 (t,q)k
.
t
3. Peak Isolator Deformation/Peak Isolator Deformation of Uncontrolled Structure, J3 (q)=
maxkdi (t,q)k
t,i
.
maxkdˆi (t,q)k
t,i
maxkdf (t,q)k
t,f
4. Peak Inter-story Drift/Peak Inter-story Drift for Uncontrolled Structure, J4 (q) = .
maxkdˆf (t,q)k
t,i
5. Peak Absolute Floor Acceleration/Peak Absolute Floor Acceleration for Uncontrolled
maxkaf (t,q)k
Structure, J5 (q) = max a (t,q) .
t,f

t,i
k f k
° °
°P °
max°
° Fk (t,q)°
°
t
6. Peak Device Force/Peak Base Shear, J6 (q) = k
maxkV0 (t,q)k .
t

7
7. RMS isolator deformation/RMS isolator deformation for the uncontrolled structure,
maxkσd (t,q)k
i
J7 (q) = maxkσdˆ(t,q)k
.
i
maxkσa (t,q)k
8. RMS acceleration/RMS acceleration for the uncontrolled structure, J8 (q)=
f
maxkσâ (t,q)k .
f
 
P  TRq 
 Fk (t,q)dk (t,q)dt
k 0
9. Total Energy Absorbed by Devices/Energy Input into the Structure, J9 (q) = .
Tq
R
<V0 (t,q)Ug (t,q)>dt
0

where, i =isolator number, 1,..., Ni (Ni =8); k=device number, 1, ..., Nd (Nd =8); f =floor
number, 1,...,Nf ; q =earthquake number: 1,...,6; t = time, 0 ≤ t ≤ Tq ; h·i = inner product;
k·k =vector magnitude incorporating NS and EW components.
Further details, data files and Matlab and Simulink files can be found at
http://www.ruf.rice.edu/~nagaraja/baseisolationbenchmark.htm.
ACKNOWLEDGMENTS
Funding for this project provided by the National Science Foundation, NSF-CAREER
Grant, with Dr. S. C. Liu and Dr. P. Chang as the Program directors, is gratefully acknowl-
edged.
REFERENCES
Gavin, H.P., (2001), “Control of Seismically-Excited Vibration using Electrorheological Mate-
rials and Lyapunov Methods,” IEEE Trans. on Control Sys. Tech., vol. 9, no. 1, pp. 27–36.
MATLAB (1994), The Math Works, Inc., Natick, Massachusetts.
Nagarajaiah, S., Reinhorn, A.M., and Constantinou, M.C., (1991a), ”3D-BASIS:Nonlinear
dynamic analysis of three dimensional base isolated structures-Part II,”Rep. No. NCEER-91-
0005, Nat. Ctr. for Earthquake Engrg. Res., State University of New York, Buffalo.
Nagarajaiah, S., Reinhorn, A.M., and Constantinou, M.C., (1991b), ”Nonlinear dynamic analy-
sis of 3-D-base-isolated structures,”J. of Str. Engrg., ASCE, Vol. 117(7), pp. 2035-2054.
Narasimhan, S., and Nagarajaiah, S., (2002), ”3DBASIS-SAIC: Nonlinear Dynamic Analysis
of smart base isolated structures, ”Mechanics Research Report No.53, Dept. of Civil and
Env. Engrg., Rice University.
Ohtori, Y., Christenson, R.E., Spencer, B.F., and Dyke, S.J., (2001), ” Benchmark problems
in seismically excited nonlinear buildings, ” http://www.nd.edu/~quake, University of Notre
Dame.
Park, Y.J., Wen, Y.K., and Ang, A.H.S., (1986), ”Random vibration of hysteritic systems under
bi-directional ground motions,” Earthquake Engrg. Struct. Dyn., Vol. 14(4), pp. 543-557.
Rosenbrock, H.H., (1964), ”Some general implicit processes for the numerical solution of
differential equations.” Computer J., Vol. 18, pp. 50-64.
Sahasrabudhe, S., and Nagarajaiah, S.,(2001), ”Sliding Isolated Buildings with Smart
Dampers," Proc. Structures Congress, ASCE, Washington, D.C., CDROM.
Spencer, B. F., Johnson, E. A., and Ramallo, J. C. (2000), ”Smart Isolation for Seismic Con-
trol," JSME Int. J., Series C, 43 (3).
Yang, J.N., Agarwal, A., Samali, B., and Wu, J.C., (2000), ”A benchmark prob-
lem for response control of wind excited tall buildings.” Proc. 14th Eng. Mech.
Cong., ASCE, UT Austin, CDROM, http://www.eng.uci.edu/~anil/benchmark.html and
http://www-ce.engr.ccny.cuny.edu/people/faculty/agrawal/benchmark.html.

View publication stats

You might also like