DEVELOPMENT OF EUROCODES AND RECENT RESEARCH FOR EARTHQUAKE RESISTANT MASONRY CONSTRUCTION

Transcription

1 DEVELOPMENT OF EUROCODES AND RECENT RESEARCH FOR EARTHQUAKE RESISTANT MASONRY CONSTRUCTION M. Tomaževič Slovenian National Building and Civil Engineering Institute, Dimičeva 12, 1000 Ljubljana, Slovenia ABSTRACT The need for research to provide experimental basis for the requirements of Eurocodes for the design of innovative masonry structures in seismic zones is discussed. Robustness of masonry units and type of masonry bond are important parameters which define the behaviour of masonry walls when subjected to seismic loads. If the units are brittle and the bond between units is not adequate, the known relationships between the strength and ductility properties of masonry walls change significantly. If not avoided or at least taken into consideration properly, the brittleness of units and nonhomogeneity of structural walls may lead to overestimated design resistance values. Consequently, the actual structure may face risk of collapse although it is designed for earthquake loads according to code. In the design of masonry structures, energy dissipation capacity is taken into account by force reduction, behaviour factor q. The shaking table tests have shown that behaviour factor does not depend only on the masonry system, but also on the quality of materials and structural configuration. Therefore, the values of behaviour factor q cannot be simply determined from ductility tests of masonry walls. KEYWORDS: seismic resistance, masonry structures, units, brittleness, masonry bond, structural behaviour factor. 1. INTRODUCTION AND SCOPE OF RECENT RESEARCH In order to eliminate technical obstacles to trade and to harmonise technical specifications, the Commission of the European Community in 1975 decided on an action programme in the field of construction and started the initiative to establish a set of harmonised technical rules for the design of construction works which would ultimately replace the national rules in force in the Member States. Given the mandate by the Commission, CEN, European Committee for Standardisation, is preparing a family of standards for the design of structures, structural Eurocodes, which will replace the national codes at the end of this decade. The initial expectations were optimistic. However, it turned out that the harmonisation of structural design codes is not an easy task. As the situation is now, it will take more than three decades from the decision to harmonise the design codes to the implementation of Eurocodes and official withdrawal of existing national codes. According to recent plans, the years 2007/08 will be the years of withdrawal of national codes in the member states of EU.

2 In different stages of preparation, the philosophy of Eurocodes remained the same, but their form changed. At the last stage, common general rules and requirements to be respected in all member states are given in the harmonised standards, whereas requirements, specific for each country should be specified in the country s National Annex. The National Annex may only contain information on those parameters which are left open in the Eurocode for national choice. However, in no case the complementary information given in the Annex should be in contradiction with the requirements of the basic document. Eurocode 6: Design of masonry structures - Part 1-1: Common rules for reinforced and unreinforced masonry structures (Eurocode 6, 2003) specifies the general design rules for masonry structures. The general policy of Eurocodes that the use of no construction product available on the market should be eliminated but the relevant limits which ensure that the buildings built with such products fulfil the essential requirements of European Directive for Construction Products, especially mechanical resistance and stability and safety in the case of fire, need to be specified, is followed. Consequently, traditional and innovative masonry products and technologies of construction are considered in structural design rules, whereas additional expert consideration should be required by the designer in the case of unusual forms of construction or design conditions. It should be mentioned, however, that structural masonry in Europe is mainly limited to the construction of family, single and two-storey dwellings in unreinforced and confined masonry systems, and masonry infills in reinforced concrete structural systems. Clay is the predominant material. In seismic zones additional rules, specified in Eurocode 8: Design of structures for earthquake resistance - Part 1: General rules, seismic actions and rules for buildings (Eurocode 8, 2003), apply, which in most cases overrule the requirements for the design of masonry structures in non-seismic areas. Most of these requirements are given in a general form, leaving the National Annexes the task to provide specific values of parameters, depending on the country s specific conditions. It is not the aim of this contribution to discuss the contents of Eurocodes. However, in order to better understand the recent need for research in masonry, some of the most important parameters, which are expected to be defined in National Annexes to Eurocode 8, are listed as follows: type of masonry units with sufficient robustness alternative classes of perpend (head) joints in masonry, maximum value of ground acceleration for the use of unreinforced masonry, the values of structural behaviour (force reduction) factor q for different masonry structural systems. These parameters cannot be determined unless systematic experimental research has been carried out. Whereas not so long ago there have been no problems with robustness (or brittleness) of masonry units, and only fully filled head joints have been allowed for the construction of masonry structures in seismic zones, innovation and optimisation of technology of masonry construction introduced new types of masonry units and construction technologies, which have so far not been regulated by the codes. Namely, being pushed by market competition, masonry industry improved the thermal properties of masonry units and developed new, faster and cheaper technologies of construction. As a result of such development, hollow masonry units with very thin shells and webs are produced, and construction technologies are introduced where traditional head joints, fully filled with mortar, are replaced by either ungrouted or partly grouted head joints or mechanical interlocking between masonry units. The innovative proposals have been developed in the 2

3 countries not prone to seismic hazard. However, though not significantly influencing the collapse mechanisms when subjected to gravity loads, these innovations significantly influence the behaviour of masonry structures of all systems in seismic conditions since they reduce the robustness of masonry units (due to thin shells and webs) and homogeneity of masonry walls (due to masonry bond) as structural elements. New methods have been also developed for the verification of seismic resistance of masonry structures. In order to take the advantage of these methods and optimise the resistance of the building without risking the reduction of structural safety against earthquake loads, experimental research is also needed to determine the values of structural behaviour factor q. It is therefore of interest of European masonry industry to provide evidence that newly developed masonry units and construction technologies are suitable for the use in earthquake prone regions. In fact, several experimental research projects have been recently financed by the industry to study the real situation and possibly withdraw the limitations given in the codes or propose possible improvements in the technology. Since unreinforced and confined masonry represent most frequently used masonry construction systems in Europe, masonry industry also showed interest and financed experimental research to provide information regarding the possible reduction of design seismic loads of these masonry construction systems. The results of recent research, carried out at Slovenian National Building and Civil Engineering Institute, and possible conclusions will be summarized in this contribution. 2. ROBUSTNESS OF UNITS According to Eurocode 8 (EC 8), masonry units should have sufficient robustness in order to avoid local brittle failure when subjected to seismic loads. The requirement is obvious, however, the definition and criteria for robustness are lacking. In the previous version of EC 8, the percentage of holes and the minimum thickness of shells in the case of hollow clay units have been limited to 50 % and 15 mm, respectively. As past research and experiences indicate, the requirement would guarantee solid behaviour of units and prevent unexpected behaviour of structural walls due to local brittle failure of units when subjected to lateral seismic loads. In the present version, however, the selection of suitable units according to grouping specified in Eurocode 6 (EC 6) is up to the National Annexes. Group 1 units are represented by solid or almost solid units, whereas in the case of clay units belonging to Group 2, the amount of vertical holes is limited to 55 % of the volume and the minimum thickness of shells and webs to 8 mm and 5 mm, respectively. This means that the shape of Group 2 clay units varies in a wide range from almost solid to highly perforated hollow units. Without adequate quantification and testing it will be impossible to propose a selection. Although the robustness of units has not yet been studied systematically, it recently proved to be an extremely important parameter which governs the seismic behaviour of all systems of masonry construction where hollow units are used. This has been the main conclusion made on the basis of the analysis of test results obtained within the framework of two research projects, recently carried out. The aim of the first project was to study the influence of the amount of reinforcement on the lateral resistance and ductility of reinforced masonry walls. In the second case, however, the effect of different vertical bond types on lateral resistance and ductility of plain masonry walls has been investigated. Recent research results have clearly indicated that design recommendations, developed for the case of masonry walls where no 3

4 brittle behaviour of units is expected, have no validity in the case where the walls fail in a non-ductile mode because of local brittle failure of units. 2.1 Robustness of units and effect of reinforcement In the case of reinforced masonry sufficient anchorage and bond, as well as strength of the units should be provided to utilize the tension capacity of reinforcing steel. To underline the importance of robustness of masonry units in the case of reinforced masonry, the results of a research project, where the suitability of a specially designed clay hollow masonry unit with holes to accommodate vertical reinforcement has been investigated, will be summarized (Tomaževič and Lutman, 1997). A series of tests of reinforced masonry walls, designed to fail in bending, has been carried out. The shape and dimensions (175x290x190 mm - length x width x height) of the units used for the construction of specimens were in conformity with the requirements of EC 6 for Group 2 clay masonry units. The volume of the holes was 44 % of the gross volume of the unit, whereas the thickness of shells and webs was 12 mm and 8 mm, respectively. The ratio between the total thickness of webs in a section and the dimension of the unit across the same section was 28 % which is more than required by EC 6 (16 %). Two groups of four walls with different unit strength have been tested. The amount of vertical reinforcement, placed in the holes at the ends of walls and grouted with concrete, varied as indicated in Table 1. Horizontal bed joint reinforcement was the same in all cases. In addition to reinforced walls, two unreinforced walls of each unit strength class have been made and tested as referential specimens. Wall designation Table 1. Designation and reinforcement of tested walls Strength of masonry units Vertical steel Vertical reinforc. ratio Horizontal steel Horizontal reinforc. ratio H MPa H MPa 2 φ 14 2x0.056 % 2 φ6/20 cm % H MPa 2 φ 20 2x0.115% 2 φ6/20 cm % H2-0 9 MPa H MPa 2 φ 20 2x0.115% 2 φ6/20 cm % H MPa 2 φ 28 2x0.225% 2 φ6/20 cm % The walls have been 96 cm long, 140 cm high and 29 cm thick. General purpose mortar of the strength class M2.5 has been used for the construction. The specimens have been built on r.c. foundation blocks, fixed to the strong floor during the testing. In the case of reinforced walls, deformed steel reinforcing bars, 14, 20, or 28 mm in diameter of the strength class M400, have been placed vertically at the ends of the walls. The reinforcing bars have been anchored into the foundation block at the bottom and into the bond-beam at the top of the wall. Stirrups, placed in each horizontal mortar bed joint, have been bent around the vertical reinforcement at the ends to improve the anchoring. Stirrups have been made of smooth reinforcing steel, 6 mm in diameter, of the strength class M250. Two specimens of each type have been tested. 4

5 The walls have been tested as vertical cantilevers, with foundation block fixed onto the strong floor by means of steel bolts (Figure 1). They have been subjected to constant vertical load and cyclic lateral displacements, repeated three times at each displacement amplitude. Constant compressive stress of 20 % of the compressive strength of masonry has been induced in the walls horizontal section during the lateral resistance tests. Figure 1. Typical testing arrangement for lateral resistance test of masonry walls Experimentally obtained lateral resistance - displacement hysteresis envelopes are presented in Figs. 2 and 3 for the walls type H1 and H2, respectively. In the case of reinforced specimens, no difference in the resistance can be observed as a result of different amount of vertical reinforcement. There is, however, an improvement in lateral resistance and ductility with regard to referential unreinforced walls of the same quality, but this is attributed to horizontal bed joint, and not vertical reinforcement placed in the holes at the vertical edges of the walls. This can be concluded on the basis of the measured strain in horizontal and vertical reinforcement. Figure 2. Lateral load - displacement envelopes for walls H1: 0 - referential unreinforced wall Figure 3. Lateral load - displacement envelopes for walls H2: 0 - referential unreinforced wall Although the specimens have been designed to fail in bending, brittle local failure of units, i.e. buckling and crushing of thin shells and webs, resulted into predominant ultimate shear behaviour and collapse of the walls. Consequently, only a small part of the available tension capacity of vertical reinforcing bars has been utilized. As can be seen in Table 2, where the experimental values of lateral resistance H exp are correlated with predicted values of flexural capacity H u,cal, the predicted values by far overestimated the actual resistance of the walls. 5

6 Typical damage patterns at ultimate state are shown in Figs. 4 and 5. As can be seen, local buckling and crushing of the shells of masonry units prevailed. On the basis of the analysis of test results the conclusion can be made that in the case where local brittle failure of hollow masonry units takes place, code specifications regarding the amount of steel and calculation of section's capacity do not provide reliable results. In the best case, the ductility and shear resistance capacity of such walls will be slightly improved due to horizontal steel placed in the bed joints. However, vertical steel placed in the holes at the edges of the walls will not improve the behaviour. Although the reinforced walls have been designed properly by following the requirements and recommendations of the code, the actual resistance will be only slightly greater than in the case where the walls were not reinforced at all. Figure 4. Brittle shear failure of wall due to local brittle failure of units Figure 5. Local buckling and crushing of shells and webs of masonry units Table 2. Comparison between experimentally obtained H exp and calculated values of flexural capacity H u,cal of tested walls Experimental Calculated Wall h H exp M exp H u,cal M u,cal exp/cal designation (m) (kn) (knm) (kn) (knm) H H H H The models used for the calculation of lateral in-plane resistance of reinforced masonry walls' sections are reliable under the condition that sufficient bond between mortar and steel as well as mortar and blocks is provided and that masonry units are able to carry additional compression and shear which is transferred from the steel into the units as a result of seismic loads. Yielding of steel at ultimate state is assumed in the calculations of sectional capacity. However, if local brittle failure of masonry units takes place and bond is degraded under cyclic loading before the tension capacity of reinforcing steel is attained, the design value of 6

7 resistance capacity of the reinforced masonry wall's section will overestimate the actual resistance capacity. Although designed for earthquake loads by taking into account code specifications, the actual degree of seismic safety of a masonry structure under consideration could be lower than required and/or verified by calculation Robustness of units and bonding patterns Traditionally, only the construction of masonry walls with fully grouted (filled) head joints has been allowed in seismic zones. By definition given in EC 6, perpend joints can be considered to be filled if mortar is provided to the full height of the joint over a minimum of 40 % of the width of the unit. In the new draft of EC 8, however, a note states that the National Annex will select which of the three classes of perpend joints (fully grouted, ungrouted, and ungrouted with mechanical interlocking between the units) will be allowed to be used in a country. Whereas only fully grouted perpend joints have been allowed in the previous version of EC 8, it seems that the changes in the new EC 8 draft are the result of the pressure of masonry industry to introduce new technologies of masonry construction in seismic zones, without any reliable experimental evidence to prove that these technologies are suitable and acceptable. Therefore, a research project has been proposed and accepted by masonry industry where the influence of different systems of filling the perpend joints on the seismic behaviour of the walls has been investigated. Four different types of filling the perpend joints have been studied (Figure 6): Fully filled perpend joint (walls series BN - referential) Dry, unfilled perpend joint (walls series BG) Partly filled perpend joints with mortar in the pockets (walls series BP) Dry, grove and tongue perpend joint (walls series BZ) Figure 6. Schematic presentation of the investigated types of perpend joints 7

8 Figure 7. Masonry units for the construction of walls of series BN and BG The walls, designed to fail in shear, have been 125 cm long, 150 cm high and 30 cm thick. Grade M10 clay hollow units with dimensions 245x300x240 mm (length x width x height), with 12 mm thick shells, 8 mm thick webs and 50 % of holes per volume of the unit (Group 2 units according to EC 6) have been used, same in all cases except for the head joint face (Fig. 7). Grade M5 general purpose mortar has been used for the construction of specimens, built on r.c. foundation blocks and tested in the same way as described in 2.1 and shown in Fig. 1. Three specimens of each type have been tested under constant vertical load, which induced vertical stresses in the horizontal section amounting to 1/3 of the compressive strength of the masonry (Bosiljkov et al., 2004). All walls failed in shear, as expected. However, brittle local failure of units, i.e. buckling and crushing of thin shells and webs was the predominant phenomenon which determined the failure mode. The walls failed in a non-ductile, brittle mode, soon after the maximum resistance has been attained. The test results are summarized in Tables 3 and 4. In Table 3, mechanical characteristics of the tested types of walls, such as compressive f and tensile strength ft, as well as modulus of elasticity E and shear modulus G, are evaluated. Table 3. Mechanical characteristics of the tested types of masonry walls Type f (MPa) f t (MPa) E (MPa) G (MPa) f t /f G/E BN BG BP BZ It can be noticed that the ratio between the tensile and compressive strength of masonry f t /f of all tested types is low and did not exceed 4 % of the value of compressive strength (usually 6-8 %). The ratio between the values of shear modulus G, evaluated on the basis of the effective stiffness of the walls measured during lateral resistance tests, and modulus of elasticity E, evaluated on the basis of compression tests, did not exceed The ratio decreased with the increased compressive strength of masonry. 8

9 In Table 4, the lateral resistance and displacement capacity indicators are given for the tested types of wall, defined as the ratios between the lateral resistance and displacement values at different limit states, such as crack (d cr, H cr ), maximum resistance (d Hmax, H max ), and ultimate limit (d u, H u ). Table 4. Lateral resistance and ductility capacity indicators of the tested types of walls Series H cr /H max H du /H max d cr /d Hmax d u /d Hmax d u /d cr BN BG BP BZ A conclusion can be made that no significant influence of different types of filling the perpend joints can be observed on the lateral resistance and displacement capacity of the tested walls. The ratio between the lateral load acting on the walls at the initiation of diagonal cracking and maximum resistance H cr /H max was close to This means that the occurrence of diagonal shear cracks in masonry characterizes the attainment of lateral resistance of the walls. In the particular case studied, once the diagonal cracks occurred, the resistance of the walls at increased imposed displacements amplitudes started to degrade. Relatively large resistance degradation has been observed in all cases, i.e. small values of H du /H max ratio have been obtained despite the relatively small ultimate displacements, hence indicating the brittle character of the behaviour of the tested walls at shear failure. The displacement capacity of the tested walls in terms of displacement capacity indicators is small, below the expected values for unreinforced masonry walls. The actual values of mechanical properties of masonry, obtained by testing, are compared with the values, calculated by considering EC 6 equations and procedures, in Table 5. Table 5. Correlation between experimental and EC 6 predicted values of shear strength and shear modulus of masonry Series f v exp (MPa) G exp (MPa) f vk EC6 (MPa) G EC6 (MPa) f v exp /f vk EC6 G exp /G EC6 BN BG BP BZ It can be seen that the code predicted values of mechanical properties which define the seismic resistance of walls at shear failure (shear strength f vk and shear modulus G) overestimate the actual, experimentally obtained values in almost all cases. Taking into consideration the observed resistance and displacement capacity as well as mechanical properties of the tested wall types, it can be concluded that in all cases the behaviour of the specimens subjected to cyclic lateral loading has been governed by the 9

10 premature local brittle failure of units. Consequently, the failure of the walls subjected to cyclic lateral loading occurred at a stage where the type of filling the vertical, head joints did not yet influence the behaviour of the walls. The failure mechanisms in all cases depended on the characteristics of masonry units and did not depend on the way of construction of the wall specimens, i.e. the type masonry bond or the way of how the head joints have been filled. Therefore, no firm general conclusion can be made as regards the influence of types of masonry bond on the seismic behaviour of masonry walls. Additional research is needed, and masonry units with sufficient robustness to avoid local brittle failure should be used for the construction of test specimens in order to ensure that the masonry bond and not the units influence the behaviour of the walls when subjected to cyclic lateral load. 3. STRUCTURAL BEHAVIOUR FACTOR q With exception of the so called simple buildings, i.e. buildings which fulfil severe limitations regarding the height, structural configuration and quality of materials, the stability of masonry structures for vertical and seismic loads should be verified by calculation. As is the case of other types of structures, the seismic resistant design of masonry structures of all systems by EC 8 is based on: no collapse requirement, and damage limitation requirement. Taking into account the regularity of masonry buildings whose response is not significantly affected by contribution from higher modes of vibration (e.g. Abrams 1988; Tomaževič and Weiss 1994), lateral force method of analysis will provide adequate results. Following this method, the seismic base shear force F b for each horizontal direction in which the building is analysed, is determined by EC 8 as follows: F b = S d (T 1 ) m λ, (1) where: S d (T 1 ) = the ordinate of the design spectrum at period T 1, T 1 = fundamental period of vibration of the building for lateral motion in the direction considered, m = total mass of the building above the foundation or above the top of a rigid basement λ = correction factor, accounting for the fact that in building with at least three storeys and translational degrees of freedom in each horizontal direction, the effective modal mass of the 1st mode is smaller than the total building mass. Masonry buildings are rigid structures with natural periods of vibration ranging between periods where the EC 8 response spectrum is flat. Therefore, the ordinate of the design spectrum for masonry buildings can be determined by: 2.5 Sd (T) = ag S η, (2) q 10

11 where: a g = design ground acceleration on type A ground (rock or rock-like formation), S = soil factor, η = damping correction factor (η = 1 for 5% viscous damping), q = structural behaviour factor. As specified in EC 8, the capacity of structural system to resist seismic actions in the nonlinear range generally permits the design for forces smaller than those corresponding to a linear elastic response. To avoid explicit inelastic structural analysis in design, the capacity of the structure to dissipate energy through mainly ductile behaviour of its elements and other mechanisms is taken into account by performing an elastic analysis based on a response spectrum reduced with respect to the elastic one by introducing the behaviour factor q. In a qualitative and simplified way, the definition of behaviour factor q is explained in Fig. 8, where the seismic response curve of an actual structure, idealized as a linear elastic - perfectly plastic envelope, is compared with the response of a perfectly elastic structure having the same initial elastic stiffness characteristics. Figure 8. Definition of structural behaviour factor q As a result of the energy dissipation capacity of the actual structure, expressed by the global ductility factor µ u = d u /d e, there is no need for the structure to be designed for strength, i.e. for the expected elastic load H e. The structure is designed for the ultimate design load H du and the ratio between the two is called the behaviour factor q: q = H e /H du. (3a) In other words, behaviour factor q is an approximation of the ratio of the seismic forces that the structure would experience if its response was completely elastic, to the minimum seismic forces that may be used in the design with a conventional elastic model, still ensuring a satisfactory response of the structure. Following the definition in Fig. 1, structural behaviour factor can be also expressed in terms of the global ductility factor µ u = d u /d e as follows: q = (2 µ u - 1) 1/2. (3b) 11

12 A range of values of q factor for different systems of masonry construction is proposed in the recent draft of EC 8: for unreinforced masonry: q = for confined masonry: q = for reinforced masonry: q = Since the proposed values have limited experimental background (Tomaževič and Weiss, 1994), a research project has been recently carried out to investigate the seismic behaviour and determine the range of values of structural behaviour factor for selected types of unreinforced masonry buildings Experimental program and description of tests Six models representing buildings with two different structural configurations and two different types of masonry materials have been tested on a simple uni-directional seismic simulator, a two-storey terraced house with main structural walls orthogonal to seismic motion (models M1 - Fig. 9) and a three-storey apartment house with uniformly distributed structural walls in both directions (models M2 - Fig. 10). Four models of the first and two models of the second type have been tested. In the case of the terraced house, two models have been built as either partly or completely confined masonry structures (Table 6). Table 6. Shaking-table tests - description of tested models Designation Type Material Remark M1-1 Terraced house Calcium silicate no confinement M1-2 Terraced house Hollow clay unit no confinement M1-1c Terraced house Calcium silicate confined staircase walls M1-1d Terraced house Calcium silicate fully confined walls M2-1 Apartment house Calcium silicate no confinement M2-2 Apartment house Hollow clay unit no confinement Because of the limited capacity of earthquake simulator installed at ZAG, models have been built at a 1:5 scale from special model materials designed to fulfil the requirements of complete model similitude. The correlation between the model and prototype characteristics of masonry has been verified by testing a series of model and prototype-size walls under compression and a combination of compression and cyclic lateral loading. Nevertheless, it should be borne in mind, that by testing small scale masonry models, only the global behaviour and mechanism of the buildings behaviour can be adequately simulated, and not the behaviour of structural details. The models have been built on foundation slabs which have been fixed before the tests to the steel platform of the shaking table by means of bolts. In order to fulfil the requirements for similitude of dynamic behaviour, additional masses (steel bricks) have been fixed to each floor slab to simulate the effect of the live load. The models have been instrumented with displacement meters (LVDTs) and accelerometers fixed on the model at both edges and centre of each floor slab. 12

13 The north-south component of the earthquake acceleration record obtained at Petrovac during the April 15, 1979, earthquake in Montenegro, with peak ground acceleration of 0.43 g has been used to drive the shaking-table. The intensity of shaking was controlled by adjusting the maximum amplitude of the shaking-table displacement, obtained by numerical integration of the accelerogram used as the input in each successive test run, scaled according to the laws of model similitude. The analysis of shaking-table motion, carried out during one of our previous studies (Tomaževič and Weiss, 1994) has shown that the absolute acceleration spectra of the shaking-table motion, normalized with regard to maximum acceleration, are in good agreement with one of the previous versions of the EC 8 response spectrum. All models have been tested by subjecting them to step-wise increasing intensity of the shaking-table motion in each subsequent test run until models final collapse. During the tests, displacement and acceleration responses of the models at each floor have been measured. The behaviour of he models during testing has bee video-taped for further analysis of damage propagation. After each test run, the models have been inspected for damage, and the cracks have been marked and photographed. Also, the changes in dynamic properties of the models have been determined by analysing the records of free vibrations obtained by hitting the top slab of the model with hammer Test results All models failed in shear, as expected. Regardless to the structural type and configuration, shear cracks developed in structural walls in the direction of seismic motion, subsequently leading to stiffness and strength degradation and final collapse of the models. The unreinforced terraced house model M1-1 and apartment house model M2-1, made of materials simulating calcium silicate masonry units collapsed immediately after the first damage occurred, whereas respective models M1-2 and M2-2, made of materials simulating hollow clay units, though damaged, withstood additional shaking before collapse. The behaviour of partly and fully confined terraced house models M1-1c and M1-1d was significantly improved. Typical damage to the models just before collapse is shown in Figs. 9 and 10 for a terraced and apartment house model, respectively. Figure 9. Confined terraced house model M1-1c just before collapse Figure 10. Apartment house model M2-2 just before collapse 13

14 As can be seen, the damage to structural walls was concentrated in the first storey, so that typical shear type mechanism of seismic behaviour prevailed. Very little damage to the walls in the upper storeys has been observed at the moment of collapse in all cases, including confined terraced house models M1-1c and M1-1d. As a result of this mechanism, relative displacements of the upper storeys in the non-linear range of vibration were very small compared to the first storey drift. As the deformations and the amount of damage were small, the amount of dissipated, hysteretic energy in the upper storeys did not exceed 5 % of the energy dissipated in the first storey. Taking this into consideration, the conclusion can be made that the resistance envelope of the first storey determines the seismic behaviour of the tested structures. On the basis of the recorded displacement and acceleration response time histories and taking into account the masses of the models, concentrated at each floor level, the maximum values of the base shear developed in the models during the individual phases of testing, have been calculated. The values have been expressed in a non-dimensional form in terms of the base shear coefficient (BSC), which is the ratio between the base shear resisted and the weight of the model. The values are plotted against the first storey rotation angle (the ratio between the relative storey displacement and storey height), hence obtaining the lateral resistance - displacement envelopes of a critical storey in a non-dimensional form (Fig. 11). BSC 2 1,6 1,2 0,8 Models M1 Model M1-2 Model M1-1c Model M1-1d Model M1-1 BSC 0,8 0,6 0,4 Models M2 0,4 0,2 Model M2-1 Model M ,01 0,02 0,03 0,04 0,05 Rotation angle 0,0 0 0,005 0,01 0,015 0,02 Rotation angle Figure 11. Experimentally obtained base shear coefficient - storey rotation angle relationships As can be seen, the values of storey rotation angle where the stiffness of the models has significantly changed as a result of the damage occurring to structural walls (damage limit), are very close in all cases. The values of 0.25 % have been measured in the case of the terraced house models M1 and the values of 0.3% in the case of the apartment house models M2. It can be also seen that the damage limit values of storey rotation angle coincide or are very close to the values where the maximum resistance has been attained. Regarding the influence of masonry materials on the seismic behaviour of the tested buildings, it can be seen that the models of both, terraced house and apartment house structural type, made of model materials simulating calcium silicate masonry units (models M1-1 and M2-1) exhibited substantially more brittle behaviour than the models of the same type, but made of model materials simulating hollow clay units. However, there has been not much difference observed as regards the resistance. The confining the structural walls with vertical r.c. confining elements in the case of the terraced house models M1-1c and M1-1d 14

15 proved to be a successful measure to improve the seismic behaviour of the terraced house type of structure as regards both lateral resistance and displacement capacity. 3.3 Structural behaviour factor q In order to evaluate the values of behaviour factor q, the basic definition given in EC 8 and the simple philosophy, explained in Chapter 1, has been followed. Typical examples are shown in Figs. 12 and 13 for the terraced and apartment house models, respectively. The values of behaviour factors q, resulting from experimental envelopes, are presented in Table 7. In the calculations of the elastic response of the models, effective stiffness of the model structure, i.e. the measured stiffness at the occurrence of the first significant damage to structural walls (damage limit), has been taken into account, and not the initial stiffness of the model, measured before the shaking-table tests. EAVEK, a commercial computer program for seismic analysis of multi-storey buildings, has been used. Following the definition of behaviour factor q, the response of the elastic structure subjected to shaking-table motion during the testing phase in which the maximum resistance of the model has been attained, has been calculated. 1,0 0,8 M ,5 M1-1d BSC 0,6 0,4 0,2 R025 R005 R010 R050 0,0 0 0,01 0,02 0,03 0,04 Rotation angle Experimental Elastic response Idealized BSC 2 1,5 R075 R100 1 R050 R025 R150A 0,5 0 R010 R ,01 0,02 0,03 0,04 Rotation angle Experimental Elastic response Idealized Figure 12. Base shear coefficient - storey rotation angle relationships obtained for plain and confined terraced house models BSC 1,2 1 0,8 0,6 0,4 0,2 R125 R100 R125 R075 R050 R025 R010 R005 Experimental Elastic response Idealized M ,01 0,02 0,03 0,04 Rotation angle BSC 1,2 1 0,8 0,6 0,4 0,2 R100 R075 R050 R025 R100 R100 Experimental Elastic response Idealized M2-2 R005 R ,01 0,02 0,03 0,04 Rotation angle Figure 13. Base shear coefficient - storey rotation angle relationships obtained for apartment house models 15

16 The elastic response in terms of maximum elastic base shear has been compared with the maximum experimental base shear value BSC max as well as with the value BSC u obtained by the idealization of experimental resistance envelope, following the definition given in Eq. 3a. As can be seen in the diagrams in Figs. 12 and 13, the experimentally obtained envelopes have been idealized as bilinear elastic-plastic relationships. In the cases where no sudden collapse has occurred, 20 % of strength degradation has been allowed as a measure to evaluate the idealized global ductility µ u of the structure without risking collapse. However, it has to be noted that substantial damage to structural walls of the models has occurred at that stage. Therefore and in order to fulfil also the damage limitation requirement, for the evaluation of behaviour factor q on the basis of global ductility of the structure (Eq. 3b), only part of the available displacement capacity has been taken into account, limited by the displacement value where severe damage to structural walls occurs. This value has been arbitrarily chosen to be 3-times the value of storey rotation at the damage limit Φ u = 3 Φ dam. Typical damage to structural walls at the maximum permissible storey rotation angle (ultimate damage limit) is shown in Fig. 14. Table 7. Values of structural behaviour factor q evaluated from experiments Model q = BSC e /BSC max q = BSC e /BSC u q = (2 µ u - 1) 1/2 M M M1-1c M1-1d M M Fig. 14. Typical damage to structural walls at ultimate damage limit (model M1-1c) It can be seen that, generally speaking, the evaluation of behaviour factor q on the basis of the observed ductility capacity of the models resulted into higher values than the simple 16

17 correlation of theoretical elastic and observed base shear responses. It can be also seen that despite the differences observed in the behaviour of unreinforced masonry models during shaking-table tests (brittle behaviour of models M1-1 and M2-1 made of calcium silicate units against ductile behaviour of models M1-2 and M2-2 made of hollow clay units - see Fig. 11), the values of behaviour factor q, evaluated on the basis of simple definition, are of the same order of magnitude for the cases of both, terraced and apartment house structural types. In the case of the terraced house models with confined structural walls (models M1-1c and M1-1d), however, the observed improved behaviour resulted also in the increased evaluated values of behaviour factor q. Although the resistance and displacement capacity can be used as a measure of energy dissipation capacity, taking advantage of the measured response data, energy dissipation capacity of the tested models has been calculated on the basis of the measured storey shear - storey drift (relative storey displacement) hysteresis loops. The results of calculations are given in Table 8, where for each of the tested models the input energy, induced to the system during shaking by hydraulic actuator (energy demand - Bertero and Uang, 1992) and dissipated hysteretic energy, are presented. Cumulative values of input and dissipated hysteretic energy from the beginning to the end of the shaking-table tests, are given in the table, as well as the ratio between both.. Table 8. Relationships between the cumulative input and dissipated hysteretic energy at the end of shaking-table tests Model Input energy Dissipated hysteretic energy E inp (Nm) E hys (Nm) E hys /E inp M M1-1c M1-1d M M In order to make conclusions on the basis of data presented in Table 8, one has to take into consideration that, although both parameters are a function of structural response, input energy (energy demand) is a function of masses of the structure, ground (shaking-table) acceleration time history and velocity response of the structure, whereas the amount of dissipated hysteretic energy is determined mainly by damage propagation mechanism. Therefore, only the comparison of data obtained for the models of the same structural configuration, is reasonable. In this regard, the differences in the observed behaviour of apartment house models M2-1 and M2-2, made of calcium silicate and hollow clay units, respectively, can be also explained by the differences in energy dissipation capacity. Model M2-2 needed 4-times more input energy E inp than model M2-1 to cause collapse. At the same time, energy dissipation capacity E hys of model M2-2 was almost 8-times greater than energy dissipation capacity of model M2-1. As a result, the difference in E hys /E inp ratio is also significant. It is assumed that similar observation could have been made for the case of unreinforced terraced house models M1-1 and M1-2, if the response records of model M1-1 were available in adequate form for the analysis. 17

18 As the data for terraced house model M1-1 are missing, it is not possible to evaluate the effect of partial and complete confinement of structural walls with regard to referential model in terms of energy dissipation capacity. However, the difference between the partly and fully confined models M1-1c and M1-1d can be clearly seen. 2.6-times more input energy has been needed in the case of fully confined model M1-1d to cause collapse than in the case of partly confined model M1-1c, and 2.8-times more energy has been dissipated. Obviously, this resulted in almost the same E hys /E inp ratios in both cases. The decision about which values of behaviour factor q to recommend for the design of the tested types of buildings is therefore not a simple one. Namely, following the simple definition of behaviour factor according to EC 8, the differences in the assessed values for unreinforced buildings of both structural configuration types are not significant (Table 7). However, significant differences have been observed in the behaviour of the models of both structural types, made of the calcium silicate units on the one hand and those made of hollow clay units on the other. As can be seen by comparing the experimentally obtained resistance envelopes, the behaviour of the models, made of hollow clay units was significantly more ductile than the behaviour of the models made of calcium silicate units. The differences can be also seen by comparing the calculated input and dissipated hysteretic energy balance (Table 8). Obviously these observations need to be taken into consideration when making the final proposal. However, it can be concluded that the obtained values are well within the range of EC 8 proposed values for unreinforced and confined masonry structural systems. 4. CONCLUSIONS AND PLANS FOR FUTURE RESEARCH As the recent experimental investigations indicated, the development and innovation in masonry brought to the front the robustness of masonry units as one of the basic parameters, which determine the seismic behaviour of masonry walls and structures. In the case of the local brittle failure of masonry units, the failure mechanisms of the walls subjected to seismic loads change, so that, because of the changed basic relationships, the validity of most practically used calculation procedures for seismic resistance verification of unreinforced and reinforced masonry structures becomes questionnable. This has been also recognized in the new draft of EC 8, where the requirement that masonry units, used for the construction of masonry structures in seismic zones, should have sufficient robustness to avoid local brittle failure, is given. However, no quantitative criteria to fulfill such requirement are given in EC 8. Therefore, additional experimental research is needed to provide the basis and criteria for the classification of hollow masonry units regarding the robustness. It is planned that, as a result of a recently initiated research project, such criteria will be determined and a testing method for simple evaluation of robustness will be developed. The influence of mortar strength and the way of laying the units will be also studied, along with the influence of different types (classes) of head joints (fully and partly grouted, ungrouted, mechanical interlocking) on the homogeneity of structural walls, built with masonry units which will comply with the criteria for sufficient robustness. As regards the energy disipation capacity of masonry structures and EC 8 suggested values of structural behaviour factors q, the experimental study indicated that the ranges of values of factor q proposed in EC 8 for different masonry systems, are adequate. However, the study also indicated that the values depend not only on the system of construction, but also on the properties of masonry materials and structural configuration of the building under 18

19 consideration. Therefore, experimental research is needed for the assessment of a particular value for a particular structural type within the recommended range of values. However, although such tests are helpful, the values of behaviour factor q cannot be assessed by means of only ductility tests of structural walls. It has to be also stressed that the values recommended on the basis of the recent experimental study (q = 1.5 for unreinforced houses of both tested types and q = 2.0 for confined masonry houses) are valid for regular masonry structures, with uniformly distributed walls in both orthogonal directions and along the height. In the particular case studied, the structural configuration of the apartment house buildings fulfilled the requirements for regularity, whereas the configuration of the terraced house building was on the edge of the acceptable limit. Requirements should be provided for minimum amount and position of structural walls for the case that the behaviour factor q = 1.5 is used in the design of terraced house type of buildings. 5. ACKNOWLEDGEMENTS This paper discusses the results of several research projects, financed by the Ministry of Education, Science and Sports of the Republic of Slovenia, Chamber of Commerce and Industry of the Republic Slovenia, and associations of brick masonry producers from Slovenia (Wienerbeger Opekarna Ormož and Goriške opekarne), Austria (Verband Österreichischer Ziegelwerke), Germany (Deutsche Gesellschaft für Mauerwerksbau), Italy (Associazione Nationale Degli Industriali dei Laterizi) and Switzerland (Verband Schweizerische Ziegelindustrie). Their financial and professional contribution to the outcome of the projects is gratefully acknowledged. REFERENCES 1. Abrams, D.P. Dynamic and static testing of reinforced concrete masonry structures. The Masonry Soc. J., 7 (1), 1988, pp. T18-T Bertero, V. and Uang, C.M. Issues and future directions in the use of an energy approach for seismic resistant design of structures. In Nonlinear seismic analysis and design of reinforced concrete buildings, P.Fajfar and H.Krawinkler, eds., Elsevier Applied Science, London, England, 1992, pp Bosiljkov, V., Tomaževič, M. and Lutman, M. Optimization of shape of masonry units and technology of construction for earthquake resistant masonry buildings. Research Report ZAG, Ljubljana, Eurocode 6: Design of masonry structures, Part 1-1: Common rules for reinforced and unreinforced masonry structures. pren , CEN, Brussels, Eurocode 8: Design of structures for earthquake resistance, Part 1: General rules, seismic actions and rules for buildings. pren , CEN Brussels, Tomaževič, M.and Lutman, M. Influence of reinforcement and block strength on seismic behavior of reinforced masonry walls. Proc., 11th IBMaC, Vol.1, Shanghai, 1997, pp Tomaževič, M. and Weiss, P. Seismic behavior of plain and reinforced-masonry buildings. J. of Str.Engrg., ASCE, 120 (2), 1994, pp