Académique Documents
Professionnel Documents
Culture Documents
CONCRETE FRAME
By S. G. Buonopane,1 Associate Member, ASCE, and R. N. White,2 Fellow, ASCE
ABSTRACT: Seismic evaluation of a two-story, two-bay reinforced concrete frame infilled with masonry was
performed by pseudodynamic testing of a half-scale specimen. The second-story infill included window openings.
The specimen was subjected to four tests of increasing magnitude based on the Taft ground motion. Explicit
numerical integration with a small time step, soft-coupled load system, and an iterative actuator control algorithm
Downloaded from ascelibrary.org by Indian Institute of Technology Roorkee on 07/23/18. Copyright ASCE. For personal use only; all rights reserved.
limited the displacement errors normally associated with pseudodynamic testing of stiff structures. The final
sequence of tests produced diagonal cracking in the upper story, but primarily bed joint shear cracking in the
lower story. Relations between the type of observed cracking and story drift-story shear response are explored.
Compressive strut mechanisms are defined from the experimental values of moments and axial forces in the
frame and infill panel strains. Estimates of story stiffness from several simple strut models were found to bound
the experimentally measured values for both the first- and second-story walls prior to significant damage. Fric-
tion-based analytical estimates of panel shear strength were found to underestimate the measured strength and
to be sensitive to the assumed coefficient of friction. Available methods for estimating shear strength that neglect
infill-frame interaction were found to largely underestimate measured shear strength.
Compressive Secant
Specimen Number of strengtha modulusb
Material type specimens (MPa) (MPa)
(1) (2) (3) (4) (5)
Concrete, 4 ⫻ 8 in. 8 30.6 21,900
28 day cylinder
Concrete, 4 ⫻ 8 in. 13 40.0 25,900
15 month cylinder
Masonry single block 12 16.1 8,410
Masonry 3 course 15 10.1 7,550
Downloaded from ascelibrary.org by Indian Institute of Technology Roorkee on 07/23/18. Copyright ASCE. For personal use only; all rights reserved.
prism
Mortar, 2 ⫻ 4 in. 14 13.9 12,700
type S cylinder
a
Masonry compressive strengths over net area.
b
Secant modulus to 45% of compressive strength.
ically where possible, and the frequencies from the free vibra-
FIG. 1. Elevation of Half-Scale Infill Specimen for PSD Testing tion tests were cross-checked by eigenvalue analysis using the
measured stiffness matrix and an estimated mass matrix. De-
tailed results of all preliminary testing appear in Buonopane
tailing and with nonductile (GLD) detailing. Another substan- (1997).
tial experimental and analytical research program is reported
in Zarnic and Tomazevic (1984, 1985), where cyclic tests were PSEUDODYNAMIC TEST METHOD
performed on one-half and one-third scale reinforced concrete
frames infilled with brick and block. Both unreinforced and Pseudodynamic testing combines features of quasi-static
reinforced infills were tested, as well as panels with window testing, shake table testing, and numerical time history analysis
and door openings. Results have been used to propose some in order to realistically simulate the nonlinear behavior of
innovative nonlinear strut models, including multiple strut structures that exhibit varying stiffness, here caused by dam-
schemes for panels with openings. Tests of unreinforced infill age incurred during the excitation. Whereas numerical simu-
at one-fourth scale in steel frames by Mosalam (1996, 1997) lations rely on hysteretic rules to trace the changing stiffness
include quasi-static and PSD tests on multistory, multibay based on evolving system parameters, such as displacements
specimens and infills with window and door openings. or interstory drifts, PSD testing accounts for the nonlinear ef-
fects through experimentally measured restoring forces. The
measured restoring forces replace the product of the stiffness
SPECIMEN INSTRUMENTATION AND PROPERTIES
matrix and displacement vector in the equation of motion. The
Displacement transducers (DCDTs) mounted on an external remainder of the PSD algorithm follows from established nu-
reference frame measured the story displacements at the center merical integration techniques.
of the interior beam-column joints and displacement at the The PSD test method was first proposed and used by Jap-
center of the base beam. Eighty strain gauges on the reinforc- anese researchers in 1975, and its mathematical formulation
ing steel provided strain values for calculation of moments and and development have been reviewed in Takanashi and Na-
axial forces in the frame members. Three strain gauge delta kashima (1987) and Mahin et al. (1989). Recent advances in
rosettes were applied on each of the first-story masonry panels PSD-specific numerical techniques (Shing et al. 1996) and in
to measure principal strains and directions. Twenty DCDTs actuator control schemes (Seible et al. 1996) have resulted in
measured total length changes across main and off diagonals wider use of PSD testing for stiff, multi-story specimens.
of the masonry, and across window openings. Eight additional A review of PSD literature reveals certain unique advan-
DCDTs positioned near the panel corners measured interface tages and disadvantages compared with quasi-static or shake
gap openings between frame and infill. Story displacements table testing. Considering first the advantages:
were applied at each story level using a 245 kN capacity, ⫾75
mm stroke hydraulic actuator. The actuators were connected 1. For MDOF systems, no assumptions on the distribution
to the beam stubs through compression-only, ball-and-cup con- of seismic forces among the DOFs need be made.
nections, thereby minimizing rotational restraint and allowing 2. Effects of damage on behavior are physically modeled
realistic reinforcing at the points of load application (Fig. 1). with no numerical assumptions regarding degradation
An extensive materials testing program was conducted dur- necessary.
ing construction and PSD testing of the specimen, including 3. Controlled testing speed allows for data acquisition from
mix design development of the frame concrete and various extensive instrumentation with modest electronics and
masonry tests (Buonopane 1997). Important material proper- careful recording of important information such as crack
ties of the as-built concrete and masonry appear in Table 1. trajectories.
Each CMU has a measured gross area of 17,800 mm2, net area 4. Full-scale and large specimens may be tested with equip-
of 9400 mm2, and mortared thickness of 34.4 mm. All rein- ment requirements not much different than necessary for
forcing steel is indicated in Fig. 1, and main longitudinal steel quasi-static testing.
has a nominal yield stress of 275 MPa. 5. Specimen mass need not be accurately reproduced in the
Free vibration and static flexibility tests were performed on lab as it is modeled numerically.
the concrete frame before the infill was built and repeated 6. Unique substructure tests are possible, where part of the
again after construction of the infill. The free vibration tests prototype is built and tested in the lab, while the re-
were used to estimate modal frequencies and natural damping mainder is modeled numerically within the time integra-
ratios. Results of these preliminary tests were verified analyt- tion loop.
JOURNAL OF STRUCTURAL ENGINEERING / JUNE 1999 / 579
produced without real-time PSD testing. iteration step k, and is a reduction factor to prevent over-
3. Controlled testing speed and small integration time steps shooting of the target. The iteration is halted when the differ-
for numerical accuracy may require excessive testing ence d s ⫺ d̂ s(k) falls below a tolerance level.
times. Seible et al. (1996) demonstrate that intentional soft cou-
4. Test response is often specific to a particular input mo- pling may be used to overcome several experimental problems
Downloaded from ascelibrary.org by Indian Institute of Technology Roorkee on 07/23/18. Copyright ASCE. For personal use only; all rights reserved.
M= 冋5080
0
0
3330 册 kg (1)
olution limit of the actuator. Such fine displacement control
can be especially useful during preliminary testing when ex-
tremely low-level excitations are used to verify the PSD al-
Using this mass matrix, the expected modal frequencies during gorithm and software without damaging the specimen.
low-level PSD testing are 13 Hz and 28 Hz. As the specimen The success of this control algorithm depends largely upon
is damaged, the modal frequencies will approach those of the the use of a well-determined matrix Q. Previously, Q has been
bare frame, 3 Hz and 9 Hz. The modal damping ratios were assembled mathematically based on independently measured
specified as 2% and 5% to account for only natural damping, stiffnesses of the specimen and flexible load chain elements.
as PSD testing includes the effects of frictional and hysteretic In the present test setup, soft coupling occurred as a result of
damping through the measured force feedback. Rayleigh several flexible loading system components and also was de-
damping is an acceptable assumption since the system has only pendent upon the direction of loading; therefore, Q was de-
two DOFs with roughly equal magnitude damping. The damp- termined by direct measurement, accounting for flexibility dis-
ing matrix was assumed constant throughout the PSD testing tributed anywhere in the loading system, symmetric or not.
sequence. Control of the actuator displacements and measurement of
An explicit Newmark algorithm ( = 0, ␥ = 0.5) was used both actuator and structure displacements allow assembly of
to integrate the equation of motion, with a small time step to Q based on (3). Conveniently, such measurements may be re-
ensure stability and accuracy. The application of explicit in- corded during a typical stiffness test. For this specimen, Q
tegration is limited by the stability criterion was measured before each PSD test. Typically, the plots of
actuator displacements versus structure displacements showed
n(⌬t) < 2 (2) both nonlinearity and asymmetry, but a reasonable linear es-
with n the largest circular frequency of the structure. The time timate gave acceptable performance in the PSD tests. Before
step of the input record was approximately 0.0035 s, resulting the Taft 0.10g test
in a criterion of 0.62. Studies of error propagation in PSD tests
by Shing and Mahin (1990) for various integration methods
suggest that, for this structure and time step, explicit integra-
Q= 冋 4.69
⫺1.57
⫺1.24
2.52 册 (5)
tion has comparable error propagation characteristics to other while for the Taft 0.80g test
冋 册
more complicated integration techniques.
1.73 ⫺0.30
Q= (6)
Soft Coupled Load System and Actuator Control ⫺0.53 1.31
A soft coupled load system uses flexible elements in the showing the expected convergence of Q to the identity matrix.
load chain connecting the actuator to the specimen, resulting Severe stiffness degradation occurred during the Taft 0.80g
in actuator displacements that are significantly greater than the test; nevertheless, the single estimated Q of (6) provided sat-
corresponding specimen displacements, and it requires a spe- isfactory actuator control throughout that test.
cific actuator control scheme. This method, introduced at the
University of California at San Diego (Seible 1996), is briefly PSD Displacement Errors
reviewed here.
In PSD testing the structure must be moved to a target dis- As pseudodynamic testing combines experimental and nu-
placement vector dictated by the numerical integration before merical processes, error from both sources may affect results.
restoring forces can be measured, but target specimen dis- Experimental errors enter the numerical process through force
placements from the numerical integration cannot be used di- and displacement feedback and may rapidly build up and over-
rectly as actuator command displacements. The actuator com- whelm the numerical process. Errors in displacement control
mand displacement vector, d a, is related to the target structural may be classified as overshooting or undershooting. Over-
displacement vector, d s, by shooting the target displacement may cause irreversible dam-
d a = Qd s (3)
age to the specimen which then affects the remainder of the
test. Undershooting may require excessive iterations to reach
where Q, the displacement amplification matrix, is a measure the target displacements, resulting in testing times beyond
of relative stiffnesses of structure and load system. When the practical limits and in load reversals near the target displace-
580 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 1999
accurately distinguish between under- and overshooting. Ex- FIG. 2. Taft Response Spectrum (Scaled to PGA of 1.0g)
amination of the experimental error data from these PSD tests
(Buonopane 1997) supports this conclusion. Undershoot errors ground motion for this series of PSD tests. The Taft response
consistently occurred more frequently and with greater mag- spectrum in Fig. 2 (scaled to a peak ground acceleration of
nitude, yet the error frequency spectra showed varied distri- 1.0g) shows that the trend of the spectrum rises between the
butions including nearly uniform and peaks near high or low fundamental period of the undamaged specimen (0.076 s) and
modal frequencies. No direct relationship between peaks in the the bare frame (0.302 s). Thus, damage to the infilled frame
error frequency spectra and the occurrence or magnitude of will increase the seismic demand on the structure. Since the
under- or overshooting errors was observed. specimen is half scale, the time scale of the input motion is
compressed by a factor of 公0.5; all times cited refer to this
Displacement Control Tolerances compressed time scale.
The actuator control algorithm of (4) prevents time integra- Verification with Low-Level Excitation
tion from proceeding unless a certain predefined tolerance
A low-level excitation with the Taft ground acceleration
level is met. With a tight tolerance band, experimental dis-
scaled to a peak ground acceleration (PGA) of 0.10g was used
placement errors entering the numerical integration must be
to verify the accuracy and stability of the PSD algorithm. Story
small. Even for actuator control which consistently under-
drift–story shear plots showed that the structure remained es-
shoots to protect the specimen from unintended damage, a
sentially undamaged, with hysteresis similar in magnitude to
tight tolerance apparently can limit excessive energy addition
that observed during preliminary static testing. No visible
and maintain cumulative error within acceptable bounds.
cracking was observed. Numerical simulation with a bilinear
Table 2 lists the tolerances used in the final sequence of
stiffness model and no degradation was used to verify the re-
PSD testing. The measured displacement is considered ac-
sults of the Taft 0.10g test. Measured and bilinear displacement
ceptable if it is within the specified tolerance above or below
time histories for both stories for 4 s of strong motion exci-
the target displacement. Tolerance expressed as a percentage
tation show good correspondence between the measured and
of maximum measured displacement allows for comparison
bilinear cases, with most local extremes reproduced very
across the tests of different excitation levels and to tests by
closely (Fig. 3). Hysteretic behavior and the frequency spectra
other researchers. For instance, results reported by Seible et
also showed good agreement. The primary conclusion from
al. (1996) show percentages of about 0.5% to 1.0% for tests
the bilinear modeling is that experimental errors in displace-
on a five-story masonry structure. In the Taft 0.35g test the
ment and force feedback, present in the low-level test but not
tolerance was extremely low relative to peak displacement,
in the bilinear model, have been controlled and reduced to the
and at several time steps the control algorithm failed to con-
point which they do not significantly affect the performance
verge within the iteration limit. In these unconverged steps,
of the PSD test.
the maximum error carried forward in the numerical integra-
tion was about 1.4% of peak displacement, a level sufficiently Final Pseudodynamic Tests
low to not adversely affect the remainder of the test. This
testing program, using tight actuator tolerances with a soft The final pseudodynamic testing sequence subjected the
coupled load system and actuator control scheme, suggests specimen to a series of three additional Taft excitations of PGA
PSD error propagation can be confined within acceptable 0.35g, 0.55g, and 0.80g. The Taft 0.35g excitation tested the
bounds by practical means. behavior of the PSD control algorithm with nonlinear stiffness
caused primarily by separation between infill and frame. Sev-
eral of the DCDTs positioned to measure relative movement
RESULTS OF PSEUDODYNAMIC TESTING of frame and infill exhibited gap-opening behavior.
The Taft-Lincoln School S69E record of the event at Kern The final tests at 0.55g and 0.80g produced significant dam-
County, California, on July 20, 1952, was selected as the age and degradation of the infill and frame, comparable to that
which might occur during major seismic events, such as a
TABLE 2. Actuator Control Tolerances and Displacement Er-
maximum considered earthquake. The ‘‘NEHRP recom-
rors mended provisions’’ (1997) map spectral acceleration contours
for a seismic event with 2% probability of occurrence in 50
Tolerance as years (2,500 year return) for 0.2 and 1.0 s period structures.
Test Tolerance Maximum percentage Maximum Spectral accelerations in the Los Angeles area generally range
PGA semibandwidth displacement of maximum error from 1.50g to 2.50g; in the San Francisco area, from 1.50g to
(g) (⫻10⫺3 mm) (mm) displacement (⫻10⫺3 mm) 2.00g. The peak contour surrounding Charleston, South Car-
(1) (2) (3) (4) (5)
olina, is 1.50g; the New Madrid area, 3.00g. From the re-
0.10 1.850 0.213 0.87 1.850 sponse spectrum (Fig. 2), the spectral acceleration at a period
0.35 1.850 2.121 0.09 29.21 of 0.2 s (about midway between the undamaged infilled frame
0.55 9.40 5.090 0.18 9.40
0.80 19.05 11.834 0.16 19.05
and bare frame periods) for a Taft 0.55g excitation is 1.9g,
and for a Taft 0.80g excitation is 2.8g. Thus the final two PSD
JOURNAL OF STRUCTURAL ENGINEERING / JUNE 1999 / 581
FIG. 3. Measured and Bilinear Displacement Time Histories for Taft 0.10g Excitation
Total hysteretic
PGA energy First-story Second-story
(g) (N-m) percentage percentage
(1) (2) (3) (4)
0.10 6.67 63 37
0.35 399 65 35
0.55 1593 66 34
0.80 7129 81 19
Downloaded from ascelibrary.org by Indian Institute of Technology Roorkee on 07/23/18. Copyright ASCE. For personal use only; all rights reserved.
blocks directly adjacent to the column, allowed a significant strain field consistently fell between 40⬚ and 45⬚ below hori-
shear to develop over a relatively short portion of the column. zontal, compared with the main diagonal angle of about 37⬚.
The top moment is 1160 kN-mm (132%) and the midheight The compressive strains generally increased with increasing
moment, 730 kN-mm (83%). Estimating the magnitude of PGA. However, for the Taft 0.80g test at 4.66 s (peak base
shear as the change in moment over half the column height shear) the east panel had a negligible increase in strain over
(630 mm) gives a shear of 30 kN. The actual shear will prob- the Taft 0.55g test, while the west panel had a significant in-
ably be larger than this value as lateral interaction force near crease. This difference between the east and west panels may
be due to the state of cracking at 4.66 s. At this time the east
wall had major bed joint failures near the top and bottom of
the panel, whereas the west wall had only one such failure
near the bottom. The ability of the east panel to develop a
significant corner-to-corner force was severely reduced by hor-
izontal slip along the cracked bed joints, caused by the hori-
zontal component of the strut force itself. The strut resistance
lost in the east panel may have been compensated by higher
strain in the less damaged west panel.
For the Taft 0.80g test at 10.76 s (peak first-story displace-
ment), severe damage had occurred in both first-story masonry
panels, including several full-width bed joint cracks as well as
numerous diagonal and stepped cracks. The increase in first-
story drift seen in the story drift–story shear plot may be as-
sociated with the failure of the infill to mobilize an effective
compressive strut mechanism because of the extensive damage
and many prominent horizontal slip planes.
␣h =
2冑 4 4Ec Ic h
E m t sin 2
(8)
FIG. 11. Frame Forces for Taft 0.80g Test at Time of Maximum A similar expression gives the horizontal contact length, ␣ L.
First-Story Displacement (10.76 s) Based on an assumed strut width, the axial strut stiffness is
value.
Paulay and Priestley (1992) recommend a conservatively high
strut width of one-fourth of the panel diagonal length, resulting
in a 527 mm wide strut and a story lateral stiffness of 108 kN/
mm (127% of measured). FEMA-273 recommends a strut width
of
冉 冊
⫺0.4
⫺0.4 h
w = 0.175rinf ( 1 hcol) = 0.175d (9)
2␣ h