Influence of Clay Block Masonry Properties on the Out-of-Plane Behaviour of Infilled RC Frames

In order to determine the characteristics that govern the out-of-plane behaviour of masonry infills, two groups of wall specimens were built and tested in the 
 laboratory. Specimens were assembled and tested as described in EN 1052-2 provisions and constitute of flexural strength for a plane of failure parallel and perpendicular 
 to the bedjoints specimens. By obtaining data from experiments, numerical micromodels were developed to predict their mechanical behaviour. A calibration procedure 
 undertaken and results obtained from the experimental campaign were found to be in agreement with those obtained from the numerical models. Additionally, former inplane infilled frame numerical models were tested with acquired out-of-plane calibrated material model. No significant difference was found.

Reinforced concrete (RC) frames infilled with unreinforced masonry units (URM) is a common structural practice in seismically active South Europe (Booth & Key, 2006). European earthquake design provisions Eurocode 8 (CEN, 2004b) regard wall infill/panels as an secondary elements, i.e. they do not contribute to overall seismic behaviour. However, it was known that infills contribute in seismic behaviour of RC frames even in the late 1960's. From there, interest in seismic behaviour of infilled frames has grown (Asteris et al., 2017;Trapani et al., 2015;Asteris et al. 2016) on separate fields of in-plane (IP) loading, out-of-plane (OOP) loading and their combination (IP + OOP). A large amount of experimental and analytical studies have been done in the field of IP, the same cannot be stated for the OOP and especially for IP + OOP interaction (Asteris et al., 2017;Wang et al. 2016). Moreover, the OOP field is based on analytical research of arching action, and numerical, i.e. computational research is scares and is based on membrane and strut with centred mass models.
Consequently, this paper is a part of OOP research with the intention to account properties that determine behaviour of infills subjected to OOP loading. Accordingly, 20 masonry wall specimens were tested and numerical micro models calibrated to account the experiment.

METHODS, MATERIALS AND RESULTS OF TESTED
The experiment preparation and testing was done in accordance to EN 1052-2 (British Standards Institution, 2016. Two testing groups were made: Group I: flexural strength for a plane of failure parallel to the bedjoints, and Group II: flexural strength for a plane of failure perpendicular to the bedjoints (parallel to headjoints). The recommendation of 10 wall test specimens for each Group was adopted in favour of statistical significance (Sorić, 2016). Wall specimens are made from whole and halflength blocks ( fig.2).
Firstly, hollow clay masonry units ( fig.1a) were cut in half of their height ( fig.1b) to emulate the units used as an infill in RC frames testing from (Penava, 2012) and units that will be used in further experiments.
Pretested properties of clay blocks, mortar and wall specimens are presented on the table 1.
Test setup of masonry wall specimens can be seen on a figure 2, 4c & 4f. The average dimensions of 10 specimens in each group as well as test setup dimensions are shown on figure 2. Testing was conducted with an increasing monotonic load on a 4-point (2 line reactions + 2 line loads) load setup on Controls 50-C1201/BFR by 50-C1200/8 apparatus.  It was expected that Group I will fail by separating two rows of blocks on bedjoint at the mid-height of the specimen. Hence, reaching tensile strength of the mortar. On the other hand, two possible failures were expected for the Group II. Those include: a) separation of blocks by mortar (blocks are undamaged) or b) failure trough the specimens (blocks failed). The b) failure is more likely to happen as fmt > fbh.
a) Group I b) Group II Figure 2 Test setup mesurement

Experimental results
Averaged results of the conducted test can be seen on table 2 & its distribution on figure 3. Figure 3 shows the minimum (MIN), maximum (MAX) and mean strength (AVG) with its variation within standard deviation (straight lines), i.e. fx ± s. Flexural strength was calculated by equation 1 from (British Standards Institution, 2016). Group I failed by separation of block rows by the bedjoint at the specimens mid-height (fig4d&e). Group II failed by failing clay blocks (fig4g&h), hence, through the whole wall specimen. (1)

Numerical model
Numerical models were assembled and tested using Atena 3D Eng (Cervenka Consulting, 2015). A threedimensional micromodeling approach was used, constructed from three-dimensional solid and twodimensional contactinterface (zero thickness) elements Line of loading Line of support multiplied by their length of their span, the force corresponds to 0.5 kN/step force each. Furthermore, solid elements beyond the supports in the numerical model ( fig.6) were discarded in order to gain faster calculation time. It is to be noted that the calculation with solids continuing beyond the supports was carried out, and no significant differences was observed from those without solids beyond supports.

Numerical material models and calibration
Numerical material models (tab.3&4) were adopted form (Penava, Sigmund, & Kožar, 2016) and modified during the calibration. The CC nonlinear cementitious 2 material model from table 3 was used for modelling clay masonry units, hence, solid elements. Likewise, CC interface material model from table 4 was used to model the mortar joints, hence, 2D interface gap elements. The interlocking effect of mortar filling the voids of opposite blocks and thereby locking them is modelled by interlocking function (fig.7).
The mentioned models from (Penava et al., 2016) acquire properties of clay blocks in direction of voids, however, during the analysis of the results from conducted numerical tests they were inadequate for modelling of Group II, i.e. the response was higher than measured by experiments. To that end, changes to tensile strength and tension softening function was introduced.
Tensile strength was changed from that in the direction of voids ft = 1.80 MPa to that of perpendicular to the voids ft = 0.38 MPa as the OOP loading caused failure of the clay blocks in direction perpendicular to voids. The displacement tension softening function through trial and error was adjusted from d = 0.010 mm to d = 0.001 mm. Fracture energy calculation depends upon tensile strength (eq.1) (Vos, 1983), however it was left unchanged, i.e. as if tensile strength in eq.1 was is in the direction of voids. If tensile strength in eq.1 was changed to be perpendicular to the voids, a predeveloped failure occurs in both Groups.

Numerical test results
With changes to the material models, the results from numerical tests are shown on figure 8 and table 5. Table 5 shows the force at failure and maximal principal stress obtained from figure 8.    Figure 8a it can be observed that numerical model of Group I had failure by discontenting bedjoints, i.e. mortar tensile failure. Figure 8c shows failures and cracking of the clay blocks.

General information
Having material model properties changed, previous work with unreinforced masonry infilled (URM) RC frames (Anić, Penava, Legatiuk, & Sarhosis, 2017; was questioned. Hence, the modifications to the infill units were implemented into the infilled frame model in order to measure the possible alterations. In short, the reinforced concrete (RC) frame has a designated medium ductility class (DCM) by Eurocode 2 provisions (CEN, 2004a), boundary conditions with numerical test setups are presented on figure 9. The model was subjected as in previous works (in-plane pushover method). For more details on the infilled frame please refer to the  article.

CONCLUSIONS AND DISCCUSION OF THE RESULTS
By comparing numerical and experimental results of Group I & II, differences force-wise were calculated as 9.55% for Group I and 7.32% for Group II. Group II has stress-wise difference of 2.63%.
Based on flexural testing of masonry wall specimens a numerical model was compiled and calibrated.
Calibration included modifying tension strength and displacement in tension softening function. Tension strength was changed from the value in direction of voids to the value perpendicular to voids. The calibration has proven adequate enough to have high correlation with the experiments. It is to be noted that the calibration was carried out in favour of Group II as Group I due the specific failure mode (reaching tensile strength of mortar) had agreement with the experiments from beginning.
Additionally, an infilled framed was tested in order to observe the validity due to changes in material model of clay blocks. It was shown that the changes did not drastically affect the outcome force ( fig.10), crack and stress wise ( fig.11).
In summation, the following conclusions can be drawn: a) Wall specimens had failure modes as predicted, Group I had failure along bedjoints due to reaching mortar tensile strength. Group II failed along the blocks, reaching tensile strength of the blocks in direction perpendicular to the voids. b) In order to simulate OOP bending, a mix of mechanical properties had to be implemented into the material models.