7shaking Table Tests Of 1-4 Reduced-scale Models Of Masonry Infilled Reinforced Concrete Frame Buildings

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS Earthquake Engng Struct. Dyn. 2001; 30:819–834 (DOI: 10.1002/eqe.39)

Shaking table tests of 1:4 reduced-scale models of masonry in lled reinforced concrete frame buildings  Roko Zarni c1;∗;† , Samo Gostic2 , Adam J. Crewe3 and Colin A. Taylor 3 1 University

of Ljubljana; Faculty of Civil and Geodetic Engineering; 1001 Ljubljana; Jamova c. 2; P.O. Box 3422; Slovenia 2 Civil Engineering Institute ZRMK; DimiÄ ceva 12; Ljubljana; Slovenia 3 University of Bristol; Department of Civil Engineering; Earthquake Engineering Research Centre; Queen’s Building; University Walk; Bristol BS8 1TR; U.K.

SUMMARY Two models of masonry in lled reinforced concrete frame buildings were tested at the shaking table. Models were built in the reduced scale 1:4 using the materials produced in accordance to modelling demands of true replica modelling technique. The rst model represented a one-storey box-like building and the second one the two-stories building with plan shaped in the form of a letter H. Models were shaken with the series of horizontal sine dwell motions with gradually increasing amplitude. Masonry in lls of tested models were constructed of relatively strong bricks laid in weak mortar. Therefore, typical cracks developed and propagated along mortar beds without cracking of bricks or crushing of in ll corners. Data collected from tests will be used in future evaluation, veri cation and development of computational models for prediction of in-plane and out-of-plane behaviour of masonry in lls. The responses of tested models can be well compared with global behaviour of real structures using the modelling rules. The similarity of local behaviour of structural elements, e.g. reinforced concrete joints, is less reliable due to limitations in modelling of steel reinforcement properties. The model responses showed that buildings designed according to Eurocodes are able to sustain relatively high dynamic excitations due to a signi cant level of structural overstrength. Copyright ? 2001 John Wiley & Sons, Ltd. KEY WORDS:

∗ †

in lled frame; building model; true replica; shaking table; non-linear response

 Correspondence to: Roko Zarni c, University of Ljubljana, Faculty of Civil and Geodetic Engineering, 1001 Ljubljana, Jamova c. 2, P.O. Box 3422, Slovenia. E-mail: [email protected]

Contract=grant sponsor: British Council; contract=grant number: ALIS Link No. 20 Contract=grant sponsor: Ministry of Science & Technology, Republic of Slovenia; contract=grant number: J2-7659792-96 Contract=grant sponsor: European Economic Community; contract=grant number: Supplementary agreement ERBCIPDCT 940089 to the main contract number CHGECT 920010.

Copyright ? 2001 John Wiley & Sons, Ltd.

Received 25 November 1997 Revised 23 May 2000 Accepted 26 September 2000

820

 R. ZARNI C ET AL.

INTRODUCTION The reinforced concrete frames with masonry in lls is the structural system that still attracts many research eorts. The experience gained from the 1999 earthquakes in Turkey [1] shows that irregular distribution of in lls, neglecting the interaction between the frame structure and in lls and weakness of structural components may cause the collapse of the entire buildings. The structures may even be designed according to the best practice, but the overall response to seismic action is still not enough understood. The real capacity of these structures and their ability to withstand moderate and strong earthquakes needs to be evaluated using accurate models for predicting the behaviour of real structures subjected to in-plain and out-of-plain loads. The extensive experimental and analytical research on masonry in lled concrete frames has been carried out worldwide in recent 50 years in order to establish design procedures that would realistically predict structural behaviour during an earthquake. Dierent models have been developed and veri ed mainly using the results of static, cyclic or pseudo-dynamic tests. However, the complete understanding of behaviour of structures can be obtained only through taking into account the dynamic nature of their seismic response. Learning from earthquakes can be well combined by learning from shaking table testing of structural models. Therefore, extensive collaborative research on in lled frames is in progress on several universities and institutes in Europe. The part of it is shaking table testing of reduced scale models of buildings constructed of masonry in lled reinforced concrete frames. The framework of the European Consortium of Earthquake Shaking Tables [2] and European Commission support to researchers from Central and Eastern European countries (PECO) made possible to perform tests that are described in this paper. The tests were carried out on shaking table at University of Bristol, UK within the collaborative research project with University of Ljubljana, Slovenia. The main objective of paper is to present the testing method. The aim of experimental research was to get more information about the behaviour of masonry in lled frames exposed to seismic ground motion since there is still a lack of the available experimental data in this topic. PREVIOUS AND CURRENT RESEARCH Some valuable information on dynamic behaviour of in lled frames can be obtained from research performed and published in the last decade. The signi cant problems associated with dynamic testing of scaled models that were noti ed by researchers concern the modelling scaled material properties, the nancial restrictions leading to testing of limited number of specimens and limitations of specimen size due to capacity of available shaking tables. That is mainly the reason for relatively small number of executed and published experiments of this kind. The most signi cant of them are brie y summarized including some of ongoing research that will be published in the near future. Manos et al. [3] reported on two models of two-storey reinforced concrete buildings with masonry in lls that were constructed in 1:9 scale and tested on shaking table with N–S El Centro 1940-based scaled excitation along the plane of in lls. The mechanical characteristics of materials used for construction of models corresponded to the characteristics of weak prototype materials. The 1:3 and 1:9 scaled models of one-bay one-storey in lled frames were also tested by cyclic loading to study the in uence of scaling on structural response. The authors concluded Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

821

that the correctness of small-scale simulation was proved in terms of envelope curves and thus the shaking table test results can be used for development of computational models. Kwan and Xia [4] tested on shaking table three the third-scale four-storey models of three dierent types of structures: reinforced concrete shear wall structure, masonry in lled reinforced concrete structure and concrete in lled steel frame structure. Models were constructed of materials with almost prototype characteristics. Therefore, heavy loading had to be applied atop of each oor. Models were excited along the in-plane direction by the loading based on the N–S El Centro 1940 record. The main conclusions obtained from tests were related to general behaviour of structures including development of failure mechanisms and to the in uence of damages on changing of dynamic characteristics of the tested models. The main conclusion was that more tests are needed for better understanding of behavioural mechanisms and that the integrity of in lled frame structure can be very in uenced by out-of-plane collapse of in lls. Klinger et al. [5] tested two sets of half-scale models of the so-called “weak frames” and “strong frames”. Models were constructed of prototype materials and therefore relatively highly loaded with added masses and axial post-tensioning of columns. The rst set was intended to represent buildings designed by the 1956 ACI Codes and the second one the buildings designed by the 1989 ACI Codes. The specimens, consisted of two parallel one-storey onebay frames, were tested separately in plane and out-of-plane on shaking table. Initially, a bare frame specimen was tested in plane until certain level of damage was produced. The frames were in lled afterwards and again tested in plain until the in lls cracked. Finally, one in lled frame was tested by out-of-plane excitation until sever cracking occurred. In the conclusions of paper it is clearly stated that “the test results con rmed previous indications from quasi-static tests that in lled frames can signi cantly increase the stiness, strength, and energy dissipation capacity of framed structures, even under conditions involving simultaneous in plane and outof-plane inertial forces”. This conclusion is showing that knowledge gained from static and pseudo-dynamic tests can be well combined with knowledge gained from dynamic tests in process of developing of computational models. Vintzileou et al. [6] contributed the important experimental data on the behaviour of sixstorey reinforced concrete frames with masonry in ll designed according to Eurocode 8 [7]. Four test specimens were constructed in 1:5.5 reduced scale from materials having modelled mechanical properties following the true replica modelling rules. The single frames were excited in-plane with base accelerations modelled after the 1940 N–S El Centro record by increasing intensity in subsequent test runs. Between test runs the dynamic characteristics were determined by random vibration tests. The model structures were observed to behave in a similar manner as expected for a full-scale structure. Comparing the design and achieved base shear it was found that the overstrength in average magnitude of 3 was achieved. The overstrength is a phenomenon that has been widely observed in laboratory testing of models. Several have identi ed the factors contributing to the overstrength such as material overstrength, capacity design philosophy, minimum construction requirements, slab contribution, better workmanship that assumed, near to ideal conditions for construction of models in laboratory environment, etc. However, some of these parameters can also in uence the overstrength of real structural system what might explain some cases of dierent behaviour of equally designed structures during an earthquake. The research presented in this paper is continuing in the mainframe of the European project where six universities and institutes are studying dierent aspects of problems of in ll frame Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

822

 R. ZARNI C ET AL.

Figure 1. Overall view of instrumented specimens.

testing and modelling. Two 1:4 reduced scale models have recently been tested at shaking table of University of Bristol, U.K. by researchers from University of Ljubljana, Slovenia [8]. These models of three-storey two-bay in lled frame buildings constructed according to true replica modelling rules were tested by excitation based on Tolmezzo, Italy, 1976 record. The data obtained from tests proved to be of great importance for evaluation and veri cation of previously developed inelastic computational model. Another model was tested on shaking table of ISMES Institute in Seriate, Italy [9]. It was constructed in the form of two-storey single-bay full-scale building. The intention of test was to study the in uence of polymeric grid laid in horizontal mortar joints of masonry in ll on the behaviour of in lled frame structure. The detailed analysis of test results is still in progress. It is obvious that the interest for dynamic testing of in lled frame models is growing and that more needed data on their behaviour under earthquake loading will be available in future. DESIGN OF BUILDING MODELS Two models of simple buildings were constructed following the rules of true replica modelling. It means that not only dimensions but also material properties were scaled following the reduced scale 1:4. The rst model named “B” represents a box-type one-storey building. The second model named “H” was built as a two-storey building with plan in form of letter H. Two parallel two-bay two-storey in lled frames connected by two-storey single-bay in lled frame formed the structural system. The con guration of models (Figure 1) enables study of both in-plane and out-of-plane behaviour of in lls on the same model. Limitation in designing of models was the shaking table dimensions and capacity. Models were designed by scaling down the geometric and material properties from prototype structure. Design gravity loads of 0:5 kN=m2 (dead) and 2 kN=m2 (live) was taken into account. The prototype structures were designed according to Eurocode 8 as medium ductility class structures taking into consideration the design peak acceleration of 0:3 g and design q-factor of 3.75. In selection of the design q-factor the status of structural irregularity as well as particular structural system were taken into account. The additional design measures for in lled frames as described in Eurocode 8 were also implemented. Critical Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

823

MASONRY INFILLED RC FRAME BUILDING TESTS

Table I. Geometric properties of models. Property (Units: mm)

Model B

Storey height Longitudinal frame span Transversal frame span Beam=column dimension Main reinforcement Reinforcement ratio Thickness of slab Thickness of in ll

1380

640

730 80=80 4? 6 mm 1.77% 40 30

Prototype H

B

730

5440

Scale H

2560

2920

2920 320=320 4? 32 mm 3.14% 160 120

4 4 4 4 5.3 1.8 4 4

regions of columns where shear reinforced by stirrups following the rules related to masonry in lled reinforced concrete frames. Longitudinal reinforcement was anchored in steel plates on the face of beam–column joints. This solution was chosen considering uncertainties related to relatively small cross sections of joints and properties of modelled materials: microconcrete and model ? 6 mm mild steel bars. Transversal beams and slabs were caste after casting of main frames. Therefore, the reinforcement was left out from beams of longitudinal frames to provide connection with slab and transverse frame’s beams. The connecting reinforcement was made of two U-shaped bars for beam connections and in form of steel mesh to provide anchorage of slabs to the frame beams. Steel mesh made of wires ? 2 mm on 19 mm was used for reinforcement of the slab. Reinforcement bars were placed through holes in the steel plates and welded. In lls were considered to be light dividing walls in prototype structure. The principles of true replica modelling are generally known [10] and therefore there is no need for further explanation of scaling theory. In the case of herein reported testing of dynamic behaviour of masonry in lled reinforced concrete frames, the minimal requirements were: • Preservation of Newton’s law of inertial forces on prototype and model, • Use of adequate stress–strain relationship, i.e. Hook’s law in elastic range of all materials. In Table I are presented the main dimensions of models and compared with dimensions of prototype constructions. The selection of model reinforcement was dictated by the use of S220 6-mm bars that are available on market. They represent S400 32-mm bars of prototype structure. The scale factors for reinforcement diameters and reinforcement ratio were derived according to rules of true replica modelling. The only exception from strict rules of modelling was made in scaling of mild steel reinforcement (Table II) because of the well-known problems associated with steel reinforcement modelling. According to experiences obtained by structural modelling, it is far more important to target the appropriate reinforcement-to-concrete bond than to scale the elastic stress in the reinforcement. Another problem that appeared in construction of models was the relatively high strength of microconcrete. However, the properties of masonry were satisfactorily attained, as it can be seen from Table II. Masonry bricks were made from a mixture of ne-grained clay rubble, Portland cement, lime and water. Model mortar was a mixture of quartz sand, Portland cement, lime and water. It was decided to model the masonry using relatively weak mortar and strong blocks. All-over properties of in lled frames were comparable to prototype ones because all main parameters were properly scaled according to possible properties that can be found Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

824

 R. ZARNI C ET AL.

Table II. Modelled and prototype material properties. Material property (units: kg=m3 , MPa)

Modelled value

Prototype value

Scale

Masonry Compressive strength of brick Compressive strength of mortar Compressive strength of masonry Compressive strain at max. stress Ref. tensile strength of masonry Density of masonry Masonry module of elasticity Masonry shear modulus

6.06 0.46 2.0 0.38% 0.11 1600 590 310

24.24 1.84 8.0 0.38% 0.44 1600 2360 1240

4 4 4 1 4 1 4 4

Concrete Compressive strength Compressive strain at maximal stress Modulus of elasticity Density

13.9 0.09% 11700 1900

55.6 0.09% 46800 1900

4 1 4 1

Main reinforcement of frame Yield strength Yield strain Modulus of elasticity Bar-to-concrete bond strength

250 0.15% 200 000 0.24

550 0.15% 200 000 0.71

2.2 1 1 2.97

Table III. Scale factors of parameters related to response of tested models. Parameter Force Mass Displacement Time Velocity Acceleration Frequency Scale

64

64

4

2

2

1

0.5

in real structures. The comparison of modelled and corresponding prototype characteristics of materials is given in Table II. The parameters that are associated with response of structure were also modelled following the true replica modelling as presented in Table III. TEST PROCEDURE Tests were carried out as a series of exploratory tests and main test runs on the shaking table at the University of Bristol in December 1995. Exploratory tests were performed as low-level random inputs before each main test run to measure the model resonant frequencies. During the main test runs models were shaken with single-direction, horizontal two sines dwell motions with frequencies approximately 10 per cent lower than were models’ rst resonant frequencies. This means that during each test run, dierent structure has been tested with dierent simulated earthquake, but with the similar resonant eect. This simulates the circumstances that may occur in the extreme natural situation when the earthquake is followed by strong aftershocks as it happened during Turkey 1999 earthquakes. The rst test run was performed with low Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

825

Figure 2. The normalized envelope of two sine dwells.

acceleration of 0:1 g to check the testing procedure and measuring equipment. Further values of amplitudes were chosen according to observations of model behaviour at previous test run and approximation of response in the next test run. The sinusoidal signal and its envelope used for motion control of the shaking table are shown in Figure 2. The signal was composed of three connected parts with acceleration amplitudes in magnitude ratio 0.5:0.25:1.0 and in numberof-cycles ratio 20:10:20. The duration of each next test increased because the frequencies of sine dwell motions decreased. After transition of models from elastic to inelastic response, the changing of their resonant frequencies caused resonance eect with the excitation frequencies. The Model B was loaded with three dierent loading schemes and Model H with one loading scheme. Two lead blocks of 25 kg xed atop of each storey of models represented the 1:64 scaled live load of prototype structure. The second loading scheme of Model B was equal to the rst with addition of steel plate having 1000 kg mass. The third loading scheme included two more steel plates of 1000 kg each. The rst, lowest steel plate was xed on steel tubes, which were welded on steel anchor plates atop of the columns. The other two plates were xed to the rst one (Figure 3). Exploratory test proved that steel tubes had no in uence on dynamic characteristics of tested building model. Additional steel plates in second and third loading scheme were needed to obtain inertial forces high enough to crack the masonry in lls and reinforced concrete frames. They also generated the additional amount of vertical compressive stress in the frame columns. Considering scale factors, the mass of each 1000 kg steel plate was equal to the vertical load caused by four additional stories. The Model B with added masses in form of 1000 kg steel plates was considered a SDOF system representing the rst storey of in lled frame structure that was partly damaged by an earthquake. To obtain complete data about the excitation of the model, the shaking table accelerations have been measured in all three translational directions, as it is shown in Figures 1 and 4. It is clear that some spurious motions appeared although the table was excited in single direction. This is a common problem of shaking table tests caused by imperfection control systems, although it is usually not reported. However, the spurious motions should be taken into account when the tests results are used for validation of analytical models. The currently available models are not yet suitable for three-dimensional analysis of in lled frame structures. The data on spurious motions of table obtained from herein-described tests may be helpful for future developing of 3D computational models of in lled frames. Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

826

 R. ZARNI C ET AL.

Figure 3. Southeast views of Model B (left) and Model H (right) loaded with lead blocks.

Figure 4. Envelopes of shaking table acceleration time histories in X (A1), Y (A2) and Z (A3) direction measured during test run B#11 (left) and test run H#18 (right).

During testing of models, it was also observed that the programmed input acceleration amplitudes dier from measured accelerations of shaking table what was observed in the case of frequencies. During the elastic behaviour of model, the dierences between input and output amplitudes were relatively small. During the inelastic behaviour of model, the dierences were moderate and very signi cant during test runs when models transit elastic to inelastic response. That was observed during test run B#26 of model B (Figure 10) and test run H#8 of model H (Figure 15) when the drastic change in models’ behaviour occurred. Extensive damages developed both in in lls and in frames what resulted in great alternation of model stiness during dynamic excitation. The dierences between the programmed input and actual amplitudes of shaking table accelerations occurred due to limited abilities of shaking table control system. The 1980s control system that was in use at Bristol in time of testing in lled frame models relied heavily on the feedback of displacement from LVDTs in the actuators, even though they are known to Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

827

Figure 5. Excitation frequencies and accelerations of shaking table as recorded during testing of model B.

Figure 6. Excitation frequencies and accelerations of shaking table as recorded during testing of model H.

become unreliable at high frequency. Recently, the new controlling system has been installed [11]. The system is based on minimal control synthesis (MCS) algorithm developed rst by Stoten in 1989 [12] and used successfully by him in process plant control and robotics. The advantage of new control of shaking table is in the successful combination of accurate feedback of displacements at low frequencies with the accurate feedback of acceleration signals at high frequencies, to produce a composite measure of displacement, which is accurate across the 0–100 Hz frequency spectrum. The new system solves the above discussed problems with spurious motions of shaking table and dierences between programmed input and actual shaking table motion. The frequencies and acceleration amplitudes of driving signals were chosen separately for each test run as it can be seen from diagrams on Figures 5 and 6. Responses of models were measured with accelerometers (SETRA) and magnetic transducers (INDIKON). Transducers were mounted on sti steel frame that was xed on shaking table. Natural frequencies of sti steel support frame were about four times higher than initial natural frequencies of models. The diagonal deformations of in lls were measured by LVDT transducers that were stretched over 1=3 length of diagonal span of in lls. Strain gauges were glued on longitudinal 6 mm-reinforcement bar 50 mm from column or beam-ends. The layout of instruments is shown on Figure 1. Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

828

 R. ZARNI C ET AL.

Figure 7. Alternation of rst natural frequencies, peak accelerations and damping of model B.

Figure 8. Comparison of shaking table (A1) and model (A8) acceleration time history (left) and displacements (D1) of upper joint of southern in lled frame vs acceleration (A8) (right) as recorded during the test run B#11.

TEST RESULTS Behaviour of Model B The dynamic characteristics of model excited on shaking table are summarised in Figure 7 where the response frequencies, response peak accelerations and damping of model are presented. The rst natural frequencies and damping were obtained from exploratory tests that preceded main test runs. The peak accelerations measured atop of model were obtained from acceleration time histories recorded during main test runs (Figures 9 and 10). Testing of model B was carried in three series each characterized with dierent mass attached atop of model. During the rst series, the cracks mostly developed in masonry in lls. The initial cracks appeared in longitudinal in ll during test run B#6 when response of the model was still in elastic range (Figure 8). Major cracks and damages of in lls developed during test runs B#17 and B#19. Spurious motions of shaking table and structural imperfection of model due to nonsymmetric development of cracks in in lls caused the rotation of model around its vertical axis. Consequently, during the test run B#19 when peak acceleration of 2.7 g was attained, the Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

829

Figure 9. The crack and damage patterns as developed in the rst 19 test runs.

Figure 10. Comparison of shaking table (A1) and model (A8) acceleration time history (left) and displacements (D1) of upper joint of southern in lled frame vs acceleration (A8) as recorded during the test run B#26.

upper parts of out-of-loaded in lls started falling out. The pattern of cracks developed during this test run is shown in Figure 8. Although the damages of in lls were serious, the frame structure behaved almost elastically. Therefore, it was decided to proceed tests with lower excitation accelerations and added mass of 1050 kg to develop damages in frame structure and to examine its inelastic behaviour. The rst natural frequencies of model started to decay faster after the test run B#19 due to added load and further development of damages in in lls and plastic hinges in the joints of frames. The major damage occurred during test runs B#23, B#26 and B#28 when the excitation accelerations were in range between 0.6 and 1.0 g and excitation frequency in range between 21 and 12 Hz. During the test run B#26, damages lowered the rst natural frequency of model to the level of excitation frequency. Therefore, the resonant eect developed and caused the dierences between excitation accelerations of shaking table (A1) and response acceleration of model (A8) as shown on Figure 10. Consequently, the model properties dramatically changed due to sudden lowering of its stiness (Figure 10). The third series of tests were carried out with model loaded with 3050 kg mass to nd out the ultimate load-bearing capacity of in lled frames. The intention of this test series was to obtain data on behaviour of heavily damaged structure, which was afterwards repaired and strengthened. Strengthened model was retested in order to study the eect of strengthening of frames by jacketing [13]. The rst natural frequencies of model strongly decreased due to Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

830

 R. ZARNI C ET AL.

Figure 11. Comparison of predicted and experimentally obtained relations base shear vs storey drift of model B.

Pushover prediction: M0 - bare frame model B M1 - inÿlled frame model M2 - inÿlled frame model M3 - inÿlled frame model

- 50 kg added mass B with 50 kg mass B with 1050 kg mass B with 3050 kg mass

Experimentally obtained response of Inÿlled frame model B with:

50 kg added mass atop of model  1050 kg added mass atop of model  3050 kg added mass atop of model

added mass (Figure 7) and slightly decayed due to development of damages in frame joints during test runs B#32 to B#44. The behaviour of model during shaking table tests was analysed by comparison of attained base shear forces and diagrams of predicted storey resistance (Figure 11). They were obtained by pushover analysis of model using DRAIN-2DX program [14] that has been upgraded by two new elements for inelastic structural analysis of in lled frames [15]. The data on material mechanical properties that were used in computational prediction were obtained by testing of materials used for construction of model. According to predicted behaviour of tested model, the attained base shear forces during the rst series of tests (B#5–B#19) were still in elastic range concerning the reinforced concrete frames in spite of crack development in masonry in lls. The response of model during second series of test runs was near to prediction. From diagram it is clearly seen that during the test run B#26 and B#29 were reached higher base shear forces than predicted. During these, two tests majority of heavy damages developed both in longitudinal in lls and in frames. During the third series of tests (B#32–B#44) the deformation of frames were less constrained by in lls due to extensively developed damages. Therefore, the structure of model behaved almost as bare frame structure. The experimentally reached base shear forces were higher than expected for bare frame probably due to the in uence of damaged in lls and overstrength of frames. The overstrength of structural models as well as prototype structures is common phenomenon observe also by other researchers as mentioned before in brief review of published research. Nevertheless, it is very dicult to distinguish between the in uence of the overstrength phenomenon and the level of accuracy of computational model. The observed level of similarity between predicted and experimentally obtained values may be judged as satisfactory taking into account the variety Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

831

Figure 12. Alternation of rst natural frequencies, peak accelerations and damping of model H.

Figure 13. The crack and damage patterns as developed in the rst 18 test runs.

of problems associated with testing of scaled structures and mathematical modelling of in lled frame structures where the large scatter of material properties is one of the main problems. BEHAVIOUR OF MODEL H In Figure 12 are presented the response frequencies, response peak accelerations and damping that was measured during testing of model H. Testing was carried out with only one con guration of masses attached atop of each oor representing the scaled actual live load. The rst natural frequency of model deteriorated during development of cracks and damages in masonry in lls and frames. The main development of damages occurred during test runs H#8, H#10 and H#13. During later test runs the deterioration of the rst natural frequency stabilized because most of the damages developed in previous runs. The natural frequencies and damping was obtained from exploratory test runs that followed the main test runs. During the rst three main test runs, cracks developed in the masonry in lls of the rst storey along the horizontal mortar joints (Figure 13). During later test runs the cracks and damages developed also in the upper storey of in lled frames. The in uence of spurious motions of shaking table is not of the same magnitude as in the case of Model B. The Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

832

 R. ZARNI C ET AL.

Figure 14. Shaking table (A1) and model (A5 and A7) acceleration time history envelopes (left) and displacements (D3) of upper joint of southern in lled frame vs acceleration (A7) (right) as recorded the test run H#6.

Figure 15. Shaking table (A1) and model (A5 and A7) acceleration time history envelopes (left) and displacements (D3) of upper joint of southern in lled frame vs acceleration (A7) (right) as recorded during the test run H#8. Table IV. The acceleration ampli cations in subsequent test runs of Model H. Acceleration ampli cation ratio

Test run H#6

H#7

H#8

H#10

H#13

H#14

H#16

H#18

H#20

A7=A1 A5=A1

3.1 2.3

3.8 2.7

4.1 2.6

4.8 3.9

6.6 5.5

5.4 4.3

3.5 2.9

2.3 2.0

1.8 1.5

pattern of cracks was similar in both longitudinal frames and transversal in lls did not suer out-of-plane damages. The radical change of model response from elastic to inelastic occurred during the test run H#8 when stiness of model changed (Figures 14 and 15). The similar behaviour of model was observed during the next test run H#10 when the lower storey in lls suered additional damages and reinforcement in lower joints of frames started to yield. The intensity of excitation of tested model can be expressed in term of acceleration ampli cation ratio (Table IV). It was calculated from peak values of table accelerations (A1) and model accelerations measured on its rst (A5) and second storey (A7). The highest ampli cations of table accelerations were observed during the test run H#13 due to strongest resonant eects. The resonant eect was also observed during test runs H#10 and H#14. Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

MASONRY INFILLED RC FRAME BUILDING TESTS

833

Figure 16. Comparison of predicted and experimentally obtained relations base shear vs storey drift of model H.

In Figure 16 the relationship between peak base shear and corresponding storey drift attained during particular test run of Model H. These values are compared with diagrams obtained by inelastic pushover analysis of in lled and bare two-storey two-bay frames that were equal to tested ones. The analysis was carried out by DRAIN-2DX as in the case of Model B. The magnitude of the peak base shear force acting on model H recorded during test run H#10 was higher than the predicted base shear capacity of model. That illustrates the overstrength of model regarding to computation prediction. However, the dierence between predicted and experimentally detected base shear was relatively small. Therefore, the computational model can be considered as appropriate. The base shear load obtained by following test runs was in the range between the predicted values for bare and in lled frames. That can be explained as the behaviour of model with strongly damaged in lls where the reinforced concrete frame was carrying the majority of load induced by shaking table excitation.

CONCLUDING REMARKS In general, testing of in lled frame building models constructed in reduced scale following the true replica modelling rules can provide valuable information on overall response of structure and variation of their dynamic characteristics due to damage development. The model response can be well compared with global behaviour of real structures using the modelling rules. The similarity of local behaviour of structural elements, e.g. reinforced concrete joints is less reliable due to limitations in modelling of steel reinforcement properties. Test procedure was following the idea of excitation of partly damaged models with most severe dynamic action induced by resonant eect what can be compared with action of strong earthquake aftershocks. The model responses showed that buildings designed according to Eurocodes are able to sustain relatively high dynamic excitations due to signi cant level of structural overstrength. The usefulness of obtained experimental results was demonstrated by validation of the non-linear computational model developed by authors of this paper that proved its already known reliability. The intention of presented experimental research is to help further research of in lled frame behaviour and development of design methodologies for engineering practice. Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834

834

 R. ZARNI C ET AL. ACKNOWLEDGEMENTS

The shaking table tests were carried out within European Community PECO Programme that made possible to researchers from Slovenia to have access to large-scale facilities within ECOEST=PREC8 framework (grant no. ERBCIPDCT 940089 within ERBCHGEST 9200=0). The research was also supported by the British Council through ALIS Link No. 20 and by the Ministry of Science and Technology of the Republic of Slovenia (grant no. J2-7659-792-96). These supports are gratefully acknowledged. REFERENCES 1. Ansal A, Bardet JP, Bray J, Cetin O, Durgunoglu T, Erdik M, Kaya A, Ural D, Yilmaz T, Youd L. Initial geotechnical observations of the August 17, 1999, Izmit earthquake. A Report of the TurkeyUS Geotechnical Reconnaissance Team, September 3, 1999. http:==www.koeri.boun.edu=earthqk=earthqk.html http:==www.eerc.berkeley.edu=turkey=report.html 2. Severn RT. ECOEST — European consortium of earthquake shaking tables — Overview. Proceedings of the 10th European Conference on Earthquake Engineering, Duma G (ed.). A.A. Balkema: Rotterdam, 1995; 2987 – 2992. 3. Manos GC, Yasin B, Triamataki M. Experimental and numerical simulation of the in uence of masonry in lls on the seismic response of reinforced concrete framed structures. Proceedings of the 5th U.S. National Conference on Earthquake Engineering, EERI, vol. II, Chicago, IL, USA, July 10–14, 1994; 817– 826. 4. Kwan AKH, Xia JQ. Shaking-table tests of large-scale shear wall and in lled frame models. Proceedings of Institution of the Civil Engineers, Structures and Buildings 1995; 110:66 – 77. 5. Klinger RE, Rubiano NR, Bashandy TR, Sweeney SC. Evaluation and analytical veri cation of shaking table data from in lled frames. The Masonry Society Journal 1997; 15(2), The Masonry Society, USA: 33– 41. 6. Vintzileou E, Yong Lu, Zhang GF. Seismic behaviour of multi-storey r=c frames tested on an earthquake simulator. Proceedings of the 6th SECED Conference on Seismic Design Practice into the Next Century, Booth E (ed.). A.A. Balkema: Rotterdam, 1998; 443 – 450. 7. Commission of the European Communities. Eurocode 8 — Design Provisions for Earthquake Resistance of Structures. CEN=TC 250=SG8: Brussels, 1994. 8. Gostic S. Models of reinforced concrete frames with masonry in lls. Ph.D. Thesis, FGG University of Ljubljana, 2000 (in Slovenian). 9. Lozza S. INCO-COPERNICUS EUROQUAKE: towards European integration in seismic design and upgrading of building structures, Task 5-Design and execution of shaking table validation experiments, First phase. ISMES Report, Seriate, Italy, November 1999. 10. Noor FA, Boswell LF (eds.). Small Scale Modelling of Concrete Structures. Elsevier Applied Science Publishers Ltd.: London, 1992. 11. Severn RT. Structural Response Prediction using Experimental Data, The Sixth Mallet-Milne Lecture, Balkema: Rotterdam, 1997. 12. Stoten DP. The minimal control synthesis identi cation algorithm. International Journal of Control 1993; 58(3):685 – 696. 13. Dritsos SE, Vandoros KG, Taylor CA. Shaking table tests on a retro tted, small scale, reinforced concrete model building. Seismic Design Practice into the Next Century, Booth (ed.). Balkema: Rotterdam, 1999; 525 – 533. 14. Prakash V, Powell GH, Campbell S. DRAIN-2DX base program description and user guide ver. 1.10. Report No. UCB=SEMM -93=17, University of California at Berkeley: Berkeley, California, 1993.  15. Zarni c R, Gostic S. Masonry in lled frame as an eective structural sub-assemblage. Proceedings of Workshop on Seismic Design Methodologies for the Next Generation of Codes, Krawinkler H, Fajfar P (eds). Balkema: Rotterdam, 1997; 335 – 346.

Copyright ? 2001 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2001; 30:819–834