Benchmark Control Problems For Seismically Excited

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/228889524
Benchmark Control Problems for Seismically Excited Nonlinear Buildings
Article in Journal of Engineering Mechanics · April 2004

DOI: 10.1061/(ASCE)0733-9399(2004)130:4(366)
CITATIONS READS
739 4,478
4 authors, including:
Yasuki Ohtori Billie F. Spencer

Central Research Institute of Electric Power Industry University of Illinois, Urbana-Champaign
15 PUBLICATIONS 835 CITATIONS 588 PUBLICATIONS 31,605 CITATIONS
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Billie F. Spencer on 18 September 2014.
The user has requested enhancement of the downloaded file.

Benchmark Control Problems for
Seismically Excited Nonlinear Buildings
Y. Ohtori1, R.E. Christenson and B.F. Spencer, Jr.

Department of Civil Engineering and Geological Sciences
University of Notre Dame, Notre Dame, Indiana 46556-0767, USA
and
S.J. Dyke
Department of Civil Engineering
Washington University, St. Louis, Missouri 63130, USA
Abstract
This paper presents the problem definition and guidelines of a set of benchmark control
problems for seismically excited nonlinear buildings. Focusing on three typical steel structures, 3-
, 9- and 20-story buildings designed for the SAC project for the Los Angeles, California region,
the goal of this study is to provide a clear basis to evaluate the efficacy of various structural con-
trol strategies. A nonlinear evaluation model has been developed that portrays the salient features
of the structural system. Evaluation criteria and control constraints are presented for the design
problems. The task of each participant in this benchmark study is to define (including sensors and
control algorithms), evaluate and report on their proposed control strategies. These strategies may
be either passive, active, semi-active or a combination thereof. The benchmark control problems
will then facilitate direct comparison of the relative merits of the various control strategies. To
illustrate some of the design challenges, a sample control strategy employing active control with a
linear quadratic Gaussian (LQG) control algorithm is applied to the 20-story structure.
Introduction
The protection of civil structures, including material content and human occupants, is, with-
out doubt, a world-wide priority. The extent of protection may range from reliable operation and
occupant comfort to human and structural survivability. Civil structures, including existing and
future buildings, towers and bridges, must be adequately protected from a variety of events,
including earthquakes, winds, waves and traffic. The protection of structures is now moving from
reliance entirely on the inelastic deformation of the structure to dissipate the energy of severe
dynamic loadings, to the application of passive, active and semi-active structural control devices
to mitigate undesired responses to dynamic loads. However, even in controlled structures, it can
be expected that large seismic events, such as the 1994 Northridge and the 1995 Kobe earth-
quakes, will cause structural members to exceed the elastic limit.
1. On leave from the Central Research Institute of Electric Power Industry of Japan.
page 1
In the last two decades, many control algorithms and devices have been proposed for civil
engineering applications (Soong 1990; Housner, et al. 1994; Soong and Constantinou 1994;
Fujino, et al. 1996; Housner, et al. 1997; Spencer and Sain 1997), each of which has certain
advantages, depending on the specific application and the desired objectives. The seismic
response of nonlinear structures to severe earthquakes has also been studied and control algo-
rithms for these nonlinear structures have been proposed by a number of researchers (Masri, et al.
1981; Masri, et al. 1982; Reinhorn, et al. 1987; Yang, et al. 1992; Yang, et al. 1994; Yang, et al.
1995; Pantelides and Nelson 1995; Chang, et al. 1996; Bani-Hani and Ghaboussi 1998; Tsai, et
al. 1998; Barrosa 1999). It is evident that different researchers use different structures and differ-
ent criteria to show the efficacy and effectiveness of their own particular control strategies. At the
present time, structural control research, as a whole, is greatly diversified with regard to these spe-
cific applications and desired objectives. A common basis for the comparison of various algo-
rithms and devices does not currently exist. Determination of the general effectiveness of
structural control algorithms and devices, a task which is necessary to focus future structural con-
trol research and development, is challenging.
Ideally, each proposed control strategy should be evaluated experimentally under conditions
that closely model the as-built environment. However, it is impractical, both financially and logis-
tically, for all researchers in structural control to conduct even small-scale experimental tests. An
available alternative, denoted “software testbeds” by Caughey (1998), is the use of consensus-
approved, high-fidelity, analytical benchmark models to allow researchers in structural control to
test their algorithms and devices and to directly compare the results.
The American Society of Civil Engineers (ASCE) Committee on Structural Control has rec-
ognized the significance of structural control benchmark problems. The Committee developed a
benchmark study, focusing primarily on the comparison of structural control algorithms for three-
story building models. The initial results of this study were reported at the 1997 ASCE Structures
Congress, held in Portland, Oregon (Spencer, et al. 1997; Balas 1997; Lu and Skelton 1997; Wu,
et al. 1997; Smith, et al. 1997). Additionally, a more extensive analysis of these benchmark struc-
tural control problems formed the basis for a special issue of Earthquake Engineering and Struc-
tural Dynamics (Spencer, et al. 1998 a, b). Several of these algorithms have been experimentally
verified at the University of Notre Dame’s Structural Dynamics and Control/Earthquake Engi-
neering Laboratory (Baker, et al. 1999).
Building on the foundation laid by the ASCE Committee on Structural Control, the next
generation of benchmark structural control studies were initiated by the Working Group on Build-
ing Control during the Second International Workshop on Structural Control held December 18–
20, 1996, in Hong Kong (Chen 1996). As stated by the Working Group, the goal of this effort was
to develop benchmark models to provide systematic and standardized means by which competing
control strategies, including devices, algorithms, sensors, etc., can be evaluated. This goal drives
the next generation of structural control benchmark problems, and its achievement will take the
structural control community another step toward more readily allowing implementation of inno-
vative control strategies for dynamic hazard mitigation.
As an outgrowth of the workshop in Hong Kong, two benchmark problems for the control of
buildings were developed for presentation at the 2nd World Conference on Structural Control held
June 28–July 1, 1998, in Kyoto, Japan. The first, detailed in Yang, et al. (1999), proposed a bench-
mark problem for wind excited buildings. The second benchmark problem, detailed in Spencer, et
al. (1999), was the next generation benchmark control problem for seismically excited buildings.
These two benchmark studies were successful, although the structural models were considered to
page 2
remain perfectly elastic. Large magnitude earthquakes can, however, cause material yielding in
the structural elements of civil structures, resulting in nonlinear responses. At the First World
Conference on Structural Control held in Pasadena, the necessity of taking into account structural
nonlinearity was identified. During the 2nd World Conference on Structural Control, as a result of
the success of the linear benchmarks presented, it was decided to pursue a nonlinear analysis.
The benchmark problem presented herein is an extension of the seismically excited next
generation benchmark problem presented at the 2nd World Conference on Structural Control to
include nonlinear responses and to address other building heights. For this nonlinear benchmark
study, 3-, 9- and 20-story steel buildings designed for the SAC2 project are the benchmark struc-
tures considered. Descriptions of these structures are discussed in the next section. High-fidelity
nonlinear models of the structures are developed and are designated as the nonlinear evaluation
models. In the context of this study, the evaluation models are considered to be true models of the
structural systems. Several evaluation criteria, measuring the effectiveness of the control strate-
gies to reduce undesired responses of the evaluation model to ground excitation, are given, along
with the associated control design constraints. Researchers/designers participating in this bench-
mark study can propose control strategies for any of the three benchmark buildings and shall
define (including sensors and control algorithms), evaluate and report the results of their proposed
control strategies for each building considered. The location on the structure must be specified for
each control device and sensor employed. Passive, active and semi-active devices, or a combina-
tion thereof, may be considered. Although the dynamics of the control actuators are important to
the control design problem (Dyke, et al. 1995), for this initial nonlinear benchmark problem, the
dynamics of control devices are not considered. Appropriate models for sensors must be identi-
fied. For active and semi-active control strategies, control algorithms must also be specified. For
illustrative purposes, a complete sample control design, similar to that employed in the next gen-
eration benchmark study, is presented for the 20-story benchmark building. Although this sample
control system is not intended to be competitive, it demonstrates how one might identify the sen-
sors and control devices employed, build a design model, develop a controller and evaluate a com-
plete control system design.
Benchmark Structures
The 3-, 9- and 20-story structures used for this benchmark study were designed by Brandow
& Johnston Associates3 for the SAC Phase II Steel Project. Although not actually constructed,
these structures meet seismic code and represent typical low-, medium- and high-rise buildings
designed for the Los Angeles, California region. These buildings were chosen because they also
serve as benchmark structures for the SAC studies and, thus, will provide a wider basis for the
comparison of results. The specifications for each of these buildings are discussed in the follow-
ing paragraphs.
2. SAC is a joint venture of three non-profit organizations: The Structural Engineers Association of California
(SEAOC), the Applied Technology Council (ATC) and California Universities for Research in Earthquake
Engineering (CUREE). SAC Steel Project Technical Office, 1301 S. 46th Street, Richmond, CA 94804-4698.
http://quiver.eerc.berkeley.edu:8080/.
3. Brandow & Johnston Associates, Consulting Structural Engineers, 1660 W. Third St., Los Angeles, CA 90017.
page 3
3-Story Benchmark Building
The three-story (3-story) benchmark structure is 36.58 m (120 ft) by 54.87 m (180 ft) in
plan, and 11.89 m (39 ft) in elevation. The bays are 9.15 m (30 ft) on center, in both directions,
with four bays in the north-south (N-S) direction and six bays in the east-west (E-W) direction.
The building’s lateral load-resisting system is comprised of steel perimeter moment-resisting
frames (MRFs) with simple framing between the two furthest south E-W frames. The interior
bays of the structure contain simple framing with composite floors.
The columns are 345 MPa (50 ksi) steel. The columns of the MRF are wide-flange. The lev-
els of the 3-story building are numbered with respect to the ground level (see Figure 1). The 3rd
level is the roof. Typical floor-to-floor heights (for analysis purposes measured from center-of-
beam to center-of-beam) are 3.96 m (13 ft). The column bases are modeled as fixed (at the ground
level) to the ground.
The floors are composite construction (i.e., concrete and steel). The floor system is com-
prised of 248 MPa (36 ksi) steel wide-flange beams acting compositely with the floor slab. In
accordance with common practice, the floor system, which provides diaphragm action, is assumed
to be rigid in the horizontal plane. The inertial effects of each level are assumed to be carried
evenly by the floor diaphragm to each perimeter MRF, hence each frame resists one half of the
seismic mass associated with the entire structure.
The seismic mass of the structure is due to various components of the structure, including
the steel framing, floor slabs, ceiling/flooring, mechanical/electrical, partitions, roofing and a
penthouse located on the roof. The seismic mass of the first and second levels is 9.57×105 kg
(65.5 kips-sec2/ft) and the third level is 1.04×106 kg (71.0 kips-sec2/ft). The seismic mass of the
entire structure is 2.95×106 kg (202 kips-sec2/ft). The 3-story N-S MRF is depicted in Figure 1.
Building Plan
NOTES
Beams (248 MPa):
beam sizes as indicated in figure.
Columns (345 MPa):
N
column sizes same throughout elevation.
Restraints:
columns fixed at base;
A A Connections:
indicates a moment resisting connection;
indicates a simple (hinged) connection.
Elevation Dimensions:
all measurements are center line;
A–A floor-to-floor heights 3.96 m (13′-0″);
3rd (16) W24x68 (17) W24x68 (18) W24x68 (19) W21x44 (20) bay widths (all) 9.15 m (30′-0″).
Seismic Mass:
2nd (11) W30x116 (12) W30x116 (13) W30x116 (14) W21x44 (15)
including steel framing, for both N-S MRFs;
9.57×105 kg;
W14x68 (weak axis)
1st–2nd levels
1st (6) W33x118 (7) W33x118 (8) W33x118 (9) W21x44 (10)
3rd level 1.04×106 kg;
W14x257
W14x311
W14x257
Ground (1) (2) (3) (4) (5)

entire structure 2.95×106 kg.
Figure 1. 3-Story Benchmark Building N-S MRF.
page 4
The nine-story (9-story) benchmark structure is 45.73 m (150 ft) by 45.73 m (150 ft) in plan,
and 37.19 m (122 ft) in elevation. The bays are 9.15 m (30 ft) on center, in both directions, with
five bays each in the north-south (N-S) and east-west (E-W) directions. The building’s lateral
load-resisting system is comprised of steel perimeter moment-resisting frames (MRFs) with sim-
ple framing on the furthest south E-W frame. The interior bays of the structure contain simple
framing with composite floors.
The columns are 345 MPa (50 ksi) steel. The columns of the MRF are wide-flange. The lev-
els of the 9-story building are numbered with respect to the ground level (see Figure 2). The ninth
level is the roof. The building has a basement level denoted B-1. Typical floor-to-floor heights (for
analysis purposes measured from center-of-beam to center-of-beam) are 3.96 m (13 ft). The floor-
to-floor height of the basement level is 3.65 m (12 ft) and for the first floor is 5.49 m (18 ft).
Building Plan NOTES

Beams (248 MPa):
Ground– 2nd level W36x160;
3rd – 6th level W36x135;
N 7th level W30x99;
8th level W27x84;
9th level W24x68.
Columns (345 MPa):
A A
column sizes change at splices
corner columns and interior columns the same,
Elevation respectively, throughout elevation;
A–A
Restraints:
columns pinned at base;
9th (61) (62) (63) (64) (65) (66) structure laterally restrained at 1st level.
Splices:
W14x257
(57)
8th (55) (56) (58) (59) (60) denoted with ;
are at 1.83 m (6 ft) w.r.t. beam-to-column joint
7th (49) (50) (51) (52) (53) (54)
Connections:
W14x283
indicates a moment resisting connection.

6th (43) (44) (45) (46) (47) (48)
indicates a simple (hinged) connection.
5th (37) (38) (39) (40) (41) (42)
Dimensions:
W14x370
4th (31) (32) (33) (34) (35) (36) basement level height 3.65 m (12′-0″);
Ground level height 5.49 m (18′-0″);
3rd (25) (26) (27) (28) (29) (30) 1st – 8th level heights 3.96 m (13′-0″);
bay widths (all) 9.15 m (30′-0″).
W14x455
2nd (19) (20) (21) (22) (23) (24)

Seismic Mass:
including steel framing, for both N-S MRFs;
1st (13) (14) (15) (16) (17) (18) 1.83 m typ
Ground level 9.65×105 kg;
1st level 1.01×106 kg;
W14x500
Ground (7) (8) (9) (10) (11) (12) 2nd – 8th level 9.89×105 kg;
9th level 1.07×106 kg;
B-1 (1) (2) (3) (4) (5) (6)
entire structure (above ground) 9.00×106 kg.
page 5
The column lines employ two-tier construction, i.e. monolithic column pieces are connected
every two levels beginning with the first level. Column splices, which are seismic (tension) splices
to carry bending and uplift forces, are located on the first, third, fifth and seventh levels at 1.83 m
(6 ft) above the center-line of the beam to column joint. The column bases are modeled as pinned
and secured to the ground (at the B-1 level). Concrete foundation walls and surrounding soil are
assumed to restrain the structure at the ground level from horizontal displacement.
The floor system is comprised of 248 MPa (36 ksi) steel wide-flange beams acting compos-
itely with the floor slab as in the 3-story building. Similar to the 3-story building, each frame
resists one half of the seismic mass associated with the entire structure.
penthouse located on the roof. The seismic mass of the ground level is 9.65×105 kg (66.0 kips-
sec2/ft), for the first level is 1.01×106 kg (69.0 kips-sec2/ft), for the second through eighth levels is
9.89×105 kg (67.7 kips-sec2/ft) and for the ninth level is 1.07×106 kg (73.2 kips-sec2/ft). The seis-
mic mass of the above ground levels of the entire structure is 9.00×106 kg (616 kips-sec2/ft). The
9-story N-S MRF is depicted in Figure 2.

The twenty-story (20-story) benchmark structure is 30.48 m (100 ft) by 36.58 m (120 ft) in
plan, and 80.77 m (265 ft) in elevation. The bays are 6.10 m (20 ft) on center, in both directions,
with five bays in the north-south (N-S) direction and six bays in the east-west (E-W) direction.
The building’s lateral load-resisting system is comprised of steel perimeter moment-resisting
frames (MRFs). The interior bays of the structure contain simple framing with composite floors.
The columns are 345 MPa (50 ksi) steel. The interior columns of the MRF are wide-flange.
The corner columns are box columns. The levels of the 20-story building are numbered with
respect to the ground level (see Figure 3). The building has two basement levels. The level directly
below the ground level is the second basement (B-1). The level below B-1 is the second basement
(B-2). Typical floor-to-floor heights (for analysis purposes measured from center-of-beam to cen-
ter-of-beam) are 3.96 m (13 ft). The floor-to-floor heights for the two basement levels are 3.65 m
(12 ft) and for the ground level is 5.49 m (18 ft).
The column lines employ three-tier construction, i.e. monolithic column pieces are con-
nected every three levels beginning with the first level. Column splices, which are seismic (ten-
sion) splices to carry bending and uplift forces, are located on the first, fourth, seventh, tenth,
thirteenth, sixteenth and eighteenth levels at 1.83 m (6 ft) above the center-line of the beam to col-
umn joint. The column bases are modeled as pinned and secured to the ground (at the B-2 level).
Concrete foundation walls and surrounding soil are assumed to restrain the structure at the ground
level from horizontal displacement.
The floor system is comprised of 248 MPa (36 ksi) steel wide-flange beams acting compos-
itely with the floor slab as in the 3-story building. Similar to the 3-story building, each frame
resists one half of the seismic mass associated with the entire structure.
penthouse located on the roof. The seismic mass, including both N-S MRFs, of the ground level is
5.32×105 kg (36.4 kips-sec2/ft), for the first level is 5.63×105 kg (38.6 kips-sec2/ft), for the second
level to 19th level is 5.52×105 kg (37.8 kips-sec2/ft), and for the 20th level is 5.84×105 kg (40.0
kips-sec2/ft). The seismic mass of the above ground levels of the entire structure is 1.11×107 kg
(760 kips-sec2/ft). The 20-story N-S MRF is depicted in Figure 3.
page 6
Building Plan
Elevation
A–A
20th (133) (134) (135) (136)
W21x50 (137) (138) N
15x15
t = 1.27 cm
19th W24x84
(127) (128) (129) (130) (131) (132)
W24x62
A A
18th (121) (122) (123) (124) (125) (126)
t = 1.91 cm
15x15
W24x117
17th (115) (116) (117) (118) W27x84 (119) (120)

NOTES
16th (109) (110) (111) (112) (113) (114)
Beams (248 MPa):
B-2 – 4th level W30x99;
t = 1.91 cm
15th (103) (104) (105) (106) (107) (108) 5th – 10th level W30x108;
11th – 16th level W30x99;
W24x131
15x15
14th (97) (98) (99) (100) W24x131 (101) (102) 17th – 18th level W27x84;
19th level W24x62;
13th (91) (92) (93) (94) (95) (96) 1.83 m typ
20th level W21x50.
(89) t = 2.54 cm
Columns (345 MPa):
12th (85) (86) (87) (88) (90)
column sizes change at splices
W24x192
corner columns and interior columns the same,

15x15
11th (79) (80) (81) (82) W30x99 (83) (84)

respectively, throughout elevation;
10th (73) (74) (75) (76) (77) (78)
box columns are ASTM A500 (15×15 indicates
a 0.38 m (15 in) square box column with wall
t = 2.54 cm
9th (67) (68) (69) (70) (71) (72) thickness of t).

Restraints:
W24x229
15x15
8th (61) (62) (63) (64) (65) (66)

columns pinned at base;
structure laterally restrained at Ground level.
7th (55) (56) (57) (58) (59) (60)
Splices:
denoted with ;
t = 2.54 cm
(49) (53)
6th (50) (51) (52 (54)
are at 1.83 m (6 ft) w.r.t. beam-to-column joint
W24x229
15x15
5th (43) (44) (45) (46) W30x108 (47) (48) Connections:

indicates a moment resisting connection,
4th (37) (38) (39) (40) (41) (42) indicates a simple (hinged) connection.
Dimensions:
3rd (31) (32) (33) (34) (35) (36)
15x15
t = 3.18 cm
basement level heights 3.65 m (12′-0″);

W24x335
2nd (25) (26) (27) (28) (29) (30)

Ground level height 5.49 m (18′-0″);
1st– 19th level heights 3.96 m (13′-0″);
1st (19) (20) (21) (22) (23) (24)
bay widths (all) 6.10 m (20′-0″).
t = 5.08 cm
15x15
Seismic Mass:
W24x335
Ground (13) (14) (15) (16) (17) (18) including steel framing, for both N-S MRFs;
Ground level 5.32×105 kg;
B-1
5.63×105 kg;
(7) (8) (9) (10) (12)
W30x99 (11)
1st level
B-2 (1) (2) (3) (4) (5) (6)
2nd –19th level 5.52×105 kg;
20th level 5.84×105 kg.
entire structure (above ground) 1.11×107 kg.
page 7
Evaluation Models
This benchmark study will focus on an in-plane (2-D) analysis of the benchmark structures.
The frames considered in the development of the evaluation models are the N-S MRFs (the short,
or weak, direction of the buildings) for the structures described in the previous section. Passive,
active and/or semi-active control devices can be implemented throughout these N-S frames of the
3-, 9- and 20-story structures and their performance assessed using the evaluation models in this
section and the evaluation criteria identified in the Control Design section.
Based on the physical description of the 3-, 9- and 20-story structures described in the previ-
ous section, in-plane finite element models of the N-S MRFs have been developed. Structural
member nonlinearities are included to capture the inelastic behavior of buildings during strong
earthquakes. The beams and columns of the structures are modeled as plane-frame elements, and
mass and stiffness matrices for each of the structures are determined. A bilinear hysteresis model
is used to characterize the nonlinear bending stiffness of the structural members. The damping
matrix is determined based on an assumption of Rayleigh damping. This process is described in
further detail in the following paragraphs.
Nodes are located at beam-to-column joints. Elements are created between nodes to repre-
sent the beams and columns in the structure. The beam members extend from the center-line of
column to center-line of column, thus ignoring the column panel zone. Inertial loads, accounting
for the seismic mass of the floor slabs, ceiling/flooring, mechanical/electrical, partitions, roofing
and penthouse are uniformly distributed at the nodes of each respective level assuming a lumped
mass formulation.
The 9- and 20-story building frames contain column splices. The column joint of the splice
story is located 1.83 m (6 ft) above the center-line of the beam. For simplicity the spliced columns
are modeled as having uniform properties over the story height equal to the weighted average of
the upper and lower column properties of that story. There is no node modeled at the splice.
Each node has three degrees-of-freedom (DOFs): horizontal, vertical and rotational. The 3-,
9- and 20-story structures have 60, 198, and 414 DOFs prior to application of boundary condi-
tions/constraints, respectively. Global DOF n is the p-th local DOF (i.e., horizontal: p=1, vertical:
p=2, rotational: p=3) of the q-th node, which is indicated in the figures of the previous section,
and is given by n = 3 ( q – 1 ) + p .
Each element, modeled as a plane frame element, contains two nodes and six DOFs. The
length, area, moment of inertia, modulus of elasticity and mass density are pre-defined for each
element. The elemental lumped mass and stiffness matrices are determined as functions of these
properties (Sack 1989; Cook, et al. 1989). Global mass and stiffness matrices are assembled from
the elemental mass and stiffness matrices by summing the mass and stiffness associated with each
DOF for each element of the entire structure. Rotational inertia is ignored; thus, rotational mass is
assigned a small value. The DOFs corresponding to fixed boundary conditions are then con-
strained by eliminating the rows and columns associated with these DOFs from the global mass
and stiffness matrices.
The simulation of the benchmark buildings is developed to represent control of the entire
structure, including both N-S MRFs and the entire mass of each structure. Researchers/designers
should recognize that the control strategies applied within this study represent the structural con-
trol of the entire benchmark building being considered.
The first three natural frequencies of the 3-story benchmark evaluation model are: 0.99,
3.06, and 5.83 Hz. The first five natural frequencies of the 9-story benchmark evaluation model
are: 0.443, 1.18, 2.05, 3.09, and 4.27 Hz. The first 10 natural frequencies of the 20-story bench-
page 8
mark evaluation model are: 0.261, 0.753, 1.30, 1.83, 2.40, 2.44, 2.92, 3.01, 3.63 and 3.68 Hz.
These results are consistent with those found by others who have modeled this structure (Barrosa
1999; Spencer, et al. 1998).
Assuming the structures respond in the elastic range, transfer functions for the displacement
and absolute acceleration at the top of each building from ground acceleration can be determined.
These transfer functions are provided in Figure 4. The first three mode shapes for the 3-, 9- and
20-story building models are shown in Figure 5.
50
Magnitude (dB)
−50
Acceleration
Displacement
−100
−1 0 1
10 10 10
frequency (Hz)
(a) 3-Story Building Model
50
Magnitude (dB)
−50 Acceleration
Displacement
−100
−1 0 1
10 10 10
frequency (Hz)
(b) 9-Story Building Model
50
Magnitude (dB)
−50
Acceleration
Displacement
−100
−1 0 1
10 10 10
frequency (Hz)
(c) 20-Story Building Model
Figure 4. Typical Transfer Functions from Ground Acceleration to Absolute Roof
Acceleration (Solid) and Ground Acceleration to Relative Horizontal Roof Displacement
(Dashed) for the (a) 3-Story; (b) 9-Story; and (c) 20-Story Building Models.
page 9
Mode 1 (0.99 Hz) Mode 2 (3.06 Hz) Mode 3 (5.83 Hz)
(a) 3-Story Building Model

(b) 9-Story Building Model

(c) 20-Story Building Model
Figure 5. Mode Shapes for the (a) 3-Story; (b) 9-Story; and (c) 20-Story Building Models.
page 10
Nonlinear Analysis Tool
During large seismic events, structural members force
can yield, resulting in nonlinear response behavior that (moment)
may be significantly different than a linear approxima-
EI2
tion. To better represent the nonlinear behavior, a bilin-
ear hysteresis model, as shown in Figure 6, is used to
model the plastic hinges, the points of yielding, of the EI1 deformation
3-, 9- and 20-story building structural members. The (curvature)
bilinear bending properties are predefined for each d
structural member. These plastic hinges, which are
assumed to occur at the moment resisting column-
beam and column-column connections, introduce a
material nonlinear behavior of these structures. While Figure 6. Bilinear Hysteresis Model
MATLAB® (1997) has a wide range of numerical tools for Structural Member Bending.
for control design and analysis, it does not have built-in
integration schemes that can effectively determine the time domain response of such nonlinear
structures. Therefore, an efficient implementation of the Newmark- β time-step integration
method was developed in MATLAB for this purpose (Ohtori and Spencer 1999). This section dis-
cusses the nonlinear algorithm and its implementation.
The Newmark- β method, which is reviewed in detail by Subbaraj and Dokainish (1989), is
used to solve the incremental equations of motion. The incremental equations of motion for the
nonlinear structural system take the following form
M∆U̇˙ + C∆U̇ + K∆U = – MG∆ẋ˙g + P∆f + ∆F err (1)
where, M , C , and K are the mass, damping and stiffness matrices of the building, ∆U is the
incremental response vector, G is a loading vector for the ground acceleration, ∆ẋ˙g is the ground
acceleration increment, P is a loading vector for the control forces, ∆f is the incremental control
force and ∆F err is the vector of the unbalanced forces. The unbalanced force is the difference
between the restoring force evaluated using the hysteresis model and the restoring force assuming
constant linear stiffness at time t during the time interval ( t, t + ∆t ). This unbalanced force is
handled at the next time step. That is, the unbalanced force is added as external force at the next
time step.
Because the floor slab is assumed to be rigid in the horizontal plane, the nodes associated
with each level have the same horizontal displacements. Therefore, the dependent (slave) horizon-
tal DOFs on each floor slab can be expressed in terms of a single active horizontal DOF. That is,
the displacements ∆U can be expressed using those of the active nodes, ∆U act , that include all
vertical, all rotational and one horizontal DOF per level. This relation can be written as
 ∆U act 
∆U =   = T R ∆U act (2)
 ∆U slv 
in which T R is a transformation matrix for expressing the full response vector in terms of the
active degrees of freedom.
page 11
The reduced damping matrix, Ĉ , can now be determined using an assumption of Rayleigh
damping as
T T
Ĉ = c 1 ⋅ T R MT R + c 2 ⋅ T R KT R (3)
where c 1 and c 2 are chosen such that modal damping coefficients ζ 1 = ζ 5 = 0.02 . The damping
in each mode, according to Rayleigh damping, is given by
2 2
ζi = ζ1 ( ω1 ω5 + ωi ) ⁄ ( ωi ω1 + ω5 ) (4)
where ω i is the natural frequency of the i-th mode. A plot of the damping coefficients in the
respective modes for the 20-story building model is shown in Figure 7.
The following two expressions, proposed by Newmark (1959), are employed to solve the
incremental equations of motion in Eq. (1):
U t + ∆t = U t + ∆tU̇ t + ( ∆t )  --- – β U̇˙t + βU̇˙t + ∆t

2 1
(5)
2 
U̇ t + ∆t = U̇ t + ∆t [ ( 1 – γ )U̇˙t + γ U̇˙t + ∆t ] (6)
in which ∆t is the calculation time interval, { } t + ∆t and { } t are the responses at t + ∆t and t ,
respectively, and β and γ are the Newmark parameters. When β = 1 ⁄ 4 and γ = 1 ⁄ 2 , the cal-
culations are unconditionally stable. Substituting Eq. (5) and Eq. (6) into Eq. (1), the following
equations can be derived:
T T
T R K D T R ∆U act = T R ∆F D (7)
where
0.05
0.045
0.04
0.035
Damping Coefficient
0.03
0.025
0.02
0.015
0.01
0.005
0
0 5 10 15 20 25
Circular Frequency [Rad/sec]
Figure 7. Damping Coefficients for the First 10 Modes of the 20-Story Building.
page 12
1 γ
K D = ---------------2-M + --------- C + K t (8)
β ( ∆t ) β∆t
γ
∆F D = – MG∆ẋ˙g +  ------M +  ------ – 1 ∆tC U̇˙t
1
 2β  2β  
(9)
γ
+  --------- M + --- C U̇ t + P∆F + ∆F err
1
 β∆t β 
T –1 –1
C = ( T R ) ĈT R (10)
and where K t is the tangent (instantaneous) stiffness matrix of the structure at time t .
To determine the tangent stiffness, K t , a concentrated plasticity model is implemented to
model the material nonlinearity of the structural members. The concentrated plasticity model is
appropriate for steel structures such as the steel moment resisting frame structures being consid-
ered here. The spread plasticity model forms the basis for the concentrated plasticity model. The
matrix coefficients of the spread plasticity model were derived by Lobo (1994) and rewritten for
improved numerical stability by Valles, et al. (1996). These models use a flexural formulation to
derive the element stiffnesses of the nonlinear elements. With the concentrated plasticity model
(Clough and Johnson 1966; Giberson 1967), the members are assumed to be elastic and yielding
occurs only at the ends. The parameters α AA and α BB are introduced so as to express the inelastic
spring at the end of each member. For the concentrated plasticity model, the following stiffness
matrix is used for expressing the material nonlinearity
–1
MA  f AA f AB  θA 
  =   (11)
 MB  f BA f BB  θB 
In the Eq. (11), the flexibility coefficients of the elements are:

L 1
f AA = ----------------------------------- f AA ′ + ---------------- (12)
12EI 0 EI A EI B ( GA )L
L 1
f AB = f BA = ----------------------------------- f AB ′ + ---------------- (13)
L 1
f BB = ----------------------------------- f BB ′ + ---------------- (14)
where
f AA ′ = 4EI A EI B + ( EI 0 – EI A )EI B α AA (15)
f AB ′ = – 2E I A EI B (16)
f BB ′ = 4EI A EI B + ( EI 0 – EI B )EI A α BB (17)
page 13
in which M A and M B are the moments at the ends A and B of the element, respectively; θ A and
θ B are the rotations at the ends A and B; EI A and EI B are flexural stiffness at the ends of the
members at A and B, respectively; EI 0 is the center stiffness of the member; GA is the shear
stiffness of the member; and L is the length of the member.
MATLAB Implementation of Nonlinear Analysis

The nonlinear evaluation model requires a MATLAB-based program, implemented as a SIM-
®
ULINK system function (S-function), to calculate the nonlinear response of the model. This
S-function performs the nonlinear dynamic analysis using the previously discussed Newmark- β
method. The accuracy of the nonlinear evaluation model was verified with the commercial struc-
tural analysis program IDARC2D (Ohtori and Spencer 1999).
Control Design
The task of the researchers/designers in the benchmark study control design problems is to
define an appropriate passive, active, or semi-active control strategy, or a combination thereof. It
is left to the researchers/designers to define the type, appropriate model and location of the control
device(s)/sensor(s) and to develop appropriate control algorithms. The evaluation model, how-
ever, will remain invariant to the various applied control strategies. By using a specific building
model and common evaluation criteria, various control strategies can be compared directly, fulfill-
ing the purpose of these benchmark problems. In this section an overview of the control design
process, with respect to the task of the researcher/designer, is offered, and the interface with the
various models of the simulation are identified.
Overview of Researcher/Designer Tasks

As part of the MATLAB-based nonlinear benchmark study, four programs and four data sets
are employed directly. The data flow between the MATLAB-based programs and data sets is shown
in Figure 8. Shaded tasks are predefined and remain invariant to the different control strategies,
clear tasks are to be defined by each researcher/designer. Each proposed control strategy requires
the researcher/designer to define an input/output data file to identify the input location of control
forces onto the evaluation model and the type and location of output responses of the evaluation
model for evaluation, measurement and connectivity purposes. Additionally, the researcher/
designer must define sensor and controller models for their proposed control strategy. Each pro-
posed control strategy is applied to a nonlinear evaluation model and simulated for seismic excita-
tion. The simulator for the nonlinear evaluation model is illustrated in Figure 9. To run the
nonlinear evaluation model simulation, the researcher/designer must include SIMULINK blocks for
their sensor(s) and controller/control device(s). The final task of each researcher/designer is to
evaluate their proposed control strategy using the evaluation criteria discussed in the following
Evaluation Criteria section.
page 14
Structural Data Input/Output Data Ground Accel Data Uncontrolled Data
Build
Building Model
Run Evaluate Evaluation
Simulation Strategy Criteria
Define
Control Strategy
Figure 8. Schematics of the Nonlinear Benchmark Evaluation Model.

(Shaded Tasks are Predefined; White Tasks are Defined by Researcher/Designer)
Benchmark Control Problems

for Seismically Excited Nonlinear Buildings
by
Y. Ohtori, R.E. Christenson, B.F. Spencer
and S.J. Dyke
El Centro
Hachinohe
ye ye
Northridge −K− xddg
ym
Kobe
intensity
f
Ground Accel yc
Nonlinear Evaluation
Model
t_out
Clock
f yc
yf yf u u ys ys ym
Control Devices Controller Sensors
Figure 9. SIMULINK Block Diagram for Vibration Control Simulator.
page 15
Interfacing with the Evaluation Model
The evaluation model for the 3-, 9- and 20-story structures takes the general form
ẋ = g 1 ( x, f , ẋ˙g ) (18)
y m = g 2 ( x, f , ẋ˙g ) (19)
y e = g 3 ( x, f , ẋ˙g ) (20)
y c = g 4 ( x, f , ẋ˙g ) (21)
where x is the state vector, y m is the vector corresponding to the measured outputs, y e is the vec-
tor corresponding to the building response evaluation outputs that are used for evaluation of the
system, and y c is the vector of output responses that are used as inputs to control device models.
The specific response quantities considered in y m , y e , and y c should be specified in the user
defined data file for outputs. It should be noted here that information regarding the building dam-
age is saved directly by the nonlinear evaluation model and need not be specified in the vector y e .
Equation (18), in general a nonlinear equation, is solved using the Newmark- β method and the
nonlinear evaluation model described previously.
Interfacing with the Sensor Models

For active/semi-active control strategies to interface with the benchmark building model
defined in Eqs. (18)-(21), the measured outputs of the evaluation model must be measured by sen-
sors. Researchers/designers should develop models for the sensors which take the form
s s
ẋ = g 5 ( x , y m, t ) (22)
s s
y = g 6 ( x , y m , v, t ) (23)
s
where x is the continuous time state vector of the sensor(s), v is a measurement noise vector and
s
y is the continuous time output of the sensor(s). All measured responses have units of Volts.
Interfacing with the Controller

For active/semi-active control systems, the corresponding controller is required to take the
form
c c s
x k + 1 = g 7 ( x k, y k, k ) (24)
c s
u k = g 8 ( x k, y k, k ) (25)
page 16
c s
where x k is the discrete state vector of the control algorithm at time t = kT , y k is the sampled
input to the control algorithm (discretized measured output from the sensor model), and u k is the
discrete control command from the control algorithm.
Interfacing with the Control Device Models

For this initial nonlinear benchmark problem, ideal control actuators may be assumed. The
control force and device information would take the form
f = g 9 ( y c, u , t ) (26)
y f = g 10 ( y c, u, t ) (27)
where f is the continuous time force output (in Newtons) of the control device(s) applied to the
structure, y f is the vector of control device responses used for evaluation purposes and u is the
continuous controller output (sample and hold of u k ) in Volts.
Evaluation results should be presented for a control strategy using idealized actuators. How-
ever, if a researcher/designer wants to include actuator dynamics, to interface with the benchmark
building model the control device model(s) should take the form
a a
ẋ = g 11 ( x , y c, u, t ) (28)
a
f = g 9 ( x , y c , u, t ) (29)
a
y f = g 10 ( x , y c, u, t ) (30)
a
where x is the continuous time state vector of the control device. Note that for passive control
devices, g 9 , g 10 and (if applicable) g 11 will not be functions of u .
Evaluation Criteria
In order to evaluate proposed control strategies, two far-field and two near-field historical
records are selected: (i) El Centro. The N-S component recorded at the Imperial Valley Irrigation
District substation in El Centro, California, during the Imperial Valley, California earthquake of
May 18, 1940. (ii) Hachinohe. The N-S component recorded at Hachinohe City during the Toka-
chi-oki earthquake of May 16, 1968. (iii) Northridge. The N-S component recorded at Sylmar
County Hospital parking lot in Sylmar, California, during the Northridge, California earthquake
of January 17, 1994. (iv) Kobe. The N-S component recorded at the Kobe Japanese Meteorologi-
cal Agency (JMA) station during the Hyogo-ken Nanbu earthquake of January 17, 1995. The
absolute peak acceleration of the earthquake records are 3.417, 2.250, 8.2676, and 8.1782 m/sec2,
respectively. The earthquake records are shown in Figure 10.
Additionally, this benchmark study will consider various levels of each of the earthquake
records including: 0.5, 1.0 and 1.5 times the magnitude of El Centro and Hachinohe; and 0.5 and
1.0 times the magnitude of Northridge and Kobe. This is a total of 10 earthquake records to be
considered in the evaluation of each control strategy.
page 17
5 3
Acc. [m/sec2]
Acc. [m/sec ]
2
0 0
−1
−2
−5 −3
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35
Time [sec.] Time [sec.]
(a) El Centro (b) Hachinohe
10 10
Acc. [m/sec ]
5
Acc. [m/sec ]
2
2
0 0
−5 −5
−10 −10
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time [sec.] Time [sec.]
(c) Northridge (d) Kobe

Figure 10. Time Histories of the Near and Far Field Historical
Earthquake Records Used in the Benchmark Study:
(a) El Centro; (b) Hachinohe; (c) Northridge; and (d) Kobe.
The evaluation criteria are divided into four categories: building responses, building dam-
age, control devices, and control strategy requirements. The first three categories have both peak
and normed based criteria. Small values of the evaluation criteria are generally more desirable.
The first category of the evaluation criteria is related to the building responses. The first
three criteria are based on peak interstory drift ratio ( J 1 ), level acceleration ( J 2 ) and base shear
( J 3 ):
 di(t ) 
 max -------------- 
J 1 = Elmax  t, i hi  (31)
Hachinohe 
-
Centro ------------------------------
max
Northridge
Kobe  δ 
 max ẋ˙ai ( t ) 
J 2 = Elmax  t, i
Centro --------------------------------
 (32)
Hachinohe  max 
Northridge ẋ
˙ a
Kobe
 
 max ∑ m i ẋ˙ai ( t ) 
J 3 = Elmax  t i  (33)
Hachinohe  ------------------------------------------ 
Centro
max
Kobe  
Northridge
Fb
over the range i = [ 1, 3 ] , i = [ 1, 9 ] and i = [ 1, 20 ] for the 3-, 9- and 20-story buildings
respectively (includes only above ground levels), where d i ( t ) is the interstory drift of the above
ground level over the time history of each earthquake, h i is the height of each of the associated
page 18
max
stories, δ is the maximum interstory drift ratio of the uncontrolled structure calculated by
max
equation of max d i ( t ) ⁄ h i , ẋ˙ai ( t ) and ẋ˙a are absolute acceleration of the i-th level with and
t, i
without control devices respectively, m i is the seismic mass of the i-th above ground level and
max
F b is the maximum base shear of the uncontrolled structure for each respective earthquake.
The next three criteria are based on normed building responses. The interstory drift ( J 4 ),
level acceleration ( J 5 ), and base shear ( J 6 ) are defined in their normed based forms as
 di(t ) 
 max ---------------- 
J 4 = Elmax  i hi  (34)
Hachinohe 
-
Centro ----------------------------
max
Kobe 
Northridge
δ 
 max ẋ˙ai ( t ) 
J 5 = Elmax  i -
Centro ------------------------------
(35)
Northridge ẋ˙ a
Kobe
 
 ∑ m i ẋ˙ai ( t ) 
J 6 = Elmax  i  (36)
Hachinohe  ------------------------------- 
Centro
max
Kobe  
Northridge
Fb
where the norm, . , is computed using the following equation
1 tf 2
t f ∫0 ˙
- [ ] dt
. ≡ --- (37)
and t f is a sufficiently large time to allow the response of the structure to attenuate. In this bench-
mark study the duration of 100 sec is adopted for the El Centro, Hachinohe, and Northridge earth-
max
quakes and 180 sec for the Kobe earthquake. The normed values for δ ≡ max d i ( t ) ⁄ h i ,
max max i
ẋ˙a , and F b are the maximum normed interstory drift ratio, absolute acceleration and
base shear force corresponding to the uncontrolled structure excited by each respective earth-
quake.
The second category of the evaluation criteria assesses the building damage. These criteria
have been added because of the nonlinear character of this benchmark study. Both ends of each
element are considered in these criteria to assess the yielding, excluding simply supported beam
elements (as well as the fixed-hinged beam elements in the 9-story structure). The seventh and
eighth evaluation criteria are based on peak responses while the ninth and tenth are normed-based
criteria. The evaluation criteria for the ductility factor ( J 7 ) and dissipated energy of the curvatures
at the end of members ( J 8 ) are defined as
page 19
 φ j(t ) 
 max --------------- 
J 7 = Elmax  t, j φ yj  (38)
Hachinohe 
-
Centro -------------------------------
max
Kobe 
Northridge
φ 
 
 ∫ dE j 
 max ------------------- 
J 8 = Elmax  t, j F yj ⋅ φ yj  (39)
Hachinohe 
-
Centro -----------------------------------
max
Kobe  
Northridge
E
 
where φ j is the curvature at the ends of the j-th element (member), ∫ dE j is the dissipated energy
at the ends of the member during the respective earthquake, φ yj and F yj are the yield curvature
max max
and the yield moment at the end of the j-th member, respectively, and φ and E are the
maximum curvatures and the maximum dissipated energy (maximum of all element ends and over
time) of uncontrolled structure.
The ninth evaluation criterion ( J 9 ) is the ratio of the plastic connections sustained by the
structure while controlled and uncontrolled and is given by
C
Nd 
J 9 = Elmax  ------
N d 
- (40)
Hachinohe 
Centro
Northridge
Kobe
C
where N d is the number of damaged connections (member ends) without control and N d are the
number of damaged connections with control. Evaluation criteria J 8 and J 9 only have meaning
for structures undergoing plastic deformations and are, therefore, undefined (should not be calcu-
lated or reported) when the uncontrolled building remains elastic.
The tenth evaluation criterion ( J 10 ) is the normed ductility factor and is given by
 φ j(t ) 
 max ---------------- 
J 10 = Elmax  j φ yj  (41)
Hachinohe 
-
Centro -----------------------------
max
Kobe 
Northridge
φ 
where φ j ( t ) and φ
max
are the normed curvature at the member ends with and without control,
respectively. The curvature ( φ j ), normed curvature ( φ j ( t ) ), dissipated energy ( ∫ dE j ) and infor-
mation to determine the number of plastic hinges ( N d ) are determined within the nonlinear S-
function during the simulation and saved to a specified data file.
The third category is related to the control devices. This category assess the required perfor-
mance of the devices. Peak criterion J 11 , J 12 , and J 13 show control force, control device stroke,
and power used for control
page 20
 max f l ( t ) 
J 11 = Elmax  t, l -
Centro ------------------------------
(42)
Hachinohe  W 
Northridge
Kobe
a
 max y l ( t ) 
J 12 = Elmax  t, l
Centro -------------------------------
 (43)
Northridge x
Kobe
 max ∑ P l ( t ) 
 t l  (44)
J 13 = Elmax 
Centro -------------------------------------

Hachinohe  max 
ẋ W
Kobe  
Northridge
where f l ( t ) is the force generated by the l-th control device over the time history of each earth-
quake, W is the seismic weight of the building based on the above ground mass of the structure
a
(excluding the mass below or at the ground level), y l ( t ) is the displacement across the l-th control
max
device during the earthquake, x is the maximum uncontrolled displacement of the levels rela-
tive to the ground, P l ( t ) is a measure of the instantaneous power required by the l-th control
max
device and ẋ is the maximum uncontrolled velocity of the levels relative to the ground. For
a a
active control devices P l ( t ) ≡ ẏ l ( t ) f l ( t ) , where ẏ l ( t ) is the velocity across the l-th control
device. For semi-active devices P l ( t ) is the actual power required to operate the device. For the
passive control devices the power required is zero.
The fourteenth evaluation criterion ( J 14 ) is a measure of the total power required for the
control of the structure and is defined as
tf
 1 
 ∑ --- t
- ∫ P l(t ) 
 l f0 
J 14 = Elmax  ---------------------------
max
-
(45)
Hachinohe  ẋ W 
Centro
Kobe  
Northridge
 
With passive control devices, the J 14 criterion is zero.

The last category for evaluating the performance is related to the control strategy of each
method. The fifteenth ( J 15 ), sixteenth ( J 16 ) and seventeenth ( J 17 ) evaluation criteria are
J 15 = Number of control devices (46)
J 16 = Number of required sensors (47)
c
J 17 = dim ( x k ) (48)
page 21
The J 15 criterion is the total number of control devices implemented to control the bench-
mark building. The J 16 criterion is the total number of control sensors used for the control strat-
egy. The J 17 criterion assesses the computational resources required, which is represented by the
c
dimension of the discrete state vector x k , required for the control algorithm.
A summary of the evaluation criteria is presented in Table 1. The maximum uncontrolled
responses required to calculate the evaluation criteria are given Tables 3-5, located in the Appen-
dix A. All seventeen criteria should be reported for each proposed control strategy. The various
levels of the respective earthquakes should each be considered in determining the evaluation crite-
ria.
Researchers/designers are encouraged to include other criteria in their results if, through
these criteria, important features of their control strategy can be demonstrated.
Table 1: Summary of Evaluation Criteria for the Nonlinear Benchmark Problem.
Interstory Drift Ratio Level Acceleration Base Shear

 di(t )   max ẋ˙a i ( t )   
J 3 = max  t ∑ i ai
 max ------------- - t, i
J 2 = max  --------------------------------
-  max m ẋ˙ ( t ) 
J 1 = max  t, i hi  El Centro  max  
El Centro  -------------------------------  Hachinohe ẋ˙a El Centro  i
-
Northridge Hachinohe ------------------------------------------
Northridge   
Hachinohe max
δ Kobe Northridge Fb
max
Kobe Kobe
Normed Interstory Drift Ratio Normed Level Acceleration Normed Base Shear
 di(t )   max ẋ˙a i ( t )   
J 6 = max  ∑ i ai
 max --------------- - J 5 = max  ------------------------------  m ẋ˙ ( t ) 
-
i
J 4 = max  i hi  El Centro  max  
El Centro  -----------------------------  Hachinohe ẋ˙a El Centro  i 
Hachinohe -------------------------------
Northridge 
Northridge
Northridge 
Hachinohe max max
δ Kobe Fb
Kobe Kobe
Ductility Dissipated Energy Plastic Connections

  C
 φ j(t )   Nd 
 max -------------- - ∫ dE j 
 max ------------------- J 9 = max  ------- 
J 7 = max  t, j φ yj  J 8 = max  t, j F yj ⋅ φ yj  El Centro  N d 
El Centro  --------------------------------  El Centro  ------------------------------------- 
Hachinohe
Northridge
Northridge 
Hachinohe max
φ Hachinohe 
Northridge E
max  Kobe
Kobe Kobe  
Normed Ductility Control Force Control Device Stroke

 φ j(t )   max f l ( t )  a
 max ----------------  J 11 = max  ------------------------------  max y i ( t ) 
= max  j φ yj  El Centro 
t, l
- t, i
J 12 = max  ------------------------------
-
J 10 W  El Centro  
El Centro  ------------------------------  Hachinohe Hachinohe x
max
Northridge
Northridge 
Hachinohe max
φ Kobe Northridge
Kobe
Kobe
Control Power Normed Control Power Control Devices

J 15 = Number of control devices
 max ∑ P l ( t )   tf 
 --- 
∑ t ∫ l 
 t l  1
- P ( t )
J 13 = max  --------------------------------------  Sensors
El Centro  max   l f0 
Northridge
Hachinohe ẋ W  J 14 = max  ----------------------------
max
- J 16 = Number of required sensors
Kobe El Centro  ẋ W 
Northridge 
Hachinohe
Kobe  
 
Computational Resources
c
J 17 = dim ( x k )
page 22
Control Implementation Constraints and Procedures
To make the benchmark problem as representative of the full-scale implementation as possible and
to allow for direct comparison of the results submitted to the study, the following constraints and
procedures are specified:
1. The measured outputs directly available for use in determination of the control action are the
absolute horizontal acceleration and the interstory drift of each level of the structure. Al-
though absolute velocity measurements are not available, they can be closely approximated by
passing the measured accelerations through a second-order filter similar to that described in
Spencer, et al. (1998a) but modified as appropriate for the bandwidth of the structures here.
2. The digitally implemented controller should have a sampling time between
0.001 ≤ T ≤ 0.01 seconds.
3. The A/D and D/A converters on the digital controller have 16-bit precision and a span of ± 10
Volts.
4. Each of the measured responses contains an RMS noise of 0.03 Volts, which is approximately
0.3% of the full span of the A/D converters. The measurement noises are modeled as Gaussian
rectangular pulse processes with a pulse width of 0.01 seconds.
5. No hard limit is placed on the number of states of the control algorithm, although the number
of states should be kept to a reasonable number as limited computational resources in the dig-
ital controller exist. The designer/researcher should justify that the proposed algorithm(s) can
be implemented with currently available computing hardware.
6. The control algorithm is required to be stable.
7. The performance of each control design should be evaluated for each of the earthquake
records provided (i.e., El Centro, Hachinohe, Northridge and Kobe) and each of the levels
specified for each earthquake. Performance should be evaluated assuming ideal actuators.
8. The closed loop stability robustness for each proposed active control design should be dis-
cussed.
9. If control device dynamics are included, the control signal to each control device has a con-
straint of 10 Volts for each respective earthquake; i.e., max u i ( t ) ≤ 10 V.
t, i
10. The capabilities of each control device employed in the design should be discussed, and the
designer/researcher should provide a justification of the availability of the device. Additional
constraints unique to each control scheme should also be reported (e.g., maximum displace-
ment, velocity, or force capacity of control devices). The control device model(s) should take
the form of Eqs. (26)-(27) (i.e. actuator dynamics are neglected) when determining the evalu-
ation criteria.
11. Researchers/designers should submit electronically a complete set of MATLAB files that will
produce the evaluation criteria specified in this problem statement. For more details, see the
README file included with the downloaded benchmark data on the benchmark homepage
(http://www.nd.edu/~quake/).
page 23
Sample Control System Design
To illustrate some of the constraints and challenges of this benchmark problem, a sample
control system is presented. Although the structure’s response might be nonlinear during a severe
earthquake, the sample control system is designed for the nominal linear structure. The sample
control system design is included to serve as a guide to the participants in this study and is not
intended to be a competitive design. The sample control system design will provide a structural
control strategy for the 20-story benchmark building, including both N-S MRFs of this structure.
The sample control system is a type of active system. The control actuators are located on each
story of the structure to provide forces to the building. Feedback measurements are provided by
accelerometers placed at various locations on the structure. In this section, the accelerometers
chosen for the sample control system are described and models for the accelerometers are dis-
cussed. The control actuators are assumed to be ideal (i.e., the dynamics of the actuators are
neglected) for the sample control design. A linear quadratic Gaussian (LQG) control algorithm is
designed based on a reduced order model of the system. Simulations are performed and evalua-
tion criteria are determined and recorded here. The results of the sample control system design are
discussed.
Sensors
Because accelerometers can readily provide reliable and inexpensive measurements of the
absolute accelerations at arbitrary locations on a structure, the sample control system is based on
acceleration feedback (Spencer, et al. 1994). A total of five acceleration measurements are
selected for feedback in the control system: measurements on levels 4, 8, 12, 16 and 20.
A wide variety of accelerometers are available, many with a natural frequency at least an
order of magnitude above the dynamics of this structure. Thus, each accelerometer is modeled as
having a constant magnitude and phase. The magnitude of the output of each accelerometer is 10
V / g (where 1g = 9.81 m/sec2), which is the sensitivity of the sensor. The measurement noise,
prescribed in control implementation constraint 4, is included in the sensor model. Thus, in the
form of Eqs. (22)-(23) the sensors can be modeled with g 5 = 0 and
s
y = Ds ym + v (49)
where D s = ( 10 ⁄ 9.81 )[I] V / (m/sec2). Based on the measurements selected for feedback in the
T
sample control strategy, y m = [ ẋ˙a4 ẋ˙a8 ẋ˙a12 ẋ˙a16 ẋ˙a20 ] in which the numeric subscript
denotes the level.
For simulation purposes, the sensor block shown in Figure 11 is used to represent the five
accelerometers used in the sample control system design.
Control Devices
Control actuators are placed throughout the above ground stories of the 20-story benchmark
building, connecting adjacent levels. The size of the actuators are limited to provide maximum
control forces of 1000 kN. Actuators with this capacity are readily available. To provide larger
control forces at a particular location, multiple actuators can be employed. The control actuators
for this sample control strategy are placed on each level of the structure, and a total of 25 actuators
are used to control the 20-story benchmark building. Four actuators are located on the ground
page 24
Sensor
Gains (Ds) ym
K 1
Sum
ys
1
Measurement
Noise (v)
Figure 11. SIMULINK Block Representing the Sensors in the Sample

Control System Design.
level, two actuators are located on each of the first and second levels, and one actuator is on each
of the third through 19th levels of the structure. Each actuator is implemented in the structure
using a chevron brace configuration, in which the actuator is horizontal and rigidly attached
between the two consecutive levels of the building. Thus, the actuators placed on the first level
will produce equal and opposite control forces on the first level and second level. In the analysis
the compliance of the bracing is neglected. In the form of Eqs. (26) - (27), the force vector for the
25 control devices located on 20 levels of the building and producing 39 control forces (20 equal
and opposite forces, less the force applied to the ground level) can be modeled as
f = Kf u (50)
u
yf = (51)
yc
where K f is a matrix that accounts for multiple actuators per level and loads the control devices
forces (per level) onto the building levels.
For simulation purposes, the control device block shown in Figure 12 is used to represent
the ideal actuators used in the sample control system design. Note that the force gain, K f , in Fig-
ure 12 is the gain matrix from Eq. (50) that converts the control signal to control force, accounts
for multiple actuators per level, and loads the actuator forces onto the 20 levels of the structure.
This gain block represents the idealized actuator models.
Ideal Actuator (Kf)
f u
1 K 2
Gain_ctr
yf
−K− yc
2 Mu
1
Figure 12. SIMULINK Block Representing the Control Devices in the

Sample Control System Design.
page 25
Control Design Model
Because the evaluation model is quite large, a reduced order model of the system, desig-
nated the design model, is developed for purposes of control design. This 20-state reduced order
design model of the 20-story structure is given as
d d
ẋ = A d x + B d u + E d ẋ˙g (52)
d
y sd = D s ( C md x + D md u + F md ẋ˙g ) + v (53)
d
y ed = C ed x + D ed u + F ed ẋ˙g (54)
d T
where x is the design state vector, y md = [ ẋ˙a4 ẋ˙a8 ẋ˙a12 ẋ˙a16 ẋ˙a20 ] is the vector of mea-
T
sured responses in Volts, y ed = [ ẋ˙a1 … ẋ˙a20 ] is the vector of regulated responses (lateral accel-
erations at each of the above ground levels), u is the control signal for the control force of the
individual control devices neglecting actuator dynamics (note that of the 25 actuators, only 20
independent control forces are required, with the four actuators on the ground floor all exerting
the same force, the two on the first level exerting the same force, and the two on the second level
exerting the same force), and A d , B d , E d , C md , C ed , D md , D ed , F md and F ed are the reduced
order coefficient matrices. Notice that this design only makes use of the absolute accelerations of
five levels of the structure for measured feedback, although additional measurements are available
for feedback.
Control Algorithm
To illustrate some of the challenges of this benchmark problem, a sample linear quadratic
Gaussian (LQG) control design is presented. To simplify design of the controller, ẋ˙g is taken to be
a stationary white noise, and an infinite horizon performance index is chosen that weights the
accelerations of the levels, i.e.,
τ
1  T 2
∫  ( Ced x
d d
Ĵ = lim --- E + D ed u ) Q ( C ed x + D ed u ) + Ru dt
τ→∞ τ 
0 (55)
where R is a [ 20 × 20 ] matrix with equal weighting placed on each actuator force (i.e.,
R = ( 1 ⁄ 16 ) [ I ] )and the weighting matrix Q was chosen to be a [ 20 × 20 ] matrix with equal
9
weighting placed on each of the level accelerations (i.e., Q = 3 ×10 [ I ] ). Further, the measure-
ment noises are assumed to be identically distributed, statistically independent Gaussian white
noise processes, and S ẋ˙g ẋ˙g ⁄ S vi vi = γ g = 25 .
The separation principle allows the control and estimation problems to be considered sepa-
rately, yielding a control law of the form (Stengel 1986; Skelton 1988)
d
u = – K x̂ (56)
˜
page 26
d
where x̂ is the Kalman Filter estimate of the state vector based on the reduced order design
model, including the actuator models. By the certainty equivalence principle (Stengel 1986; Skel-
ton 1988), K is the full state feedback gain matrix for the deterministic regulator problem given
by ˜
–1 T
K = R̃ ( Ñ + B d P ) (57)
˜ ˜
where P is the solution of the algebraic Riccati equation given by
˜
T –1 T
0 = P Ã + Ã P – P B d R̃ B d P + Q̃ (58)
˜ ˜ ˜ ˜
and
T –1 T
Q̃ = C ed QC ed – ÑR̃ Ñ (59)
T
Ñ = C ed QD ed (60)
T
R̃ = R + D ed QD ed (61)
–1 T
Ã = A d – B d R̃ Ñ (62)
Calculations to determine K were done using the MATLAB routine lqry.m within the control tool-
box. ˜
The Kalman Filter optimal estimator is given by
d
x̂˙ = A d x̂ + B d u + L ( y s – C md x̂ – D md u )
d d
(63)
–1 T T
L = [R ( γ g F md E d + C md S )] (64)
˜
where S is the solution of the algebraic Riccati equation given by
T
0 = SA + A S – SG S + H (65)
˜ ˜ ˜ ˜
and
T T –1 T
A = A d – C md R ( γ g F md E d ) (66)
˜ ˜
T –1
G = C md R C md (67)
˜ ˜
T 2 T –1 T
H = γ g E d E d – γ g E d F md R F md E d (68)
˜ ˜
T
R = I + γ g F md F md (69)
˜
page 27
Calculations to determine L were done using the MATLAB routine lqe.m within the control tool-
box.
Finally, the controller is put in the form of Eqs. (24)–(25) using the bilinear transformation
(Antoniou, 1993) to yield the following compensator
c c s
xk + 1 = Ac xk + Bc yk (70)
˜ ˜
c s
uk = Cc xk + Dc yk . (71)
˜ ˜
Calculations to determine the discrete time compensator were performed in MATLAB using the
c2dm.m routine within the control toolbox. The control algorithm is implemented in simulation
with the block illustrated in Figure 13. The A/D and D/A converter blocks are comprised of a sat-
uration block in series with a quantizer, enforcing control implementation constraints 3 and 9.
u ys
y(n)=Cx(n)+Du(n)
1 u(t) u(k) y(k) y(t) 1
x(n+1)=Ax(n)+Bu(n)
D/A Converter A/D Converter
Discrete Controller
Figure 13. SIMULINK Block Representing the Control Algorithm

in the Sample Control System Design.
Evaluation of Sample Control Design

The nonlinear 20-story benchmark model controlled with the linear sample control strategy
described is evaluated. The sample control strategy is applied to the 20-story benchmark model.
The responses of the controlled structure are determined for each of the 10 earthquake records
using the MATLAB simulation discussed herein. Time histories of the acceleration response of the
20th level of the building for uncontrolled and controlled cases are shown in Figure 14 along with
the maximum absolute acceleration and maximum drift ratio profiles over the height of each
building. Only the original earthquake magnitudes are presented in this figure. It is clear from the
time histories that the control strategy adds significant damping to the building system as well as
provides moderate reduction of the maximum accelerations and drift ratios.
The structure’s responses are used to calculate the performance indices (evaluation criteria)
identified in this paper in order to discuss the performance. The performance of the sample con-
trol design is reported in Table 2. In this sample control strategy, the floor accelerations, both peak
and normalized, are reduced substantially from the uncontrolled cases. The drift evaluation crite-
ria are also reduced slightly, though the base shear criteria remain unaffected. The control force
evaluation criteria are reasonable. The building damage evaluation criteria show overall improve-
ment. The amount of energy dissipated through structural yielding is reduced substantially. The
control device criteria appear reasonable. The number of sensors and control devices are also rea-
sonable as well as the computational resources required to implement to 20-state controller. The
sample control strategy provides adequate reduction of the evaluation criteria; however, let it be
page 28
6 4
Absolute Acceleration
of the Roof (m/sec2)

4 (a) El Centro 3
(b) Hachinohe
2
2
1
0 0
−1
−2
−2
−4
−3
−6 −4
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time (sec) Time (sec)

20 20 20 20
18 18 18 18
16 16 16 16
Above Ground Level
Above Ground Level

14 14 14 14
12 12 12 12
10 10 10 10
8 8 8 8
6 6 6 6
4 4 4 4
2 2 2 2
0 0 0 0
0 1 2 3 4 5 0 2 4 6 8 0 1 2 3 4 5 0 2 4 6 8
−3 −3
x 10 x 10
Max. Accel. (m/s2) Max. Drift Ratio Max. Accel. (m/s2) Max. Drift Ratio
10
8
8
6 (c) Northridge 6
(d) Kobe
4
4
2 2
0 0
−2 −2
−4
−4
−6
−6
−8
−8
−10
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time (sec) Time (sec)

20 20 20 20
18 18 18 18
16 16 16 16
Above Ground Level
Above Ground Level
14 14 14 14
12 12 12 12
10 10 10 10
8 8 8 8
6 6 6 6
4 4 4 4
2 2 2 2
0 0 0 0
0 2 4 6 8 10 0 0.005 0.01 0.015 0.02 0 2 4 6 8 10 0 0.005 0.01 0.015 0.02 0.025
Max. Accel. (m/s2) Max. Drift Ratio Max. Accel. (m/s2) Max. Drift Ratio
Figure 14. Uncontrolled (Dotted) and Controlled (Solid) Responses:

(a) El Centro; (b) Hachinohe; (c) Northridge; and (d) Kobe.
page 29
Table 2: Earthquake Evaluation Criteria for the Sample Control Strategy.
Earthquake El Centro Hachinohe Northridge Kobe

Max Value
(intensity) (0.5/1.0/1.5) (0.5/1.0/1.5) (0.5/1.0) (0.5/1.0)
7.4732e-01 8.8310e-01
J1 8.5857e-01 8.1611e-01
7.4831e-01 8.8697e-01 9.4211e-01
Peak Drift Ratio 9.4211e-01 7.2778e-01
7.4762e-01 9.0656e-01
6.4758e-01 7.4578e-01
J2 8.0678e-01 7.0209e-01
6.4609e-01 7.4292e-01 9.0432e-01
Peak Level Accel. 9.0432e-01 8.3914e-01
6.6412e-01 8.3269e-01
7.8004e-01 9.7714e-01
J3 8.8472e-01 9.2500e-01
7.8198e-01 9.8191e-01 1.0656e+00
Peak Base Shear 9.6937e-01 1.0656e+00
9.0885e-01 1.0091e+00
6.6243e-01 8.8464e-01
J4 7.2391e-01 6.4772e-01
6.6287e-01 8.8399e-01 9.2948e-01
Norm Drift Ratio 9.2948e-01 2.3033e-01
6.6964e-01 9.0272e-01
5.6300e-01 6.5815e-01
J5 5.9233e-01 5.7887e-01
5.5989e-01 6.5162e-01 7.1293e-01
Norm Level Accel 6.3662e-01 7.1293e-01
5.7763e-01 6.6149e-01
7.2441e-01 8.4922e-01
J6 7.7585e-01 6.8868e-01
7.2328e-01 8.4788e-01 8.5766e-01
Norm Base Shear 8.4075e-01 8.4043e-01
7.2916e-01 8.5766e-01
7.7193e-01 9.5539e-01
J7 7.2787e-01 6.8769e-01
7.7294e-01 9.5920e-01 9.7763e-01
Ductility 9.7763e-01 6.8796e-01
7.2222e-01 9.4328e-01
-- --
J8 2.2047e-01 1.4443e-01
-- -- 7.1441e-01
Dissipated Energy 5.4786e-01 3.2294e-01
7.8186e-02 7.1441e-01
-- --
J9 5.4167e-01 3.0769e-01
-- -- 9.0625e-01
Plastic Connections 9.0625e-01 8.0952e-01
3.7209e-01 7.9070e-01
7.3335e-01 8.4741e-01
J 10 6.3239e-01 7.7728e-01
7.3344e-01 8.4703e-01 9.4374e-01
Norm Ductility 9.4374e-01 2.2667e-01
6.5577e-01 8.8958e-01
1.6809e-03 1.9219e-03
J 11 6.7064e-03 4.9702e-03
3.3460e-03 3.5460e-03 8.9063e-03
Control Force 8.1814e-03 8.9063e-03
5.0115e-03 5.0604e-03
page 30
Table 2: Earthquake Evaluation Criteria for the Sample Control Strategy.

Max Value
(intensity) (0.5/1.0/1.5) (0.5/1.0/1.5) (0.5/1.0) (0.5/1.0)
7.2168e-02 7.5420e-02
J 12 7.8181e-02 1.2611e-01
7.2060e-02 7.5770e-02 1.2611e-01
Device Stroke 1.0307e-01 1.1417e-01
7.2510e-02 8.0511e-02
1.2680e-03 8.2600e-04
J 13 4.0316e-03 4.0289e-03
2.4704e-03 1.6910e-03 8.8103e-03
Control Power 5.1187e-03 8.8103e-03
3.8787e-03 2.5968e-03
5.4943e-05 4.0700e-05
J 14 7.9195e-05 7.6991e-05
1.0279e-04 7.3444e-05 1.6110e-04
Norm Control Power 1.2176e-04 1.4546e-04
1.6110e-04 1.1333e-04
J 15
25 25
Control Devices
J 16
5 5
Sensors
J 17
20 20
Comput’l Resources
stated again that this control strategy is intended to provide an example of the procedure for this
benchmark problem and is not a competitive design.
Closure
The models and data for the nonlinear benchmark control problem for seismically excited
buildings are available in a set of MATLAB files. Included are scripts which build the evaluation
models of the 3-, 9- and 20-story buildings, perform the sample control design and run the simula-
tion. These files are available on the World Wide Web at the following URL:
http://www.nd.edu/~quake/
If you cannot access the World Wide Web or have questions regarding the benchmark problem,
please contact the senior author via e-mail at: [email protected].
To increase the value of this effort to the community, participants in the nonlinear bench-
mark study are requested to submit their control designs electronically for inclusion on the bench-
mark homepage cited previously. This electronic submission should be in the form of an m-file
script and the associated data that, when run, produces the evaluation criteria used defined in the
problem statement. The file Eval_NLBM.m included with the sample control strategy is an exam-
ple of the required evaluation m-file script. See the README file included with the downloaded
benchmark data for more details.
page 31
Acknowledgements
The authors gratefully acknowledge the partial support of this research by the National Sci-
ence Foundation under grants No. CMS 95-00301 and CMS 95-28083 (Dr. S. C. Liu, Program
Director) and the Central Research Institute of Electric Power Industry in Japan. The authors
would also like to acknowledge the generous efforts of Prof. Erik A. Johnson, University of
Southern California Department of Civil and Environmental Engineering, in editing and provid-
ing detailed comments on this paper and the corresponding MATLAB programs.
References
Antoniou, A. (1993). Digital Filters: Analysis, Design, and Applications, McGraw-Hill, Inc., New
York, pp.444–446.
Baker, G.A., Johnson, E.A. and Spencer, B.F., Jr. (1999). “Control Strategies for a Structural Con-
trol Benchmark Study: Verification by Experiment.” In the Proceedings of the 13th ASCE En-
gineering Mechanics Division Specialty Conference, Johns Hopkins University, Baltimore,
June13–16, 6 pages, (CD-ROM).
Balas, G.J. (1997). “Synthesis of Controllers for the Active Mass Driver System in the Presence of
Uncertainty.” Proceedings of the ASCE Structures Congress XV, 1997.
Bani-Hani, K., and Ghaboussi, J. (1998). “Nonlinear Structural Control Using Neural Networks.”,
Journal of Engineering Mechanics, Vol. 124, No. 3, pp.319–327.
Barrosa, L. R., (1999). “Performance evaluation of vibration controlled steel structures under seis-
mic loading”, Ph.D. thesis of Stanford University.
Caughey, T. K., (1998). “The Benchmark Problem.” Earthquake Engineering and Structural Dy-
namics, Vol. 27, pp.1125.
Chang, K. C., Chen, S. J., and Lai, M. L. L., (1996). “Inelastic behavior of steel frames with added
viscoelastic dampers.” Journal of Structural Engineering, Vol 122, No. 10, pp.1178–1186.
Chen, J-C. Ed. (1996). Proceedings of the 2nd International Workshop on Structural Control: Next
Generation of Intelligent Structures, Research Center, The Hong Kong University of Science
and Technology (see http://cwis.usc.edu/dept/civil_eng/structural/welcome.html).
Clough, R.W. and Johnson, S. (1966). “Effect of stiffness degradation on earthquake ductility re-
quirements.” Proc. of the Japan Earthquake Engineering Symposium, Tokyo.
Cook, R.D., Malkus, D.S., Plesha, M.E. (1989). Concepts and Applications of Finite Element Anal-
ysis, John Wiley & Sons, New York.
Dyke, S.J., Spencer, B.F., Jr., Quast, P. and Sain, M.K. (1995). “The Role of Control-Structure In-
teraction in Protective System Design.” Journal of Engineering Mechanics, ASCE, Vol.121,
No. 2, pp.322–338.
Fujino, Y., Soong, T.T. and Spencer, B.F., Jr. (1996). “Structural Control: Basic Concepts and Ap-
plications.” Proceedings of the ASCE Structures Congress XIV, Chicago, Illinois, pp. 1277–
1287.
Giberson, M.F. (1967). “The response of nonlinear multistory structures subjected to earthquake
excitation.” Ph.D. Dissertation, Caltech.
Housner, G.W., et al., Masri, S.F., and Chassiakos, G.A., Eds. (1994). Proceedings of the First
World Conference on Structural Control, International Association for Structural Control, Los
Angels.
Housner, G.W., Bergman, L.A., Caughey, T.K., Chassiakos, A.G., Claus, R.O., Masri, S.F., Skel-
ton, R.E., Soong, T.T., Spencer, B.F., Jr. and Yao, J.T.P. (1997). “Structural Control: Past,
Present and Future,” Journal of Engineering Mechanics, ASCE, Vol. 123, No. 9, pp. 897–971.
page 32
Lobo, R.F. (1994). “Inelastic Dynamic Analysis of Reinforced Concrete Structures in Three Di-
mensions.” Ph.D. Dissertation, Dept. of Civil Engrg. New York State University at Buffalo.
Lu, J., and Skelton, R.E. (1997). “Covariance Control Using Closed Loop Modeling for Struc-
tures.” Proceedings of the ASCE Structures Congress XV, 1997.
Masri, S.F., Bekey, G.A. and Caughey, T.K. (1981). “Optimum Pulse Control of Flexible Struc-
tures.” Journal of Applied Mechanics, Vol. 48, pp.619–626.
Masri, S.F., Bekey, G.A. and Caughey, T.K. (1982). “On-Line Control of Nonlinear Flexible Struc-
tures.” Journal of Applied Mechanics, Vol. 49, pp.877–884.
MATLAB (1997). The Math Works, Inc., Natick, Massachusetts.
Newmark, N.M. (1959) “A method of computation for structural dynamics” Journal of Engineer-
ing Mechanics Division, ASCE, Vol. 85, pp.67–94.
Ohtori, Y. and Spencer, B.F., Jr. (1999). “A MATLAB-Based Tool for Nonlinear Structural Analy-
sis”, In the Proceedings of the 13th ASCE Engineering Mechanics Division Specialty Confer-
ence, Johns Hopkins University, Baltimore, June13–16, 6 pages, (CD-ROM).
Pantelides, C. P., and Nelson, P. A. (1995). “Continuous Pulse Control of Structures with material
Nonlinearity.”, Earthquake Engineering and Structural Dynamics, Vol. 24, pp. 263–282.
Reinhorn, A.M., Manolis, G.D., and Wen, C.Y. (1987). “Active Control of Inelastic Structures.”,
Journal of Engineering Mechanics, Vol. 113, No. 3, pp.315–333.
Sack, R.L., (1989). Matrix Structural Analysis, PWS-Kent Pub. Co., Boston.
Skelton, R.E. (1988). Dynamic System Control: Linear Systems Analysis and Synthesis. Wiley,
New York.
Smith, H.A., Breneman, S.E. and Sureau, O. (1997). “H–infinity Static and Dynamic Output Feed-
back Control of the AMD Benchmark Problem.” Proceedings of the ASCE Structures Con-
gress XV, 1997.
Soong, T.T. (1990). Active Structural Control: Theory and Practice, Longman Scientific and Tech-
nical, Essex, England.
Soong, T.T. and Constantinou, M.C., Eds. (1994). Passive and Active Structural Vibration Control
in Civil Engineering, CISM Lecture Note, Springer-Verlag, New York.
Spencer, B.F., Jr., Suhardjo, J. and Sain, M.K. (1994). “Frequency Domain Optimal Control Strat-
egies for Aseismic Protection.”, Journal of Engineering Mechanics, ASCE, Vol. 120, No.1,
pp.135-159.
Spencer, B.F., Jr., Dyke, S.J., and Deoskar, H.S. (1997). “Benchmark Problem in Structural Con-
trol.”, Proceeding of the ASCE Structural Congress XV, Vol. 2, pp.1265-1269.
Spencer, B.F., Jr., Dyke, S.J. and Deoskar, H.S. (1998a,b). “Benchmark Problems in Structural
Control – Part I: Active Mass Driver System; Part II: Active Tendon System.” Earthquake En-
gneering. and Structural Dynamics, Vol. 27, No. 11, pp.1127–2247.
Spencer, B.F., Jr. and Sain, M.K. (1997) “Controlling Buildings: A New Frontier in Feedback,”
IEEE Control Systems Magazine: Special Issue on Emerging Technologies (Tariq Samad
Guest Ed.), Vol. 17, No. 6, pp. 19–35.
Spencer, B.F., Jr., Christenson, R.E. and Dyke, S.J. (1999). “Next Generation Benchmark Control
Problems for Seismically Excited Buildings.” Proc., 2nd World Conf. on Structural Control,
(T. Kobori, et al., eds., Wiley), Vol.2, pp.1135–1360.
Stengel, R.F. (1986). Stochastic Optimal Control: Theory and Application. Wiley, New York.
Subbaraj, K. and Dokainish, M.A. (1989). “A Survey of Direct Time-Integration Methods in Com-
putational Structural Dynamics - II. Implicit Method.” Computers and Structures., Vol.32, No.
6, pp.1387–1401.
Tsai, C. S., Chen, K. C., and Chen, C. S. (1998). “Nonlinear Behavior of Structures with Added
Passive Devices.”, Proceedings of Second World Conference on Structural Control, Kyoto, Ja-
pan, Vol. 1, pp.309–318.
page 33
Valles, R.E., Reinhorn, A.M., Kunnath, S.K., Li, C. and Madan, A. (1996). “IDARC2D Version4.0:
A Computer Program for the Inelastic Damage Analysis of the Buildings.” Technical Report
NCEER-96-0010, Nat. Ctr. for Earthquake Engrg. Res., Buffalo, New York.
Wu, J.C., Agrawal, A.K. and Yang J.N. (1997). “Application of Sliding Mode Control to a Bench-
mark Problem.” Proceedings of the ASCE Structures Congress XV, 1997.
Yang, J. N., Li, Z., and Liu, S.C. (1992). “Control of Hysteretic System Using Velocity and Accel-
eration Feedbacks.” Journal of Engineering Mechanics, Vol.118, No. 11, pp.2227–2245.
Yang, J. N., Li, Z., and Vongchavalitkul, S. (1994). “Generalization of Optimal Control Theory:
Linear and Nonlinear Control.”, Journal of Engineering Mechanics, Vol. 120, No. 2, pp.266–
283.
Yang, J. N., Wu, J. C., and Agrawal, A. K. (1995). “Sliding Mode Control for Nonlinear and Hys-
teretic Structures.”, Journal of Engineering Mechanics, Vol. 121, No. 12, pp.1330–1339.
Yang, J. N., Wu, J. C., Samali, B., and Agrawal, A. K. (1999). “A Benchmark Problem for Re-
sponse Control of Wind-Excited Tall Buildings.” Proc., 2nd World Conf. on Structural Con-
trol, (T. Kobori, et al., eds., Wiley), Vol.2, pp.1408–1416.
page 34
Appendix A: Uncontrolled Responses
Table 3: Uncontrolled Response Quantities for the 3-Story Benchmark Building
(integration time step = 0.005 sec)
(intensity) (0.5/1.0/1.5) (0.5/1.0/1.5) (0.5/1.0) (0.5/1.0)
1.0529e-02 8.7308e-03
max 1.8433e-02 1.9929e-02
δ 1.5034e-02 1.4491e-02
2.5292e-02 3.6526e-02
1.7136e-02 2.0449e-02
max 4.2820e+00 3.9299e+00
ẋ˙a 7.2971e+00 7.3332e+00
6.5389e+00 6.4800e+00
(m/sec2)
9.5898e+00 1.2035e+01
7.4044e+00 6.9300e+00
max 8.9267e+06 6.8201e+06
Fb 1.2542e+07 1.1148e+07
1.0120e+07 9.6532e+06
(N)
1.4225e+07 1.3369e+07
1.0978e+07 1.1374e+07
1.3148e-03 1.7714e-03
max 7.3560e-03 4.5279e-03
δ 2.4586e-03 3.1946e-03
9.1753e-03 1.0031e-02
4.2633e-03 4.4445e-03
max 5.6856e-01 7.3930e-01
ẋ˙a 7.1020e-01 8.7806e-01
9.2382e-01 1.2113e+00
(m/sec2)
1.0539e+00 1.1339e+00
1.2678e+00 1.4789e+00
max 1.0311e+06 1.3891e+06
Fb 1.2341e+06 1.5471e+06
1.6609e+06 2.2709e+06
(N)
1.9531e+06 1.9071e+06
2.2803e+06 2.7837e+06
1.1186e+00 8.6224e-01
max 3.0339e+00 2.3134e+00
φ 1.8119e+00 1.7610e+00
4.1131e+00 3.9848e+00
2.2866e+00 2.2747e+00
1.8401e+00 --
max 3.7655e+01 1.0430e+02
E 9.1945e+01 1.0729e+02
1.8877e+02 2.5011e+02
1.8260e+02 2.7485e+02
2.0000e+00 --
2.4000e+01 2.4000e+01
Nd 2.2000e+01 2.2000e+01
3.3000e+01 3.2000e+01
2.2000e+01 2.4000e+01
1.3509e-01 1.7583e-01
max 1.3877e+00 5.7743e-01
φ 3.8657e-01 3.7202e-01
1.4097e+00 1.3249e+00
6.7595e-01 6.0460e-01
max 1.0491e-01 9.0144e-02
x 1.6690e-01 1.9294e-01
1.5073e-01 1.4653e-01
(m) 2.7151e-01 3.0690e-01
1.7202e-01 1.8510e-01
max 7.8298e-01 5.3705e-01
ẋ 1.1126e+00 1.3988e+00
9.5484e-01 8.8999e-01
(m/sec) 1.4105e+00 1.8750e+00
1.2267e+00 9.2899e-01
page 35
(intensity) (0.5/1.0/1.5) (0.5/1.0/1.5) (0.5/1.0) (0.5/1.0)
8.5680e-03 6.9973e-03
max 1.5453e-02 2.1616e-02
δ 1.5088e-02 1.3336e-02
2.0871e-02 2.9315e-02
2.1311e-02 1.9413e-02
max 3.2884e+00 2.5913e+00
ẋ˙a 5.6327e+00 6.9249e+00
5.4153e+00 5.1598e+00
(m/sec2)
7.3395e+00 1.0097e+01
6.1457e+00 6.0303e+00
max 1.1581e+07 9.9777e+06
Fb 1.7985e+07 1.4707e+07
1.7337e+07 1.6710e+07
(N)
2.1233e+07 2.0813e+07
1.9400e+07 1.8261e+07
1.5064e-03 1.1758e-03
max 3.1206e-03 3.7216e-03
δ 2.4850e-03 2.2736e-03
3.7980e-03 6.6937e-03
3.6308e-03 4.1847e-03
max 5.7698e-01 3.7648e-01
ẋ˙a 8.2840e-01 7.0597e-01
9.7837e-01 6.9060e-01
(m/sec2)
9.3598e-01 8.9825e-01
1.2604e+00 9.0154e-01
max 1.8315e+06 1.6650e+06
Fb 4.1269e+06 1.7684e+06
2.8483e+06 3.1243e+06
(N)
4.4983e+06 2.2577e+06
3.3887e+06 4.2257e+06
8.7286e-01 7.5391e-01
max 2.1660e+00 2.4609e+00
φ 2.1145e+00 1.8784e+00
2.8399e+00 3.1837e+00
3.1185e+00 2.0809e+00
-- --
max 3.2348e+01 7.4119e+01
E 1.5704e+01 1.6693e+01
8.9588e+01 2.4428e+02
4.0527e+01 4.3983e+01
-- --
6.3000e+01 2.6000e+01
Nd 4.9000e+01 5.7000e+01
7.5000e+01 7.6000e+01
6.9000e+01 7.1000e+01
1.4043e-01 1.2737e-01
max 3.7622e-01 4.5631e-01
φ 3.0245e-01 3.0150e-01
5.2675e-01 1.3378e+00
6.2741e-01 6.7259e-01
max 2.0010e-01 1.8627e-01
x 4.4937e-01 3.0704e-01
3.8509e-01 3.7228e-01
(m) 6.1972e-01 4.7760e-01
4.8825e-01 4.6111e-01
max 8.3503e-01 6.5607e-01
ẋ 1.6355e+00 1.1205e+00
1.3814e+00 1.1492e+00
(m/sec) 2.3196e+00 1.9751e+00
1.4688e+00 1.1501e+00
page 36
(intensity) (0.5/1.0/1.5) (0.5/1.0/1.5) (0.5/1.0) (0.5/1.0)
3.5418e-03 3.2055e-03
max 1.0899e-02 1.0807e-02
δ 7.0835e-03 6.4111e-03
1.8820e-02 2.0481e-02
1.0640e-02 9.1585e-03
max 2.6696e+00 1.8291e+00
ẋ˙a 6.0076e+00 7.4945e+00
5.3392e+00 3.6582e+00
(m/sec2)
8.5107e+00 9.9741e+00
7.7856e+00 4.7152e+00
max 3.6882e+06 3.1784e+06
Fb 1.0363e+07 1.0079e+07
7.3765e+06 6.3568e+06
(N)
1.4257e+07 1.1423e+07
9.5282e+06 9.1234e+06
6.7334e-04 6.0306e-04
max 1.8994e-03 1.4874e-03
δ 1.3467e-03 1.2061e-03
6.6722e-03 7.3706e-03
2.0001e-03 1.7664e-03
max 3.8851e-01 2.7580e-01
ẋ˙a 8.0270e-01 7.5885e-01
7.7702e-01 5.5161e-01
(m/sec2)
1.0136e+00 1.0029e+00
1.1286e+00 8.0448e-01
max 7.6393e+05 7.0643e+05
Fb 2.1527e+06 1.6258e+06
1.5279e+06 1.4129e+06
(N)
2.3912e+06 2.0298e+06
2.2728e+06 2.0821e+06
4.6630e-01 4.2780e-01
max 1.6753e+00 1.9229e+00
φ 9.3259e-01 8.5560e-01
3.3601e+00 3.1987e+00
1.5032e+00 1.3970e+00
-- --
max 2.0605e+01 1.8701e+01
E -- --
8.6226e+01 2.8269e+02
1.9465e+01 6.0923e+00
-- --
9.6000e+01 7.8000e+01
Nd -- --
1.9200e+02 1.6800e+02
8.6000e+01 8.6000e+01
9.8637e-02 9.3745e-02
max 3.6294e-01 2.4164e-01
φ 1.9727e-01 1.8749e-01
1.4056e+00 1.4228e+00
3.3098e-01 2.7843e-01
max 1.5219e-01 1.7397e-01
x 4.8940e-01 3.1409e-01
3.0439e-01 3.4795e-01
(m) 7.5036e-01 5.1734e-01
4.5352e-01 4.9178e-01
max 4.5783e-01 4.5122e-01
ẋ 1.6616e+00 1.3145e+00
9.1566e-01 9.0244e-01
(m/sec) 2.0390e+00 1.8458e+00
1.3007e+00 1.2684e+00
page 37
Appendix B: Nomenclature
Nomenclature
Ã – matrix used in the solution of the algebraic Riccati equation
A – matrix used in the solution of the algebraic Riccati equation
˜
A d, B d , E d – state space matrices for the reduced order design model
α AA , α BB – concentrated plasticity parameters
β – Parameter for Newmark- β method
C – Global damping matrix
Ĉ – Damping matrix for reduced system
C md, D md, F md – state space matrices for the reduced order evaluation model used to specify the
measured responses
C ed, D ed, F ed – state space matrices for the reduced order evaluation model used to specify the
regulated responses
c 1 , c 2 – parameters for Rayleigh damping
D s – sensor gain
d i – interstory drift of above ground levels (m)
∆F err – vector of the unbalanced force (N)
∆F D – equivalent force of the incremental equation of motion (N)
∆f – incremental control force (N)
∆t – time interval for Newmark- β integration (sec)
∆U – incremental response vector
∆U act – incremental response vector for active degrees of freedom
∆U slv – incremental response vector for slave degrees of freedom
∆ẋ˙g – ground acceleration increment (m/sec2)
max
δ – maximum interstory drift ratio for each respective earthquake
max
δ – the maximum normed interstory drift ratio corresponding to the uncontrolled
structure excited by each respective earthquake
E [ ] – expected value operator used in the infinite horizon performance index for the
.
sample LQG
max
E – maximum dissipated energy of the uncontrolled structure
page 38
EI A , EI B , EI O – flexural stiffness at end A, end B, and center of member, respectively.
ζ i – damping in the i-th mode of the evaluation model
max
Fb – maximum base shear (N)
max
Fb – the maximum normed uncontrolled base shear for each respective earthquake (N)
F yj – yield moment at the end of the j-th member
f – vector of forces produced by the control device(s) (kN)
f AA , f AB , f BA , f BB f AA' , f AB' , f BB' – flexibility coefficients
f l – control force of the l-th control device (N)
G – loading vector of ground acceleration to the structure
G – matrix used in the solution of the algebraic Riccati equation
˜
GA – shear stiffness of member
g – gravitational constant (9.81 m/sec2)
g 1 (. ) → g 4 (. ) – functions defining the evaluation model
g 5 (. ), g 6 (. ) – sensor dynamics for the sensor(s)
g 7 (. ), g 8 (. ) – functions defining the feedback controller
g 9 (. ) → g 11 (. ) – functions defining the control device(s)
γ – parameter for Newmark- β method
γ g – ratio of autospectral densities for Gaussian white noise
H – matrix used in the solution of the algebraic Riccati equation
˜
h i – height of the i-th level (m)
I – identity matrix
J i – i-th evaluation criteria
Ĵ – infinite horizon performance index

K – global stiffness matrix of structure
K t – tangent stiffness matrix of structure at time t
K D – equivalent dynamic stiffness matrix
K f – matrix to account for multiple actuators and to load control forces onto building
K̂ – global stiffness matrix for reduced system

K – full state feedback gain matrix for the deterministic regulator problem
˜
k – discrete time step index
page 39
k AA , k AB , k BA , k BB – stiffness matrix coefficient
L – length of member (m)
L – Kalman Filter optimal estimator
l – index of control devices
M – global mass matrix
M̂ – global mass matrix for reduced system
m i – seismic mass (including framing) of the i-th level
Ñ – matrix used in the solution of the algebraic Riccati equation

N d – number of damaged connections of the uncontrolled structure
C
N d – number of damaged connections of the structure with the applied control strategy
n – index of global DOFs
P – vector defining the loading of control forces onto the structure
P – solution of the algebraic Riccati equation
˜
p – index of local DOFs
Φ̂ – mode vector for reduced system
P i – measure of the instantaneous power required by the active control actuators
P i – measure of the total power required by the ith actuator for the entire seismic event
φ j – curvature at the ends of the j-th member
φ yj – yield curvature at the ends of the j-th member
max
φ – maximum curvature of the uncontrolled structure
max
φ – normed maximum curvature at the member ends
Q – weighting matrix of responses for LQG control design
Q̃ – matrix used in the solution of the algebraic Riccati equation
q – index for node number
R – weighting matrix of control forces for LQG control design
R̃ – matrix used in the solution of the algebraic Riccati equation
R – matrix used in the solution of the algebraic Riccati equation
˜
S – solution of the algebraic Riccati equation
S vi vi – autospectral density function of measurement noise
S ẋ˙g ẋ˙g – autospectral density function of ground acceleration
T – sampling time (sec)
page 40
T R – transformation matrix for rigid floor
t – continuous time step index (sec)
t f – sufficiently large time to allow the response of the structure to attenuate (sec)
U t – response vector at time t
u – continuous vector control output (sample and hold of u k ) (Volts)
u k – vector control output at time t = kT for the feedback compensator (Volts)
v – measurement noise vector for the sensors
W – seismic weight of the structure (N)
ω i – natural (undamped) frequency of the i-th mode of the evaluation model
x – state vector for the evaluation model
a
x – continuous time state vector for the control device
x i – i-th state of evaluation model
c
x k – state vector for the discrete feedback compensator at time t = kT
max
x – maximum uncontrolled displacement relative to ground for each respective earth-
quake (m)
max
x – the maximum normed uncontrolled displacement of the levels for each respective
earthquake (m)
d
x – continuous state vector for the design model
d
x̂ – Kalman Filter estimate of the state vector for the design model
s
x – state vector of the sensor model
max
ẋ – maximum uncontrolled velocity relative to ground (m/sec)
ẋ˙ai – absolute acceleration of the ith level (m/sec2)
max
ẋ˙a – maximum uncontrolled absolute roof acceleration (m/sec2)
max
ẋ˙a – the maximum normed absolute acceleration of the uncontrolled structure excited
–3 ⁄ 2
by each respective earthquake ( m ⋅ sec )
ẋ˙g – absolute acceleration of the ground (m/sec2)
a
y l – displacement of the l-th control device (m)
a
ẏ l – velocity of the l-th control device (m/sec)
y c – vector of connector responses of evaluation model
page 41
y cd – vector of connection responses of reduced order design model
y e – vector of evaluation responses of evaluation model
y f – vector of control device responses
y m – vector of measured responses of evaluation model
y md – vector of measured responses of design model
s
y – vector of responses of sensor model
s
y k – discrete vector of responses of sensor model
y ed – vector of regulated responses of design model
. – absolute value operator
. – normed operator
∫ dEj – dissipated energy at the ends of the j-th member.
page 42
View publication stats

Benchmark Control Problems For Seismically Excited

Uploaded by

Copyright:

Available Formats

Benchmark Control Problems For Seismically Excited

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Benchmark Control Problems For Seismically Excited

Uploaded by

Copyright:

Available Formats

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

Benchmark Control Problems for Seismically Excited Nonlinear Buildings

Article in Journal of Engineering Mechanics · April 2004

Yasuki Ohtori Billie F. Spencer

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Y. Ohtori1, R.E. Christenson and B.F. Spencer, Jr.

Ground (1) (2) (3) (4) (5)

Figure 1. 3-Story Benchmark Building N-S MRF.

Building Plan NOTES

indicates a moment resisting connection.

2nd (19) (20) (21) (22) (23) (24)

Figure 2. 9-Story Benchmark Building N-S MRF.

20-Story Benchmark Building

17th (115) (116) (117) (118) W27x84 (119) (120)

corner columns and interior columns the same,

11th (79) (80) (81) (82) W30x99 (83) (84)

9th (67) (68) (69) (70) (71) (72) thickness of t).

8th (61) (62) (63) (64) (65) (66)

5th (43) (44) (45) (46) W30x108 (47) (48) Connections:

basement level heights 3.65 m (12′-0″);

2nd (25) (26) (27) (28) (29) (30)

Figure 3. 20-Story Benchmark Building N-S MRF.

Mode 1 (0.44 Hz) Mode 2 (1.18 Hz) Mode 3 (2.05 Hz)

Mode 1 (0.26 Hz) Mode 2 (0.75 Hz) Mode 3 (1.30 Hz)

M∆U̇˙ + C∆U̇ + K∆U = – MG∆ẋ˙g + P∆f + ∆F err (1)

U t + ∆t = U t + ∆tU̇ t + ( ∆t )  --- – β U̇˙t + βU̇˙t + ∆t

U̇ t + ∆t = U̇ t + ∆t [ ( 1 – γ )U̇˙t + γ U̇˙t + ∆t ] (6)

In the Eq. (11), the flexibility coefficients of the elements are:

f BB ′ = 4EI A EI B + ( EI 0 – EI B )EI A α BB (17)

MATLAB Implementation of Nonlinear Analysis

Overview of Researcher/Designer Tasks

Figure 8. Schematics of the Nonlinear Benchmark Evaluation Model.

Benchmark Control Problems

Control Devices Controller Sensors

Figure 9. SIMULINK Block Diagram for Vibration Control Simulator.

Interfacing with the Sensor Models

Interfacing with the Controller

Interfacing with the Control Device Models

(c) Northridge (d) Kobe

where the norm, . , is computed using the following equation

With passive control devices, the J 14 criterion is zero.

J 15 = Number of control devices (46)

J 16 = Number of required sensors (47)

Interstory Drift Ratio Level Acceleration Base Shear

Ductility Dissipated Energy Plastic Connections

Normed Ductility Control Force Control Device Stroke

Control Power Normed Control Power Control Devices

Figure 11. SIMULINK Block Representing the Sensors in the Sample

Figure 12. SIMULINK Block Representing the Control Devices in the

Figure 13. SIMULINK Block Representing the Control Algorithm

Evaluation of Sample Control Design

of the Roof (m/sec2)

Time (sec) Time (sec)

Above Ground Level

Time (sec) Time (sec)

Above Ground Level

Figure 14. Uncontrolled (Dotted) and Controlled (Solid) Responses:

Earthquake El Centro Hachinohe Northridge Kobe