Analysis of wood diaphragms and trusses.

Part I: Diaphragms'
RICARDO0. FOSCHI
We.ttertl Fore51 Prot/iii~l.sLo/~orciIor.!.,Cotrtrt/ititl Forc.cl,:\. Scrl,ic,e,Dt~~~tirlttrort
c<f'Fi.\lrrr.ie.strtrr/ I / IG~~ ~ . i r o t ~ t , ~ o r ~ .
Vot1c.orr1.c,t.,B.C., Ctitroclo V6T 1x2
Received January 3, I977
Accepted May 17. 1977
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

An 21nkrlysis for wood diaphtx~msis presented taking into account the deformation of the
connections. The lokrcldeformat~oncha~xcteristicsfor the conncctol-s are ass~~mecl to be non-
1ine;rr. A numerical example is presented and comparisons itre made with experiment;~lresults on
a 20-ft by 60-ft (6-m by 18-m) plywood and decking roofdiaphrkrgrn. The agreement between the
test results and the predictions of the analysis is shown to be good.

L ' a ~ ~ t e uanalyse
r Lln cadre en bois avec diaphragrnes tenant compte de la nun-linearite des
connexions. Un exemple n~lrne~.iq~le est compare h Llne expel-ience e f f e c t ~ ~ sur
e e un catlre de bois
couvelt tle contreplaq~~e et mesurant 20 pi x 60 pi (6 m x 18 m). La correlation entre les r i s ~ ~ l t a t s
theoriques et exp6riment;rux est adeqll;rte.
Can. .I.Civ. Eng., 1,315(1977)

Introduction tions between frame members, and the cover-

Structural components act as diaphragms frame connections. Each of these ekments will
now be discussed.
For personal use only.

when they withstand loads in their own plane

(Fig. 1 ) i s in the case of roof or wall systems The Cover
loaded by the action of the wind or by the The cover may consist or* plywood sheets,
forces generated during earthquakes. The struc- lumber decking, or combinations of both. T h e
tura! analysis of diaphragms is not a simple two-dimensional state of stress is considered to
problem, as the system is highly indeterminate be one of plane stress and a cubic isoparametric
and, as in thc case of wood diaphragms, the finite element, as shown in Fig. 2, is used for
deformation of connectors must be taken into the analysis. T h e displacements u and v in the
account.
x and y directions, respectively, are expressed
Design guidelines for wood diaphragms are in terms of the nodal displaccments u i and vi
mostly based on cxperience and testing, and a according to
com~rehensive literature survev has recentlv
been' published (Carney 1975). Most of thk
diaphragms tested have been rectangular and
without openings, and it is difficult to extrapo-
late from thesc tests conclusions applicable to
other shapes, other loadings, or other construc-
tion configurations.
T h e objective of this paper is to present the where the shape functions Ni are given by
formulation fcr the structural analvsis of wood Zienkiewicz (1971). The material of t h e cover
diaphragms implemented in the computer pro- is assumed to b e elastic and orthotropic, with
gram SADT recently developed at the Western principal axes of elastic symmetry s' and y'
Forest Products L a b o r a t o r y . V o u r different (Fig. 2 ) . The derivation of the stiffness matrix
and basic structural elements are considered in [K],,corresponding to the element is shown by
the analysis: the cover, the frame, the connec- Zienkiewicz ( 197 1) . T h e components of this
matrix may be obtained by numerical intcgra-
'Presented to the meeting of the International tion and, in the cxarnple presented in this
Union of Forest Research Organizations, Wood paper, a 4 x 4 Gaussian quadrature scheme
Engineering Group, Aalborg, Denmark, June 17-18, was used. The higher-order cubic finite ele-
1976.
'Available upon request from the Western Forest
ments allow the representation of the cover
Products Laboratory, 6620 NW Marine Dr., Van- with few such elements, a s will be shown in
couver, B.C., Canada V6T 1x2. a numerical example.
 CAN. J . CIV. E N G . VOL. 4, 1977
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

FIG. 2. C o v e r finite elenlent.

FIG. 1. W o o d diaphragm. L

The Frame
Consider a frame member of length L be-
tween two nodes i and j (Fig. 3 ) . Three de-
grees of freedom are attached to each node:
For personal use only.

displacements u and v in the x and y direc-

tions, respectively; and rotation 6' (positive
when counterclockwise). The deformation of
the member is assumed to be represented by w,
the lateral deflection, and li, the axial displace-
ment. These two functions are taken, in terms
of the local coordinate s, as follows:
I
[2] cll = Fl(s)wi+ F2(s)Oi + F3(s)wj + F4(s)Oj
[31 Ir = Hl(s)iii + H2(s)ij
where x
F,(s) = 1 - 3(s/L)l + ~(s/L)~ FIG.3. Framing member.

approximation to the deformation. The same

applies to the axial displacement ii, here taken
as a linear function of s.
Connections between Frame Members
Figure 4 shows a connector (idealized as a
H2(s) = s/L spring) linking nodes i and j, the latter belong-
The derivation of the stiffness matrix [KIf ing to member I. The load-deformation prop-
corresponding to the member is shown, for erties of the connector are assumed to be
example, by Przemieniecki (1968). Since the known for three different loading conditions:
lateral deflection w is taken as a cubic function ( a ) for a load P,,., that produces a relative
of s, the deformation of the member will be displacement Au between i and j along the mem-
only approximate if the length L has loads ber axis x f ; ( b ) for a load P , , that produces a
directly applied along it. For the case in which relative displacement A v between i and j along
nodal loads as well as member loads are pres- the axis y f ; and ( c ) for a moment M that pro-
ent, the length L may represent a segment of duces a relative rotation A6' between i and j.
the actual member length, giving a better All three load-deformation relationships are
 FOSCHI. I 347

assumed to be nonlinear and to be represented

as
Y'
[51 Px, = (mo,, + m,,,lAul)
x [I - exp (-k,,lAullmoll>l
[61 P,, = (moo + m,,lAul)
x [I - exp (-k,lAulimo,>l
I
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

[71 M = (moo + m1olAeI)

The parameters kt,, k,, and ko correspond to Load

the initial connector stiffness, the parameters
ml to the stiffness for large relative displace-
ments or rotations, and the mo to the intercepts
of the asymptotes with slope m l (Fig. 4).
T h e relative displacements Au and A V and
the relative rotation A0 can be written as

tan-' k
For personal use only.

Deformation
[lo] A0 = QOTticf FIG.4. Load-deformation property, frame connec-
tor.
where the vectors Q,,, Q,, Q,, and ti,, are given
by
cos 4/' -sin

-1

Let U,,, U,, and U, be, respectively, the work

required to deform the connector from 0 to
IAul, 0 to IAvl, and from 0 to IAel. Consider,
for example, the work U,,. It can be determined
from
El21 U,, = So'*"' Px.(z)dz Before introducing the vector SU,, into the
global system, it is convenient to separate a
that linear from a nonlinear part. Thus, 6 U,, can
be written as
[I31 au,,lalAul = P,,(lAul)
The application of the principle of virtual
work for the derivation of the global system of
equations requires the derivatives of (Ut, + U, + where the matrix [KIcfis given by
U,) with respect to each of the components of
the vector ti,,. These derivatives can be as-
sembled into a vector SUcf which, using [8],
[9], and [lo], is given by and the vector W,,, using [13], by
 348 C A N . J. C I V . ENG. VOL.. 4. 1977
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

Connections between Cover and Frame

Members
Figure 5 shows a cover finite element con-
nected to a frame member with end nodes
i and j. Let P and P' be a pair of points with
the same x-y coordinates, but such that P
belongs to the cover and P' to the frame. The
relative displacement between points P and P'
is, therefore, the slip of the connection at the
location of P. Consider Fig. 6. Let ~ i be i the
relative displacement of P with respect to P' FIG.6. Cover-frame connection deformation.
along the member axis x', and let AW be the
relative displacement between the same pair of expressed as
For personal use only.

points along the axis y', normal to the member.

Assume now that P' has moved with respect to
P an amount 141 in the direction making an
anglc a with the s' axis. The force F required with the parameters k, mo, and m l functions
to produce this deformation will, in general, of a. These parameters may be determined in
depend upon the angle a and will be non- terms of those found from two basic tests: one
linearly related to la!.Let this relationship be for a = 0" (force applied along the member
axis) and another for a = 90" (force applied
perpendicular to the member axis). Assuming,
for example, that the initial stiffness measured
for a = 0 is k,, and for a = 90" is k,., then

An analogous relationship can be used for the

parameters mo and ml knowing the correspond-
ing values moil,mop,rnl:,, and m ~ , .
T h e relative displacements Aii and AW can
b e calculated by using [I], [2], a n d [3]. Thus,
from [I], the displacements of P a r e

where the nondimensional coordinates 6 and 7

corrcsponding to point P can be obtained by,
for example, using the x-y coordinates of
points i and C (the centroid of t h e element)
and the local coordinate s along the member
FIG.5. Cover-frame connection. axis x':
 FOSCHI. I 349

t = (xi + s cos # - +)/a Using [ 2 ] and [ 3 ] for the displacements of P',

[21I the relative displacements ~ i =i iip - iipj and
7) = (yi + s sin # - y,)/b A W = wp - wp. can now be determined. These
Therefore, with respect to the xl-y' axes, can be written in the form

ii, = x
12

k= 1
(Nkukcos# + Nkuksin #) [23I
Aii = Q2SCc

[221 AW = QcTSCc
w, = x
12
(- Nkuksin /J + Nkukcos/J)
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

k= 1 where the vectors Q,, Q,, and 6,, are given by

- H I cos y~! F1 sin $

- H I sin # - F, cos #
0 -F2
- H z cos # F, sin #
- H z sin I/J
0
N , cos #
N 1 sin #

*
For personal use only.

#
N12 cos
N , , sin #
- N 1 2 sin
N12 cos *
Consider again Fig. 6 . Assume that for a [27 I s u c c = [Klcc6cc - Wcc
variation in the Si component of the vector ticc, with the matrix [K],.,given by
point P' undergoes a virtual displacement dz
in the direction r77. The corresponding virtual
work dUCcdone by the forces F is given, for
the entire length of contact L, by
Accordingly, from [26], the vector Wcc is

where p is the density of connectors (e.g. num-

ber of nails per unit length). Since, from Fig.
6, dz cos + = AM) and dz sin + = ~ ( A w ) ,
[ 2 5 ] may also be expressed as
L
The integrations required by [ 2 8 ] and [ 2 9 ] are
1261 %; J
aucc -
- 0
Aii a(Aii)
~ ~ ( 1(TG
~ 1 1
performed numcrically, using for example a
Gaussian quadrature scheme along the length
L of the member.
Global System of Equations and Solution
This expression represents the contribution Let 6 be the vector containing all the un-
from the connection to the equation for the knowns, i.e. the displacements and rotations
unknown S i in the global system. The deriva- for the frame joints plus the displacements for
tives [26] form a vector SUcc which, before the cover finite element nodes. The individual
assembly into the global system, is conveni- stiffness matrices [ K ] , , [ K ] , , [K],,, and [ K ] , ,
ently split into a linear and a nonlinear part. are assembled into a global matrix [ K ] ; the
Thus, let vectors Wc, and Wcc are similarly assembled
 C A N . J . CIV. ENG. VOL. 1. 1977

.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

60 f t -
FIG.7. Diaphragm used for testir
into a global vector W. If R is the vector of nails/m) at the chords ( 7 ) , 0.250 nails/in.
loads, and since W is a function of the de- (9.84 nails/m) at the endposts (G), and 0.167
formation vector 6, the global system has the nails/in. (6.57 nails/m) at the inner posts
form ( 8 ) . The boards in the decking were of ran-
dom length, with very limited slant nailing be-
tween rows. The overlaid plywood was #-in.
For personal use only.

This system may be solved by iterations ac- (9.5-mm) C-D interior grade Douglas fir, with
cording to thc face grain oriented at 45' as shown in Fig.
7 ( 1 ). The plywood was nailed to the decking
with 8d common nails, with a density large
with the initial 'elastic' approximation enough to effectively prevent plywood buckling
during deformation.
The frame consisted of nominal 2-in. by
The matrix [K] is kept constant during the 10-in. (50-mm by 254-mm) Douglas-fir lum-
iterative process? making the procedure fall in ber, collstruction grade. The chords ( 3 ) con-
the category of initial stress methods as dis- tained only one board except at spliced sec-
cussed by Zienkicwicz ( 197 1) . The computer tions; the endposts ( 4 ) were made from two
program SADT uses an acceleration technique boards nailed together; and the inner posts
as described by Irons and Tuck (1969) and ( 5 ) consisted of four nailed boards. The end
Jennings ( 1971 ) . The stopping criterion used connections were made with 1Gd common
is given by the following ratio of vcctor norms nails ( 9 ) and four nails were used per post.
Four equal loads Po were applied on the
[331 116i+l - ~ i 1 1 ~ / 1 1 ~ i 1 1E ~ plane of the diaphragm and at the inner posts,
where c is a certain tolerance, e.g. 0.001 with point of application on the frame and not
on the decking. Deflections were monitored at
Numerical Example points A and B of Fig. 7.
Figure 7 shows a 20-ft by 60-ft (6-m by The experimental results available for this
18-m) diaphragm section sheathed with lum- diaphragm were used to test the structural
ber decking ( 2 ) and overlaid plywood ( 1 ) . analysis. As a simplification, the decking-
This diaphragm was built and tested at Oregon plywood assembly was considered as a unit,
State University. A comprehensive test report disregarding the deformation of the nails used
is available (Johnson 1971). The decking con- to kcep these two elements together. Further-
sisted of nominal 3-in. by 12-in. (76-mm by more, no consideration was given to the nailing
305-mm) laminated ponderosa pine boards, between the boards making u p the posts.
with actual thickness of 2.25 in. (57 m m ) . It Finally, it was assumed that, because of the
was nailed to the frame with 30d common random length of the boards in the decking
nails, with densities of 0.125 nails/in. (4.92 and the very limited slant nailing, the contribu-
 HI. I 35 1

tion of the decking to the stiffness of the cover board into the end grain of the posts. The
was negligible. initial stiffness k,: can be estimated by using the
The cover was therefore regarded as a procedure described by Wilkinson ( 1972).
2.625-in. (66.7-mm) thick plate with elastic The ultimate load mo, (assuming that ml,. =
properties derived from the contribution of the 0 ) can be obtained by following the method
plywood alone. Along the xi-yi axes oriented at discusscd by Larsen (1 973 ) . Thus,
45", the following were the elastic parameters
chosen:
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

The rotational parameters m , , ~mlo,

, and ko
can be derived by using the withdrawal data
[35] and assuming that rotation takes place
about the edge of the post. Therefore, accord-
ing to the width of the post, a different set of
with E L ,and E,,l the moduli of elasticity, G,t,t parameters is obtained for connections 0
the shear modulus, and IJ,O,I and v,,),, the Pois- and 9:
son ratios. The parameters [34] were derived
117,~ = 2400 1b.in. (270 N.m) for 0;
by equating the extensional and shear stiffness
of the 2.625-in. (66.7-mm) plate with that of moo = 1200 lb-in. (135 Nsm) for 9;
the ;-in. (9.5-mm) plywood, assuming a
For personal use only.

veneer thickness of 0.1 in. (2.5 mm) for the ko = 360 000 lb.in./rad
faces, 1.5 x loGIb/in." 10.4 x lo6 kN/m2) (40 700 N.m/rad) for 0 ;
1371
for modulus of elasticity along the grain and ko = 90 000 lb.in./rad
75 000 Ib/in."518 000 kN/m" for the shear (10 200 N.m/rad) for 9;
modulus 'through thickness'.
The framing lumber was assumed to have a
modulus of elasticity E = 1.54 x 10" Ib/in."
(10.6 x 10" kN/m2), corresponding to con- Finally, the parameters nz,,;,, m,:,, k:,, m,,,.,
struction grade Douglas fir in accordance with m,,,, and k,, for the cover-frame ccnnections
the Canadian Standard CSA-086 (Canadian can be obtained by considering that, for ex-
Standards Association 1976). Actual dimen- ample, a 30d common nail is deformed along
sions for the 2-in. by 10-in. lumber were 1.5 the frame member parallel to the grain of
in. (38 mm) by 9.25 in. (235 mm). Douglas fir when driven through 2.25 in. (57
The connections 9 and 0 (Fig. 7 ) between mm) of poilderosa pine decking and deformed
frame members consisted of four 16d common perpendicular to the grain of the decking. For a
nails driven into the end grain of the posts. deformation perpendicular to the frame mcm-
Based on information obtained from the ber, the nail bends in the direction normal to
USDA Wood Handbook (U.S. Department of thc Douglas-fir grain and parallel to that of the
Agriculture 1974), for side-grain withdrawal pine decking. Using Wilkinson's ( 1972) and
and using a 5 0 % reduction factor for end- Larsen's ( 1973 ) methods, the following pa-
grain withdrawal, the following parameters rameters were derived:
were obtained:
mo, = 400 lb (1780 N) per nail
k, = 10 000 lb/in. (1750 kN/m) per nail
[35] k,, = 40 000 lb/in. (7000 kN/m)

moc = 600 lb (2670 N) per nail

The parameters mo,, ml,., and kc correspond to kc = 10 000 lb/in. (1750 kN/m) per nail
the lateral deformation of 16d common nails
driven through a 1.5-in. (38-mm) Douglas-fir
 352 C A N . J . CIV. ENG. VOL. 4. 1977

Each bay in Fig. 7 was considered a single forces in the chords obtained from the analysis
finite element for a total of five. Each framing and those determined from simple statics. At
member was divided into four segments giving Po = 10 000 Ib (44.5 kN), the maximum ten-
a total of 6 4 members for the analysis. The- sile force in the bottom chord was found t o
oretical estimates for the deflection of the dia- be 1 4 529 Ib (64.6 kN). If the bending mo-
phragm were obtained in increments of Po = ment is assumed to be transmitted solely by
2000 Ib (8.9 k N ) . Figure 8 shows a compari- the chords, a simple calculation from statics
son between the predictions from the analysis gives a maximum chord force of 1 8 000 Ib
and the measured deflections at points A and (80.1 k N ) . The conservatism i n the latter is
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by 99.243.105.24 on 01/09/15

B of Fig. 7. Good agreement is obtained, re- due to the contribution of the cover, but the
producing the nonlinear behaviour observed in difference is not excessive: about 2 4 % .
the test. The deflections measured at points A
and B correspond to frame deflections and Conclusions
almost all of the observed nonlinearity is due A structural analysis for wood diaphragms
to the nonlinear behaviour of the nailing be- has been discussed. It has been implemented in
tween the cover and the framing. the computer program SADT a n d incorporates
Although the parameters chosen for the the basic features of wood structural assem-
analysis represent realistic estimates based on blies: orthotropic plate action and nonlinear
available literature, all could have been deter- connection behavior. The analysis has been
mined by carrying out simple tests on connec- shown to give reliable estimates for diaphragm
tions. A limited study was carried out to inves- deformations and it is capable of providing ap-
tigate the effect that variability in the connec- proximations for ultimate loads based on con-
For personal use only.

tors' properties has in the overall diaphragm nection yielding. The required conneetion
deflections. It was found that this effect is rela- load-deformation input data can be obtained
tively minor when compared with the effect of from testing or may be obtained, approxi-
variability in the stiffness properties for the mately, from information already in the litera-
cover. ture.
I t is interesting to compare, finally, the

14.0-
C A N A D I AST,\NDARDS
N ASSOCIAI-ION. 1976. Canadian
Stanclarcl CSA-086.
C A R N E YJ.. M. 1975. Bibliography on woocl ancl plywood
12.0- diaphragms. ASCE J. Struct. Div. 101 (STII), pp.
2423-2436.
IRONS.B. M. and TUCK,R. C. 1969. A version of the
10.0- Aitkenaccelelr~tor.forcomputeritelation.Int. J. Numer.
Methocls Eng. 1, pp. 275-277.
S . 1971. Accelerating the convergence of mz~t-
J E N N I N C A.
rix iterative processes. J . Inst. Math. Its Appl. 8, pp.
99-1 10.
-
Y
Point
JOHNSON. J . W. 197 1. Lateral test of a 20- by 60-foot roof
section sheathed with plywoocl overlaid on decking. Re-
A B port T-29, School of Forestry, Oregon State Univ., Cor-
vz~llis,OR.
Test(9): o WILKINSON T.. L. 1972. Analysis of nailed joints with
dissimilar members. ASCE J. Struct. Div. 98 (ST9). pp.
Theory : ---- - 2005-20 13.
U .S. DEPAR.ISMEN.S OF A C I I I C U L ~ S U
1974.
R E . Wood hand-
book. Handbook No. 72, p. 7-2.
LARSEN,H. J. 1973. The yield load of bolted and nailed
joints. Proc. Int. Union For. Res. Org., IUFRO5,South
Africa, pp. 636-654.
0.1 0.5 1.0 P H Z E M I E N I E CJ K
. SI ,. 1968. Theory of matrix str~~ctural
D e f l e c t i o n ( in.) analysis. McGlaw-Hill Bookcompany, New York, NY.
Z I E N K I E W I C0 Z. ,1971. The finite element method in en-
FIG.8. Comparison between test and theory (1 kip gineering science. McGraw-Hill Book Company, Lon-
= 4450 N; 1 in. = 25.4 m m ) . clon, Englancl.