Journal of Constructional Steel Research 67 (2011) 398–406

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Bolted steel slip-critical connections with fillers: II. Behavior

Mark D. Denavit a , Daniel J. Borello a , Jerome F. Hajjar b,∗
a
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA
b
Department of Civil and Environmental Engineering, Northeastern University, Boston, MA, USA

article info abstract

Article history: Research has been conducted to better understand the effect of fillers in bolted steel connections. In
Received 11 May 2010 a companion paper, the results of sixteen experiments on bolted steel slip-critical connections with
Accepted 4 October 2010 fillers are presented along with proposed design recommendations. In this paper, detailed behavior of
the specimens is documented through an examination of deformation and strain response. Additionally,
Keywords: mechanisms are proposed that clarify key aspects of the behavior of bolted connections with fillers,
Steel connections
including prediction of slip and shear strengths. A stochastic analysis, using order statistics, is employed
Slip-critical bolted connections
Bolted bearing connections
to quantify the detrimental effects of multiple possible slip surfaces on expected slip strength. The use
Filler plates of multiple plies and the effects of developing the filler plate are investigated with respect both to
the experimental results and the proposed behavioral mechanisms. The results indicate that the use
of multiple plies exacerbates the detrimental effects on slip strength and, to a lesser extent, on shear
strength. Furthermore, filler development reduces and in many cases eliminates the reduction in slip and
shear strengths.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction Borello et al. [5,6] provide a full description of an experimental

Often in bolted steel construction, it is necessary to connect
members of different depths. In these cases, filler plates are used
to provide a common faying surface. The use of filler plates has
an influence on the behavior of the bolted connection including
the slip strength and ultimate shear strength. Although prior
experimental research on bolted connections with fillers is limited,
important trends in behavior have been identified. In a series
of tests by Lee and Fisher [1], the slip strength of specimens
with fillers was found to be approximately 20% less than that of
control specimens without fillers. A series of tests by Frank and
Yura [2] observed a slip strength reduction of approximately 17%
for a connection with a single ply filler and an approximately
46% reduction for a connection with a multiple ply filler. Also
observed in this series of experiments was a reduction in the shear
strength depending on filler thickness. This finding served as the
basis for the bolt shear strength reduction formula in the AISC
Specification [3]. A recent study by Dusicka and Lewis [4] observed
similar trends for fillers up to 25 mm (1 in.) thick, but found that the
strength of connections with 51 mm (2 in.) thick fillers was greater
than that of thinner fillers, indicating that detrimental effects of
adding fillers reach a peak and decrease for larger filler thicknesses.
∗ Corresponding address: Department of Civil and Environmental Engineering,
400 Snell Engineering Center, 360 Huntington Avenue, Northeastern University,
Boston, MA 02115, USA. Tel.: +1 617 373 3242; fax: +1 617 373 4419.
E-mail address: jf.hajjar@neu.edu (J.F. Hajjar).
program that was undertaken to further investigate the effects of
various configurations of filler plates on the behavior of bolted steel
slip-critical connections. Sixteen full scale specimens were tested
to failure. Each specimen consisted of two wide-flange members,
connected by two 51 mm (2 in.) thick splice plates, with filler
plates provided where required to provide a constant connection
depth (e.g., Fig. 1). The bottom column for all specimens was a W14
× 730. The top column was a W14 × 159, W14 × 455, or W14
× 730. The W14 × 159 and W14 × 455 top column specimens
required a filler plate of 95 mm (3 3/4 in.) and 41 mm (1 5/8 in.)
respectively. Instrumentation included linear variable differential
transformers (LVDTs) and strain gages affixed at various locations
on the specimens and the load cell of the 13.3 MN (3000 kip)
testing machine. The specimens were identified based on the top
column nominal weight (in lbs per foot), development (n—none,
h—half, f—full), and unique details. Where duplicate specimens
were tested, an additional specimen number was added to the end
of the designation. For example, the second undeveloped specimen
with a W14 × 159 top column was identified as 159n2.
This paper describes the behavior observed in the experimental
program and proposes mechanisms to explain the slip and
shear strengths documented in the companion paper [6]. Load-
deformation response is the primary means of evaluating behavior,
with the overall response progressing from a stiff linear response
to the stage of slip followed by bolt shear failure. Specimen 455h,
with a W14 × 455 top column and half developed filler plate, is
representative of typical behavior of all of the specimens and is
presented in greater detail in this paper.

0143-974X/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2010.10.001
 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406 399

Table 1
Slip load and sequence of slip per faying surface.
Specimen Top column/filler plate slip Filler plate/splice plate
load and sequence slip load and sequence
North (kN) South (kN) North (kN) South (kN)

730-std 7549a (1) 7549a (1) – –

730-over 7268a (1) 7268a (1) – –
159f 10782 (3) 10782 (3) 5445 (1) 5445 (1)
159h 7549 (1) 7549 (3) 7549 (4) 7549 (1)
159n1 8358 (1) 8358 (1) 8358 (3) 8358 (3)
159n2 7580 (1) 7580 (1) 7580 (4) 7580 (1)
455f 6090 (4) 6090 (3) 6090 (1) 6090 (1)
455h 5227 (3) 5498 (4) 5227 (1) 5227 (1)
455n1 6174 (4) 6174 (3) 6174 (1) 6174 (1)
455n2 6374 (1) 6374 (3) 6374 (4) 6374 (1)
159n-2ply1 4559 (3) 4559 (2) 5333 (4) 2927 (1)
159n-2ply2 5996 (1) 5996 (1) 6143 (3) 6143 (3)
159h-TC 9088 (3) 9088 (3) 7233 (1) 7233 (1)
159n-TC 6921 (3) 5738 (1) 5738 (1) 6921 (3)
159f-weld –b –b 7495 (1) 7495 (1)
159h-weld 11165 (3) 11165 (3) 7188 (1) 7188 (1)
(x) Denotes xth surface to slip.
a
Slip between top column and splice plate.
b
Slip was not achieved.

considering stochastic effects was employed to investigate these

Fig. 1. Typical specimen.
2. Effect of fillers on slip behavior
The displacement and strain measurements (Figs. 2 and 3) show
that the behavior of the bolted connections prior to slip is typically
linear. Prior to slip, there is no relative motion between the splice
plate and filler plate or the filler plate and the top column (Fig. 2).
During this stage, the strain measurements indicate relatively
uniform introduction of force into the splice plate, as well as slight
bending in the splice plates in the gap between the top and bottom
columns (Fig. 3). The bending is due to the natural eccentricities
existing in the test specimen and has little effect on the behavior
of the connection.
Upon reaching the slip load, there was a sudden increase in
displacement, corresponding to the slip of at least one of the faying
surfaces. The slip event often lasted tens of seconds as the load
in the testing machine was stabilized. During this period, stress
within the connection was redistributed, as indicated by the offsets
in the strain measurements (Fig. 3). Table 1 shows the order in
which the various surfaces of each specimen slipped. The order of
slip was determined by examining the measured displacements.
Frequently, the slip of multiple surfaces occurred within the data
sampling period (0.1 s). These surfaces were denoted to have
slipped at the same time (e.g., specimen 730-std). In other cases
(e.g., specimen 159h), two surfaces slipped at the same load, but
the displacement data indicate which of the two surfaces slipped
first.
While the average test-to-predicted ratio exceeded unity in this
series of sixteen experiments, three specimens slipped prior to the
predicted load. One, a specimen with multiple ply fillers (159n-
2ply1), failed at 52% of the predicted load [5,6]. Additionally, prior
research [1–3] indicates that the use of fillers has a detrimental
effect on the slip resistance of bolted connections. An analysis
observations.
Significant uncertainty is observed in measured values of the
slip coefficient [7,8]. When the slip coefficient is considered a
randomly varying quantity in connections with more than one
faying surface, failure may occur at a load less than would be
indicated by a deterministic analysis assuming a single faying
surface. Thus, the more slip surfaces there are, the more likely a
lower value of the slip coefficient will be present for one of the
slip surfaces. Therefore, it is more likely that initial slip of the
connection will be at a lower load than in similar connections
with fewer slip surfaces. As one example of how to address
the detrimental effect of additional slip surfaces, the AASHTO
Specification limits the number of plies in fillers 6 mm (0.25 in.)
thick or greater to at most two, unless approved by the engineer [9].
Statistical data of the slip coefficient are obtained from
experimental tests. These tests are, in general, conducted with
two slip surfaces, such as the ancillary tests presented in [5,6]
(Fig. 4). Many of the slip tests summarized by Grondin et al. [8]
follow a similar testing scheme. A distinction needs to be made
whether the slip coefficient from these tests is representative of
the lower of the two slip surfaces (Assumption A) or the average of
the two slip surfaces (Assumption B). A related distinction needs
to be made whether failure of the main connection is defined as
when the lowest slip strength of any surface in the connection
is exceeded (Assumption C) or as when the lowest slip strengths
from both sides of the connections are exceeded (Assumption D).
These two alternatives relate to the ability of the connection to
withstand the eccentricities incurred when one slip surface fails
before the other. If the connection is capable of supporting these
eccentricities (e.g., has thick splice plates in the configuration of the
main specimens), then the slip strength is not realized until the slip
resistance of both sides is reached. If the connection is not capable
of supporting those eccentricities (i.e., has thin splice plates), then
movement occurs on one surface when the lowest slip resistance
of either side is reached.
It is reasonable for these tests to assume that the measured
slip coefficient from the test is actually the average of the slip
coefficients of the two surfaces, rather than the lowest value of
slip coefficient from the two surfaces (Assumption B). This was
consistently observed in the ancillary tests presented in [5,6], for
example, where the displacement measurements and observations
clearly showed that one surface did not typically fail prior to the
400 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406
(a) Load vs. top column displacement.
(c) Load vs. fill/column relative displacement.
Fig. 2. Displacement response of specimen 455h.
other. Following this assumption, the slip coefficient mean, mµ ,
and standard deviation, σµ , of one surface are:
mµ,one surface = mµ,two surfaces
√ (1)
σµ,one surface = 2σµ,two surfaces
For blast-cleaned surfaces the mean slip coefficient is 0.525
with a standard deviation of 0.101 [8]. These values are based
predominantly on experiments in which the slip strength is the
average of the slip strength of the two surfaces being tested. Thus,
the mean and standard deviation of any one slip surface is given by
Eq. (2):
mµ,one surface = 0.525
√ (2)
σµ,one surface = 2 0.101 = 0.143
The slip coefficient, as a random variable, is assumed to follow
a normal distribution. If one further assumes that the clamping
force is deterministic, the slip strength will also follow a normal
distribution. There exists variation in the bolt pretension that
provides the clamping force. However, statistical data derived
from Grondin et al. [8] indicate that the uncertainly in the slip
coefficient is significantly greater. Also, the statistical data on the
average clamping force depend on the bolt type and method of
pretensioning. Thus, the variable effect of bolt pretension is not
addressed in this analysis. Note that for this analysis, neglecting
the uncertainty in the average bolt pretension results in higher
expected slip strengths. The concept of order statistics [10,11] can
be used to determine the statistical properties of the lowest value
of a set of randomly varying quantities. In the simplest case of
undeveloped fillers having the same surface preparation as the
connected elements, the statistical data are the same for each slip
(b) Load vs. splice plate (middle) displacement.
(d) Load vs. splice/fill relative displacement.
surface. The cumulative distribution function (CDF) of the lowest
slip strength of multiple slip surfaces in this case is written in terms
of the CDF of a single slip surface (FS ,one surface (x)) and the number
of slip surfaces, n, Eq. (3):
FS ,one side (x) = 1 − [1 − FS ,one surface (x)]n (3)
For example, with a single-ply filler, the number of slip surfaces per
side, n, is 2. Eq. (3) is used to find the CDF of the lowest of the two
slip strengths on one side, based on using the values obtained from
Eq. (2). It is noted that should the surface preparation of the fillers
be different than that of the connected elements, this analysis
would need to be altered and changes in slip strength are possible.
From the CDF, the probability distribution, mean (mS ,one side ), and
standard deviation (σS ,one side ) of the lowest slip strength on each
side can then be determined using standard statistical approaches.
For the purposes of this study, failure of the primary connection
test specimens is defined as the sum of the lowest slip strengths
from each side of the connection, recognizing that each side may
have multiple slip surfaces (Assumption D). This implies that
both sides of the connection fail at approximately the same load.
This assumption is consistent with observed behavior for most of
the specimens [6] (for the one exception, 159n-2ply1, slip first
occurred only on one surface, not on both sides). The average and
standard deviation of the sum of the lowest slip strengths from
each side are thus computed using Eq. (4):
mS ,connection = 2mS ,one side
√ (4)
σS ,connection = 2σS ,one side
This resulting mean is the expected connection slip strength.
The expected slip strength can be determined for any number
of plies in an undeveloped connection by changing n in Eq. (3).
The results of this process (using the published statistical data
 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406 401

(a) Load vs. fill plate strain. (b) Load vs. fill plate strain.

(c) Load vs. splice plate (below bolt row 6) strain. (d) Load vs. splice plate (below bolt row 4) strain.

(e) Load vs. splice plate (below bolt row 1) strain. (f) Load vs. inside face of splice plate strain.

Fig. 3. Strain response of specimen 455h.

Table 2
Statistical multi-ply slip strength reduction.

Number of Number of surfaces per Expected strength, σS ,connection µeffective = expected strength/clamping Percent reduction from
plies side mS ,connection (kN) (kN) force deterministic (%)

0 1 6445 1290 0.525 0.0

1 2 5454 1028 0.444 15.4
2 3 4955 930 0.404 23.1
3 4 4635 872 0.377 28.1
4 5 4399 832 0.358 31.7

for the slip coefficient of Class B surfaces presented above) are
shown in Table 2. The percent reduction in the slip strength is
computed with respect to a deterministic analysis. The results are
also shown in Fig. 5, along with results from similar analyses using
different combinations of the assumptions for defining slip failure.
The choice of assumptions is important, as a range in calculated
results exists between the different combinations. In addition to
being consistent with experimental observations, the combination
of Assumptions B and C provides the same result as a deterministic
analysis for the case of no fillers.
402 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406

1.6
Lee & Fisher

1.4 Frank & Yura

Borello, Denavit,

Slip Test-to-Predicted Ratio

1.2 & Hajjar
Statistics Model
1.0

0.8

0.6

0.4

0.2

0.0
0 1 2 3
Number of Plies

Fig. 4. RCSC definition of slip load. Fig. 6. Undeveloped fillers slip strength test-to-predicted ratio vs. number of plies.

If the widths of the fill and connected element are the same, as they
60%
AssumptionA/C are for all tests considered, then thicknesses may be used:
AssumptionA/D
Percent Change in Expected Slip Strength

50% AssumptionB/C tfiller

AssumptionB/D
Ru,development = Ru,connection (6)
tfiller + tconnected element
40%
Defining the ratio of the thickness of the filler to the thickness of
30% the connected element as ρ(=tfiller /tconnected element ) and assuming
that all bolts have the same strength, the number of bolts required
20% to fully develop the filler can be determined using Eq. (7):
ρ
10% Nb,development = Nb,connection (7)
1+ρ
0% These additional bolts are intended to allow for a uniform
distribution of stress across the combined cross section, but
-10% they also increase the clamping force between the filler and the
connected element. This added clamping force can be modeled in
-20%
0 1 2 3 4
the statistical model by modifying Eq. (3) as Eq. (8):
Number of Plies FS ,one side (x) = 1 − [1 − FS ,undeveloped (x)][1 − FS ,developed (x)]n (8)

Fig. 5. Comparison of assumptions in statistical analysis of slip strength.
The addition of fillers reduces the expected slip strength
depending of the number of slip surfaces. The strength reduction
is plotted against test-to-predicted ratios for connections with
undeveloped fillers from Lee and Fisher [1], Frank and Yura [2], and
this study [5,6] in Fig. 6. The statistical model identifies the trends
well, although there is significant scatter in the data.
2.1. Developed fillers
Developed fillers extend beyond the splice plates and are
connected with additional fasteners. Providing additional fasteners
allows the filler plate to act more integrally with the connected
element. Unlike undeveloped fillers, the statistical properties of
each slip surface in developed fillers are not the same since there
are a greater number of bolts between the filler and connected
element. However, the surface preparation of all faying surfaces
is still assumed to be the same. The number of bolts required to
fully develop a filler can be calculated as follows: full development
is achieved when the strength of the connection of the filler
extension to the connected element is sufficient to uniformly
distribute the total force across the combined cross section,
satisfying Eq. (5):
Afiller
Ru,development =
Afiller + Aconnected element
where n is the number of plies of developed filler and the mean and
standard deviation of the CDF for each developed surface reflect the
increased clamping force, Eq. (9):
1 + 2ρ
mS ,developed = mµ,one surface Nb Tb,avg
1+ρ
(9)
1 + 2ρ
σS ,developed = σµ,one surface Nb Tb,avg
1+ρ
Fig. 7 shows the results of statistical analyses with one and
two ply developed fillers of varying thicknesses. The results of
statistical analyses with one and two ply undeveloped fillers are
shown as horizontal lines, since these values do not change with
filler thickness. For filler thicknesses approaching zero, the percent
reduction of expected slip strength approaches the expected
strength of an undeveloped filler, since very few additional bolts
are required and hence the added clamping force is very little.
As the filler thickness becomes very large, the percent reduction
of expected slip strength becomes relatively small. For these
connections, although the number of added bolts is substantial,
there is still a statistical possibility that slip will occur on the
developed surface, thus the percent reduction does not reach zero.
The statistical model indicates that the added clamping force
of the development bolts reduces the possibility of slip on the
developed slip surface and thus increases the expected strength of
developed connections. The experimental data presented in [5,6]
Ru,connection (5) provide mixed results for supporting these trends. The data

30%
Percent Change in Expected Slip Strength
20%
10%
0%
-10%
-20%
-30%
0 0.5 1 1.5
ρ =tfill /tconnected element
Fig. 7. Percentage change of expected slip strength for connections with
undeveloped fillers, developed fillers, or extended connections.
in Table 1 provide evidence that development is beneficial: (1) both
sides of specimen 159f slipped between the top column and
the filler plate at a much higher load than between the filler
plate and the splice plate; (2) one side of specimen 455f slipped
between the top column and the filler plate at a higher load
than between the filler plate and the splice plate; and (3) the six
developed specimens (159f, 159h, 455f, 455h, 159f-weld, 159h-
weld) consistently slipped first along the undeveloped slip surface
between the filler plate and the splice plate, whereas specimens
with no development sometimes failed first along the slip surfaces
between the filler plate and top column (e.g., specimen 159n1). In
addition, as the number of bolts developing the fill decreases, the
evidence that these bolts mitigate slip on the surface between the
filler and the top column decreases. This is seen in a comparison
of the slip order and loads for specimens 159f versus 159h, 455f,
and 455h. This indicates that if one is developing a thin filler,
the few additional bolts help very little to mitigate slip along the
surface between the filler and the top column. It is also likely
that once there is slip in the bolts on the surface between the
filler and the splice plate, the bolt pretension is diminished and
thus slip along the other slip surfaces is more likely to follow
immediately. However: (1) specimens 159n1 and 159n2 slipped at
higher loads than specimen 159h and than the first slip occurrence
in specimens 159f, 159f-weld, and 159h-weld; and (2) specimens
455n1 and 455n2 slipped at higher loads than specimens 455h
and 455f. Nevertheless, these results comparing the slip loads
between developed and undeveloped specimens could be due to
the inherently uncertain nature of slip. For example, due to some
randomness in the bolt pretension despite the use of control bolts
on one side of each connection.
Instead of having distinct development bolts, it is permissible
to extend the connection to accommodate the required number
of development bolts [3]. As a conservative action, it is expected
that extending the joint will result in a larger expected strength
for all cases, since all surfaces benefit from the additional clamping
force. Fig. 7 shows a comparison of percentage change in expected
strength between a connection with a developed filler and one
where the joint has been extended to accommodate the additional
development bolts. While the difference is modest for very thin
fillers, extending the joint provides a significant strength increase
above only developing the filler for thick fillers. The extended
joint exceeds the deterministic strength of the connection for thick
fillers, in that the percentage change is positive.
M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406 403
One Ply Undeveloped
One Ply Developed
One Ply Extended
Two Ply Undeveloped
Two Ply Developed
Two Ply Extended
2 2.5 3 3.5 4
Fig. 8. Undeveloped specimens: load vs. top column displacement.
3. Effect of fillers on the shear behavior
After the first slip event the connection continues to deform
while sustaining approximately the same load until shear and
bending are engaged in the bolt, seen as an increase in stiffness in
the displacement response (Fig. 2). The specimens were assembled
in negative bearing, allowing for the greatest slip deformation.
For a connection without fillers, this slip deformation is equal to
approximately twice the bolt hole tolerance or 16 mm (5/8 in.).
For specimen 730-over, the increase in stiffness is observed at
approximately this displacement (Fig. 8). For a connection with
thick fillers, the bending stiffness of the bolt within the filler is
relatively low. Deformation equal to approximately three times
the bolt hole tolerance or 24 mm (15/16 in.) is needed before
the connection stiffness significantly increases, such as is the case
for specimen 159n1 (Fig. 8). However, for a connection with thin
fills, the bending stiffness is significant and an increase in stiffness
is seen at a deformation level between two and three bolt hole
tolerances. This is the case for specimen 455n1 (Fig. 8). After
this increase in stiffness the bolts are primarily being engaged in
shear. As the load increased, specimen 455h produced repeated
pinging noises which increased in frequency until shear failure.
These noises were common among most specimens and are likely
associated with additional small slip events, bolts coming into
bearing with the bolt holes, or possibly initiation of fractures
within the bolts. Load continued to increase until shear failure,
softening was observed in both absolute and relative displacement
responses prior to failure, indicating ductile behavior.
During the period after initiation of slip and prior to shear
failure, a number of bolts in several specimens prematurely failed
through the threads at the face of the nut. No correlation was
found with overall specimen performance and these premature
bolt failures, as the shank of the bolt often stayed in the hole.
Due to the bending of the bolts within the filler plate, the
presence of the filler decreased the stiffness and increased the
deformation during slip (Fig. 8) as compared to specimen 730-over
which did not have a filler. The displacement of the top column
at shear failure was approximately the same for specimens with
either thickness of filler. However, the stiffness in the latter portion
404 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406
730-over 159n1 455n1
Fig. 9. Specimen bolt deformation (730-over, 159n1, and 455n1).
of the loading was lower for the specimens with the 41 mm (1 5/8
in.) fillers.
The shear behavior of bolted filler connections can be charac-
terized in more depth by studying two counteracting phenomena
further. The first phenomenon is the bending in the bolt due to the
relative movement of the shear planes. In an idealized connection
without fillers, there is no bending of the bolts, simply pure shear
at the shear plane between the faying surfaces (see the deforma-
tion at the end of specimen 730-over in Fig. 9). With fillers, the
original faying surfaces are separated, which has the effect of pre-
venting a clear definition of the shear plane and adding bending to
the bolt (Fig. 9, which shows the difference in bolt bending in the
159 versus the 455 specimens). This phenomenon is documented
in prior filler research, notably Frank and Yura [2], who developed
a strength reduction formula for bolted connections with undevel-
oped fillers based on filler thickness.
The second phenomenon is the resistance to bending provided
by the filler plate as the bolt jams in the hole. Two situations
illustrate the extremes of this behavior (Fig. 10). The first situation
is a filler plate with a very large hole, such that no part of the
bolt is in contact with filler plate. In this case, the effect of the
filler plate is to maintain separation between the other plates
(e.g., splice plate and column). In addition to not being as stiff as
a connection without fillers, the bending in the bolt will result in
Fig. 10. Shear mechanism deformation modes.
a lower strength due to the interaction of shear and bending. The
second situation is a rigid filler plate with the hole precisely the
same size as the bolt, leaving no tolerance. In this case, the bolt will
be completely restrained from bending and the behavior will be
essentially the same as two uncoupled shear connections without
fillers and no reduction in strength will be observed.
This mechanism is consistent with experimental observations
of shear strength. Thin fillers allow bending in the bolt sufficient
to reduce the shear strength, while thick fillers only allow minimal
bending, not significantly reducing the shear strength. This is seen,
for example, in the residual plastic deformation in the in the bolt at
the shear plane between the top column and the filler plate (Fig. 9).
While the shear deformation is similar, the rotation is greater for
the bolt from the specimen with the thinner filler.
The two phenomena may be extended to the behavior of
multiple ply fillers. Consider two filler connections with the same
total filler thickness. The first is made of a thick filler and a thin
filler, and the second is made of several thin fillers. For the first
connection, the portion of the bolt inside the thick filler will be
restrained against bending while the portion of the bolt inside the
thin filler will not. The behavior of such a connection would be
much like two uncoupled shear connections (one with a filler).
For the second connection, since the filler plates are able to move
relative to each other, no part of the bolt will be restrained against
bolt bending and it will behave as though there were one filler with
a larger hole. As a result, if a small filler needs to be added for fit up
in the field, this model does not predict a large reduction in shear
strength since this situation would be more like the case of thick
filler and thin filler.
3.1. Developed fillers
As noted in [5,6] the shear strength results of the specimens
with a W14 × 455 top column indicate that a connection with
developed fillers does not suffer the same detrimental effects in
shear as one with undeveloped fillers. However, due to a lack
of experimental data, further direct indications are unavailable.
To examine the effect of development further, one can consider
an undeveloped or partially developed filler connection as fully
developed for a fewer number of bolts. For this analysis, the bolts
are considered to be separated into those that resist shear and
those that develop the filler. The total number of bolts remains
the same; however, using common terminology, there are fewer
bolts in the connection (those that resist shear). The number of
 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406
2.5
Frank & Yura
Borello, Denavit,
& Hajjar
Effectively Fully Developed Bolts
2.0
Shear Test-to-Predicted Ratio for
1.5
1.0
0.5
0.0
0.0 20.0 40.0 60.0
Filler Thickness (mm)
Fig. 11. Shear strength test-to-predicted ratio vs. fill thickness for effectively
developed bolts.
bolts that are assumed to resist shear (those ‘‘in the connection’’
or ‘‘effectively fully developed’’) are less than the total number of
bolts by the number of bolts required to fully develop the bolts
that resist shear. This is analogous to the third option provided to
designers in Section J5 of [3], where the size of the joint may be
extended to accommodate the number of bolts required to develop
the filler.
Defining %EFD as the percentage of bolts in the undeveloped,
partially developed, or fully developed filler connection that are
considered to resist shear when the connection is considered
fully developed, i.e., effectively fully developed, from Eq. (6) the
following is obtained:
(1 − %EFD)Ru,connection + Ru,development,provided
ρ However, as the bolt jams within the bolt hole as the thickness
= %EFD Ru,connection
1+ρ
The first term on the left side of Eq. (10) represents the portion of
the total number of bolts assumed to be development bolts. The
second term on the left side represents the strength of the bolts
provided to explicitly develop the filler. This term is for specimens
with bolts, or other means of development, between the filler
extension and connected element. Solving for %EFD,
1+ρ
 
Ru,development,provided
%EFD = 1+
1 + 2ρ Ru,connection
The limits of this equation are logical: when ρ approaches zero,
i.e., very thin fillers, nearly no bolts are required to develop the
filler, i.e., %EFD approaches 1 for undeveloped fillers. When ρ
approaches infinity, i.e., very thick fillers, the same number of
bolts is required to develop the filler as in the joint, i.e., %EFD
approaches 0.5 for undeveloped fillers. %EFD has been defined
such that multiplying the predicted slip and shear strengths
of the undeveloped, partially developed, or fully developed
filler connections by %EFD would result in predicted values
of the effectively fully developed connection. The strength is
reduced due to the allocation of some of the connection strength
towards connection development. Test-to-predicted ratios using
predictions based on effectively fully developed strengths are
plotted in Fig. 11.
The shear strength test-to-predicted ratios based on the
effectively fully developed prediction, shown in Fig. 11, indicate
that if the connection is fully developed by extending the joint,
the test-to-predicted ratio is always very near or greater than
unity. This indicates that the detrimental effects of fillers on shear
405
strength are mitigated by developing the filler and extending
the connection to include the development bolts. The shear
strength test-to-predicted ratios increase with filler thickness
in Fig. 11 because as the filler thickness is increased, the
number of development bolts increases. These bolts offer some
shear resistance that is conservatively neglected in the predicted
strength.
4. Conclusions
Experimental research has been conducted on steel bolted
connections with fillers. Through a detailed description of the
experimental observations of such connections, the various
mechanisms influencing behavior were illustrated. The effect of
fillers on the slip strength and shear strength was discussed.
Further, the effect of multiple plies, developed fillers, and
80.0 100.0 effectively developed extended connections has been discussed.
A stochastic approach was developed to quantify the detrimen-
tal effect of additional slip surfaces on the slip strength of the con-
nection. Key findings from this model are:
• The reduction in slip strength increases with increasing number
of plies.
• Developing the filler negates some, but not all, of these detri-
mental effects depending on how many additional develop-
ment bolts are required.
• Extending the connection to accommodate the development
bolts can completely negate the detrimental effects of addi-
tional slip surfaces, depending on how many additional devel-
opment bolts are required.
Through the counteracting mechanisms of bolt bending and
jamming within the bolt hole, the detrimental effects of fillers
on shear strength were described. Noting these mechanisms,
experimental data were analyzed to determine that:
• For thin fillers, the connection shear strength reduces as a
function of the thickness of the filler due to bolt bending.
(10)
of the filler increases, the reduction reaches a minimum and
eventually vanishes.
• Multiple ply fillers delay the jamming of the bolt within the
hole, thus adding to the detrimental effects.
• Developing the filler and extending the connection to accom-
modate the development bolts were both shown to be viable
solutions to the detrimental effects on the shear strength.
Acknowledgements
(11)
This work was supported by the American Institute of Steel
Construction, W&W Steel Corporation, and the University of
Illinois at Urbana-Champaign. In-kind funding was provided by
Lohr Structural Fasteners. The authors thank G.A. Rassati and
J. Swanson of the University of Cincinnati for conducting the
ancillary bolt shear and tension tests; Prof. P. Dusicka of Oregon
State University and Prof. G. Grondin of the University of Alberta for
sharing data related to their research on bolted steel connections;
Prof. J. Philips, Director of the 3,000,000 lb Testing Machine at
the University of Illinois at Urbana-Champaign, for his extensive
contributions to this research; Prof. J. Song of the University of
Illinois at Urbana-Champaign for his contributions to the statistical
assessment of slip strength; and K. Elam, D. Foley, G. Banas,
T. Prunkard, M. Bingham, and M. Parkolap of the University of
Illinois at Urbana-Champaign for their pivotal contributions to
the execution of the experiments. The authors also thank the
members of the AISC Technical Advisory Panel (TAP) for their
excellent contributions to this research, including T. Schlafly (TAP
Chair and AISC Director of Research), C.J. Carter, and C.J. Duncan,
AISC; S. Armbrust, T. Winneberger, and W. Lindley, W&W Steel
406 M.D. Denavit et al. / Journal of Constructional Steel Research 67 (2011) 398–406
Corporation; L.S. Muir and W.A. Thornton, Cives Engineering
Corporation; L.A. Kloiber, LeJeune Steel Company; Prof. G. Grondin,
University of Alberta; G. Heathcock, FabArc Steel Supply, Inc.; and
J.M. Fisher, Computerized Structural Design, S.C.
References
[1] Lee JH, Fisher JW. Bolted joints with rectangular of circular fillers. Report
No. 318.6. Bethlehem (PA): Fritz Engineering Laboratory, Department of Civil
Engineering, Lehigh University; June 1968.
[2] Frank KH, Yura JA. An experimental study of bolted shear connections, Report
No. FHWA/RD-81/148. Federal Highway Administration. Washington (DC): US
Department of Transportation; December 1981.
[3] American Institute of Steel Construction (AISC). ANSI/AISC360–05: specifica-
tion for structural steel buildings. Chicago (IL): AISC; 2005.
[4] Dusicka P, Lewis G. High strength steel bolted connections with filler plates.
Journal of Constructional Steel Research 2010;66:75–84.
[5] Borello DJ, Denavit MD, Hajjar JF. Behavior of bolted steel slip-critical
connections with fillers. Report No. NSEL–017. Newmark structural laboratory
report series. Urbana (IL): Department of Civil and Enviromental Engineering,
University of Illinois at Urbana-Champaign; 2009.
[6] Borello DJ, Denavit MD, Hajjar JF. Bolted steel slip-critical connections with
fillers: I. Performance. Journal of Constructional Steel Research 2011;67(3):
379–88.
[7] Kulak GL, Fisher JW, Struik JHA. Guide to design criteria for bolted and riveted
joints. 2nd ed. New York (NY): John Wiley & Sons; 2001.
[8] Grondin G, Jin M, Josi G. Slip critical bolted connection — a reliability
analysis for design at the ultimate limit state. Edmonton (Alberta, Canada):
The American Institute of Steel Construction, Department of Civil and
Environmental Engineering, University of Alberta, April 2008.
[9] American Association of State Highway and Transportation Officials
(AASHTO). LRFD Bridge Design Specifications with interim revisions, AASHTO.
Washington (DC). 2008.
[10] David HA. Order statistics. New York (NY): Wiley; 1970.
[11] Song J, Der Kiureghian A. Bounds on system reliability by linear programming.
Journal of Engineering Mechanics 2003;129(6):627–36.