You are on page 1of 7

BEHAVIOR AND MODELING OF WOOD-PEGGED TIMBER FRAMES

By William M. Bulleit,1 L. Bogue Sandberg,2 Members, ASCE, Matthew W. Drewek,3


Tonia L. O’Bryant4

ABSTRACT: Traditional timber connections with wood pegs are encountered in the renovation and rehabilitation
of historic wood structures and in the construction of new structures where rustic appearance and traditional
methods are desired. This study examined four types of traditional connections: mortise and tenon, mortise and
Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

tenon with a shoulder, mortise and tenon with a shoulder and a knee brace, and fork and tongue. A total of 60
specimens, primarily white oak and Douglas fir, were tested. About a quarter of the specimens were assembled
and allowed to season under simulated in-service load. All specimens were tested under simulated gravity load
up to failure or to about five times typical design load. Some knee-braced specimens were retested to simulate
lateral loads. Test results showed that joint behavior is primarily linear. A tightly fitting joint carries load with
less peg damage than a loose joint allows. A mortise and tenon with a shoulder performs better than a mortise
and tenon or a fork and tongue. At typical occupancy sustained live plus dead load levels, the effects of warping
and checking caused by drying shrinkage overshadow any damage caused by the sustained load. Analysis of
structures employing these connections requires that the behavior of the wood-pegged joints be included in a
reasonable manner. In frames without knee braces, the joints can be modeled as pinned connections. In frames
with knee braces, the post-to-beam and knee-brace connections can be modeled as pinned, although the reduction
in effective axial stiffness of the knee brace caused by the wood pegs must be included. The effects of shear
deformations should be incorporated into the analysis.

INTRODUCTION Framers Guild of North America has recently published a


workbook in which many of these papers and other material
The need to understand the structural behavior of wood- have been gathered (Timber 1995). Research sponsored by the
pegged timber connections has become more pressing since Timber Frame Business Council has examined wood dowel
the 1980s. This need has been partially driven by the trend connections, including tests and work toward using the Eur-
toward more renovation and rehabilitation of historic wood opean yield model to predict the capacity of the joints
structures. Another significant contributor to this need is the (Schmidt and MacKay 1997). Further joint tests have also
increased use of traditional construction methods in new struc- been performed by Reid (1997).
tures. The development and growth of the Timber Framers This paper presents the results of a project funded by the
Guild of North America (TFGNA) from a seedling organiza- National Science Foundation with material contributions by
tion in 1985 to its present size of almost 1,000 individual the Timber Framers Guild of North America. Both experi-
members and more than 100 firms is an indication of the re- mental and analytical work are discussed as well as further
newed interest in timber framing. work on frame analysis. The experimental portion of the study
The literature on the design and analysis of timber frame is discussed further in Sandberg et al. (1996) and O’Bryant
structures is relatively limited. Benson and Gruber (1980) dis- (1996). The analytical portion is discussed further in Bulleit
cuss structural analysis and design in an elementary fashion et al. (1996), O’Bryant (1996), and Drewek (1997).
and only for members. The analysis of frames is not addressed.
Brungraber (1985) tested joints and a few frames. He per- EXPERIMENTAL PLAN
formed two-dimensional finite-element analyses on some joint Four types of joints were included in the simulated gravity
details and proposed a three-spring joint model for frame anal- load test program. All employed 25 mm (1 in.) diameter red
ysis. His model would have required testing of many different oak pegs. The four joint details were: (1) mortise and tenon;
joint configurations to determine the spring constants for each (2) mortise and tenon with a shoulder (also called a housed
joint type. He compared his model to his frame tests with good mortise and tenon); (3) mortise and tenon with a shoulder and
results. Later work by Weaver (1993) showed that Brungra- a knee brace; and (4) fork and tongue. These details are illus-
ber’s frames were insensitive to joint behavior, so almost any trated schematically in Figs. 1 – 4. For each of the four joint
reasonable model worked well in predicting frame behavior. types, there were seven white oak specimens, six Douglas fir,
Kessel at al. (1988) reported on the reconstruction of an eight- and one each of Eastern white pine and salvaged Douglas fir.
story timber frame building in Germany. They modeled the Knee braces were of the same species as the corresponding
joints as pinned connections, but no tests were performed to
examine the adequacy of the model. Recently, Kessel and Au-
gustin (1995a and b, 1996) published their experimental stud-
ies on the load capacity of wood-pegged joints. The Timber
1
Prof., Dept. of Civ. and Envir. Engrg., Michigan Technol. Univ., 1400
Townsend Dr., Houghton, MI 49931-1295.
2
Prof., Dept. of Civ. and Envir. Engrg., Michigan Technol. Univ., 1400
Townsend Dr., Houghton, MI.
3
Struct. Engr., Cleveland-Cliffs, Inc., Ishpeming, MI 49849.
4
Struct. Engr., Black and Veatch, Ann Arbor, MI 48105.
Note. Associate Editor: David Rosowsky. Discussion open until June
1, 1999. To extend the closing date one month, a written request must
be filed with the ASCE Manager of Journals. The manuscript for this
paper was submitted for review and possible publication on July 7, 1998.
This paper is part of the Journal of Structural Engineering, Vol. 125,
No. 1, January, 1999. 䉷ASCE, ISSN 0733-9445/99/0001-0003 – 0009/
$8.00 ⫹ $.50 per page. Paper No. 18707. FIG. 1. Schematic of Mortise and Tenon Joint

JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999 / 3

J. Struct. Eng., 1999, 125(1): 3-9


for 16 months while subjected to a load of 8 kN (1,800 lb)
distributed over the middle third of the beam. One specimen
of each joint type of white oak and Douglas fir were assembled
while fairly green and allowed to dry along with the loaded
specimens. The remaining specimens were allowed to season
unassembled for a minimum of one year prior to testing. All
seasoning was done at uncontrolled indoor ambient conditions
with monitoring of temperature and relative humidity (RH)
(approximately 65% RH in summer and 35% RH in winter).
The short-term load tests were conducted as shown in Fig.
Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

5. The cylinders were connected to maintain equal load and


were equipped with separate load cells. The tie rods also had
a load cell. For testing the fork and tongue specimens, the tie
rods were attached to the post immediately below the beam.
FIG. 2. Schematic of Mortise and Tenon Joint with Shoulder Deformations were monitored with linear variable differential
transducers (LVDTs) at the positions shown in Fig. 6. These
provided data on the deflections of the beam at the load points
and the relative horizontal, vertical, and rotational movement
between the beam and post. Load and displacement readings
were taken at load increments of 2.2 kN (500 lb) per cylinder
or more frequently if failure appeared imminent. A PC-based
data acquisition system was used. Load cells and LVDTs were
all calibrated as wired into the system, prior to and during the
series of tests. Maximum load per cylinder was restricted by
cylinder limitations to 67 kN (15,000 lb).
A second series of tests were conducted to evaluate knee-
brace tensile behavior, because tensile knee brace forces occur
under lateral loads. The specimens used for these tests were
ones that had survived the main test with only minor damage
to the main joint tenon and to the pegs. To load the knee braces
in tension, the beam end was cut off, reducing the length from
the face of the post to about 1.83 m (72 in.). The specimen
was turned upside down and configured as shown in Fig. 7.
FIG. 3. Schematic of Knee Brace Connection
Only load, tie rod force, and deflection at the load were re-
corded. New red oak pegs were used for the knee-brace con-
nections, and new or undamaged pegs were used to attach the
beam to the post.

FIG. 4. Schematic of Fork and Tongue Joint

main members, with the exception that red oak knee braces
were used with three of the white oak specimens.
Each specimen consisted of a 3.7 m (12 ft) beam framing FIG. 5. Test Frame Configuration
into a post at one end of the beam with the far end simply
supported. The posts for the first three types were 2.2 m (7.3
ft) long, with the beam connected at about 1.4 m (4.5 ft) from
the bottom. The fork and tongue specimens had the beam con-
nected at the top of a 1.4 m (4.5 ft) post. Cross sections of
the beams and posts were about 180 ⫻ 230 mm (7 ⫻ 9 in.).
The knee braces were about 83 ⫻ 133 mm (3.25 ⫻ 5.25 in.),
installed at 45⬚. They ended in a tenon and were pegged into
the post and beam at 0.71 m (28 in.) from the inner face of
the perpendicular main member. The mortises and tenons were
cut while the members were in the green condition, with the
exception of the salvaged Douglas fir. All this work was per-
formed by experienced craftspeople working from drawings
prepared by the Timber Framers Guild of North America.
Two specimens of each joint type of white oak and Douglas
fir were assembled while still fairly green and allowed to dry FIG. 6. Location of LVDTs

4 / JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999

J. Struct. Eng., 1999, 125(1): 3-9


Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

FIG. 7. Knee Brace Test Configuration

EXPERIMENTAL RESULTS FIG. 9. Peg Failure Modes


A large quantity of data was generated and recorded during
the various tests. Only selected results can be presented here. sides of the tenon. In this mode, the root of the tenon was
Further details can be found in O’Bryant (1996). peeled out of the top of the beam. This occurred when the gap
beneath the tenon was small and was accompanied by consid-
erable crushing in bearing on the bottom of the tenon. A com-
Seasoning under Load
bination of the two modes described above was seen in several
The seasoning under load did not produce any significant specimens. In most specimens, there was some combination
effect on joint behavior. The specimens were monitored for of bearing, shear, or flexural failure in the pegs, as shown in
changes in vertical and horizontal gaps at the joints, for rela- Fig. 9. Further details are available in Sandberg et al. (1996).
tive rotation and vertical movement between beam and post,
and for beam deflection. Any load effects on gaps and relative Mortise and Tenon with Shoulder Behavior
movement seemed to be overwhelmed by the effects of shrink- These joints are favored for beam-to-column connection of
age. Even the beam deflections were not consistent with the load-bearing beams. The results clearly showed the benefit of
expectation of creep-induced deflection during drying. In sev- the extra effort required to fabricate the recess (housing) in
eral cases, beams actually moved upward due to drying-in- the post and the shoulder behind the tenon. Virtually all of
duced warping, rather than down due to creep. There was con- these specimens failed in flexure or flexure/shear in the beam,
siderable twisting and severe checking in many of the white away from the joint. There was some tension-perpendicular
oak specimens. These problems were less evident in the Doug- splitting similar the first failure mode described for the mortise
las fir. The extent of warping and checking of loaded speci- and tenon without a shoulder, as shown in Fig. 8(a). Again,
mens did not differ from that exhibited by the specimens being this resulted from a sizable gap between the bottom of the
stored without load in the same room. shoulder and the sill of the tenon. Damage to pegs was similar
to that seen in the plain mortise and tenon. The peg damage
Unshouldered Mortise and Tenon Behavior in these specimens was caused by a combination of relative
beam-to-post vertical and rotational movements. Further de-
These joints are generally intended for use in non – loading
tails are available in Sandberg et al. (1996).
bearing situations. Their performance indicated that this is
wise. Significant damage usually occurred in the joint prior to Behavior of Mortise and Tenon with Shoulder and
flexural or shear failure of the beam. One joint failure mode, Knee Brace
Fig. 8(a), was a prying, perpendicular-to-grain tensile failure
in the tenon, generally through the upper peg hole. This tended The addition of the knee brace caused several changes in
to occur when there was a gap of 6 mm (0.25 in.) or more joint behavior. First, the compression force induced in the knee
between the bottom of the beam and the sill of the mortise. brace by the loads caused a tensile force in the beam at the
The other mode, Fig. 8(b), was a rolling shear failure in the joint. This tended to pull the tenon and shoulder out of the
radial-tangential plane extending out into the beam on both post. The result was either a peg failure in bending or shear,
a block shear failure of the portion of the tenon between the
upper peg and the end of the tenon, or a splitting failure of
the tenon at the upper peg hole. Knee-brace response was
highly dependent on the degree of contact between the ends
of the knee brace and the faces of the beam and post. If a gap
was present at either face, all knee-brace force was carried
through the peg at that end. This led to peg damage similar to
that discussed previously. Relatively little horizontal move-
ment occurred in the beam/column joint until the knee brace
went into bearing at both ends. Then, horizontal movement
increased rapidly with increasing load. In the majority of
cases, white oak specimens reached the maximum load of 67
kN (15,000 lb) without catastrophic failure of any component.
Douglas fir and eastern white pine specimens generally failed
in bending or a combination of bending and shear between the
knee brace attachment point and the roller supported end of
FIG. 8. (a) Tension Perpendicular-to-Grain Failure; (b) Rolling the beam. Further details are available in Sandberg et al.
Shear Failure (1996).
JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999 / 5

J. Struct. Eng., 1999, 125(1): 3-9


Fork and Tongue Behavior
This type of joint is most commonly used as a ridge joint
that joins two rafters. For this reason, it must be noted that
the testing configuration used in this study does not fully du-
plicate actual joint force conditions. The primary omission was
that the forked post did not apply a force perpendicular to itself
and acting in compression against the end of the beam.
In many respects, the fork and tongue behavior resembled
that of the mortise and tenon joints without a shoulder. Be-
cause the edge and end distances on the peg holes were
Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

greater, the fork and tongue specimens had less of a tendency


to suffer the tension perpendicular to grain failure of Fig. 8(a).
Further details are available in Sandberg et al. (1996).

Knee Braces in Tension


FIG. 10. Rotation of Beam Relative to Post
Aside from peg failures, a number of the specimens with
knee braces showed little or no damage in the main joint and ing transmitted through the overall connection (main joint and
the knee brace ends. In several cases, the beam tenon damage knee brace). However, observations during the tests and anal-
that did occur was such that it would not significantly influ- ysis of the data show that the amount of moment resistance is
ence behavior if the axial force, shear force, and moment on highly dependent on how tightly the joints fit together, partic-
the joint were reversed. This permitted ‘‘recycling’’ of these ularly the knee-brace joints.
joints as specimens to test the tensile capacity of the knee
braces. The testing procedure used was discussed in the ‘‘Ex- ANALYSIS TECHNIQUE
perimental Plan’’ section.
The specimens with white oak knee braces were all limited Joint Modeling
by knee-brace peg failure. The pegs failed by some combi-
nation of bending and/or a very steep, brash shear failure. Con- A joint model is required in order to analyze a frame. De-
siderable bearing crushing of the peg was also evident. The termining a model that includes as much relevant behavior as
Douglas fir and red oak knee braces were limited by peg fail- possible is the first step in this, followed by simplification of
ure in a few cases, but were more likely to suffer a block shear the model if feasible. Initially, an attempt was made to develop
failure of the material between the peg hole and the end of a model that included the pegs, modeled as springs related to
the knee-brace tenon, referred to as a ‘‘relish’’ failure in the the bending stiffness of the peg, and all contact points between
timber framing community. the mortise and tenon, where contact was modeled using a
Given recorded data for the cylinder and tie rod forces, it pressure bulb analogy (Weaver 1993). This approach to mod-
was possible to estimate the force in the knee brace as a func- eling the joint behavior was unsuccessful, although it did lead
tion of applied load. This was done by assuming the main joint to the second modeling approach, and use of it showed that
acted as though the beam was pinned into the column. Based the frames tested by Brungraber (1985) were insensitive to
on this, ultimate knee-brace forces for six oak specimens av- joint behavior.
eraged 23.5 kN (5,280 lb) with a coefficient of variation The second approach was to model the combination of two
(COV) of 0.22. Half of these knee braces were red oak; two pegs and internal joint contact as two translational springs, one
of which had block shear failures through the end of a knee vertical and one horizontal, and a rotational spring. The details
brace tenon at a below average load. The remaining failures of this approach can be found in Bulleit et al. (1996).
were all in the pegs. The corresponding strength values for the A third approach using a commercial matrix analysis code,
six Douglas fir specimens were 15.9 kN (3,580 lb) with a COV SAP IV (Bathe et al. 1973) and later SAP90 (1992), is dis-
of 0.24. Three of these failed by block shear failures through cussed below. This method uses no specialized elements that
the end of a knee-brace tenon, two by peg breakage, and one are unavailable in standard commercial analysis codes. It uses
by a split in the beam at the peg hole. only beam-column elements, with shear effects included.

Load-Displacement Behavior Frame Modeling

All of the joint types showed remarkably linear response up The test specimen for a mortise and tenon with a shoulder
to about a third of ultimate load. The mortise and tenon, mor- and a knee brace is shown in Fig. 5. A frame model of this
tise and tenon with shoulder, and fork and tongue showed test specimen is shown in Fig. 11. To minimize clutter, only
beam deflections and relative joint rotations that would indi-
cate these joints transmit very little moment and are function-
ing essentially as hinges. This probably should not be surpris-
ing when the relative stiffness of the pegs is compared with
that of the beam and post. An idea of the degree of moment
transfer from beam to post can be made by looking at the
relative rotation between beam and post. Fig. 10 shows the
range of experimental rotations versus load in the linear range.
The rotations have been normalized by the ratio of specimen
beam modulus of elasticity over average modulus. The range
falls very close to the plot for a simple beam, indicating very
little moment transfer.
The knee-braced joints are more complex, as discussed pre-
viously. It is evident in these specimens that moment was be- FIG. 11. Model of Test Frame

6 / JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999

J. Struct. Eng., 1999, 125(1): 3-9


those nodes and elements referred to in the text are shown in Frame Model versus Test Results
the figure. Node numbers are circled. Nodes 14 and 15 are the
Frame models similar to that described above were devel-
load points. Element 18 is the tieback system and is made of
oped for the other three cases tested: mortise and tenon with-
steel; elements 6, 8, 12, and 17 are very stiff members that
out a shoulder, mortise and tenon with a shoulder, and fork
account for eccentricity of load. For example, because the pegs
and tongue. For these three cases, deflection behavior and
are not oriented on the centerline of the post, element 6 locates
bending moments were well predicted using the model. How-
the pegs at their correct place and transfer forces and moments
ever, further analyses showed that the beam behavior was well
between that location and the post centerline. Elements 8, 12,
predicted by assuming that the beam was simply supported on
and 17 are similar. The stiffness of these members needs to
both ends. As discussed above, this was also supported by
be about 10,000 times greater than the most flexible element
examination of the rotational behavior of the test specimens.
connected to it. Element 7 models the region that includes the
Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

The analysis results using the model in Fig. 11 can be com-


shoulder and the tenon from the end of the beam to the cen-
pared with the test results at 22.24 kN (5,000 lb) of load at
terline of the pegs. The average of the cross-sectional prop-
each load point. This is acceptable because the behavior of the
erties of the tenon and shoulder were used for this element.
test specimens was quite linear through much of the load
Elements 3 and 14 have cross-sectional properties that account
range. The test results discussed are for specimens fabricated
for the reduction in post area caused by the mortise. Elements
from Douglas fir-larch. There were seven specimens for the
1, 2, and 15 have section properties for the post, and elements
case under consideration here, so the test results used for com-
13 – 16 have section properties for the beam. The elements are
parisons are an average over the seven specimens. The MOE
six degree-of-freedom beam-column elements. The stiffness
used in the analysis was 11 GPa (1,600 ksi). This is the value
matrices for all elements include the effect of shear deforma-
for No. 1 DF-L posts and timbers (Supplement 1991). The 3%
tions (Ghali and Neville 1972). In all analyses, the shear mod-
reduction in MOE incorporated into the NDS Supplement val-
ulus was assumed to be MOE/16 (Timber 1994), where MOE
ues was not removed due to the small effect that this would
is the modulus of elasticity in bending.
have on the final results.
In Fig. 11, the artificial members that model the pegs in the
The posts and beams were all approximately 180 mm (7 in.)
knee brace are elements 9 and 11, with moment releases at
wide and 230 mm (9 in.) deep. The pegs were red oak, and it
nodes 8 and 12, respectively. The pegged connections on the
was assumed that the peg MOE was the same as the timber
knee brace are now pins, but the relatively flexible peg is dealt
MOE. This is not too unreasonable when the approximate na-
with by modifying the cross-sectional area of members 9 and
ture of the peg stiffness calculation is considered. The peg
11. This is done by equating the axial deformation of the short
diameter was 25 mm (1 in.) and the peg span was 127 mm (5
element to the bending deflection of the peg under a unit con-
in.). The short elements at the end of the knee brace were each
centrated load at midspan. Fig. 12 shows a peg modeled as a
71 mm (2.8 in.) long. Thus, Aeff was 32 mm2 (0.05 in.2).
beam for determination of the short element axial stiffness.
For comparison purposes, at 22.24 kN (5,000 lb) per load
The concentrated load is the force applied to the peg by the
point, the average deflection at the load point nearest the knee
tenon, set equal to 1.0 when determining the peg deflection
brace (node 14 in Fig. 6) was 10.4 mm (0.41 in.), and at node
used in the stiffness calculation, and the reactions are located
15 it was 10.7 mm (0.42 in.); the horizontal movement at the
at the midpoint of each side of the mortise (or, in timber fram-
postbeam interface caused by the prying action of the knee
ing parlance, at the midpoint of each of the mortise cheeks).
brace was 0.8 mm (0.03 in.). The force in the tieback, which
The element axial stiffness is the stiffness of an equivalent
was measured using strain gauges, was 5.03 kN (1,130 lb).
linear spring located at midspan of the peg. Performing this
Using A9 and A11 — the areas of members 9 and 11,
calculation gives an effective area of the short element of
respectively — equal to 32 mm2 (0.05 in.2) and A7 equal to 64
Aeff = 48Epeg Ipeg Lend /Ekb L3peg (1) mm2 (0.10 in.2), the deflections at nodes 14 and 15 were 10.4
mm (0.41 in.) and 10.4 mm (0.41 in.), respectively; the post-
where Epeg = MOE of the peg; Ipeg = moment of inertia of the
beam interface displacement was 0.8 mm (0.03 in.), and the
peg; Lpeg = span of the peg, as described previously and shown
tieback force was 4.96 kN (1,115 lb). The tieback force and
in Fig. 12; Ekb = MOE of the knee brace; and Lend = length of
displacements compare quite well with the test results. Some
a knee brace end element (element 9 or 11 in this case). Each
further discussion of sensitivity analyses of this test frame can
of the elements at the ends of the knee brace will have an area
be found in Bulleit et al. (1996).
determined using (1).
The pegged joint at the postbeam connection is modeled FULL-SCALE FRAME ANALYSIS
using a moment release at node 7 in element 7. If it is desirable
to include the tendency for the beam to be pried out of the Drewek (1997) analyzed two frames using the joint model
mortise when the knee brace is in compression, then the tenon discussed above. The software package SAP 90 (SAP 90 1992)
element, element 7 in Fig. 11, can have its area reduced using was used in the analyses. One of the frames was fairly straight-
(1). For that situation, the final effective area will be two times forward and will not be discussed here. The second one was
the value found in (1), because there are two pegs acting in taken from plans given in the TFGNA workbook (Timber
parallel. This was done in the model of the test specimens. 1995) and is shown in Fig. 13.
The observations and analyses discussed above have led to
basic guidelines for modeling traditionally connected timber
frames. The guidelines are:
1. Frame members should be modeled as beam-columns
with the effects of shear included.
2. Joints should be assumed to carry no moment, that is,
free to rotate.
3. Eccentricity of force should be included. For instance,
when the pegs are not located at the centroid of a column
(post) or beam, the eccentricity between the pegs and the
column or beam centroid should be included in the anal-
FIG. 12. Peg Stiffness Model ysis.
JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999 / 7

J. Struct. Eng., 1999, 125(1): 3-9


Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

FIG. 13. Full-Scale Frame

properties into this region. The axial stiffness of element 7 is


controlled by the state of stress in element 8. If element 8 is
in compression, then contact may be assumed, and the axial
stiffness becomes the same as the beam, i.e., element 8. If the
beam is in tension, then the axial stiffness of element 7 is
controlled by the pegs, and the pegs are modeled as short
beams with a concentrated load acting at the center of the
tenon, as discussed above. The other section properties of this
element should be based on the tenon dimensions. Again, there
FIG. 14. (a) Model of Post-Beam Connection; (b) Model of
are indications (Drewek 1997) that the properties, other than
Knee Brace-Beam Connection axial stiffness, may be based on the beam dimensions, i.e., a
reduction only in the axial stiffness when the beam is in tension.
4. Where contact between one member and another may Next consider the knee brace/beam connection circled and
have a significant effect on the behavior of the frame, labeled B in Fig. 13. This joint is detailed in Fig. 14(b). Ele-
the effect of contact should be examined. ment 11, with nodes 12 and 13, is used to model the peg
effects. There is a moment release at node 12. The axial stiff-
This fourth guideline generally requires that the frame be an- ness is based on the cross-sectional area of the knee brace
alyzed more than once for each load case. First, an analysis when the knee brace is in compression and is in contact with
must be performed to discern which members are in tension. the beam and post. It is based on the single peg in bending,
Members in tension should have the axial stiffness of the con- as described above, when the knee brace is in tension, or if
nections on that member based on the peg behavior only; i.e., the knee brace is in compression and not in contact with the
no contact effects should be included. This is particularly im- beam or post. The other properties are based on the knee-brace
portant for members that act as braces, including knee braces cross section, although since there is a member release on each
and collar ties. Second, braces in compression should be in- end of the knee brace, its behavior is controlled by the axial
cluded in the analysis by accounting for the possibility that stiffness of the elements that make up the knee brace.
they may or may not be contact with the adjacent member. As an example of the effect of the knee brace, consider
This last guideline is a direct result of the behavior seen in positive moment at point C of the left floor beam in Fig. 13.
the simulated gravity load tests of the knee-brace specimens, For the case of a 1.44 kPa (30 psf) floor load on the beam
where contact between the beam and the knee brace had a and a 2.39 kPa (50 psf) snow load on the roof, the positive
significant effect on the behavior of the frame and was not moment is 3,480 N-m (30,800 in.-lb) when the knee brace is
consistently present until fairly high load levels. in compression and in contact, and 8,220 N-m (72,750 in.-lb)
The post-beam connection circled and labeled A in Fig. 13 when the knee brace is not in contact. This is a significant
is modeled like the post-beam connection discussed above [see difference. If workmanship is excellent and shrinkage is neg-
Fig. 14(a)]. The elements labeled 3 and 4 are beam-column ligible, then, under service loads, contact would be likely. But
elements that model the reduced cross section at the mortise, in our tests, loads at service load levels did not always produce
as discussed above, but some recent work (Drewek 1997) in- contact. This indicates that, until better knowledge is available
dicates that it may be adequate to continue the full post section about contact behavior in actual frames, bracketing of load
8 / JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999

J. Struct. Eng., 1999, 125(1): 3-9


effects is important. Similar behavior also occurs for moments findings, conclusions, or recommendations expressed in this publication
in the rafters where the collar tie acts as a compression brace. are those of the writers and do not necessarily reflect the views of the
National Science Foundation. The writers would like to thank the Timber
Considering a 0.96 kPa (20 psf) basic wind pressure acting Framers Guild of North America for providing the test specimens used
from the left with no floor or snow load, the tensile force in in this project.
the left exterior knee brace, D in Fig. 13, is 6,850 N (1,540
lb) if contact is assumed for all members in compression, and
8,380 N (1,883 lb) if knee braces in compression are assumed APPENDIX. REFERENCES
not to be in contact. This is a potentially significant difference, Bathe, K.-J., Wilson, E. L., and Peterson, F. E. (1973). SAP IV: a struc-
particularly as it affects the pegs connecting the knee brace to tural analysis program for static and dynamic response of linear sys-
the post and beam. tems. National Technical Information Service, Springfield, Va.
The load effects that are sensitive to contact are also sen- Benson, T., and Gruber, J. (1980). Building the timber frame house.
Downloaded from ascelibrary.org by UFMG - Universidade Federal de Minas Gerais on 08/07/18. Copyright ASCE. For personal use only; all rights reserved.

sitive to peg stiffness. The simple beam model described pre- Charles Scribner’s Sons, New York.
Brungraber, R. L. (1985). ‘‘Traditional timber joinery: a modern analy-
viously for modeling peg stiffness worked well when it was sis,’’ PhD dissertation, Stanford University, Palo Alto, Calif.
used in analysis of our test frames, but recent single peg joint Bulleit, W. M., Sandberg, L. B., O’Bryant, T. L., Weaver, D. A., and
tests (Reid 1997) showed that the effective span of the peg is Pattison, W. E. (1996). ‘‘Analysis of frames with traditional timber
longer than that predicted in the aforementioned approach. connections.’’ Proc., 1996 Int. Wood Engrg. Conf., Vol. 4, Omnipress,
This means that the model used in this study might overesti- Madison, Wis., 232 – 239.
mate the stiffness of the pegs. The effect of this overestimation Drewek, M. W. (1997). ‘‘Modeling the behavior of traditional timber
frames,’’ MS thesis, Michigan Technological University, Houghton,
on the forces and moments was relatively small (<5%) for the Mich.
frame examined in this study. Ghali, A., and Neville, A. M. (1972). Structural analysis: a unified clas-
Suggested areas of further research include: (1) simplifica- sical and matrix approach. International Textbook Co., Scranton, Pa.
tion of frame analysis by minimizing the number of elements Kessel, M. H., and Augustin, R. (1995a). ‘‘Load behavior of connections
required for accurate analysis, e.g., removal of the short stiff with pegs I.’’ M. Peavy and R. Schmidt, translators, Timber Framing,
elements; (2) further enhancement of the peg model; (3) joint 38, 6 – 9.
Kessel, M. H., and Augustin, R. (1995b). ‘‘Load behavior of connections
capacity determination for use in joint design; and (4) full- with pegs II.’’ M. Peavy and R. Schmidt, translators, Timber Framing,
scale frame tests for comparison with analysis procedures. 39, 8 – 10.
Kessel, M. H., Speich, M., and Hinkes, F.-J. (1988). ‘‘The reconstruction
CONCLUSIONS of an eight-floor timber frame house at Hildesheim (FRG).’’ Proc.,
1988 Int. Timber Engrg. Conf., Forest Products Research Society, Mad-
A significant amount of testing and analysis has led to basic ison, Wis., 415 – 421.
guidelines for the analysis of wood-pegged timber frames. The O’Bryant, T. L. (1996). ‘‘Modeling the behavior of pegged timber con-
guidelines are as follows: nections,’’ MS thesis, Michigan Technological University, Houghton,
Mich.
1. Frame members should be modeled as beam-columns Reid, E. H. (1997). ‘‘Behavior of wood pegs in traditional timber frame
connections,’’ MS thesis, Michigan Technological University, Hough-
with the effects of shear included. ton, Mich.
2. Joints should be assumed to carry no moment. Sandberg, L. B., Bulleit, W. M., O’Bryant, T. L., Postlewaite, J. J., and
3. Eccentricity of force should be included; for instance, when Schaffer, J. J. (1996). ‘‘Experimental investigation of traditional timber
the pegs are not located at the centroid of a column (post) connections.’’ Proc., 1996 Int. Wood Engrg. Conf., Vol. 4, Omnipress,
or beam, the eccentricity between the pegs and the column Madison, Wis., 225 – 231.
or beam centroid should be included in the analysis. SAP90: a series of computer programs for the finite element analysis of
structures. (1992). Computer and Structures, Inc., Berkeley, Calif.
4. Where contact between one member and another may Schmidt, R. J., and MacKay, R. B. (1997). Timber frame tension joinery.
have a significant effect on the behavior of the frame, University of Wyoming, Laramie, Wyo.
the effect of contact should be examined. Supplement to the national design specification for wood construction.
(1991). National Forest Products Association, Washington, D.C.
Further research into both analysis and design of these types Timber construction manual. (1994). 4th Ed., American Institute of Tim-
of frames is still necessary. ber Construction, Englewood, Colo.
Timber frame joinery and design workbook. (1995). Timber Framers
ACKNOWLEDGMENTS Guild of North America, Bellingham, Wash.
Weaver, D. A. (1993). ‘‘Modeling the behavior of traditionally connected
This material is based upon work supported in part by the National timber frames,’’ MS thesis, Michigan Technological University,
Science Foundation under Grant Number MSS-9108064. Any opinions, Houghton, Mich.

JOURNAL OF STRUCTURAL ENGINEERING / JANUARY 1999 / 9

J. Struct. Eng., 1999, 125(1): 3-9

You might also like