Professional Documents
Culture Documents
Design Models of Light Frame Wood Buildings Under Lateral Loads
Design Models of Light Frame Wood Buildings Under Lateral Loads
Design Models of Light Frame Wood Buildings Under Lateral Loads
Abstract: In this paper, different methods of lateral force distribution and design are described and compared with the results of
Downloaded from ascelibrary.org by Politehnica University of Timisoara on 05/22/17. Copyright ASCE. For personal use only; all rights reserved.
experiments on a full-scale woodframe test house, and with a detailed three-dimensional analysis using a Finite Element 共FE兲 model. The
primary motive for this analysis and investigation is to provide the basic understanding required for the development of improved design
procedure共s兲 for light-frame wood buildings subjected to lateral loads. Improved procedures for lateral load distribution should be
analytically correct in their philosophy, and offer reasonable trade-offs between the uncertainties and omissions in the design process and
simplicity for the end-user. In this paper, eight methods are used to predict the distribution of design wind forces to the walls of an
example L-shaped single-story woodframe house, in a moderate hurricane prone environment. The results are compared and the differ-
ences in the predictions between the various methods are highlighted. Four of these methods are further compared against the results of
physical experiments conducted on a full-scale test house. The FE model was generally the most accurate in predicting the experimentally
measured load distribution, followed by the ‘‘plate method’’ and the ‘‘rigid beam method’’ described herein.
DOI: 10.1061/共ASCE兲0733-9445共2004兲130:8共1263兲
CE Database subject headings: Wooden structures; Lateral loads; Load distribution; Framed structures; Full scale tests; Models.
Overview and Description of Lateral Force Because of the growing interest in Europe to use light-frame
Distribution Methods wood construction in multistory residential buildings, there has
also been a strong interest there recently to evaluate and improve
A light-frame wood building is an assemblage of several compo- the design methods for light-frame wood construction 共Andreas-
nents and diaphragms with repetitive members such as walls, son 2000兲. A general approach to design against wind and earth-
floors, and roof systems connected by intercomponent connec- quake loads—citing recent research results and trends, and iden-
tions such as bolts, metal straps, or proprietary connectors form- tifying key issues that should be considered in analysis and
ing a three-dimensional highly indeterminate structural system. design—can also be found in Foliente 共1998兲, Foliente and Kasal
Because little is known about how the load is shared and distrib- 共2000兲, and Cobeen 共1997兲. Based on the results of the CUREE-
uted in such a complex structural system, simplifying assump- Caltech woodframe project 共Hall 2000兲, numerous improvements
tions are made in analysis and design. This simplification can in earthquake design of woodframe buildings are expected in the
result in either over- or under-designed elements, resulting in ei- near future.
ther over-conservative 共and therefore uneconomical兲 or poten- In the following sections, a critical commentary of common
tially non-conservative structures. Shear wall forces in a light- design assumptions, and a description of some currently used and
frame wood building, for example, may be over-predicted by some proposed methods for determining the distribution of lateral
130% or under-predicted by 60% using the tributary area method, loads to the shear walls in woodframe buildings are given.
in which the building is idealized as a flexible beam on rigid
supports 共Kasal and Leichti 1992兲.
Common Design Assumptions
There is an increasing interest in a systems approach to the
analysis and design of light-frame buildings. The forces induced Current analytical and design methods used for lateral load dis-
by high winds and earthquakes act on the whole building and are tribution are based on a number of assumptions. In the following,
then distributed to the LFRS within the building. The design of we provide comments on some of the commonly used assump-
individual subsystems and components are therefore highly de- tions:
pendent on the accuracy and reliability of methods used to dis- • Diaphragms are flexible or rigid—As a horizontal load-
tribute global building loads to the various components of the resisting member subjected to lateral loads, any diaphragm
building. This step in the analysis and design process is so critical will have a finite deflection. However, its flexibility is usually
that if an inappropriate method of distributing these loads is cho- defined relative to the rest of the structure. That is, the magni-
sen, the ensuing effort spent for a detailed engineering design of tude of expected diaphragm deflection relative to shear wall
the components may yield completely inaccurate results. Thus, it deformation is frequently used to characterize a diaphragm as
is possible to have an ‘‘engineered’’ house that is potentially less rigid or flexible. The International Building Code 共IBC 2000兲
safe than the one that has not been engineered; or in other words, uses the relative deflection of shear walls and diaphragms to
the fact that a house has been engineered can create a false sense determine the diaphragm type. The rule is that a diaphragm is
of confidence of safety and performance. rigid when the lateral deformation of the diaphragm is less
Fig. 2. Description of different load distribution methods 关共a兲 to 共h兲 in plan view兴
than two times the average story drift. Thompson 共2000兲 dem- load-resisting elements in the house. However, by neglecting
onstrated that for most woodframe buildings, the diaphragms the stiffness and contribution of interior walls, the exterior
would be considered as rigid due to significantly larger shear walls may, in some cases, be over-designed.
wall deformations as compared with diaphragm deflection. • Shear wall stiffness is approximately equal—This assumption
The relatively rigid nature of diaphragm behavior in typical is almost never met. Assuming equal shear wall stiffness for
light-frame wood buildings has been observed and docu- all walls can lead to both over-designed and under-designed
mented both experimentally and analytically 共Phillips et al. walls in the same house.
1993; Leichti and Kasal 1993; Paevere et al. 2003兲. • Shear walls behave linearly—This assumption can be used
• Interior walls have unknown or negligible capacity—The lat- where small wall deformations are expected. If walls are
eral capacity of interior walls, which are mostly sheathed with loaded beyond the ‘‘elastic’’ limit, this assumption will yield
nonstructural panels, is commonly ignored in design, although
both over- and under-designed walls.
these walls can be designed as load-bearing walls. Research
• The building does not rotate when loaded laterally—This as-
studies have shown that gypsum sheathed walls can carry sig-
sumption, used in simple beam-type models of the building’s
nificant in-plane loads 共Gad 1997; Karacabeyli and Prion
1999兲. Therefore, if the interior walls are effectively or mar- LRFS, as shown in Figs. 2共c and e兲, neglects different wall
ginally connected to a diaphragm, their influence should be stiffnesses and does not account for the offset between center
included in building performance analysis. of stiffness and applied load. In a nonlinear system, the center
• External walls carry most or all of the loads—Exterior walls of stiffness will change its location.
are designed as load-bearing walls and are frequently sheathed These assumptions are exploited to varying degrees by different
with structural boards. These are considered the main lateral design and analytical methods for lateral force distribution. Some
as a series of flexible beams on rigid supports, with varying line story drift, this assumption holds true for most light-frame wood
loads to account for changes in house geometry. This method buildings. This method has been discussed in detail by Kasal and
does not consider the effect of stiffness on load distribution and is Leichti 共1992兲. Linear springs may be used to represent the shear
based on assumptions, which are rarely met in practice, as noted walls when the loads and corresponding responses are small. The
above. It has been shown that this approach can lead to inaccurate method is not suitable where strong nonlinear response of the
results for certain plan configurations 共Kasal and Leichti 1992兲
system is expected, such as for earthquake loading, where the
and can lead to both conservative and nonconservative results.
structure can be loaded well beyond its elastic limit.
stiffness removed兲. The plate properties are calculated using Plate (E⫽11,300 MPa) 1.22 4.81 17.05 6.62 29.7
the parallel axis theorem to incorporate the offset of sheath- Plate (E⫽1130 MPa) 1.57 6.05 15.65 6.44 29.71
ing with respect to neutral axis; Finite element model 1.46 6.15 14.61 7.64 29.86
3. Actual masses of individual components are included as
mass densities and are calculated based on actual geometry
of these components 共additional masses such as the chimney the behavior of the house. Disturbingly, there is a large difference
can be included兲; in the calculated values for two of the most common approaches,
4. The roof, which can include mass, is modeled as a folded i.e., relative stiffness and tributary area methods. This disparity
plate 共i.e., no individual roof trusses or joists are modeled兲. supports earlier findings by Kasal and Leichti 共1992兲, and raises
More complex roof models 共linear or nonlinear兲 can be in- some serious questions about the design methods used in light-
cluded; frame construction. Which of the values are conservative and
5. Intercomponent connections can be modeled as nonlinear which are nonconservative? Which method gives the most accu-
springs, with either hysteretic or standard nonconservative rate distribution?
nondegrading behavior; Apart from the tributary area and total shear methods, all of
6. The model can be loaded dynamically; and the lateral force distribution methods examined herein require an
7. The in-plane shear stiffness of each of the shear walls is estimate of the stiffness characteristics of the walls 共i.e. elastic
modeled using a nonlinear spring. This element is energeti- stiffness or full load-displacement relationship兲. Estimation of the
cally equivalent to a detailed FE model of a shear wall. The wall stiffness is crucial, since the methods of force distribution
detailed shear wall model includes nonlinear hysteretic ele- are very sensitive to the values used. Various simplified methods
ments for nails and shell elements for studs and sheathing. are available for estimating wall stiffness characteristics
The model can include material plasticity for the shell ele- 共NAHBRC 2000兲, including the perforated shear wall method
ments. Alternatively, the properties for the energetically and the segmented shear wall method. However, the load dis-
equivalent spring can be obtained from shear wall tests. placement relationships calculated using these methods resulted
Due to its complexity, this model is most appropriate as a re- in estimates that were an order of magnitude different, when com-
search tool, and can be used to study the behavior of light-frame pared with experimentally derived values 共Paevere et al. 2003兲.
buildings subjected to various static and dynamic loads, to con- Of note, the walls in the full-scale test house do not satisfy the
duct parametric studies as well as to generate inputs for other anchorage requirements for the perforated shear wall or seg-
simpler models. The FE mesh used to represent the example light- mented shear wall method as outlined in the Residential Struc-
frame house is shown in Fig. 2共i兲. tural Design Guide. Additionally in some cases, the walls were
also outside the recommended aspect ratio limits. Since the meth-
ods listed in the Residential Structural Design Guide proved to be
Numerical Comparison of Lateral Force Distribution unsatisfactory in determining realistic stiffness values, another ap-
Methods proach was needed since realistic estimates of the stiffness of the
shear walls are necessary to experimentally validate the various
In order to compare numerically the differences between the pro- methods of lateral force distribution.
posed and currently used methods described above, each method The approach used to estimate the stiffness was based on FE
was used to calculate the distribution of lateral load to the North- modeling and experimental testing. FE models of all the walls
South oriented walls of an example L-shaped woodframe house, were developed, based on those described in Kasal et al. 共1994兲.
under a design wind load of approximately 30 kN 共6.75 kips兲. The FE simulations generated monotonic force displacement
Overall details of the house used in the examples are given in Fig. curves to ultimate load based solely on the material properties and
1, and a full description of the house is given in Paevere et al. the connection properties of the individual joints, as well as
共2003兲. The wind load provisions in the Residential Structural boundary conditions for the walls to simulate either a free stand-
Design Guide 共NAHBRC 2000兲 provided the basis for the design ing wall or a wall in a house. Unfortunately, the boundary condi-
wind load. The wind load was calculated assuming a 120-mph tions of the individual walls within the house are unknown. The
共192-kph兲 zone with an exposure ‘‘B’’—which is equivalent to a boundary conditions on the perimeter of the walls are a function
suburban exposure condition based on ASCE 7-98 共ASCE 1999兲. of the intercomponent connections. These are not well docu-
Table 1 lists the results of the different lateral force distribu- mented and difficult to establish. The work of Groom 共1992兲 is
tion methods for the four walls aligned with the North-South axis one of the very few studies available today that investigate the
in Fig. 1共b兲. The variation among the calculated reaction forces performance of intercomponent connections in a light-frame
using the different methods is quite large. This variability is un- building. These connections are nonuniform and load history de-
derstandable since the methods examined cover a range of com- pendent. The nonuniformity means that although the general na-
plexity and are based on different underlying assumptions about ture of the connections 共linear or nonlinear, type of hysteresis,
Overall details of the house are given in Fig. 1, and a detailed Importantly, the testing procedure is used purely to compare
description of the house and the experimental setup are given in and validate the load distribution methods examined, and not to
Paevere et al. 共2003兲 and Paevere 共2002兲. The four design proce- evaluate the response to the design wind load. The four imposed
dures used in the experimental validation were displacement fields were not intended to represent the real build-
1. Rigid beam model; ing displacement under a wind load, but rather to cover a range of
2. Plate model with ‘‘rigid’’ diaphragm, E⫽11,300 MPa implicitly assumed displaced shapes for the different lateral force
共1640 ksi兲; distribution methods.
3. Plate model with ‘‘flexible’’ diaphragm, E⫽1,130 MPa In Table 5, the measured load distributions, for the four differ-
共164 ksi兲; and ent displacement cases, are compared against calculated values
4. Finite element model using four different load distribution methods. Table 5, starting
In the full-scale experiments, four different displaced shapes were from the far-left column includes the experimentally measured
applied to the house, and the reaction forces under the four North- reaction forces in each wall, the measured load in each wall as a
South oriented walls were measured. The displacement fields ap-
percentage of the total load for all four walls, the measured load
plied to the house corresponded to the four different fundamental
in each wall as a percentage of the measured capacity of each
assumptions concerning the deflection of the house in the North-
wall 共from Table 2兲, the calculated reaction forces from each of
South direction, under a design wind load. The displacements
the four methods examined, and the percentage error for each
were determined after calculating the design wind load for each
method. The error is calculated as error⫽(predicted value
wall for the respective cases and then calculating the subsequent
⫺measured value)/(measured value)⫻100%. The percentage of
displacement. A plan view of these displacement fields is shown
graphically in Fig. 4. The four different displaced shapes, were total load in the second column of Table 5 does not include the
applied incrementally to the house at the top plate level of the reaction forces in the cross walls 共i.e., the East-West oriented
four North-South oriented walls, until the total applied load was walls兲, which are perpendicular to the direction of loading. The
equal to the design wind load 共30 kN or 6.75 kips兲, or until the measured capacities of the walls were taken from later testing that
displacement in any wall was greater than 1.9 mm 共0.075 in.兲, examined ultimate and post-ultimate behavior 共Paevere et al.
whichever occurred first. Full details of the load and displacement 2003兲.
measurement systems used are given in Paevere 共2002兲. The dis- The FE model provided the most consistent and accurate esti-
placement limit of 1.9 mm 共0.075 in.兲 was set to minimize struc- mates of the load distribution, for all walls under each loading
tural damage, in order to preserve the house for further experi- condition examined. The plate method and the rigid beam method
mentation 共Paevere et al. 2003兲. The values of the applied were less accurate than the FE model, but provided reasonable
displacements and the measured reaction forces under walls W1 results considering their simplicity. All of the methods examined,
to W4 are summarized in Table 4. The sum of the reaction forces can either under- or over-estimate the experimentally measured
is provided in the right-hand column. Note that for applied dis- load in individual walls to some extent. All four methods exam-
placement cases three and four, the design wind load could not be ined predicted experimental values to within ⫾20%, apart from
achieved in the experiment without exceeding the prescribed dis- wall W4 under applied displacement case three. The load levels in
placement limit of 1.9 mm. wall W4 under this displacement profile were very small. The
Table 4. Experimentally Applied Displacements and Measured Load Distribution in Test House
Applied displacement 共mm兲 Measured reaction 共kN兲
Applied displacement case W1 W2 W3 W4 W1 W2 W3 W4 Sum
1-关Fig. 4共a兲兴 1.2 1.2 1.1 1.1 0.8 4.3 14.0 6.6 25.7
2-关Fig. 4共b兲兴 1.9 1.7 1.3 1.2 1.2 5.8 14.3 6.8 28.1
3-关Fig. 4共c兲兴 1.9 1.4 0.8 0.1 1.3 5.2 10.1 0.4 17.0
4-关Fig. 4共d兲兴 1.3 1.7 0.7 0.5 0.8 5.2 9.2 3.8 19.0
resulting errors were large under this case, however when com- hurricane prone environment. The results were compared and the
pared to the capacity of the wall the percent error is negligible. differences in the predictions from the different methods were
highlighted. Compared to a detailed nonlinear FE model of the
Comments on Experiment house under examination, the plate model was the most accurate
in predicting the load distribution, followed by the relative stiff-
Notably, the internal walls such as W3, the stiffest wall in the ness model followed by the three different beam methods. The
house at low levels of displacement, are often not structurally tributary area method predicted with the least accuracy, although
connected to the roof and ceiling diaphragm, but are connected to it is one of the most commonly used approaches by engineers
the rest of the structure via the end walls and nonstructural con- today. For the example house under consideration, the methods
nections. This effect can only be modeled properly using the FE which are based on a more rigid diaphragm, but also allow con-
model. The experiments and analysis presented here were limited trol of the degree of flexibility 共i.e., the two plate methods兲 gave
in magnitude to 1.9 mm displacement. At such small displace-
the best results. This finding is consistent with experimental re-
ment levels, the effect of tolerances and geometrical imperfec-
sults in Paevere et al. 共2003兲 which indicate the roof and ceiling
tions may be exaggerated even though the induced displacements
diaphragm for the house had some degree of flexibility, but is
were large enough to achieve a 30 kN 共6.75 kips兲 total reaction
effectively rigid compared to the walls.
共design wind level兲 in some of the displacement cases. Given
Four of the load distribution methods were further investigated
such low displacement levels the roof and ceiling diaphragms
may not have been fully engaged. with the results of physical experiments conducted on a full-scale
test house. The FE model was generally the most accurate in
predicting the experimentally measured load distribution, justify-
Conclusions ing its use as the reference for the comparisons between different
design models. This is followed by the plate method and then the
In this paper, different methods of lateral load distribution have rigid beam method. All four methods examined generally pre-
been described and compared with the results of experiments on a dicted experimental values with reasonable accuracy 共within
single-story full-scale woodframe test house, and with a detailed ⫾20%).
three-dimensional analysis using a FE model. For the type of load and house configuration studied in this
Eight different methods were used to predict the distribution of work, the plate method and the rigid beam method are the most
design wind forces to the walls of the house, for a moderate practical, since they both predict with acceptable accuracy, with-
Agriculture. The writers would like to acknowledge the technical wall design of light-frame structures.’’ J. Struct. Eng., 118共12兲, 3350–
advice and assistance provided by Jay Crandell from the NAHB 3361.
Research Center. Kasal, B., Leichti, R. J., and Itani, R. Y. 共1994兲. ‘‘Nonlinear finite-element
model of complete light-frame wood structure.’’ J. Struct. Eng.,
120共1兲, 100–119.
References Leichti, R. J., and Kasal, B. 共1993兲. ‘‘A look at the structure as a three-
dimensional system.’’ FPS Proc. 7312, Forest Products Society, Madi-
Andreasson, S. 共2000兲. ‘‘Three-dimensional interaction in stabilisation of son, Wis.
multi-story timber frame building systems.’’ Rep. TVBK-1017, Divi- NAHBRC. 共2000兲. Residential structural design guide: 2000 Edition,
sion of Structural Engineering, Lund Univ., Lund, Sweden. U.S. Department of Housing and Urban Development, Rockville, Md.
ASCE. 共1999兲. Minimum design loads for buildings and other structures. Paevere, P. J. 共2002兲. ‘‘Full-scale testing, modelling and analysis of light-
ASCE 7-98, Reston, Va. frame structures under lateral loading.’’ PhD thesis, Dept. of Civil and
Cobeen, K. 共1997兲. ‘‘Seismic design of low-rise light-frame wood Environmental Engineering, The Univ. of Melbourne, Parkville, Vic-
buildings—State-of-the-practice and future directions.’’ Earthquake toria, Australia.
performance and safety of timber structures, G. C. Foliente, ed., For- Paevere, P. J., Foliente, G., and Kasal, B. 共2003兲. ‘‘Load-sharing and
est Products Society, Madison, Wis., 24 –35. redistribution in a one-story woodframe building.’’ J. Struct. Eng.,
Foliente, G. C. 共1998兲. ‘‘Design of timber structures subjected to extreme 129共9兲, 1275–1284.
loads.’’ Prog. Struct. Eng. Mater., 1共3兲, 236 –244. Phillips, T. L., Itani, R. Y., and McLean, D. I. 共1993兲. ‘‘Lateral load
Foliente, G. C., and Kasal, B., eds. 共2000兲. ‘‘Wind safety and perfor- sharing by diaphragms in wood-framed buildings.’’ J. Struct. Eng.,
mance of wood buildings.’’ Forest Products Society Publication No. 119共5兲, 1556 –1571.
7258, Forest Products Society, Madison, Wis. Thompson, D. S. 共2000兲. ‘‘Wood diaphragm and shear wall deflections.’’
Gad, E. F. 共1997兲. ‘‘Performance of brick-veneer steel-framed domestic Proc., ASCE Structures Congress 2000, Advanced Technology in
structures under earthquake loading.’’ PhD thesis, Dept. of Civil and Structural Engineering, CD ROM, M. Elgaaly, ed., ASCE, Reston,
Environmental Engineering, Univ. of Melbourne, Australia. Va.