Академический Документы
Профессиональный Документы
Культура Документы
PARTIAL RESTRAINTS3
By Zia Razzaq, 1 M. ASCE and Antoun Y. Calash 2
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
ABSTRACT: An inelastic theoretical study of the effect of biaxial partial end re-
straints on the response of hollow rectangular steel nonsway columns with or
without biaxial crookedness and residual stresses is presented. The partial end
restraints considered have linear, elastic-plastic, or trilinear moment-rotation
characteristics. The fundamental total equilibrium equations for the problem are
derived. An algorithm is presented based on a coupling between an iterative
tangent stiffness approach and a finite-difference scheme for evaluating the spatial
response of the materially nonlinear columns. The effect of partial end re-
straints, as well as the individual and combined influence of crookedness and
residual stresses is also explained by means of load-deflection, and bending
stiffness degradation curves. Several interesting conclusions are drawn regard-
ing the behavior of columns as affected by end restraints, imperfections and
slenderness.
INTRODUCTION
PROBLEM
758
n i t i a l l y crooked column
IP=0I
Deflected column
(P/0)
TTZ
Vi = voi sin — (4)
Fig. 2 shows the rectangular cross section of the column with width B,
759
The total normal strain at any point (x,y) on a cross section is ex-
pressed as follows:
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
P= - <rtdA-
d,dA - vYdA (8)
J A,r
JA JA
JA„f
Ae = dA (14)
JA,
761
(18)
Pp = aY dA ...
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
(19)
he = y2dA ...
JAe
(20)
lye = x1 dA ...
(21)
hye = xydA ..
J A,
Mm= <srydA (22)
J A,
BOUNDARIES
Ay !
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
0i I°I
|m|
Kl
k
a] ]
! ! to
le a l le b l Pe|
SOLUTION
The algorithm given by Razzaq (6) for planar analysis of partially re-
strained /-section columns is extended herein to solve Eqs. 11-13 for
spatial analysis of hollow rectangular columns. The procedure is based
on satisfying the compatibility of lateral deflections, u and v, by an it-
erative finite-difference scheme through the flexural Eqs. 12 and 13 while
satisfying the axial force and biaxial flexure equilibrium Eqs. 11-13 at
several cross sections along the column length by the tangent stiffness
approach given in Ref. 7. The second order finite central difference ver-
sion of Eqs. 12 and 13 applied at N equally spaced nodes over the in-
ternal [0,L] and the boundary conditions given by Eq. 33 leads to the
following system of simultaneous equations:
[KM] = {M} (35)
in which [Kt] = tangent stiffness coefficient matrix of the order 2N + 4;
{A} = a lateral deflection (u and v) vector; and {M} = a "load" vector. A
column with equal end restraints is analyzed with N nodes over the
interval [0,L/2] and using the symmetry conditions outlined previously.
In the elastic range, [Kt] and {M} are explicitly defined, and Eq. 35 can
be solved directly. In the inelastic range, the coefficients in [Kt] become
dependent upon the inelastic cross-sectional properties at various nodes
along the column length and are evaluated using the procedure in Ref.
7. The overall solution scheme is given as follows:
The axial load increment, SP, is held constant throughout the elastic
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
APPROACH CORROBORATION
765
in which urt = maximum tensile residual stress shown in Fig. 2; and Pmax
= dimensionless column maximum load given by
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 07/05/13. Copyright ASCE. For personal use only; all rights reserved.
Pmax = ~ (38)
in which Pmax = column maximum load. For these columns, initial crook-
edness was taken as uoi = L/1,000 and vai = L/100,000, although no mea-
surements of initial crookedness are given for columns 4-11 in Ref. 4.
The maximum difference between the P max values is less than 4%.
COLUMN BEHAVIOR
(39)
100,000
b= (40)
v ;
1,000
For the specific column studies presented, the following four different
initial spring stiffness values are used: k1 = 0.0 kip-in./rad (0.0 kN-cm/
rad); k2 = 5,397.22 kip-in. /rad (61,004.8 kN-cm/rad); k3 = 15,506.94 kip-
in./rad (175,274.9 kN-cm/rad); and k4 = 15 x 1015 kip-in./rad (169.54 x
1015 kN-cm/rad). The value of kx represents a pinned condition. The k2
value is provided by a 5 x 5 x 0.1875 in. (12.7 x 12.7 X 0.4762 cm)
hollow square restraining beam with a length, Lb = 12 ft (3.66 m), and
equals 2EIb/Lb, in which /,, = moment of inertia of the beam. The k3
value is obtained similarly by a 7.0 X 7.0 x 0.375 in. (17.78 x 17.78 X
0.952 cm) hollow square restraining beam. The stiffness, k4, simulates
a nearly fixed condition. The values of the plastic spring moments and
stiffnesses for piecewise-linear restraints to define the relationships in
Figs. 4(b-c) are given later.
Hollow Square Imperfect Columns with Equal Linear Rotational End
Restraints.—Table 2 summarizes the results for a total of 72 columns
766
numbered (CN) 12-83. Referring to Figs. 1 and 4(a), the end spring stiff-
nesses for these columns are defined as follows:
=
fc ksx = KBy = kTx = kTy (41)
In this table, the P m a x values corresponding to three slenderness ratios
(L/r), and four k values are presented for nearly straight, uniaxially
crooked, and biaxially crooked columns with or without residual stresses.
Columns 12-35 are nearly straight («„,- = a; voi = a); columns 36-59 are
uniaxially crooked («„,- = a; voi = b) a n d columns 60-83 are biaxially crooked
(w oi = v_oi = b ) .
The P max for the nearly straight columns 12-23 without residual stresses
are all close to unity for various k a n d L/r values. With residual stresses,
however, the effects of k a n d L/r on P m a x become noticeable for columns
24-35. The effect of L/r on P m a x is most p r o n o u n c e d w h e n k = 0 a n d
diminishes with increasing k. The e n d spring m o m e n t s for columns 1 2 -
35 at P max were all negligibly small.
The uniaxially crooked columns 3 6 - 4 7 become inelastic only in the
vicinity of P m a x , at the midspan. The P m a x for these columns are affected
both by k a n d L/r. The effect of L/r on P m a x diminishes considerably with
increasing k. The effects of k and L/r on P m a x are more p r o n o u n c e d w h e n
residual stresses are included. The m a x i m u m e n d spring m o m e n t de-
velops for column 59 at P m a x a n d is 0.097 M y , in which
My = ay x sectional m o d u l u s about y (42)
The biaxially crooked columns 60-71 without residual stresses also ex-
hibit dependence u p o n k and L/r. With residual stresses, this depen-
dence is further magnified. The maximum e n d spring m o m e n t develops
for column 83 at P m a x and is also 0.097 My .
The relationships between the dimensionless load given by
and the total midspan deflection, Vc, in the y direction for the p i n n e d
767
\[Kt]
D (44)
\[K,]\
in which \[Kt]\0 = determinant for zero axial load. The results in Fig. 5
show that residual stresses alone have a less detrimental effect on the
column strength than uniaxial crookedness alone. The biaxial crooked-
ness alone appears to be less detrimental than the combined effect of
uniaxial crookedness and residual stresses for columns 61 versus 49.
However, a comparison of P max for columns 48, 50-59, to those for col-
umns 60, 62-71 (see Table 2), respectively, shows that the biaxial crook-
edness alone has almost the same effect on the column strength as the
uniaxial crookedness and residual stresses combined. Also, the biaxially
crooked columns are weaker than the uniaxially crooked columns. Fig.
6 shows that the D-P relationship remains identical in the elastic range
for the six columns under consideration. Beyond the elastic range, the
D-P curves follow different descending paths, depending upon the type
of imperfections, and end at different locations along the P axis. A com-
parison of the inelastic portions of the D-P curve for column 25 (with
residual stresses) to that of column 37 (with uniaxial crookedness) re-
veals that although the former one experiences earlier stiffness degra-
dation due to early yielding, it is still stronger than the later one. The
/
CN 13 (NS.NRI
- t
• CN 25 INS.WR)
CN37 l U C W R I ' ^
CN61 (BCNR)-"""
j \CN49(UC,WRI
/ CN 73 [BC.WRI
"
/
NS=Nearly Straight
UC=Uniaxiolly Crooked
BC=Biaxially Crooked
NR=No Residual Stresses
WR=With Residual Stresses
768
"kill it" faster than column 25. In general, however, t h e relative column
strengths m u s t be d e p e n d e n t u p o n the relative degrees of crookedness
and residual stresses, and the ensuing distributions of yielding along the
column length. Figs. 7 and 8 respectively, show the P-Vc a n d D-P curves
for columns 19, 31, 43, 55, 67 a n d 79, with L/r = 53.73; k = k3; and for
various combinations of crookedness a n d residual stresses. These curves
for the partially restrained columns are similar in nature to those in Figs.
5 and 6, respectively, for the p i n n e d columns, with two exceptions. The
restrained imperfect columns carry higher loads, and their bending stiff-
ness degradation occurs at a slower rate with increasing P than that of
the p i n n e d columns.
The effect of uniaxial crookedness on P m a x may be observed by com-
paring the results in Table 2 for columns 12-23 to 36-47, respectively.
Also, the effect of biaxial crookedness on P m a x m a y be observed by com-
paring the results of columns 12-23 to 6 0 - 7 1 , respectively. The detri-
mental effect of uniaxial or biaxial crookedness is intensified by both an
increase in L/r and a decrease in k. This conclusion applies also to the
crooked columns with residual stresses.
Of all the hollow square imperfect columns analyzed in the presence
of equal linear rotational end restraints (Table 2), those with both biaxial
crookedness and residual stresses are the weakest. Fig. 9 presents re-
lationships between P m a x a n d k for these columns. It is seen that each
-\
XN 31 INS.WR)
\Q( \CN55(UC,WR)
It^ \ \ CN79IBC.WR)
\ CNi3(UC,NR!
CN 67 [BC.NRI
NS = Nearly Straight
UC Uniaxially Crooked
BC= Biaxially Crooked
NR= No Residual Stresses
WR= With Residual Stresses
\
769
^ L/r=35.82
\
^ L/r=5373
\L/r=71.64
k k, =kT = k
Bx= Tx By Ty
k (10J kip-in/rad)
of the three solid curves in this figure becomes relatively flat beyond
some moderate value of k.
Effect of Spring Stiffness Ratio on the Strength of Hollow Square
Imperfect Columns.—The results for columns 84-89 in Table 3, and for
columns 73, 76, 79, and 82 in Table 2 may be used to study the effect
of varying the end restraint stiffness ratio, kx/ky, on P max , such that
kx = kBx = kTx (45)
Kj = kBy = kTy (46)
These columns have an L/r = 53.73, and possess both biaxial crooked-
ness and residual stresses. Fig. 10 shows Pmax versus kx curves for var-
ious constant values of ky . As kx increases, the difference between Pmax
values increases as ky is varied from zero to infinity.
The results for columns 90-95 in Table 3, and for columns 73, 76, 79
770
k„MO kip~in/rod]
and 82 in Table 2 may be used to study the effect of the ratio, kB/kT,
between the top and bottom end restraint stiffnesses, on P m a x , such that
kB = kBx = kBy (47)
KT — KT ^Ty < (48)
kg (10 kip-in/rod]
"m "of
a, a a, b b,a b, b
k v„ CN p
x
CN p CN P,„ax CN p
A
max * max max
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
fci 0.0 96 0.987 104 0.924 112 0.878 120 0.858
fci 0.5 97 0.878 105 0.869 113 0.808 121 0.799
k2 0.0 98 0.998 106 0.935 114 0.896 122 0.887
k2 0.5 99 0.933 107 0.910 115 0.855 123 0.844
k3 0.0 100 0.998 108 0.949 116 0.919 124 0.917
h 0.5 101 0.976 109 0.942 117 0.905 125 0.894
h 0.0 102 0.998 110 0.971 118 0.962 126 0.955
h 0.5 103 0.998 111 0.967 119 0.946 127 0.942
772
"0.0 0.1 0.2 0.3 0.4 0.5 0.6 07 "0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.2 02
(a) (W
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.20
k (103 kip-in/rad)
1. Linear: k = k3.
2. Elastic-Plastic: k = k3; and \mv\ = m„.
3. Trilinear: ka = k3; kb = 0.5 k3; kc = k1; \m„\ = 0.5 m„; and \mb\ = m0.
CONCLUSIONS
ACKNOWLEDGMENT
APPENDIX.™REFERENCES
1. Ballio, G., and Campanini, G., "Equivalent Bending Moments for Beam-Col-
umns," Journal of Constructional Steel Research, Vol. 1, No. 3, May, 1981, pp.
13-23.
2. Chen, W. F., and Atsuta, T., Theory of Beam Columns, Vol. 2, McGraw-Hill,
New York, N.Y., 1977.
3. Johnston, B. G., ed., Guide to Stability Design Criteria for Metal Structures, 3rd
Ed., John Wiley and Sons, Inc., New York, N.Y., 1976.
4. Jain, A. K., Goel, S. C , and Hanson, R. D., "Inelastic Response of Restrained
Steel Tubes," Journal of Structural Division, ASCE pp. 897-910.
5. Razzaq, Z., and McVinnie, W. W., "Rectangular Tubular Steel Columns Loaded
Biaxially," Journal of Structural Mechanics, Vol. 10(4), 1983, pp. 475-493.
6. Razzaq, Z., "End Restraint Effect on Steel Column Strength," Journal of Struc-
tural Engineering, ASCE, Vol. 109, No. 2, Feb., 1983, pp. 314-334.
7. Santathadaporn, S., and Chen, W. F., "Tangent Stiffness Method for Biaxial
Bending," Journal of the Structural Division, ASCE, Vol. 98, No. ST1, Jan., 1972,
pp. 153-163.
776