Connections: Art, Science, and Information
in
the Quest for Economy
and
Safely
DR. WILLIAM A. THORNTON
INTRODUCTION Connections are an extremely important part of the final configuration of
a
steel structure. Many, if not most, collapses are caused by inadequate connections. The constructed cost of
a
steel frame is heavily dependent on the connections used, both the type of connection and how they are configured. Yet, connections are often an afterthought. Commercially available software will pretty much automatically design the members of the frame, but there is no commercially available software that will do the same for the connections. In fact, the frame design software chooses optimal members, usually least weight, with no regard whatsoever as to how these members will impact on the connections. The emphasis in engineering schools is similar to that of commercially available design software, i.e., it is on the design of members. Very little work is done on connections at the undergraduate level and probably also at the graduate level. Connections are considered by many professors as essentially trivial in a mathematical sense. Very sophisticated and mathematically elegant solutions can be prescribed for member and frame design; e.g., lateral torsional buckling of members, and the direct stiffness method for frames. Connec
tions,
on the other
hand,
are thought
to
be designed
by no
more sophisticated analysis than counting bolts and determining weld lengths. This is not true except for the simplest connec
tions.
While there are essentially only three types of members in a structural frame (beams, columns, and beamcolumns), there is an almost infinite variety of connections depending on frame geometry. For this reason, connection design is actually more interesting than member design, because this great variety often requires the designer to rely on intuition and art as well as science. As mentioned above, connections are often an afterthought. In many engineering offices, once the frame is designed and on paper, the drawings are ready to be released for construction. Connections are handled by a series of typical details and general notes which refer to AISC manuals. Typical notes for shear connections might make reference to full depth connections or the Uniform Design
Dr. William A. Thornton is chief engineer, Cives Steel Co. and president, Cives Engineering Corp., Roswell, GA.
Load (UDL): For moment connections, the notes might say use stiffeners and doublers as required, and design for the strength of the beams. For bracing connections, a typical detail might be shown with a statement to design for all eccentricities. It is the primary purpose of this paper to show by anecdotal examples, i.e., war stories, that this approach to connections can be both uneconomical and unsafe. A secondary purpose is to give the motivation behind a new method of bracing design called the Uniform Force Method. Examples will be given of shear connections, bracing connec
tions,
and moment connections. SHEAR CONNECTIONS Shear connections
are a
subject where information
is
the primary quantity lacking on most jobs. These connections have been heavily studied over the years, and other than questions regarding ductility and robustness, are well understood. Shear connections are the most common connection on all
jobs.
Ideally, the engineer should give the shear for every beam end. While this may appear to be a lot of extra work, it is not as difficult as it first seems since the loads are known from sizing the beams. Why not put them on the drawing? (In addition to helping the fabricator, having the loads used in the srcinal design right on the drawing is very handy for future renovations.) If the loads are shown for every beam end, there is very little room for error, and the connections will be as economical as possible. However, instead of actual
loads,
most jobs these days have one or more of the statements regarding shear connections: ã Item
1.
All shear connections shall contain the maximum possible number of rows of bolts; ã Item 2. Design all shear connections for onehalf UDL; ã Item
3.
Design all shear connections for the shear capacity of the beam; ã Item
4.
Minimum design loads for standard rolled shapes, unless noted otherwise: W8 C8 . . W10C10 W12C12 W14C15 W16 . . . W18 . . . 10 kips W21 65 kips 15 kips W24 75 kips 25 kips W27 90 kips 35 kips W30 125 kips 45 kips W33 140 kips 55 kips W36 175 kips
132 ENGINEERING JOURNAL / FOURTH QUARTER /1995
Let's consider each of these.
Iteml
Item
1
requires full depth connections. The fabricator assumes the engineer has reviewed his design and the capacity of these connections will exceed the actual loads in all cases. But in many cases, these will be uneconomical, as with long span beams. In other cases, they may be unsafe. Suppose a beam has a large cope, as when connecting a small beam to a large one (see Figure 1). This may greatly reduce the capacity of the full depth connection because of the reduced beam section. Has the engineer considered this, or has he reviewed his drawings by checking the actual load against the capacity of
a
full depth connection on an uncoped beam? It is very likely that he has done the latter. As a second example, consider steel at different elevations. Figure 2 shows a full depth connection for the upset W 18x35. The capacity of this full depth connection is 20k whereas a true full depth connection for the W 18x35 (Figure 3) is 49k. Will the engineer realize this if he specifies use the maximum possible number of rows ?
Item 2
If infill beams frame near the ends of
a
main beam, the UDL method can be unsafe. If beams are short, it will be uneconomic. Figure 4 shows a floor framing plan. All beam shear connections are contractually required to be designed for onehalf UDL. The three W10x22s framing between the W36xl70 and the W36x230 are 3ft. long. The onehalf UDL reaction is 61.8k Of course, this is ridiculous—but the fabricator is contractually obliged to supply it if the engineer insists, and we have done jobs where the engineer did just that. Figure 5 shows the resulting connection. Note that the shear capacity of a W10x22 is only 35.4 kips, so designing for 61.8 kips is doubly ridiculous and leads to a discussion of Item 3.
Item 3
Item 3 requires the connection to develop the shear capacity of the beam, but this is impossible with the usual shear connections (single clips, double clips, shear end plate, shear tab) unless the beam is haunched or web doublers are used. Also, since most beams are coped, just what is the shear capacity of the beam? Is it the uncoped capacity (35.4 kips for the W 10x22 shown previously) or should the capacity of what is left be used? It's clear that Item
3
is ambiguous, which can lead to errors affecting safety as well as result in ridiculous designs. In Figure 6, the W 10x22 of Figure 4 has end connections good for 35.4 kips, which means the W10x22 is capable of supporting 35.4 tons Obviously, these W10x22s
are
just intended to reduce the unbraced length or provide decking support. If a real load of 35.4 tons must be carried, a short W 18x35 with five rows of bolts would be cheaper and safer.
Item 4
While at first glance, Item 4 appears to be innocuous, try to develop 15k in the W 10x22 shown previously. Figure 7 results.
Single Angles and UDL
The uniform design load UDL is
a
great crutch of the engineer because it allows him to issue design drawings without putting the beam reactions on the drawings. Instead, often the fabricator is told to design the beam end connections for onehalf UDL, or some other percentage to account for composite design, unless greater reactions are shown. Unless concentrated loads are located very near the beam ends, UDL reactions are generally very conservative. Because the reactions are too large, extremely strong connections, such as double framing angles, will often be required. Single angles, because the bolts are in single shear, will have about half the strength of double clips for the same number of rows of bolts. But if actual reactions are given, it
CARRYING DEAM
\
r
INFILL BEAM
jr— 1J
jjr#]
Hr
I
Hr
I
nr
l
COPE 7% X 7% CAPACITY = 20 KIPS
W18 x
35
1
BOLTS A325N
>U
0
W30
x 173
Fig. 1. Usual beam to beam connection top of steel at common elevation. Fig. 2. Full depth
connection for upset beam.
ENGINEERING JOURNAL/FOURTH QUARTER/ 1995 133
will almost always be found that a single angle connection will work, perhaps with a couple of extra rows of bolts. Figure 8 is part of an industrial building with dead load of
140
psf and live load of 250 psf. Beam 1, of Figure
8
is shown in Figure 9. The total load on Beam
1
is 82 kips and the actual reactions are thus 41 kips. The onehalf UDL reaction is 45 kips, which is pretty close. Now look at the connections. The minimum double clip connection on this coped beam
has
four rows and is good for
81
kips, almost twice the actual reaction. Many designers routinely require full depth connections, i.e. six rows. The six row double dip connection is good for 166 kips, almost three times the actual reaction. However, a five row single angle is good for 52 kips, which is okay for the actual and the onehalf UDL reactions. As this example illustrates, single angles will work even in heavy industrial applications, and they are much less expensive than double clips, especially for erection. In Figure 10, the connections for this W24x55 beam have the same strength and have a differential cost of
10
for fabrication. But, including erection, the single angle beam costs approximately $25 less than the double clip beam. For a 30story building, 200 ft.x200 ft. with 25ft. bays and
200 beams
per floor with single angles, there is a savings of 200 x 30 x 25 = $150,000. Returning to Figure 8, suppose Beam
1
is subjected to the same load of 82 kips total, but 32 of the 82 is a concentrated load located at midspan, such as from a vessel. Figure 11 shows the actual reaction of the beam, now a W24x76, is still 41 kips, while the onehalf UDL reaction is 56 kips—which is 37 percent greater than the actual reaction. This means while a five row single angle connection
is
okay for the actual reaction, a six row connection with a capacity of 66 kips would be required for the onehalf UDL reaction. Figure 12 shows the disparity between actual and onehalf UDL reactions for Beam 2. Again, single angles are sufficient.
BRACING CONNECTIONS
Bracing connections are subject where the art and science of connection design can be used to achieve a safe and efficient design. They are also an area where missing or misleading information can lead to drastically unsafe connections or connections which are grossly overdesigned.
COPE l
3
'« X 7
3
'4
r
W18 x 35 CAPACITY = 49 KIPS
/
W30 x 173 BOLTS A325N
3
'4 0
COPE 2 x
6
l
U
COPE
2\
x 8''i
Fig. 3. Normal full depth connection. Fig. 5. Section AA of Figure 4 W 10x22 to carry
V2
UDL
=
61.8 kips reaction each end.
ENGINEERS NOTE: DESIGN BEAM END CONNECTIONS FOR AXIAL FORCES SHOWN ON PLANS 246 ^, 246
COPE 2 x 6
1
' COPE 2% x 8''i
W12«96
1*8$*)
i
'
p
\
L
£3«3>'
2
(*3S
K
)
TYP THIS BAT
D
WT6 I8&75
K
UN0)TYPTHIS BAY
Fig. 4. Ambiguous forces for connection design.
1 \'3C> x 170
]
r*
r*
=U li=
W10 x 22
—/\ A
^ / \
'J
PL
x
ã 1
ã 4 ã 4 ã 4 ã j
[ DOLT W36
r
1 1 .vu I I
p
Fig. 6. Section AA of Figure 4 W10x22 to carry maximum uncoped shear capacity of 35.4 kips as reactions at each end.
134 ENGINEERING JOURNAL / FOURTH QUARTER / 1995
Art and Science
Figure
13
shows
a
bracing connection design method which satisfies all of the requirements
for
equilibrium for
the
gusset, the beam,
and the
column.
It
includes consideration
of all
eccentricities and
it
is simple
to
use because all forces acting between
the
gusset
and the
beam
and
column
are
known before the size
of
the gusset is known.
It
has been referred
to
as
the
KISS method
by
a
detailer who
was
impressing upon his people the necessity
of
getting the shop drawings out the door. Thus,
the
sarcastic comment
to
Keep
It
Simple, Stupid ,
or use the
KISS method. Unfortunately, while this method
is a
boon
to the
detailer,
it
is a
bane
to
the fabricator and
owner.
It results in large and expensive
connections.
Also, the engineer
and
owner
do not
like
it
when,
if
there
are
four gussets
in a
building panel, they almost meet
at
the
center. Also,
the
load paths through this gusset, beam,
and
column are very unnatural
and
inefficient as will
be
shown. Beginning about 15 years ago, AISC began to address this problem with a research program
at the
University of Arizona. This program resulted
in
published work
by
Richard (1986) which contained figures similar
to
Figure 14.
In
this Figure, the resultant forces
on the
gusset edges
for
a
wide variety
of
gusset edge support conditions
are
seen
to
fall within
the
envelopes shown. The edge resultants appear
to
intersect with the line
of
action
of
the brace
at a
point
on
this line
on the
other side
of
the working point (WP) from
the
gusset. Note that
no
couples were required
in
Figure
14.
This data from Richard
is
the
genesis
of
the
author's development
of
what has come
to be
called
the
Uniform Force Method (Thornton
1991,
1992
and
AISC
1992,
1994).
The
method
is
shown
in
Figure
15A.
Figure
15B
shows
a
force distribution which captures
the
essence
of
the
distributions given fuzzily
in
Figure
14.
In
other words,
a
force structure
is
imposed
on
Richard's data.
In
order
to
test
the
efficacy
of
this structure, the data of six full scale tests were filtered through
it.
The tests were performed by Chakrabarti and Bjorhovde, (1983,1985) and Gross
and
Cheok (1988, 1990). Typical test specimens are shown
in
Figures
16 and
17.
The
limit states considered in
the
filtering process
are
given
in
Tables
1 and 2.
Table
3
shows the results. For the Chakrabarti/Bjorhovde tests, excellent agreement
is
achieved.
The
ratio
of
test capacity
to
predicted capacity
is
close
to but
slightly larger than unity
as
COPE 2 x
6
U
COPE
2\ x B
l
't
A
W36 x 170
fe
ã*
BOLTS A32SN> 4 0
*\
W36 x 230
3*0
4
,— 3.25 Wft W21x68
v
/
~r
i i
.i i r
V
250
BEAM
1
SeoiON
UZ)*66 Lc*o»
IWoftM
^FT
3Z3
COMC.
hA o
&TAL fclFS
82
RlACTlCNi
ACTUAL
bps
41 fcUDL
ke5
i5
C.o**cncki5
DOUBLE. CLIPS
MIKI ROMS
Of
(tori)
Rows
CAP
i

81
 6  M6
SINGLE
CUP
CAP
ãb

S2
BOLTS
7
8
0 A
325N,
CUPS
4 X 3»
2
X 3
8
Fig.
7.
Section AA of Figure
4
W10x22
to
carry 15 kip reaction each end. Fig.
9.
Comparisons for beam 1 of Figure 8
—
uniform
load.
(EH
W33X1I9
JL±.
W33xitg g4
i
g4
SINGLE CLIPS DOUBLE CLIPS
U24>55
T
rvows
5AME. STKENGTM
4 ^ows
FABRICATION
 $io
PER BEAM LESS FOR SINGLE CUPS
ERECTION

$ 15 PER BEAM LESS FOR SINGLE CUPS TOTAL COST REDUCTION
$25 PER
BEAM USING SINGLE CLIPS
Fig.
8.
Partial plan of industrial building floor. Fig. 10. Cost comparison same strength single and double clips.
ENGINEERING JOURNAL
/
FOURTH QUARTER
/
1995
135