Semi-Rigid Frame Design Methods

for Practicing Engineers

JOHN E. CHRISTOPHER and REIDAR BJORHOVDE

ABSTRACT connections in simple frames is negligible. Thus, while all

Design of semi-rigid (PR) frames focuses on behavior semi-rigid frames are considered as PR construction, not
characteristics of non-linear connections, including their all PR construction exhibit semi-rigid behavior. This dis-
substantially different loading and unloading characteris- tinction becomes evident from the study of the connection
tics. Moment-rotation connection representations, such as moment-rotation relationships.
the three-parameter power model, facilitate the calculation Significant benefits can be gained by incorporating con-
of stiffness data required for frame analysis. In this pa- nection behavior into the design of partially restrained
per, the connection characteristics are described in terms (PR) building frames when the connections have semi-
of linearized connection stiffnesses that are calculated on rigid properties. These benefits offer the potential for re-
the basis of expected connection loads. This allows for the ducing costs. The economy of semi-rigid frames is the
use of first-order analysis to determine structural stabil- result of reduced beam moments, reduced beam deflec-
ity, serviceability and member load effects. The design tions, and increased restraint at column ends. The effect
method detailed in this paper includes the concurrent se- is most pronounced in braced (sway-prevented) frames,
lection of connections and member sizes. The LRFD ap- where requirements for beam and column sections are re-
proach of AISC is utilized, including the provisions that duced as a result of the more realistic connection behavior.
rely on amplification factors to account for second-order The case for unbraced (sway-permitted) frames is not as
effects. Member section checks are made with unbraced clear, because the increased connection flexibility leads to
length K-factors determined from the alignment charts, increased drift and reduced column restraint. This can re-
using modified relative distribution factors to account for sult in higher material costs, but there is still a potential
connection flexibility. New connection limit states are for improved economy. This is particularly due to the fact
quantified by the useful range of expected connection that semi-rigid connections utilize simpler details which
deformation. are more economical to fabricate and erect.
The difficulty in designing semi-rigid frames lies pri-
marily in the non-linear connection behavior, where the
1.0 INTRODUCTION
stiffness depends on the connection deformation. Signifi-
The AISC Specification1 identifies two construction types: cant progress has been made in solving this problem with
FR and PR, which refer to the degree of restraint that the the introduction of advanced methods of frame analysis. 713
connections offer in structural steel frames, namely fully However, these methods require the use of computer pro-
restrained and partially restrained, respectively. While FR grams that are not yet suitable for office use, nor are they
construction represents building frames with "rigid" con- readily available. Also, at this time Chen's methods do
nections, PR construction represents frames with simple not include the unloading behavior of the connections,
"pinned" connections as well as semi-rigid connections. which is radically different from their loading behavior.
The distinction between these two types of PR construc- A method based on using expected connection loads to
tion (simple and semi-rigid) is that the connections in a estimate equivalent linearized connection stiffnesses is pre-
semi-rigid frame are able to develop moments and rota- sented here to provide design and analysis techniques for
tional deformations that significantly affect the member semi-rigid frames. Using the estimated linearized connec-
forces and displacements, while the moment response of tion stiffness values allows the use of simple, first-order
analysis techniques, which are familiar to practicing engi-
neers. The approach also allows the use of computer soft-
John E. Christopher is a structural engineer with STV, Incor- ware that is based on linear-elastic elements, and currently
porated, Pittsburgh, PA.
in common use and familiar to most design offices.
Reidar Bjorhovde is President of The Bjorhovde Group, Tuc- A second major problem associated with designing
son, AZ. semi-rigid frames is how to quantify the strength limit

12 ENGINEERING JOURNAL/ FIRST QUARTER / 1999

state, or connection moment capacity. Tests 18,24 have points. However, some connections exhibit M-<f> re-
demonstrated that the ultimate connection moment is of- lationships that make a unique definition of Rki very
ten well beyond the useful range of expected connection difficult.
deformation, far exceeding serviceability limits. A new • Secant stiffness, Rks—the effective stiffness of the con-
definition of connection limit states that is related to ser- nection based on an expected moment. R^ = Mel<f>e,
viceability requirements is presented in this paper as a ra- where Me and <f>e are illustrated in Figure 1.
tional approach to this issue. The limit states are based on • Tangent stiffness, R^—the instantaneous stiffness,
a practical expected limit to connection deformation. which usually decreases as the moment increases.
With these problems addressed and satisfied, it is possi- Such behavior does not apply for the full range of
ble to use the AISC LRFD format, following the approxi- M-<f> response when "work hardening" affects the
mate method that relies on amplification factors to account connections.6
for second-order effects. This method is particularly suited • Unloading stiffness (reversing loads), Rku—the un-
to the need to account for the abrupt change in connection loading rate that is approximately linear until reach-
stiffness that is associated with load reversal. ing zero moment. It is the result of subsequent,
In part, since North American design codes address the superimposed loads that produce rotations opposite in
use of semi-rigid connections in only a limited way, design direction to the initial response. This phenomenon has
engineers have been reluctant to use them. However, with been observed by Popov and Pinkney28 and others; the
continued research, code criteria for semi-rigid frames are unloading stiffness is commonly assumed to equal the
likely to become more detailed. Although more research is initial stiffness, R^.
required to explore fully the behavior of semi-rigid frames,
Figure 1 also illustrates the dependence of the secant stiff-
enough information is available now to implement the tech-
ness, Rks, and the tangent stiffness, /?&, on the expected
nology in normal design practice.
connection moment and deformation, as represented by
point a.
2.0 CHARACTERISTICS OF SEMI-RIGID
CONNECTIONS 2.1 Connection Models
The databases of beam-to-column connection tests 18,24 in- Since it is not practical to depend on the availability of
dicate a generalized, non-linear behavior as illustrated test data for a myriad of connection details, analytical
in Figure 1, where point a represents an expected con- models are a convenient substitute. In an early attempt to
nection moment, Me, and corresponding deformation, <f)e. resolve this problem, Kennedy22 derived a polynomial ex-
This particular M-<f> curve has been constructed from test pression for the flexible end-plate connection. Later, Frye
data for a flush end-plate beam-to-column connection.24 and Morris16 presented a polynomial model to represent the
The illustration indicates the key connection stiffness loading behavior of a variety of connection types. Although
parameters: suitable for many connections, the problem with the Frye
and Morris model is that it gives negative stiffness val-
• Initial stiffness, R^—the initial slope of the M-</> curve,
ues for certain connection types and ranges of M-</>. Since
which is nearly constant for the first few data
this is physically unacceptable, Jones et al. 20 achieved an
accurate correction, substituting a 5-spline curve to fit ex-
perimental M-<f) data.
.Initial Stiffness, Rki Kishi et al. 25,26 presented the three-parameter power
model to represent the M-</> behavior of connections uti-
Me ^ lizing bolted angle elements. The M-0 curve is defined by
Equation (1).
Tangent Stiffness, Rkt Rki<t>
M = \ln (1)
Unloading Stiffness, Rku = Rki
$0

Secant Stiffness, R^
where
1
0.0 10.0 20.0 30.0 40.0 50.0 60.0
R^ = initial stiffness of the connection
3
Rotation, <j>(10 rad)
n = shape factor
</>o = reference plastic rotation, calculated as <£o
Mu/Rki
Fig. 1. Moment vs. Rotation for a Flush End-Plate Connection24 Mu = ultimate moment capacity of the connection

ENGINEERING JOURNAL / FIRST QUARTER / 1999 13

 The tangent stiffness is the first derivative of Equa- 2. The connection moment is given as a non-dimensional
tion (1): ratio, compared to the fully plastic moment of the
beam:
Rki
Rkt — (2)
M = MIMn (4)
1 + (<*/<fc>)n
where
Thus, the loading portion of the M-</> curve for these M = connection moment
types of connections may be mathematically expressed in Mp = plastic moment of the beam
terms of three simple parameters, which may be calculated
from the material properties and detail dimensions by using 3. The non-dimensionalized connection deformation is
methods such as those presented by Chen et al.7 These given by Equation (5)
parameters are the initial stiffness, the ultimate moment 4 = MP (5)
capacity, and the shape factor for the M-<f> curve.
A key problem associated with the three-parameter where (f> = connection deformation, and <j>p is de-
power model is that for many connection details, the fined by
ultimate moment capacity, MU9 is often not practical to 5dMD
define as a strength limit state. This is because these con- <t>P = (6)
EI
nections do not exhibit a clearly defined plateau in the
M-(j> curve. It is the plastic bending rotation of the beam, illus-
trated in Figure 2(b).
2.2 Connection Classification The product (a* d) is termed the characteristic length, or
Beam-to-column connections are produced in many types the length of the beam that will have the same stiffness as
and sizes with a wide range of responses. This offers op- the initial connection stiffness. The value offivein Equa-
portunities for the designer to satisfy a variety of structural tion (6) was chosen because it represented the middle of the
needs and conditions. It also presents a problem in deter- semi-rigid range for reference length, as determined from
mining when a particular connection should be treated as a a review of the Kishi and Chen24 database.6
simple pin, because of its highflexibility;as a rigid connec- Eurocode 3 12 includes a connection classification system
tion, because of its high stiffness and strength; or as a semi- that requires evaluations to determine what frame behav-
rigid connection, because of a combination of stiffness and ior to consider in the analysis: simple, rigid or semi-rigid.
strength that differs significantly from "rigid" or "pinned." Identifying these frame behavior characteristics are similar
With the advent of the semi-rigid category in design prac- to the objectives of the nondimensional system presented
tice, there is a need for simple criteria to determine how in Figure 2(a), but the EC3 system can result in a different
stiff is "rigid," and howflexibleis "pinned." Bjorhovde et conclusion.
al.6 offered a simple test, illustrated in Figure 2(a), to clas-
sify the properties of a connection as "rigid," "semi-rigid,"
or "flexible." This is a non-dimensional system that uses
the strength and stiffness of the connection in comparison
to the beam properties. The illustration of Figure 2(a) in-
1 . 2 -I
dicates a connection in the semi-rigid category, where the
curve lies in the area between the "rigid" and "flexible" 1.0-
boundaries. The implementation of this system uses terms Rigid \ ^Simplified Ductility Requirement
0.8 -
that are defined as follows:
1. The non-dimensional characteristic length factor for Jo.6- a =2
/ S
the initial connection stiffness is given as 0.4-
\jS - Se mi-Rigid

EI 0.2-
Oii = (3)
Rkid Wy/^a = 10 | Flexible |

where 0 0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 3.3 3.6
E = modulus of elasticity
/ = moment of inertia of the beam
Rki = initial connection stiffness
d = beam depth Fig. 2(a). Non-Dimensional Classification of Connections6

14 ENGINEERING JOURNAL / FIRST QUARTER / 1999

 Using either the Bjorhovde system or the EC3 system where
offers the following benefits: Fy = the yield strength of the beam steel (ksi)
• A means to determine if semi-rigid design and analysis 81 h = the specified inter-story drift ratio
techniques are appropriate and necessary.
In Equation (8) the value of 0.0008^ reflects the simple
• The ability to assess the connection behavior in relation
beam end rotation resulting from a uniformly distributed
to member behavior.
load that causes a plastic moment to develop at mid-span.
• The opportunity to select semi-rigid connections con-
If 6R is the practical limit of beam end rotation, then
currently with member sizes during the design process.
a semi-rigid connection should be expected to have the
The term a, the general term for the effective charac- ability to deform safely throughout the range of OR while
teristic length factor, is also useful in calculating member resisting moment and shear. Thus, the connection defor-
forces and column stability, since it provides a way to quan- mation requirement, or in effect the ductility requirement,
tify the effective connectionflexibilityat various degrees of </>My, should equal 6R. Using Equation (8), connections sup-
connection loading. For example, the characteristic length porting steel beams with a yield stress of Fy = 36 ksi
factor associated with the connection secant stiffness is should expect to see a limit of about 0.03 radians, and with
given as Fy = 50 ksi, 0.040 radians of rotational deformation. As
a comparison to these numbers, Thornton31,32 assumed an
EI end rotation as 0.03 radians for a maximum expected value
<*c = (7)
Rksd for most cases, and considered 0.07 radians as very large.
Also, Deierlein13 noted that the moment at 6 = 0.02 ra-
2.3 Connection Limit States dians corresponds fairly well to the nominal connection
Examination of many M-<£ test curves does not indicate a strength. As a comparison to the Bjorhovde classification
clear point for defining the ultimate connection strength. system, the results of Equation (8) indicate a connection
The M-<f> curves for many of the connections are still rising deformation about 30% less than the simplified ductility
at the end of the test. For this reason Christopher9 identi- requirement illustrated in Figure 2(a).
fied a practical connection deformation, and then defined Substituting^ = 0#, from Equation (8) into the three-
the corresponding connection moment as the strength limit parameter power model equation gives the strength limit
state. Considering the connection deformation for a sim- state as
ple beam with a span-to-depth ratio of 30 or less as an ex- Rkifauj
treme case, the upper bound of the expected beam end Muj = \ln (9)
rotation, OR, including beam flexibility and inter-story 1 +( M
drift, is given as
0R = 0.0008Fy + S/h (8) This new identification of the strength limit state removes
a major obstacle to implementing semi-rigid frame design.

2.4 Some Reliability Issues

Variation in the values of the initial connection stiffness,
the connection moment capacity and the shape factor used
in the three-parameter model affect the reliability of ex-
pected connection strength and behavior. Thus, the use of
the three-parameter power model itself, or any mathemat-
ical substitution for actual connection data, affects relia-
bility. Sellier et al.30 present approximate coefficients of
variation for each of these parameters, as given in Table
1. These data are approximate because certain assump-
tions were made regarding connection property statistics.
Very limited statistical test data are available; the study by
Rauscher and Gerstle29 is one of very few.
These sources of variation are of concern for two rea-
sons:
1. The concern for the individual connection perfor-
Fig. 2(b). Beam Moment-Rotation Curve mance and capacity.

ENGINEERING JOURNAL / FIRST QUARTER /1999 15

 2. The effect of these variations on member forces as of unbraced frames. This method considers that moment
they are calculated from analyses that rely on the in- connection properties depend on the loading direction and
teraction of connection stiffness and member stiff- history. The moment connections in Disque's model de-
ness. velop a plastic hinge when they are loaded in one direc-
tion, and respond elastically when unloaded in the opposite
Further investigation is therefore required to verify these
direction.
data, and to determine if there is a need to modify the LRFD
Clearly, connection behavior is affected by the history
resistance factors when designing a semi-rigid frame. Until
and direction of sequentially applied loads. Therefore, it is
this work is complete, it is assumed that a resistance factor
necessary to use the representative connection stiffnesses
of 0.9 is appropriate for the connections in a semi-rigid
for each loading stage for calculating member forces in the
frame design, and the practical limit state for semi-rigid
analysis of semi-rigid frames.
connection design strength, M'n, is
A similar case can be made when considering semi-rigid
ML = 0.9MUJ (10) frame stability. Column behavior in semi-rigid frames is
affected by the degree of end-restraint offered by the con-
2.5 Frame Sensitivity to History and Direction of nection resistance to column rotation. For example, con-
Connection Loads sider a braced (sway-prevented) frame, with beam loads
gradually applied until column buckling occurs. As the
If expected values of connection moments can be accu-
beams are loaded, their end rotations develop connection
rately estimated, then linearized connection stiffness val-
moments. When column buckling develops from additional
ues can be calculated and used in a frame analysis to
load, column rotation occurs at the beams. Examination of
determine member forces, displacements and column sta-
these beam-to-column connections indicates that at the ini-
bility. However, observation of the typical M-qb curve
tiation of buckling, those on one side of the columns con-
represented by Figure 1 suggests that the sequence and
tinue to load, while those on the opposite side will unload.5
direction of the connection loads are needed in addition
Thus, evaluating frame stability also requires consideration
to the estimate of expected connection moments. For ex-
ample, consider an unbraced (sway-permitted) semi-rigid of connection loading history and direction, i.e. the initial
frame that is first subjected to gravity loads, and subse- loading from beam end rotation, and the subsequent load-
quently to lateral loads due to wind or seismic action. In ing caused by column buckling. Chen and Lui8 confirmed
this case the connections at the beam ends rotate as they this observation by the analysis of a semi-rigid braced "tee"
are first loaded by the gravity loads. With the application of sub-assemblage subjected to sequential loading.
lateral load, the leeward beam ends continue to rotate in the Since braced frame buckling is characterized by sin-
same direction, while the windward beam ends begin to un- gle curvature of the columns, there is a regular, alternat-
load. During the subsequent loading stage of the frame, the ing pattern of connection loading-unloading behavior as
connection behavior is significantly affected, in the sense illustrated by Figure 3. The following observations apply
that the loading rate, R^, at the leeward beam end is much to a semi-rigid braced frame with repeating patterns of
less than the unloading rate, Rkh at the windward beam
end.
Disque14 recognized that, in general, the loading char- K/ f r
acteristics of semi-rigid connections are much different u u L L u U L L
from their unloading characteristics, and he proposed the
method of "directional moment connections" for design L L u u L L u u

u u L L U U L L

Table 1
L L U U L L u u
Coefficients of Variation for
Connection Parameters 30
• r / /
Key :
Connection Parameter cov L = Connectioris that load at columri buckling w thtf A = Ru, tangent stiffness

Ultimate Moment Capacity, Mu 18% U = Connections that unload at colli mn buckling w i t h Rk = Rk„ initial stiffnes s.

Initial Stiffness, Rki 25%

Shape Factor, n 15%

Fig. 3. Connection Behavior at Column Buckling in a Braced Frame

16 ENGINEERING JOURNAL / FIRST QUARTER / 1999

member sizes:9 3.0 EXPECTED VALUES OF CONNECTION
MOMENTS
1. All columns are offered restraint from symmetrically
supported beams. Expected values of connection moments can be determined
2. Beams offer restraint based on the combined stiffness from beam-line diagrams such as the one shown in Fig-
of the beam and the connection in alternating posi- ure 6. The basic assumption of this illustration is that the
tions. This stiffness is either the initial or the tangent beam end rotation, 6, is equal to the connection deforma-
stiffness for the connection. tion, (/>. Theoretically, this assumption is violated by ro-
3. Exterior columns are restrained unequally at the top tations of the supporting columns, but in most cases the
and bottom, i.e. at one end the restraint is based on column rotation is small, and the assumption will result in
the tangent stiffness, and at the other end the initial good initial estimates of the connection deformation. These
stiffness applies. estimates provide ways to calculate beam end moments
4. Interior columns are restrained equally at the top and and connection stiffness values.17 The beam-line diagram
bottom, i.e. the restraint is the sum of that based on is therefore a useful tool for selecting preliminary beam
tangent stiffness on one side of the column, and initial sections and beam-to-column connections concurrently.
stiffness on the other side. A frame analysis, using linearized connection stiff-
ness values determined from the expected connection mo-
ments, provides a means to refine the preliminary design.
The results of a similar study of an unbraced semi-rigid
frame also indicate a pattern to connection behavior that is
sensitive to the history and direction of connection loads.9
This study utilized the analysis of an unbraced "tee" sub-
assemblage as illustrated in Figure 4. The initial connection 24'-0 .P 24'-0

load due to gravity loads on the beams is represented by

Figure 4(a), and the subsequent vertical and lateral loads r^~\ ^ H
are shown in Figure 4(b). After the application of the initial, O W14x34 W14x34
0' |
first-stage gravity load, w, a second-stage lateral load, //,
and proportional axial load, P, were applied gradually until S e mi- Rigid / ^ ^
failure occurred. The results indicated that the connection W8 v9/ I 15'- 0
on the windward beam end unloaded at the higher initial
stiffness rate, while the connection on the leeward beam end <
continued to load at the lower reduced rate. Since unbraced
frame buckling is characterized by double curvature of the
J >L •

columns, this behavior results in a regular, repeating pattern

of connection loading-unloading behavior as illustrated in Fig. 4(b). Unbraced "Tee"—Load Sequence 2
Figure 5.

u LIU LI

24'-0 24'-0

U WWU . H H i l i l
CL W14x34 W14x34 ~
U U u U
Semi-Rigid
Connections -W8x24 15'-0
Key:
ri/ ri/
L = Connections that load at column buckling with Rk = RkL, reduced stiffness.

^s: U = Connections that unload at column buckling with Rk - Ru, initial stiffness.

Fig. 5. Connection Behavior at Column Buckling in an Unbraced

Fig. 4(a). Unbraced "Tee"—Load Sequence 1 Frame

ENGINEERING JOURNAL / FIRST QUARTER / 1999 17

This analysis accounts for the column rotations in cal- where Ms - the reactive moment at the beam end that is
culating the connection deformations. An additional sub- equal to the connection moment.
sequent frame analysis serves to verify the expected The equation of the beam-line for a symmetrically sup-
connection moment by demonstrating that the connection ported beam with a uniformly distributed load is given as
deformations are consistent with the connection moments
determined from the M-<£ curves. This process of concur- wl Msl
0 = (13)
rent member and connection design serves to shorten the 24EI 2EI
iterative design and analysis process by providing an ac- Substituting <f> = 0 from Equation (12) into Equation (13)
curate initial estimate of connection moments. results in expressions for the beam moments in terms of the
applied load and the connection stiffness:
3.1 Using the Beam-Line Diagram to Estimate Ms = mswl2 (14)
Connection Moments 2
Mf = mfwl (15)
The connection flexibility ratio, w, which links connec-
tion behavior and beam stiffness, was recognized by where M/ = the moment at the beam mid-span and where
Geschwindner17 to simplify the expression for the con- 1
nection moment in term of the beam stiffness. It is a con- ms = (16)
venient tool for estimating beam end moments, and it is
T20T2uj
defined by mf = 0.125-m, (17)
EI ad Equations (16) and (17) are plotted in Figure 7 to il-
u= Rl (11) lustrate the usefulness of the connection flexibility ratio
k T in the design process, where the connection can be se-
where lected to minimize the beam moments. Using the fixed-end
moment as a reference point, Geschwindner presented a di-
a = EIldRk = the characteristic length factor for agram that is similar to Figure 7. Kotlyar27 gives equations
connection flexibility in terms of the connectionflexibilityratio for beams with
E = modulus of elasticity other loading conditions and semi-rigid connection support
d = beam depth configurations in a useful and convenient table.
/ = span length
/ = moment of inertia of the beam
3.2 Concurrent Selection of Connections
Rk = stiffness of the connection
and Beam Sizes
Utilizing the intersection of the beam-line and the M-<j> If reasonable assumptions can be made for the connec-
curve gives tion stiffness values, in terms of the connection flexibility
ratio, u, then the beam moments may be calculated, and
Msad
</> = (12)
EI

Secant Stifthess, R^ , at factored bad, wu

0.025
Beam -Line at factoredtoad,wu
0.000 -h -+- H h -+- -h -h- -+- H- H
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1—
10 15 20 25 30 35 40 Connection Flexibility Ratio, u
3
Rotation, <p or 0 (10 rad)

Fig. 7. Optimum Connection Flexibility (Symmetrically Supported

Fig. 6. Beam-Line Diagram with M-<j> Curve10 Beam)

18 ENGINEERING JOURNAL / FIRST QUARTER / 1999

preliminary beam sections may be selected. If gravity loads This example serves as a guide to make reasonable as-
are assumed to govern the beam sizes, then us, the con- sumptions for values of us to use for concurrently selecting
nection flexibility ratio corresponding to the connection preliminary beam and connection sizes.
secant stiffness, RkS9 should be assumed. The problem,
however, is knowing what values of us are reasonable 3.3 Structural Analysis Using Expected
assumptions for this purpose. As in the design of any Connection Loads
statically indeterminate structure, the selection of one
component affects the behavior and resulting load effects First-order structural analysis of semi-rigid frames depends
of the adjacent components. This phenomenon requires a on using connection stiffness values which are consistent
trial and adjustment process. Semi-rigid braced frames are with the M-<f> behavior. Generally, the semi-rigid connec-
also statically indeterminate, and the selection of the tion behavior affects braced and unbraced frames in signif-
connection property, us, in the design process also re- icantly different ways.
quires trial and adjustment.
To help in the process of identifying reasonable as- 3.3.1 Braced Frames
sumptions for us, an example is provided in Figure 8, The secant stiffness of the connections provides the means
which shows a family of M-<j> curves for the same type of to distribute the applied loads to the members. Calcula-
semi-rigid connection supporting a beam with a uniformly tions for member forces do not need to consider load-
distributed load represented by the beam-line. For this ing history, because the connections are always loaded
example, the beam is a 25 ft long W21X44 carrying a total in the same direction until column buckling occurs. A
(factored) distributed load of 4.25 k/ft. Each M-<t> curve, simple first-order analysis which includes the approxi-
generated by the three-parameter power model, represents mate connection stiffness therefore will provide reasonably
a top and seat angle connection with double web clip an- accurate values for the member forces.
gles. The web clip angles are L4 X 3V2 X \6 X 0'-5y2", This approach works because there is very little column
and the top and seat angles are L4 X 4 of variable thick- rotation at the interior beam-to-column connections. How-
ness and 0'-6V2" long. Thus, the intersections of the M-<f> ever, in semi-rigid braced frames, significant column rota-
curves and the beam-line represent the expected connec- tion occurs at the connections to the exterior columns. This
tion moments and deformations, depending on the top and can generally be ignored in a preliminary analysis, and in a
seat angle thickness, and assuming no column rotation. Ta- later analysis the connection stiffness can be adjusted up-
ble 2 presents the expected beam moments, Mf at the span ward to account for the column rotation and the associated
center, and Ms at the supports, and end rotations, <ps, as a decrease in connection moment.
response to the variation in top and seat angle thickness, t. Christopher9 demonstrated that preliminary selection of
Table 3 provides the connection stiffness values, Rk, and beams framed to exterior columns should be made by as-
the corresponding values of the connection flexibility ra- suming pinned connections at the exterior ends, while the
tios, w, which are calculated from the expected connection interior ends are supported with semi-rigid connections.
moments given in Table 2. This agrees with the column stability analysis developed
by Bjorhovde5. The interior beams should be selected by
assuming symmetrical supports (equal connection stiffness
at each end). The connection design should take place con-
currently with the beam selection.
This approach is possible by assuming a value for us,
-0.250 which corresponds to the connection secant stiffness. As
-0.313
noted from the above example, the range of 0.4 < us < 1.0
-0.375
-0.438 for a top and seat angle with double clip angles repre-
-0.500 sents a relatively stiff connection. However, this depends
-0.625 on the beam section and span length. Higher values of us
-0.750
correspond to less connection stiffness. Generally, us will
affect the resulting beam end moments, which may then
be calculated to permit the selection of preliminary beam
10 15 20 25 30 35 sizes. Connection element details can be selected that result
Rotation, 6; Deformation, <j> (10 rad) in a value of us that is close to the assumed value.
Selection of preliminary column sections may be made
by simply assuming unbraced length K-factors of 1.0 for
Fig. 8. Influence of Connection Behavior on Beam Response11 exterior columns and 0.85 for interior columns. A more

ENGINEERING JOURNAL / FIRST QUARTER / 1999 19

accurate frame stability evaluation to determine unbraced process requires a different approach to the analysis. If a
length K-factors is appropriate after connection stiffnesses loading sequence is assumed to start with gravity loads
are adjusted for the selected connection details to account and end with lateral wind or seismic loads, then there is a
for column rotations in a frame analysis. change in connection stiffness with each loading sequence,
With the preliminary selection of beams, connections, as some connections continue to load while others unload.
and columns, this analysis can be performed by making This behavior is illustrated by Figures 10(a), (b) and (c),
a linear-elastic computer model of the frame. The con- where the initial connection loading is represented by point
nection behavior is best represented using zero length "a" resulting in M; and <£,- on the M-<f> diagram. Point "b"
elements, but where the software does not include such in Figure 10(a) illustrates the response following a subse-
elements in its library, adjunct (dummy) members can rep- quent increased loading. The effective connection stiffness
resent the connection stiffness in a variety of possible ar-
rangements. An example is shown in Figure 9. The results
of this preliminary analysis provide the means to evaluate
the connections' performance, for ductility demand, for Adjunct Torsion Bar
comparison of expected connection moments to those cal-
culated in the analysis, for member section checks, and fi-
nally for stability of the frame. Afinalverification analysis Beam
is required to account for refining adjustments in the con-
nection stiffness values and changes in the member sizes.
• Column
3.3.2 Unbraced Frames
The sensitivity of semi-rigid unbraced frames to direc-
tional changes of the connection moments in the loading Fig. 9. Connection Model Using Torsion
Elements

Table 3
Variation of Connection Behavior with Top and Seat 800-I
b
Angle Thickness,11 t _ Muj- T\
a ^ — "
Stiffness, Rk (in-k/rad) Flexibility Ratio, u
? 600 :
•1\
Angle 4.25k/ft 4.25k/ft J. 400 - \ <t>a
Thickness
\ y Beam line
f(in) "to RkL Rkl Us UL u, 1
0.7500 198461 11777 1523175 0.411 6.919 0.054 1 200-
0.6250 111697 8606 836525 0.730 9.469 0.097 i i i i
i i i | i I
0.5000 61445 6188 414105 1.326 13.169 0.197 U 1
n -() 5 10 15 20 fa 25 30
0.4375 47137 5171 278375 1.729 15.758 0.293
0.3750 46617 2658 181492 1.748 30.656 0.449
Connection Deformation, 0 (10~3 rad)
0.3125 39781 1410 115693 2.048 57.812 0.704
0.2500 33264 1005 73949 2.450 81.064 1.102

Fig. 10(a). Loading After Initial Deformation

Table 2
Expected Floor Beam Response To Connection Behavior11

Top & Seat Angle

Thickness, f (in) Rigid 0.750 0.625 0.500 0.438 0.375 0.313 0.250 Pinned

Ms (in-k) 2656 1459 1080 727 596 591 521 450 0

Mi (in-k) 1328 2526 2904 3257 3388 3394 3463 3534 3984

< M X 1000 Rad) 0.00 7.35 9.67 11.84 12.64 12.67 13.10 13.54 16.30

20 ENGINEERING JOURNAL / FIRST QUARTER / 1999

associated with the increase of load from "a" to "b" is de- necessary to confirm the correct use of R^ as the oper-
termined as ative connection stiffness for connections at the windward
beam ends. This test is to determine which is greater
in magnitude—the initial connection moment resulting
from gravity loads, M,-, or the subsequent change in con-
where Ma and <f>a are defined in thefigure,and Muj (Equa- nection moment, Ma. Thus, the unloading stiffness at wind-
tion 9) is defined as the practical limit state. Equation (18) ward beam ends is equal to Rkh if and only if, \Ma\ < \Mt\.
also implies that the connection limit state is reached at the Although possible, from the experience of the authors,
end of the second sequence, when the additional moment it is unlikely that \Ma\ > |M,-|, even for small values of ini-
reaches Ma. Thus, Mt + Ma = Muj. tial gravity load. However, for such cases the conservative
Point "c" in Figure 10(b) indicates a subsequent de- approach is simply to ignore the (diminished) benefit of
creased loading. The effective connection stiffness associ- considering Rki in these connections.
ated with a loading reversal from "a" to "c" is the same Implied in this approach is the need to superimpose the
as the initial connection stiffness, R^. In some cases it analysis for gravity load with the analysis for lateral load.
may be possible for the unloading connections (at wind- Superposition of load effects on non-linear structures is not
ward beam ends) to unload completely and then con- legitimate because compatibility of force and deformation
tinue to load in the opposite direction, as illustrated in is violated. However, if the load-deformation relationship
Figure 10(c). Because of this possibility, a simple test is is modified for each loading event, superposition of non-
linear structures can be achieved, or approximated, using
linearized stiffness values based on the expected loads.
Barakat and Chen3,4 recognized this concept when they
presented afirst-orderelastic analysis method consistent
with the AISC LRFD approach, although their procedure
neglects the effect of load reversal.

3.4 Evaluating Semi-Rigid Frame Stability Using

Expected Connection Loads
The solutions provided by Julian and Lawrence21 to de-
termine column unbraced length factors are the basis for
0 5 10 15 20
the well-known alignment charts for braced and unbraced
3
frames.2 The use of these solutions in building column de-
Connection Deformation, <j) (10" rad)
sign is well established common practice. However, struc-
tural configurations often conflict with the assumptions
used in the Julian and Lawrence solutions. This is par-
Fig. 10(b). Unloading After Initial Deformation (\Ma\ < \Mt\) ticularly true for semi-rigid frames when the beams are
notrigidlyconnected to the columns. The relative stiffness
distribution factor, given as

r _ X(EIc/lc)
G (19)
- YiMW
can be modified to accommodate these violations of the
assumptions required for the original solution. Such mod-
ifications have been presented by Yura33 and Johnston19
for certain structural configurations, by considering the ef-
fective stiffness of the beams framing into the column ends.
If the effective stiffness of a beam and its semi-rigid con-
nections can be determined from the expected connec-
tion loads, this same approach to modify G can be utilized
for semi-rigid frames. This approach was developed by
Bjorhovde5 for semi-rigid frames and presented by Christ-
opher,9 using Driscoll's15 solution for the effective stiff-
Fig. 10(c). Unloading After Initial Deformation (\Ma\ > |M,-|) ness of a beam with semi-rigid connections at each end.

ENGINEERING JOURNAL / FIRST QUARTER / 1999 21

 A list of effective stiffness values are presented in Ta- work needs to be undertaken to include other connec-
ble 4,9 based on the equation for the modified relative stiff- tion arrangements. In particular, the database includes few
ness distribution factor, given as connection arrangements which connect to the column
web.
G = ^ (20)
4.2 Reliability
where
Reliability studies are required to investigate and calculate
SRF = the Stiffness Reduction Factor2 to account for a statistically based resistance factor. The suggestion by
inelastic column behavior the authors of using <f> = 0.9 is based on current practice
fjbc = column stiffness modification factor in Table 4, for connection design, but it does not necessarily account
to account for pinned orfixedcolumn ends for the statistical variation of connection properties, model
C* = modified beam stiffness in Table 4, representing accuracy, and the resulting associated variation in member
the effective stiffness of an assembly of a beam load effects. Such statistical data are very sparse.
and its connections The semi-rigid connection properties are sensitive to the
The values in the table are expressed in terms of the con- connection detail dimensions. This fact is evident from
nectionflexibilityratio, u. the procedure presented by Chen et al.7 to calculate MM,
Rki, and n, the parameters that define connection behav-
ior for the three-parameter power model. Sellier et al.30
4.0 LIMITATIONS AND RESEARCH NEEDS presented the results of studies that show the sensitivity
of connection behavior from variation of connection fea-
4.1 Connection Types tures. This sensitivity indicates the need for the design en-
The research to develop the use of the three-parameter gineer to maintain control of the connection details during
power model to represent connection behavior is presently the preparation of shop fabrication drawings. The usual
limited to connections that have bolted connection ele- shop drawing approval review is an appropriate opportu-
ments which are attached to the column flange. Similar nity to compare the connection dimensions with those used

Table 4
Modifications to G, The Relative Stiffness Distribution Factor

Frame Configuration Braced Frame Unbraced Frame

a) Semi-rigid connections of ~ Elg/lg ~ Elg/lg

equal stiffness at each 1 +2u 1 +6t7
end.

b) Semi-rigid connection at
near end, pin at far end. '-SXStt) H S
c) Semi-rigid connection at
near end, fixed at far end.
'-GX3&) Hi)(^)
d) Semi-rigid connections of N/A See Note 1.
unequal stiffness at each
end.

e) Column pinned at far end. fic = 1.5 fic = 0.5

f) Column fixed at far end. Mc = 2.0 /Ltc = 0.667

Notes:
(2UB + ^Elg/lg
1. cr
^2uAUB + 4uA + 4uB + 1
where "A" refers to the near end of the beam, and
"ET refers to the far end of the beam.
9
2. u = * where Rk = Effective connection stiffness.
3. u = * for a pinned connection.
4. u = 0 for a fixed (rigid) connection.

22 ENGINEERING JOURNAL / FIRST QUARTER /1999

in the connection property calculations, and then make cor- connection assembly, based on the expected connec-
rections if necessary. tion moment, is used to calculate the effective column
restraint.
4.3 Standardized Connections 7. The loading sequence which results in connection
The use of standard connection design can be a signif- load reversal is considered in the analysis to deter-
icant asset in the implementation of semi-rigid frame mine member forces and frame stability.
design. Standard connection dimensions and shear capac- 8. The LRFD approach of AISC, which relies on am-
ities are already included in the AISC Manual.2 In an plification factors to account for second-order effects,
attempt to fill the need for semi-rigid connection informa- is appropriate for design and analysis of semi-rigid
tion, Kim and Chen 23 have provided a table of properties frames because of the following considerations:
for top and seat-angle connections with double web clip an-
• The design and analysis technique of separately
gles using the terms of the three-parameter power model,
evaluating Mnt and M\t load effects allows superpo-
MM, Rkh and n. Additional, similar work is necessary for
sition of sequentially applied loads when the semi-
other connection arrangements. To better utilize the design
rigid frame behavior is non-linear.
procedure outlined in Appendix A, additional information
• The criteria for the LRFD Specification are satis-
is needed to present a list of characteristic length factor
fied.
values for <z,, for the initial connection stiffness, and that
for the average, expected or relative values of as, for the These criteria make it possible for practicing engineers
secant connection stiffness. The availability of this infor- to implement the use of semi-rigid frame technology into
mation will reduce the trial and adjustment process asso- design practice using tools and procedures already in place.
ciated with selecting a connection detail. This information As an aid to the proper use of these design techniques,
will also enable selection of connections that will have the recommendations for connection stiffness values for the
engineer's anticipated effect on member forces and frame design of braced and unbraced semi-rigid frames are listed
stability. in Table 5. The design procedure for semi-rigid unbraced
frames is provided in Appendix A.
5.0 SUMMARY AND CONCLUSIONS
1. Connection behavior is estimated using mathemati-
cal models which are calculated from the material REFERENCES
properties and detail dimensions. The three- 1. American Institute of Steel Construction, Load and
parameter power model is appropriate and practical Resistance Factor Design Specification for Structural
for this purpose. Steel Buildings, Chicago, IL, 1993.
2. New connection limit states are defined by the prac- 2. American Institute of Steel Construction, Manual
tical, maximum expected connection deformation, as of Steel Construction, Load and Resistance Design,
given by Equation (8). The required connection duc- 2nd Ed., Vol. I & II, Chicago, IL, 1994.
tility should be assured by tests of similar connection 3. Barakat, M. and Chen, W. R, "Practical Analysis
types as recorded in the databases. of Semi-Rigid Frames," AISC Engineering Journal,
3. Initial, preliminary member section sizes are selected Vol. 27, No. 2,1990, pp. 54-68.
using the expected values of connection moments de- 4. Barakat, M. and Chen, W. R, "Design Analysis of
termined by the beam-line diagrams superimposed on Semi-Rigid Frames: Evaluation and Implementation,"
the connection M-</> curves. AISC Engineering Journal, Vol. 28, No. 2, 1991,
4. Connections are concurrently selected with the mem- pp. 55-64.
ber section sizes to reduce the iterative verification 5. Bjorhovde, R., "Effect of End Restraint on Column
process for the connection deformations. This veri- Strength — Practical Applications," AISC Engineer-
fication process ensures that the deformations deter- ing Journal, Vol. 21, No. 1, 1984, pp. 1-13.
mined in the frame analysis are consistent with those 6. Bjorhovde, R., Colson, A. and Brozzetti, J., "Classi-
determined from the M-</> curves. fication System for Beam-To-Column Connections,"
5. Identifying expected values of connection moments ASCE Journal of Structural Engineering, Vol. 116,
serves to estimate effective linearized connection No. 11, November, 1990, pp. 3059-3076.
stiffness values. These can be used in the structural 7. Chen, W. R, Goto, Y., and Liew, J. Y., Stability De-
analysis to determine member forces. sign of Semi-Rigid Frames, John Wiley and Sons, New
6. Column stability is evaluated using the effective York, NY, 1996.
end restraint at the beam-to-column connections. 8. Chen, W. R and Lui, E. M., "Columns with End
The combined stiffness of the beam and semi-rigid Restraint and Bending in Load and Resistance

ENGINEERING JOURNAL / FIRST QUARTER / 1999 23

 Factor Design," AISC Engineering Journal, Vol. 22, 15. Driscoll, G. C , "Effective Length of Columns with
No. 3,1985, pp. 105-132. Semi-Rigid Connections," AISC Engineering Jour-
9. Christopher, J. E., Semi-Rigid Frame Design and nal, Vol. 13, No. 4,1976, pp. 109-115.
Analysis Techniques, Ph.D. Dissertation, University 16. Frye, M. J. and Morris, G. A., "Analysis of
of Pittsburgh, Pittsburgh, PA, 1996. Flexibly Connected Frames," Canadian Journal of
10. Christopher, J. E. and Bjorhovde, R., "Some Key Cri- Civil Engineering, Vol. 2, No. 3,1975, pp. 280-291.
teria for Semi-Rigid Frame Design," Innovations in 17. Geschwindner, L. F , "A Simplified Look at Par-
Structural Design: Strength, Stability, and Reliabil- tially Restrained Beams," AISC Engineering Jour-
ity. Proceedings of the T. V. Galambos Symposium, nal, Vol. 28, No. 2,1991, pp. 73-78.
Structural Stability Research Council, Lehigh Uni- 18. Goverdhan, A. V, "A Collection of Experimental Mo-
versity, Bethlehem, PA, 1997, pp. 123-134. ment Rotation Curves Evaluation of Predicting Equa-
11 Christopher, J. E. and Bjorhovde, R., "Response Char- tions for Semi-Rigid Connections," M.S. Thesis, Van-
acteristics of Frames with Semi-rigid Connections," derbilt University, Nashville, TN, 1983.
Journal of Constructional Steel Research, Vol. 46:1- 19. Johnston, B. G. ed., Structural Stability Research
3, Paper No. 141,1998. Council, Guide to Stability Design Criteria for Metal
12 Commission of the European Communities (CEC), Structures, 3rd Ed., Wiley Interscience, New York,
Eurocode 3: Design ofSteel Structures, Brussels, Bel- NY, 1976.
gium, 1992. 20. Jones, S. W., Kirby, P. A., and Nethercot D. A.,
13. Deierlein, G. G., "An Inelastic Analysis and Design "Columns with Semi-Rigid Joints," ASCE Journal of
System for Steel Frames with Partially Restrained the Structural Division, Vol. 108, No. ST2, February,
Connections" in Bjorhovde, R., Colson, A., Haaijer, 1982, pp. 361-372.
G., and Stark, J. W. B., eds., Connections in Steel 21. Julian, O. G. and Lawrence, L. S., "Notes on J
Structures II: Behavior, Strength, and Design, AISC, and L Nomograms for Determination of Effective
Chicago, IL, 1992, pp. 408-115. Lengths," unpublished report, Jackson and Moreland
14. Disque, R. O., "Directional Moment Connections— Engineers, Boston, MA, 1959.
A Proposed Method for Unbraced Steel Frames," 22. Kennedy, D. J. L., "Moment-Rotation Characteristics
AISC Engineering Journal, Vol. 12, No. 1, 1975, of Shear Connections," AISC Engineering Journal,
pp. 14-18. Vol. 6, No. 3, 1969, pp. 105-115.

Table 5
Linearized Stiffness Values

Unbraced Frame
Non-Sway Sway
Calculation Objective Braced Frame Mnt

Pu and Mu (Notel) Pks Rks Rki and RkL

Member Forces (Note 2)

Rki and flw Rki and RM

Euler Buckling Load for B, (Note 3) (Note 3)

Pe2 (Note 4) Rki and RkL

Euler Buckling Load for E^ (Note 2)

Pn Rki and R« Rki and RkL

Nominal Compressive Strength (Note 3) (Note 2)

Notes:
1. Rks is calculated from the beam-line diagram or expected moment value.
2. Refer to the repeating pattern of connection stiffness resulting from double
column curvature. (Fig. 5)
3. Refer to the alternating pattern of connection stiffness resulting from single
column curvature. (Fig. 3)
4. Actual connection loading stiffness is RM at Euler buckling, but RkL is more
convenient and conservative.

24 ENGINEERING JOURNAL / FIRST QUARTER / 1999

23. Kim, Y. and Chen, W. E, "Design Tables for Top- APPENDIX A
and Seat-Angle with Double Web-Angle Connec- DESIGN PROCEDURE
tions," AISC Engineering Journal, Vol. 6, No. 35, The following procedure, which incorporates the criteria
1998, pp. 50-75. outlined in this paper, gives the necessary design steps for
24. Kishi, N. and Chen, W. R, "Data Base of Steel an unbraced (sway permitted) semi-rigid frame such as
Beam-to-Column Connections," Structural Engin- that shown in Figure 11.
eering Report No. CE-STR-86-26, School of Civil
1. Select span-to-depth ratios for the beams and col-
Engineering, Purdue University, West Lafayette,
umns to accommodate architectural requirements.
IN, 1986.
2. Assume a value for us, the connection flexibility ratio
25. Kishi, N., Chen, W. E, Matsuoka, K. G. and based on the connection secant stiffness R^, for each
Nomachi, S. G, "Moment-Rotation Relation of Top- semi-rigid connection. Decreasing us corresponds to
and Seat-Angle with Double Web-Angle Connec- increasing the connection stiffness. For guidance in
tions," in Bjorhovde, R., Brozzetti, J., and Colson, assuming uS9 refer to the example in Section 3.2.
A., eds., Connections in Steel Structures: Behaviour, For example, assuming us = 0.25 and 0.70 for the
Strength, and Design, Elsevier, New York, NY, roof and floor beam connections, respectively, for the
1988, pp. 121-134. frame illustrated in Figure 11 will provide relatively
26. Kishi, N., Chen, W. E, Matsuoka, K. G. and No- stiff connections to limit drift.
machi, S. G, "Moment-Rotation Relation of Sin- 3. Calculate beam moments (both interior and exterior
gle/Double Web-Angle Connections" in Bjorhovde, spans) using Figure 7 and Equations (14) to (17) for
R., Brozzetti, J., and Colson, A., eds., Connections the assumed values of us, and select preliminary beam
in Steel Structures: Behaviour, Strength, and Design, sizes.
Elsevier, New York, NY, 1988, pp. 135-149. 4. Design connections to accommodate beam shear reac-
27. Kotlyar, N., "Formulas for Beams with Semi-Rigid tions, and to behave according to the assumed values
Connections," AISC Engineering Journal, Vol. 33, of us.
No. 4,1996, pp. 142-146.
• Use Table 9-2 in Vol. II of the AISC Manual2 to
28. Popov, E. P. and Pinkney, R. B., "Cyclic Yield Rever- select web clip angles.
sal in Steel Building Connections," AISC Engineer- • By trial and adjustment, select top and seat angle
ing Journal, Vol. 8, No. 2, 1971, pp. 66-79. sizes and dimensions that will result in secant stiff-
29. Rauscher, T. R. and Gerstle, K. H., "Reliability ness values that are consistent with the assumed val-
of Rotational Behavior of Framing Connections" in ues of us for the loading combination that produces
Bjorhovde, R., Colson, A., Haaijer, G., and Stark, J. the maximum gravity load for the beams. The as-
W. B., eds., Connections in Steel Structures II: Behav- sumptions for us in the example of step 2 above
ior, Strength, and Design, AISC, Chicago, IL, 1992, correspond to the following connection details:
pp.218-224. • Connection for W14 roof beams—(1.2D + 0.5L +
30. Sellier, A., Mebarki, A., Colson, A., and Bjorhovde, 1.6Lr)
R., "Reliability of Steel Structures with Semi-Rigid Top and seat angles—L6 X 6 X V2 X 0 ft.-6 in.
Connections" in Bjorhovde, R., Colson, A., and Zan- with a gage distance of 2 in.
donini, R., eds., Connections in Steel Structures III: Web clip angles—2-L4 X 3 V2 X V4 X 0 ft.-5 V2 in.
Behaviour, Strength, and Design, Pergamon - Else- with a gage distance of 2 in.
vier, New York, 1996, pp. 331-348.
31. Thornton, W. A., "A Rational Approach to Design of
wL=1.125k/ft ^H = 2.81k
Tee Shear Connections," AISC Engineering Journal, -Q& -qp- -Q^— ,
94
Vol. 33, No. 1,1996, pp. 3 ^ 3 7 . wn
wL=1.250k/ft
32. Thornton, W. A., "Estimates of Ductility Require- b- -cp- H = 5.63 kf
ments for Simple Shear Requirements" in Bjorhovde, «n

R., Colson, A., and Zandonini, R., eds., Connections

in Steel Structures III: Behaviour, Strength, and De- (-Indicates Semi-Rigid Connection A36 Steel

sign, Pergamon - Elsevier, New York, NY, 1996, 4 Bays @ 25' -0 = lOO'-O
pp. 201-209.
Drift limit = 8/h = 1/300
33. Yura, J. A., "The Effective Length of Columns in Un-
braced Frames," AISC Engineering Journal, Vol. 8,
No. 2,1971, pp. 37-^2. Fig. 11. Semi-Rigid Unbraced Frame Example1-

ENGINEERING JOURNAL / FIRST QUARTER /1999 25

 • Connection for W21floorbeams—(1.2D + 1.6L + where
0.5Lr)
h = adjunct member torsion constant
Top and seat angles—L6 x 6 X V2 X 0 ft.-6 in.
L = assumed adjunct member length
with a gage distance of 2 in. G = shear modulus
Web clip angles—2-L4 X 3 V2 X V4 x 0 ft.-11V2 in.
with a gage distance of 2 in. Identify adjunct member property, /*, for each of the
• Determine the M-<j> characteristics of the connec- connection stiffness values representing the follow-
tions by using the procedures outlined by Chen et ing:
al.7 for the three-parameter power model.
• Check connection behavior for its classification • Initial stiffness
as rigid, semi-rigid, or flexible, according to the • Secant stiffness, for each gravity load combination
Bjorhovde et al.6 classification method represented such as
by Figure 2(a).
(1.2D + 1.6L + 0.5Lr)
• Calculate the connection limit states using Equa-
tions 8,9 and 10. (1.2D + 0.5L + 1.6Lr)
• Calculate the linearized stiffness values, R^, Rkt • For reduced stiffness using:
and RkL, for the connections. Note that there is one
set of values for each loading combination to be con- _ Muj - Mj
K
kL T T- (lo)
sidered in the analysis. <Puj ~ <Pi

5. Select preliminary column sizes using estimated axial For example, the initial stiffness of the floor
loads and moments resulting from beam reactions as beam connection, Rki, noted above in step num-
follows: ber 4 is 1,551,000 in-k/rad for a W21 beam. This
stiffness was calculated from Chen et al.7 The
• Assume the beams are rigidly connected to the corresponding stiffness of the adjunct members
columns at their "windward" ends, and pinned at in Figure 9 is 1391 in.4 when their length is as-
their "leeward" ends, as illustrated in Figure 12. sumed to be 10 in.
• Assume leeward exterior columns are governed by Other examples of calculated dummy member
gravity load combinations. properties representing these connection stiffness
• Assume windward exterior columns and interior values is given in Table Al for L = 10 in. and
columns are governed by wind and gravity load G = 11150 ksi. Flexural moment of inertia val-
combinations, and consider drift limits as well as ues Iy and Iz are assumed as 1000 in4 to minimize
strength requirements. errors caused by displacements of the beam ends
6. Create connection models using adjunct (dummy) due toflexureof the dummy members.
members with properties calculated according to con- 7. Assemble afirst-orderlinear-elastic computer model
nection stiffness. Using the adjunct torsion bar system of the frame, incorporating the connection mod-
illustrated by Figure 9, the torsional stiffness of the els with adjunct members as determined in Step 6.
adjunct members are calculated as The use of any commercially available 3-D analy-
sis program is suitable. Prepare the model for two
loading conditions, utilizing the approximate method
h = ^ (Al)
outlined in the AISC LRFD Specification to account
for second-order effects:
• Case I, the non-sway frame condition—subjected
to gravity loads to determine Mnt loads using the
connection secant stiffness, /?&, for all semi-rigid
p _- - — H
connections in the frame.
• Case II, the sway frame condition—subjected to lat-
m H
eral loads to determine Mtt loads using the initial
connection stiffness, R^, for all "windward" beam
-1 L -1 L -J L JL J end connections, and R^L for all "leeward" beam
end connections.
For example, primary load cases and subsequent fac-
Fig. 12. Unbraced Frame with Wind Load tored load combinations that are appropriate for the

26 ENGINEERING JOURNAL / FIRST QUARTER / 1999

 design problem illustrated by Figure 11 are listed as beam ends. If |M,-| < \Ma\, unloading connection
follows: stiffness must be adjusted to insure compatibility
between connection moment and connection defor-
• Gravity load 1—Applied to the non-sway frame mation, considering the connection loading history.
(1.2D + 1.6L + 0.5Lr) Evaluate story drift using the amplification factor,
• Gravity load 2—Applied to the non-sway frame
(1.2D + 0.5L + 1.6Lr) 1
• Gravity load 3—Applied to the non-sway frame B2 =
(1.2D + 0.5L + 0.5L r ) 1 -
ZPe2
• Wind load—Applied to the sway frame (W)
where 2 P — ^(D + L)at service loads.
Case I—Gravity load (1 or 2)
Thus, the amplified drift, 8 = B28'x, where 8'x =
Case II—Combined wind and gravity load (1.2D +
story drift from a first-order elastic analysis. Note that
0.5L + 0.5Lr + L3W)
connection properties used to calculate 8'x and Pe2
(1.2D + 0.5L + 0.5L r )£i + (l.3W)B2
should be calculated from beam-line diagrams using
8. Evaluate connection response using computer model
service level loads.
results as follows:
10. Review beam response using the computer model
• Determine if connections are modeled correctly by results to evaluate strength and serviceability limit
checking connection stiffness as R^, Rks or RkL = states.
{Mbeam)l(QBeam ~~ ® Column)- 11. Evaluate column sizes and frame stability using
• Show that connections do not exceed the strength computer model results. Determine unbraced length
limit state, M'n. factors from the alignment charts shown on page 6-
• Check connection deformation with the ductility 186 of the AISC Manual.2 Modify the relative stiff-
limit, <f>Uj. If the factored connection load effect, ness distribution factor, G, to account for column
Mu < Muj, then there is also adequate ductility, and end restraint offered by the beams using Table 4
no need to check the connection deformation against and
a connection ductility limit. Review the database for
tests of similar connection types to assure that (f>uj Z(g/ c //c)(SRF)/i c
G = (20)
can be reached.
• Verify that the connection unloading stiffness is Calculate Pe\ for the non-sway frame:
equal to the initial connection stiffness, /?&, by - At interior columns (see Figure 3):
showing that the initial connection moments, Af,-,
due to gravity loads are greater than subsequent EI8llg Elgllg
wind moments, Ma, at each of the "windward"
IC* = 1 + 2ut + 1 + 2M,
(Table 4)

Table A1
Summary of Adjunct Member Properties

Rk A lx ly \z Model
Adjunct Members in-k/rad in2 in4 in4 in4 (Type)
Roof Beams
(1.2D+1.6L + 0.5Lr)
Rks 164600 100 148 1000 1000 Non-sway
RkL (leeward ends) 9548 100 8.56 1000 1000 Sway
Rki (windward ends) 712300 100 639 1000 1000 Sway
(1.2D + 0.5Z.+ 1.6M
Rks 58260 100 52.3 1000 1000 Non-sway

Roof Beams
(1.2D+1.6Z. + 0.5M
Rks 129800 100 116 1000 1000 Non-sway
RkL (leeward ends) 9949 100 8.92 1000 1000 Sway
Rki (windward ends) 1551000 100 1391 1000 1000 Sway
(1.2D + 0.5Z-+1.6L,)
Rks 227300 100 204 1000 1000 Non-sway

ENGINEERING JOURNAL / FIRST QUARTER /1999 27

 At exterior columns (see Figure 3), noting the where A refers to the near end, and B refers to the
pattern of connection stiffness at alternating far end of the beam for either initial stiffness, Rki
levels: or reduced connection stiffness, RkL as indicated
by Figure 5.
Zc* = 1 + 2ui
(Table 4) • Determine combined column moments from Case
I and Case II analysis, by
or
Mu = BxMnt + B2Mlt AISC Eq. (Cl-1)
EIgHg
IC* = 1 +2ut
(Table 4) • Apply the AISC LRFD interaction Equations (Hl-
la) and (HI-lb) to evaluate column strength and
where
stability.
EL
ut = , corresponding to tangent stiff- 12. Modify preliminary member sizes and connection
Rktlg details of the frame to meet limit states require-
ness, and ments and/or to improve economy. This involves
ElQ repeating steps required to calculate the modified
w = , corresponding to unloading stiff- connection properties for the computer model.
Rkilg 13. Review the modified computer model results to
ness. verify the frame design using steps 8 through 12
- Determine unbraced length factors. to evaluate connection response, story drift, beam
- Calculate amplification factor, B\ and column response, column strength and stability. This step
strength, P n , for each of the columns using the is also necessary to verify the design by showing
unbraced length factors determined from above. that the calculated connection moments are close
• Calculate Pe2 and Pn for the sway frame. Consider to the expected moments used in the beam-line
the unequal connection stiffness at each end of the diagrams.
restraining beams. Use column end restraint as:
(2uB + l)EIg'gL/lg
(Table 4)
ir = z \2UAUB + 4UA + 4wfl + 1

28 ENGINEERING JOURNAL / FIRST QUARTER / 1999