You are on page 1of 8

A 3D Model for Collapse Analysis of Soft-story Light-frame Wood

Buildings

Weichiang Pang1, Ershad Ziaei2, and Andre Filiatrault3

ABSTRACT: This paper presents the development of a new three-dimensional (3D) numerical model,
developed as part of the NEES-Soft project, which is capable of modelling the collapse mechanism of 3D
light-frame wood buildings under earthquake motions. In the new 3D model, the in-plane and out-of-plane
flexibilities of floor and roof diaphragms are characterized using two-node 12 degrees-of-freedom (DOF)
frame elements with corotation formulation and the shear walls are modelled using two-node 6-DOF link
elements. To reduce the computational overhead of the model, a novel condensation technique using shape
functions is employed to condense the DOFs of the link elements. This 3D model has been used to perform
incremental dynamic analyses (IDA) for a three-story building with tuck-under parking garage in the first
floor. The IDA results revealed that the building is prone to side-sway collapse in the first-story.

KEYWORDS: Light-frame wood; Corotation; Collapse Risk; Soft-story; Geometric Nonlinearity

1 INTRODUCTION with three degrees-of-freedom (DOFs) in the


horizontal plane [1] or rigid plates with 6 DOFs in
123
The nonlinear dynamic response of light-
three-dimensional (3D) space [2] (Figure 1). In
frame wood buildings during earthquakes is
addition, the framings of shear walls are modelled
extremely difficult to model and analyze. Unlike
as pinned connected rigid elements. Due to these
typical reinforced concrete and steel frame simplified modelling assumptions, which lead to
buildings where the main structural members are inaccurate boundary conditions, these simplified
well defined, light-frame wood buildings consist of
models are not able to accurately predict incipient
many interconnected framing members and
collapse of light-frame wood buildings.
redundant elements that make the load paths and
structural elements not readily identifiable. A recent
study [3] has shown that bearing contacts between
the framing (e.g. stud-to-sill plate and sill-plate-to-
foundation), uplift of hold-downs and shear slip of
anchor bolts have significant influence on the
lateral behaviour of wood shear walls and
buildings.
The current state-of-the-art numerical
models for light-frame wood buildings are
formulated based on small rotation theory and the
diaphragms are either assumed to be rigid plates

1
Weichiang Pang, Assistant Professor, Glenn Department of (a) (b)
Civil Engineering, Clemson University, Clemson, South
Carolina, USA. Email: wpang@clemson.edu Figure 1: a) Pseudo 3D or 2D pancake model, and (b)
2
Ershad, Ziaei, PhD Student, Glenn Department of Civil 3D rigid plate model.
Engineering, Clemson University, Clemson, South Carolina,
USA.
3
Andre Filiatrault, Professor, Department of Civil, Structural
and Environmental Engineering, University at Buffalo,
Buffalo, New York, USA.
This paper presents the formulation of a new
three-dimensional (3D) model developed as part of EA L
the NEES-Soft project [4]. The new model 0 k z1
0 0 k 1y
addresses most of the deficiencies of the 0
aforementioned simplified models and is able to 0 0 GJ sym.
L

0 0 k y4 0 k y5 (1)
capture the collapse mechanism of light-frame
K mf EA
0 k z3 0 0 0 k z5
wood buildings. This 3D model is an extension of L 0 0 0 0 0 EA
L
two detailed 2D shear wall models [6][8] developed 0 k z2 0 0 0 k z4 0 k z1
for collapse analysis of light-frame wood shear 0 0 k y2 0 k y3 0 0 0 k 1y
0 0 0
walls. L 0 0 0 0 0
GJ GJ
L
0 0 k y4 0 k y6 0 0 0 k y4 0 k y5
0 k z3 0 0 k z6 0 k z 0 0 0 k z
3 5
2 MODEL FORMULATION 0
12 EI i 6 EI i
The 3D model is developed using a co- ki1 ki4
L3 (1 i ) L2 (1 i )
rotational formulation and large displacement 12 EI z
12 EI i (4 i ) EI i y
theory. The floor and roof diaphragms are modelled ki2 ki5 GL2 Azs
L3 (1 i ) L(1 i )
using co-rotational 3D, two-node, 12-DOF frame 12 EI y
6 EI i (2 i ) EI i z
elements which accounts for geometric nonlinearity ki3 ki6
L (1 i )
2
L(1 i ) GL2 Ays
(
Figure 2). Using a co-rotational formulation where the subscript i denotes the element x, y, and
allows proper consideration of the in-plane and out- z axes. A is the cross-sectional area. Ais is the shear
of-plane motions of the diaphragms under large area normal to direction i. E is the modulus of
deformations. The lateral stiffness of shear walls elasticity. L is the element length. G is the shear
and axial stiffness of wall studs are modelled using modulus. I and J are the bending and torsional
3D, two-node, 6-DOF link elements. In order to moment of inertias, respectively.
reduce the computation time, shape functions of the slave node
zs
frame elements are utilized to eliminate the DOFs z1 z2
Rx1 Rz1 Rxs Rzs Rz2 y2
of the link elements. The condensed global stiffness x1 xs x
matrix is independent of the number of link y1 Ry1 ys Rys Ry2 R 2
x2

elements. It depends only on the number of frame z


y
elements. This condensation technique has been x shape function
successfully used to model the collapse mechanism
Figure 3: 3D frame element.
of a pair of shear walls tested on a shake table
[5][6]. The corresponding geometric stiffness and
consistent mass matrices are given by Eqns. (2) and
(3), respectively:
0
3D 6-DOF 0 36
Shear Wall Element 0 0 36
0 0 0 0 sym. (2)
0 0 3L 0 4 L2
3D 2-node 12-DOF P 4 L2
K Gf e 0 3L 0 0 0

Beam Element 30 0 0 0 0 0 0 0
0 36 0 0 0 3L 0 36
0 0 36 0 3L 0 0 0 36

0 0 0 0 0 0 0 0 0 0
0 0 3L 0 L2 0 0 0 3L 0 4 L2

Figure 2: 3D Model. 0 3L 0 0 0 L 0 3L 0 0 0 4 L
2 2

2.1 3D Frame Element 140


0 156 (3)
The 3D frame element has two nodes with six 0 0 156 sym.
0 0 0 140 J A
DOFs at each node (three translations in the 0 0 22 L 0 4 L2
me 0
element x, y, z directions and three rotations about Mf
420 70
22 L
0
0
0
0
0
0 4 L2
0 0 140


the element x, y, z axes) which can be used to 0 54 0 0 0 13L
0 13L 0
0 156
0 0 54 0 0 156
model tension, compression, torsion and bending 0 0 0 70 J A 0 0 0 0 0 140 J A
0 0 13L 0 3L2 0 0 0 22 L 0 4 L2
effects (Figure 3). The 3D frame element material 0 13L 0 0 0 3L2 0 22 L 0 0 0 4 L2
stiffness matrix is [10]: where Pe is the axial load (tension is positive) and
me is element mass. It should be noted that the
variation of axial loads is tracked during the
analysis and the geometric stiffness matrix is
updated at each time step to account for the where the {Dc} are the deformations associated
geometric nonlinearity due to large deformations. with the 6 DOFs of the link element and {Df} are
the nodal deformations of the 3D frame element.
2.2 3D Link Element
The link element is a zero-length two nodes 2.4 Global Stiffness and Mass Matrices
element with six uncoupled DOFs. The link The global stiffness is given by the following
element stiffness matrix is: expression:
nf nc
kx
K K mf KGf N1T K c j N1 N 2T K c j N 2
i
(7)
0 ky sym. i j

(4) where nf and nc are the total numbers of frame and


0 0 kz
Kc link elements, respectively. It should be noted that
0 0 0 k Rx the size of the global stiffness matrix is independent
0 0 0 0 k Ry
of the number of link elements. It depends only on
0 0 0 0 0 k Rz the number of frame elements. The interpolation
matrix, N, is used to condense the DOFs of the link
where the first three and last three diagonal terms elements. The N1T Kc N1 and N2T Kc N 2 terms in Eqn.
correspond to the translational and rotational DOFs,
(7) describe the influences of a link element on the
accordingly.
two frame elements that are connected together by
Slave Node the link element (Figure 4). The global mass matrix
Undeformed on Frame 2 is:
frame 1 Deformed
frame 1 nf n

c m m
M M f c N1T N1 c N 2T N 2 (8)
j 2 2
i
i
z x Link Element
Slave Node where mc is the mass of the link element.
on Frame 1
y 2.5 Corotation
Figure 4: Link Element.
The 3D model is developed based on a co-
2.3 Nodal Condensation rotational formulation for geometric nonlinear and
large displacement analysis. In co-rotational
In order to reduce the computation time, formulation, the total deformation of the framing
shape functions of the frame elements are utilized members is decomposed into two components, a
to eliminate the DOFs of the link elements. The rigid body motion and relative deformations with
shape functions are given in the following 612 respect to the rotated coordinate system (Figure 5).
interpolation matrix: At each time step, the global stiffness matrix is
assembled based on the rotated coordinate system
N1 N5 y N5 z 0 N7 z N7 y N 2 N5 y N5 z 0 N9 z N9 y (5) of the individual frame and link elements.
0 N3 0 N1 z 0 N6 0 N4 0 N2 z 0 N8

0 0 N3 N1 y N 6 0 0 0 N 4 N 2 y N8 0
N
0 0 0 N1 0 0 0 0 0 N2 0 0 y y y
0 0 N5 0 N7 0 0 0 N5 0 N9 0
0 N5 0 0 0 N7 0 N5 0 0 0 N9

1 r (r 1)2 (r 2) (r 1)(3r 1)
N1 N4 N7
2 4 4
1 r 3(r 2 1) L(r 1)(r 1) 2 x
N2 N5 N8 x
2 2L 8 z z

(r 1) (r 2)
2
L(r 1)2 (r 1) (r 1)(3r 1) y y
N3 N6 N9 Wall studs:
4 8 4 1) CoRot = 1 .

2) kx = bilinear (to carry gravity loads)


where r = 2x/L + 1. The interpolation matrix is used 3) All other stiffness components are zero
to determine the deformations of any arbitrary point Shear stiffness:
1) CoRot = 0
(slave node) within the 3D frame element (see 2) kx = MSTEW (shear stiffness)
Figure 3). 3) All other stiffness components are zero
Dc 61 N 612 D f (6)
121
Figure 5: Corotational model.
3 THREE-STORY BUILDING
As part of the NEES-soft research project, a
full-scale three-story woodframe building with
tuck-under parking garage in the first story will be
constructed and will be tested using hybrid
simulation technique at the University at Buffalo
NEES site. This type of buildings, known as soft-
story buildings, are susceptible to collapse in the
first story due to the open space condition in the
first story.
Figure 6 shows the floor plans of the proposed
test building. This test building will be constructed
according to the pre-1970s construction. The
framing members are assumed to be 2x6
dimensional lumbers. Many of these buildings are
zero-lot line buildings. It is assumed that the
exterior of the zero-lot wall line is covered with
horizontal wood sheathing only while other wall
lines are sheathed with plaster on wood lath. The (a)
interior of the exterior walls in the first story is
assumed to be unsheathed while the interior in the
upper stories is assumed to be sheathed with
gypsum wall boards (GWB). It should be noted that
GWB is not a typical original pre-1970s
construction material. However, many of these
buildings have been renovated in recent years and it
is not uncommon to have the interior sheathing
replaced by GWBs.
A preliminary 3D model has been constructed
to evaluate the seismic performance of this three-
story building. The effective seismic weights at the
first floor, second floor and roof diaphragm are
estimated to be 100, 116, and 65 kN, respectively.
The 3D model presented in this paper is part of the
on-going numerical studies of the NEES-soft
project. It has been used to conduct a preliminary
analysis to simulate and predict the mechanism that
induces the side-sway collapse of the three-story (b)
building.

(c)

Figure 6: NEES-soft hybrid test building; (a) first- and (b)


upper stories plan views (c) 3D view.
FPL 1956 -Plaster on wood lath
4 WALL HYSTERESIS MODELS 2

As discussed in the previous section, the 1.5 Test Data - wall11


Model/Fit
sheathing materials of the walls in the three-story 1
building are horizontal wood sheathing, plaster on

Top of Wall Force, kN


wood lath and gypsum wall board. Table 1 shows 0.5

the hysteretic parameters for the CUREE hysteretic


0
model (also known as the modified Stewart
hysteretic model) [1] determined for each of these -0.5
sheathing materials per unit wall width (i.e. per
meter of full-height wall width) and the -1

corresponding hysteresis loops are shown in -1.5


Figures 7 to 9. The parameters for walls with
horizontal wood sheathing and walls with plaster on -2
-30 -20 -10 0 10 20 30
wood lath are determined using the wall test data Top of Wall Displacement, mm
conducted in the 1950s in the Forest Products Figure 8: Hysteresis of walls sheathed with plaster on
Laboratory [9]. The gypsum wall board parameters wood lath.
are obtained from the wall tests conducted in one of
the CUREE project subtasks [11]. These wall CUREE1.4.4 -Gypsum Wall Board
2
parameters are used to construct the 3D model
Input Data
shown in Figure 11 to represent the contribution of 1.5
Model/Fit
each of these sheathing materials on the overall
1
seismic resistance of the building. Direct
Top of Wall Force, kN

summation or superposition of the lateral strength 0.5


contributions of the different layers of wall
sheathing materials is used in the model. 0

-0.5
Table 1 Wall hysteresis parameters (per meter of wall).
K0 r1 r2 r3 r4 F0 F1 -1
kN
/mm kN kN mm
-1.5
Horizontal
Wood 0.279 0.06 -0.20 1.02 0.007 1.61 0.20 83.8 0.85 1.07
Sheathing -2
-150 -100 -50 0 50 100 150
Plaster on
1.233 0.02 -0.33 1.02 0.007 5.77 0.69 17.0 0.80 1.05 Top of Wall Displacement, mm
Wood Lath
Gypsum
Wall Board
0.512 0.22 -0.02 1.04 0.006 0.66 0.10 10.4 0.85 1.35 Figure 9: Hysteresis of walls sheathed with gypsum wall
board.

.
FPL 1956 - Horizontal wood sheathing 5 EQUIVALENT VISCOUS DAMPING
2
Test Data - wall2A Most of the damping will be accounted for in
1.5 Test Data - wall20 the CUREE shear wall hysteresis model. However,
Test Data - wall33 the CUREE hysteresis model has been shown to
1
overestimate the energy dissipation at small
Top of Wall Force, kN

Model/Fit
0.5 deformations. To ensure that the model does not
over-estimate energy dissipation particularly at
0 small deformations, a small amount of equivalent
-0.5 viscous damping was included in the numerical
model (2% Rayleigh damping assigned to modes 1
-1 and 2).
-1.5
6 MODE SHAPE
-2
-150 -100 -50 0 50 100 150 Modal analysis was performed to estimate the
Top of Wall Displacement, mm fundamental periods of the building. The periods
Figure 7: Hysteresis of walls sheathed with horizontal for the first three modes were estimated to be 0.33s,
wood sheathing. 0.31s and 0.23s. Figure 10 shows the deformed
shape of the building for the first mode. As can be
seen, the building exhibits a primarily soft-story the median spectral acceleration at the fundamental
drift in the longitudinal direction. period of the building, Sa(T1). The ground motions
350 Mode 1 Period = 0.329s were scaled in accordance to the procedure outlined
in the FEMA P-695 report (formerly known as the
300 ATC-63 project) [12]. The 22 bi-axial ground
motions were scaled using a single scaling factor,
250 computed as the ratio of the scaled median spectral
acceleration to the median spectral acceleration of
200
the normalized ground motion ensemble [12].
The building is deemed to have collapsed under
z

150
a particular ground motion intensity when the
100
tangent-to-initial slope ratio of the IDA curve is
less than a prescribed threshold. In this study,
50
different threshold values (0.2, 0.4, 0.6 and 0.8)
were used to define the collapse points. The 20%
0 tangent slope (i.e. tangent-to-initial slope = 0.2) is
-500 typically used to define the global collapse capacity
0 point [14].
500
Figures 14 and 15 show the collapse fragility
50 100 150 200 250 300
x
-150 -100 -50 0
y
curves using the median spectral acceleration at the
Figure 10 First mode shape, T = 0.33 s. fundamental period as the intensity measure (IM)
and using the resultant drift as the demand measure
7 INCREMENTAL DYNAMIC (DM), respectively. Also shown in these figures are
ANALYSES the lognormal distribution parameters for the
collapse fragility curves. Using the 20% tangent
Incremental dynamic analyses (IDA) [13] were slope approach, the IDA results confirmed that the
performed using the 22 bi-axial ground motions building is susceptible to collapse in the first story
developed in the ATC-63 project [12]. Figure 11 under a design level earthquake event which
shows the deformed shape of the three-story roughly corresponds to a 500-year return period
building at incipient of collapse under the ATC-63 earthquake. The design Sa value for a typical
bi-axial ground motions [ATC-63 Far-Field EQ ID location in Southern California is about 1g for
21, 1971 San Fernando Earthquake]. Figure 12 periods between 0.2s to 0.6s. From Figure 14, it
shows the transverse displacement time history can be clearly seen that, there is slightly more than
responses of two corner points on the roof level 50% chance that the three-story building considered
(see Figure 11). The differences between the in this study would collapse at the design
transverse displacements of these two corners point earthquake level. Figure 15 reveals that the median
indicate that significant torsional effects in the collapse drift capacity of the building is
building. approximately 13% (inter-story drift).
Node 4
Node 1 Roof Drift in Y direction - node#1&4 - EQ17
0.05
400 node#4
0.04 node#1

300 0.03

0.02
Z 200
z

0.01
Y Roof Drift

100
0

0 -0.01
-100 600
0 400 -0.02
100 torsion
200
X 200 -0.03
300 0
400 -200 Y -0.04
y
x
Figure 11: Structure deformed shape at incipient -0.05
collapse. 0 5 10 15 20 25
Time(s)

Figure 12 Torsion in the upper stories.


Figure 13 shows the IDA curves of the three-
story building. The intensity measure of the IDA
curves (i.e. the vertical axis) is defined in terms of
Maximum resultant Inter-Story Drift
3.5
Incremental dynamic analyses were performed
EQs g=0o using an ensemble of 22 bi-axial earthquake
3
x2 y2 EQs g=90o motions on a three-story building. The preliminary
16%
h results of the IDA analyses showed that the
Median Sa at Natural Period (g)

2.5 median
84% building has a high risk of collapse (more than 50%
2 chance) under a design level earthquake.
1.5
k1 ACKNOWLEDGEMENT
1 The material presented in this paper is based upon
work supported by the United States National
0.5
Science Foundation under Grant No. CMMI-
0
1041631 (NEES Research). Any opinions, findings,
0 5 10
Max. Resultant Interstory Drift (%)
15
and conclusions or recommendations expressed in
Figure 13 IDA curves for the maximum resultant inter-
this material are those of the author(s) and do not
story drifts. necessarily reflect the views of the National
Science Foundation.
Collapse Diagram - Max Resultant Inter-Story Drift
1
k/k1<0.2
REFERENCES
0.9
k/k1<0.4 [1] Folz, B., and Filiatrault, A. (2001). A Computer
0.8
k/k1<0.6
Program for Seismic Analysis of Woodframe
0.7
k/k1<0.8
Structures, CUREE Report W-21, Task 1.5.1,
Consortium of Universities for Research in
Collpase Probability

0.6
Earthquake Engineering, Richmond, CA.
0.5

0.4
[2] Pei, S., and van de Lindt, J.W. (2009). Coupled
Shear-Bending Formulation for Seismic Analysis of
0.3
Stacked Shear Walls Systems, Earthquake
0.2 Engineering and Structural Dynamics, 38: 1631-
0.1 1647.
0
0 0.5 1 1.5 2 2.5 3
[3] Christovasilis, I.P., Filiatrault, A., and Wanitkorkul,
Median S a at Natural Period(g)
A. (2009).Seismic Testing of a Full-Scale Two-
Figure 14 Collapse fragility curves in terms of the Story Light-Frame Wood Building: NEESWood
spectral acceleration. Benchmark Test, , MCEER Technical Report
MCEER-09-0005, Multidisciplinary Center for
Collapse Diagram - Max Resultant Inter-Story Drift
1 Earthquake Engineering Research, University at
0.9
Buffalo, New York, USA, 216 p..
0.8 [4] van de Lindt, Symans, M.D., Pang, W., Shao, X.,
0.7
k/k1<0.2 and Gershfeld, M. (2012), Seismic Risk Reduction
k/k1<0.4
for Soft-story Woodframe Building: The NEES-Soft
Collapse Probability

0.6
Project, 121th World Conference on Timber
k/k1<0.6
0.5 Engineering, Auckland, New Zealand, Jul 16-19.
k/k1<0.8
0.4
[5] van de Lindt, Pei, S., Pang, W., and Shirazi, S.M.
0.3 (2012), Collapse Testing and Analysis of a Light-
0.2
frame Wood Garage Wall, ASCE J. of Structural
Engineering, in press.
0.1

0
[6] Pang, W., and Shirazi, S.M. (2010) Next
0 2 4 6 8 10 12
Max. Resultant Interstory Drift (%)
14 16 18 20
Generation Numerical Model for Non-linear In-
plane Analysis of Wood-frame Shear Walls, 11th
Figure 15 Collapse fragility curves in terms of the World Conference on Timber Engineering, Trentino,
maximum resultant inter-story drift.
Italy, June 20-24.
[7] Pang, W., and Shirazi, S.M. (2012) A Co-rotational
8 SUMMARY AND CONCLUSIONS Model for Cyclic Analysis of Light-frame Wood
A new model has been developed to model the Shear Walls and Diaphragms, ASCE J. of
soft-story collapse meachanism of three- Structural Engineering, in press.
dimensional three-story light-frame wood [8] Christovasilis, I.P. and Filiatrault, A. 2010. Two-
buildings with tuck-under parking garage. Dimensional Seismic Analysis of Multi-Story Light-
Frame Wood Buildings, 9th US National & 10th
Canadian Conference on Earthquake Engineering:
Reaching Beyond Borders, Toronto, Canada, Paper
No. 69, 10 p.
[9] Trayer, G.W. (1956), The Rigidity and Strength of
Frame Walls, FPL Report No. 896, Forest Products
Laboratory, Madison, WI.
[10] Cook, R.D., Malkus, D.S., and Plesha, M.E. (1989)
Concepts and Applications of Finite Element, 3rd
Edition, John Wiley & Sons, New York.
[11] Pardoen, G. C., Walman, A., Kazanjy, R. P.,
Freund, E. and Hamilton, C. H., (2003). Testing and
Analysis of One-Story and Two-Story Shear Walls
Under Cyclic Loading, CUREe Publication No. W-
25, CUREE-Caltech Woodframe Project.
[12] ATC (2009). Quantification of Building Seismic
Performance Factors, FEMA P695, Applied
Technology Council, Redwood City, CA.
[13] Vamvatsikos, D. and Cornell, C.A. (2002).
Incremental Dynamic Analysis, Earthquake
Engineering and Structural Dynamics, 31: 491-514.
[14] FEMA (2000). Recommended seismic design
criteria for new steel moment-frame buildings.
Report No. FEMA-350, SAC Joint Venture, Federal
Emergency Management Agency, Washington, DC.

You might also like