Professional Documents
Culture Documents
h i g h l i g h t s
" Experimental study of the second order bending failure mode of full-scale unreinforced brickwork walls.
" Experimental analysis of the effect of different slenderness and eccentricities.
" Test setup for testing hinged full-scale masonry walls in laboratory.
" Numerical simulation of walls under eccentric compressive loads taking into account geometric and material non linearity.
" Comparison from different experimental programs, analytic standards solutions and the proposed numerical simulation.
a r t i c l e i n f o a b s t r a c t
Article history: The eccentric in-plane loading of masonry walls involves complex bending performance that includes
Received 19 September 2012 second-order effects. In this work, a bidimensional (2D) simplified micro-model for the analysis of this
Received in revised form 21 January 2013 type of failure is developed. An experimental investigation based on 20 tests of full-scale unreinforced
Accepted 25 January 2013
masonry walls is performed. The tests are characterised by slenderness and load eccentricity. The analyt-
Available online 28 February 2013
ical methods of Eurocode-6 and ACI-530 are compared with experimental data from the present investi-
gation, other experimental results available in the literature and simulation results from the numerical
Keywords:
model.
Buckling failure
Unreinforced brickwork wall
Ó 2013 Elsevier Ltd. All rights reserved.
Experimental tests
Simplified micro-modelling
1. Introduction masonry and its fragile response to tension. Although the tensile
strength of masonry walls is neglected in most codes (see [2,3]),
Although the use of structural masonry has been dramatically some authors (see [4–6]) have reported that this property has a
reduced since the widespread introduction of concrete and steel significant influence, even at small values, on the load-bearing
structures, there is still a large number of existing buildings in capacity of slender or eccentrically loaded walls.
use that are structurally composed of load-bearing brick masonry To understand the structural behaviour of compressed brick-
walls. Assessing the strength of these buildings is a part of the work walls, several experimental research studies have been pub-
maintenance work required to enhance their useful life. Among lished reporting empirical analytical calculation methods for the
the possible failure modes, buckling failure must be addressed to assessment of the wall load-bearing capacity. The application
verify the safety of the walls from sudden collapse. range of such methods is limited to specific support and load con-
The in-plane load-bearing capacity of unreinforced brickwork ditions [4,7–13]. Recently, the development of modern computa-
walls depends on both the material properties and the geometry tional methods (e.g., FEA) has fostered the proposal of a large
of the structure. The typical effects of eccentricity and slenderness number of numerical models to simulate the behaviour of
have been widely analysed and are reported in many works [1]. masonry. Most of the proposed computational methods for the
The behaviour of eccentrically loaded unreinforced masonry walls analysis of masonry structures have been formulated using
is complex because of the non-linear compressive behaviour of the micro-modelling [14–16] or homogenised macro-modelling
approaches [17–20].
⇑ Corresponding author. Tel.: +34 937398728; fax: +34 937398994. In the experimental field, several studies have reported on the
E-mail address: ernest.bernat@upc.edu (E. Bernat). load-bearing capacity of masonry walls. The works of da Porto
0950-0618/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.conbuildmat.2013.01.025
2 E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13
et al. [21] and Cavaleri et al. [22] are especially significant, with re- Some authors ([6,46,47]) have demonstrated that the most widely utilised
structural masonry codes are too conservative when applied to the analysis of very
spect to the study of the out-of-plane behaviour of load-bearing
slender or very eccentrically compressed unreinforced masonry walls. Comparing
walls. The similarity between full-scale and small-scale tests has the experimental values from the tests with the result of the application of widely
been widely studied ([23]), and full research programmes have used standards is another specific objective of the present work.
been completed considering only small-scale walls ([6,23–25]). Numerical models have been proposed to overcome the limitations of the stan-
Compressive full-scale tests on masonry walls or columns are not dards in calculating the structural response of masonry walls. However, most of the
proposed numerical methods have been implemented in-house, using proprietary
very common. Among the studies on full-scale walls, those of Wat-
software that is not freely available to outside practitioners. Conversely, the present
stein and Allen [26] and Kirtschig and Anstötz [27] are remarkable research is based on a finite element approach that is available in a commercial
because of the use of specific systems to model neatly hinged sup- software environment. Validating and proving the accuracy of this available numer-
ports (i.e., rotationally free) at the wall ends. In these cases, the ical method is another objective of this research. In addition to the experimental re-
sults obtained in the present study, the model is also validated using experimental
clarity and simplicity of the support condition allow the study of
data available in the literature.
the influence of the load eccentricity and the wall slenderness with The available data are extended by using the numerical tool to perform a com-
no uncertainty in the actual boundary conditions. prehensive numerical simulation. The experimental and numerical results are com-
One of the parameters that affect the buckling failure and the pared to two standards (EC-6 [2] and ACI 530 [3]) to assess the accuracy of the
load-bearing capacity of masonry walls is the masonry modulus standards.
determined from the flexural values according to the method pro- Series Height, H (mm) Width, b (mm) Thickness, t (mm) Walls #
posed in [24]. H 2700 900 132 1–3, 5–9
The masonry was tested to obtain the compressive strength, fc, M 1650 900 132 10–16
the modulus of linear deformation in the direction perpendicular S 1000 900 132 18–20
to the bed joints, E, and the bonding strength between bricks and F 2700 900 132 4
T 1800 900 270 17
mortar joints, fxt. Specimens consisting of five stacked bricks were
used to obtain values for fc and E, and samples of two stacked
bricks were used for the other tests. The values obtained for each
wall series are summarised in Table 2. The results show that the the vertical direction to allow for free descending displacement on
value of the modulus of linear deformation is low compared with the top of the wall during the test.
the correlations between E and fc provided by the codes (E/fc = A steel beam was installed between the jack and the upper
1000 in [2] and E/fc = 700 in [3]). The measured values are closer hinge to distribute the load along the wall width. The connection
to those found by other researchers (see [31,51–53] and Table 3), between the wall and the test machine was made using two steel
although the measured values are still significantly lower. Using plates at each end. These plates held the wall in contact with the
low-strength mortars and actual brick-laying conditions (less con- hinges and allowed different load eccentricities to be set. Any
trolled than in typical laboratory conditions) may explain the eccentricity between 0 mm and 80 mm, at 5 mm intervals, was
unexpectedly low modulus of deformation. possible. The contact between the wall and the test system con-
Five different geometries of full-scale unreinforced masonry sisted of a 20-mm M2 mortar layer to distribute the load and avoid
walls (H, F, M, T, S) were tested as summarised in Table 4. Twenty points of high stress. The details of the connection system are
walls were tested. All walls were built by professional bricklayers shown in Fig. 2. Eight potentiometers were installed to control
in the vertical position to emulate the actual fabrication process. the rotation of both hinges (two sensors for each hinge) and the
The bricks were wetted before being placed in position to assure descending movement of the upper hinge (four sensors on the steel
good adherence. The alignment of each row was checked during distribution beam). Depending on the test, two or four laser dis-
the construction. The walls were fabricated on an auxiliary steel placement sensors were used to measure the out-of-plane defor-
tool utilised for the transport, elevation and positioning into the mation. The mid-height deformation was always measured with
test setup. Walls W#1–W#9 were fabricated and air-cured under the deformation at 1=4 and 3=4 height in the case of walls W#10–
indoor atmospheric conditions, whereas the rest of the walls were W#20.
fabricated and air-cured under outdoor atmospheric conditions. The tests were force-controlled. All data were recorded at 50 Hz.
An ordinary video camera was used to record all tests, and a high-
velocity camera was used to record the W#10–W#20 tests to cap-
3.2. Wall test setup and procedure ture the details of the failure mode.
The testing was executed as follows: (a) The wall was placed in
The test setup is described in Fig. 1. To produce accurate and the test position using an overhead travelling crane and a pallet
clear boundary conditions, the walls were prepared with a pair of truck, affixing the wall to the loading system in the theoretically
steel hinges as part of the test setup. The lower hinge was laid desired position; (b) The actual geometry was measured with a
on the floor (H and F series) or placed over a structural steel beam laser sensor, including local and global geometric deviations. Two
(M, S and T series) to increase the height of the wall’s base, whereas out-of-plane measurements, one at each border of the wall, were
the upper hinge was connected to a hydraulic jack with a compres- obtained every three masonry rows; (c) The sensors were put in
sion load capacity of 1000 kN. The upper hinge could move only in place; (d) The surrounding area was protected from the projection
of rubble; and (e) The test proceeded until the wall collapsed. The
Table 2
load increased continuously at a uniform rate until the sudden col-
Characterisation test results of masonry. lapse point.
Wall # Mortar Modulus of linear Compressive Bonding
deformation, E strength fc strength fxt 3.3. Experimental results
(MPa) (MPa) (MPa)
1 M1 780 18.2 0.23 Table 5 summarises the test conditions and results for the wall
2–5 M2 12.9 0.36 series. The geometry type (as outlined in Table 4) and slenderness
6–9 13.7
(Eq. (1)) are indicated for each wall. The mid-height eccentricity ra-
10–20 10.8
tio (em/t), calculated as the ratio of the initial mid-height eccentric-
Table 3
E/fk ratio from literature and experimental tests.
Source of information Modulus of linear deformation, E (MPa) Compressive strength, fc (MPa) E/fc
[31] Clay brickwork series 2 21,700.00 26.04 833.33
[31] Clay brickwork series 1 8675.00 13.82 627.50
[51] Concentric loading 2172.50 12.92 168.19
[52] Concentric brickwork type 1 1867.00 13.50 138.30
[52] Concentric brickwork type 2 1700.00 13.25 128.30
[52] Concentric brickwork type 3 2313.00 7.53 307.04
[51] Concentric load 1440.00 11.70 123.08
Mortar M1. Wall 1 780.00 18.20 42.86
Mortar M2. Walls 2–5 780.00 12.90 60.47
Mortar M2. Walls 6–9 780.00 13.70 56.93
Mortar M2. Walls 10–20 780.00 10.80 72.22
4 E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13
Fig. 1. Test setup (a) for walls from series S, (b) for walls from series M and (c) for walls from series H.
ity (em) to the thickness (t) of the wall, is also listed. A dimension- displacement of the upper hinge at the maximum load, vmax, both
less measure of the hinge vertical alignment (d/t), describing the measured from the effective height (Hef), are shown in the last col-
overall initial rotation of the wall, is also calculated and shown in umns of Table 5. Hef is the real distance between the axes of the
Table 5. Fig. 3a shows the geometric configuration of a typical wall hinges.
of the current campaign. With respect to the experimental results,
k ¼ Hef =t ð1Þ
the load-bearing capacity of the wall, Pmax, and the ratio (Eq. (2)) of
the maximum load to the maximum theoretical uniformly distrib- / ¼ Pmax =Pu ð2Þ
uted load (Eq. (3)), are included. The mid-height out-of-plane
deformation at the maximum load, hmax, and the descending Pu ¼ b t fc ð3Þ
Fig. 2. Detail of the connection between the wall and the test system at the lower end (left) and the upper end (right).
E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13 5
Table 5
The results of the tests on unreinforced masonry walls under eccentric compressive loads.
Wall Geometry k e/t (%) d/t (%) Pmax (kN) U (%) hmax/Hef 103 vmax/Hef 103
W#1 H 22.3 4.2 0.0 169.3 7.8 4.5 2.5
W#2 H 22.2 14.9 0.0 65.7 4.3 5.9 1.4
W#3 H 22.1 8.5 7.6 133.8 8.7 9.2 2.4
W#4 F 15.0 3.9 4.9 578.6 37.8 12.7 10.8
W#5 H 22.1 10.6 5.3 239.8 15.6 5.1 2.2
W#6 H 21.7 16.1 9.5 30.0 1.8 4.9 0.5
W#7 H 21.8 5.4 4.5 134.7 8.3 6.7 1.9
W#8 H 22.3 0.3 6.1 129.4 8.0 5.6 1.5
W#9 H 21.9 9.8 6.4 109.8 6.7 4.5 1.3
W#10 M 14.1 8.1 8.0 423.9 33.0 8.2 6.0
W#11 M 14.3 18.1 7.6 371.2 28.9 9.2 5.5
W#12 M 13.9 5.0 3.8 471.1 36.7 8.9 6.0
W#13 M 14.1 25.6 1.5 83.8 6.5 9.4 2.2
W#14 M 14.1 16.0 0.0 518.5 40.4 9.0 5.4
W#15 M 14.0 24.3 2.3 236.7 18.4 10.1 3.2
W#16 M 14.1 12.5 2.3 408.2 31.8 8.5 4.5
W#17 T 6.8 66.9 9.1 491.1 38.3 8.8 4.8
W#18 S 9.1 15.8 2.3 803.4 62.6 2.9 5.4
W#19 S 9.2 22.9 0.0 686.1 53.5 4.5 8.7
W#20 S 9.3 25.0 2.3 152.6 11.9 7.6 4.3
The walls were tested so that the eccentricities at the lower and ing capacity. As the results in Fig. 8 show, the descending displace-
upper ends were equal. Although a constant eccentricity along the ment of the wall top section at the instant of failure depends on the
wall height was expected under these conditions, additional acci- initial eccentricity at mid height (which, if the alignment of the
dental eccentricities appeared because of construction imperfec- hinge is not taken into account, is equal to e/t d/2t). Practically,
tions. The actual wall shape was measured after the placement of the experiments show that, for a fixed slenderness, the vertical dis-
each wall in the test position. The actual mid-height eccentricity, placement required to cause the collapse of the wall decreases
em, was calculated based on the measurements. Construction with an increase in eccentricity. Slenderness also affects this rela-
imperfections also caused the hinges to be imperfectly aligned, tionship. The results differ between walls from series H, M and S, as
requiring the measurement of the initial horizontal distance shown in Fig. 8. For example, an increase of the normalised initial
between the upper and bottom axes (d). In Tables 5 and 7–9, if d eccentricity (without the hinge-misalignment effects) from 4.2% to
is positive the upper hinge was displaced in the out-of-plane direc- 20.8% causes a reduction in the vertical displacement needed to
tion (the right side in the configuration of Fig. 1) with respect to the reach failure from 2.5% of the effective height to 0.5% in H-series
lower hinge. The contribution of the initial alignment (d) to the walls (i.e., a reduction of 80%). Similar behaviour is observed for
eccentricity at mid-height was systematically estimated as d/2 for the M-series walls. As shown in Fig. 8, an increase in the norma-
the present experimental campaign. lised initial eccentricity from 3.1% to 24.8% causes a 63% reduction
The walls were compressed until failure. Three different col- in the vertical descending movement needed to cause the failure of
lapse mechanisms were observed. The most common failure mech- an M-series wall. Similar to the results in Fig. 7, these results are
anism observed was the formation of a single horizontal crack at consistent with the fact that most of the walls, as mentioned, failed
approximately mid-height with a sudden out-of-plane displace- because of geometric instability.
ment and collapse of the entire wall. Tensile failure caused the for- The experimental results show that the initial shape (eccentric-
mation of a natural joint and was the mechanism leading to the ity) affects both the load-bearing capacity and the vertical defor-
corresponding wall collapse. This mechanism was observed for mation at the instant of failure. This effect increases as the
all walls (Fig. 4) except W#4, W#14 and W#17. Walls W#14 (M slenderness of the wall increases.
series) and W#17 (T series) failed in a mixed mode, i.e., a combina- The results are also scattered. The scatter could be attributed to
tion of joint opening at the tension side with masonry crushing at the lack of homogeneity in the component material properties, the
the compression side (Fig. 5). Wall W#4 failed by out-of-plane manual procedure used to build the walls and/or the variation in
shear near the upper edge of the wall. The results of Wall W#5 the curing processes. Tests performed on the material components
were not included in the analyses because they show an anoma- (summarised in Table 1) resulted in coefficients of variation
lous response that is inconsistent with the rest of the H-series slightly higher than those found in literature. The vertical position
walls. of the walls during the construction and the manual brick laying
The experimental results are summarised in Fig. 6. As expected, may also explain some of the scatter compared with other test
the results show that the load-bearing capacity decreases signifi- studies (e.g. [6]). In addition, imperfections in the wall geometry
cantly with increases in the load eccentricity and the slenderness. were measured in the present study and may partially explain
For an actual eccentricity of approximately 16% of the wall thick- the scatter in the results. These imperfections are acceptable
ness, changing the slenderness from 9.1 to 14.1 or 21.7 causes a according to the current manufacturing standards. The air curing
loss of the load-bearing capacity of 35.5% or 96.3%, respectively. conditions may also explain part of the scatter. Some of the walls
Similarly, the load-bearing capacity of the walls in the H series were built and cured in summer outdoor conditions, whereas oth-
(with a theoretical slenderness of approximately 21) decreases ers were produced in winter conditions, both indoors and
by approximately 75% if the eccentricity is increased from 0.3% outdoors.
to 16.1% of the wall thickness. The load-bearing capacity decreases The scatter in the results is comparable to the scatter typically
by approximately 80% if the eccentricity is reduced from 5% to observed in similar studies [26,27]. Although the scatter is signifi-
25.6% of the wall thickness for M-series walls. cant, these tests more closely represent actual masonry structures
The results in Fig. 7 exhibit a linear relationship between the than strict laboratory-controlled tests. The data obtained in this
vertical displacement of the walls and their maximum load-bear- study are representative of real structures. Nonetheless, scatter
6 E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13
Table 6
Geometric values used in the numerical simulation of the experimental campaign from [26].
Case et (mm) eb (mm) em (mm) d (mm) Hef (m) t (mm) hrow aprox. (mm)
HB-10-0 0 0 0 0 1.15 93 73
HB-20-0 2.12
HB-30-0 2.98
HB-40-0 3.95
HB-10-1/6 15.5 15.5 1.15
HB-20-1/6 2.12
HB-30-1/6 2.98
HB-40-1/6 3.95
HB-10-1/3 31.0 31.0 1.15
HB-20-1/3 2.12
HB-30-1/3 2.98
HB-40-1/3 3.95
Table 7
Geometric values used in the numerical simulation of the experimental campaign from [27].
Case et (mm) eb (mm) em (mm) d (mm) Hef (m) t (mm) hrow aprox. (mm)
5.6-0 0 0 0 0 0.635 115 105
11.1-0 1.25
18.8-0 2.12
27.7-0 3.12
5.6-1/8 14.4 14.4 14.4 0.635
11.1-1/8 1.25
18.8-1/8 2.12
27.7-1/8 3.12
5.6-1/4 28.8 28.8 28.8 0.635
11.1-1/4 1.25
18.8-1/4 2.12
27.7-1/4 3.12
5.6-1/3 38.3 38.3 38.3 0.635
11.1-1/3 1.25
18.8-1/3 2.12
27.7-1/3 3.12
be modelled accurately in all cases. A minimum 1-mm eccentricity experimental load-bearing capacity of the walls presented in
is always used to provide an initial imperfection to activate the [26,27] and the numerical results for different slenderness and
geometric non-linear response. All the models (see Fig. 3g–i) have eccentricities. Table 8 presents the geometric data for the simula-
been defined with the actual geometry, measured as discussed in tion of the walls tested in the current research. It must be noted
Section 3.2 for the present experimental campaign and available that previous researches do not provide comprehensive informa-
in [26,27] for the comparison cases tested by other authors. Tables tion about the lateral or vertical displacement of the walls at the
6–8 summarise all geometric data used in numeric models. The collapse load. In contrast, the work herein presented includes a
description of the geometry (see Fig. 3d–f) takes into consideration comparison of the numerical predictions of the lateral deformation
the eccentricity at each end of the wall (eb and et which are the of the walls at mid-height (Fig. 8) and the vertical descending
eccentricity at the bottom and top of the wall respectively), the movement of the top of the specimen with the corresponding
hinge alignment (d) and the actual out-of-plane geometric imper- experimental values (Fig. 7).
fections which determine the mid-height eccentricity, em. The
effective height (Hef), wall thickness (t) and average masonry row
4.2. Application of the standard codes EC-6 and ACI-530
height (hrow) are also presented in Tables 6–8.
A mapped mesh of 8-node quadrangular elements is used in
The load-bearing capacity of the walls is calculated using two
combination with the interface elements. The parameters required
standard codes, EC-6 and ACI-530 ([2,3]), which have been applied
by the numerical model are the modulus of elasticity (E), the Pois-
to calculate the load-bearing capacity of walls in the literature
son’s coefficient (t) and the compressive strength (fc) of the ma-
([26,27]). The comparison of the standards’ analytical results with
sonry. The model requires also the bonding strength (fxt) of the
the experimental results is used to determine the accuracy of these
brick–mortar interface and the fracture energy of the brick–mortar
two widely used codes.
interface of the first fracture mode (GIf ) as input data. In the lack of
The codes are compared using the experimental values of the
direct experimental information, the fracture energy has been esti-
modulus of linear deformation, E, and the masonry compressive
mated based on a linear fitting, based on experimental values from
strength, fc (Table 9). Safety factors are not considered because
[60], relating the fracture energy of the brick–mortar interface, GIf
the purpose of the calculation is the direct comparison with exper-
with the corresponding bonding strength fxt. The resulting correla-
imental data.
tion allows the estimation of the fracture energy as GIf ¼ 36:65 fxt ,
where GIf is given in N/m and fxt in MPa. Table 9 summarises the
values of all variables used for each simulation case. Tables 6 and 5. Comparison of results and discussion
7 show the comparison of the numerical predictions with all the
experimental results for the cases taken from the literature In this section, the experimental results obtained by the
[26,27]. Figs. 9 and 10 show a graphical comparison between the present study and by other researchers are compared with the
8 E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13
Fig. 4. Failure mode of wall W#10. Opening of a joint and mechanism formation.
Fig. 5. Failure mode of wall W#14. Crushing of the compressed side mixed with the tensile opening of two joints to form a mechanism.
predictions of the numerical model and with codes EC-6 [2] and ends. The same numerical approach is compared with the experi-
ACI-530 [3]. mental results of the tests described in Section 3. The goal of this
Experimental results from other researchers [26,27] are ana- comparison is to evaluate the accuracy of the simplified micro-
lysed, and the numerical model is validated by comparison to these model in predicting the load-bearing capacity of walls.
tests. These studies involve 28 different combinations of slender- The methods proposed by EC-6 [2] and ACI-530 [3] for the
ness and eccentricity (see Tables 6 and 7). All of them are compres- determination of the wall capacity are also applied and compared
sion tests on full-scale unreinforced masonry walls hinged at both with the experimental and numerical results.
E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13 9
Fig. 6. Experimental relationship between eccentricity and load-bearing capacity Fig. 8. Experimental relationship between actual eccentricity (not taking into
for three different theoretical values of slenderness corresponding to wall geom- account hinge alignment) and the descending vertical displacement at maximum
etries H, M and S. load for three different theoretical values of slenderness corresponding to wall
geometries H, M and S.
e = 0, the error was 6.4%; for e = t/3, the error was 19.5%. The slen-
derness did not seem to affect the accuracy of the model, as good
results were obtained regardless of the value of the slenderness.
On the whole, the model slightly overestimates the load-bearing
capacity of the walls. By contrast, the EC-6 code produces conser-
vative results, underestimating the capacity of the walls for all
cases of [26] (with an average relative error of 62.1%), while the
ACI-530 code shows greater accuracy, with an error of 45.3%, and
a tendency to underestimate the load-bearing capacity of the most
eccentrically loaded walls. Conversely, the ACI-530 code overesti-
mates the load-bearing capacity of the concentrically loaded walls.
In Fig. 10, the results from tests on walls with calcium silicate
units reported in [27], are compared with the results obtained with
the numerical model. Good agreement between experimental and
numerical results is shown, with an average relative error of 28.4%.
This error is within the range of the scatter in the experimental
Fig. 7. Experimental relationship between downward displacement at maximum results in [27]. The agreement is particularly good in the cases with
load and load-bearing capacity for three different theoretical values of slenderness lowest eccentricities and/or slenderness (4.2% for e = 0 and k ¼ 5:6)
corresponding to wall geometries H, M and S.
in contrast with the trend obtained for the results reported in [26].
The numerical model seems to overestimate the load-bearing
Experimental data from two sources, namely Watstein and capacity of the walls tested by Kirtschig and Asntötz [27]. The
Allen [26] and Kirstchig and Anstötz [27] were chosen for the com- EC-6 standard code conservatively underestimates the strength
parison because of the clear boundary conditions of the tests (the of almost all walls compared with the data in [27] (Fig. 10). In all
walls were hinged at both ends), the detailed information regard- the comparisons, EC-6 overestimates the load-bearing capacity in
ing the material properties and the fact that all walls were at full only two cases, those with the highest eccentricity. The ACI-530
scale. standard code overestimates most of the ultimate loads. It shows
Table 9 summarises the parameters used in each simulation. For more accurate results for the cases with low load eccentricity
Poisson’s coefficient, which is of little relevance for the type of 2D and overestimates the load-bearing capacity for the cases with
modelling utilised, a common value taken from the literature the greatest load eccentricity. Compared with these data, the aver-
[61,62] was used. age relative error of the EC-6 standard code is 32.3% and that of the
The brick–mortar bonding strength (fxt) for the walls tested in ACI-350 standard code is 28.9%.
[27] was estimated because it was not provided by the authors. The comparison of the numerical model with the results from
In this case, the value of fxt was chosen to satisfy the ratio fc/fxt [26,27] shows that the largest influence of the Young’s modulus
equal to the experimentally obtained value in the present study (E) over the axial load-bearing strength of the walls occurs when
(fc/fxt = 40). The rest of the parameters in Table 9 were obtained the load is eccentrically applied and second-order bending appears.
from the experimental tests. As the Young’s modulus increases, the deformation of the wall
The series of walls labelled as ‘‘HB’’ in [26] was chosen for com- from second-order effects decreases because of the increased stiff-
parison. The numerical model provided a very satisfactory predic- ness of the wall. A larger axial load is thus needed to develop sec-
tion of the corresponding ultimate wall load-bearing capacities. ond-order bending failure, and buckling becomes more likely.
The average relative error was 10.5%. Some calculated results were Similarly, larger flexural brick–mortar bonding strength (fxt) con-
higher than the experimental ones and others were lower, so the tributes by providing more strength against second-order bending
model appears to be balanced (see Fig. 9). It predicts without either failure. The load-bearing capacity of the walls can also be improved
a conservative or an unsafe bias. However, the results with the by increasing the masonry compressive strength (fc). However, the
least error were in the cases with the lowest eccentricities. For effect of the compressive strength is more apparent when the pre-
10 E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13
Table 8
Geometric values used in the numerical simulation of the current experimental campaign.
Case et (mm) eb (mm) em (mm) d (mm) Hef (m) t (mm) hrow aprox. (mm)
W#1 20 20 5.6 0.0 2.95 132 60
W#2 20 20 19.6 0.0 2.93
W#3 10 10 1.2 10.0 2.92
W#4 0 0 0 6.5 1.98
W#5 5 5 7.0 7.0 2.92
W#6 25 25 33.7 12.5 2.86
W#7 10 10 13.1 6.0 2.87
W#8 10 10 7.6 8.0 2.94
W#9 5 5 21.5 8.5 2.89
W#10 10 10 0.2 10.5 1.87
W#11 20 20 13.9 10.0 1.89
W#12 10 10 1.6 5.0 1.84
W#13 30 30 31.8 2.0 1.86
W#14 20 20 21.2 0.0 1.86
W#15 30 30 29.0 3.0 1.85
W#16 20 20 19.5 3.0 1.86
W#17 80 80 88.0 12.0 1.91 282
W#18 20 20 20.8 3.0 1.20 132
W#19 30 30 30.3 0.0 1.22
W#20 30 30 33.0 3.0 1.22
Table 9
Values of the parameters for the numerical model.
6. Conclusions
pressive behaviour model (linearly elastic and perfectly plastic). [17] Ainsworth M, Mihai LA. An adaptive multi-scale approach to the modelling of
masonry structures. Int J Numer Method Eng 2009;78:1135–63.
Future work to include an improved compressive stress–strain
[18] Chen S, Moon F, Yi T. A macroelement for the nonlinear analysis of in-plane
relationship in the model is recommended. Further future works unreinforced masonry piers. Eng Struct 2008;30:2242–52.
could also be oriented toward incorporating the out-of-plane shear [19] Mistler M, Anthoine a, Butenweg C. In-plane and out-of-plane homogenisation
in the analysis of unreinforced masonry walls. of masonry. Comput Struct 2007;85:1321–30.
[20] Ramalho M, Taliercio A, Anzani A, Binda L, Papa E. A numerical model for the
To obtain acceptable results with a finite element analysis, a description of the nonlinear behaviour of multi-leaf masonry walls. Adv Eng
realistic Young’s modulus must be used because it determines Softw 2008;39:249–57.
the deformation leading to the geometric instability associated [21] da Porto F, Mosele F, Modena C. Experimental testing of tall reinforced
masonry walls under out-of-plane actions. Constr Build Mater 2010;24:
with buckling failure. The formulations included in the standard 2559–71.
codes ([2,3]) are highly dependent on the value of Young’s modu- [22] Cavaleri L, Failla A, La Mendola L, Papia M. Experimental and analytical
lus. Given the variability and limited information on the value of response of masonry elements under eccentric vertical loads. Eng Struct
2005;27:1175–84.
Young’s modulus in masonry, it is necessary to do further work [23] Roca P. Simplified methods for assessment of masonry shear-walls. 6°
on the study of this essential parameter and its influence of the Congresso Nacional de Sismologia e Engenharia Sísmica; 2004. p. 101–18.
behaviour of slender and eccentrically loaded walls. [24] Charry J. Estudio experimental del comportamiento de paredes de obra de
fábrica de ladrillo ante la acción de cargas laerales; 2010.
These results emphasise that a realistic description of the joint [25] Murthy C, Hendry a. Model experiments in load bearing brickwork. Build Sci
behaviour is essential to accurately predict the load-bearing capac- 1966;1:289–98.
ity of unreinforced masonry walls, especially when the failure [26] Watstein D, Allen MH. Structural performance of clay masonry assemblages
built with high-bond organic-modified mortars. In: Second international brick
mode is primarily second-order bending.
masonry conference: structural clay products institute; 1970. p. 99–112.
[27] Kirtschig K, Asntötz W. Buckling tests on masonry. In: Proceedings of the 9th
international brick/block masonry conference. Berlin: IBMAC; 1991. p. 202–9.
Acknowledgements [28] Knutsson H. Vertical load bearing masonry – the danish approach. Mason Int
1991;5:23–6.
The authors wish to acknowledge Dr. Marco Antonio Pérez, Mr. [29] Maurenbrecher AHP. Effect of the test procedures on compressive strength of
masonry prisms. In: 2nd Canadian masonry symposium. Ottawa: National
Juan José Cruz, Mr. Francesc Puigvert and Mr. Christian Escrig for Research Council Canada; 1980. p. 119–32.
providing essential support in the experimental programme and [30] Maurenbrecher AHP. Compressive strength of eccentrally loaded
Dr. Jordi Marcé for his assistance in the development of the numer- masonry prisms. In: NRCC, editor. Third Canadian masonry
symposium. Edmonton: National Research Council Canada; 1983.
ical model.
[31] Maurenbrecher AHP. Axial compression tests on masonry walls and prisms. In:
This study has been partially performed within the project BIA Third North American masonry conference. Texas: National Research Council
2006-04127, funded by the Spanish Ministry of Science and Inno- Canada; 1985. p. 17.
[32] Maurenbrecher AHP. Compressive strength of hollow concrete blockwork. In:
vation, and the project MURETEX, funded by CDTI – Spanish Min-
Proceedings of the fourth Canadian masonry symposium. New
istry of Economics and Competitiveness, whose assistance is Brunswick: Institute for Research in Construction of National Research
gratefully acknowledged. Council Canada; 1986. p. 997–1009.
[33] Gazzola EA, Drysdale RG, Essawy AS. Bending of concrete masonry wallettes at
different angles to the bed joints. In: Proceedings of the third north american
References masonry conference. Arlington, Texas: Construction Research Center Civile
Engineering Department University of Texas; 1985. p. 15.
[34] Sinha B, Ng C, Pedreschi R. Failure criterion and behavior of brickwork in
[1] Mann W. Basics of design of masonry walls subjected to vertical loads and
biaxial bending. J Mater Civil Eng 1997;9:70–5.
buckling according to the present draft of Eurocode EC6 and comparison with
[35] Yokel FY. Stability and load capacity of members with no tensile strength. J
experimental data. In: Proceedings of the 9th international brick/block
Struct Div 1971;97:1913–26.
masonry conference. Berlin: Deutsche Gesellschaft fur Mauerwerksbau eV;
[36] Dawe JL, Liu Y. Analytical modeling of masonry load-bearing walls. Can J Civil
1991. p. 1281–91.
Eng 2003;30:795–806.
[2] Standardization ECf. Eurocode 6: design of masonry structures. Part 1-1:
[37] Lourenço PB. Analysis of masonry structures with interface elements. Theory
General rules for buildings. Rules for reinforced and unreinforced masonry;
and applications. TU-DELFT report no 03-21-22-0-011994. p. 34.
1997. p. 150.
[38] Salerno G, de Felice G. Continuum modeling of periodic brickwork. Int J Solids
[3] Committee MSJ. Building code requirements for masonry structures. ACI 530-
Struct 2009;46:1251–67.
05. Farmington Hills: American Concrete Institute; 2005.
[39] Martini K. Finite element studies in the two-way out-of-plane failure of
[4] Lu M, Schultz A, Stolarski H. Analysis of the influence of tensile strength on the
unreinforced masonry. In: Institute EER, editor. Proceedings of the 6th national
stability of eccentrically compressed slender unreinforced masonry walls
conference on earthquake engineering. Seattle; 1998. p. 1–12.
under lateral loads. J Struct Eng 2004;130:921.
[40] Cecchi A, Sab K. A homogenized Reissner–Mindlin model for orthotropic periodic
[5] Brencich A, Corradi C, Gambarotta L, Mantegazza G, Sterpi E. Compressive
plates: application to brickwork panels. Int J Solids Struct 2007;44:6055–79.
strength of solid clay brick masonry under eccentric loading. In: Proceedings
[41] Mercatoris BCN, Bouillard P, Massart TJ. Multi-scale detection of failure in
British Masonry Society; 2002. p. 37–46.
planar masonry thin shells using computational homogenisation. Eng Fract
[6] Sandoval C, Roca P, Bernat E, Gil L. Testing and numerical modelling of buckling
Mech 2009;76:479–99.
failure of masonry walls. Constr Build Mater 2011;25:4394–402.
[42] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry walls,
[7] Dajun D. Studies on brick masonry under compression. Mater Struct
Part II: Structural examples. Comput Struct 2006;84:181–95.
1997;30:247–52.
[43] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry walls,
[8] Defalco A, Lucchesi M. No tension beam-columns with bounded compressive
Part I: Failure surfaces. Comput Struct 2006;84:166–80.
strength and deformability undergoing eccentric vertical loads. IJMS
[44] Zucchini A, Lourenço PB. Homogenization of masonry using a micro-
2007;49:54–74.
mechanical model: Compressive behaviour. In: Soares CAM, editor. III
[9] Hansen L, Nielsen MP. Stability of masonry columns. orbitdtudk. Lyngby:
European conference on computational mechanics solids, structures and
Technical University of Denmark, Department of Civil Engineering, Section for
coupled problems in engineering. Lisbon, Portugal: ECCOMAS; 2006.
Structural, Engineering; 2003. p. 147.
[45] Reyes E, Casati M, Gálvez J. Experimental scale model study of cracking in brick
[10] Mura I. Stability of nonlinear masonry members under combined load. Comput
masonry under tensile and shear stress. Mater Construcc 2008;58:69–83.
Struct 2008;86:1579–93.
[46] Jager W, Bergander H. Comparison of buckling safety of masonry walls
[11] Romano F, Ganduscio S, Zingone G. Cracked nonlinear masonry stability under
according to EC6 and German standards. Proc Br Mason Soc 1998:279–83.
vertical and lateral loads. J Struct Eng 1993;119:69–87.
[47] Lijdens A, Villegas L. Aspectos del diseño de muros portantes de fábrica simple
[12] Schultz A, Mueffelman J, Ojard N. Critical axial loads for transversely loaded
según diferentes normas: excentricidades, reducción de la capacidad portante,
masonry walls. In: 12th int brick/block masonry conf proc; 2000. p. 1633–46.
e hipótesis de cargas a considerar. Rev Obras Pub 1994;141:43–53.
[13] Yokel FY. Strength of load bearing masonry walls. J Struct Div
[48] 83 CAC. UNE-EN 1015-11:2000/A1:2007. Métodos de ensayo de los morteros
1971;97:1593–609.
para albañilería. Parte 11: Determinación de la resistencia a flexión y a
[14] Orduña A. Seismic assessment of ancient masonry structures by rigid blocks
compresión del mortero endurecido. Madrid: AENOR; 2007.
limit analysis. PhD thesis 2003. University of Minho, Portugal. <www.civil.
[49] 41 CAC. UNE-EN 772-1:2002. Métodos de ensayo de piezas para fábrica de
uminho.pt/masonry> 2003. p. 171.
albañilería. Parte 1: Determinación de la resistencia a compresión. Madrid:
[15] Sacco E. A nonlinear homogenization procedure for periodic masonry. Eur J
AENOR; 2002.
Mech – A/Sol 2009;28:209–22.
[50] 41 CAC. UNE-EN 772-11:2001/A1:2006. Métodos de ensayo de piezas para
[16] Zucchini A, Lourenço PB. Mechanics of masonry in compression: results from a
fábrica de albañilería. Parte 11: Determinación de la absorción de agua por
homogenisation approach. Comput Struct 2007;85:193–204.
E. Bernat et al. / Construction and Building Materials 43 (2013) 1–13 13
capilaridad de piezas para fábrica de albañilería, en hormigón, piedra natural y [57] Lourenço PB, Rots JG. Multisurface interface model for analysis of masonry
artificial, y de la tas. Madrid: AENOR; 2006. structures. J Eng Mech 1997;123:660–8.
[51] Brencich A, Corradi C, Gambarotta L. Eccentrically loaded brickwork: [58] Milani G, Lourenço PB, Tralli A. 3D homogenized limit analysis of masonry
theoretical and experimental results. Eng Struct 2008;30:3629–43. buildings under horizontal loads. Eng Struct 2007;29:3134–48.
[52] Brencich A, Felice GD. Brickwork under eccentric compression: experimental [59] Alfano G, Crisfield MA. Finite element interface models for the delamination
results and macroscopic models. Constr Build Mater 2009;23:1935–46. anaylsis of laminated composites: mechanical and computational issues. Int J
[53] Brencich A, Gambarotta L. Mechanical response of solid clay brickwork under Numer Method Eng 2001;50:1701–36.
eccentric loading. Part I: Unreinforced masonry. Mater Struct 2005;38:257–66. [60] Pluijm R. Material properties of masonry and its components under tension
[54] Attard M, Nappi A, Tin-Loi F. Modeling fracture in masonry. J Struct Eng and shear. In: Neis VV, editor. 6th Canadian masonry symposium. Saskatoon,
2007;133:1385. Saskatchewan, Canada; 1992. p. 675–86.
[55] Brasile S, Casciaro R, Formica G. Multilevel approach for brick masonry walls – [61] Oliveira D. Mechanical characterization of stone and brick masonry. report 00-
Part I: A numerical strategy for the nonlinear analysis. Comput Method Appl M Dec/E-4, Universidade do Minho, Departamento de Engenharia Civil,
2007;196:4934–51. Guimarães, Portugal. Guimarães, Portugal; 2000. p. 1–77.
[56] Brasile S, Casciaro R, Formica G. Multilevel approach for brick masonry walls – [62] Rosas Rodríguez JG, Villegas L, Lorenzo Esperante D. The numerical models
Part II: On the use of equivalent continua. Comput Method Appl M front to the behaviour of masonry elements tested in laboratory. Informes
2007;196:4801–10. Construcc 2008:53.