A Practical P-Delta Analysis Method

for Type FR and PR Frames

ERIC M. LUI

I n the stability design of frameworks, attention must be order elastic analysis. In using this approximate method,
given to the additional moment developed as a result of the designer must perform two separate first order analy-
member (F-8) and frame (P-A) amplification effects from ses: A non-sway analysis in which artificial supports are
compressive axial loads. These secondary moments may provided to ^Le frame to prevent it from displacing later-
have a deleterious effect on the strength and stability of ally; and a sway analysis in which the frame is allowed to
the frame and so they must be reckoned with in the analy- displace laterally (Fig. 1). The first analysis gives M^^ that
sis and design processes. These are various approaches is multiplied by Bi to account for the P-8 effect and the
ranging from very simple to rather rigorous by which a second analysis gives Mi^ that is multiplied by B2 to ac-
designer can employ to account for these second-order count for the P-A effect. These two magnified moments
effects. In particular, the AISC/LRFD Specification^ are then added algebraically to obtain the design moment
adopts the so-called moment magnifier method in which (required flexural strength) M„ for the member in ques-
moment amplification factors B^ and B2 are introduced to tion. The advantage of this method is that it is straightfor-
account for the additional moments due to P-8 and P-A ef- ward and can be implemented easily in design since only
fects, respectively, as an alternative to a complete second- first-order analysis is required. The disadvantage of this

w. w.
1 ( j \ f f f t ! t
H, H. -H -H,
^

W2 1 Wo

' ' ^ f i f
H -H
H2 H,

wi 1 w, +
11
H
1
'1 f f f
H. m H,

777 777 TJT. /;///;/

Original frame Non-sway frame Sway frame
analysis for Analysis for
Mu

Eric M. Lui is Assistant Professor of tfie Department of Civil

Engineering, Syracuse, New York
Fig. 1. AISC/LRFD procedure for first-order frame analysis

THIRD QUARTER/1988 85
method is that it is appHcable only to rectangular rigid demonstrated in subsequent sections, the approach is
(Type FR) frames. This restriction arises as a result of the based on a firm theoretical background and provides good
assumptions and simplifications used in the structural results for frames subject to normal loading conditions.
model in deriving the expression for B2. Another disad-
vantage is that the additional moments arising from the
P-b and P-A effects may not be synergistic, since they nec-
THEORETICAL BASIS OF THE METHOD
essarily do not occur at the same location. The use of this
method will, therefore, in some cases overestimate the re- Consider the beam-column of flexural rigidity EI with no
quired strength for the member leading to an uneconomi- relative joint translation subjected to an axial force of F,
cal design. Furthermore, for multi-story, multi-bay end moments M^, M^ and an arbitrary lateral load of w as
frames, there is the question of where to place the artifi- shown in Fig. 2. The differential equation governing the
cial supports for the non-sway analysis, since placement of small displacement behavior of this member is given by
these supports at different locations will lead to slightly (2a).
different but noticeable results.
The aforementioned disadvantages can be circum- (1)
dx^ dx^
vented if one uses the preferred method contained in the
LRFD Specification by performing a direct second-order Upon rearrangement, Eq. 1 can be written as
elastic analysis to obtain M,,. In using this direct approach,
the designer often relies on computer programs capable of (2a)
performing second-order structural analysis that account dx^
for both the deterioration of member flexural rigidity where
under axial compression (F-8) and equilibrium of the de-
flected structure (P-A). d^
w' = -P (2b)
Although second-order structural analysis programs are dx^
available in the market, extreme care must be exercised in If we compare Eq. 2a with the differential equation of a
using these programs. Special attention must be paid to beam (Fig. 3),
such factors as modeling of the structure and setting toler-
ance limits for the analysis. It is important that a designer
possess some basic knowledge of nonlinear analysis and EI (3)
dx^
behavior of the structure before any attempt is made to
accept the results of the analysis. it is seen a beam-column differs from a beam by only
In this paper, a simple and practical method of second- the extra term w' given by Eq. 2b. Based on this obser-
order frame analysis using first-order structural analysis vation, we can, therefore, account for the beam-column
technique is proposed. The proposed method accounts for effect by applying a pseudo lateral in-span load of
the F-8 and P-A effects by the use of a fictitious or pseudo w' = -P(d^yldx^) to the member (Fig. 4). In other words,
lateral load. This method has the advantage over other ex- the beam-column shown in Fig. 2 can be analyzed as a
isting methods in that no special consideration is required beam shown in Fig. 4 provided that a fictitious lateral load
for the modeling of the structure—and the approach is ap- w' equal to the negative product of the axial force P and
plicable not only to rectangular rigid (Type FR) frames but the second derivative of the displacement d^yldx^ is ap-
also to non-rectangular, semi-rigid (Type PR) frames. As plied to the member.

i 1 I t t \

Fig. 2. Beam-column Fig. 3. Beam

86 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 If small displacements are assumed, the following rela- tain yet better values for P and M. Thus, an expression for
tionship holds the pseudo lateral load at the i-th cycle of calculation can
be written as
(4)
EI M,- (7)
EI
where M is the bending moment (considering second-
order effects) in the member. If the member in question is subject to relative joint
In view of Eq. 4, the fictitious or pseudo lateral load ex- translation (Fig. 5a), a secondary overturning moment
pressed in Eq. 2b can now be written as equals the product of the axial force P and the relative
joint displacement A induced in the member. This so-

-"(I EI
M (5) called P-A moment traditionally is taken into account by
replacing the axial force P by a pair of equal and opposite
Equation 5 indicates the pseudo lateral load can be ob-
tained from the second-order moment diagram by scaling
it down by a factor of P/EI. Since P and M are not known
in advance for a member in a structure, the exact value of
w' cannot be evaluated directly. However, as a first ap-
proximation, the first-order values for P and M can be
used. Using these values, an approximate value for w' can
be evaluated using

(6)
!-;)"•
in which the subscript 1 refers to the first cycle of calcula-
tion. By loading the member with this pseudo lateral load
together with the actual loadings, a first-order analysis
again can be performed on the structure from which a bet-
ter estimate of P and M can be obtained. Using these new
values P and M, a better approximation for w' can be eval-
uated. Using this updated value of w' and the actual load-
ings, another first-order analysis can be performed to ob-

(a)
w ' - - P ^ = ^

i i t f t r t *

P-A Effect
A (b) (c)
Fig. 4. Beam with pseudo-lateral, in-span load Fig. 5. Beam-column with relative joint translation

THIRD QUARTER/1988 87
shear forces equal PA/L (L is the length of the member) to obtain a psuedo joint load to be appUed to the struc-
acting at the ends of the member. Strictly speaking, this ture. For instance, if two columns meet at a joint, the
approach is not exactly correct. The pitfall lies in the fact pseudo joint load H' is obtained from
that by subjecting the member to a pair of end shears of
PA/L, only the P-A effect can be accounted for, whereas H' =
the P-h effect, which is also present in the member cannot \dxl B lower story \dxjA upper story (10)
be accounted for. In what follows, it will be shown that
with slight modification, the proposed pseudo load This pseudo joint load must be applied in such a direction
method can also be applied to members with relative joint as to cause an increase in deflection of the structure from
translation. The method accounts for both the P-b and its original configuration.
P-A effects simultaneously for these members.
To illustrate the rationale behind the method, let's con-
PSEUDO LATERAL LOAD METHOD
sider Fig. 5b in which the axial force P acting on the mem-
ber has been decomposed into components. The compo-
Member With No Relative Joint Translation
nent that acts along the chord (dashed-dotted line) of the
members is given by To demonstrate how the pseudo lateral load can account
for the member instability effect, cosider the beam-
P cos 9. P at the A-th end (8a) column in Fig. 6. The member is subjected to an axial force
and of P and a concentrated mid-span lateral load of Q. Under
the load Q, the member will deflect. The axial force will
P cos e^ ^ F at the B-th end (8b) act through this lateral displacement creating additional
respectively, and the component that acts perpendicular deflection and moment. This phenomenon is referred to
to the chord is given by as the P-8 effect. The maximum moment of this member
occurs at mid span and has a theoretical value (2) of
P sin 8^ - P OA - H~/')^ ^^ ^^^ ^'^^ ^"^ ^^^^ kL
2
and -''^max tan k_L M. (11)
/dy 2
P sin e^ - F e^ =- P at the B-th end (9b)
\dx where M^ is the first-order moment given by QL/4 and
respectively. The component of P that acts along the k = vP/EI. It is now desirable to obtain an approximate ex-
member chord (i.e. Eqs. 8a and 8b) gives rise to the P-h ef- pression for Mmax using the proposed approach and com-
fect and the component of P that acts perpendicular to the pare it with Eq. 11.
member chord (i.e. Eqs. 9a and 9b) contributes to the P-A To begin, let's ignore the axial load P and analyze the
effect of the member. It has been shown earlier the P-b ef- member subjected to Q only using first-order analysis
fect can be accounted for by applying the pseudo lateral technique. The result of this analysis is in Fig. 7a. Next,
load of w' to the member. To account for the P-A effect, scale the moment diagram of Fig. 7a by the factor P/EI to
what we need to do is to apply a pair of pseudo shear obtain the pseudo lateral load as shown in Fig. 7b. Load
forces equal P{dy/dx)^ and P{dyldx)B at the A-\h and B-ih the member by this pseudo lateral load together with the
ends of the member, respectively. This is shown in Fig. 5c. actual lateral load, analyze the beam again using first-
Note that by subjecting the member to a pseudo lateral order analysis. The moment at midspan (i.e. M^^^) now
load of w' in conjunction with the pseudo end shears, both becomes
the P-h and P-A effects can be accounted for simultane-
ously.
Since the end slopes (dy/dx)^ and {dyldx)^ are not
known in advance, it is necessary to use approximate val- a
ues from a first-order analysis. These values can be im-
proved upon in subsequent cycles of analyses.
It is important to note the directions of these pseudo — P-x
loads. The pseudo lateral load must be applied in the di-
rection to amplify the displacement of the member with L/2 L/2
respect to its chord, while the pseudo end shears must be
applied in the direction to cause the member to rotate in t
the same sense as the P-A moment.
If more than one member is connected at a joint, the y
pseudo end shears for these members must be combined Fig. 6. Beam-column subjected to concentrated mid-span load

88 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 Table 1. Comparison of {Mmax I Mo) exact with
M„ l^Y^ikLf M„ (12) iMmaxlMo)approx. for Supported Beam-Column

MmaxIMo
Table 1 shows a comparison of the theoretical value for kL
Mmax expressed in Eq. 11 with the approximate value Exact Approximate
given in Eq. 12 for the range of kL commonly encoun- Eq. 11 Eq. 12 Eq. 13
tered in nonsway frames. It can be seen that reasonably (one cycle) (two cycles)
good approximations are obtained for small kL values. 0.20 1.003 1.003 1.003
For large kL values, the approximate values begin to devi- 0.40 1.014 1.013 1.014
ate from the exact values. This derivation becomes more 0.60 1.031 1.030 1.031
pronounced as kL increases. To obtain a better approxi- 0.80 1.057 1.053 1.057
1.00 1.093 1.083 1.092
mation, a second iteration is performed. The moment dia- 1.140 1.120 1.137
1.20
gram obtained from the first iteration is scaled by the fac- 1.40 1.203 1.163 1.195
tor PIEL the resulting pseudo lateral load together with 1.60 1.287 1.213 1.268
the actual lateral load are then loaded on the member as 1.80 1.400 1.270 1.357
shown in Fig. 7c. Under these loadings, it can be shown by 2.00 1.557 1.333 1.467
elementary structural analysis that the mid-span moment
(M^a^) is given by

M. 1 + — (kLf + — (kLy M. (13)

12 ^ ^ 120 ^ ^
A comparison of Eq. 13 with Eq. 11 is shown on Table 1.
As shown, Eq. 13 gives a better approximation than Eq.

POL
4EI

2 3
P QL
48(EI)^

^
^
=m

+ +3
PQL PQL
48EI 48EI

(a) (b) +
PV^
480(EI)'

(c)

Fig. 7. P- 6 analysis by the pseudo-lateral load method

THIRD QUARTER/1988 89
12 to Eq. 11, especially for higher kL values. This process P will then act through this sidesway displacement creating
can be continued further until the desired accuracy is ob- additional sidesway and moment. This phenomenon is re-
tained. This is not attempted here. Instead, it will be ferred to as the F-A effect. Note that in addition to this
shown in the forthcoming that there is a physical signifi- P-A effect, the JP-8 effect is also present since the member
cance to the pseudo lateral load used in the proposed also deflects with respect to its chord. Using structural sta-
method. bility theory, it can readily be shown that the theoretical
If we expand Eq. 11 using Taylor series expansion, we moment at the fixed-end of the member is given by
obtain
Mf = (sec kL) M,, (15)
1 kL^^2_ kLY
M. 1 + M. where k = VPIEI
2 / 15 2 /

or

M„ (14)
'^Ti^'^^^'-Tr,^''^'^-
Upon comparison of the above equation with Eqs. 12
M L
and 13, it can be seen readily that by successive applica- o
tion of the pseudo lateral load, additional terms in the EI
Taylor Series expression of the theoretical value of M^ax
will be obtained. Thus, for each cycle of calculation, a bet-
ter approximation to Mj^^x will be obtained, since one
more term in the series will be generated. For ordinary
structures under normal loading conditions, the value of
kL will be small. Convergence of the series will be fast and
so one iteration (two cycles of calculations) will usually
suffice.

Member With Relative Joint Translation

Consider the cantilever beam-column shown in Fig. 8. This

member is subjected to a concentrated moment of M^ and
a vertical force of P at its free end. Under the action of
M„, the members will deflect sidesway. The vertical force (a)

+ +
'^^

PM L PM L
o o
EI 2EI

Fig. 8. Cantilever beam-column Fig. 9. P-h analysis by pseudo-lateral load method

90 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 To obtain an approximate value for Mp using the pro- real load M^ in Fig. 9b. Under these loadings, M/r is calcu-
posed method, the member is analyzed as a beam with the lated to be
concentrated moment M^ acting alone as shown in Fig. 9a.
The result of this analysis indicates that {dyldy)/^ = 0 and Mp 1 + \ {kLf M„ (16)
{dyldx)^ = M^L/EI. Consequently, in order to account for
the P-A effect, an end shear of PM^LIEI is applied to the A comparison of this approximate value of Mp with the
member. To account for the P-h effect, the moment dia- exact value expressed in Eq. 15 for a range of kL typical of
gram of Fig. 9a is scaled by the factor PI EI to obtain a sway frames is presented in Table 2. Good agreements are
pseudo lateral load oi w\ = PM^IEL These pseudo loads observed at low kL values. For high kL values, the ap-
are shown acting on the member in conjunction with the proximate value deviates from the exact value. However,

2 2 2 2
PM P M L P M L
o o o
^^ (EI)2 2(EI)^

+ +
PM L*" P'M L»
EI
3(EI)'

+ + +
4-
2 4
(c) PM L*" P M L
o o o
2EI
6(EI)^ 24(EI)'
Fig. 9. (continued)

THIRD QUARTER/1988 91
 Table 2. Comparison of (MFlMo)exact with regarding the structure geometry and member end condi-
{MplM,). for Cantilever Beam-Column tions were made in the derivation of the method, the
method thus is applicable not only to rectangular Type FR
MmaxIMo frames but also to non-rectangular Type PR frames. This
kL
Exact Approximate represents an obvious advantage over the moment magni-
Eq. 15 Eq. 16 Eq. 17 fier method contained in the current LRFD Specification
(one cycle) (two cycles) by which only rectangular Type FR frames can be handled.
0.20 1.020 1.020 1.020 In using the proposed method, it is necessary for the de-
0.40 1.086 1.080 1.085 signer to carry out first-order frame analysis on the struc-
0.60 1.212 1.180 1.207 ture. A number of first-order frame analysis schemes are
0.80 1.435 1.320 1.405 available in the literature. Examples are: the slope-
1.00 1.851 1.500 1.708 deflection method, moment distribution method and ma-
trix stiffness method. In all these methods, the analyst is
required to obtain values for the fixed-end moments to ac-
count for any in-span loads that might be present in the
improvement can be made by performing another cycle of member. Since in-span member loads are always present
calculation. This new cycle of calculation is depicted in in the pseudo lateral load method of frame analysis, a
Fig. 9c. The moment at the fixed-end is now brief discussion of these fixed-end forces is necessary and
will be presented in the following section.
M, 1 + \ {kLf + ^ {kLf M. (17)

It can be seen in Table 2 that Eq. 17 gives a better approxi- FIXED-END FORCES
mation of Mp than Eq. 16. Table 3a lists the general expressions for the fixed-end
A reader should recognize Eqs. 16 and 17 are the Taylor forces of several pseudo lateral load distributions that are
Series expansion of Eq. 15. Hence, the pseudo loads for most commonly encountered. Recall the pseudo lateral
sway members bear the same physical significance as that load is obtained from the moment diagram of a first-order
for non-sway members in that each set of new pseudo loads analysis, it follows that the order of the pseudo lateral load
represents an additional term in the power series expan- expression will be two higher than that of the real load.
sion of the exact solution. Thus, Case 1 in Table 3a represents the pseudo lateral
load for a member subjected to end moments only. Case
2 represents the pseudo lateral load for a member sub-
APPLICABILITY OF THE METHOD jected to a concentrated lateral load and end moments.
In the preceeding section, two simple examples were Case 3 represents the pseudo lateral load for a member
shown to demonstrate the use of the proposed method. It subjected to a uniformly distributed lateral load and end
should be mentioned that the method can easily be ex- moments.
tended to structures composed of an assemblage of frame If a member is subjected to a real lateral load whose
members. The steps that need to be followed are: order is higher than a uniformly distributed load (i.e. zero
order), the moment expression will have an order higher
1. Perform a first-order analysis on the structure. than two. For such cases, it may be preferable to obtain an
2. Construct bending moment diagram for each and every approximate rather than an exact expression for the fixed-
member of the structure. end moments of the resulting pseudo lateral load. This is
3. Obtain pseudo lateral in-span loads for all the members shown as Case 4 in Table 3b. The member is first divided
of the structure by scahng the moment diagram of the into n segments. An equivalent concentrated load is then
member in question by the factor PI EI where P is the calculated for each segment from the equation
axial force and EI is the flexural rigidity of the member.
For members subjected to sway, additional end shears w'iL
G, (18)
and, hence, pseudo joint loads must be calculated and n
applied to the joints of the structure according to Eqs.
9a, 9b and 10. In Eq. 18, w- is the value of w' at point /. Once Qf is calcu-
lated, the fixed-end forces can be evaluated from the given
4. Load the structure by these pseudo loads together with
expressions.
the actual loadings. Perform another first-order analy-
The expressions shown in Tables 3a and 3b are applica-
sis on the structure.
ble only if rigid connections are present at both ends of the
5. Repeat steps 2 to 4 if desired.
members. If semi-rigid connections exist at either end, the
By following this procedure, secondary moments which fixed-end forces must be modified to account for the pres-
must be accounted for in design can be obtained by first- ence of these connections. Denoting Rj^^ and Rf^^ as the
order analysis technique. In addition, since no assumption connection stiffness at the ^-th and B-\h ends of the mem-

92 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 Table 3a. Fixed-end Forces
[(6 - gQp + (4 - OLM + 2air] L" - [ ( 4 - a2)p + (6 - oi2)q + 2 a2r] L^
Mp M.
120 120

_ [(4 - a3)p + (2 - 0^3)^ + 2a3r] L M^^ + M^^ [(2 - oi^)p + (4 - a4)^ + la^r] L Mp^ + MpB
Vj.B =
12

Load Case Values of a's

1
q

^VA(
"^r—T—
M
1 fV.,. «! = a2 = a3 = a4 = 0

T "^ f ^ "''
^A 1 1 ^FB
2

r
a i = 2a^ (10 - 15a + 63^) + Sb^ (5 - 4b)

q oi2 = 3a^ (5 - 4a) + 2b2 (10 - 15b + 6b2)

p y^
: [ f \ ' r ^ a3 = a^ + Sab + 2b2
^ aL
1
bL
n FB
L a4 = 2a2 + Sab + b^
^A

3
r ^ p a r a b c) l a

""^^ q ^1 ^ Ct2 = 4
P r - ^
a3 = a4 = 2
j \ \
i
1
L/2
L ' '
1 L/2 ^ 1 0 '-
^FA'

hers respectively, the fixed-end moments for the member In these expressions, Mf^ and M/r^ are the fixed-end mo-
are modified to (3) ments of the member with rigid connections and Mp^ and
Mf-s are the fixed-end moments of the members with semi-
AEI \ 2EI rigid (PR) connections. The subscripts A and B refer to
LR, LR, the A-th and B-th ends of the member, respectively. Once
M'PA - (19a)
4£/ E/ \ 4 /£/ Mfr^ and MpB are calculated, the modified fixed-end
1 + 1 +•
J^^kB/ ^kARkB \ ^ shears Vf^ and Vps can be evaluated by considering mem-
ber equilibrium.
2EI AEI
L/? kA NUMERICAL EXAMPLES
M^«=- 7 (19b)
AEI AEI In this section, several frame examples will be given to
1 +
LRkA
1 +•
^Rks) RkARkB \ L f) demonstrate the validity of the proposed method.

THIRD QUARTER/1988 93
 Table 3b. Fixed-end Forces

Load Case Fixed-end Forces

^0
^Qi afb,
Mp , Mp

FA 4=1 t: A MFB
io 1 2 i n V)
l.Qiaf(ai + 35,)
Vj.^ =

" V
FA FB

Example 1 approximation to the more exact method of second-order

Figure 10a shows the geometry and loading conditions of frame analysis with just one iteration. For the purpose of
a pinned-base portal frame. All members of the frame are comparison, the moment magnifier method {B^ and B2
made of W12x65 sections. The loadings shown are fac- method) recommended in the L R F D Specification is also
tored loads. The frame is braced against out-of-plane presented. It can be seen that for this simple portal frame
bending. It is desired to determine the maximum mo- the results obtained using the proposed method are very
ment in the column and the beam considering second- comparable to those obtained using the L R F D moment
order effects. magnifier method.
To proceed with the proposed method, a first-order
analysis of the frame is performed from which value of the Example 2
axial force P and bending moment M are obtained for In this example, a three-story frame shown in Fig. 11a will
each individual member. Next, obtain the pseudo lateral be analyzed. As in the preceding example, all loads shown
in-span load for each member by scaHng the moment dia- are factored loads and the frame is braced against out-of-
gram down by the factor PI EL For members that experi- plane bending. Following the same procedure as before, a
ence relative joint translations (i.e. the columns), addi- set of pseudo lateral in-span loads is obtained by scaling
tional end shears equal to the product of P and the end the first-order moment diagrams by the factor PI EI and a
slope must be calculated and transformed into pseudo set of pseudo joint loads is obtained by first calculating the
joint loads. Load the frame with these pseudo loads to- end shears according to Eqs. 9a and 9b and then trans-
gether with the real loads as shown in Fig. 10b. Perform forming them to joint loads according to Eq. 10. Using
another first-order analysis on the frame. The results of these pseudo loadings in conjunction with the real load-
this analysis are presented in Table 4 in which maximum ings Fig. l i b , another first-order analysis is performed on
moments in the column and the beam are shown for vari- the frame. The results for the maximum column moments
ous methods of analyses. obtained using the proposed method and several other
As can be seen, the proposed method gives an excellent methods are shown in Table 5.

94 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 Table 4. Moment in Leeward Column and Beam Table 5. Moment in Leeward Columns of the
of the Simple Portal Frame (kip-in) Three-story Frame (kip-in)

Method of Analysis Column Beam Method of Analysis 1st Story 2nd Story 3rd Story

First-order 2506 2702 First-order 833 686 576

Analysis Analysis

Second-order 2537 2711 Second-order 845 695 579

Analysis Analysis

Proposed Method 2537 2710 Proposed Method 844 695 579

(1 iteration) (1 iteration)

Si, S2 Method 2535 2710 Si, 82 Method 963 717 587

0 .0883 k / i n

2.98 • - W14
1 X 130 - '

A m
A
X
CO
13 ft
(a) Frame geometry and loading 0^148 k/in

^f ' ' ' ' f

5.46
W21 X 4 4
LO

X 13 ft
00

0 . 148 k/in

' ' ' ' ^

5.46 W21 X 4 4

00
13 ft
14 0.1942 X
CO

TT 77" 77 7 7
UNITS FOR DISTRIBUTION LOADS:
25 ft
xlO"^ K/IN.

(a) Frame geometry and loading

(b) Frame subjected to real and pseudo loads

Fig. 10. Numerical example—simple portal frame Fig. 11. Numerical example—three-story frame

THIRD QUARTER/1988 95
 0 .0883 k / i n Although only one iteration is used, the proposed
method gives excellent results when compared to a
2.98 '
_ .1 1 .J second-order analysis. The LRFD moment magnifier
method, however, overestimates the column moments.
The overestimation is rather pronounced for the first- and
second-story column moments. This conservatism is at-
tributed to the fact that the maximum nonsway moment
0.148 k / i n and the maximum sway moment do not coincide at the
same location. For instance, for the first- and second-story
5.46
1 ' '' f 1'
1 1, J columns, (M„^)max and (Mi^)^.^^ occur at different ends of
the member. Depending on the degree of conservatism,
the LRFD moment magnifier method could lead to a less
economical design.

0.148 k / i n Example 3

1 In the preceding examples, the frames analyzed were all

5.46
1
-1 1 "iJ rectangular in geometry and the connections were as-
sumed to be rigid. To demonstrate the proposed method is
also appUcable to non-rectangular, semi-rigid frames, the
gable frame shown in Fig. 12 will be analyzed. All mem-
bers of the frames are made of W8x40 sections. The
frame is assumed to be braced against out-of-plane bend-
ing. While the connection at the apex of the frame is as-
77 rr Rea / loads 77 sumed to be rigid, the connections at the eaves of the
frame are modeled as semi-rigid. The connection stiffness

0.0553
1.128
r T-^ 0.0412 2.20

0.472
0.0663

0-0406 0.0635 0.0345 7.33

0.947
0.0195
0.0841
0.0649

1.26 0.711 0.024^/

0.323 0.00672

0.100

//// Pseudo loads ////

3.874 9.68
(Units for distributed loads: x 10~^ Klin.)
(b) Frame subjected to real and pseudo loads
Fig. 11. (continued)

96 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 Table 6. Moment in Girder and Leeward Column 0.167 k/in
of the Semi-rigid Gable Frame (kip-in)

Method of Analysis Girder Column

1 :: : i

3 ft

First-order 869 1167

Analysis
12 ft

Second-order 890 1187

Analysis

12 ft 12 ft
Proposed Method 890 1186
(1 iteration)
(a) Frame geometry and loading

0.167 k/in

assumed for these connections is R^ = 28525 kip-in./rad. J i i i

The frame is analyzed using a semi-rigid frame program
developed by the author. The results obtained for the
maximum moment in the girder and the column using sev- 12
eral different analysis schemes are presented in Table 6. It
can be seen that even with just one iteration, the proposed
method gives very good results for the maximum girder
and column moments. The LRFD moment magnifier
method is not used for this example, since the magnifica-
tion factor B2 is applicable only to rectangular rigid
frames.
Real loads

2.84

2.44
2.95

0.988

1.51

0.179

IUNITS FOR DISTRIBUTION LOADS:

_ . XlO"^K/lN. ^
1.06 rmr yfrr 6.97
Pseudo loads
(b) Frame subjected to real and pseudo loads
Fig. 12. Numerical example—gable frame

THIRD QUARTER/1988 97
 SUMMARY AND CONCLUSIONS NOMENCLATURE
A simple and effective method of second-order frame =Nonsway moment amplification factor
analyses was presented in this paper. The method ac- B2 = Sway moment amplification factor
counts for the F-8 and P-A effects by the use of pseudo E =Modulus of elasticity
loads obtained from a first-order analysis of the frame. H' =Pseudo joint load
The method is quite versatile in that it is applicable riot I = Moment of inertia
only to rectangular Type FR frames but also to nonrectan- k --VP/EI
gular Type PR frames. The validity of the method has L = Length
been demonstrated and it is therefore recommended for M =Moment
general use. Mp =Fixed-end moment
M'p = Modified fixed-end moment for the presence of
semi-rigid connection
=Maximum moment
Mo =First-order moment
p =Axial force
REFERENCES Q =Transverse force
American Institute of Steel Construction, Inc. Load Rk = Connection stiffness
and Resistance Factor Design Specification for Struc- = Fixed-end shear
tural Steel Buildings 1986, Chicago, III. Vf = Modified fixed-end shear for the presence of
Chen, W.F. and E.M. Lui Structural Stability- semi-rigid connection
Theory and Implementation Elsevier Science Publish- w = Distributed load
ing Co., 1987, New York, N.Y. w' = Pseudo lateral in-span load
Lui, E.M. and W.F. Chen Steel Frame Analysis with 8 = Deflection of member with respect to its chord
Flexible Joints Journal of Constructional Steel Re- A =Deflection of member with respect to its initial
search, to be published, Essex, England. position

98 ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION