You are on page 1of 9

Compurers & Sfrucfures Vol. 53, No. 4, pp. 839-847.

1994
Copyright Q 1994 Ekvier Science Ltd
Pergamon 0045-7949(94)E0205-G Printed in Great Britain. All rights wscrvod
co45-7949194 97.00 + 0.00

STABILITY ANALYSIS OF FLEXIBLY CONNECTED


THIN-WALLED SPACE FRAMES
G. E. Bla~ford
Department of Civil Engineering, University of Kentucky, Lexington, KY 40506-0046, U.S.A.

(Receiued 30 June 1993)

Abstract-The elastic stability (bifurcation) response of space frames, including warping deformation and
flexible (semi-rigid) connections, is considered. Calculation of the elastic stability load is based on a linear
eigenvalue analysis. A parametric study of the effect of connection rigidity for a portal space frame of
variable geometry subjected to gravity and various wind loads is presented to provide insight into the
stability behavior of space frames.

Ettouney and Kirby[4] introduced the concepts of


continuous warping and partially restrained warping.
With the introduction of load and resistance factor They added a warping degree of freedom to the
design (LRFD) for steel structures, the analy- conventional six degrees of freedom at each beam
sis/design of frame structures needs to include a element node to account for the warping effects and
reasonable model of the joint behavior since only a warping restraint factor (ratio of the warping
partially restrained (RR) and fully restrained (FR) moment or bimoment of the partially restrained
connections are considered. It is the purpose of this warping connection to the bimoment obtained for a
paper to present some elastic, space frame stability fixed warping connection) to prescribe the degree of
(bifurcation) results which include the influence of warping restraint at each end of the member. Yang
flexible connection behavior and warping defor- and McGuire [l] introduced the concept of the warp-
mation. A stability load evaluation through a linear ing indicator (ratio of the warping displacement of
eigenvalue calculation can be very valuable in deter- the partially restrained warping connection to the
mining the structure load carrying capacity. The warping displacement obtained for an unrestrained
eigenvalue and eigenvector calculation can also or free warping connection) to prescribe the degree of
provide insight into the critical structural response warping restraint at each end of the member. Both
mode. the warping restraint factor, aR, and the warping
The elastic and geometric space frame element indicator, a,, vary between zero and one. However,
stiffness matrices developed by Yang and the numerical values are at the opposite ends of the
McGuire [ 1,2] for doubly symmetric beam sections, restraint spectrum, i.e., a, = 1 for free warping and
which also include warping defo~ation behavior, ar = 0 for fixed warping, whereas aR = 0 for free
are used to discretize the beam members. A separate
warping and aR = 1 for fixed warping. Yang and
connection element is used to model flexural joint
McGuire pointed out that the warping indicator
behavior and the warping resistance of the joint
should be easier to measure ex~rimentally, although
connections is modeled after the developments of
no such experiments have been performed. The sol-
Yang and McGuire (11. Sample results are presented
ution generated by Yang and McGuire is used to
for space frame structures composed of doubly sym-
model the warping restraint behavior in terms of a
metric thin-walled members to demonstrate the influ-
warping fixity factor, ar, which is defined as one
ence of column o~entation and semi-rigid connection
minus the warping indicator, i.e., uF = 1 - a,.
behavior on the linear elastic bifurcation load. Lastly,
The beam element elastic stiffness equations for the
conclusions pertinent to the stability behavior of
element of Fig. 1 can be represented as
flexibly connected space frames are presented.

STIFFNESS EQUATIONS

A limited number of investigators have included in which IA”) = @,, , q, uziI Q,,+@,,, fL , s2, uy2,u,, ,
warping deformation in their analysis of three-dimen- Ox,,@,,, @,,, x, , x2>=,the element displacement vector,
sional member response. Barsoum and Gallagher [3] which includes the conventional beam displacements
considered warping effects in the analysis of torsional u and 0 along with a rate of twisting or warping
and torsional-flexural stability beam problems. displacement 2, i.e., x = df?Jdx; {Fe) = (F,, , I;,, ,

839
840 G. E. Blandford

Fig. 1. Beam element displacements and forces.

F:,>M.y,,&I, M:,> Frzt 41, Fzz,Mxz, Myz, Mz,, 4, placements of node 1, or vice versa. (This slaving
Bz)‘, the element force vector corresponding to the procedure simply enforces transverse displacement
element displacement vector {A”}, which includes the continuity between the beam and column connected
conventional beam forces F and M along with the elements). The only connection deformation that can
bimoment B; subscripts x, y, and z signify local take place is torsion and bending. Three stiffness
coordinate directions; subscripts 1 and 2 signify the coefficients are used to account for torsion and
node number; [kr] is the 14 x 14 beam element elastic bending. The connection element elastic stiffness
stiffness matrix of Yang and McGuire [l] for doubly equations are expressed as
symmetric beams and includes the warping resistance
at the beam connections; [ 1, { } and ( ) signify
rectangular, column and row matrices, respectively;
and superscript T denotes matrix transpose. Yang where
and McGuire’s elastic beam stiffness matrix is exactly
the same as obtained by Barsoum and Gallagher [3]
with the exception that a diagonal warping restraint
matrix is added to the elastic stiffness matrix of
eqn (1). (connection element stiffness matrix)
Node point warping torsion displacements, in-

1
cluded in the element elastic stiffness matrix, do not k “I
follow the usual coordinate transformation rules. I” = k0) .
Ettouney and Kirby [4] recommended expressing the
standard node point displacements (u and 0) in terms [ ku:
of the global coordinate system and leaving the
{P} = (OX,, O,,, &,, 0.r2r$*, f322)T, the element dis-
warping deformation x in the local coordinate sys-
placement vector shown in Fig. 2;
tem. This scheme has been adopted, but a discontinu-
ity in the member warping displacements exists at the {M’} = CM,, , MvI, Ml,, Wz. M,.,, ML2>T. the el-
ement force vector consistent with the element dis-
structure joints. Ettouney and Kirby [+] and Yang
placements of Fig. 2; and the subscripts on k signify
and McGuire [l] used static condensation on the
the rotational stiffness component. Since doubly sym-
discontinuous element warping displacements. Static
metric beam elements are considered, no coupling
condensation is not required in the present formu-
between major and minor axis bending or torsion and
lation provided a beam-to-column connection el-
bending takes place as may happen with singly
ement is used at each joint.
symmetric or unsymmetric sections. However, the
Modeling the connection behavior involves using a
connection element can be modified to represent the
two node element with zero length in terms of the
coupled behavior. Standard coordinate transform-
rotational degrees of freedom 0,) 0, and I!?~,as shown
ation procedures are applied to eqn (2) to express the
in Fig. 2. It is assumed that no axial or transverse
element equations in the global coordinate system.
deformation occurs within the connection element.
The connection element allows a natural method
As a result, the translational displacements of node 2
for modeling warping deformation discontinuity at
for the element are slaved to the translational dis-
the member joints. Discontinuity is provided by
having a beam warping displacement at one end (e.g..
node 2) of the connection element and a column
warping displacement at the other end (e.g.. node 1).
Since two distinct nodes are provided at the same
location (zero length element) to model the warping
displacements, as shown in Fig. 3, there is no need to
perform static condensation. Warping displacements
are modeled in the local coordinate system and
Fig. 2. Connection element displacements and forces. assembly follows the usual direct stiffness procedure.
Stability analysis of thin-waft space frames 841

somewhat complex and time consuming. Conse-


quently, a determinant search technique is used to
ensure convergence to the neighborhood of the lowest
positive eigenvalue prior to switching to the inverse
iteration strategy.
Al m To ensure that all the displacement variables are of
Fig. 3. Warping discontinuity between joint connected a similar magnitude, the element matrices [kE] and
elements. [k,] are scaled as

If a joint is prevented from expe~encing warping


deformation (fixed warping condition) then the nodal
warping displacements are simply set to zero and no I&J = IwlT’[kJ[wl (5b)
connection element is required to model discontinu-
ous warping behavior. in which [0(1)]=diag(l, 1, 1, l/i, l/f, l/L 1, 1, 1, l/1,
The beam and connection element stiffness l/i, I/l, l/1’, 1/12)and [is the maximum element length
matrices are assembled to yield the structure stiffness for the discretized structure.
equations via direct stiffness assembly, i.e.,
Determinate search
[&I{4 = V’L (3) Determinate search techniques consist of ‘search-
ing’ for the eigenvalue, or load factor A, that satisfies
in which [&I is the structure elastic stiffness matrix, the equation
(A) is the structure dispIa~ement vector, and (P) is
the structure node point force vector. Equation (3) is p(ll) = det([R,] + A[&]) = 0. (6)
used to calculate the unknown node point displace-
ments. Once the displacements are computed, the To reduce floating point errors, the calculation of
beam element forces are calculated using eqn (1). The p(A) is scaled as
beam element forces are then used to calculate the
14 x 14 element geometric stiffness matrix [&fSSgiven
in Yang and McGuire [Z] for each beam element P@) = fi dk(W,, (7)
i=l
(superscripts ss signify a semi-tangential represen-
tation for both the bending and torsion moments). in which diag(i) is a diagonal entry in the factorized
Transformation of the beam element geometric stiff- form of the global stiffness matrix, &]f ;C[&], q is
ness matrix is performed in exactly the same fashion the number of equations in the system and
as described for the beam element elastic stiffness s, = (det[Q)“2. The search is initiated by calculating
matrix. No geometric stiffness exists for the connec- the determinant of the elastic stiffness matrix, which
tion element. Consequently, direct stiffness assembly is defined by & = 0, as shown in Fig. 4, and
only on the beam element geometric stiffness matrices ~(1, = 0) = 1 based on the scaling used in eqn (7). An
is required to construct the structure geometric stiff- elastic analysis is performed using the applied struc-
ness matrix, [K&l. ture loading from which the member forces are
calculated and used to evaluate the geometric stiffness
EIGENVALUE SOLUTION matrix, [&]. Referring to Fig. 4, a new determinant
p(h,) is evaluated with 2. = 1,. Secant iteration is then
The linear eigenvalue problem consists in solving used to obtain a new choice for A, shown as 1, in
the equation Fig. 4. The secant iteration equation for iteration i is

for the smallest positive eigenvalue (bifurcation load


multiplier) & and the eigenvector (buckling mode)
(4 I_. A solution algo~thm has been developed com- P(X)
bining a determinant search routine based on secant t
iteration with a ‘shifted’ inverse iteration strategy [5]. P(XtyO)
Application of the inverse iteration strategy can
result in a negative eigenvalue which is not physically
possible in structural systems. Since the lowest posi-
tive eigenvalue, or load magnitude, is desired for
structural design considerations, some adjustment is P(X,) --

required to obtain the desired results. Modifications h


x0 4wc A’, x”, x
can be applied directly within the inverse iteration
strategy through ‘shifting’ [6], however, they become Fig. 4. Secant iteration.
842 G. E. Blandford

Convergence to the correct value 1: (shown as A, in p(l) < 1 x 10-8. (9b)


Fig. 4) is normally automatic; superscript D signifies
determinant search and subscript c signifies a con- Inverse iteration

verged value. Inverse iteration provides an effective method for


To ensure that a good secant step is performed determining a structure’s eigenvector while simul-
from I,, ~(1,) must be less than 0.90, otherwise i, is taneously evaluating the eigenvalue I to the desired
recursively doubled until this condition is satisfied. tolerance. Using the calculated determinant search
Upon satisfying this condition, iz is calculated based eigenvalue to shift the structure stiffness matrix,
on eqn (8). inverse iteration is applied to the ‘shifted’ eigenvalue
It is possible within the determinant search pro- problem expressed as
cedure for the predicted eigenvalue to exceed the
critical eigenvalue, i.e., the eigenvalue begins converg- KKI + ZKI) + ~t_KGll~~~
= 04~ (10)
ing to a higher buckling mode than the lowest mode
which is desired. Schematically the higher modes are where 1: is the previously determined approximation
represented by Lf and is in Fig. 4. Further modifi- of the structure eigenvalue obtained from determi-
cation of the basic determinant search algorithm nant search and 1: is the eigenvalue of the ‘shifted’
eliminates the possibility of converging to any eigen- matrix to be obtained from inverse iteration.
value other than 1, of Fig. 4. Based on the Sturm Formulation of the inverse iteration algorithm is
sequence [6], any eigenvalue i, resulting from secant expressed as
iteration generating one or more negative diagonals
in the factorized form of the global stiffness matrix, ]Q{4}i+ I = -x+ ,K?l{41, (11)
[&I + L,[&], is greater than the eigenvalue corre-
sponding to the lowest buckling mode. To guarantee in which
convergence to A, in Fig. 5, 1, resulting from a secant
iteration causing one or more negative diagonals will @I = WEI+ w&1~ (12)
not be used in subsequent iterations. Instead, a value
i, will be used such that 2, = 1, ~, + (Ai- i.,_ ,)/2 Iterations on the eigenvector (4) begin with
where i,_ 1 is the last ‘good’ approximation of 1 in {4}i=, = {I} where (I} is a vector in which all entries
eqn (6); a ‘good’ i,_, being one for which equal one. Introducing {y}, = [&]{4}i into eqn (11)
p(L!_ ,) < 0.90p(l,_2) with no negatives on the diag- yields the iteration strategy on {&jl+, expressed as
onal of the factorized global matrix. Secant iteration
is not applied again until E.,= i, yields no negatives on [Q{+)i+l ’ {Y>d. (13)
the diagonal.
Approaching the solution of the eigenvalue prob- Defining
lem by limiting i to the region 0 < 1 < i, guarantees
convergence to 1:. Additionally, it effectively ‘shifts’ {Y)i+, = K1{4h+1 (14)
the structure stiffness matrix into the neighborhood
of &. With the proper ‘shifting’ completed, the and substituting appropriately into eqn (10) yields
inverse iteration algorithm provides an efficient
method by which final convergence is achieved to the
desired eigenvalue and its corresponding eigenvector.
The transition from determinant search to inverse
iteration occurs when The vector {y},+, is updated as

(I,-LJ<o,ol {y},+, =sign(&+,)%, (16)


(94
/I, ’ 6 !+I

in which ai+, = (+),+i{j}i+,; sign(Gi+,)= 1 for


or
6i+,>0;sign(6,+,)=-1for6,+,<0;and] Isignify
absolute values. Iterations with eqns (1 l)-(15) con-
p(X) tinue until convergence:

(A+, -A)<t
(17)
Ai,, ’ i’
in which Lk= li+, is the eigenvalue of the ‘shifted’
matrix [a and c1 is the desired tolerance on the
eigenvalue, tl = lo-” in which s is the desired number
of eigenvector significant digits. The critical eigen-
Fig. 5. Secant iteration with backtracking line search. value and scaled eigenvector are calculated as
Stability analysis of thin-walled space frames 843

n,,=np+n; (184

{()..=sign(ai+,)$?&. (18b)
in which 0 < v < 1 is the connection flexural fixity
,+I
factor (v = 0, simple connection; v = 1, rigid connec-
NUMERICAL RESULTS tion); and the section properties are taken to be the
connected beam properties. A warping restraint fac-
Based on the outlined procedure, a computer pro- tor 0 < uF < 1 is used to represent the degree of flange
gram has been prepared to investigate the stability warping restraint, e.g., CI~= 0 means the flanges are
response of space frames including warping defor- free to warp whereas for ur = 1 the flanges are fixed
mation and flexible connections. A parametric study
against warping.
on the effect of connection rigidity for a portal space The numerical results in this paper utilize three
frame is presented.
connection conditions: v = (I~ = 1, k,, = lOI* (rigid
The frame layout and load conditions for the
case); v = 0.625, ur = 0.5, k, = 3GJ/L (semirigid);
portal space frame are shown in Fig. 6(a). Column
and v = 0.25, aF = 0, kox = GJ/L (flexible), unless
orientations has been arranged into the three different
stated otherwise. Wind load multipliers are restricted
patterns shown in Figs 6(b), 6(c) and 6(d), which were
to CI,/I = 0 or 0.25. Based on the results of Carlberg
also considered by Razzaq and Naim [7]. The girder
et al. [8], a four element/member discretization is used
members are oriented such that the weak cross-sec-
for all analyses.
tion axis coincides with the global y-axis. Two wide Previous results utilizing the portal space frame of
flange sections, a W12 x 53 (A = 15.6in2,
Fig. 6(a) [8] focused on using an IMSL routine to
Z:; = 425 in4, ZyJ= 95.8 in4, .Z = 1.58 in4, and
calculate the buckling loads for a lateral twist loading
C,,.= 3160 in6) and a W36 x 160 (A = 47.0 in*,
case (loading such that a resultant torque T = 2yPL,
Z2-_= 9750 in4, Z, = 295 in4, J = 12.4in4 and
0 < y < 1, goes through the frame centroid) for a
C,. = 90,200 in6) are used for all the members in the
parametric study on the influence of the full range of
space frame with an elastic modulus of 29,000 ksi and
connection and loading conditions, i.e., 0 < v ,< 1,
a shear modulus equal to 11,600 ksi. Warping defor-
O,<a,<l, GJ/L<kox<lO’*, and O<T,<2PL. A
mation can be significant for I-section members if
major disadvantage of using IMSL is that all eigen-
pL < 2 (e.g., Yang and McGuire [1]) where
values and eigenvectors are calculated rather than
p = (GJ/EC,,,)“*. The W12 x 53 is less susceptible to
just the critical pair. Furthermore, IMSL does not
warping deformation (PL = 2.00) as compared to the
utilize the sparseness of the stiffness equations. The
W36 x 160 section (PL = 1.05) for L = 144in.
determinant search/inverse iteration procedure re-
Bending connection stiffness coefficients (k, and
sults in the same critical solution as obtained using an
k,) are calculated as
IMSL eigenvalue/eigenvector routine.
Table 1 presents the critical buckling load ratios
( = P,,IP,; P, s n*EI,/L*, an Euler buckling load)
for the gravity load case (c( = ,!I = 0) using the
W12 x 53 and IV36 x 160 sections with H = L for
the three column patterns of Fig. 6. Table 1 shows
that column pattern A is the least stiff frame and
BP- column pattern C results in the stiffest frame.
Table 1 also shows that the gravity loaded frame
composed of the stiffer IV36 x 160 section members
is much less sensitive to the connection condition
than the frame composed of the W12 x 53 sections,

0 Triple Node
Table 1. Critical buckling load ratios for the cubic portal
space frame subjected to gravity loading and a = /I = 0
Ia) Frame Geometry ond Loadlng Buckling load ratio
Connection Column
detail pattern w12 x 53 W36x 160
A 0.927 0.988
Rigid B 1.130 1.421
C 1.909 3.945
A 0.854 0.976
Semi-rigid B 0.982 1.184
C 1.653 3.203
(b) Pattern A (Cl Pattern 6 (d) Pottrrn C A 0.659 0.932
Fig. 6. Portal space frame geometry, loading, and column Flexible B 0.714 0.999
patterns. C 1.207 3.018
844 G. E. Blandford

Table 3. Critical buckling load ratios for the cubic portal


space frame subjected to gravity loading plus a = 0 and
b = 0.25
Buckling load ratio
Connection Column
.--w-l cc-cl
detail pattern w12 x 53 W36 x 160
(a) A: W12x53 (b) A: W36x160 A 0.900 0.979
Rigid B 0.948 1.150
C 1.590 3.019
A 0.591 0.844
Semi-rigid B 0.576 0.763
C 0.737 1.406
A 0.435 0.689
Flexible B 0.422 0.655
C 0.544 0.934
(c) B: W12x53 (d) B: W36x160

load cases for the cubic portal space frame. Compar-


ing the results of Table 2 with Table 1 shows that for
c( = 0.25, column pattern A essentially exhibits no
change in its eigenvalue magnitude for a rigid connec-
tion case using WI2 x 53 members and for all con-
nection cases using W36 x 160 members. Otherwise,
(e) C: W12x53 (f) C: W36x160 the influence of wind loading reduces the structure
Fig. 7. Critical buckling modes for gravity loading and rigid buckling load (compare Tables 2-4 with Table 1), as
connections. should be expected. Comparing Tables 2 and 3 shows
a reduced buckling load capacity for both column
even though both sets of results are similarly scaled. patterns A and B, whereas column patterns C is
An explanation of this observation is that the ratio of invariant to a single direction wind load as rep-
the W36 x 160 to W12 x 53 strong axis moments of resented by u = 0.25, /J = 0 and c( = 0, /3 = 0.25.
inertia is 22.94 whereas the ratio of the weak axis Table 4 shows that applying a wind load in both
moments of inertia is only 3.08. lateral directions (a = fi = 0.25) results in the smallest
Eigenvector plots for the eigenvalues of Table 1 are buckling loads. Tables 2-4 also show that even
given in Fig. 7 for the rigid connection analyses. column pattern C exhibits a relatively large buckling
These plots for the other connection cases are similar load decrease with decreasing connection stiffness,
to those shown. The solid circles in all the eigenvector which differs from the results of Table 1. Application
of the lateral load forces reduces the self-bracing
plots identify the girder nodes and the member
orientations match those of Fig. 6. Letters A, B, C effect of column pattern C.
Figures 8-10 present the eigenvector (buckling
define column patterns A, B, C. Figure 7 shows that
mode) plots corresponding to the eigenvalues of the
the frame composed of W12 x 53 members experi-
rigid connection cases in Tables 2-4. Figures 8 and
ences greater twisting deformation than the frame
9 show the buckled mode shapes for the frames
composed of W36 x 160 members, subjected to grav-
composed of W 12 x 53 and W36 x 160 members
ity loading only.
subjected to gravity as well as wind loading in the
Tables 2-4 give the critical buckling ratios for the
z-axis direction (a = 0.25) and x-axis direction
frames subjected to gravity loading plus various wind
(fi = 0.25). Buckling modes for the semi-rigid and
Table 2. Critical buckling load ratios for the cubic portal
space frame subjected to gravity loading plus a = 0.25 and Table 4. Critical buckling load ratios for the cubic portal
&I=0 space frame subjected to gravity loading plus a = p = 0.25
Buckling load ratio Buckling load ratio
Connection Column Connection Column
detail pattern w12 x 53 W36x160 detail pattern Wl2 X 53 IV36 x 160
A 0.927 0.988 A 0.899 0.978
Rigid B 1.093 1.407 Rigid B 0.929 1.143
C 1.588 3.019 C 1.448 2.789
A 0.785 0.976 A 0.588 0.836
Semi-rigid B 0.721 1.158 Semi-rigid B 0.568 0.752
C 0.737 1.406 C 0.721 1.271
A 0.489 0.931 A 0.433 0.686
Flexible B 0.488 0.988 Flexible B 0.418 0.654
C 0.544 0.934 C 0.529 0.93 I
Stability analysis of thin-walled space frames 845

,‘i
.

/ C-C-4 ---ccl
0 ,
/ (a) A: W12~53 (b) A: W36x160
(a) A: a=0.25 (b) A: paO.25
Pr
c 1

(c) 8: W12x53 (d) B: W36x160


(c) B: a=0.25 (d) B: /S-O.25

(e) C: W12x53 (f) C: W36x160


(e) C: a=0.25 (f) C: /?=0.25 Fig. 10. Critical buckling modes for gravity plus full lateral
Fig. 8. Critical buckling modes for the frames composed of loading with rigid connections.
W12 x 53 members with rigid connections subjected to
gravity plus lateral loads.
mode shown in Fig. 8(a) to a lateral beam buckling
flexible connection cases are similar to the rigid mode as shown in Fig. 8(f). Both Figs 8 and 9 show
connection cases except for column pattern A. that the buckling mode changes significantly with the
Column pattern A for a = 0.25 goes from the sway change in direction of the applied lateral forces;
compare the left and right columns of each figure.
Figure 10 presents the buckling mode shapes for

a
gravity plus full lateral loading (i.e., tl =/I = 0.25)
and the rigid connection eigenvalues of Table 4.
ccc
Column patterns A (Figs IOa, and lob) essentially
buckles in a sway mode. Column patterns B and C for
the WI2 x 53 member frame buckle in a twist mode;
ccecl
column pattern B for the W36 x 160 member frame
also twists, but the larger size columns provide
(a) A: a=O.25 (b; A: @=0.25
greater rigidity in the strong axis direction. The
W26 x 160 member frame in column pattern C (Fig.
10(f)) essentially provides internal bracing resulting
in lateral buckling of each girder.

Table 5. Gravity loaded critical buckling load ratios for the


cubic portal space frame with semi-rigid connections and
various warping fixity factors (+)
(c) 8: a-0.25 (d) B: /I=O.25
Buckling load ratio
Warping Column
detail pattern WI2 x 53 W36 x 160
A 0.854 0.976
UF= 1.0 B 0.984 1.189
C 1.654 3.897
A 0.854 0.976
CIF= 0.5 B 0.982 1.184
(a) C: a==0.25 (f) C: /ho.25 C I.653 3.203
Fig. 9. Critical buckling modes for the frames composed of A 0.854 0.976
W36 x 160 members with rigid connections subjected to aF = 0.0 B 0.982 1.183
gravity plus lateral loads. C 1.653 3.078
846 G. E. Blandford

Table 6. Gravity plus full lateral loaded critical buckling


load ratios for the cubic portal space frame with semi-rigid
connections and various warping fixity factors (L+) +ASR; *AFL
I BSR ; l BFL
Buckling load ratio
A CR ; t CSR : l CFL
Warping Column _________~
detail pattern WI2 x 53 W36 x 160
A 0.593 0.843
ar = I .o B 0.578 0.793
c 0.733 1.388
A 0.588 0.836
UF = 0.5 B 0.568 0.752
C 0.721 1.271
A 0.587 0.835 H/L
UF = 0.0 B 0.565 0.739 (a) W12x53
C 0.717 1.192
OAR; +ASR; l AFL
q BR : x BSR ; l EFL
Tables 5 and 6 present the buckling load ratios for a CR : * CSR : . CR
the gravity and gravity @us full lateral loading cases
assuming semi-rigid bending connection stiffness and
a variety of warping fixity factors. Tables 5 and 6
show that the buckling loads for the cubic portal .;2
space frame composed of W12 x 53 members is
s
insensitive (less than a 2% change) to the degree of ml
connection warping restraint. Frames composed of 1 I I
W36 x 160 members arranged in a column pattern A 850 1 .oo 1.50
are also insensitive to the degree of connection warp- H/L
ing restraint. Column pattern B exhibits a greater (b) W36x160
warping restraint influence (maximum buckling load Fig. 12. Buckling load ratios versus column height for the
reduction z 5% for the full lateral load case). Unlike gravity plus full lateral loaded frames.
column patterns A and B, column pattern C is
sensitive to the degree of warping restraint for both
7 gravity and full lateral loading. There is a substantial
0 AR ; + ASR; l AFL
reduction in the frame buckling load for column
6 q BR ; H BSR ; l BFL
,g pattern C when the connection warping restraint is
A CR ; l CSR ; . CFL
:5 changed from fixed (ur = I) to semi-rigid (uF = 0.5);
17.8% for gravity loading and 8.4% for gravity plus
full lateral loading. A smaller reduction occurs as the
connection is further reduced to be unrestrained in
warping (aF = O), i.e., 3.9% for gravity loading and
6.2% for gravity plus full lateral loading.
Figures 11 and 12 show the influence of column
height on the gravity and gravity plus full lateral
8.5 1.0 1.5
H/L
loaded frames composed of W12 x 53 and
(a) ~2x53 W36 x 160 members using the three column patterns
(A, B and C) and three connection details (R = rigid,
16
OAR; + ASR ; l AFL SR z semi-rigid, and FL = flexible) discussed at the
0 14 0 BR ; Y BSR ; . BFL beginning of the numerical results section. The grav-
.s 12 A CR ; li CSR ; 4 CFL ity loading results (Fig. 11) show that the column
pattern strength is arranged as A < B 4 C. Thus,
-cl 10
changing column heights did not change the gravity
58 loading buckling strengths from the results provided
2’6
‘2
in Table 1 for H = L.
2 4 Figure 12 shows that the column pattern buckling
m 2 strengths are arranged as: A < B < C for the rigidly
(R) connected frames with H = L and H = 1.5L
8 .5 1.0 1.5 (same as the gravity loaded cases); B < C < A for
H/L both the semi-rigidly (SR) and flexibly (FL) con-
(b) W36x160 nected frames composed of W36 x 160 members with
Fig. 11. Buckling load ratios versus column height for the H = 0.5L;and B < A < C for all other frames sub-
gravity loaded frames. jected to gravity plus full lateral loading.
Stability analysis of thin-walled space frames 847

n
C- axis column orientations results in increased buckling
b 4 load capacity for the gravity loaded portal space
L 1 frame structures. Decreasing the column height natu-
1 \ rally led to an increase in buckling strength and
.-.-. increasing the column height resulted in a buckling
strength reduction.
(a) A: W12x53 (b) A: W36x160 For the point force cases considered, column pat-
tern orientation has been shown to be a significant
C-
parameter in determining the importance of warping
deformation on the critical elastic frame buckling

(c) B: W12x53 (d)


El B: W36x160
load. Cross-section properties have also been shown
to determine the significance of warping deformation
on the buckling loads, but not to the same extent as
column pattern orientation of the portal space
frames.
Column orientation and connection rigidity have
been shown to strongly influence the elastic critical
buckling mode, particularly if lateral point loading is
i also considered. Column pattern A (columns all

*i
aligned in the same direction) is the most susceptible
to buckling mode changes, whereas column pattern C
Ll -0-c (alternating strong and weak axis orientation for each
of the four columns) is the least susceptible to
(e) C: W12x53 (f) C: W36x160 buckling mode changes. Due to changes in the buck-
Fig. 13. Critical buckling modes for gravity plus full lateral ling modes, frames subjected to both gravity and full
loading with semi-rigid connections and H = OSL. lateral loading (joint point forces in both lateral
directions) did not exhibit consistent frame buckling
Figure 13 shows the critical buckling modes for the strengths for the different column heights and con-
semi-rigid connected frames subjected to gravity plus nection stiffnesses considered.
full lateral loading with H = OSL. Critical buckling
modes for the flexibly connected frames are similar to REFERENCES
those shown in Fig. 13 and the rigidly connected
frames are similar to those shown for the WI2 x 53 1. Y. B. Yang and W. McGuire, A procedure for analyzing
section, even the frames composed of W36 x 160 space frames with partial warping restraint. Int. J.
Numer. Merh. Engng 20, 1377-1390 (1984).
members. The right-hand column of Fig. 13 shows 2. Y. B. Yang and W. McGuire. Joint rotation and
that the W36 x 160 member frames with semi-rigid geometric non-linear analysis. J. srrucf. Engng, ASCE
and flexible connections do not experience any llZ(ST4), 879-905 (1986).
sidesway or twisting deformation in the critical buck- 3. R. S. Barsoum and R. H. Gallagher, Finite element
analysis of torsional and torsional-flexural stability
ling modes for any of the column patterns. All the problems. Int. .I. Numer. Meth. Engng 2, 335-352
girders participate in the buckling response for (1970).
column pattern A. Basically, only two girders partici- 4. M. M. Ettouney and J. B. Kirby, Warping restraint in
pate for column pattern C and oniy one girder for three-dimensional frames. J. struct. Dirt. ASCE
column pattern B. The reduced girder participation in 107(ST8), 1643-1656 (1981).
5. G. E. Blandford, S. T. Wang and C. D. Hill, Stability
the lateral beam buckling response is the reason and post-buckling analysis of steel space trusses. Proc.
column pattern A exhibits a greater buckling strength SSRC Conf. Computer Technology Applied to Structural
for both the semi-rigid and flexible connection cases Stability, April 26-27, Minneapolis, MN. pp. 309-320
with H =0X and IV36 x 160 members. (1988).
6. K. J. Bathe, Finite Element Procedures in Engineering
Analysis, Chap. 1I. Prentice-Hall, Englewood Cliffs,
CONCLUSIONS N.J. (1982).
7. Z. Razzaq and M. M. Naim, Elastic instability of
The results presented demonstrate the influence of unbraced space frames. J. struct. Dia., ASCE 106(ST7).
column orientation, connection rigidity, and column 1389-1400 (1980).
height on the elastic buckling response of a portal 8. R. C. Carlberg Jr, G. E. Blandford and S. T. Wang,
Stability analysis of steel space frames with flexible
space frame subjected to point gravity loading at the
connections and partial warping rigidity. Stabifify of
beam-to-column connection locations and gravity Bridges, Proc. 1990 SSRC Annual Technical Session,
plus lateral loading. Alternating the weak and strong April 9-l I, St Louis, pp. 121-131 (1990).

You might also like