Engineering Structures 24 (2002) 719–734 www.elsevier.com/locate/engstruct Seismic performance of a 3-story RC frame in a low-seismicity region Han-Seon Lee * , Sung-Woo Woo Department of Architectural Engineering, Korea University, Seoul 136-701 Received 16 April 2001; received in revised form 7 November 2001; accepted 14 November 2001 Abstract The objectives of the research stated herein are to investigate the seismic performance of a 3-story reinforced concrete (RC) ordinary moment-resisting frame, which has not been engineered to resist earthquake excitations, and to evaluate the reliability of the available static and dynamic inelastic analysis techniques. A 1:5 scale model constructed according to the Korean practice of nonseismic detailing and the similitude law was subjected to a series of the shaking table motions simulating Taft N21E component earthquake ground accelerogram. Due to the limitation in the capacity of the used shaking table, a pushover test was performed to observe the ultimate capacity of the structure after earthquake simulation tests. The model showed the linear elastic behavior under the Taft N21E motion with the peak ground acceleration of 0.12g, representing the design earthquake in Korea. The model revealed fairly good resistance to the higher levels of earthquake simulation tests though it was not designed against earthquakes. The main components of its resistance to the high level of earthquakes appear to be (1) the high overstrength, (2) the elongation of the fundamental period, (3) the minor energy dissipation by inelastic deformations, and (4) the increase of the damping ratio. The drifts of the model under these tests were approximately within the allowable limit. Analysis of the results of the pushover test reveals that the model structure has the overall displacement ductility ratio of 2.4 and the overstrength coefficient of approximately 8.7. The evaluation of the accuracy of analytical simulation by IDARC-2D leads to the conclusion that while global and local behaviors can be, in general, simulated with limited accuracy in the dynamic nonlinear analysis, it is easy to obtain a fairly high level of accuracy in the prediction of global behavior in the static nonlinear analysis. 2002 Elsevier Science Ltd. All rights reserved. Keywords: Reinforced concrete; Earthquake simulation test; Pushover test; Nonlinear analysis; Overstrength; Nonseismic detailing 1. Introduction Recently, minor earthquakes have occurred over 20 times a year in Korea. The earthquake of December 13, 1996 at Yeongweol in Korea, which is known to have a magnitude of 4.5 on the Richter scale imposed signifi- cant nonstructural damages [1]. These recent earth- quakes indicate that the Korean Peninsular is no longer safe from seismic hazard. If a severe earthquake such as the 1995 Kobe earthquake should occur in Seoul, the damage would be tremendous. Most building structures in Korea, which are normally medium- to low-rise reinforced concrete (RC) frames, * Corresponding author. Tel.: +82-2-3290-3337; fax: +82-2-921- 7947. E-mail address: [email protected] (H.-S. Lee). 0141-0296/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII:S0141-0296(01)00135-3 have not been engineered to resist major or moderate earthquakes. Therefore, should any major earthquake occur, the damage or collapse of not only general com- mercial buildings, but also public-service buildings such as police offices, communication centers and hospitals, would implement very large life and economic losses as well as cause the critical interference with the function of the nation. Several researchers in the eastern and central United States have already performed research, including earth- quake simulation tests, on the seismic capacity of gravity–load-designed RC structures [2–5]. However, attention concentrated mainly on the detail practice in North America, particularly according to American Con- crete Institute (ACI) 318 code [6]. The practice in reinforcement detailing and construction in Korea is somewhat different from that of the US, which will be explained in the section on the design of the model.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Engineering Structures 24 (2002) 719–734
www.elsevier.com/locate/engstruct
Seismic performance of a 3-story RC frame in a low-seismicityregion
Han-Seon Lee *, Sung-Woo Woo
Department of Architectural Engineering, Korea University, Seoul 136-701
Received 16 April 2001; received in revised form 7 November 2001; accepted 14 November 2001
Abstract
The objectives of the research stated herein are to investigate the seismic performance of a 3-story reinforced concrete (RC)
ordinary moment-resisting frame, which has not been engineered to resist earthquake excitations, and to evaluate the reliability of
the available static and dynamic inelastic analysis techniques. A 1:5 scale model constructed according to the Korean practice of
nonseismic detailing and the similitude law was subjected to a series of the shaking table motions simulating Taft N21E component
earthquake ground accelerogram. Due to the limitation in the capacity of the used shaking table, a pushover test was performed to
observe the ultimate capacity of the structure after earthquake simulation tests.
The model showed the linear elastic behavior under the Taft N21E motion with the peak ground acceleration of 0.12g, representing
the design earthquake in Korea. The model revealed fairly good resistance to the higher levels of earthquake simulation tests though
it was not designed against earthquakes. The main components of its resistance to the high level of earthquakes appear to be (1)
the high overstrength, (2) the elongation of the fundamental period, (3) the minor energy dissipation by inelastic deformations, and
(4) the increase of the damping ratio. The drifts of the model under these tests were approximately within the allowable limit.
Analysis of the results of the pushover test reveals that the model structure has the overall displacement ductility ratio of 2.4 and
the overstrength coefficient of approximately 8.7. The evaluation of the accuracy of analytical simulation by IDARC-2D leads to
the conclusion that while global and local behaviors can be, in general, simulated with limited accuracy in the dynamic nonlinear
analysis, it is easy to obtain a fairly high level of accuracy in the prediction of global behavior in the static nonlinear analysis.
Recently, minor earthquakes have occurred over 20times a year in Korea. The earthquake of December 13,1996 at Yeongweol in Korea, which is known to havea magnitude of 4.5 on the Richter scale imposed signifi-cant nonstructural damages [1]. These recent earth-quakes indicate that the Korean Peninsular is no longersafe from seismic hazard. If a severe earthquake such asthe 1995 Kobe earthquake should occur in Seoul, thedamage would be tremendous.Most building structures in Korea, which are normally
medium- to low-rise reinforced concrete (RC) frames,
0141-0296/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S0141-0296 (01)00135-3
have not been engineered to resist major or moderateearthquakes. Therefore, should any major earthquakeoccur, the damage or collapse of not only general com-mercial buildings, but also public-service buildings suchas police offices, communication centers and hospitals,would implement very large life and economic losses aswell as cause the critical interference with the functionof the nation.Several researchers in the eastern and central United
States have already performed research, including earth-quake simulation tests, on the seismic capacity ofgravity–load-designed RC structures [2–5]. However,attention concentrated mainly on the detail practice inNorth America, particularly according to American Con-crete Institute (ACI) 318 code [6]. The practice inreinforcement detailing and construction in Korea issomewhat different from that of the US, which will beexplained in the section on the design of the model.
The objectives of this research are; (1) to observe theactual response of this kind of low-rise RC ordinarymoment-resisting frame with nonseismic detailing whensubjected to various levels of earthquake groundmotions, (2) to get the information on the ultimatecapacities (strength, deformability and so on) of thestructure, and (3) to provide the calibration to, or tocheck the reliability of, the available static and dynamicinelastic analysis techniques.Considering the capacity of the shaking table to be
used, the reduction scale for the model was determinedas 1:5. Using the techniques for manufacturing themodel according to the similitude requirementsdeveloped through other researches [7,8], a 1:5 scale 2-bay 3-story RC frame model was constructed. Thismodel was then subjected to the shaking table motionssimulating Taft N21E component earthquake groundmotions, whose magnitude of peak ground acceleration(PGA) was modified to approximately 0.12, 0.2, 0.3 and0.4 g. Before and after each earthquake simulation test,free vibration tests were performed to determine thechange in the natural period and the damping ratio ofthe model. The global behavior and damage pattern wereobserved. The lateral accelerations and displacements ateach story and the local deformations at the criticalregions of the structure were measured. The base shearwas measured using self-made load cells. Since the ulti-mate capacity of the structure could not be found due tothe limitation in the capacity of the shaking table, apushover test was performed to observe this capacity ofthe structure after a series of earthquake simulation tests.Based on all the test results, the interpretation on theresponse of the model is carried out.The computer code IDARC-2D [9], one of the codes
widely used in the world for the nonlinear dynamic andstatic analyses of RC frame structures, was adopted forthe analysis to investigate the correlation between theresults of (earthquake simulation and pushover) tests andanalyses. Although the time history analyses correspond-ing to all the earthquake simulation tests were performed,the correlation of experiment and analysis will be inves-tigated only for the case of earthquake simulation testTFTF04 in this paper.
2. Experiment
2.1. Design of the model
The prototype of this test model was adopted from abuilding structure for the police office, actually built andin use in Korea. The plan and elevation of the 1:5 scalemodel are shown in Fig. 1(a,b). The compressivestrength of concrete, f9c, in the prototype structure isassumed to be 20.6 MPa and the nominal yield strengthof reinforcement, 294.2 MPa. The typical sections of
members and the details regarding transverse steel,anchorage and splice are shown in Fig. 1(c–h).The important characteristics in the Korean detailing
practice are as follows: (1) the splice is located at thebottom of the column, (2) the spacing of hoops is rela-tively large, (3) seismic hooks are not used, (4) con-finement reinforcements are not used in beam–columnjoints, and (5) the special style of anchorage in the joints.That is, the length of tension and compression anchorageare usually 40 and 25 db respectively, from the criticalsection, where db means the nominal diameter ofreinforcement. Moreover, the length of the tail in thehook is included in this anchorage length and the tailsof the anchorage of the bottom bars in beams usuallydirect downward into the exterior columns as shown inFig. 1(h).
2.2. Model reinforcement and model concrete
It is essential to maintain the similitude in the materialproperties between prototype and model reinforcement.However, it was difficult to make the cross sections ofthe model reinforcement conform exactly to the simili-tude law. So, the yield forces rather than yield stresseswere selected as the target to be achieved in annealingthe model reinforcement. An electric furnace with a 3-zone vacuum tube was designed and used. The defor-mation on the surface of the model reinforcements wasmade using a deforming device. Reinforcing bars D22and D10 in the full-scale structure match D3 and D2 inthe model, respectively. The target yield forces derivedfrom similitude requirements are shown in Table 1. Theachieved average yield force of 5.0 kN is approximately10% less than the target yield force in the case of D3.The model concrete was made using type I Portlandcement. The average strength of the model concrete atthe time of testing was about 21.6 MPa.
2.3. Instrumentation and experimental setup
The used shaking table is 3×5 m with one degree offreedom. The data were acquired simultaneously at therate of 300 Hz in 32 channels. Displacement transducers,accelerometers and load cells were used to measure thelateral displacement and the angular rotations in someends of beams and columns, acceleration at each story,and shear forces on the columns of the first story. Fig.2(a) and Plate 1 show the view of the experimental setupand instrumentations for the shaking table test. Theinstrumentation was confined to only one frame due tothe limitation in the number of available channels exceptthe load cells at the first-story columns. To measure storydrifts, a reference frame was used as shown in Fig. 2.The white-noise (random-vibration) test, which was per-formed before the earthquake simulation tests, indicatedthat the natural frequency of the reference frame is
B19; (g) Section B2–B29; and (h) Anchorage detail of beam bars in exterior joint.
approximately 40 Hz. Therefore, the reference framewas considered to have the sufficient rigidity to measureaccurately the drifts of the model. The load cells weredesigned and manufactured following the reports of El-Attar et al. [4] and Bracci et al. [2] and calibrated byusing a universal testing machine (UTM).The volume of the model is reduced to 1/53 of the
prototype while the similitude law, premising the agree-ment in the stress–strain relation of the material betweenthe prototype and the model, requires the mass to be
reduced to 1/52 [10]. Therefore, compensation for thedifference in the mass was artificially made by addingconcrete blocks as shown in Fig. 2(a). The effectiveweight of the model with concrete blocks is estimatedto be 100.6 kN while the weight ideally required by thesimilitude law is 96.1 kN. The error is approximately5%. The hinges between the model and the concreteblocks were designed to transmit only the vertical andhorizontal forces, excluding the moments, due to themass of the concrete block.
Fig. 2. Instrumentation and experimental setup. (a) Earthquake simulation test; and (b) Pushover test.
The experimental setup for the pushover test is shownin Fig. 2(b). Steel plates, each of which has the dimen-sion W×D×L=9×4×35 cm (0.097 kN) were used as theartificial mass. The effective weight of the model withsteel plates is estimated to be 100.7 kN. The roof driftwas obtained by averaging the two measured values inboth Frame A and Frame B. Two different data acqui-sition systems were used due to the limitations in thenumber of available channels in the data acquisition sys-
tems, but the measured data were synchronized witheach other’s system by comparing the displacements oftransducers T5 and T6 installed at the same location inFig. 2(b) but belonging to the different data acquisitionsystem. The experimental results were interpreted,assuming that the behavior of Frame A represents thatof the whole model structure in both earthquake simul-ation test and pushover test. The displacements at thesecond floor (T2) and third floor (T3) were those meas-
Plate 1. Experimental setup for shaking table tests.
ured in the middle of both frames, and the lateral forcedistribution was maintained in the shape of an invertedtriangle by using the whiffle tree.
2.4. Experimental program
Since there was no recorded strong-motion accelerog-
ram as of the time of this experiment in Korea, the inputmotion to the shaking table was derived from the
recorded Taft N21E component by adjusting the peak
ground acceleration (PGA) and compressing the timescale by the factor of Î5 according to the similitude law[10]. The most important reason for this selection was
that the test results in this study can be compared easilywith those of several previous earthquake simulation
tests [2–4,11] which used the same accelerogram, and
the other one is that damage potential of Taft N21Ecomponent accelerogram seems to be relatively high.
The design earthquake defined in Korean seismic code
[12] is the event expected in the recurrence period of475 years (the probability of exceedance of 10% in 50
yr). The elastic response spectra of the input Taft N21E(PGA = 0.12 g) motion and the output table motion are
shown in Fig. 3. From this figure, it can be found that
Fig. 3. Response spectrum for input and output table motions and
design spectra compressed by the scale of 1:Î5 (soil factor S 5 1.2).
the fidelity of the shaking table is satisfactory. The com-
parison of the design spectrum according to the Koreanseismic code [12] with the response spectrum of the out-
put motion indicates that the dynamic amplification
implicit in the design spectrum of this code is underesti-mated in the range of short periods whereas it is overesti-
mated in the range of long periods. The final design
spectrum by UBC [13] is quite similar to the Koreandesign spectrum though the used R factors are different.
The program of tests is shown in Table 2. Each testhas the significance stated in the column containing
remarks. For earthquake simulation tests, the model
structure did not show serious damage even after testTFTF04. However, due to the limitation in the capacity
of the shaking table, it was impossible to implement a
higher level of earthquake simulation test. Therefore, inorder to get the information on the capacities (strength,
deformability and so on) of the model structure, a push-
over test (or monotonically-increasing lateral-load statictest) was conducted.
3. Results of earthquake simulation tests and
interpretation
3.1. Global responses
Free vibration tests were performed by enforcing aninitial lateral displacement of approximately 1–2 mm at
the roof of the model, and releasing. From these tests,the natural periods and damping ratios of the model were
obtained by using the Fourier transform and logarithmic
decrement method. Table 3 shows that natural periodsand damping ratios tend to increase as the model experi-
ences higher levels of ground motions.
The drifts at the roof were measured at two locations.The test results indicate that almost negligible torsional
behaviors occurred in this model for both earthquake
simulation and pushover tests. Therefore, the measureddata in one of the two plane frames are assumed to rep-
resent the behaviors of both frames. Table 4 summarizes
the measured maximum response quantities. Time his-tories of the floor accelerations for TFTF04 are given in
Fig. 4, which indicates that the response has the pre-
dominant period approximately equal to the period ofthe first mode of the model.
Table 4 shows the maximum interstory drift indices
(I.D.I.) at the time when the roof undergoes themaximum drift. For the design earthquake of Korea
(TFTF012), the model has a maximum I.D.I. of 0.26%which is much less than the maximum allowable value
of 1.5%. However, the maximum I.D.I. for the test of
TFTF04 appears to be 1.68%. Therefore, regarding driftcontrol, the behavior of the model appears to be satisfac-
tory. From the profiles of measured drift and the
interstory drift indices in Fig. 5, it can be noted that the
connected columns (R6 and R7). On the contrary, the
angular rotations at the beams (R1 and R2) are generallymuch less than those at the columns (R3 and R4) in the
case of the interior joint.
Fig. 8 shows the distribution of shear forces in thecolumns. It is interesting to note that column (1) and
column (3) have a directional bias in shear forces due
to the influence of the axial compressive force on theshear resistance.
4. Correlation of analytical and experimental
dynamic responses
4.1. Analytical model
The computer code IDARC-2D [9], one of the codes
widely used in the world for the nonlinear dynamic and
static analyses of RC framed structures, was adopted.This code utilizes a global Takeda-like model. The
objective of this analytical study is to find the mostappropriate values of parameters in the analysis byIDARC-2D to simulate the responses given by experi-ments, and then to evaluate the degree of accuracy inthe obtained simulations. Eventually, this evaluation ofreliability for IDARC-2D will lead to the more carefulinterpretation of analysis results for other types of RCframe structures such as schools, hospitals, and the like.The material models used to derive the relation
between moment and curvature at critical sections areshown in Fig. 9. The program RESPONSE, developedby Felber and Andreas [14], was used to obtain theenvelope curve for the moment–curvature relations, asdepicted in Fig. 10. The effective width (410 mm) ofACI code 318-95 [6] was used to model the relationbetween the curvature and the moment for the time his-tory analyses.The hysteretic model incorporates stiffness degra-
dation (HC, a), strength deterioration (HBD, HBE, b),non-symmetric response, slip-lock (HS, g), and a trilin-ear monotonic envelope. The model traces the hysteretic
Fig. 8. Time histories of column shears under TFTF04.
behavior of an element as it changes from one linearstage to another, depending on the history of defor-mations. For a complete description of the hystereticmodel, refer to Park et al. [15].Aycardi, Mander, and Reinhorn [5] used a 5 0.5,b 5 0.04, and g 5 0.7 based on the experimental testresults for elements. Stiffness degradation is severer inthe model than in the prototype, and an α value between0.5 and 1.0 has been used in the analytical model, whichis scaled as 1/4 or below [9]. In this study, the samevalue of hysteretic parameters as adopted by Aycardi etal. are used to simulate the test results.The artificial weight (concrete blocks in this case)
loaded to compensate the mass according to the simili-tude law acts as concentrated loads on the girders.Hence, the girders are divided into three elements whichalso properly represent the change in the reinforcementof the sections.
4.2. Global responses
The parameters determining hysteretic behavior in theM–f relation have actually been adjusted to simulatemost closely the response, particularly the roof drift.Though the roof drift history shown in Fig. 11 is thebest simulation among all the trials, there is still a dis-crepancy of 4.11 mm (about 14%) in maximum values,and in phase in the latter part. Though the peak drift atthe roof in the analysis is smaller than in the experiment,the peak response acceleration at the roof of the analysisappears to be a little larger than that of the experiment.Nevertheless, the analysis reveals a smaller maximumbase shear than the experiment in Fig. 12. This discrep-ancy in the maximum base shear between analysis andexperiment can be attributed to the stronger effects ofthe second or higher mode in the analysis. In general,the histories of story drifts are similar in shape in bothcases of analysis and experiment, indicating that the firstmode governs, though the peak values differ.
4.3. Local responses
The time histories of the column shear in Fig. 13imply that the analysis could not simulate the bias in theshear force caused by the increase of shear stiffness dueto the increase of the axial compressive forces in col-umns. Fig. 14 compares the time histories of angularrotations at the bottom of the second-story columnobtained from both the experiment and analysis. Theangular rotations were calculated by multiplying the cur-vature at the end of members in analysis with the samelengths over which the rotations were measured in theexperiment. Generally, the magnitude of the angularrotation in the analysis is much smaller than that in theexperiment. The discrepancy in magnitude is considered