You are on page 1of 14

Proc. Instn Civ.

Engrs,Part 2,1982,13, June, 4 4 1 4 5 4

8577 Non-linear analysis of


multistorey infilled frames

TE-CHANG LIAUM, DIC. PhD, MICE*


KWOK-HUNG KWAN, BSc(Eng)*

In this Paper, multistorey infilled frames with or without connectors are analysed by a finite
element approach which takes into account non-linearities of the materials andthe
structural interface between the infill and the frame. Where connectors are provided, the
non-linear shear-slip characteristic of the structural interface is accounted for. Where
connectors are not provided, separation, slip and development of friction are allowed. In
addition, theinitial lack of fit at the structuralinterface is also considered. In order to predict
failure and study theentire range of load-deflexion behaviour, cracking, crushing and
non-linear stress-strain behaviour of the infill and yielding of the frame are all taken into
account. Typical model structures are analysed and compared with experimental results to
demonstrate theversatility and accuracy of non-linear analysis.

Notation
D elasticity matrix of infill material
E Young's modulus of infill material
F" F, normal and shear forces at interface
1

h eccentricity of interface
k , k, normal and shear stiffness of interface
T transformation matrix for anisotropy of infill material
a the angle between the normal of the crack surface and the
x-axis
shear strain
gap width at the interface due to initial lack of fit
relative normal displacement and relative shear displacement
(slip) of interface
previous and new residual slips at interface
normal strain
coefficient of friction at interface
Poisson's ratio of infill material
normal stress
shear stress
Introduction
The earliest attempt to analyse infilled frames without connectors by the finite
element method seems to be that of Mallick and Severn' in 1967. In their analysis,
separation at the structural interface was allowed where tensile normal stress was
developed. It was assumed that slip would occur along the structural interface
remaining in contact so that therewouldbe either limiting friction atthe
unseparated interface or no friction at the separated interface. Furthermore, the

Written discussion closes 16 August 1982, for publication in Proceedings, Part 2.


* University of Hong Kong.
441
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
LIAUW AND KWAN

axial deformation of the frame was neglected and the part of the frame in contact
with the infill wasassumed to remain straight. In 1977 Mallick and Barua2refined
the method of analysis by taking into account the axial deformation of the frame,
but it was still assumed that slip would occur and hence limiting friction would be
induced along the unseparated interface. This has the weakness that the limiting
friction might not be mobilized if slip were restrained at the contact area.
2. In 1977 Riddington and Stafford Smith first introduced short, stiff linking
members as interface elements whereby the linking members at the separated
interface were removed. At areas of contact, two possibilities were allowed
(a) no s l i p t h e short, stiff linking member forced both nodes at the interface
to have identical displacement so that friction was developed with no
slip
(b) no friction-the linking matrix was of a form which forced the two nodes
to have equal displacements only normal to the interface so that slip
occurred with no friction.
Unfortunately, this interface model was too crude, especially in the case where
friction was developed with no slip.
3. In 1978 King and Pandey4 used the friction elements originally developed
by Goodman in their analysis. With the useof these friction elements, more
accurate interface behaviour can be simulated by adjusting the element properties
according to the interface condition. Their idealized elastoplastic frictionslip
model whichwas based on shear box tests was much more realistic than the
previous models. In the computation,they employed the incremental load-tangent
stiffness-incremental stress method. However, the reliefof friction due to a re-
duction of normal stress or separation has not been taken into account in their
iteration method.
4. The improvements offered by providing connectors at thestructural
interface have long been recognized.5p6However, there have beenvery few
attempts on the finite element analysis of infilled frames with connectors. Mallick
and Garg tackled the problem in 1971, assuming the frame to remain in contact
with the infill along all of the interface except at the unloaded corners, and the
interaction forces between the frame and the infill to consist only of normal forces,
i.e. the shear force at the interface was neglected.This last assumption is not really
acceptable because the main contribution of the connectors at the interface is in
the transmission of the lateral shear to the infill.
5. Experimental study by various investigators has shown the existence of a
considerable non-linear phase before collapse. Thoughthe behaviour of the
structures near failure is inevitably complex, it is of utmost importance in
evaluating strength and ductility under strong wind and seismic excitation.
6. As part of a research project where the non-linear behaviour of various
types of infilled frames and their performances are studied, this Paper aims at a
deeper understanding of the non-linear behaviour of typical model structures and
presents the non-linear analysis of infilled frames both with or without connectors.

Method of analysis
7. An iterative procedure with incremental displacement (simulating a loading
process under displacement control) is used. During the iteration process,
prescribed displacement is applied to thestructure in increments. At each
displacement step, the structure is analysed by using the secant stiffness at the
442

Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
ANALYSIS OF MULTISTOREY INFILLED FRAMES

~ previous displacement level; after the stresses are evaluated, the secant stiffness is
modified according to the new stress state.
8. The more popular tangent stiffness-incremental stress method is not
employed here for the following reasons.
(a) Since cracking of the infill material is to be allowed, the stress-strain curve
of the infill material is discontinuous leading to undefined tangent stiff-
ness at these discontinuities.
(b) When no connectors are provided, the friction at the structural interface
may be relieved when the normal forces are somehow reduced or even
eliminated by separation. This relief of stresses would induce stress
redistribution in the structure for which the tangent stiffness-
incremental stress method becomes unsuitable. This redistribution of
stresses would give no problem if secant stiffness were adopted because
the stresses are always recalculated.

Finite element formulation


9. The infilled frame structure is modelled by using three types of elements
(Fig. 1): the panel elements which constitute the infilled panels, the frame elements
which constitute the skeletal frame, and the interface elements which simulate the
behaviour of the structural interface.
Interface element
10. The interface elements have three degrees of freedom at the nodes
connected to frame elements and two at nodes connected to panel elements. Since
the skeletal frame is replaced by frame elements residing at the axis of the frame, a
rigid rod is added between the axis of the frame and the interface to account for the
eccentricity (Fig. 2).
11. The interface behaviour is described by the equations

1
f f f f
Frame
~

/
//
Panel

L
4
Fig. 1. The finite elements (i = interface element, p = panel element, f = frame
element)
443
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
LIAUW AND KWAN

Interface
Panel AXIS of frame

Fig. 2. Structural interface between infilled panels and frame

12. Taking into account the eccentricity, the equations of equilibrium with
respect to local co-ordinates become

13. Where connectors are not provided, separation and slip are allowed by
adjusting their interface properties according to the interface condition. When the
normal force of the interface is tensile (i.e. An + Af > 0),separation is assumed to
have occurred and the normal stiffness k, is then taken as zero. Since a separated
interface cannot take up any shearing force, the shear stiffness k, is also taken as
zero. For an unseparated interface, the normal stiffness is perfectly rigid (i.e. kn is
taken as very large) and shear forces are taken up by friction, the behaviour of
which is idealized as shown in Fig. 3.
14. The shear force (friction) is given by

Fa=
I k s ( ~ s - ~ r )
PIFnI
if I F ~GpIFnI
I
if k a ( 4 - 4 ) > P I Fn I (4)
1 -PlFnl if k,(A, -- Ar) < - p I F, I
in which A, is the residual slip of the interface when the shear force of the interface
at the previous iteration step is released. It is also the slip when a formerly opened
interface closes again. After each iteration step, the residual slip is renewedto
444
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
ANALYSIS OF MULTISTOREY INFILLED FRAMES

F,
t

Fig. 3. Friction-slip relation of interface with no connectors

F,

Fig. 4. Shear-slip relation of interface with connectors

A: = As - (FJkJ (5)
15. Initially, when load is first applied, interface elements with no lack of fit are
assumed to be unseparated and interface elements with lack of fit are assumed to
be separated. Furthermore, the initial value of the residual slip of all interface
elements istaken aszero.
16. Where connectorsare provided, thenormal stiffnessisassumed to be
perfectly rigid while the shearslip relation, which is to be independent of the
normal stress, is represented by a multilinear interpolation (Fig. 4). Unlike the
interface without connectors, both initial lack of fit and residual slip are zero. It
should be noted that k, is now the secant shear stiffness, the definition of which is
shown in Fig. 4. It is also assumed that after shear failure which occurs when the
magnitude of slip exceeds a certain limit, the interface behaves as if there are no
connectors.

Panel element
17. In the panels, constant strain, three-noded, triangular plane-stress elements
having two degreesof freedom at each node areused.
18. In tension, the panel material is idealized as a linearly elastic, brittle ma-
445
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
L l A U W A N D KWAN
terial, assuming that cracks occur when the principal tensile stress exceeds the
tensile strength. Bearing in mind that cracks may be closed and reopened again, a
cracked element is checked in every iteration step for closure and reopening: a
crack is closed and opened up if the stress normal to the crack surface is compress-
ive and tensile respectively.
19. Before cracking, the material is assumed to be isotropic and the elasticity
matrix, which is defined by

[g =D[ ;]
is givenas

E
- Ev
- 0
1-v2 1-VZ

Ev
- E
-
D= 0
1-v2 1-v2

20. After cracking, due to the presence of the crack surface, the material
becomes anisotropic. For a cracked element, whichis opened, the Youngs
modulus normal to the crack surface and the shear modulus parallel to the crack
surface are taken as zero. Hence the elasticity matrix for a cracked element which
is opened becomes

where
[:::l
D = T 0 E 0 T

T= [ cosz a
sin2 a
sin a cos a
sin2 a
cos2 a
-sin a cos a
-2 sin a cos a
2 sin a cos a
cos a -sin2 a 1
in which a is the angle between the normal of the crack surface and the x-axis.
When a crack is closed, the Youngs modulus normal to the crack surface is
restored to the uncracked value and the shear stress of the crack surface is assumed
to be taken upby friction in a way similar to interface elements.
21. In-compression, the panel material exhibits extensive non-linearity in the
stress-strain relation. As the infill panels are two-dimensional, the panel material is
under biaxial stresses. However, the biaxial stress state is approximately uniaxial
(i.e. one principal stress is much smaller than the other) as can be inferred from the
following. When no connectors are provided, the panels act as diagonal struts so
that they are subjected to nearly uniaxial stresses. In the case when connectors are
provided, the numerous and extensive cracks remove the stresses normal to the
446
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
ANALYSIS OF MULTISTOREY INFILLED FRAMES

crack surface so that the stresses approximate to the uniaxial state again. In both
cases, this has been confirmed by strain gauge measurements of the panel stresses
in the Authors experimental investigation.
22. If it is assumed that the stress states are approximately uniaxial, the
analysis can be simplified with the following assumptions. .
(a) The anisotropyis due solely to cracking. Thus for uncracked elements the
material is isotropic; and for close cracked elements the Youngs
modulus normal and parallel to the crack surface are equal. Hence in
every case (whether uncracked, open cracked or close cracked) only a
single valueof Youngs modulus is needed.
(b) The Youngs modulus varies asa function of the larger compressive
principal strain as in the uniaxial case.
(c) The Poissons ratio remains constant at any stress level.
(d)The panel is crushed when the principal compressive strain exceeds the
ultimate limit. Once the material fails in compression, it can never
regain any strength.
23. One very important simplification with the above assumptions is that the
secant stiffness matrix is symmetric.

Frame element
24. The frame elements are standard prismatic bending elements having three
degrees of freedom at each node. The material is idealizedas elastoplastic. In each
iteration step, the axial force and bending moment are integrated from the normal
stress in the section and thesecant stiffness is modifiedaccordingly. The stiffness is
assumed constant along the length of the element and is taken as that of the
midsection.

Results and discussion


25. The results of two typical models, chosen from a model study undertaken
by the Authors, are presented here. The two models were identical except that one
was without connectors and the other was provided with connectors. Details of the
models are shown in Fig. 5.
26. The compressive stress-strain curve of the microconcrete infill, which was
measured from cylinder tests, is shown in Fig. 6. The Poissons ratio and tensile
strength of the microconcrete were taken as 0.2 and one tenth of compressive
strength respectively. Youngs modulus and the plastic moment of the mild steel
frame were 202 KN/mm2 and 839 Nm. The shear-slip relation of the connectors
was obtained by push-out tests, the results of which are shown in Fig. 7. Shear
stiffness and coefficient of friction at the interface without connectors were meas-
ured by shear box test and found to be 145 N/mm2 and 0.42respectively.

Load-deflexion characteristics
27. Loading was applied at the level of top beams under displacement control.
The load4eflexion curves of the analytical and experimental models for deflexion
up to 2% of the height of the models are shown in Fig. 8.
28. In the model without connectors, initial lack of fit was observed in the
experimental investigation. Shrinkage of the infill material, which was the main
cause of the lack of fit, was measured to be 0.02%. Gap width at the interface
corresponding to this measured shrinkage was used in the analysis and was found
447
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
LIAUW AND KWAN

61 0

22 X 22 Solid
mild steel bar

Microconcrete
infill 305
610x305~22
22

305

22

305

22

Fig. 5. Details of models

Strain:96
Fig. 6. Compressive stress-strain curveof infill material (solid line = cylinder test result;
dotted line = curve used in analysis)

448

Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
A N A L Y S I S O F M U L T I S T O R E Y I N F I L L E D FRAMES

Slip: mm
Fig. 7. Shearslip characteristic of connectors (solid line = push-out test result; dotted
line = curve used in analysis)

80[
60

0 5 2010 15 25
Deflexion at top beam: mm
( b)
Fig. 8. Load-deflexion curves of models: (a) without connectors; (b) with connectors
449
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
LIAUW AND KWAN

to give very good agreement in the initial part of the load-deflexion curves. The
stiffness increased sharply after the frame had gained firm contact with the panels.
Due to the high stress concentration and the consequent yielding at the
compressive corners of the infill, the stiffness gradually decreased until the corners
were crushed at peak load. After crushing, the model continued to sustain
substantial loading (more than 85% of the peak load) for a very large range of
deflexion.
29. For the model with connectors, the analytical and experimental load-
deflexion curves showed similar patterns. However, the measured lateral deflexion
was in general larger than the analytical result by about 12%. Analysis showed
that the model started cracking at about 2 mm deflexion whilein the experimental
investigation, cracks were only observable at deflexion greater than 3 mm. The
compressive corners of the infilled panels were crushed at about7 mm (analytical)
or 8 mm (experimental) deflexion. After crushing, the model continued to with-
stand even higher load until the infill-beam connection yielded. Finally, the model
failed by shear failure of the infill-bottom beam connection and the subsequent
tension failure of the infill-column connection at the column nearer to the load
together with the compressive failure of the infill at the infill-column interface
further away from the load.

Panel stresses
30. Panel stresses of both models are shown in Fig. 9. Since the panel stresses
were similar in different storeys, only those of the third storey are shown. It can be
seen from both the analytical and experimental results that the panel stresses were
approximately uniaxial+onfirming the assumption made in the analysis.
31. When the connectors were not provided, the stress concentration at the
compressive corners was very high. This high stress concentration led to early
crushing of the compressive corners, resulting in low strength of the structure.
When connectors were provided, the panel stresses were different before and after
theformation of cracks. Before cracking, each panel acted as two diagonal
struts-ne in tension and one in compression. Hence stress concentration
occurred at all the four corners. In the experimental model, the tensile stresses
measured at the tensile corners were smaller than expected, showing that cracks
might have already formed there, though they were not observed. After cracking,
due tothe failure of the tensile corners, each panel acted as a single diagonal strut,
resulting in slightly higher stress concentration at the compressive corners. Since
part of the bracing action of the panels was transmitted to the frame through the
connectors, stress concentration wasless serious than in the model without
connectors.
Frame stresses
32. Frame stresses of the theoretical models at different stages of loading are
shown in Figs 10 and 11.
33. Before crushing of the corners of the infilled panels, frame moment in both
models was negligible. Afterwards, due to the propagation of the contact area
away from the corners, the frame moment increased significantly until they finally
yielded. Frame moment in the model without connectors was much greater than in
that with connectors. It is likely that the strengthof infilled frames without connec-
tors was verymuch dependent on the bending strength of the frame.
34. Though in general frame moment was very much reduced (compared with
450
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
ANALYSIS OF MULTISTOREY INFILLED FRAMES

-1 2
K;;
-30 -1 54
r15 -4

pq
Experimental Experimental Experimental

57
-48

Analytical Analytical
I Analytical
I
Fig. 9. Principal stresses of third storey panel (in lO- unit load/mm2; positive values
are compressive and negative values are tensile): (a) without connectors at 4.8 mm
deflexion(no cracks formed);(b) with connectors at 1.8 mm deflexion (before
cracking); (c) with connectors at 4.8 mm deflexion (after cracking)

6 6-

1 0

~19.2 At 4.8 mm topdeflexion(beforecrushing of panel corners)

At 14.4mm top deflexion (after crushing of panel corners)

Fig. 10. Bending moments ofthird storey frame (inunitload;mmshownon


compression side): (a) without connectors; (b) withconnectors

451
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
.
LIAUW AND KWAN

At 4.8 mm top deflexion (before crushing


of panel corners)

At 14.4 mm top deflexion (after crushing of panel corners)


(4 ( b)
Fig. 11. Shear forces of third storey frame (in fraction of loading applied): (a) without
connectors; ( b ) with connectors

that of bare frames) by the bracing action of the infilled panels, frame shear might,
however, be increased rather than reduced. When connectors were not provided,
very high shear concentration occurred near the joints. These high local shear
forces were detrimental to the structure and could only be dealt with by strength-
ening the joints or by providing connectors. When connectors were provided, the
shear concentration near the joints was very much reduced because part of the
bracing action was transmitted to the frame through connectors.

Distribution of lateral shear


35. In order tostudy the distribution of lateral shear in the models, the fraction
of lateral shear taken by the columns is plotted against deflexion in Fig. 12. The
contribution of friction and connectors at theinfill-beam interface calculated from
those of the columns are also plotted. In the experimental models, shear of the
columns was measured by strain rosettes near the joints. Due to the diffrculties of
high stress gradient there, the measured shear force?, which were the averaged
values within the sizeof the strain gauges,were consistently lower than the
theoretical values. Hence only analytical results are considered in the following
discussion.
36. When connectors were not provided, more than 80% of the lateral shear
was taken up by column 2 on the far side of the load. Column 1, however, took
only negligible shear. The contribution of the friction at the infill-beam' interface
increased from about 10% to about 25% as the corners of the infilled panels were
crushed.
37. When connectors were provided, due to the development of cracks, the
452
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
ANALYSIS OF MULTISTOREY INFILLED FRAMES
100'
_ _ - - - -- - - ---
'. - _ - - - - - - -
_-X\

'\
C 80- Column 2 ._c
0
c
3
n
-2
.-
L

60-

40
a, Frictlon at interface

0 ---
-1 0 - _ _ - L ------
(a)

B 401

0
15
-- 5
- - - _10_ _ - _ _ _ --- .J

-l 0 Deflexion: mm
(b)
Fig. 12. Distribution of lateral shear in the models: (a) without connectors; (b) with
connectors (solid line = experimental; dotted line= analytical)

fraction of the lateral shear taken by column 1 decreased to negligible values while
that taken by column 2 increased significantly and then remained at about 43%
until the cornersof the infilled panels were crushed. Afterthat, the contributionof
the columns decreased with the connectors taking a largerfraction of shear. The
contribution of the connectors, from about 50% to 65%, was much greater than
the friction.

Conclusion
38. A non-linear finite element method has been presented for the analysis of
multistorey infilled frames with or without connectors. The non-linear analysis is
453
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.
LlAUW AND KWAN
simplified bytaking advantage of the uniaxial feature of the panel stresses. In both
models, with or without connectors, the analytical prediction shows good
agreement with the experimental results throughout thewhole range of deflexion.
39. In the model without connectors, non-linearity arises mainly fromthe
gradual softening of the compressive corners of the panels and the consequent
propagation of the area of contact and the yielding of the frame. Their effects on
stiffness degradation, which would cause subsequent stress redistribution of the
structure, are significant. In the model with connectors, non-linearity arises mainly
from cracking, softening of the infill material and yielding of the infill-beam con-
nection. The importance of taking account of these non-linear effects is shown in
the variationof stiffness and stress distribution in both the panels and the frame.
40. Thestructural behaviour of infilled frames isvery much improved by
integrating the infilled panels with the frame using connectors, of which the
contribution in resisting lateral shear is much greater and more reliable than
friction. With the connectors provided, the stress concentrations in compressive
corners of panels and the magnitudes of frame shear and moment at the joints are
considerably reduced.

References
1. MALLICKD. V. and SEVERN R. T. The behaviour of infilled frames under static loading.
Proc. Instn Ciu. Engrs, 1967,38, Dec., 639-656.
2. MALLICKS. K.and BARUAH. K. Behaviour of mortar infilled steel frame under lateral
load. Bldg & enuiron., 1977.12.263-372,
3. RIDDINGTON J. and STAFFORD SMITHB. Analysis of infilled frames subject to raking with
design recommendations. S t r u t . Engr, 1977,52, No. 6,263-268.
4. KINGG . J. W. and PANDEY P. C. The analysis of infilled frames using finite element. Proc.
Instn Ciu. Engrs, Part 2, 1978.65, Dec., 749-760.
5. MALLICKD. V. and GARGR. P. Effect of openings on the lateral stiffness of infilled
frames. Proc. Instn Ciu. Engrs, 1971,49, June, 193-210.
6. LIAUWT. C. and LEES. W. On the behaviour and the analysis of multi-storey infilled
frames subject to lateral loading. Proc. Instn Ciu. Engrs, Part 2, 1977, 63, Sept.,
641-656.
7. MALLICKD. V. and SEVERN R. T. Discussion on The behaviour of infilled frames under
static loading. Proc. Instn Ciu. Engrs, 1968,41, Sept., 205222.

454
Downloaded by [ Universidad Industrial De Santander] on [21/10/17]. Copyright ICE Publishing, all rights reserved.

You might also like