Modelling of Reinforced Concrete Core Walls

Under Bi-directional Loading

R. Constantin & K. Beyer

Ecole Polytechnique Fdrale de Lausanne (EPFL), Switzerland

SUMMARY:
Reinforced concrete core walls with open sections are commonly used in practice as a lateral load resisting
system for multi-storey buildings. This type of walls has mainly been modelled in the past using simplified
models such as plastic hinge models or equivalent frame models. Such models are well suited for addressing the
global flexural behaviour of core walls and are commonly used for design purposes. However, shear behaviour
or transfer of stresses between the web and the flanges, for example, are either not captured correctly with such
models or they are not even addressed. These issues are particularly important when assessing the core wall
behaviour under diagonal loading. In order to estimate correctly the above mentioned quantities, a 3D multilayered shell element model for U-shaped walls is set-up. The local as well as the global behaviour is examined
and the behaviour of the wall under diagonal loading is investigated.
Keywords: U-shaped Wall, Shell Model, Diagonal Loading Direction, Out-of-plane Bending

1. INTRODUCTION
Many reinforced concrete (RC) buildings are designed to resist the lateral seismic loads by means of
RC core walls. Such walls commonly accommodate staircases or elevator shafts and provide an
important part of the lateral load resisting capacity of the building. Common types of RC core walls
are the open section core walls, such as U-shaped walls. Their behaviour is complex and relatively
unknown as compared to rectangular walls, although they are widely used in practice. Experimental
data for such walls is relatively scarce and recent (Reynouard and Fardis, Eds. 2001; Beyer et al.,
2008a). Hence rules in design codes were mainly based on experimental results of rectangular walls.
However rules for rectangular walls cannot be directly extrapolated to core walls as the behaviour of
core walls is considerably more complex than the behaviour of rectangular walls, e.g. with respect to
the shear stress distribution between flanges and web. A better understanding of the seismic behaviour
of RC core walls is therefore required. A few numerical studies of the inelastic behaviour for U-shaped
walls have been performed (Ile and Reynouard, 2005; Mazars et al., 2006; Beyer et al., 2008b; Beyer
et al., 2008c) with emphasis on the global response of the wall.
This paper investigates the global as well as the local behaviour of U-shaped walls under complex bidirectional lateral loading. To achieve this, a detailed numerical model is set-up and results are
compared to experimental data. The paper gives first a brief overview of the experimental test that is
used as reference for the modelling. Next, the set-up of the numerical model is described and the
experimental and numerical results are compared in terms of global and local behaviour. Finally, the
load transfer of the forces from the wall to the foundation is investigated when the wall is loaded in the
diagonal direction.

2. REFERENCE TEST
Two half scale U-shaped walls have been tested under bi-directional quasi-static cyclic loading at ETH
Zurich (Beyer et al., 2008a). The two test units differed with regard to their wall thickness: the first
test unit (TUA) had a wall thickness of 150 mm while the second one (TUB) had a wall thickness of
100 mm. The reference test used herein is TUA. The cross-section and elevation details for the TUA,
the labelling of different wall sections and the displacement loading history applied at the top of the
wall are shown in Fig. 2.1 and 2.2, respectively.
(c)

(c)

Figure 2.1. (a) Labelling of different wall sections and lines of action of the actuators, (b) bi-directional
displacement loading history and (c) wall instrumentation (taken from Beyer et al. [2008a]). All dimensions are
in mm.
(b)

650
75

150

150

150

25
25

collar (840 mm high,

300 mm thick)

(c)
wall

300

TUA

1050

150
100 100

tie 6 s = 50
tie 6 s = 50

wall

125

shear reinforcement
6 s = 125

7 12

300
footing
(without blocks)

125

ties 6 s = 50

86

25

150
150
13
00

shear span of the

EW actuator = 3350 mm

6 shear keys
(100 60 mm, 40 mm deep)

150

shear reinforcement
6 s = 125

shear span of the two

NS actuators = 2950 mm

4 12

150

100

12 6

100

840

(a)

footing
(without
blocks)

105

Figure 2.2. TUA geometry: (a) Cross-section, (b) shear keys and (c) elevation (taken from Beyer et al. [2008a]).
All dimensions are in mm.

The test units were subjected to a bi-directional loading history according to the loading pattern shown
in Fig. 2.1b:
reverse cycle parallel to the web (Position A and B, also termed EW cycle) ,
reverse cycle parallel to the flanges (Position C flange ends in compression and Position D
web in compression, also termed the NS cycle),
reverse cycle (Position E one flange end in compression and Position F one corner in
compression, also termed the diagonal cycle)
sweep cycle : O A G D C H B O
The pattern was repeated at displacement ductility levels of = 1, 2, 3, 4, 6 and 8 until failure. The
cycles at = 1 were preceded by four force-controlled cycles at 25%, 50%, 75% and 100% of the
predicted first yield lateral force.

During the entire test, the axial force was maintained constant and the rotation of the top of the wall
(the wall collar) was restrained by imposing equal displacements at the level of the NS-W and the NSE actuators (Fig. 2.1a). To investigate the evolution of the rotational stiffness, some small twists were
applied at Positions O, A, B, C and D during cycles at ductilities = 1 and 4; the rotational stiffness
is, however, not investigated within the scope of this paper. For further details on the experimental
data the reader is referred to Beyer et al. [2008a].

3. INELASTIC SHELL MODEL FOR U-SHAPED WALLS

The numerical analyses have been performed using the nonlinear finite element analysis software
VecTor4 developed at the University of Toronto (Wong and Vecchio, 2002). This software is
dedicated to the analysis of RC plates and shell structures. A smeared rotating crack formulation is
employed for the reinforced concrete which is based on the Modified Compression Field Theory
(Vecchio & Collins, 1986) and on the Disturbed Stress Field Theory (Vecchio, 2000).
TUA was modelled using multi-layered rectangular shell elements which account for the out-of-plane
shear response (Polak and Vecchio, 1993). A number of eight concrete layers of equal thickness were
used to model the thickness of the wall. The foundation was not modelled but instead all the degrees of
freedom of the nodes at the wall base were restrained. The collar of the wall was modelled with an
offset from the mid-surface of the wall so that the loading would be applied at the correct position. The
resulting mesh is shown in Fig. 3.1.

(a)

(b)

Figure 3.1. Geometry of the shell element model: (a) 3D mid-surface of the mesh and points of load application
and (b) cross-section mesh. All dimensions are in mm unless otherwise specified in the figure.

The material properties determined from material tests on reinforcement bars and concrete cylinders
were used as input parameters for the numerical model. Rupture of the reinforcement bars was,
however, not accounted for in the model. The longitudinal and the transversal reinforcements were
both modelled as smeared. In addition, for the confined concrete areas of the wall, out-of-plane
reinforcement was assigned since the concrete model can account for the tri-axial behaviour of
concrete. Perfect bond was assumed between the concrete and the reinforcement. The resulting
concrete and reinforcement stress-strain relationships are plotted in Fig. 3.2.
The loading was applied by imposing displacements at the positions of the actuators from the
experimental test as shown in Fig. 3.1a and Fig. 2.1a. The NS-E and NS-W displacements were kept
equal throughout all loading steps and patterns, restraining the rotation at the top of the wall.

(a)

80

(b)

D12
D6

600

70
60
50

Stress [N/mm2]

Stress [N/mm2]

400
200
0
200

40
30
20
10

400

Concrete

0
Reinforcement

600
0.08

0.06

0.04

0.02

0
0.02
Strain []

0.04

0.06

10
0

0.08

0.5

1.5
Strain []

2.5

3
3

x 10

Figure 3.2. Material models: (a) reinforcement model and (b) concrete model: unconfined concrete in cyclic
compression.

4. COMPARISON OF NUMERICAL AND EXPERIMENTAL RESULTS

4.1. Global behaviour: force-displacement hysteresis
The global behaviour of the wall is studied in terms of force-displacement hysteresis plotted
individually for each loading direction: the EW direction (Positions A and B), the NS direction
(Positions C and D) and the diagonal direction (Positions E and F). Results from pushover and cyclic
analyses are compared to the experimental results (Fig. 4.1). Separate cyclic analyses were performed
along these three directions; a full displacement load history was not performed so far because of large
computational expense and convergence problems.
(a)

500

300

100
0
100

300

Pos.A

80

60

40

20
0
20
40
Top displacement EW [mm]

(c)

2000
1500

60

100
0
100
200

400

80

100

Pos.D

200

300

EW cycles

400
500
100

Experimental
Numerical cyclic
Numerical pushover

300

Pos.B

200

200

500
400

Actuator force NS [kN]

Actuator force EW [kN]

(b)

Experimental
Numerical cyclic
Numerical pushover

400

NS cycles
Pos.C

500
100

80

60

Experimental
Numerical cyclic
Numerical pushover

40

20
0
20
40
Top displacement NS [mm]

60

80

100

120

Pos.F

SRSS Moment [kNm]

1000
500
0
500
1000
Diag. Cycles

1500
2000
80

Pos.E
60

40

20
0
20
40
SRSS displacement [mm]

60

80

100

Figure 4.1. Comparison of the results obtained from VecTor4 with the experimental results: Force-displacement
hystereses for cycles in the EW and the NS directions (a-b). Moment-displacement hystereses for the diagonal
direction (c). Pushover analyses results are also included (a-c).

There is a good match of the numerical results with the experimental ones in terms of force capacity
and hysteresis shape for the loading directions EW (parallel to the web) and NS (parallel to the
flanges). For the diagonal direction, the pushover analysis as well as the cyclic analysis overestimates
by approximately 25% the experimental values of the moment capacity. It was, however, argued
(Beyer et al., 2008b) that the moment capacities reached during the experiments for the diagonal
direction were not the ultimate moment capacities of the wall for that direction. Within the cycles of
one ductility level, the wall was loaded in diagonal direction after it had been loaded in the two
principal directions (EW and NS direction). Hence, when the wall was loaded in the diagonal
direction, the stiffness was reduced and as a result the full moment capacity was not reached at
Positions E and F. To investigate whether the shell element model is able to capture the behaviour in
the diagonal direction, it is therefore necessary to perform an analysis with the full displacement
loading history.
4.2. Local behaviour
The local behaviour of the wall was investigated in terms of the vertical strain profile at Position E and
in terms of curvature profiles at Positions A, B, C, D, E and F at ductility = 3.0.
(a)

200 mm

200 mm
200 mm
50 mm

(b)
West flange

Web

East flange

170 mm
170 mm
170 mm

Scale:10mm/m

Figure 4.2. Comparison of the results obtained from VecTor4 with the experimental results at Position E during
the = 3.0 cycles (at Position E the west flange end is in compression): (a) vertical strain profile of the wall
experimentally determined from Demec measurements and (b) vertical strain profile of the wall numerically
determined.

4.2.1. Vertical strain profile

The vertical strain profile computed from Demec measurements (Fig. 4.2a) is compared to the
numerically determined vertical strain profile (Fig. 4.2b). The Demec measurements were taken on the
inside faces of the wall, recording the vertical strain values over equal vertical distances of 200 mm
and starting 50 mm above the base as it is shown in Fig. 4.2a. The numerical vertical strain values
were computed as average vertical strains of the elements in the model mesh shown in Fig. 3.1a. All
elements in the mesh had a height of 170 mm, as shown in Fig. 4.2b and the average vertical strains
were computed from the vertical displacements of the four corner nodes of each rectangular element.
Considering the vertical strain values shown in Fig. 4.2 and the heights over which they have been
computed the numerical values match in general and the experimental values of the vertical strains
rather well. The vertical strains computed from the numerical model tend to concentrate stronger
towards the base while the experimental ones are more distributed over the height of the wall. Note,
however, that the experimental strains computed from the base crack opening are not included in Fig.
4.2a. At the end of the west flange, it can be, however, noticed that the experimental strains indicate
vertical tensile strains for the flange end which is compression, whereas the numerical strains are
compressive strains. This can be explained by looking at the deformed shape of the wall at Position E
at the peak of the = 3.0 cycle (Fig. 5.3). Near the base of the wall (Fig. 5.3d) the west flange
presents horizontal out-of-plane bending, with the inside part of the flange end in tension while the
outside part is in compression. The Demec measurements were taken on the inside face of the wall,
hence the tensile vertical strains. The numerical strain values are computed from the vertical
displacements of the nodes in the mid-surface of the wall, which explains the compressive strains. The
overall vertical strain profile from numerical results is found therefore to be consistent with the
experimental one.
4.2.2. Curvature profile
The experimentally determined curvatures, from LVDT measurements, are plotted against the
numerically determined curvatures in Fig. 4.3. Experimental values of the curvatures were determined
from the LVDT chains along the four outside edges of the wall (Fig. 2.1c). The numerical values of the
curvatures were determined from the vertical displacements of the nodes in the mesh at the same four
edges of the wall where the LVDT chains were positioned.
Numerically determined curvatures for the EW direction (Position A and B) fit the experimental
curvature values with some discrepancies: the predicted curvature profile indicates the presence of
important shear cracks in the web, which were not noticed during the experiments. For Positions D and
F the results match fairly well, while for Positions C and E the predicted curvatures underestimate the
experimental values of the curvatures at heights of about 0.2 of the wall height.

5. BEHAVIOUR UNDER DIAGONAL LOADING

Beyer et al. (2008a) argued that the diagonal loading direction is the most complex one in terms of
load transfer mechanism. Therefore in order to gain a better understanding of the behaviour of the wall
under diagonal loading, the transfer of forces from the wall to the foundation, as well as the deformed
shape of the wall, was investigated. The following section presents these results obtained with the
numerical model described in Section 3, with emphasis on Position E, when the west flange end is in
compression.
5.1. Transfer of forces from wall to foundation
The distribution of reaction forces at the base of the wall, in Position E at a displacement ductility of
= 3, is presented in Fig. 5.1. As expected, the shear forces are transferred from wall to foundation
mainly within the compressed zones. The distribution of the vertical reaction forces indicates that the
compressed zones develop at the end of the west flange, as expected, but also at the corner of the west
flange and the web. The assumption of plane sections remaining plane is therefore no longer valid.

Figure 4.3. Comparison of the results obtained from VecTor4 with the experimental results during the = 3.0
cycles at Positions A, B, C, D, E and F: curvatures numerically determined from the nodal displacements of the
mesh elements (VT4) compared with curvatures experimentally determined from LVDT measurements.

Table 5.1 summarises the distribution, between different wall sections, of the total applied shear force
in the X and in the Y direction, transferred from the wall to the foundation. For the X direction, the
reactions at the corner nodes between the flanges and the web are counted towards the web, while for
Y direction, the same corner nodes are counted towards the flanges.
+

+
+

Y
X
Nodal reaction forces in the X direction

X
Nodal reaction forces in the Y direction

+ = tension force
= compression force
Nodal vertical reaction forces

Figure 5.1. Numerically determined qualitative distribution of reaction forces at the base of the wall at Position
E at the peak of the = 3 cycle (at Position E the west flange end is in compression).
Table 5.1. Numerically determined distribution of reaction forces at the base between the web and flanges at
Position E at the peak of the = 3 cycle (at Position E the west flange end is in compression).
Reaction component (Position E)
West flange
Web
East flange
Shear force in the X direction
15%
83%
2%
Shear force in the Y direction
124%
~0%
-24%

As expected, the shear force in the X direction is mainly transferred through the web (83% of the
applied shear force in the X direction). The west flange takes approximately 15% of the shear force in
the X direction as an out-of-plane shear force, mainly through the compressed zones next to the corner
and at the free end of the west flange.
The west flange transfers to the foundation approximately 124% of the total shear force applied in Y
direction while the reaction forces in the east flange are of the opposite sign and account for 24% of
the applied shear force. The large shear force in the Y direction that is transferred by the west flange to
the foundation is the result of two effects: 1) the east flange is under tension and therefore has limited
in-plane shear carrying capacity and 2) since the rotation at the top of the wall is restrained, a torsional
moment in the wall appears, which produces additional in-plane shear forces in the wall sections.
+

+
+

Y
X
Nodal reaction forces in the X direction

X
Nodal reaction forces in the Y direction

+ = tension force
= compression force
Nodal vertical reaction forces

Figure 5.2. Numerically determined qualitative distribution of reaction forces at the base of the wall at Position
F at the peak of the = 3 cycle (at Position F the corner of the web and of the east flange is in compression).
Table 5.2. Numerically determined distribution of reaction forces at the base between the web and flanges at
Position F at the peak of the = 3 cycle (at Position F the corner of the web and of the east flange is in
compression).
Reaction component (Position F)
West flange
Web
East flange
Shear force in the X direction
-1%
94%
7%
Shear force in the Y direction
16%
21%
63%

For Position F, when the corner between the web and the east flange is in compression, the distribution
of reaction forces is shown in Fig. 5.2 with the amount of shear force transferred by each wall section
to the foundation summarised in Table 5.2. The distribution of reaction forces indicates that the
compressed zone is now limited only to the compressed corner, where most of the shear forces are
transferred from the wall to the foundation. In this case the assumption of the plane sections
remaining plane still holds approximately. As indicated in Table 5.2, almost the entire shear force in
the X direction is transferred through the web, as expected. It is interesting to notice that
approximately 20% of the shear force in the Y direction is transferred to the foundation through the
web as an out-of-plane force. As in the case of loading to Position E, the out-of-plane shear force is
transferred through the part of the wall which is under compression, namely the east corner of the web.

5.2. Deformed shape of the wall

The deformed shape of the wall at Position E at a displacement ductility of = 3 is shown in Fig. 5.3.
The mid-surface of the mesh elements (undeformed and deformed) is plotted three-dimensionally and
at representative cross-sections of the wall over its height: at fractions of 0.06, 0.2 and 1 of the wall
height, termed as cross-section 0.06 (CS0.06), cross-section 0.2 (CS0.2) and cross-section 1 (CS1),
respectively. It can be seen that the web as well as the west flange deform in out-of-plane bending
close to the base of the wall (Fig. 5.3c, d, e).

Again at the base of the west flange, the nodes of the mesh deform in an arched shape (Fig. 5.3b)
suggesting an arch action between the two compressed zones at the corner and the end of the west
flange. At the top of the wall the rotation is restricted (Fig. 5.3f).
Undeformed
Deformed

(a)

(b)
0.5

3
2.5

1.5
[m]

[m]

0.5

0.5
0
0
0.5

1
1.5

1.5

[m]

0.5

0.5
1.5
1.5

[m]

0.5

0.5

0.5

[m]

0.5

Magnification factor-50x

(c)

(d)

0.5

0.4
0.3
0.2
0.5

East flange

Web

West flange
0.5

[m]

1.5

[m]

[m]

0.5

0
1

[m]

1.5
1.5
Magnification factor50x

(e)

(f)

0.5

0.5

0.5

[m]

[m]

0.5

0.5

1.5
1.5

[m]

0.5

0.5

[m]

0.5

1.5
1.5

[m]

Magnification factor25x

Figure 5.3. Numerically determined deformed shape of the wall at Position E at the peak of the = 3 cycle (at
Position E the west flange end is in compression): (a) 3D deformed shape, (b) 3D deformed shape view from
the top, (c) 3D deformed shape of the element nodes for CS0.06, (d) CS0.06, (e) CS0.2 and (f) CS1. The
displacements were magnified by factors of 50 or 25 as indicated below the subfigures.

6. CONCLUSIONS AND OUTLOOK

The objective of this article was to investigate the diagonal behaviour of RC U-shaped walls under bidirectional lateral loading. For this purpose a shell element model was set-up and used for
investigating the load transfer of forces from the wall to the foundation under diagonal loading.
Predicted global force capacities from separate cyclic analyses along each direction matched fairly
well with the experimental results, except for the diagonal loading direction where the predicted
moment capacity was higher than in the experiments. A full displacement loading history needs to be
applied to check whether the model can accurately predict the experimental values of the moment for
the diagonal loading direction, this being the next step in validation of the shell element model.
Local behaviour in terms of vertical strains was checked for the diagonal direction, at Position E, and
found to be predicted in a satisfactory manner. The curvatures were checked for several loading
directions and some discrepancies were found at Positions A and B (EW cycles). Shear and flexural
deformations should be compared separately to find the source of these differences.
It was also found that under diagonal loading, approximately 15% to 20% of the shear force is
transferred to the foundation as an out-of-plane force either by the flanges (when the flange end is in
compression), or the web (when the corner of the web and of the flange is in compression).
Furthermore, at diagonal loading with one flange end in compression, the assumption of plane
sections remaining plane was found to be no longer valid. Compression zones appeared at the end of
the flange and also at its corner with the web. These zones were responsible for the transfer of the
shear forces from the wall to the foundation. A complex deformed shape with out-of-plane bending in
the web and in the west flange suggest that further efforts need to be directed in understanding the load
path of forces through the wall from the load application points to the foundation for the diagonal
loading direction.
ACKNOWLEDGEMENT
The authors would like to thank to Prof. F.J. Vecchio for making the VecTor4 software available to us as well as
to T. Hrynyk for all his kind help with the VecTor4 software.

REFERENCES
Beyer, K., Dazio, A. and Priestley, M.J.N. (2008a). Quasi-static tests of two U-shaped reinforced concrete walls.
Journal of Earthquake Engineering 12:7, 1023-1053.
Beyer, K., Dazio, A. and Priestley, M.J.N. (2008b). Inelastic wide-column models for U-shaped reinforced
concrete walls. Journal of Earthquake Engineering 12:S1, 1-33.
Beyer, K., Dazio, A. and Priestley, M.J.N. (2008c). Elastic and inelastic wide-column models for RC nonrectangular walls. Proceedings of the 14th World Conference on Earthquake Engineering, Beijing, China.
Ile, N. and Reynouard, J.M. (2005). Behaviour of U-shaped walls subjected to uniaxial loading and biaxial cyclic
lateral laoding. Journal of Earthquake Engineering 9:1, 67-94.
Mazars, J., Kotronis, P., Raguneau, F. and Casaux, G. (2006). Using multifiber beams to account for shear and
torsion applications to concrete structural elements. Computer Methods in Applied Mehcanics and
Engineering 195, 7264-7281.
Polak, M.A. and Vecchio, F.J. (1993). Nonlinear analysis of reinforced concrete shells. Journal of structural
Engineering 119:12, 3439-3462.
Reynouard, J.M. and Fardis, M.N., Editors (2001). Shear wall structures. Thematic report No. 5, Lisbon,
Portugal.
Vecchio, F.J. and Collins, M.P. (1986). The modified compression-field theory for reinforced concrete elements
subjected to shear . ACI Journal 83:2, 219-231.
Vecchio, F.J., Lai, D., Whim, W. and Ng, J. (2001). Disturbed stress field model for reinforced concrete:
Validation. Journal of Structural Engineering 127:4, 350-358.
Wong, P. and Vecchio, F.J. (2002). VecTor2 and Formworks Manual. University of Toronto, Department of
Civil Engineering, Canada.