Engineering Structures 33 (2011) 287297

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Cyclic out-of-plane behaviour of tall reinforced masonry walls under P effects

Francesca da Porto , Flavio Mosele, Claudio Modena
Department of Structural and Transportation Engineering, University of Padova, Via Marzolo 9, 35131 Padova, Italy

article

info

abstract
On the basis of experimental and numerical results, this paper discusses the out-of-plane behaviour of tall load-bearing reinforced masonry walls in a large-displacement regime, under the influence of vertical loads (P effects). Reinforced masonry systems can be advantageous for erecting one-storey commercial and industrial buildings 68 m high. These structures are often provided with deformable roofs and, as a consequence, in case of seismic actions out-of-plane forces can be significant in the wall behaviour and can lead to the onset of second order effects. For this structural configuration, a special set-up for outof-plane cyclic tests was designed and built, to assess experimentally two reinforced masonry systems, gather information on their structural behaviour, and calibrate momentcurvature relationships to be implemented in numerical models. These models allowed the test results to be extended to different dead loads, degrees of wall slenderness, and quantities of reinforcement. The research validated the effectiveness of such systems and identified some limitations and procedures for modelling and design. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 11 September 2009 Received in revised form 20 September 2010 Accepted 4 October 2010 Available online 18 November 2010 Keywords: Reinforced masonry Cyclic tests Out-of-plane tests P effects Momentcurvature relationships Single-storey buildings Commercial buildings Industrial buildings

1. Introduction Developments in structural masonry have been driven towards systems which have faster and cheaper construction processes and are based on new types of unit, mortar and/or grout, and ancillary components or reinforcement [1]. Improvements for unreinforced masonry have focused on replacing traditional head and bed joints with new types of joints with faster assemblage and better thermal performance [25]. Concurrently, developments for reinforced and confined masonry have focused on solving the lack of tensile strength in masonry while significantly improving resistance, ductility and energy dissipation capacity [68]. The general aim is improving the in-plane behaviour of walls, as the basic principles of conceptual design of structures for earthquake resistance are based on box-type behaviour. With this assumption, horizontal seismic actions are transferred to walls parallel to load direction (Fig. 1(a); see also [9]). Reinforced and post-tensioned masonry solutions have recently been proposed also for one-storey buildings, such as those for commercial and industrial purposes, as they can fulfil several functions, including structure, fire protection, thermal and sound insulation, weather protection, and proper sub-division of space [10]. For these types of constructions, reinforced loadbearing masonry walls, compared with other solutions including

Corresponding author. Tel.: +39 049 8275631; fax: +39 049 8275631. E-mail address: daporto@dic.unipd.it (F. da Porto).

framed structures, ensure that controlled thermo-hygrometric conditions are respected, with no use of insulating or coating materials [1]. However, for such slender walls, the effects of transverse loads, such as wind loading, earth pressure, and inertial forces from seismic excitations, have been recognised as significant. In addition, in such types of buildings, roofs are very often built with deformable structures. Deformable diaphragms hardly transfer horizontal loads to in-plane walls, and out-ofplane forces affect walls perpendicular to the earthquake direction. Concurrently, orthogonal walls, which could constitute good support to out-of-plane deformations, provided their adequate connection, become ineffective, as they are mostly located at very large distances [11]. Hence, during seismic events, roofs may be considered as elastic restraints, with variable stiffness along the wall extension, and this can significantly increase the local out-of-plane displacements of walls (Fig. 1(b)). The out-of-plane response of walls becomes critical, as the large displacements which may occur at the top of these structures during earthquakes, introduce second-order effects (P effects) and problems due to instability. In case of cantilever walls, fixed at the bottom and free to rotate at the top, the out-of-plane behaviour is dominated by vertical flexure. Prior to cracking, reinforced masonry walls behave as unreinforced walls. After initial cracking, wall stiffness decreases, but at the same time, vertical reinforcement starts to work and the walls continue to carry loads up to and beyond steel yielding. These walls exhibit good deflection, ductile failure, and good values of earthquake damping [12]. The flexural capacity of the

0141-0296/$ see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2010.10.004

288

F. da Porto et al. / Engineering Structures 33 (2011) 287297

List of symbols d dcr dLmax dmax du Em Es fmt h L Lcr Ldmax Lmax Lu M M1 M2 t Displacement (mm). Deflection/displacement at cracking of section (mm). Deflection/displacement at maximum load (mm). Maximum deflection/displacement (mm). Ultimate deflection/displacement, at strength degradation of 20% (mm). Elastic modulus of masonry (N/mm2 ). Elastic modulus of steel bars (N/mm2 ). Tensile strength of bed joint-unit interface (N/mm2 ). Wall height (m). Load (kN). Load at cracking of section (kN). Load at maximum deflection/displacement (kN). Maximum load (kN). Ultimate load, corresponding to strength degradation of 20% (kN). Bending moment (kN m). First-order moment (kN m). Second-order moment (kN m). Wall thickness (m). Yielding strain of masonry ( ). Ultimate strain of masonry ( ). Yielding strain of steel ( ). Ultimate strain of steel ( ). Curvature (mm1 ).

my mu sy su

section is usually evaluated by means of the strength design approach, in which the compressive force in masonry is evaluated by means of an equivalent stress block, and an elastoplastic stressstrain material model can be adopted for steel [23]. The American standard [13] and Eurocode 6 [14] also adopt this design assumption, with small variations. Nevertheless, the eccentricity of vertical loads and/or lateral loading can significantly reduce wall capacity, in relation to member slenderness. Experimental tests and theoretical models to evaluate reduction factors for the compressive strength of unreinforced walls, in relation to wall slenderness and eccentricity of vertical loads, have been developed since the 1950s [15,16], although the influence of lateral loading on the stability of slender unreinforced walls has been studied numerically [17] and experimentally [18,19] only in recent times. The response of reinforced masonry to second-order effects and the additional bending moments caused by eccentricity is certainly more effective than that of unreinforced masonry [20]. However, systematic work taking into account the influence of P effects on slender reinforced masonry walls, cannot be found in the literature. Recently, slender post-tensioned masonry walls, were tested under static out-of-plane loading [10] but with pin supports at both ends, which is more suitable to simulate wall-to-floor connections in the hypothesis of stiff horizontal diaphragms. Experimental tests, taking into account the presence of deformable roofs, were carried out by Ewing and Kariotis [21] and Simsir et al. [22], and aimed at evaluating the out-of-plane response of unreinforced or strengthened masonry walls. The former applied dynamic displacement histories, simulating the effects of floor diaphragm flexibility, at the top and bottom of unreinforced and strengthened masonry walls having various degrees of slenderness and with constant axial load [21]. The latter carried out shaking-table tests to evaluate the dynamic out-of-plane response of slender unreinforced masonry walls, under constant and centred axial loads, with different degrees of stiffness of

flexible diaphragms [22]. They observed that, despite diaphragm flexibility significantly increasing the out-of-plane displacement response of slender walls, these performed very well due to the formation of a hinge at their base, and no mid-height collapses occurred. In addition, most building codes do not have a consistent approach which also takes into account P effects in masonry structures. In particular, Italian and European codes [23,14] do not consider them explicitly, but fix slenderness limits for pin support conditions at both ends and for reinforced masonry, without providing alternative rational design procedures. In this context, two reinforced masonry systems for application in singlestorey commercial and industrial buildings have recently been developed [24] and tested [25,26,20]. The main aim of our experimental research was to study, by means of quasi-static cyclic procedures, the out-of-plane behaviour under P effects of such systems used for erecting tall load-bearing walls 68 m high, experimentally reproduced with cantilever boundary conditions (Fig. 1(c)). For deeper insights and to determine possible design limitations, we subsequently developed numerical models taking into account both material and geometrical non-linearity involved in the problem. In particular, material non-linearity is related to the flexural response of the reinforced masonry walls, which can be idealised by trilinear loaddeflection relationships, obtained by monotonic and/or cyclic out-of-plane loading tests. These three limit states directly derive from the MomentCurvature (M ) relationship of the section [27]. We thus calibrated M relationships on the basis of simple tests, and implemented them in finite element models used to carry out parametric analyses. These took into account the effect of different dead loads, degrees of wall slenderness, and quantities of reinforcement on the outof-plane behaviour. This paper presents experimental results obtained during cyclic out-of-plane testing of real-scale structures, describes the modelling approach and discusses the results of numerical analyses. 2. Experimental programme 2.1. Construction systems Two similar reinforced masonry systems were tested, both based on vertically perforated clay units and concentrated vertical reinforcement. The first system (Fig. 2(a)) is made with H-shaped units, allowing limited vertical reinforcement, aligned on the midplane of the wall, to be inserted in central holes with dimensions 72 65 mm (rmH). The second system (Fig. 2(b)) is built with alternate C- and H-shaped units (rmC). The C units have large central cores (140 110 mm), which allow more vertical reinforcement to be used. In these larger holes, instead of having a single bar aligned on the mid-plane of the wall as in rmH, it is possible to have an even number of bars, not aligned on the wall mid-plane (compare Fig. 2(a) and (b)). Hence, these uncoupled vertical reinforcement bars have a lever arm which further increases the resisting moment of the section. Considering the geometry of the C-shaped units (Fig. 2(b)), these can be laid after the vertical reinforcement has been placed in position, providing the required overlap of head joints at the same time. Hence, the construction process is simplified, as H-shaped units need to be threaded onto reinforcement bars (rebars). In addition, overlap of vertical reinforcement can be avoided, as high interstorey rebars can be used. All units have overall dimensions for length, width and height, of 250 380 190 mm, and the percentage of voids is 45% (Fig. 2). Both units showed similar behaviour in compression, with mean compressive strength on the gross area and mean

F. da Porto et al. / Engineering Structures 33 (2011) 287297

289

Fig. 1. Seismic response of single-storey building with rigid (a) and deformable (b) roof; static scheme adopted for testing (c).

Fig. 2. Reinforced masonry systems: rmH (a) and rmC (b).

elastic modulus of 12 N/mm2 and 11.6 kN/mm2 , respectively. The vertical reinforcement of rmH was made of one 16-mm diameter rebar (see Fig. 2(a)), and that of rmC was composed of threedimensional trusses, made of four 12-mm diameter rebars and 6-mm diameter stirrups every 400 mm (see Fig. 2(b)). Both types of vertical reinforcement were spaced at regular intervals (780 mm), for a corresponding reinforcement percentage of 0.08% in rmH and 0.18% in rmC. The horizontal reinforcements were made of two 6-mm diameter rebars, distributed alternatively at intervals of 400 mm. Hence, the percentage of horizontal reinforcement on the wall cross-section was about 0.04%. All reinforcements were made of B450C ribbed rebars, in either cold-drawn or hot-rolled steel, having yielding stresses between 380 and 570 N/mm2 and elastic moduli between 180 and 220 kN/mm2 . A special mortar was developed to be used for both bed joint laying and for filling in the cavities for vertical reinforcement. Starting from a general-purpose M10 mortar, this mortar had hydraulic binder, aggregates with a maximum diameter of 4 mm, and additives. Mean compressive strength and the elastic modulus of the mortar were respectively 11.3 N/mm2 and 11.4 kN/mm2 . Further details on systems and materials are given in [25,20,28]. 2.2. Test programme The main objective of the experimental programme was to assess the behaviour of two reinforced masonry systems under out-of-plane cyclic actions. Basic mechanical characterisation of systems was carried out in the laboratory by uniaxial compression and flexural tests. Uniaxial compression tests were carried out monotonically on specimens with average dimensions of 1000 1000 380 mm (length, height, width). The load increment rate was about 0.5 kN/s [29]. Specimens were instrumented with ten displacement transducers to record vertical, horizontal and transverse displacements.

A four-point bending test set-up was adopted for flexural tests, as in [30]. Spacing between load lines (600 mm) was kept smaller than prescribed, to avoid shearing at supports. Tests were carried out under displacement control, at a rate of 0.04 mm/s. Four loadingunloading cycles were performed on each specimen. Specimens were instrumented with 17 displacement transducers to record deflections, displacements, and the potential slip of vertical rebars. Strain gauges were installed on reinforcements, in the zone of constant bending moment. The tests were repeated on two series of specimens, rmH and rmC, described in the previous section. Each of them was carried out on three specimens for each masonry type, for a total number of twelve tests. Cyclic out-of-plane tests on two real-scale structures, one for each masonry type, were then carried out. Table 1 lists the main tests carried out on the test walls. Mechanical characterisation of materials preceded tests on masonry specimens. Twenty-six compression tests were carried out on units, 80 compression and flexural tests on mortar, and 30 tensile tests on reinforcement. During mortar development, a large variety of physical and chemical tests were also carried out, to check workability, bleeding properties, bulk density, air content, and workable life. The complete description of the test programme and the results can be found in [20,28]. 2.3. Main results of basic mechanical characterisation Briefly, the compressive behaviour of the two systems was very similar, with mean compressive strength on the gross cross sectional area and mean elastic modulus of 4.6 N/mm2 and 3.3 kN/mm2 , respectively. The differences between the rmH and rmC test series were always lower than 10%. During flexural tests, the masonry walls attained three main limit states: cracking of section, for corresponding values of load and deflection Lcr , dcr , maximum resistance (Lmax , dLmax ) and

290 Table 1 Test matrix. Test Uniaxial compression Flexural test Cyclic out-of-plane test Name rmH-C rmC-C rmH-F rmC-F rmH-R rmC-R

F. da Porto et al. / Engineering Structures 33 (2011) 287297

No of tests 3 3 3 3 1 1

Dimensions (mm) 1030 380 990 1165 380 990 1030 380 2790 1165 380 2790 2070 380 6010 1945 380 6010

Vertical reinforcement 216 (116@780 mm) 812 (412@780 mm) 216 (116@780 mm) 812 (412@780 mm) 316 (116@780 mm) 1212 (412@780 mm)

Reinforcement percentage (%) 0.10 0.20 0.10 0.20 0.10 0.18

Table 2 Flexural test results. Specimen Av. rmH-F Av. rmC-F Lcr (kN) 17.5 16.9 dcr (mm) 1.9 1.3 Lmax (kN) 69.2 132.2 dLmax (mm) 17.4 23.4 Lu (kN) 55.3 105.5 du (mm) 74.7 56.9 Lcr/max () 0.25 0.13 dcr/Lmax () 0.11 0.06 du/Lmax () 4.27 2.35

ultimate limit state (Lu , du ), determined by strength degradation of 20%. Table 2 lists these values for the two reinforced masonry types, and shows significant load and displacement ratios. Behaviour in the uncracked phase was similar in the two systems; the first flexural cracks opened on the three central bed joints, for similar values of displacements and loads. Conversely, the following cracked phase was strongly influenced by the higher reinforcement percentage of rmC, which yielded different failure mechanisms. RmC specimens were characterised by simultaneous yielding of rebars and crushing of units, with mid-span deflection of about 23 mm and a resisting moment of about 70 kN m. RmH specimens presented yielding of rebars only, with mid-span deflection of about 17 mm and resisting moment of about 39 kN m. Crushing of units occurred only later, in the post-peak phase, due to higher imposed deflections. The higher strength of rmC, attained with a different failure mode, influenced the deformation capacity. At the ultimate limit state, rmC specimens reached a mid-span deflection of 57 mm, versus the 75 mm of rmH. 3. Cyclic out-of-plane tests 3.1. Test set-up Each specimen for real-scale cyclic out-of-plane tests was made of two reinforced masonry walls, 6 m high and 2 m long, connected at the top by a heavy slab. Fig. 3 shows the test set-up. The heavy slab was hinged at the top of the two walls and was used to transfer horizontal displacements, imposing congruence at the wall tops and simulating the effect of roof dead load. Bottom reinforced concrete bond beams were fixed to the reaction slab. Hence, walls were tested in cantilever boundary conditions, following the scheme of Fig. 1(c). In the case of uniform load distribution, for the building types taken into account, roof dead load on walls changed, according to span and roof type, from 10 to 30 kN/m. The heavy slab had a mass of 10 ton, corresponding to about 25 kN/m. This load, close to the maximum in the roof dead-load range, was chosen to emphasise instability effects due to out-ofplane deformations of walls (P effects). Lateral out-of-plane cyclic displacements with increasing amplitude were applied at 7 m from ground level, at a constant frequency of 0.004 Hz and until a top displacement of 250 mm. The final part of the test, in the case of rmC, was carried out under monotonic loading, again under displacement control. Besides load cell and displacement transducers for measurement and control of the hydraulic jack, each specimen was instrumented with 64 sensors. Sixteen inductive wire sensors, placed at four elevations along the walls, including the wall top, measured wall deflections; 16 displacement transducers, at the base of the walls and at the bottom bond beam, checked possible displacements or rotations.

The location of these sensors on the walls corresponds to the white dots in Fig. 3(b) and (c). Lastly, 32 strain gauges were mounted on the vertical rebars, at seven levels along the walls. Test specimens, procedure, test set-up and instrumentation, and results, are extensively described in [20,28]. 3.2. Test results Fig. 4 shows cyclic loaddeflection diagrams and their envelopes for the two specimens, rmH (a) and rmC (b). Each diagram of Fig. 4 gives the total load measured on a specimen, and the top displacements of the two walls constituting that specimen. Fig. 3(b) and (c) shows rmC and rmH deflections at maximum displacement. These diagrams show that, during out-of-plane cyclic tests, the specimens again attained three main limit states. Table 3 lists load and displacement values at these limit states, and shows significant load and displacement ratios. Until reaching the first limit state, corresponding to cracking of section, for values of load and displacement Lcr , dcr , flexural behaviour was similar in both reinforced masonry systems. When the first horizontal cracks opened at the interface between units and bed joints, load Lcr was 28% higher in rmC than in rmH, due to uncoupling of rebar and the higher tensile strength of joints. Further performance was deeply influenced by reinforcement percentage, rmC cracked stiffness being definitely higher than that of rmH. In both cases, maximum resistance (Lmax , dLmax ) was attained when the increase in the second-order moment was higher than that of the resisting moment. This occurred when the resisting sections lost their stiffness, due to rebar yielding and/or crushing of the compression zone. RmH specimen attained Lmax when rebar yielding started, at a top displacement of about 165 mm (2.8% of wall height). The test was terminated at the maximum displacement dmax of 198 mm (3.3% of wall height), corresponding to a strength degradation of 11%, as the degradation velocity was becoming high enough to endanger the residual stability of the entire structure. Hence, the displacement capacity in the post-peak phase observed for rmH during flexural tests became disadvantageous when second-order effects occurred, as the latter were predominant as soon as rebar yielding took place. The maximum capacity Lmax of rmC was attained after yielding of tension rebars and crushing of masonry units respectively, at a top displacement of about 310 mm (5.2% of wall height). The test was stopped at the maximum displacement dmax of 388 mm (6.6% of wall height), corresponding to a strength degradation of 15%. This limit state (Ldmax , dmax ) was characterised by increasing width of flexural cracks in bed joints and crushing of units, extending until the 3rd unit course from the bottom. The higher stiffness of the rmC system, which reduced the influence of second-order

F. da Porto et al. / Engineering Structures 33 (2011) 287297

291

Fig. 3. Scheme of the cyclic out-of-plane test set-up (a) and view of the two specimens at their final deflection: rmC (b) and rmH (c). Table 3 Cyclic out-of-plane test results. Specimen rmH rmC Lcr (kN) 8.0 10.3 dcr (mm) 12.2 12.2 Lmax (kN) 12.4 34.4 dLmax (mm) 164.8 310.1 Ldmax (kN) 11.1 29.3 dmax (mm) 197.4 387.6 Lcr/max () 0.64 0.30 Ldmax/max () 0.89 0.85 dcr/Lmax () 0.08 0.04 dmax/Lmax () 1.20 1.25

Fig. 4. Cyclic loaddeflection diagrams and their envelopes: rmH (a) and rmC (b).

effects, yielded values of maximum horizontal load capacity almost three times higher than those of rmH and displacement values almost twice those of rmH. The smaller influence of second-order effects is also shown by the ratios between load at cracking and maximum resistance limit states. These were 64% for rmH and 30% for rmC, indicating that the cracking phase is short and unstable in rmH, but longer and more stable in rmC. 3.3. Bending moment analysis The influence of second-order effects is usually evaluated by the stability ratio [31], which, according to Eurocodes 2 and 8, is the ratio between second- and first-order moment (M2 /M1 ). Fig. 5 shows total (M), first-order (M1 ) and second-order moments (M2 ) at the wall base of rmH and rmC specimens, plotted versus measured top displacements. M2 also takes into account the wall dead load, as it is comparable to the roof dead load, and practically increases with linear trend. Table 4 lists stability ratios and bending moments at four stages: three are the limit states described in the previous section and one is when the stability ratio reaches 10%. According to Eurocode 2, for reinforced concrete structures, and Eurocode 8, for aseismic structural design, second-order effects should be taken into account when the stability ratio is higher than 10% [32,33]. At the beginning of the tests, vertical loads provided negligible contributions to the bending moment in both reinforced masonry specimens (Fig. 5) and, at crack limit state, the stability ratio was still very low (2.5%3.4%; Table 4). When displacements increased,

wall stiffness decreased and second-order effects started to influence the flexural behaviour of specimens substantially. The rmH specimen reached the limit value of the stability ratio for top displacements of about 39 mm (0.6% of wall height), with corresponding lateral loads which were 73% of maximum load capacity. RmC reached the limit value for top displacements of about 100 mm (1.7% of wall height), with corresponding lateral loads which were 59% of maximum load capacity. The influence of second-order effects, in terms of stability ratios, was always smaller in rmC, despite the higher displacements reached. In general, at the same limit state, the stability ratio of rmC was about 30% smaller than in rmH. The total moment capacity attained in the tests (maximum M) was compared to the theoretical results computed with code formulas. These were based on the hypothesis that plane sections remain plane, assuming a stress-block constitutive law for masonry with depth of 0.8 times the neutral axis position, maximum stress of 0.85 times the masonry maximum strength, maximum strain of 3.5 for masonry and maximum admissible strain of 10 for steel. These hypotheses are common to many code procedures, such as [23,14]. The computed values, assuming unitary material partial safety factors, were 56 kN m for rmH and 106 kN m for rmC, i.e. 7% higher than the experimental value (52.3 kN m) in the case of rmH and 19% lower than the experimental value (132 kN m) in the case of rmC. At these values of moment capacity, the analyses gave values of strain in steel of 9.7 in rmH, and 3.5 in rmC. However, whereas the second value, such as the assumption of masonry crushing, are completely consistent with experimental evidence in rmC (see also [28,20]),

292 Table 4 Stability ratios at significant limit states. Limit state L/2 (kN) d (mm) M1 (kN m) M2 (kN m) M (kN m) M2 /M1 (%) Crack rmH 4.0 12.2 25.8 0.9 26.7 3.4 rmC 5.1 12.2 33.1 0.8 33.9 2.5

F. da Porto et al. / Engineering Structures 33 (2011) 287297

10% rmH 4.5 38.6 28.8 2.8 31.6 9.8 rmC 10.1 99.6 65.2 6.7 71.9 10.3

Lmax rmH 6.2 164.8 40.1 12.0 52.1 29.8 rmC 17.2 310.1 111.0 21.1 132.0 19.0

dmax rmH 5.5 197.4 35.8 14.2 50.0 39.6 rmC 14.6 387.6 94.4 26.7 121.1 28.3

Fig. 5. Moment versus displacement curves: rmH (a) and rmC (b).

the first value obtained for rmH strongly overestimated (around 400%) the experimental strain, which was around yielding values at Lmax (see also Section 3.2), and masonry crushing did not occur before Lmax , too. This is due to the fact that the maximum lateral load Lmax is reached when the increase of section capacity (M ) is equal to the increase of second order moment (M2 ), regardless of whether the maximum resistant moment (maximum M) has been reached or not. This can be seen by the different tangent points between M2 and M in the diagrams of Fig. 5(a) and (b). Hence, second order effects did not allow exploiting the full section resistance and ductility in rmH, whereas they did in rmC. Bending moment analysis thus confirmed experimental evidence, i.e., the rmC system can withstand second-order effects at a higher safety level and at higher displacements than rmH which, in similar conditions, is at the limit of its resistant capacity. 4. Numerical modelling 4.1. Modelling approach Slender reinforced masonry walls under out-of-plane actions behave similarly to beams in flexure. Therefore, they can be modelled adopting a macro-modelling approach, where masonry is taken as a continuum characterised by material non-linearity. This non-linearity was implemented by means of analytical momentcurvature (M ) relationships, starting from material properties, into finite element models which also took into account geometrical non-linearity (P effects). Under the assumption that plane sections remain plane and that the stressstrain curves of masonry and steel are known, the curvatures associated with a range of bending moments and axial loads may be determined by imposing strain compatibility and force equilibrium. Elasticperfectly plastic constitutive laws were assumed for both masonry and vertical reinforcement, on the basis of the test results described in [20]. Table 5 (set 1) lists the basic material properties used: tensile strength of interface between bed joint and unit (fmt ), elastic modulus, yielding strain and ultimate strain of masonry (Em , my , mu ) and of steel bars (Es , sy , su ). Only fmt was not directly evaluated by testing. It was thus backcalculated by imposing equilibrium at the cracking limit state

Fig. 6. M diagram for rmH under N = 0.

during both flexural and out-of-plane tests. This parameter varies between 0.33 and 0.46 N/mm2 , according to type of test, curing of specimens and type of reinforced masonry system. Fig. 6 shows the M diagram for the rmH system under null axial load, revealing the influence of tensile strength, as the dashed line is drawn for fmt = 0, whereas the black line takes into account masonry tensile strength. Tension stiffening during cracking is taken into account by assuming a linear connection between cracking and yielding limits (grey line). Each reinforced masonry wall making up the rmH and rmC specimens for cyclic out-of-plane tests was modelled with 30 beam elements (one element for each unit course), fixed at the base to reproduce cantilever boundary conditions. In the finite element model, the M relationship, analytically calculated on the basis of experimental stressstrain material laws, was assigned to all elements, so a distributed ductility model was built. Hence, in the numerical analyses, each element could reach its plastic capacity according to the actual demand; i.e., there was no pre-defined plastic hinge length as in lumped ductility models. Two elastic beams were used to model the reinforced concrete bond beams on the top of each wall. The heavy slab was modelled with an elastic beam of thickness adequate to reproduce actual slab dead load. The beam centre of mass was fixed at a height corresponding to the actual heavy slab level. The connection between walls and slab

F. da Porto et al. / Engineering Structures 33 (2011) 287297 Table 5 Parameters of constitutive laws. Test series Set 1 Set 2 fmt (N/mm2 ) 0.33 0.33 Em (N/mm2 ) 3305 3966

293

my ( )
1.4 1.4

mu ( )
2.4 3.5

Es (N/mm2 ) 200 000 200 000

sy ( )
2.55 2.55

su ( )
10 10

Table 6 Parameters for out-of-plane models. Test series rmH rmC fmt (N/mm2 ) 0.37 0.45 Em (N/mm2 ) 3966 3966

my ( )
1.4 1.4

mu ( )
3.5 3.5

Es (N/mm2 ) 180 000 220 000

sy ( )
2.11 2.58

su ( )
10 10

Fig. 7. Numerical and experimental loaddeflection diagrams (flexural tests).

Fig. 8. Numerical and experimental loadtop deflection envelopes (out-of-plane tests).

was modelled by a rigid link, free to rotate around the axis of the hinge. Hence, the model was constituted by 60 beam elements for the walls, 2 beam elements for the bond beams, and 3 elements for the slab and wall-to-slab connection. 4.2. Validation of models Before modelling the tested masonry specimens, the analytical M relationship derived from the parameters of set 1 in Table 5 was checked and calibrated against the results of simple flexural tests. This relationship was implemented into the finite element models of the rmH and rmC specimens tested under four-point bending. Fig. 7 compares the experimental and numerical loaddisplacement diagrams of flexural tests. Implementing experimental data, models can already reproduce overall behaviour fairly well, and only slightly underestimate maximum load values. Nevertheless, the results were improved by model calibration. In particular, the elastic modulus of the masonry was increased by 20% and masonry ultimate strain was increased to 3.5 (see Table 5, set 2). This calibration procedure was deemed consistent with experimental observation. Indeed, the experimental ultimate deformation of 2.4 was very probably underestimated, since spalling of units affected instrument readings in the final phase of compression tests. Furthermore, unit crushing and, more in general, the behaviour of the compression zone of masonry during flexural tests differed from the behaviour of masonry under uniaxial compression. The implemented models did not simulate strength degradation related to crushing of units, as the adopted constitutive law did not have a post-peak softening phase. This explains the increase in ultimate strain, the value of which matches proposed code values [23,14]. Actually, in the four-point bending test models, crushing of units was attained at mid-span deflections of about 25 mm, which corresponds with experimental evidence [20], when ultimate masonry strain is fixed at mu = 3.5 . Increased elastic modulus also provided better estimation of maximum strength (Fig. 7).

The analytical M relationships to be implemented in finite element models of out-of-plane tests were thus based on the mechanical parameters listed in Table 6, in which the masonry values are those calibrated according to Table 5. fmt values were directly extracted from these tests, and the mechanical parameters of reinforcement were also taken from tests carried out on rebars used for tall wall construction differing from those adopted in specimens for flexural tests due to their different construction time. Fig. 8 compares the loadtop deflection envelopes of the experimental cycles with the numerical curves, and Fig. 9 compares the experimental and numerical momenttop displacement curves. The latter show some significant limit states (yielding of steel sy , crushing of masonry mu , attainment of maximum moment M1max ) identified by the models. Maximum lateral load (i.e. M1max ), is reached with yielding of steel in rmH, and with yielding of steel first and crushing of masonry later in rmC, coherently with the experimental results. The overall behaviour was adequately reproduced, initial elastic stiffness and cracking limit state were well approximated, as was stiffness in the cracked phase. The maximum load was slightly overestimated in the case of rmH, but underestimated in the case of rmC. Displacement at maximum load was slightly underestimated in both cases (Table 7). The post-peak phase of models showed a strength degradation that was smaller than in reality. This is very probably due to the fact that the rmC model does not simulate crushing of units, and the rmH model cannot reproduce the slipping of reinforcement bars on the overlap area which was observed during the test. It is worth noting that the models achieved masonry ultimate strain (3.5 ) at 192 mm top displacement for rmH and 301 mm displacement for rmC, near the point of maximum resistance. This is again in very good agreement with experimental observations. Fig. 9 shows that the models could reproduce very well not only the overall loaddisplacement curves, but also the detailed trends of global (M), first order (M1 ) and second order (M2 ) moment, which are essential to understand the results of parametric analyses and identify the significant limit states.

294

F. da Porto et al. / Engineering Structures 33 (2011) 287297

Fig. 9. Numerical and experimental moment versus displacement curves: rmH (a) and rmC (b). Table 7 Experimental and numerical results at maximum resistance. rmH exp Lmax (kN) dLmax (mm) 12.4 164.8 mod 13.2 160.1 Error (%) 1.8 2.5 rmC exp 34.4 310.1 mod 32.9 300.8 Error (%)

4.4 3.0

5. Parametric analyses 5.1. Premise The calibrated models, reproducing the experimental behaviour of the studied reinforced masonry systems, were applied to carry out parametric analyses. The main aim of these analyses was to verify the influence of various parameters such as axial load level, wall slenderness, and percentage of vertical reinforcement on outof-plane behaviour, and providing some design limits. Analyses were performed on both reinforced masonry systems, rmH and rmC. 5.2. Variations in axial load level The axial load level was varied in the above-mentioned roof dead load range, between 5 and 30 kN/m, in 5 kN/m steps. Experimental tests were carried out under a uniform load distribution of 25 kN/m. Figs. 10 and 11 show the numerical results obtained for the rmH and rmC systems. Reduction of axial loads anticipates the attainment of the cracking limit state, as it reduces the decompression moment and slightly increases lateral load capacity and displacement, as it also reduces the second-order moment. These variations are negligible for the rmH system, and their effects are also negligible in the case of cracking limit for rmC. The different axial load levels affect the maximum lateral load of rmC more significantly, although variations remain lower than 10%. In general, axial load variation influences the resistant moment (M) of rmH more than that of rmC. Axial loads also generate second-order effects (M2 ), so that first-order moments (M1 = M M2 ) undergo slight changes in rmH, as the two effects cancel each other out, whereas the first-order moments of rmC have higher variations. This is the cause of the different behaviour in the two systems. 5.3. Variations in slenderness In this context, slenderness is evaluated as the geometrical ratio between wall height and thickness (h/t) and was 15.8 for tested specimens (h = 6 m, t = 0.38 m). This parameter was changed, keeping thickness constant and varying the wall height from 5 m until instability, in 1-m steps. Figs. 12 and 13 show the numerical results obtained for the rmH and rmC systems.
Fig. 12. Influence of slenderness, rmH.

Fig. 10. Influence of axial load level, rmH.

Fig. 11. Influence of axial load level, rmC.

Increase in slenderness strongly reduces load capacity, due to increasing P effects, with a concurring increase in corresponding displacements. Increase in slenderness also produces a decrease in stiffness, with subsequent attainment of cracking limit

F. da Porto et al. / Engineering Structures 33 (2011) 287297

295

Fig. 13. Influence of slenderness, rmC. Fig. 15. Moment versus displacement curves, limit states for rmC.

Fig. 16. Influence of slenderness on maximum lateral load. Fig. 14. Moment versus displacement curves, limit states for rmH.

states at lower loads and higher displacements. The reduction in stiffness also influences the cracked phase, which can provide a lower load capacity increase from cracking to maximum load limit. Two main limits were found by means of these analyses. The first limit was at a height of 8 m (h/t = 21.1) in both reinforced masonry systems, and marked the change from a condition where lateral load governed the behaviour, to a new condition where second-order effects dominated. In the first case (rmC), the maximum lateral load (corresponding to maximum first order moment, M1 ) was reached once the section reached its maximum moment capacity (total M = M1 + M2 ). In the second case (rmH), being the influence of second-order effects dominating, the maximum lateral load (i.e., maximum M1 ), was reached slightly earlier than the maximum section capacity (M). This means that, due to P effects, the lateral load capacity started decreasing, despite the walls not having yet reached their maximum section capacity. This can be seen comparing the analyses carried out for walls 9 m high (Figs. 14 and 15), where yielding of steel for rmH (sy ) and crushing of masonry for rmC (mu ) occurs after the attainment of maximum load (M1max ), on the contrary of what happens for shorter walls. The second limit corresponds to the sudden attainment of maximum lateral load, immediately after reaching cracking load. In rmC, this limit occurs when steel attains yielding (sy ) simultaneously or immediately after the attainment of maximum load (M1max ; Fig. 15). Therefore, the first limit entails a gradual reduction in maximum load, whereas the second limit brings instability. These two limits coincide for the rmH system, whereas in rmC the second limit is at a height of more than 12 m (h/t = 31.6), as shown in Fig. 16. The second limit is also evidenced by the sudden decrease in displacements at maximum load, due to instability (Fig. 17).

Fig. 17. Influence of slenderness on displacement at maximum load.

5.4. Variations in vertical reinforcement The percentage of vertical reinforcement was varied from 0.04% to 0.18%, in two intermediate steps (0.08% and 0.13%). The lowest value corresponded to a theoretical rmH wall with three 12-mm diameter reinforcement bars. This value is also close to the minimum reinforcement percentage, 0.05%, fixed by the Italian code [23]. The 0.08% intermediate analysis corresponded to the actual reinforcement of the tested rmH specimen (three 16-mm diameter bars over a section of 2070 380 mm). This reinforcement percentage also corresponds to the minimum value fixed by the European code [14]. The 0.13% analysis corresponded to a theoretical rmC wall with three trusses made of four 10-mm diameter rebars, which was not experimentally tested in reality. The highest reinforcement percentage again corresponded to the actual value of the tested rmC specimen (three trusses made of four 12-mm diameter bars over a section of 1945 380 mm). This value was considered as an upper limit, since the rmC system already presented balanced failure. Fig. 18 shows the numerical results obtained from the four models.

296

F. da Porto et al. / Engineering Structures 33 (2011) 287297

Fig. 18. Influence of vertical reinforcement percentage.

Fig. 19. Influence of reinforcement percentage on maximum lateral load.

Fig. 20. Influence of reinforcement percentage on displacement at maximum load.

The lowest reinforcement ratio represents a limit case, since the attainment of maximum load capacity is almost concurrent with flexural cracking. This behaviour is dominated by instability due to P effects, as after cracking of a section, the second order moment (M2 ) solely brings the masonry wall to its sectional capacity (M), without any increase of lateral load, i.e. first order moment (M1 ). In general, maximum load capacity linearly increases with the reinforcement percentage (Fig. 19). Conversely, displacement at maximum load stabilizes at around 300 mm (about 5% of the wall height) starting from a reinforcement percentage of about 0.13% (Fig. 20). This very probably occurs because displacement at maximum load is related to crushing of units, which rules the balanced failure mode of walls with higher reinforcement percentage. 6. Discussion and conclusions Two reinforced masonry systems for tall, single-storey buildings were developed from existing ones, adapting unit geometry, dimensions, shape and quantity of reinforcements to the specific structural configuration required by the building type. The use of

perforated clay units, 380 mm thick, ensured that environmental requirements for the indoor spaces of this building type were satisfied. Using H-shaped and, to even greater extent, C-shaped units, resulted in simplification and optimisation of the building process. These rmH and rmC reinforced masonry systems were experimentally tested to characterise their mechanical behaviour and gather information on material and geometry non-linearity, by cyclic out-of-plane tests with P effects. These tests confirmed the higher stiffness of the rmC system, and showed its positive influence in the particular structural configuration under study. The maximum horizontal load capacity for rmC was almost three times higher than that for rmH and, consequently, rmC displacements at maximum load and maximum displacement were almost twice those of rmH. The ductility of the rmH section under flexural loads was not exploited in out-of-plane behaviour, as instability effects dominated the latter as soon as the reinforcements started yielding. The higher percentage and uncoupling of vertical reinforcements in rmC allowed full masonry section strength to be exploited, with balanced failure during both flexural and outof-plane tests, but they also reduced the impact of second-order effects. It is also worth mentioning that rebar uncoupling in rmC, besides guaranteeing better overall behaviour, allows hardening of tension bars, which is certainly useful. Hence, the rmC system is able to resist second-order effects at a higher safety level, despite the higher displacements. This was also evidenced by the high increase in lateral load capacity from cracking to maximum load (ratios of loads at cracking and maximum resistance of 64% for rmH and 30% for rmC) and by the lower value of the stability ratio of rmC (about 30% smaller than that of rmH) at the same limit states. Top displacements that activate the influence of P effects, in terms of achievement of the 10% stability ratio, were about 39 mm (0.6% of wall height) for rmH and 100 mm (1.7% of wall height) for rmC. Notwithstanding, both systems showed acceptable performance, with out-of-plane deflection capacity, evaluated as the ratio between measured topdeflection and wall height, of 5.2% and 6.6% in rmC at maximum resistance and maximum displacement, and 2.8% and 3.3% in rmH at the same limit states. It should also be noted that the slenderness limit, fixed by seismic codes [23,33], is 15, assuming simple support boundary conditions at both ends. Conversely, the tested walls had a slenderness of 15.8 in terms of the height-to-thickness ratio, which doubles (31.6) when slenderness is evaluated on the effective height, i.e., taking into account actual cantilever boundary conditions. Experimental out-of-plane behaviour was simulated with numerical models which implemented material non-linearity, by means of M relationships calibrated on experimental data, and geometrical non-linearity. Although the simplified models did not take into account softening of masonry, they were adequate to describe the influence of second-order effects and reproduced the overall out-of-plane behaviour of tall walls very well. It is worth noting that these models were based on a few and simple experimental data, which can be extracted from a small number of tests on materials (tensile strength of bed joint-unit interface and steel reinforcement properties) and from a few tests on simple masonry assemblages (uniaxial compression and fourpoint bending tests). The models were subsequently used to carry out parametric analyses and study the influence of axial load level, wall slenderness and percentage of vertical reinforcement on the outof-plane response of the studied systems. It was found that variations in roof dead load into a realistic but also quite wide range, which takes into account various roof types and building spans, do not significantly influence the out of-plane behaviour of tall walls (effects are lower than 10%). Conversely, an increase in slenderness, produced by increasing wall height, reduces maximum loads and evidently increases the

F. da Porto et al. / Engineering Structures 33 (2011) 287297

297

influence of second-order effects. Two fundamental slenderness limits were found: one marking the change from attainment of maximum capacity exploiting full sectional resistance to failure due to instability, and the second characterised by attainment of maximum capacity near the cracking limit. The first limit occurs at a wall height of 8 m (h/t = 21.1) in both systems. The second limit practically coincides with the first in the case of rmH, but occurs at a wall height of 12 m (h/t = 31.6) in rmC. On the one hand, the two limits again reveal the smaller influence of second-order effects on rmC. On the other, the slenderness limit of 15 imposed by seismic codes [23,33], assuming simple support boundary conditions at both ends, is definitely too conservative when compared with both experimental and numerical results obtained. The minimum percentage of vertical reinforcement recommended to avoid failure dominated by second-order effects is 0.08% (rmH system). This limit matches that given by Eurocode 8 [33], although, this limit was very probably not fixed to take into account P effects. It is significant, in this regard, that the minimum percentage established by the Italian code, which is 0.05% [23] has been proven to be adequate for in-plane walls [34,20] and should therefore be adopted for those applications, but caution must be used in other applications, such as those presented in this paper. From a reinforcement percentage of 0.13%0.18%, maximum load is attained at a top displacement of about 300 mm, and ultimate masonry strain is concurrently achieved in a balanced failure mode. Hence, higher percentages of reinforcement are deemed useless or even dangerous, as they would bring the masonry section towards more brittle failure modes. In this context, Eurocodes 6 and 8 [14,33] do not provide any indications regarding the maximum reinforcement percentage, whereas the Italian code establishes a maximum limit of 1.0% [23], partly influenced by the higher percentages allowed for reinforced concrete elements and partly by the higher values that can be adopted in the case of shear wall applications. However, they are inadequate for prevalently out-of-plane loads in tall walls.

Acknowledgements This work was carried out under EU Contract COOP-CT-2005018120: Developing Innovative Systems for Reinforced Masonry Walls - DISWall and the University of Padova grant CPDA085943. The unit and RM system producer, Cis Edil s.r.l., and the mortar producer, Tassullo S.p.A., were partners in the EU project. The authors are grateful to E. Barbiero and A. Biliato for their contribution to the experimental investigation during their M.Sc. theses, and to M. Dalla Benetta for supervising experimental tests. Tests were partly carried out at the Laboratory of Structural Materials Testing of the University of Padova and partly at Cis Edil s.r.l., Luzzara, Italy. References
[1] Hendry AW. Masonry walls: materials and construction. Constr Build Mater 2001;15(8):32330. doi:10.1016/S0950-0618(01)00019-8. [2] Tomaevi M, Lutman M, Bosiljkov V. Robustness of masonry units and seismic behaviour of masonry walls. Constr Build Mater 2006;20:102839. doi:10.1016/j.conbuildmat.2005.05.001. [3] Magenes G, Morandi P, Penna A. Experimental in-plane cyclic response of masonry walls with clay units. In: Proc. 14th world conference on earthquake engineering. Paper ID: 12-03-0095. 2008. [4] da Porto F, Grendene M, Modena C. Estimation of load reduction factors for clay masonry walls. Earthq Eng Struct Dyn 2009;38(10):115574. doi:10.1002/eqe.887.

[5] da Porto F, Guidi G, Garbin E, Modena C. In-plane behavior of clay masonry walls: experimental testing and finite element modelling. J Struct Eng, ASCE 2010;136(11):137992. doi:10.1061/(ASCE)ST.1943-541X.0000236. [6] Tomaevi M. Earthquake-resistant design of masonry buildings. Series in innovation in structures and construction, vol. 1. London: Imperial College Press; 1999. [7] da Porto F, Mosele F, Modena C. Reinforced clay masonry walls under shearcompression loads: experimental behaviour. In: Proc. 11th Canadian masonry symposium. 2009 [CD ROM]. [8] da Porto F, Mosele F, Modena C. Compressive behaviour of a new reinforced masonry system. In: Materials and structures. Springer; 2010. doi:10.1617/s11527-010-9649-x. [9] Shing PB, Schuller M, Hoskered VS. In-plane resistance of reinforced masonry shear walls. J Struct Eng, ASCE 1990;116(3):61940. [10] Bean Popehn JR, Schulz AE, Drake CR. Behaviour of slender, posttensioned masonry walls under transverse loading. J Struct Eng, ASCE 2007;133(11): 154155. doi:10.1061/(ASCE)0733-9445(2007)133:11(1541). [11] Macchi G, Magenes G. Ingegneria delle strutture. Vol. III. Cap. XIII. Giangreco E. (a cura di). UTET. 2002 [in Italian]. [12] Drysdale RG, Hamid AA. Masonry structures behaviour and design. 3rd ed. Boulder (Colorado): The Masonry Society; 2008. [13] The Masonry Standards Joint Committee. Building code requirement for masonry structures. ACI530/ASCE5/TMS402. Detroit (New York, Boulder): American Society of Civil Engineers, The Masonry Society; 2005. [14] European Committee for Standardization. Eurocode 6design of masonry structurespart 11: general rules for reinforced and unreinforced masonry structures. Brussels (Belgium): EN 1996-1-1; 2005. [15] Angervo K. ber der knickung und tragfeines gedrckten pfeilers ohne zugfestigkeit. Statische technische forschungsanstalt. Publication 26, Helsinki. 1954 [in German]. [16] Chapman JC, Slatford . The elastic buckling of brittle columns. Proc Inst Civ Eng 1957;6(1):10725. doi:10.1680/iicep.1957.12395. [17] La Mendola L, Papia M, Zingone G. Stability of masonry walls subjected to seismic transverse forces. J Struct Eng, ASCE 1995;121(11):15817. [18] Griffith MG, Lam NTK, Wilson JL, Doherty K. Experimental investigation of unreinforced brick masonry walls in flexure. J Struct Eng, ASCE 2004;130(3): 42332. doi:10.1061/(ASCE)0733-9445(2004)130:3(423). [19] Bean Popehn JR, Schulz AE, Lu M, Stolarski HK, Ojard NJ. Influence of transverse loading on the stability of slender unreinforced masonry walls. Eng Struct 2008;30(11):28309. doi:10.1016/j.engstruct.2008.02.016. [20] Mosele F. In-plane and out-of-plane cyclic behaviour of reinforced masonry walls. Ph.D. thesis. Trento (Italy): University of Trento; 2009. [21] Ewing RD, Kariotis JC. Methodology for the mitigation of seismic hazards in existing unreinforced masonry buildings: wall testing, out of plane. ABK topical rep. no. 4. El Segundo (CA); 1981. [22] Simsir CC, Aschheim MA, Abrams DP. Influence of diaphragm flexibility on the out-of-plane response of unreinforced masonry bearing walls. In: Proc. 9th North American masonry conference. 2003. pp. 77990. [23] Ministry of Infrastructures. Technical standards for constructions. Rome (Italy): DM 14/01/08, 2008 [in Italian]. [24] DISWall. 20052007. COOP-CT-200518120. Developing innovative systems for reinforced masonry walls. Scientific coordinator: C. Modena. University of Padova. 2008. http://diswall.dic.unipd.it/. [25] Mosele F, da Porto F, Dalla Benetta M, Modena C. Out-of-plane behaviour of tall reinforced masonry walls. In: Proc. 2nd Canadian conference on effective design of structures. McMaster University, Hamilton. 2008. pp. 289300. [26] Mosele F, da Porto F, Modena C. Out-of-plane cyclic testing of tall reinforced masonry walls. In: Proc. 14th world conference on earthquake engineering. Paper ID: 05-04-0101. 2008. [27] Abboud BE, Hamid AA, Harris HG. Flexural behaviour of reinforced concrete masonry walls under out-of-plane monotonic loads. ACI Struct J 1996;93(3): 32735. [28] da Porto F, Mosele F, Modena C. Experimental testing of tall reinforced masonry walls under out-of-plane actions. Constr Build Mater 2010; doi:10.1016/j.conbuildmat.2010.05.020. [29] European Committee for Standardization. Methods of tests for masonry determination of compressive strength. Brussels (Belgium): EN 1052-1; 1998. [30] European Committee for Standardization. Methods of tests for masonry determination of flexural strength. Brussels (Belgium): EN 1052-2; 1999. [31] Pettinga JD, Priestley MJN. Accounting for P-delta effects in structures when using direct displacement-based design. Pavia: IUSS Press; 2007. [32] European Committee for Standardization. Eurocode 2design of concrete structurespart 11: general rules and rules for buildings. Brussels (Belgium): EN 1992-1-1; 2004. [33] European Committee for Standardization. Eurocode 8design of structures for earthquake resistance. Part 1: general rules, seismic actions and rules for buildings. Brussels (Belgium): EN 1998-1; 2004. [34] Magenes G. In-plane cyclic testing of reinforced masonry shear walls. In: Proc. of the 11th European conference on earthquake engineering. 1998.