INTERACTIVE DESIGN OF SEMIRIGID STEEL FRAMES

By Balaur S. Dhillon,1 Fellow, ASCE, and

James W. O’Malley III,2 Associate Member, ASCE

ABSTRACT: A computer-based analysis and design method is presented for the design of semirigid steel frames.
A second-order nonlinear analysis is used, which includes the effects of the flexibility of the connections and
the geometric nonlinearity of the members, in conjunction with the LRFD specification of the AISC. The Frye
and Morris polynomial model is adopted for modeling the semirigid connections. The design procedure is
iterative and interactive in nature, and gives options to the engineer, interacting with the computer, to change
member sizes and connection details for economic and practical considerations. Several examples are presented
to demonstrate the validity and effectiveness of the design procedure for semirigid frames. It is shown that
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

designs with greater economy may be realized with semirigid connections by a variation of connection stiffness
that balances the span and end moments in a beam. The examples also demonstrate the more significant effect
of connection flexibility on drift in unbraced frames.

INTRODUCTION the Frye-Morris polynomial connection model (Frye and Mor-

ris 1975) and AISC-LRFD specification (1994). The design
Beam-to-column connections are an integral element of a procedure is interactive in nature, and the design engineer re-
steel frame, and their behavior affects its overall performance tains the option of changing member sizes and/or connection
under loads. In common engineering practice, it is usually as- details to achieve economic designs.
sumed that connections are either rigid or pinned. In reality,
connections are semirigid and possess a certain degree of ro- MODELING OF CONNECTIONS
tational restraint. Connections act as a medium through which forces are
The AISC-LRFD 1994 designates two types of steel transferred from one structural member to another. The major
construction — fully restrained (FR) type and partially re- force transferred through the connection is the moment, re-
strained (PR) type. FR-type construction assumes full conti- sulting in a rotational deformation. A connection under an ap-
nuity and adequate rigidity of beam-to-column connections, plied moment rotates through angle ␪r, which is the angle be-
which retain the initial angle between intersecting members. tween beam and column from their original position. Fig. 1
PR-type construction assumes that a connection possesses mo- shows the relative moment-rotation behavior of a variety of
ment capacity somewhere between the fully fixed and per- commonly used semirigid connections (Chen and Lui 1991).
fectly pinned conditions. The stiffness or flexibility of a connection is dependent on
Most connections that fasten beams to columns using an- the geometric parameters, such as plate or angle thickness, bolt
gles, plates, welds, and bolts are deformable and exhibit a size, connection depth, etc. The moment-rotation relationship
nonlinear behavior between conditions fully fixed and per- is nonlinear for all types of connections over the entire range
fectly pinned. It is more reasonable to classify all connections of loading, and varies depending on connection flexibility, i.e.,
under the classification of semirigid, with rigid and pinned the more flexible the connection, the larger the relative rota-
conditions being special cases. Connection flexibility affects tion.
both force distribution and deformation in beams and columns The moment-rotation relationship has been the subject
of the frame, and must be accounted for in a structural anal- of several studies. Extensive experimental tests have been
ysis. This can be done by conducting a second-order analysis conducted on the commonly used connections, and a large
of the frame, one that addresses both geometric nonlinearity collection of test data reported (Kishi and Chen 1986).
and connection flexibility. The AISC-LRFD requires that the Several mathematical models have been proposed to fit the
flexibility of the connection be evaluated by a rational analysis moment-rotation curves from the experimental data. These
or empirical means. Extensive work has been done in this di- models vary widely in their complexity, ranging from the
rection, and considerable data is available from tests of con-
nections (Kishi and Chen 1986). Other studies have recently
focused on the analysis of semirigid steel frames. However,
much more work is needed, especially in the development of
computer software, before its use is accepted routinely in de-
sign office practice.
The present study aims at developing a computer-oriented
analysis and design method for unbraced steel frames with
semirigid connections. A second-order nonlinear approach is
adopted to include both geometric nonlinearity and connection
flexibility. The effects of material nonlinearity, and cyclic or
dynamic loading, are not considered. The design is based on
1
Assoc. Prof., Dept. of Civ. Engrg., Widener Univ., Chester, PA 19013.
2
Struct. Engr., IPS, Lafayette Hill, PA 19444.
Note: Associate Editor: Amde M. Amde. Discussion open until Oc-
tober 1, 1999. To extend the closing date one month, a written request
must be filed with the ASCE Manager of Journals. The manuscript for
this paper was submitted for review and possible publication on March
10, 1998. This paper is part of the Journal of Structural Engineering,
Vol. 125, No. 5, May, 1999. 䉷ASCE, ISSN 0733-9445/99/0005-0556 –
0564/$8.00 ⫹ $.50 per page. Paper No. 17957. FIG. 1. Connection Moment Rotation Curves

556 / JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999

J. Struct. Eng. 1999.125:556-564.

 simplest linear model to polynomial, exponential, and power
models.
The polynomial model proposed by Frye and Morris offers
ease of implementation and is adopted in the present study.
This model is represented by an odd power polynomial of the
form

␪r = C1(␬M )1 ⫹ C2(␬M )3 ⫹ C3(␬M )5 (1)

where ␬ = standardization constant dependent upon connection

type and geometry; and C1, C2, C3 = curve-fitting constants.
Table 1 provides values for the curve-fitting constants and an
equation for the standardization constant for each major con-
nection type. The size parameters in the equations for stan-
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

dardization constants in Table 1, for the eight types of con-

nections, are given in Fig. 2.

Plane Frame – Element Stiffness

The linear elastic analysis assumes that the deformations are
relatively small, and the equilibrium equations can be formu-
lated with respect to initial geometry. When increasing applied
loads cause significant changes in the structure’s geometry, the
equilibrium and compatibility equations are nonlinear and the
resulting stiffness matrix contains terms that are functions of
axial forces and deformations. The stiffness matrix to represent
this behavior must include the effect of geometric nonlinearity.
In addition, for members with flexible connections at the ends,
the stiffness matrix must be modified to account for the effect
of connection flexibility.
In general, the columns in frames are continuous and do not
possess any internal flexible connections. The beams have
semirigid end connections, but have small axial forces with
negligible effect of geometric nonlinearity. For simplification,
two types of elements will be used in the design of plane
frames with semirigid connections:
FIG. 2. Connection Types and Size Parameters

1. Beam-Column Element. A plane-frame element modified

2. Semirigid-Beam Element. A plane-frame element modi-
to include geometric nonlinearity.
fied to include end connection flexibility.
TABLE 1. Standardized Connection Constants
Beam-Column Element
Connection Curve-fitting
types constants Standardization constant a The stiffness matrix of a beam-column element incorporat-
(1) (2) (3) ing P-delta effects can be expressed as the sum of two matri-
Singe web-angle C1 = 4.28 ⫻ 10⫺3 ␬ = d ⫺2.4
a t⫺1.81
a g0.15 ces, [k¯ e] and [k¯ g]. [k¯ e] is the first-order-frame-element stiffness
C2 = 1.45 ⫻ 10⫺9 matrix, and [k¯ g] is called the geometric stiffness matrix. For
C3 = 1.51 ⫻ 10⫺16 the element i, the nonlinear stiffness matrix is written as
Double web- C1 = 3.66 ⫻ 10⫺4 ␬ = d ⫺2.4
a t ⫺1.81
a g0.15
angle C2 = 1.15 ⫻ 10⫺6 ¯ i = [k¯ e]i ⫹ [k¯ g]i
[k] (2)
C3 = 4.57 ⫻ 10⫺8
Top and seat an- C1 = 2.23 ⫻ 10⫺5 ␬=d ⫺1.287 ⫺1.128 ⫺0.415 ⫺0.694
t c t a l g 1.35
where
gle with dou- C2 = 1.85 ⫻ 10⫺8
ble web angle C3 = 3.19 ⫻ 10⫺12
Top and seat an- C1 = 8.46 ⫻ 10⫺4 ␬ = d ⫺1.5t ⫺0.5l a⫺0.7d b⫺1.5 Ax E
gle without C2 = 1.01 ⫻ 10⫺4
double web C3 = 1.24 ⫻ 10⫺8 L
angle 12EI
End-plate with- C1 = 1.83 ⫻ 10⫺3 ␬ = d ⫺2.4t ⫺0.4 d ⫺1.5 0 SYM.
g p b
L3
out column C2 = 1.04 ⫻ 10⫺4
stiffeners C3 = 6.38 ⫻ 10⫺6 6EI 4EI
0
End-plate with C1 = 1.79 ⫻ 10⫺3 ␬ = d ⫺2.4t ⫺0.6 L2 L
[k¯ e]i =
g p
column stiff- C2 = 1.76 ⫻ 10⫺4 ⫺Ax E Ax E
eners C3 = 2.04 ⫻ 10⫺4 0 0
T-stub C1 = 2.10 ⫻ 10⫺4 ␬=d ⫺1.5 ⫺0.5 ⫺0.7
t t l d ⫺1.1
b
L L
C2 = 6.20 ⫻ 10⫺6 ⫺12EI ⫺6EI 12EI
C3 = ⫺7.60 ⫻ 10⫺9 0 0
L3 L2 L3
Header-plate C1 = 5.10 ⫻ 10⫺5 ␬ = d ⫺2.3t p⫺1.6t ⫺0.5 g1.6
⫺6EI
p w
C2 = 6.20 ⫻ 10⫺10 6EI 2EI 4EI
0 0
C3 = 2.40 ⫻ 10⫺13 L2 L L2 L (3)
a
See Fig. 2 for identification of coefficients.
and
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999 / 557

J. Struct. Eng. 1999.125:556-564.

 0 0
6
0
5
L 2L2
P 0
[k¯ g]i = 10 15
L 0 0 0 0
⫺6 ⫺L 6 FIG. 4. Plane-Frame Element
0 0
5 10 5
⫺L2 ⫺L 2L2
冉 冊冉 冊 冉 冊冉 冊
2
L 4EI 4EI EI 4
0 0 R= 1⫹ 1⫹ ⫺ (8b)
10 30 10 15 (4) LRkA LRkB L RkA RkB
where L = length of element; Ax = cross-section area; I = Transforming (7) into a member-stiffness relationship of a
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

moment of inertia; E = modulus of elasticity; and P = axial 6-DOF plane-frame element (Fig. 4), we obtain the following
force in element. stiffness matrix for a semirigid beam element:

Semirigid-Beam Element [k]i =

The influence of connection flexibility can be accounted for Ax E

by modifying the slope deflection equations for a beam ele- L
ment to account for rotational springs. For the beam element EI
i (Fig. 3), the relative rotations of springs ␪rA and ␪rB are related 0 (sii ⫹ 2sij ⫹ sjj)
L3
SYM.
to spring stiffnesses RkA and RkB as EI EI
0 (sii ⫹ sij) sii
MA L2 L
␪rA = (5a) ⫺Ax E Ax E
RkA 0 0
L L
and EI EI EI
0 ⫺(sii ⫹ 2sij ⫹ sjj) ⫺(sii ⫹ sij) 2 0 (sii ⫹ 2sij ⫹ sjj)
L3 L L3
MB EI EI EI EI
␪rB = (5b) 0 (sij ⫹ sjj) 2 sij 0 ⫺(sij ⫹ sjj) 2 sjj
RkB L L L L
(9)
The modified slope deflection equations are obtained by re-
placing the joint rotations ␪A and ␪B by (␪A ⫺ ␪rA) and (␪B ⫺ The stiffness equation for the element is of the form
␪rB), respectively, as follows:

冋冉 冊 冉 冊册
¯ ¯ ⫹ {r̄F}
{r̄} = [k]{d} (10)
EI MA MB
MA = 4 ␪A ⫺ ⫹2 ␪B ⫺ ¯ = member end force and displacement vec-
where {r̄} and {d}
L RkA RkB
tors in member coordinates. {r̄F} = fixed-end force vector re-
MB =
EI
L
冋冉 4 ␪B ⫺
MB
RkB
冊 冉⫹2 ␪A ⫺
MA
冊册
RkA (6)
sulting from in-span gravity loads on the beam with semirigid
connections, and is given by

Eq. (6) can be reduced to {r̄F} = [0 V*

FA M*
FA 0 V*
FB M*
FB] (11)
where
EI
MA = (sii ␪A ⫹ sij ␪B)
L MFA ⫹ 6␣B MFB
* =
M FA (12a)
(1 ⫹ 4␣A ⫹ 4␣B ⫹ 12␣A ␣B)
EI
MB = (sij ␪A ⫹ sjj ␪B)
L (7) MFB ⫹ 6␣A MFA
M*
FB = (12b)
(1 ⫹ 4␣A ⫹ 4␣B ⫹ 12␣A ␣B)
where

冉 冊
where
1 12EI
sii = 4⫹ EI/L EI/L
R LRkB ␣A = and ␣B = (13)

冉 冊
kA kB
1 12EI
sjj = 4⫹ MFA and MFB are fixed-end moments for the member with fixed
R LRkA
connections. The fixed-end shears V *FA and V *
FB are obtained

2 by considering the equilibrium of the member. Eqs. (12a) and

sij = (12b) are derived in Appendix I and hold for concentrated,
R (8a)
uniform, and linear loads on members.
and
Generalized Stiffness Matrix
The beam-column and semirigid-beam stiffness matrices,
(2) and (9) respectively, are in local or member coordinates.
It is necessary to transfer them to global or structure-oriented
coordinates before a complete structure stiffness matrix can be
formed. The transformation is accomplished by the relation

FIG. 3. Beam Element with Rotational Springs ¯ i [T ]i

[k]i = [T ]Ti [k] (14)

558 / JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999

J. Struct. Eng. 1999.125:556-564.

 where Pu = required axial compressive strength; Pn = nominal
axial compressive strength; Mu = required flexural strength; Mn
= nominal flexural strength; ␾c = resistance factor for com-
pression = 0.85; ␾b = resistance factor for flexure = 0.90; and
IER = interaction equation ratio.
The nominal axial compressive strength Pn and flexural
strength Mn can be determined from the appropriate column
strength and beam strength formulas based on the AISC-LRFD
limit state requirements.
The required axial strength Pu and flexural strength Mu are
determined from analysis under factored loads, and must in-
clude the effects of geometric nonlinearity and connection
flexibility.
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

Effective Column-Length Factor

FIG. 5. Plane-Frame Element in Member (x̄, ȳ, z̄ ) and Structure A column is an integral part of a frame, where end condi-
(x, y, z ) Coordinates tions depend on the stiffnesses of the beams and girders fram-
ing into the column, and the rigidity of the beam-column con-
where [k]i = stiffness matrix for member i in structure coor- nections. An effective length factor K is used to modify the
¯ i = stiffness matrix for member i in member
dinate system; [k] actual length of the column for design purposes. The factor K
coordinate system; and [T ]i = transformation matrix for mem- is defined (AISC-LRFD 1994) as

冋 册
冑
ber i, of the form
Pe
␭ ␮ 0 0 0 0 K= (18)
Pcr
⫺␮ ␭ 0 0 0 0
0 0 1 0 0 0 where Pcr = critical load of the end-restrained column; and Pe
[T ]i = (15)
0 0 0 ␭ ␮ 0 = the Euler buckling load for pin-ended column. The effective
0 0 0 ⫺␮ ␭ 0 length factor K for the columns in a frame is determined from
0 0 0 0 0 1 the following interaction equations.

in which ␭ = cos ␣; ␮ = sin ␣; and ␣ = angle between structure For a braced frame:
and member coordinate systems (Fig. 5). The stiffness matri-
冉冊 冋 册
2
GA GB ␲ GA GB ␲/K 2 tan(␲/2K )
ces in structure coordinates for beam-column and semirigid ⫹ 1⫺ ⫹ =1 (19)
beam (14) can be expressed in explicit form (O’Malley 1997). 4 K 2 tan ␲/K ␲/K
For an unbraced frame:
DESIGN CONSIDERATIONS
GA GB (␲/K )2 ⫺ 36 ␲/K
The AISC-LRFD specifications contain provisions that con- = (20)
6(GA ⫹ GB) tan(␲/K )
sider the nonlinear nature of the structure due to the material
and geometric nonlinearities and the effect of connection flex- where GA and GB = stiffness distribution factors for Ath and
ibility. It recommends either the use of modifiers to the results Bth ends of column. The stiffness factor G (AISC-LRFD 1994)
is defined as

冘
of a linear analysis, or the use of nonlinear analysis.
Ic /Lc

冘
LRFD Design Format G= (21)
The limit-states format in general form is written as Ig /Lg

冘 ␥i Qi ⱕ ␾Rn (16)
where the summation is taken over all members connected to
the joint, and where Ic = moment of inertia of column section
where Qi = nominal load effect (i indicating load type); ␥i = corresponding to plane of buckling; Lc = unbraced length of
load factor corresponding to Qi; Rn = nominal resistance; and column; Ig = moment of inertia of beam/girder corresponding
␾ = resistance factor. to plane of bending; and Lg = unbraced length of beam/girder.
The design strength ␾Rn for each structural element must Eqs. (19) and (20) are based on assumptions that the beams
be greater than or equal to the required strength, ␥i Qi, deter- and girders are rigidly connected to the columns at the joints,
mined from appropriate load combinations recommended in all columns of a story buckle simultaneously, and members
the AISC-LRFD specification. The load factors ␥i and the re- are prismatic and behave elastically. To account for different
sistance factor ␾ account for uncertainties in the determination end conditions, the beam/girder stiffness Ig /Lg in (21) is mul-
of loads and resistances, respectively. tiplied by the following factors:

Interaction Equation • For braced frames, the factor is 1.5 for the far ends fixed,
2.0 for pinned, and 1/(1 ⫹ 2EI/LRk) for flexibly con-
The interaction equation for the members of a plane frame nected.
under bending and axial stress is of the form (AISC-LRFD • For unbraced frames, the factor is 0.5 for far ends fixed,
1994) 0.67 for pinned, and 1/(1 ⫹ 6EI/LRk) for flexibly con-
nected.
Pu Pu 8 Mu
ⱖ 0.2, IER = ⫹ ⱕ1 (17a)
␾c Pn ␾c Pn 9 ␾b Mn ITERATIVE NUMERICAL ANALYSIS
Pu Pu Mu The structural stiffness equations formed by superimposing
< 0.2, IER = ⫹ ⱕ1 (17b)
␾c Pn 2␾c Pn ␾b Mn the member stiffnesses include effects of geometric nonlinear-
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999 / 559

J. Struct. Eng. 1999.125:556-564.

 Final Design
The design procedure is implemented on the computer pro-
gram PFNLFCAD (Plane-Frame, Nonlinear, Flexible Con-
nected Analysis & Design) developed for the PR-type steel
frames (O’Malley 1997). It consists of two distinct modules,
one for analysis and one for design. The analysis module con-
sists of a number of subroutines to form structure-stiffness
equations, and solves the joint displacements and member
forces. The design module uses the results from the analysis
module to check the limit states of the members in the struc-
ture, and outputs the results. It allows the user to interact with
the computer to create an interactive, iterative analysis and
design sequence. The two modules are summarized below.
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

Analysis Module

FIG. 6. Connection Secant Stiffness through Load Increments 1. Divide applied loads into a series of small increments.
2. Calculate the load increment vector {⌬A}.
3. Calculate the member-stiffness matrix [k]i for all mem-
ity and connection flexibility. These equations are nonlinear,
bers.
and require an iterative solution procedure. The applied loads
4. Assemble the member-stiffness matrices to form the
are divided into a number of small-load increments for which
structure stiffness [S ].
the structural stiffness equations are written in the incremental
5. Solve the incremental stiffness equation (22) for {⌬D}.
form
6. Determine member end actions.
{⌬A} = [S ]{⌬D} (22) 7. Check convergence. If convergence is achieved, pro-
ceed to step 11.
where [S ] = 兺[k]i, structure stiffness matrix; {⌬A} = incre- 8. Calculate the connection secant stiffnesses.
mental load vector; and {⌬D} = incremental displacement vec- 9. Update the nonlinear terms in the member stiffnesses,
tor. The incremental equations (22) are iteratively solved using the most recent connection stiffness, and member
through a sequence of linear steps. The secant stiffness ap- forces and structure geometry.
proach (O’Malley 1997) is used for evaluating the connection 10. Repeat steps 3 – 9 until convergence is achieved.
stiffness. For each load increment, the stiffness matrix is cal- 11. At convergence, calculate accumulated displacements
culated at the start of each iterative cycle. It requires calcula- and member forces.
tion of the secant stiffness at the beginning of each cycle, and 12. Continue until the analysis is complete for all load in-
changing of reference geometry and member axial forces crements.
based on information from the previous cycle. The connection
secant stiffness corresponding to all load increments is shown
in Fig. 6. The convergence compares the three nodal displace- Design Module
ments, and the rotation of the semirigid connection in the cur-
rent cycle with those of the previous cycle. Convergence is 1. Check all members to satisfy the design specifications.
obtained when the difference between joint displacements of 2. If the design is not satisfactory for any member, revise
two consecutive cycles reaches a specified tolerance. the member properties for the inadequate or oversized
A convergent solution of a load increment forms an initial member.
estimate for the next iteration, and the iterative process con- 3. Repeat until satisfactory design is obtained.
tinues until all load increments are considered. The solutions
for all load increments are accumulated to obtain a total non- Frame Specification and Design Option
linear response.
All members are checked to satisfy the limit-state require- The program works through a datafile ‘‘NLFCDAT,’’ which
ments of the AISC/LRFD specifications. If the design is not includes structural parameters, frame geometry, member prop-
satisfactory, the member properties of the inadequate and over- erties, loading, and design options. The member properties in-
sized members are revised and reanalysis is performed. This itially correspond to preliminary estimated member sizes. The
process is repeated until a satisfactory design results. loading is in the form of factored member and joint loads.
Number of load increments and increment factors also form
DESIGN PROCEDURE the input data. The design options are indicated with an indi-
cator, zero or one. These options consist of ‘‘0’’ for analysis
Preliminary Design only option, and ‘‘1’’ for analysis and design option; ‘‘0’’ for
The proposed design approach for PR-type frames is itera- not including P-delta effect and ‘‘1’’ for inclusion of P-delta
tive and interactive. The selection of initial sizes of members effect; and ‘‘0’’ for unbrace frame, ‘‘1’’ for braced frame.
and connections effects the computation effort to achieve ec-
onomic designs. Preliminary sizes are based on simplifying Connection Information
assumptions and practical considerations. The present study
suggests the following guidelines: Eight types of connections are used in addition to the stan-
dard pinned and fixed extremes, and one with constant spring
1. Select floor and roof beams for full fixity at ends. stiffness, Rk. The connection size parameters in the equations
2. Select beam-to-column connections for 90% of the mem- for standardization constants (Table 1) may be obtained from
ber fixed-end moments. the AISC Steel Manual corresponding to the column, beam
3. Select column sizes for gravity loads only using a col- and angle sizes, and fastener spacing and diameter. Each con-
umn effective length factor of 1.3. nection type has an identification number, given below.
560 / JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999

J. Struct. Eng. 1999.125:556-564.

 1. Single web angle value. This problem is taken from Bhatti and Hingtgen (1995)
2. Double web angle to demonstrate the validity of the program PFNLFCAD and
3. Top and seat angles with double web angles show the effects of connection flexibility and geometric non-
4. Top and seat angles linearity on displacements and moments.
5. End plate The results of the analysis are presented in Tables 2 and 3.
6. End plate with column stiffener Table 2 shows that joint displacements increased by 14% with
7. T-stub P-delta effect over rigid connection and by 50% with com-
8. Header plate bined connection flexibility and P-delta effects. The moments
9. Rigid or fixed in columns (Table 3) increased by 12 – 16% due to the P-delta
10. Pinned effect, and increased by 20 – 76% due to combined connection
11. Spring rotational stiffness, Rk flexibility and P-delta effects. The moments in beams in-
creased by 14% due to the P-delta effect in the rigid frame,
Member Design but show some decrease from the rigid model to the flexible
connection model. Both results show a close comparison with
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

In the design module, the limit-states for the members are Bhatti and Hingtgen (1995).
checked for the beams and beam columns and listed in output
in the form of interaction equation ratios (IER). It also includes Two-Story, One-Bay Semirigid Frame
subroutines for the calculation of moment gradient factor, Cb
for beams, and effective length factor K for columns. The de- The two-story, one-bay frame analyzed by Chen et al. is
sign engineer interacting with the computer can select member considered to show a comparison of the polynomial model
size based on the value of IER compared with 1. An IER value used in PFNLFCAD with an exponential model. Fig. 8 shows
greater than 1 indicates that the member is inadequate and a configuration, dimensions, loading, and numbering of joints
larger section should be selected. An IER value smaller than and members. The beam-to-column connections are the top
0.9 gives the indication that the design may be improved by and seat angle with double web angles, of size parameters as
selecting a reduced shape. The iterative design continues until
the engineer is satisfied with his member and connection se-
lection.

DESIGN EXAMPLES
Several plane frames are investigated using the program
PFNLFCAD to demonstrate the validity and effectiveness of
the design procedure for semirigid frames. The examples also
show the effect of connection stiffness on member sizes and
frame drift in unbraced plane frames under static loads. ASTM
A36 steel is used in all examples. The effect of material
strength variation on the connection is not considered.

Two-Story, One-Bay Frame with Concentrated Loads

A two-story, one-bay frame is analyzed with rigid and
semirigid connections, and with and without geometric non-
linearity. Fig. 7 shows the frame configuration, dimensions,
loading, and numbering of members and joints. The connec-
tions have an initial stiffness of 88,889 kN-m/rad (786,732 in.-
kip/rad) and ultimate moment of 224.7 kN-m (1,989 k-in.) The
connection stiffness is assumed to remain constant at its initial FIG. 7. Two-Story, One-Bay Frame with Concentrated Loads

TABLE 2. Lateral Displacements, mm (in.) (Example 1)

Rigid Connections Rigid Connections Semirigid Connections

No P-⌬ Effect P-⌬ Effect P-⌬ Effect
Joint PFNLFCAD Bhatti et al. PFNLFCAD Bhatti et al. PFNLFCAD Bhatti et al.
(1) (2) (3) (4) (5) (6) (7)
3 25.7 (1.011) 25.7 (1.011) 29.6 (1.165) 29.7 (1.168) 37.4 (1.473) 37.5 (1.477)
5 26.9 (1.509) 26.9 (1.509) 43.8 (1.724) 43.9 (1.731) 58.1 (2.287) 58.2 (2.292)

TABLE 3. Absolute Maximum Moments, kN-m (in.-kips) (Example 1)

Rigid Connections Rigid Connections Semirigid Connections
No P-⌬ Effect P-⌬ Effect P-⌬ Effect
Member PFNLFCAD Bhatti et al. PFNLFCAD Bhatti et al. PFNLFCAD Bhatti et al.
(1) (2) (3) (4) (5) (6) (7)
1 163.7 (1,449) 163.8 (1,450) 186.8 (1,653) 186.9 (1,654) 184.6 (1,634) 184.6 (1,634)
2 80.3 (711) 80.3 (711) 89.8 (795) 89.8 (795) 101.9 (902) 101.9 (902)
3 163.0 (1,443) 163.0 (1,443) 189.4 (1,676) 189.5 (1,677) 196.4 (1,738) 196.5 (1,739)
4 162.4 (1,437) 162.4 (1,437) 188.6 (1,669) 188.6 (1,669) 195.5 (1,730) 195.6 (1,731)
5 80.3 (711) 80.3 (711) 89.7 (794) 89.7 (794) 101.9 (902) 101.9 (902)
6 80.3 (711) 80.3 (711) 89.8 (795) 89.8 (795) 101.9 (902) 101.9 (902)

JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999 / 561

J. Struct. Eng. 1999.125:556-564.

 follows: d = 204.6 mm (8.25 in.), t = 7.9 mm (0.3125 in.), ta Four-Story, Two-Bay Frame
= 6.4 mm (0.25 in.), la = 152.4 (6.0 in.), and g = 50.8 mm
(2.0 in.). The moment-rotation relationship for connections is A four-story, two-bay frame is considered with the frame
given in Fig. 9. configuration, dimensions, and loadings as shown in Fig. 10.
The results of the analysis are shown in Tables 4 and 5. The beam-to-column connections are T-stub, type 7. The
Table 4 shows joint displacements, and Table 5 lists the mem- beams are assumed to have full lateral support. The frame is
ber forces. These results compare favorably with Chen et al. designed with the following options:
The discrepancy in the results may be attributed to the differ-
ent models used for the semirigid connections. PFNLFCAD • Rigid connections with P-delta effect
uses the Frye-Morris polynomial model to approximate the • Semirigid (T-stub) connections with P-delta effect
moment-rotation curve, whereas Chen et al. uses the modified
exponential model for the semirigid connections. The two designs are tabulated in Table 6, in the form of
member sizes and interaction equation ratios, and the frame
drift. In general, the flexible connections in comparison to
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

rigid connections reduce beam-end moment and increase span

moments, and result in higher moments in some columns. The
semirigid connection results in lowering the overall weight of

FIG. 8. Two-Story, One-Bay Semirigid Frame

FIG. 9. Moment Rotation Curve of Connections FIG. 10. Four-Story, Two-Bay Frame

TABLE 4. Joint Displacements (Example 2)

Horizontal, mm (in.) Vertical, mm (in.) Rotation (rad)
Joint PFNLFCAD Chen et al. PFNLFCAD Chen et al. PFNLFCAD Chen et al.
(1) (2) (3) (4) (5) (6) (7)
1 0.000 0.000 0.0000 0.0000 0.00000 0.00000
2 0.000 0.000 0.0000 0.0000 0.00000 0.00000
3 25.58 (1.007) 24.00 (0.945) ⫺0.54 (⫺0.0213) ⫺0.55 (⫺0.0217) ⫺0.01012 ⫺0.00937
4 25.50 (1.004) 23.83 (0.938) ⫺0.60 (⫺0.0237) ⫺0.62 (⫺0.0243) ⫺0.00976 ⫺0.00874
5 62.26 (2.451) 56.54 (2.226) ⫺0.80 (⫺0.0315) ⫺0.79 (⫺0.0312) ⫺0.00897 ⫺0.00775
6 62.18 (2.448) 56.44 (2.222) ⫺0.89 (⫺0.0352) ⫺0.89 (⫺0.0352) ⫺0.00859 ⫺0.00711

TABLE 5. Absolute Member-End Forces (Example 2)

Axial, kN (kips) Shear, kN (kips) Moment, kN-m (kip-ft)
Member PFNLFCAD Chen et al. PFNLFCAD Chen et al. PFNLFCAD Chen et al.
(1) (2) (3) (4) (5) (6) (7)
1 157.56 (35.42) 156.84 (35.26) 23.84 (5.36) 22.73 (5.11) 83.64 (61.69) 79.23 (58.44)
2 179.26 (40.30) 180.33 (40.54) 25.80 (5.80) 25.98 (5.84) 85.99 (63.42) 82.99 (61.21)
3 32.92 (7.40) 32.43 (7.29) 5.92 (1.33) 4.05 (0.91) 15.69 (11.57) 14.18 (10.46)
4 43.90 (9.87) 44.30 (9.96) 10.59 (2.38) 12.19 (2.74) 24.39 (17.99) 29.27 (21.59)
5 14.90 (3.35) 13.79 (3.10) 135.40 (30.44) 135.98 (30.57) 26.53 (19.57) 33.10 (24.41)
6 10.41 (2.34) 12.19 (2.74) 43.90 (9.87) 44.39 (9.98) 24.39 (17.99) 29.27 (21.59)

562 / JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999

J. Struct. Eng. 1999.125:556-564.

 the frame, mainly due to reduction in beam sizes. The reduc- TABLE 7. Interaction Equation Ratios and Drift (Example 4)
tion in weight of the semirigid frame is 7.2% by comparison INTERACTION EQUATION RATIO
with the rigid frame. The savings over the rigid frame will be
more significant if we account for the additional savings in Semirigid Connection
connection costs. The drift increases from 19.3 mm (0.76 in.) t = 25.4 t = 19.1 t = 12.7
for rigid frame to 22.9 mm (0.90 in.) for semirigid frame. Rigid mm mm mm
Apparently, in some cases connection design may be con- Member Section connection (1 in.) (3/4 in.) (1/2 in.)
trolled by drift limitations. Drift can be effectively reduced by (1) (2) (3) (4) (5) (6)
increasing the connection stiffness, which implies that the de- 1, 2 W14 ⫻ 38 0.899 0.942 0.959 0.988
signer must balance overall building drift and connection stiff- 3, 4 W14 ⫻ 30 0.821 0.765 0.742 0.786
ness in developing an economic framing system. 5, 6 W14 ⫻ 30 0.777 0.707 0.689 0.673
7 W14 ⫻ 38 0.908 0.730 0.688 0.756
8 W14 ⫻ 38 0.814 0.695 0.674 0.746
Three-Story, One-Bay Frame 9 W14 ⫻ 26 0.825 0.743 0.740 0.771
Drift, mm (in.) — 28.4 (1.12) 37.3 (1.47) 40.4 (1.59) 46.5 (1.83)
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

A three-story, one-bay frame is considered to demonstrate

the design application of the program PFNLFCAD and to
show the effect of variation in connection stiffness. It also TABLE 8. Member Sizes, Interaction Equation Ratios, and
Drift (Example 4)
shows a comparison of a rigid frame with a semirigid frame.
The frame configuration, dimensions, and loading are shown Rigid Connection Semirigid Connection
in Fig. 11. The beam-to-column connection is end plate with Member Section IER Section IER
column stiffeners, type 6. The beams are assumed to be lat- (1) (2) (3) (4) (5)
erally supported.
1, 2 W14 ⫻ 38 0.899 W14 ⫻ 38 0.990
The results are presented in Table 7 in the form of member
3, 4 W14 ⫻ 30 0.821 W14 ⫻ 30 0.773
sizes and interaction equation ratios for the frame members 5, 6 W14 ⫻ 30 0.777 W14 ⫻ 30 0.717
and frame drift. The interaction equation ratios show mixed 7 W14 ⫻ 38 0.908 W14 ⫻ 30 0.898
results for the columns. The IER values for the base columns 8 W14 ⫻ 38 0.814 W14 ⫻ 30 0.876
9 W14 ⫻ 26 0.825 W14 ⫻ 22 0.935
TABLE 6. Member Sizes, Interaction Equation Ratios, and Weight, N (lb) 22,143 (4,978) 19,919 (4,478)
Drift (Example 3) Drift, mm (in.) 28.4 (1.12) 45.0 (1.77)
Rigid Connection Semirigid Connection
Member Section IER Section IER show an increase due to base moments increasing under load-
(1) (2) (3) (4) (5) ing 2. The columns in second and third stories experience
1, 2, 3 W12 ⫻ 65 0.988 W12 ⫻ 65 0.970
some reduction in moments, as reflected in lower IER values.
4, 5, 6 W12 ⫻ 53 0.926 W12 ⫻ 53 0.947 The IER values for beams show a significant reduction from
7, 8, 9 W12 ⫻ 53 0.766 W12 ⫻ 53 0.769 rigid connection to 25.4-mm (1-in.) plate semirigid connec-
10, 11, 12 W12 ⫻ 53 0.630 W12 ⫻ 53 0.636 tion, which decrease further for 19.1-mm (0.75-in.) plate con-
13, 14 W24 ⫻ 62 0.948 W24 ⫻ 55 0.956 nection when the design control shifts from beam-end mo-
15, 16 W24 ⫻ 62 0.912 W24 ⫻ 55 0.929 ments to span moments. With further ‘‘softening’’ of the
17, 18 W24 ⫻ 62 0.903 W24 ⫻ 55 0.920
19, 20 W16 ⫻ 36 0.904 W16 ⫻ 31 0.910
connection to 12.7-mm (0.5-in.) plate, the positive moments
increase, as reflected in the increase in IER values. The results
Weight, N (lb) 95,888 (21,579) 89,049 (20,019) also demonstrate an increase in drift with increasing connec-
Drift, mm (in.) 19.4 (0.762) 22.9 (0.900)
tion flexibility. The connection stiffness may be determined
based on a preestablished drift maximum.
In this problem, the end plate thickness of 19.1 mm (0.75
in.) is selected for the connection based on the lowest IER
values for the beams (Table 7), and new design is performed
to achieve a more economical design with semirigid connec-
tions. The results in Table 8 show a weight reduction of 10%
for the semirigid-frame design over the rigid-frame design.
Additional savings are realized from the lower cost of semi-
rigid connections. However, the drift has increased from 28.4
mm (1.12 in.) for the rigid frame to 45.0 mm (1.77 in.) for
the semirigid frame.
SUMMARY AND CONCLUSIONS
An integrated analysis and design procedure has been pre-
sented for the design of PR-type steel frames. Several exam-
ples were undertaken to demonstrate the validity of the design
procedure and to show the effect of connection flexibility and
geometric nonlinearity on the frame design.
Connection flexibility and geometric nonlinearity influence
the distribution of forces in the structure. Both effects contrib-
ute to a significant increase in the drift of the frame. The
semirigid connections cause a shift in moments from the beam
ends to the span; and, as a result, the overall maximum mo-
ments decrease in beams, but increase in some columns as
compared to rigid frames. The softening of the connections
FIG. 11. Three-Story, One-Bay Frame results in a large increase in the building drift.
JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999 / 563

J. Struct. Eng. 1999.125:556-564.

 The connections may be used to balance positive and neg-
ative moments in beams to effect economic designs. Drift be-
comes a critical factor in the design of unbraced steel frames
and often controls the design of connections. An economic
framing system may be achieved by balancing the building
drift with the stiffness of the connections.

APPENDIX I. FIXED-END MOMENTS FOR

SEMIRIGID BEAMS
In Fig. 12, consider the beam AB with rotational springs of
stiffness RkA and RkB at ends A and B under concentrated load
P.
MA = RkA ␪rA
Downloaded from ascelibrary.org by Universidade Da Coruna on 02/12/14. Copyright ASCE. For personal use only; all rights reserved.

MB = RkB ␪rB (23)

Remove the end springs and consider the simply supported
beam under the three loadings shown. The rotations at end A
for the three cases are
Pab(L ⫹ b) MA L MB L
␪AP = ; ␪AA = ⫺ ; ␪AB = ⫺ (24)
6EIL 3EI 6EIL
Superposition of three cases at end A yields FIG. 12. Semirigid Beam
␪rA = ␪AP ⫹ ␪AA ⫹ ␪AB (25)
APPENDIX II. REFERENCES
Substituting from (23) and (24) in (25), we have
AISC. (1994). Load and resistance factor design specification for structural
MA Pb(L ⫹ b) MA L MB L steel buildings, American Inst. of Steel Construction, Chicago, Ill.
= ⫺ ⫺ (26) Barakat, M., and Chen, W. F. (1991). ‘‘Design analysis of semirigid
RkA 6EIL 3EI 6EI
frames: Evaluation and implementation.’’ Engrg. J., 28(2), 55 – 64.
Eq. (26) can be reduced to the following form: Bhatti, M. A., and Hingtgen, J. D. (1995). ‘‘Effect of connection stiffness
and plasticity on service load behavior of unbraced steel frames.’’
2MA (1 ⫹ 3␣A) ⫹ MB = 2MFA ⫹ MFB (27) Engrg. J., 34(1), 21 – 33.
Chen, W. F., and Lui, E. M. (1991). Stability design of steel frames. CRC
where Press, Boca Raton, Fla.
Dhillon, B. S., and Abdel-Majid, S. A. (1990a). ‘‘Interactive analysis and
EI EI Pab2 Pa2b design of flexibly connected frames.’’ Comp. and Struct., 36(2), 189 –202.
␣A = , ␣B = ; MFA = and MFB = (28) Dhillon, B. S., and Abdel-Majid, S. A. (1990b). ‘‘Design of steel frames
LRkA LRkB L2 L2 with PR type connections.’’ Proc., 8th Nat. Conf. on Microcomputers
Similarly, for end B we have in Civ. Engrg., University of Central Florida, Orlando, Fla.
Frye, J. M., and Morris, G. A. (1975). ‘‘Analysis of flexibly connected
MA ⫹ 2MB (1 ⫹ 3␣B) = MFA ⫹ 2MFB (29) steel frames.’’ Canadian J. Civ. Engrg., 2, 280 – 291.
Kishi, N., and Chen, W. F. (1986). ‘‘Data Base for Steel Beam-to-Column
Eqs. (28) and (29) yield Connections.’’ CE-STR-86-26, School of Civ. Engrg., Purdue Univer-
sity, West Lafayette, Ind.
MFA ⫹ 6␣B MFB Kishi, N., Chen, W. F., Goto, Y., and Matsuoka, K. G. (1993). ‘‘Design aid
MA = (30a) of semi-rigid connections for frame analysis.’’ Engrg. J., 32(3), 90 – 107.
(1 ⫹ 4␣A ⫹ 4␣B ⫹ 12␣A ␣B) Kim, S. E., and Chen, W. F. (1996). ‘‘Practical advanced analysis for
semi-rigid frame design.’’ Engrg. J., 35(4), 129 – 141.
MFB ⫹ 6␣A MFA King, W. S., and Chen, W. F. (1993). ‘‘LRFD analysis for semi-rigid
MB = (30b)
(1 ⫹ 4␣A ⫹ 4␣B ⫹ 12␣A ␣B) frame design.’’ Engrg. J., 32(4), 130 – 140.
O’Malley, J. W. (1997). ‘‘Design of steel frames with semirigid connec-
Similarly, we can show that (30a) and (30b) hold good for tions,’’ thesis submitted toward degree of Master of Engineering, Wid-
uniform and linearly varying load cases as well. ener University, Chester, Pa.

564 / JOURNAL OF STRUCTURAL ENGINEERING / MAY 1999

J. Struct. Eng. 1999.125:556-564.